Cone Penetration Testing 2022

Cone
Penetration
Testing 2022
Editors
Guido Gottardi & Laura Tonni
CONE PENETRATION TESTING 2022
This volume contains the proceedings of the 5th International Symposium on Cone Penetration Testing
(CPT’22), held in Bologna, Italy, 8-10 June 2022. More than 450 authors - academics, researchers, practi
tioners and manufacturers – contributed to the peer-reviewed papers included in this book, which includes
three keynote lectures, four invited lectures and 170 technical papers. The contributions provide a full picture
of the current knowledge and major trends in CPT research and development, with respect to innovations in
instrumentation, latest advances in data interpretation, and emerging fields of CPT application.
The paper topics encompass three well-established topic categories typically addressed in CPT events:
- Equipment and Procedures
- Interpretation
- Applications.
Emphasis is placed on the use of statistical approaches and innovative numerical strategies for CPT data
interpretation, liquefaction studies, application of CPT to offshore engineering, comparative studies between
CPT and other in-situ tests. Cone Penetration Testing 2022 contains a wealth of information that could be
useful for researchers, practitioners and all those working in the broad and dynamic field of cone penetration
testing.
PROCEEDINGS OF THE 5TH INTERNATIONAL SYMPOSIUM ON CONE PENETRATION

TESTING (CPT’22), 8-10 JUNE 2022, BOLOGNA, ITALY
Cone Penetration Testing 2022
Editors
Guido Gottardi
Department of Civil, Chemical, Environmental and Materials Engineering, University of
Bologna, Bologna, Italy
Laura Tonni
Department of Civil, Chemical, Environmental and Materials Engineering, University of
Bologna, Bologna, Italy
Cover image: photo by Claudio Turci
First published 2022

by CRC Press/Balkema
Schipholweg 107C, 2316 XC Leiden, The Netherlands
e-mail: [email protected]
www.routledge.com – www.taylorandfrancis.com
CRC Press/Balkema is an imprint of the Taylor & Francis Group, an informa business
© 2022 selection and editorial matter, Guido Gottardi & Laura Tonni; individual chapters,
the contributors
The right of Guido Gottardi & Laura Tonni to be identified as the authors of the editorial
material, and of the authors for their individual chapters, has been asserted in accordance
with sections 77 and 78 of the Copyright, Designs and Patents Act 1988.
The Open Access version of this book, available at www.taylorfrancis.com, has been made
available under a Creative Commons Attribution-Non Commercial-No Derivatives 4.0
license.
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.
Library of Congress Cataloging-in-Publication Data
A catalog record has been requested for this book
ISBN: 978-1-032-31259-0 (hbk)

ISBN: 978-1-032-36228-1 (pbk)
ISBN: 978-1-003-30882-9 (ebk)
DOI: 10.1201/9781003308829
Cone Penetration Testing 2022 – Gottardi & Tonni (eds)
© 2022 the Editor(s), ISBN 978-1-032-31259-0
Open Access: www.taylorfrancis.com, CC BY-NC-ND 4.0 license
Table of contents
Preface xvii
Symposium organizers xix
Committees xxi
International advisory board xxiii
Reviewers xxv
Sponsors xxix
Keynote lectures
New CPT methods for evaluation of the axial capacity of driven piles 3
B.M. Lehane, E.J. Bittar, S. Lacasse, Z. Liu & F. Nadim
Material Point Method simulations of cone penetration and CPT interpretation 16
R. Salgado, V. Bisht & M. Prezzi
Practical use of shear wave velocity measurements from SCPTU in clays 28
M. Long
Invited papers
Uncertainties associated with CPT data acquisition 55
R. Soage Santos
CPT equipment: Recent advances and future perspectives 66
D.J. White
Non-deterministic interpretation and applications of CPT testing data 81
M. Uzielli
Combined use of CPT & DMT: background, current trends and ongoing developments 94
P. Monaco
Technical papers
Session 1: Equipment and procedures

Incorporation of SH source wave parameter “SH Polarization” within DST seismic trace
characterization 109
E. Baziw & G. Verbeek
v
Methodology for obtaining true cone bearing estimates from blurred and noisy
measurements 115
E. Baziw & G. Verbeek
Evaluation of parameters inducing desaturation of a piezocone: Saturation liquid viscosity
and exposure to dry sand 121
G. De Backer, R.D. Verastegui-Flores, W. Vervaele, L. Vincke & K. Haelterman
Large diameter cone penetrometers: What is an appropriate location for the transition to the
rod diameter? 127
D.A. de Lange, T.A. van Duinen & D.J. Peters
Issues related to piezocone sleeve friction measurement accuracy in soft sensitive clays 133
B. Di Buò, M. D’Ignazio, T. Länsivaara & M. Haikola
Flow cone – new CPTU add-on module trialled in Halden silt 139
A.S. Gundersen, T. Lunne, R. Stelzer, Ø. Blaker, G.W. Tucker & L. Krogh
Estimating in-situ frozen loamy soil viscosity from CPT 145
O.N. Isaev
Determination of natural stress state parameters for clay soils by using 3LSU-CPTU
penetrometer 152
O.N. Isaev, R.F. Sharafutdinov & D.S. Zakatov
Towards correlating seabed penetrometer and chirp sonar measurements 159
R. Jaber & N. Stark
Development of free fall cone penetration testing system 165
H. Kang, O. Kwon, C. Shin, J. Seo, I. Jang & M. Dong-Woo
Evaluation of statistical fluctuation of measured data from nuclear density cone
penetrometer 171
M. Karthikeyan
On the accuracy and precision of the seismic cone penetration test – a field test study on the
seismic source 177
O. Koreta, A.H. Augustesen, L. Krogh, K. Lundvig & S. Bøtker-Rasmussen
The revival of multiple pore pressure measurements in the cone penetration test 184
T. Lunne, R.K. Ghanekar, G.W. Tucker, R. Santos & L. Krogh
Comparisons CPT-DMT in soft clay at Fucino-Telespazio GeoTest site 191
D. Marchetti, P. Monaco, G. Totani, F. Totani & S. Amoroso
An innovative new 3MPa CPT – to detect and measure very small fs values 198
A.J. McConnell & E.J.C. Wassenaar
Assessment of seismic cone penetration testing for small strain shear modulus 204
N. Parasie, T. Franken & J. Peuchen
Experimental procedure for checking the saturation degree of piezocone tips 210
I. Rocchi, L. Tonni, G. Gottardi & M. Marcolongo
Using penetrometer in situ and in box-core testing to obtain design information for lazy
wave riser-soil interaction 215
O. Safaqah, H.E. Low, S. Pant, S. Ingarfield, M.F. Bransby, M.F. Randolph & Z.J. Westgate
vi
Calibration of cone penetrometers according to International Organization for
Standardization requirements 221
R. Soage Santos, E. Gómez Meyer, J. Peuchen, G. Yetginer, T. Lunne & T. Carrington
Long-term strength determination of frozen soils by CPT 226
I. Sokolov & N.G. Volkov
Efficiency examined of hands-free Cone Penetration Testing using the SingleTwist™ with
COSON 230
O. Storteboom, M. Woollard & J. Verhagen
Sustainability in CPT practice: Hybrid CPT Track-Truck 236
O. Storteboom, M. Woollard & B. Ooms
Elastic soil properties investigated using seismic tests to complement the CPT 242
O. Storteboom, M. Woollard & J.L. Rangel-Núñez
Tensiocone: A cone penetrometer with the facility to measure negative pore-water pressure 248
A. Tarantino, A. Capotosto, F. Bottaro, M. Bellio & D. Gallipoli
New portable pressiocone system for carrying out CPT+FDP tests 254
G. Vinco & M. Sacchetto
Characterisation of near-surface sediments using a blend of vertical and shallow rotational
penetrometers 259
D.J. White, S.A. Stanier & H. Mohr
Development of an enhanced CPT system for Dogger Bank 267
T.I. Yetginer-Tjelta, S. Bøtker-Rasmussen, M. Rose, T. Lunne, V. Meyer & C. Duffy
Session 2: Interpretation
CSi – a joint industry project into CPTUs in silty soils 275
A.H. Augustesen, P. Carotenuto, C. Bilici, T. Lunne, R.C.J. Lindeboom, L. Krogh, J. van den
Bosch, R. Barth, C. Erbrich, S. Ingarfield, D. Giretti, V. Fioravante, H. Dias, M.-C. Sougle,
A. Barwise, S. de Wit, D. Burbury & N. Adams
Numerical investigation of piezocone dissipation tests in clay: Sensitivity of interpreted
coefficient of consolidation to rigidity index selection 282
A. Barati Nia, D.M. Moug, A.P. Huffman & J.T. DeJong
Correlation of CPT measurements and VibroCore penetration speed for medium, calcareous
sands: A case study of the cable route survey at the North sea 288
K. Bartczak & G. De Vries
CPT data interpretation for an improved characterization of the paleosol stratigraphy in the
Po River Valley, Italy 294
I. Bertolini, M. Marchi, L. Tonni, G. Gottardi, L. Bruno & A. Amorosi
Application of CPT to the evaluation of permeability in a Po river embankment prone to
backward erosion piping 300
I. Bertolini, G. Gottardi, M. Marchi, L. Tonni, A. Bassi & A. Rosso
Dissipation tests to evaluate the equilibrium pore pressure 306
F.A.B. Danziger, G.M.F. Jannuzzi & A.V.S. Pinheiro
vii
Dynamic characteristics of the soils by Cone Penetration Tests (CPT) 311
A. Cavallaro
Validating cone penetration test in partially drained conditions using a simplified numerical
modelling method 317
R.W.L. Chia, Z.Z. Wang & S.H. Goh
Cone penetration testing to constrain the calibration process of a sand plasticity model for
nonlinear deformation analysis 325
A. Chiaradonna, T.J. Carey, K. Ziotopoulou & J.T. DeJong
Probabilistic delineation of soil layers using Soil Behavior Type Index 332
S. Collico, M. Arroyo, M. DeVincenzi, A. Rodriguez & A. Deu
Clustering analysis to improve total unit weight prediction from CPTu 339
S. Collico, M. Arroyo, M. DeVincenzi, A. Rodriguez & A. Deu
Combining CPTU and UMASW to characterise Irish offshore deposits 345
M. Coughlan, A. Trafford, M. Long, S. Donohue & S. Corrales
Cone factor from CPTU tests in very soft clays at the east of Mexico’s valley 351
J.M. De La Rosa R. & F.A. Flores López
Piezocone testing in Nordic soft clays: Comparison of high-quality databases 356
M. D’Ignazio, B. Di Buò, T. Länsivaara, J.-S. L’Heureux, P. Paniagua & J. Selänpää
Prediction of resilient modulus of cohesive subgrade soils from CPTU data using polynomial
neural networks 363
W. Duan, Z. Zhao, G. Cai, A. Wang, R. Chen, A.J. Puppala, S. Liu & S.S.C. Congress
Effect of sand bio-cementation on cone tip resistance: A numerical study 368
M.El Kortbawi, K. Ziotopoulou, J.T. DeJong & D.M. Moug
A data-driven approach to predict shear wave velocity from CPTu measurements 374
I. Entezari, J. Sharp & P.W. Mayne
Interpretation and comparison of CPT derived soil properties to static and cyclic laboratory
tests on unique fine-grained soils in Western Washington and Oregon 381
B. Exley, A. Pyrch, J. Jacoby & B. Thunder
Evaluation of cyclic softening potential using CPTu and assessment with cyclic triaxial test
results: A case study 387
K. Fakharian, M. Bahrami, M. Kashkooli, H. Vaezian & T. Bahrami
Application of integrated Game Theory-optimization subground stratification (-IGTOSS)
model to Venetian Lagoon deposits 394
M.S. Farhadi, T. Länsivaara & L. Tonni
Application of two novel CPTu-based stratification models 400
M.S. Farhadi, T. Länsivaara, J.-S. L’Heureux & T. Lunne
Numerical simulation of CPT in sands using DeltaSand and Hardening Soil models 407
M. Fetrati, V. Galavi, M. Goodarzi, S. Kreiter & T. Mörz
CPT calibration in centrifuge: Effect of partial saturation on cone resistance 414
V. Fioravante, D. Giretti, E. Dodaro, C.G. Gragnano & G. Gottardi
viii
Calibration cone penetration testing in silty soils 420
V. Fioravante, D. Giretti, T. Lunne, P. Carotenuto, A.H. Augustesen, R.C.J. Lindeboom,
L. Krogh, H. Dias, M.-C. Sougle, A. Barwise, S. de Wit & D. Burbury
Comparison between coefficients of consolidation from CPTu and laboratory tests for
Guaratiba’s soft soil, Rio de Janeiro, Brazil 427
M.M. Freire, M.E.S. Marques, M.C. Tassi & L.A. Berbert
Quantitative modelling of spatial variability of piezocone data from Venice lagoon silty soils 432
G. Gottardi, M. Ranalli, L. Tonni & M. Uzielli
Thin-layer detection from the cone resistance rate of change 438
H.B. Hammer, S. Nordal, J.-S. L’Heureux & H. Skrede
Recalculation of in-situ CPTu in intermediate soils using G-PFEM 445
L. Hauser, S. Oberhollenzer, A. Gharehaghajlou, H.F. Schweiger, R. Marte & C. Fabris
Full-flow CPT tests in a nearshore organic clay 452
S. Hov, K. Borgström & P. Paniagua
Bayesian supervised learning of 2D subsurface soil stratigraphy using limited cone
penetration tests with consideration of uncertainty 459
Y. Hu & Y. Wang
Estimation of constrained modulus from CPT measurements in case of Holocene sands 466
Z. Illés, I. Kádár, G. Nagy, A. Mahler & L. Nagy
Evaluation of complex CPTu dissipation tests of B.E.S.T. 473
E. Imre, M. Hegedűs, L. Bates & S. Fityus
Simulation of CPT penetration in sensitive clay 480
J. Isaksson, J. Yannie, M. Karlsson & J. Dijkstra
A CPT-based method for estimation of undrained shear strength of sands and transitional
soils 486
K. Kaltekis & J. Peuchen
Comparison of frozen soil strength characteristics by cone penetration and triaxial
compression testing 491
D. Lagosha, I. Sokolov & N.G. Volkov
Fincone: A study on the use of CPT in soft sensitive clays 497
T. Länsivaara, B. Di Buò, J. Selänpää, M. Knuuti & M. Haikola
CPT-based unit weight estimation extended to soft organic clays and peat: An update 503
H.J. Lengkeek & R.B.J. Brinkgreve
CPT-based classification of soft organic clays and peat 509
H.J. Lengkeek & R.B.J. Brinkgreve
Shear wave velocity – SCPTU correlations for sensitive marine clays 515
M. Long & J.-S. L’Heureux
A simplified method to incorporate the benefits of microstructure for cyclic liquefaction
analyses using the SCPT 521
K. Lontzetidis, P.K. Robertson & D.J. Morton
ix
Classification of Miocene deposits using CPT data 528
A. Makra & H. Kim
On the interpretation of piezocone dissipation testing data in clay 534
F.M. Mántaras, F.S. Pereira, E. Odebrecht & F. Schnaid
Determination of fine-grained soil parameters using an automated system 540
I. Marzouk, F. Tschuchnigg, F. Paduli, H.J. Lengkeek & R.B.J. Brinkgreve
Undrained shear strength of clays from piezocone tests: A database approach 546
P.W. Mayne & J. Peuchen
Evaluating geoparameters of Maine sensitive clay by CPTU 552
P.W. Mayne, P. Paniagua, B. Di Buò & S.S. Agaiby
CPTU evaluations in Appalachian Piedmont residual sandy silts 559
P.W. Mayne & E. Cargill
Soil unit weight prediction from CPTs for soils and mining tailings 566
T. Menegaz, E. Odebrecht, H.P. Nierwinski & F. Schnaid
Use of DMT and CPTU to assess the G0 profile in the subsoil 570
Z. Młynarek, J. Wierzbicki & P. Monaco
Physical and numerical modelling of T-CPT for mechanisms of penetration and heat transfer 577
P.Q. Mo, L. Gao, H.S. Yu, X.L. Tao & Q.Z. Ma
Soil stratigraphy from seismic piezocone data and multivariate clustering in alluvial soil
deposits: Experience in the Lower Tagus Valley region 584
F. Molina-Gómez, D. Cordeiro, C. Ferreira & A. Viana da Fonseca
Undrained strength from CPTu in brittle soils: A numerical perspective 591
L. Monforte, M. Arroyo & A. Gens
Assessment of deltaic soil behavior classification using AUT: GMD database regarding
CPTu records 598
M.H. Naghibi, A. Eslami & S. Heidarie Golafzani
A review of methods for estimating undrained brittleness index from the CPT 604
Y. Narainsamy & S.W. Jacobsz
A comparative study on CPTu-based soil classification methods: Case studies 610
T.D. Nguyen, P.S. Khin, Q.N. Pham & A.T. Vu
Characterization of young sediments using CPTu and Medusa SDMT 617
S. Oberhollenzer, L. Hauser, F. Brand, R. Marte, H.F. Schweiger, D. Marchetti & S. Pfeifer
PFEM modeling of CPTu tests in saturated structured soils 623
K. Oliynyk, M.O. Ciantia & C. Tamagnini
Effect of the scatter between CPTU measured parameters in soil classification 630
P. Paniagua & J.-S. L’Heureux
Evaluation of shear wave velocity profiles in alluvial and deltaic soils using a CPT database 636
J. Paredes & F. Illingworth
Upscaling 1 500 000 synthetic CPTs to voxel CPT models of offshore sites 641
J. Peuchen, W. van Kesteren, V. Vandeweijer, S. Carpentier & F. van Erp
x
Automated CPT interpretation with a Convolutional Neural Network 646
M. Pippi, R. Vink, J. Haasnoot & S. Bersan
Watch out for the use of global correlations and “black box” interpretation of CPTU data 651
J.J.M. Powell & L. Dhimitri
New methods for assessing Plasticity Index and Low-strain Shear Modulus in fine-grained
offshore soils 657
N. Ramsey & K.K. Tho
Cone penetration testing and interpretation in the holds of two ore-carrying vessels 664
N. Ramsey
Suction influence on CPT and DMT for some Brazilian tropical soils 670
B.P. Rocha, R.A. Rodrigues & H.L. Giacheti
Study of SPT-CPT and DP-CPT correlations for sandy soils 677
M.D. Santos & K.V. Bicalho
Correlation between SPT and CPT tests in liquefiable deposits 683
R. Shahgholian, C. Ferreira & A. Viana da Fonseca
CPTU-detection of thin clay layers in sand: Results from calibration chamber tests 690
H. Skrede, H.B. Hammer, S. Nordal & J.-S. L’Heureux
Determination of hydraulic conductivity using HPT & CPTu 697
M. Slowiok, S. Oberhollenzer, R. Marte & T. Freudenthaler
DEM-FDM coupling simulation of cone penetration tests in a virtual calibration chamber 703
Y. Song, X.Q. Gu & J. Hu
Data-driven soil profile characterization using statistical methods and artificial intelligence
algorithms 708
R.L. Spacagna, A. Baris, L. Paolella & G. Modoni
Capability of seismic CPTu and DMT in assessing propagation velocity of body waves:
A comparative study 715
S. Stacul, D. Lo Presti, N. Nenci, F. Fiera, M. Perini, D. Marchetti, E. Pagani & M. Siviero
Bayesian estimation of small-strain shear modulus from offshore CPT tests in the North Sea 722
B. Stuyts, C. Sastre Jurado, D. Gomez Bautista & A. Kheffache
Factors influencing CPTU Nkt for marine clay in Singapore reclaimed land 728
C. Tanaka, M. Angeles & J.Y. Wong
Parameters affecting the CPT resistance of reconstituted sands 734
Y. Tian & B.M. Lehane
Practical experience with cone penetration in frozen soils 741
N.G. Volkov, I.S. Sokolov & R.A. Jewell
A site-specific relationship between CPT data and fines content for fine grained soil in the
context of liquefaction analyses 747
C. Vrettos
Numerical modelling of cone penetration tests in spatially variable clays 753
Z.Z. Wang, S.H. Goh & X. Zheng
xi
Some aspects of in situ testing of clay-glacial till mixture redeposited as man-made fills 760
J. Wierzbicki, K. Stefaniak, S. Wilczyński & B. Brzeziński
Cone penetration in a thin medium dense sand layer sandwiched by different clay
layers – LDFE analysis 766
Q. Xie, Y.X. Hu, M.J. Cassidy & M. Zhou
Study on SPT N-values and relative density through various soundings in full-scale chamber
test ground 772
H. Yabe, K. Harada, T. Ito & E. Watanabe
Correcting measured CPT tip resistance for multiple thin-layer effects 778
K.M. Yost, J. Cooper, R.A. Green, E.R. Martin & A. Yerro
Spatial interpolation of consolidation property of clays from limited CPTU dissipation data 784
Z. Zhao, W. Duan, G. Cai, M. Wu, Anand J. Puppala, S. Liu & S.S.C. Congress
Session 3: Applications
Evaluation of the geotechnical behavior of mining tailings through CPTU tests in the soil
improvement process for the decharacterization of upstream heightened dams 793
J.L. Albino, T.A.T. Souza J. & L.S. Machado
VCPT: An in-situ soil investigation method to validate vibratory pile-soil interaction models 799
D. Al-Sammarraie, S. Kreiter, T. Mörz, M.O. Kluger & M. Goodarzi
Monitoring ground improvement by Rammed Aggregate Piers using a combined CPTU and
SDMT approach at a silty sand liquefaction-prone site in Emilia-Romagna 806
S. Amoroso, M.F. García Martínez, L. Tonni, G. Gottardi, P. Monaco, K.M. Rollins, L. Minarelli,
D. Marchetti & K.J. Wissmann
A CPT-based method for monotonic loading of large diameter monopiles in sand 812
S. Bascunan, K. Kaltekis, B. van Dijk & K.G. Gavin
Automatic interpretation and statistical evaluation of soil conditions for preliminary design
of offshore foundations using the cone penetration test 819
L. Berenguer Todo Bom & M. Kanitz
Effect of dynamic pile driving parameters on vibratory penetration 825
A. Bhaskar, S. Kreiter, D. Al-Sammarraie & T. Mörz
Application of a new qc averaging approach for end bearing of driven piles in sand 832
E.J. Bittar, Y. Tian & B.M. Lehane
A review of a CPT based axial capacity prediction of screw piles in sand 838
E.J. Bittar, B.M. Lehane, S. Mahdavi, A.P. Blake, D.J. Richards & D.J. White
CPT-based liquefaction ejecta evaluation procedure 844
J.D. Bray & D. Hutabarat
Settlement estimations for buildings founded on saturated silty sands from CPT and DMT
results 850
M. Cáceres, J. Fumeron, F.A. Villalobos & R. Moffat
Numerical modeling of static load test in drilled shaft using CPTu results 857
M.A. Camacho, C.B. Camacho & V.H. Miranda
xii
Characterization of geotechnical spatial variability in river embankments from spatially
adjacent SCPT 863
F. Ceccato, M. Uzielli & P. Simonini
Coupling site wide CPT profiles and genetic algorithms for whole-site offshore windfarm
layout optimization 870
J.A. Charles, S.M. Gourvenec & M.E. Vardy
CPT-based model calibration for effective stress analysis of layered soil deposits 876
A. Chiaradonna, N. Ntritsos & M. Cubrinovski
Stress increase induced by impact precast pile driving 883
V. Colella, G. Cortellazzo, A. Dei Svaldi, S. Amoroso, L. Minarelli & K.M. Rollins
CPT based liquefaction potential of flood defences in The Netherlands 889
T. de Gast, K.G. Gavin, P.D. Notenboom, R. Abraimi & C. Reale
Application of the CPT for liquefaction assessment of gravelly reclamations at the port of
Wellington 894
R. Dhakal, M. Cubrinovski & J.D. Bray
Concept design of a new CPT module for direct in situ measurement of p-y soil responses 900
A. Diambra, J. Creasey, J. Leonet, A. Conn, E. Ibraim, G. Mylonakis, D.J. White,
B. Cerfontaine, S.M. Gourvenec & D. Igoe
The use of dynamic probing tests and cone penetration tests to verify the effectiveness of
expanding polyurethane resin injections for ground improvement 907
A. Dominijanni, M. Gabassi, A. Minardi & S. Pavan
Evaluation of flow liquefaction susceptibility of a sandy-silt tailings using the CPTu 913
M.P. dos Santos Junior, R. César Gomes, S.G. Silva Ribeiro & B.G. Delgado
Prospects on data mining approach for pile geotechnical design utilizing CPT and CPTu
records: Case study: AUT database 920
A. Eslami, S. Heidarie Golafzani & S. Moshfeghi
Evaluating mitigation of kinematic moments of precast driven piles in liquefiable layers using
pre- and post-CPTu tests 926
K. Fakharian, D. Mohtashamamiri, K. Behroozian, M. Bahrami, T. Bahrami & I.H. Attar
Verification of 3D FEM analysis of ground improvement works using CPT test results 932
G.A. Faour & K.N. Khouri
A comprehensive design procedure for pile groups in liquefiable soils 938
M. Franceschini, F. Fiorelli & E. Bandiera
Deep foundations of the new Pavilion 37 – fair quarter Bologna 944
M. Franceschini, F. Fiorelli, E. Bandiera, V. Colella, G. Cortellazzo & A. Dei Svaldi
Consolidation settlement of coastal areas of the Emilia-Romagna region from cone
penetration tests 949
D. Giretti, V. Fioravante, L. Perini & L. Calabrese
A comparative study on liquefaction assessment of Rajarhat area of Kolkata by using
different approaches 955
A. Halder, K. Das, S. Nandi & K. Bandyopadhyay
xiii
Kriging analysis on CPTU data from offshore wind farm 961
R. He, J. Li, S. Yang & B. He
Evaluation of CPT-based design method for offshore pile 967
B. Huang, E.J. Bittar, Y. Zhang & X. Fu
Application of CPT based 3DFE approach for estimating monopile damping in sand 973
D. Igoe & M.B. Mohammed
The use of CPTU for driven piles designed in a backfilled opencast ‘marl hole’ in an
important post-industrial revolution area within the UK 979
D. Illingworth, C. Burton, L. Dhimitri, D. Ward & P. Shelton
Consolidation settlement prediction using cone penetration testing 985
M. Kermani & F. Esford
3D FE derivation of CPT based soil reaction curves for monopile lateral static design in sand 991
L.-M. Lapastoure & D. Igoe
CPT-based assessment of densification induced by stone column installation 998
F. Marchi, E. Zambianchi, A. Boschi, A. Mastrangelo, G. Marchi, G. Gottardi & L. Tonni
Geopier Impact technology for liquefaction risk mitigation based on CPTu investigations 1004
G. Martinez, K.J. Wissmann, M. Franceschini, E. Bandiera & F. Fiorelli
Estimating bearing capacity of polar snow using the Cone Penetration Test (CPT) 1010
A.B. McCallum & G. White
The use of CPT based metamodels to predict the performance of offshore anchor piles 1016
A Mentani, L. Govoni & F. Bourrier
Spatial interpolation of sparse PCPT data to optimise infrastructure design 1023
M.P. O’Neill, M.F. Bransby, J.P. Doherty & P. Watson
Use of CPT as a soil mixing verification tool: Some practical observations 1029
A. O’Brien & I. Murray
Liquefaction damage assessment using Bayesian belief networks 1035
L. Paolella, A. Baris, G. Modoni, R.L. Spacagna & S. Fabozzi
Geotechnical zoning of deltaic and alluvial soils of Guayaquil (Ecuador) using CPT and Nkt
calibration based on FVT 1042
J. Paredes, F. Illingworth & R. Luque
Use of piezocone with dissipation tests CPTu_Δu, in tailings dams in Mexico: Case history 1048
J.L. Rangel-Núñez, E. Ibarra-Razo & R. Flores-Eslava
Assessment of the spatial variability of a Croatian flood embankment using the cone
penetration test 1053
C. Reale, M.S. Kovačević, M. Bacic & K.G. Gavin
Prediction of bearing capacity and settlement using penetrometer design method for shallow
foundation and load transfer curves 1058
P. Reiffsteck, F. Szymkiewicz, M.A. Benz Navarrete & T.A. Luong
Probabilistic analysis of gyttja undrained strength from CPTU data for slope stability analysis 1065
S. Rios, L. Sousa, A. Viana da Fonseca, P. Milheiro-Oliveira & O. Hededal
xiv
Monitoring ground improvement using in situ tests in Guayaquil, Ecuador 1071
F. Ripalda, D. Falquez, D. Besenzon, R. Luque, F. Illingworth & S. Amoroso
Plate anchor capacity estimation through CPT tip resistance in sand 1077
A. Roy & S.H. Chow
CPT results and installation parameters for CFA piles in pyroclastic soils 1083
G. Russo, M. Ramondini, A. Vecchietti & G. Russo
Validation of CPT-based methods for estimation of footing settlement in sand 1089
V.A. Sakleshpur, M. Prezzi & R. Salgado
Derivation of SRD for driven piles from CPT data 1095
J.A. Schneider, M.F. Randolph & J.P. Doherty
New method for assessing soil liquefaction resistance using a cyclic cone penetrometer 1102
A. Sharma, P. Rapanakis, E. Incardona, C. Dano, L. Sibille, B. Chareyre & H.H. Sadrabadi
Evaluation of deformation modulus during Cone Loading Tests (CLT) and settlement of
shallow foundations 1107
A. Teyssier, M. Rispal, C. Jacquard & P. Reiffsteck
Assessment of the potential for liquefaction using CPTu in the tailings dam I – Feijão 1113
M. Tincopa & G. Carnero-Guzman
Is CPT a suitable in situ test for characterizing gravely sands? 1119
G. Togliani
A CPT-based diameter-dependent m-θ spring model for lateral pile analysis 1126
J. Tott-Buswell & L.J. Prendergast
An investigation into the use of the Vibdrive and β-methods for calculating the SRV of
offshore piled foundations 1131
M.P. Trubshaw, T. Joseph & G. Giuliani
Potential of the Cone Pressuremeter Test for obtaining stiffness degradation for offshore
wind turbine monopile foundations 1137
G.W. Tucker, C.T. Leth, L. Krogh, P. Ladefoged, T. Lunne & M. Taylor
Centrifuge study on the CPT based p-y models for the monopiles 1143
H. Wang, D.V. van Zanten, D.A. de Lange, F. Pisanò, K.G. Gavin & A. Askarinejad
Using near-surface CPT data to predict foundation skirt embedment in partially drained
carbonate sands 1149
H.M. Wroth, M.F. Bransby, C.D. O’Loughlin, M.F. Silva, M. Cocjin, N. Levy & H.E. Low
The role of cone penetration testing in the Dogger Bank offshore wind farm 1156
T.I. Yetginer-Tjelta, J. De Sordi, L. Caferri, M. Rose, C. Duffy, T. Lunne, Ø. Blaker,
S. Strandvik & V. Meyer
Procedures to evaluate seismic settlement in dry sand based on CPT Data – an update 1164
F. Yi
Author index 1171
xv
© 2022 the Editor(s), ISBN 978-1-032-31259-0
Preface
This volume contains the contributions to the 5th International Symposium on Cone Penetra
tion Testing, CPT’22, that took place in Bologna (Italy) from June 8th to 10th, 2022. The Sym
posium, built on the successful series of CPT events held in Delft (2018), Las Vegas (2014),
Huntington Beach (2010) and Linköping (1995), was organized by the Italian Geotechnical
Association (AGI) and the University of Bologna, under the auspices of the ISSMGE Tech
nical Committee TC102.
Over the last decades, the use of the cone penetration testing for site investigation has stead
ily grown around the world, causing the CPT to be currently recognized as the major in situ
testing technique in geotechnical engineering practice. Its widespread use has in turn stimu
lated the development of enhanced or new types of CPT equipment, aimed at both improving
the accuracy of measurements and extending the field of application, along with the increasing
demand for a sound understanding of penetration mechanisms and consistent interpretation
methods, suitable for the characterization of a variety of soil deposits.
The CPT symposia have always provided a unique and fruitful forum for the exchange of
new ideas and discussion on key issues within the largest gathering of world’s experts, academ
ics and non-academics, working in the broad and dynamic area of cone penetration testing.
The objective of CPT’22 was therefore to foster a lively debate on current trends in cone pene
tration testing, among Researchers, Practitioners and Manufacturers, in a friendly and cre
ative atmosphere for discussion and collaboration, as tradition of CPT events.
More than 450 Authors, coming from academic institutions, private companies and public
bodies worldwide, contributed to the peer-reviewed papers included in this volume. The Sympo
sium Organizing Committee was especially pleased with the unprecedented large number of con
tributions submitted, exceeding all expectations, and with the general high quality of the papers.
A total of 186 manuscripts were received and 170 of them were finally accepted for publica
tion in the Symposium Proceedings. The papers were sorted into three well-established theme
categories, according to the subject areas typically addressed in CPT events: Equipment and
procedures, Interpretation, Applications. The volume also includes three outstanding keynote
papers and four invited papers presented by renowned experts in specific thematic discussion
sessions on selected key topics.
A significant effort was made to provide the Authors with a rigorous and fair review of the
papers. The Editors are therefore very grateful to the almost one hundred Reviewers, for their
generous and valuable work.
We gratefully acknowledge the support of the Italian Geotechnical Association, which
organized the Symposium in cooperation with the University of Bologna. Special thanks are
extended to the AGI Secretary, Mrs Susanna Antonielli, for her tireless work and patient
assistance. We would also like to express our appreciation to all the Sponsors that helped us
in making this conference a success.
Finally, many thanks to all Keynote Lecturers, Invited Speakers and Authors for their enthu
siastic and proactive response to CPT’22, and for their contribution to this Proceedings volume.
We do hope you will find its content of valuable and long lasting use.
Guido Gottardi
Laura Tonni
xvii
© 2022 the Editor(s), ISBN 978-1-032-31259-0
Symposium organizers
Italian Geotechnical Association, AGI Alma Mater Studiorum – Università di Bologna
Under the auspices of
ISSMGE TC 102 “In-Situ Testing”
xix
© 2022 the Editor(s), ISBN 978-1-032-31259-0
Committees
ORGANIZING COMMITTEE
Laura Tonni – Chair, Università di Bologna, Italy
Sebastiano Rampello – Co-Chair, AGI Delegate, Sapienza Università di Roma, Italy
Claudio Soccodato – AGI General Secretary
Susanna Antonielli – AGI Secretary
Sara Amoroso – Università di Chieti-Pescara, Italy
Giovanni Biondi – Università di Messina, Italy
Giovanni Bosco – Università dell’Aquila, Italy
Simonetta Cola – Università di Padova, Italy
Andrea Dominijanni – Politecnico di Torino, Italy
Laura Govoni – Università di Bologna, Italy
Carmine Gerardo Gragnano – Università di Bologna, Italy
Chiara Iodice – Università della Campania Luigi Vanvitelli, Italy
Maria Clorinda Mandaglio – Università di Salerno, Italy
Michela Marchi – Università di Bologna, Italy
Luca Masini – Sapienza Università di Roma, Italy
Claudia Meisina – Università di Pavia, Italy
Lorella Montrasio – Università di Parma, Italy
Irene Rocchi –Technical University of Denmark, Denmark
Gianpiero Russo – Università di Napoli Federico II, Italy
Ermanno Pagani – Pagani Geotechnical Equipment, Italy
Daniele Spizzichino - ISPRA Istituto Superiore per la Protezione e la Ricerca Ambientale, Italy
Stefano Stacul – Università di Pisa, Italy
Marco Uzielli – Università di Firenze, Italy
Maurizio Ziccarelli – Università di Palermo, Italy
xxi
© 2022 the Editor(s), ISBN 978-1-032-31259-0
International advisory board
Guido Gottardi – Chair, Università di Bologna, Italy

Nicola Moraci – AGI President, Università “Mediterranea” di Reggio Calabria, Italy
Marcos Arroyo – Universitat Politècnica de Catalunya, Spain
Guojun Cai – School of Transportation Southeast University, China
Misko Cubrinovski – University of Canterbury, New Zealand
Jason DeJong – University of California Davis, United States
Heraldo Luiz Giacheti – São Paulo State University, Brazil
Michael Hicks – Delft University of Technology, The Netherlands
An Bin Huang – National Chiao Tung University, Taiwan
Michele Jamiolkowski – Politecnico di Torino, Italy
Junichi Koseki – University of Tokyo, Japan
Tim Länsivaara – Tampere University, Finland
Jong-Sub Lee – Korea University, Korea
Barry Lehane – The University of Western Australia, Australia
Michael Long – University College Dublin, Ireland
Diego Lo Presti – Università di Pisa, Italy
Tom Lunne – Norwegian Geotechnical Institute, Norway
Mario Manassero – Politecnico di Torino, Italy
Alessandro Mandolini – Università della Campania Luigi Vanvitelli, Italy
Diego Marchetti – Studio Prof. Marchetti, Italy
Paul W. Mayne – Georgia Institute of Technology, United States
Paola Monaco – Università dell’Aquila, Italy
Joek Peuchen – Fugro, The Netherlands
John Powell – GEOLABS Limited, United Kingdom
Mark Randolph – The University of Western Australia, Australia
Peter K. Robertson – PK Robertson Inc & Gregg Drilling Inc, United States
Rodrigo Salgado – Purdue University, United States
Fernando Schnaid – Universidade Federal do Rio Grande do Sul, Brazil
Helmut Schweiger – Graz University of Technology, Austria
Paolo Simonini – Università di Padova, Italy
Laura Tonni – Università di Bologna, Italy
Antonio Viana da Fonseca – Universidade do Porto, Portugal
Phil Watson – University of Western Australia, Australia
David White – University of Southampton, United Kingdom
xxiii
© 2022 the Editor(s), ISBN 978-1-032-31259-0
Reviewers
Glenda Abate – Università di Catania, Italy

Sara Amoroso – Università di Chieti-Pescara, Italy
Marcos Arroyo – Universitat Politècnica de Catalunya, Spain
Dusan Berisavljevic – University of Belgrade, Serbia
Silvia Bersan – Crux Engineering BV, The Netherlands
Ilaria Bertolini – Università di Bologna, Italy
Davide Besenzon – Escuela Superior Politécnica del Litoral, Ecuador
Katia V. Bicalho – Federal University of Espirito Santo, Brazil
Giovanni Biondi – Università di Messina, Italy
Giovanni Bosco – Università di Messina, Italy
Lorenzo Brezzi – Università di Padova, Italy
Vincenzo Butticé – Università di Palermo, Italy
Francesco Calvetti – Politecnico di Milano, Italy
Orazio Casablanca – Università di Messina, Italy
Zijun Cao – Wuhan University, China
Francesca Casini – Università degli Studi di Roma Tor Vergata, Italy
Francesca Ceccato – Università di Padova, Italy
Manuela Cecconi – Università di Perugia, Italy
Jared A. Charles – University of Southampton, United Kingdom
Anna Chiaradonna – Università dell’Aquila, Italy
Andrea Ciancimino – Politecnico di Torino, Italy
Matteo O. Ciantia – University of Dundee, United Kingdom
Simonetta Cola – Università di Padova, Italy
Giorgia Dalla Santa – Università di Padova, Italy
Andrea Diambra – University of Bristol, United Kingdom
Francesco Di Buccio – Università di Chieti-Pescara, Italy
Bruno Di Buò – Tampere University, Finland
Giuseppe Di Filippo – Università di Messina, Italy
Marco D’Ignazio – University of Tampere, Finland
Elena Dodaro – Università di Bologna, Italy
James Doherty – University of Western Australia, Australia
Settimio Ferlisi – Università di Salerno, Italy
Alessio Ferrari – Università di Palermo, Italy
Cristiana Ferreira – Faculdade de Engenharia da Universidade do Porto, Portugal
Aligi Foglia – Fraunhofer IWES, Germany
Fabio Gabrieli – Università di Padova, Italy
Andrea Galli – Politecnico di Milano, Italy
Domenico Gaudio – Sapienza, Università di Roma, Italy
Heraldo L. Giacheti – São Paulo State University (Unesp), Brazil
xxv
Domenico Gioffré – Università di Pavia, Italy
Daniela Giretti – Università di Bergamo, Italy
Guido Gottardi – Università di Bologna, Italy
Laura Govoni – Università di Bologna, Italy
Jürgen Grabe – Hamburg University of Technology, Germany
Carmine Gerardo Gragnano – Università di Bologna, Italy
Paolo Maria Guarino – ISPRA Istituto Superiore per la Protezione e la Ricerca Ambientale,
Italy
Michael Hicks – Delft University of Technology, The Netherlands
Chiara Iodice – Università degli Studi della Campania Luigi Vanvitelli, Italy
Maria Iovino – Università degli studi di Napoli “Parthenope”, Italy
Stefan Kreiter – University of Bremen, Germany
Tim Länsivaara – Tampere University, Finland
Valentina Lentini – Università degli studi di Enna “Kore”, Italy
Jean-Sébastien L’Heureux – Norwegian Geotechnical Institute, Norway
Stefania Lirer – Università degli Studi Guglielmo Marconi, Italy
Diego Lo Presti – Università di Pisa, Italy
Tom Lunne – Norwegian Geotechnical Institute, Norway
Claudia Madiai – Università di Firenze, Italy
Rosa Maria Stefania Maiorano – Università degli studi di Napoli “Parthenope”, Italy
Maria Clorinda Mandaglio – Università di Salerno, Italy
Alessandro Mandolini – Università degli Studi della Campania Luigi Vanvitelli, Italy
Diego Marchetti – Studio Prof. Marchetti, Italy
Michela Marchi – Università di Bologna, Italy
Luca Masini – Sapienza Università di Roma, Italy
Rossella Massimino – Università di Catania, Italy
Paul W. Mayne – Georgia Institute of Technology, United States
Claudia Meisina – Università di Pavia, Italy
Lucia Mele – Università di Napoli Federico II, Italy
Fausto Andrés Molina-Gómez – University of Porto, Portugal
Paola Monaco – Università dell’Aquila, Italy
Lorella Montrasio – Università di Parma, Italy
Diane Moug – Portland State University, United States
Zhongkun Ouyang – University of New Haven, United States
Priscilla Paniagua – Norwegian Geotechnical Institute, Norway
Paul José Pinedo Vilcahuamán – Universitat Politècnica de Catalunya, Spain
Nick Ramsey – Fugro Australia, Australia
Sara Rios – University of Porto, Portugal
Peter K. Robertson – PK Robertson Inc & Gregg Drilling Inc, United States
Irene Rocchi – Technical University of Denmark, Denmark
Fabio Rollo – Sapienza Università di Roma, Italy
Alessandra Rosati – Sapienza Università di Roma, Italy
Marco Rosone – Università di Palermo, Italy
Giacomo Russo – Università di Napoli Federico II, Italy
Gianpiero Russo – Università di Napoli Federico II, Italy
Fernando Schnaid – Universidade Federal do Rio Grande do Sul, Brazil
xxvi
James Schneider – US Army Corps of Engineers, United States
Britta Søndergaard – COWI, Denmark
Daniele Spizzichino – ISPRA Istituto Superiore per la Protezione e la Ricerca Ambientale, Italy
Nunziante Squeglia – Università di Pisa, Italy
Stefano Stacul – Università di Pisa, Italy
Francine Chantal Tchamaleu Pangop – Università di Padova, Italy
Laura Tonni – Università di Bologna, Italy
Giuseppe Tropeano – Università di Cagliari, Italy
Marco Uzielli – Università di Firenze, Italy
Phil Watson – UWA, University of Western Australia, Australia
Varvara Zania – Technical University of Denmark, Denmark
Ningning Zhang – RWTH Aachen University, Germany
Maurizio Ziccarelli – Università di Palermo, Italy
Paolo Zimmaro – Università della Calabria, Italy
xxvii
© 2022 the Editor(s), ISBN 978-1-032-31259-0
Sponsors
xxix
Keynote lectures
© 2022 the Author(s), ISBN 978-1-032-31259-0
New CPT methods for evaluation of the axial capacity of driven piles
B.M. Lehane & E.J. Bittar

University of Western Australia, Australia
S. Lacasse, Z. Liu & F. Nadim

Norwegian Geotechnical Institute, Norway
ABSTRACT: High costs associated with offshore foundation installations have provided strong impetus to
the offshore energy sector in the search for more reliable design methods. This paper provides a summary of an
Industry sponsored project that led to the development of new CPT-based design methods for the evaluation of
the axial capacity of driven piles. Particular attention was given to the need for the new methods to be applic
able to large diameter offshore piles given that many existing methods are derived by calibration with capacities
measured in static pile load tests on smaller diameter onshore piles. The basic mechanisms supporting the gen
eral format of the expressions proposed for shaft friction and end bearing in sands and clays are described. It is
shown how the new expressions, which are calibrated against a database of the most reliable load tests reported
in the literature, lead to better predictions of capacity compared to other methods and can also satisfactorily
predict the capacity of piles driven into deposits comprising interbedded layers of sand, silt and clay. Recom
mendations for the prediction of pile displacements at working loads using CPT data are also presented.
1 INTRODUCTION loading direction, re-testing and loading rate. In add

ition, the formulations of most methods do not expli
The popularity of the CPT and the similarity between citly incorporate many characteristics of driven pile
the mode of penetration of a cone and a driven pile behaviour that have been well proven in experimen
have provided strong motivation in the search for tal research programmes (discussed below).
direct correlations between the CPT qt value and axial The Offshore Energy sector recognized the advan
pile capacity. The best known of initial correlations tages of a CPT-based method for axial capacity assess
proposed for pile shaft and base resistance in a range ment but also understood the need to quantify the
of soil types were developed about 40 years ago and reliability of such a method using a database that had
include methods recommended by de Beer (1972), the backing of much of the profession. A Joint Industry
Schmertmann (1978), de Ruiter & Beringen (1979), Project (JIP) was set up in 2014 to achieve this object
Zhou et al. (1982), Bustamante & Gianeselli (1982) ive. The JIP was managed by the Norwegian Geotech
and Van Impe (1986). nical Institute (NGI) with support from the University
These and other methods have since been assessed of Western Australia (UWA) and was sponsored by
using specific databases of static load tests by work Equinor, Ørsted, Lundin, ONGC, Petrobras and DNV
ers such as Briaud & Tucker (1988), Niazi & Mayne GL. A ‘Team of Experts’ worked over a period of 3
(2013), Eslami et al. (2014), Hu et al. (2012) and years with guidance from the JIP steering committee
Amirmojahedi & Abu-Farsakh (2019). These studies to assess the suitability of about 600 pile load tests for
show that there is no consensus of opinion regarding inclusion in the database. A good CPT coverage close
the relative reliability of the methods with some to pile tests was required and only first-time load tests
methods providing good predictions for a given data on driven piles that were statically loaded to
base but poor predictions for other databases. This a displacement of 10% of the pile diameter were
lack of consensus is partly due to significant incon included. The final tests selected make up, what is
sistencies between the respective databases which referred to as, the ‘Unified database’ and comprises 71
arise due to (i) different definitions of ultimate axial test piles in silica sand and 49 pile tests in clay. Full
capacity, (ii) contradictory interpretations of load test details of the ‘Unified database’ and the steps followed
data, (iii) uncertainty in CPT qt values at test pile in its compilation are provided in Lehane et al. (2017).
locations, (iv) inclusion or exclusion of layered/ A subsequent phase of the JIP began in 2017
mixed soil deposits and different pile types and (v) (sponsored by Equinor, Lundin Norway, Ørsted,
separate assumptions relating to the effects of time, ONGC, BP, Total, ExxonMobil, EnBW, EDF,
DOI: 10.1201/9781003308829-1
3
Aramco, SSER and HDEC) with the aim of ii. τf developed in any soil horizon reduces with the
developing new ‘Unified’ methods for the predic distance of that soil horizon from the pile tip (h);
tion of the axial capacity of driven piles in sand this arises largely due to the progressive cycling
and clay. This initiative was motivated by the of sand at any particular level as installation
desire of Industry to replace the 4 CPT-based progresses.
methods for driven piles in sand included in the iii. Smaller diameter piles generate larger frictions
API (2011) recommendations with a single due to constrained dilation under shear at the
method. The term ‘unified’ is employed as the interface.
method was developed with input from the pro iv. Open-ended pipe piles generate lower shaft fric
ponents of the 4 most current CPT-based sand tions than closed-ended piles due to lower levels
methods used offshore. Although no CPT method of displacement imparted to the sand mass.
for driven piles in clay and silt are currently pro v. Ageing effects lead to increases in shaft friction
vided in API/ISO documentation, the inclusion of with time after pile installation.
such a method was seen to be an important step
These characteristics are not incorporated in Eqn.
forward to reduce dependence on laboratory
1 and prompted the development of a new generation
measurements of undrained strength required by
of CPT methods included in API (2011), namely
the existing ‘alpha’ method.
those referred to as Fugro-05 (Kolk et al. 2005),
The JIP was concluded successfully in 2021 and
ICP-05 (Jardine et al. 2005), NGI-05 (Clausen et al.
the methods developed will be incorporated in the
2005) and UWA-05 (Lehane et al. 2005). While all
next versions of the ISO-19901-4 and API
of these newer methods do not include all of the
recommendations.
characteristics listed above, each incorporates
This paper provides an overview of the new CPT
a direct proportional relationship between τf and qc
methods with a focus on justification of the basis of
and a reduction of τf with distance from the pile tip
the formulations employed. A full description of the
methods is provided in Lehane et al. (2020, 2022) (h). These dependencies were first clearly revealed
and in Nadim et al. (2021). The ‘sand’ and ‘clay’ in experiments conducted in 1989 with the Imperial
formulations were calibrated using the Unified data College instrumented pile in medium dense sand at
base and their application to interbedded deposits Labenne, France (Lehane 1992). Figure 1 plots
comprising sand, silt and clay layers is examined in bounds to a large number of equalised radial stress
this paper. Recommendations are also provided to measurements (σ'rc) obtained in these experiments
assist prediction of the displacement of driven piles and displays an obvious trend for the σ'rc/qc ratio to
at working loads using CPT data. be constant in a given soil horizon and to reduce
with increasing distance of that horizon above the
pile tip (noting that there is a direct relationship
between τf and σ'rc via Coulomb’s law).
2 CPT METHODS FOR CAPACITY OF
DRIVEN PILES IN SAND
2.1 Approaches for evaluation of shaft friction

Traditional CPT methods relate the local ultimate
friction (τf) directly to the cone resistance (qt) via
a single factor, β:
Bustamante & Gianeselli (1982), for example,

recommend β values for driven steel tubular piles of
120 and 200 in loose and dense sands respectively
and propose limiting τf values in each of these
deposits to respective maximum values of 35 kPa
and 120 kPa. These and other similar recommenda
tions reported in the literature ignore the following Figure 1. Dependence of equalised radial effective stress
well-known characteristics of driven piles in sand on qc and normalised distance from the pile tip (h/R) at Lab
observed from instrumented pile tests (Lehane et al. enne, France (Lehane 1992) [R=D/2 and τf varies directly
2020): with σ'rc].
i. Ultimate shaft friction (τf) varies is direct propor

tional to the tangent of the interface friction angle The trend shown on Figure 1, which has since
between the sand and pile shaft (tan δ) i.e. shaft been confirmed by Chow (1997) and others, is the
friction is governed by Coulomb’s law. key element of the shaft friction formulations of the
4
UWA-05, NGI-05, ICP-05 and Fugro-05 methods. CPT methods is close to unity. The spread in predic
These shaft frictions correspond to frictions that can tions for the earth pressure approach, as measured by
be developed in the 2 to 4 week period following the CoV for Qm/Qc, is also higher than for the CPT
driving as the methods’ calibrations employed static methods and is indicative of a significantly lower
load test data recorded in the same period. level of reliability. The UWA-05 and ICP-05
methods are the best performing methods with the
lowest CoVs.
2.2 Approaches for evaluation of base resistance
Given the comparable modes of penetration, all his
torical CPT methods for closed-ended driven piles in
sand relate the ultimate end bearing directly to the qc Table 1. Mean ðμÞ and coefficient of variation (CoV) of
value in the vicinity of the pile tip. However, the Qm/Qc for the Unified database of driven piles in silica
ultimate end bearing stress, defined at sand.
a displacement of 0.1D (qb0.1), is typically only
All open & closed-ended piles (total
about 50% of qc for a closed-ended piles driven in
capacity) 71 Piles
homogeneous sand deposits because a displacement Method μ CoV
of the pile tip of order of 1D is required to reach
steady state penetration conditions (as exist during API (2011), K tan 1.66 0.56
cone penetration). In addition to partial mobilisation δ ðβÞ approach
at a displacement of 0.1D, Van Mierlo & Koppejan Fugro-05 0.99 0.40
(1952) recognised the importance of the scale differ ICP-05 1.04 0.27
ence between a pile and a penetrometer. Recognition NGI-05 0.99 0.34
of this effect subsequently led to a variety of pro UWA-05 1.06 0.26
posals for qc averaging techniques, the most popular Unified Method 1.05 0.24
of which relates qb0.1 to a simple average value of qc
values in a zone extending 1.5D above and 1.5D
below a pile tip. 3 UNIFIED CPT METHOD FOR DRIVEN PILES
Compared with closed-ended piles, pipe piles IN SAND
induce lower levels of displacement (or disturbance)
to the sand near their bases during installation. The While the UWA-05 and ICP-05 methods performed
degree of partial plugging (and hence disturbance) relatively well against the ‘Unified database’,
reduces as the pile diameter increases. Consequently, a primary aim of the JIP was to develop a unified
as shown by Gavin & Lehane (2003), the end bear and un-affiliated method that had the support of
ing of large diameter piles under static loading those involved in the derivation of the API β method
(when the plug remains stationary) reduces to that of and the four API CPT methods. The first point of
a bored pile for which qb0.1 is typically about 15% of agreement was that, in line with the findings of
the qc value (Lehane & Randolph 2002). This diam Lehane et al. (1993), and others, the following
eter dependence is incorporated in a number of dif expression should form the basis of the correlation
ferent ways in the Fugro-05, ICP-05, NGI-05 and for peak shaft (τf):
UWA-05 correlations for qb0.1. No set-up of end
bearing resistance has been observed for driven piles
in sand.
where σ'rc is the stationary (equalised) radial
2.3 Performance of CPT methods in API (2011) effective stress, Δσ'rd is the increase in radial effect
The Fugro-05, ICP-05, NGI-05 and UWA-05 CPT ive stress during pile loading (attributed to dilation)
methods included in API (2011) were calibrated and δf is the constant volume sand-pile interface fric
using databases compiled specifically for the devel tion angle. On review of the state-of-the-art, it was
opment of these methods. It is therefore of interest to agreed that the Unified method should incorporate
examine their performance against the ‘Unified data considerations described in the following.
base’, which is fully supported by all researchers
involved in the four methods. The findings from this 3.1 Interface friction angle (δf)
exercise are summarised in Table 1, which presents
statistics for ratios of measured to calculated capaci In the absence of site-specific tests to measure ultimate
ties (Qm/Qc) for the four methods as well as for the interface friction angles (δf), the ICP-05 and UWA-05
earth pressure theory method in API RP2A (2011) methods propose the variation with the mean effective
and the new ‘Unified method’ (discussed below). particle size (d50) indicated by the curve shown in
The results in Table 1 show that, while the earth Figure 2. This variation was deduced from direct
pressure approach in API over-predicts pile capaci shear box tests on steel interfaces with a roughness
ties by an average of 66%, the mean Qm/Qc ratio of typical of industrial piles but that did not include pre
measured to calculated capacity for the four API shearing to large relative displacements. Yang et al.
5
(2010), and others, have since shown that crushing of tests and tests on centrifuge piles with a range of
sand at the pile tip and subsequent shearing during diameters presented in Lehane et al. (2005) show
installation reduces the grading to that of a fine sand. that the cavity strains can be relatively large and that
Interface shear angles measured in the Bishop ring the operational G value is less than G0 for typical
shear apparatus in tests that induced a large level of pile diameters in the ‘Unified database’. The follow
pre-shearing are plotted on Figure 2 confirmed the ing revised approximate expression was deduced
relatively low sensitivity of δf to the initial d50 value. using Δσ'rd measurements on jacked piles and paral
Similar tests reported by Liu et al. (2019) also show lel numerical analyses (Lehane et al. 2020):
that δf has virtually no dependence on the (non-plastic)
fines content of typical siliceous sands and on the
normal stress level. It was therefore concluded that, in
the absence of site specific ring shear interface tests,
adoption of a constant δf value of 29° is a reasonable
assumption for all piles in the load test database.
with dCPT =35.7mm, which is the usual CPT
diameter.
Most piles in the Unified database have
a diameter (D) between 350mm and 800mm. Equa
tion (4) predicts that the increase in peak friction due
to dilation (=Δσ’rd tanδf) for a 20m long pile in
medium dense sand is about 35% for D=350mm but
only 10% for D=800mm. The relative influence of
dilation is clearly an important consideration when
extrapolating from the smaller diameter piles in the
database to larger diameter offshore piles. The rela
tive influence of dilation is also greatest in looser
sands and for longer piles.
3.3 Allowance for partial plugging

Many of the pipe piles in the Unified database (with
D= 550 ± 250mm) experienced partial plugging
during pile driving. The additional shaft and base
capacity that such plugging induces (e.g. see Gavin
& Lehane 2003) needs to be accounted for correctly
in the database analysis to ensure safe extrapolation
to the capacity of full scale offshore piles.
Figure 2. Interface friction angle and d50 for roughness White et al. (2005) used a cavity expansion analogy
typical of industrial piles (centreline average roughness of
to deduce that the equalised lateral effective stress
5 to 20μm) (Barmpopoulos & Ho, 2009); Ho et al., 2011).
acting on the pile shaft (σʹrc) varies with the effective
area ratio (Are) raised to a power of between about 0.3
3.2 Increase in radial stress (Δσʹrd) during pile and 0.45 while Xu et al. (2008) present experimental
loading data showing how qb0.1 varies in proportion to Are.
Are provides a measure of the level of soil dis
The restraint to dilation at the pile shaft during pile placement in any given soil horizon, and is
loading provided by the surrounding sand leads to an defined as:
increase in radial stress on the pile shaft (Δσʹrd) and
hence to the peak shaft friction. This increase can be
assessed from cavity expansion (CE) theory, where
G is the operational shear modulus of the sand mass,
y is the dilation of the sand at the shaft interface and
y/2D is the cavity strain: where IFR is ratio of the change in plug length to
change in pile embedment and Di is the internal pile
diameter. Are varies from unity for a closed-ended or
fully plugged pile to a value of 4t/D for a large diam
eter coring pile (where t is the pile wall thickness).
The UWA-05 and ICP-05 methods assume that As the IFR is not measured routinely, the analysis of
the cavity strain is small enough for conditions to be the unified database substituted the plug length ratio
fully elastic and hence the methods equate G with (PLR) for IFR, noting that the PLR is the average
the small-strain elastic value (G0). However, data IFR during pile driving and equal to the ratio of
from constant normal stiffness direct shear interface the final plug length to the embedded pile length.
6
The PLR is primarily a function of the internal pile information, the global (best-fit) tension to compres
diameter (Di) (e.g. see Gudavalli et al. 2013) and sion ratio (fL) of 0.75 was employed in the new
the following approximate expression was derived method.
based on available records (dCPT =35.7mm):
3.7 Formulation for end bearing
The Unified method relates the ultimate end bearing
(qb0.1) with the effective area ratio (Are) for pipe
piles in line with observations of Gavin & Lehane
(2003) and as adopted in the UWA-05 method. The
The database analysis examined the influence of end bearing formulation of ICP-05 incorporates this
various exponents of Are in the search for a best-fit dependency indirectly by allowing qb0.1 for pipe
expression for τf. piles to reduce with pile diameter.
The Unified method relates qb0.1 with qp, where
qp is the end bearing resistance expected for an
3.4 Time effects ‘imaginary cone’ that has the same diameter as the
The shaft capacity of driven piles in silica sand pile being considered (or equivalent diameter for
increases with time over a period of least one year a pipe pile= (Are/π)0.5). The value of qp is deter
(e.g. Chow et al. 1998, Jardine et al. 2006, Karlsrud mined using a component of a ‘thin-layer’ procedure
et al. 2014, Gavin et al. 2015). Such increases are described by Boulanger and DeJong (2018) and its
not exhibited by bored piles and may be viewed as application in piling calculations is explained by
a recovery process following the ‘trauma’ of driven Bittar et al. (2020). Although qp is a rational and
pile installation (Lim and Lehane 2014, Anusic et al. objective way of obtaining an averaged cone resist
2019). The new CPT method is calibrated using the ance near a pile base of given diameter, its determin
Unified database comprising static load tests with ation requires use of software (freely downloadable
a median equalisation period (or set-up time) of from https://faculty.engineering.ucdavis.edu/boulan
about two weeks. It is therefore likely to underesti ger/research-interests/) which is not appealing to
mate long term capacities and over-estimate short many practitioners. The new ISO-19901-4 recom
term capacities (including driving resistance). mendations therefore suggest taking qp as the aver
age qc value in the zone extending 1.5D above and
below the pile tip, unless conditions at the pile tip
3.5 Formulation for σ'rc are highly variable. Separate studies show that qp for
The observed reduction in the equalised radial the Unified database piles is, on average, 20% higher
effective stress (σ'rc) with the distance from the than qDutch, where qDutch is the CPT resistance deter
pile tip (h) or normalised distance from the pile mined using the ‘Dutch’ averaging technique (Van
tip (h/D) can be described as a power law rela Mierlo & Koppejan 1952)
tionship (as in the ICP-05 and UWA-05 methods)
or as an exponential relationship (proposed by 3.8 Formulation of Unified CPT method in sand
Randolph et al. 1994 and Salgado et al. 2011). and predictive performance
Lehane et al. (2020) show that the bias in the
database, with respect to diameter, of the ratio of The final formulations decided upon for the Unified
measured to calculated capacities (Qm/Qc) can be CPT method for piles in sand (Zone 6 of the soil
removed when a h/D term is used instead of ‘h’ behaviour type chart) are provided in Table 2. When
(as proposed by Alm & Hamre 2001). Lehane these are applied to the Unified database, the average
et al. (2020) also show that while the power law ratio of measured to calculated capacities (Qm/Qc) is
and exponential variations with h/D lead to very 1.05 while the CoV for Qm/Qc is 0.24. These statis
similar statistics for the Qm/Qc ratios, the expo tics are a marginal improvement on those for the
nential relationship tended to predict marginally UWA-05 and ICP-05 methods (see Table 1). The
larger capacities for piles with L/D<20 compared method is, however, considered to be a significant
with the power law form. The power law form step forward as it replaces the four API CPT
was retained as many piles used in the offshore methods with full support of the authors for these
wind Industry have lower L/D values. methods and it incorporates state-of-the-art under
standing of the mechanisms controlling the axial
capacity of driven piles in sand. It should be noted
3.6 Tension compression ratio that the method potentially under-estimates axial
The database piles clearly showed that shaft friction capacity in gravelly sands (Zone 7 of the SBT chart),
measured in tension are, on average, about 75% of where the presence of gravels leads to higher aver
the shaft friction of compression piles. Limited data age qc values. The method is only applicable to piles
exist in relation to differences in the distribution of driven in a conventional manner and should not be
τf along test piles and, in the absence of other used for jacked piles or piles installed by vibration.
7
Table 2. General formulations for Unified method in
sands, silts and clays (see https://pile-capacity-uwa.com/).
Capacity estimate for piles for flexible piles in strain-

softening clays requires load-transfer analysis using load
transfer curves given in API (2011)
Sands: Zone 6 of SBT chart (Ic< 2.1)
qb0.1 = [0.12 +0.38Are]qp; for plugged base (expected when

L/D>5)
=A
L
qp can be taken as the average qc within a zone 1.5D above and
below the pile tip or determined using the procedure described Figure 3. Mean recorded variations of normalised radial
Boulanger & De Jong (12018) and Bittar et al. (2020) total stresses (Si) with h/D measured during installation of
closed-ended instrumented piles.
Clays: Zones 1,2,3 & 4 of SBT chart
4 CPT METHODS FOR CAPACITY OF

Fst =1 for clays with Iz1>0, in Zones 2, 3 and 4 on the SBT DRIVEN PILES IN CLAY
Chart (Ic ≥ 2.6)
Fst =0.5±0.2 clays with Iz1<0, in Zone 1 on the SBT Chart
D*= (D2-Di2)°.5 for an open-ended pile and D*=D for 4.1 Shaft friction
Relationships between shaft friction and the measured
and corrected cone resistances (qc and qt) for driven
p t piles in clay have been proposed for many years e.g.
below the pile tip (closed-ended/ plugged pile) or average
qt to a depth of 20t below the pile tip (large diameter,
Bustamante & Gianeselli (1982), Almeida et al. (1996),
unplugged pile) Lehane et al. (2000, 2013), Eslami & Fellenius (1997)
and Niazi & Mayne (2016). The newly proposed Uni
Silts: Zone 5 of the SBT chart (2.1< Ic <2.6) fied CPT method for clays builds on these methods but
Apply equations as for Zone 6 using corrected qc value is based primarily on observations made in high quality
determined as: instrumented pile experiments conducted by a number
qc = [3.93 Ic2 -14.78 Ic +14.78] qt of institutions, most notably at the Norwegian Geotech
nical Institute and Imperial College London.
Notation The basis of the correlation between shaft friction
h L – z, where L is pile embedment length and z is the and the (corrected) CPT end resistance (qt) for the
depth) new Unified method is the observation that, in any
D Pile outer diameter particular clay, the radial total stress developed
Di Internal diameter of a pipe pile during installation (σri) on the shaft of a closed-
ended displacement pile varies directly with qt and
fs Cone sleeve friction
reduces with the normalised distance from the pile
Fr Friction ratio (expressed as a percentage)=fs/(qt-σv0) tip (h/D). Examples of observed trends are shown on
qt Total (corrected) CPT end resistance (=qc in sands) Figure 3, which plots the mean measured variation
Qtn of the normalised radial total stress in six different
pa=ref. stress=100 kPa clays against h/D, where the normalised radial total
Ic Consistency index, function of Qtn and Fr; see stress (Si) is defined as follows and u0 is the hydro
Robertson (2009) static or ambient pore pressure:
n Stress exponent for Qtn, function of Ic and σ’v0; see
Robertson (2009)
t Pile wall thickness
SBT Soil Behaviour type; see Robertson (2009) The mean trend lines indicated on Figure 3 have
σv0 Total vertical stress a typical standard deviation of 25% and are evidently
σ'v0 Effective vertical stress dependent on the clay type. Following installation,
8
instrumentation on driven piles shows that radial total
stresses fall as excess pore pressure dissipate while
radial effective stresses (σ'r) increase. σ'r reaches
a fully equalised value (σ'rc) after equalisation of
radial total stresses and full pore pressure dissipation.
When a pile is loaded to failure after equalisation,
radial effective stresses reduce attaining a value of σ'rf
at peak local shear stress (τf).
These stages in the life of a driven pile are incorp
orated in the following expression for τf, which is
based on Coulomb’s friction law:
where
Equation (8) illustrates that the relationship between

τf and qt depends on the degree of relaxation of radial
total stresses during equalisation (expressed by Sc/Si),
the relative change in σ’r during load testing (fl) and
the interface friction angle between the pile and clay
(δf). Numerical simulations such as those of Whittle &
Baligh (1988), as well as experiments, indicate that Sc
/Si ratios are lower in sensitive and low OCR clays
where installation causes a significant degree of
remoulding and reduction in effective stress. The Sc/Si
values measured experimentally range from 0.15 in
(highly sensitive) Lierstranda clay to 1.0 in the heavily
overconsolidated London clay. Measured values of fl
are in the range of 0.8 to 1.0 in all clays but average at
about 0.6 in Lierstranda clay. The operational coeffi Figure 4. (a) Average measured variations of τf/qt with h/D
(showing envelope for all clays except Lierstranda clay),
cient of friction (tan δf) is strongly affected by the clay
(b) best-fit mean trend lines.
mineralogy, amongst other factors, and for the six
clays considered in Figure 3 varies by more than
a factor of 2 (from a tan δf value of 0.23 in London has not been the subject of extensive research. The
Clay to 0.55 in Bothkennar clay). assumption of a simple direct relationship between
The compounding effects of differences in respect qb0.1 and the average qt value near the pile tip is con
ive clays of the Sc/Si, fl and tan δf values as well as sidered adequate for practical purposes. On the basis
the Si relationship with h/D (shown on Figure 3) of a review of existing correlations, Lehane et al.
reveal a complex relationship between τf and qt. (2022a) propose the following relationships for typ
Nevertheless, when ratios of τf/qt with h/D measured ical onshore driven piles (D <0.75m)
in instrumented pile tests are plotted, as shown in
Figure 4a, it is evident that a number of compensatory
factors lead to a relatively consistent relationship
between τf /qt and h/D in all of clays apart from Lier
stranda, where the product of Sc/Si and fl is, on aver
age, about four times less than the other clays. Best-
fit mean trend lines for the instrumented data included
in Figure 4a are presented on Figure 4b and, as 5 UNIFIED CPT METHOD DRIVEN PILES IN
shown later, are consistent with best-fit expressions CLAY
for the capacities of the piles in the Unified database.
The development of the new Unified CPT method for
clays recognised the compensatory effects of factors
4.2 End bearing evident on Figure 5a but investigated potential system
The contribution of end bearing to the capacity of atic dependence on (i) clay plasticity index, (ii) soil
driven piles in clay is relatively small and therefore behaviour type (SBT), (iii) CPT sleeve friction (fs) and
9
(iv) vertical effective stress level (σ’v). The calibration
of the method with the Unified database also included
consideration of the aspects detailed below.
5.1 Dependence of τf on loading direction

Compression and tension tests on piles with identical 5.5 End bearing of large diameter pipe piles
configurations in seven different clays in the Unified
database indicate that, in general, there is not a clear Eqn. 9 was employed in the calibration of the rela
dependence of ultimate shaft friction on the load dir tively small diameter piles in the Unified database.
ection. A load direction factor (fL) of unity was found For large diameter offshore piles, the following
to provide a best fit to the full Unified database. expression is considered more appropriate; this was
based on findings of Doherty and Gavin (2011) from
measurements of twin-walled instrumented piles
5.2 Friction on open-ended piles during installation.
Instrumented pile test data reported by Doherty &
Gavin (2011) show that lower levels of soil displace
ment during installation of open-ended piles lead to
lower installation radial total stresses (σri) and lower
ultimate shaft frictions after equalisation. There is, where qt is the average corrected cone resistance
however, a considerable shortage of such data in dif value in the zone extending from the pile tip to
ferent clays and, in the absence of such data, the a depth of 20t below the pile tip (t = pile wall
regression analyses conducted for the Unified thickness)
method assumed the following potential dependen
cies (where Equation 10a is comparable to the Uni 5.6 Formulation of Unified CPT method in clay
fied CPT sand method and Equation 10b is a similar and predictive performance
format to the ICP-05 method for sands):
The formulations for the Unified clay method are
provided in Table 2. These were established follow
ing various optimisation analyses which revealed
that, as inferred from instrumentation pile test data,
τf is primarily correlated to the CPT qt value and the
normalised distance from the pile tip (h/D*). The
Lehane et al. (2017) show that Equations 5 & 6, which analyses also showed that the shaft friction generated
were derived from data for piles in sand, may also in high sensitivity clays (within Zone 1 of the SBT
be used to approximate Are values for pipe piles in clays. chart) is typically 50% of the shaft friction in other
clays (although there is significant variability). The
5.3 Progressive failure formulations developed are remarkably similar to
those deduced independently from instrumented pile
A number of piles in the clay database exhibit post- test data (compare formulae in Figure 5 with those in
peak softening of shaft shear stress. The calibration Table 2). Surprisingly, the fit to the capacities of the
of the Unified method assumed that peak shear stres database piles was not improved by consideration of
ses corresponded to the measured peak axial cap additional factors such as cone sleeve friction, plasti
acity or, if such a peak was not observed, to the axial city index and overconsolidation ratio (inferred by
resistance at a displacement of 0.1D. This approach examination of a qnet/σ’v0 term).
is considered reasonable for the majority of the data The predictive performance for the Unified clay
base (which comprised relatively rigid piles) but is database of the new clay method was compared by
moderately conservative for a few very long piles in Lehane et al. (2022a) with a number of other current
clays where strain softening can be significant. methods. This comparison is summarised in Table 3,
which also lists the main input parameters in the τf
5.4 Silt layers correlation for each method (noting that differences
in the end bearing formulations had little impact on
Occasional silt layers with a soil behaviour type (SBT) calculated capacities). Evidently the CoV of Qm/Qc
index, Ic, in the range 2.05 to 2.6 occur within the clay for the new method is a good improvement on exist
strata of the database. For these deposits, the calibration ing methods and is significantly lower than the cor
process employed the Unified sand method adopting responding CoV for the existing ‘α – su’ method
the equivalent clean sand qt value (qt,sand) for the silt. recommended by API/ISO. It should be noted, how
This was derived using the following relationship ever, that part of the reason for the substantial reduc
which is equivalent to the proposal of Robertson & tion in the CoV with respect to other methods arises
Wride (1998) but adapted to a simplified format and due to application of a separate equation for Zone 1
modified to give a correction factor of unity at Ic = 2.05: (sensitive) clays.
10
Table 3. Comparison of predictive performance for driven Sample predictions for two driven steel pipe pile
piles in clay of the Unified method with other methods (in terms case histories examined are presented in Figure 6a for
of mean and CoVof measured to calculated capacity ratios). a tension test in Oakland California on a 13.3m long,
610mm pile and in Figure 6b for a compression test in
Parameters Mean CoV for Minnevika, Norway on a 40m long, 405mm diameter
Method
controlling τf Qm/Qc Qm/Qc pile. As seen with reference to the plotted Ic profiles in
these figures, a 5.5m thick layer of Zone 3 clay is pre
API (2011) su & su/σ’v 1.05 0.43 sent along the central portion of the pile shaft whereas
Fugro-96 (Kolk & van su, h/D & the stratigraphy at Minnevika comprises alternating
1.04 0.35 1.5m to 2m thick layers of sand and clay. The calcu
der Velde 1996) su/σ’v
lated τf distributions were determined in a simple
ICP-05 (Jardine et al. OCR, σ’v, spreadsheet using the Ic dependent expressions, as sum
1.12 0.55
2005) h/D*, δ & St marized in Table 2; calculations can also be performed
NGI-05 (Karlsrud su, su/σ’v, Ip using UWA freeware at https://pile-capacity-uwa.com/.
1.1 0.36 These τf distributions lead to ratios of measured to cal
et al. 2005) & σ’v
culated capacities (Qm/Qc) of 1.15 at Oakland and 1.16
UWA-13 (Lehane at Minnevika (ignoring potential strain softening).
qt & h/D* 1.12 0.33
et al. 2013)
Fugro-10 (Van Dijk & qnet, h &

0.98 0.37
Kolk 2010) qnet/σ’v
Unified method
qt, h/D* & St 0.99 0.23
(Lehane et al. 2022a)
The Unified method in clay is calibrated on

a relatively small number of pile tests (49) and,
as may be inferred from the instrumented pile
test data, its relatively good predictive perform
ance arises partly because of compensating fac
tors. For example, interface friction angles (δ) for
clays in the Unified database range from 12° to
30°, implying a range in capacities of about 2.5
if τf varied only with tan δ. However, encour
agingly, Lehane et al. (2022a) examined the per
formance of the new method against a different
‘Test database’ comprising 24 pile tests in clay
and found a mean Qm/Qc ratio of 1.09 (i.e. a 9%
over-prediction, on average) and a CoV for Qm
/Qc of 0.22; this CoV is a substantial improve
ment on the CoV of 0.43 determined for the API
α – su method (Table 3).
Figure 5. Stratigraphy and predicted τf distribution for (a)
6 APPLICATION TO LAYERED DEPOSITS tension test on 610-mm diameter, 13.3m long pipe pile in
Oakland and (b) compression tests on 405-mm diameter,
40m long pipe pile at Minneveka.
The Unified CPT methods for sand and clay were
calibrated against test data for piles that were driven
into predominately sand or clay deposits. However,
many deposits encountered in practice contain sand, Bittar et al. (2022) show that the mean and coeffi
silt and clay layers and it is therefore important to cient of variation of the Qm/Qc values for the 23 load
assess the reliability of the methods at such sites. test database were 1.02 and 0.17 respectively i.e. the
Bittar et al. (2022) conducted such a study involving spread in predictions is less than that of the Unified
23 load tests on piles driven into mixed stratigra databases for sand and clay sites. It can therefore be
phies. Ten of these load tests were on piles with concluded that the Unified methods can be applied
diameters between 0.9m and 2m in diameter which with the same level of confidence to layered strati
is considerably larger than the mean D value of graphies as to ‘single soil type’ deposits. However
about 0.4m for the piles in the Unified database. An well instrumented test piles that accurately measure
average of 55% of the layers in these case histories the distribution of shaft friction in layered stratigra
were coarse grained and two cases involved sensitive phies are needed for verification of the method.
zone 1 soils. There is also a great shortage of skin friction data for
11
piles installed in in Zone 5 (silt) of the SBT chart.
Such data would help to resolve the discontinuity in
the expressions for τf at the boundary between Zone
4 and 5 (at Ic=2.6), where τf calculated using the
clay expression is typically double that calculated
using the clean sand correction approach for silt.
7 LOAD DISPLACEMENT RESPONSE USING

THE UNIFIED CPT METHOD
The load-displacement response of piles is normally

predicted in a load transfer analysis where the shear
stress-displacement (τ-w) springs (also called
t-z springs) at various levels along the pile shaft are
scaled in proportion to the ultimate shaft friction (τf)
and the base spring (qb-wb) is scaled with the ultimate
end bearing (qb0.1). The application of the new Unified
CPT method to this approach is examined in the
following.
7.1 Piles in sand

Lehane et al. (2020b) examined the accuracy of pile
displacement predictions for driven piles in the Uni
fied sand pile database sand using the τ-w and qb –
wb load transfer relationships recommended in API
(2011). These relationships are provided in tabular
form in API (2011) but may be expressed as follows,
where wf is the displacement to peak shear stress (τf)
and is assigned a mean value of 0.01D in API
(2011):
Although Equation (14) is not a perfect match

to the API tabulated values when qb/qb0.1>0.5, this
is not a concern as settlement predictions are
required for the serviceability limit state when
applied loads are rarely greater than 70% of the
pile capacity.
API (2011) recommends adoption of a mean wf/D
value of 0.01. However the analyses showed that use
of a constant value did not capture the tendency for
softer pile responses in tension compared with com
pression and for stiffer responses in denser sands
and longer piles. On examination of the trends
shown by the piles in the Unified database, Lehane
et al. (2020b) proposed the following expression for
wf/D:
Figure 6. Comparison of measured pile load-displacement
response with the calculated response using the Unified method
with (i) wf/D given by Equation 15 and (ii) wf/D=0.01. The sym
bols CEC and OET denote closed ended piles tested in compres
sion and open-ended piles tested in tension respectively.
12
where the constant A is 1250 in compression and over the past 30 years. The method is shown to have
625 in tension. higher reliability than existing approaches.
Comparisons of measured and calculated load- As for most empirical methods, further develop
displacement responses are plotted for typical cases ments and improvements are still to come. Such
from the Unified database on Figure 7, where calcula improvements include the need to better quantify the
tions were performed using Equations 13 and 14 com effects of time in various soil types and how the
bined with the Unified method to determine τf and resistance to pile driving as well as the long term
qb0.1. Figure 7 presents calculated responses for capacity can be assessed with more confidence.
a constant wf/D value of 0.01 and for wf/D values at More static load tests on large diameter instrumented
various levels along the pile shaft determined using piles are clearly required while greater understanding
Equation 15. The improved fit obtained using Equa of factors controlling the influence of the installation
tion 15 for tension piles and for longer piles is evident mode and friction in silt is needed.
on Figure 7. Lehane et al. (2020b) show that the error It is hoped that the ‘Unified database’ can be
in calculated displacements at 50% of the ultimate cap expanded on an ongoing basis with each additional
acity is typically about 0.002D and is less than pile test included satisfying the strict selection cri
0.005D. Equation 15 will be included in the next ver teria applied in the database development to date.
sion of the API/ISO recommendations.
ACKNOWLEDGEMENTS
7.2 Piles in clay
Lehane & Bittar (2022) repeated the exercise The authors gratefully acknowledge the funding and
described above for the Unified database of piles in support provided under a Joint Industry Project (JIP)
clay, for which API (2011) recommends the same funded by Aramco, Equinor, Lundin, Ørsted,
format of load transfer curves apart from the allow ONGC, BP, Total, ExxonMobil, EnBW, EDF and
ance for post-peak softening to a shear stress of (0.8 SSER. The significant contribution of the team of
±0.1) τf at a displacement of 0.02D. It was found experts to the JIP, namely Prof. Richard Jardine,
that a combination of the Unified CPT method in Dr. Philippe Jeanjean, Mr. Bas van Dijk, Dr. Mike
clay (Table 2) with these load transfer curves pro Rattley and Mr. Pasquale Carotenuto is much
vides good predictions of the displacements of both appreciated.
the tension and compression piles in the database.
The standard deviation of the difference between
measured and calculated displacements was 0.002D. REFERENCES
Alm T. and Hamre L. (2001). Soil model for pile driveabil
ity predictions based on CPT interpretations. Proc. 15th
8 CONCLUSIONS Int. Conf. on Soil Mechanics and Geotechnical Eng.,
1297–1302.
There are clear advantages to CPT-based design Almeida M.S., Danziger F.A., and Lunne, T. (1996). Use of
the piezocone test to predict the axial capacity of driven
approaches for the assessment of the axial capacity
and jacked piles in clay. Canadian Geotechnical J., 33
of driven piles. The CPT-based methods have the (1), 23–41.
full support of the geotechnical community as they Amirmojahedi M. and Abu-Farsakh M. (2019). Evaluation
remove subjectivity associated with parameter of 18 direct CPT methods for estimating the ultimate
assessment, provide essentially continuous profiles pile capacity of driven piles. Transport Research
of soil behaviour and allow direct rapid and auto Record, 2673(9), 127–141.
mated calculation of capacity at any CPT location. Anusic I., Lehane B.M., Eiksund G., and Liingaard M.
There has been a concerted effort over the past (2019). Evaluation of installation effects on the set-up
decade to first quantify and then increase the reliabil of field displacement piles in sand. Canadian Geotech
nical Journal, 56(4), 461–472.
ity of design methods for piles. This was made pos
API (2011). ANSI/API RP 2GEO: Geotechnical and Foun
sible by the thorough, detailed and careful dation Design Considerations. ISO 19901-4:2003
compilation of a database of static pile load tests (the (Modified), Petroleum and natural gas industries-
Unified database) that had the support of the propon Specific requirements for offshore structures,
ents of recent existing CPT methods and the support Part 4-Geotechnical and foundation design consider
of a large and active group of energy companies ations. 1st edition. Washington, DC: API Publishing
from around the world. Services.
The creation of the Unified database highlighted Barmpopoulos I. and Ho T. (2009). The large displacement
the level of uncertainty associated with design shear characteristics of granular media against concrete
and steel interfaces. Proc Research Symp. Characteriza
methods in common use and prompted the develop
tion & Behaviour of Interfaces, Atlanta, Georgia, USA
ment of the new non-affiliated ‘Unified CPT method’ Behaviour of Interfaces, 16–23.
described in this paper. The new method captures key Bittar E., Lehane B.M., Boulanger R.W., and DeJong J.
characteristics of driven piles, as observed in high (2020). CPT filter to estimate end bearing of
quality instrumented pile test programmes completed closed-ended driven piles in layered sands. Proc. Int.
13
Symp. Frontiers in Offshore Geotechnics, Austin, Jardine, R. J., Chow, F. C., Overy, R. & Standing, J. R.
Texas, 520–528. (2005). ICP design methods for driven piles in sands
Bittar E., Lehane B.M., Nadim F., Liu Z. and Lacasse S. and clays. Thomas Telford.
(2022). Assessment of the unified CPT Design Method Jardine, R. Standing, J. and Chow, F, (2006). Some obser
for Axial Capacity of Driven Piles in Interlayered vations of the effects of time on the capacity of piles
Deposits. Canadian Goetechnical Journal (submitted). driven in sand. Géotechnique, 56 (4), 227–244.
Boulanger, R.W.W., and DeJong, J.T.T. (2018). Inverse fil Karlsrud, K., Clausen, C.J.F. and Aas, P.M. (2005). Bearing
tering procedure to correct cone penetration data for capacity of driven piles in clay, the NGI approach. Proc.
thin-layer and transition effects. Proc 4th Int Symp. on Int. Symp. on Frontiers Offshore Geotechnics, Perth. pp.
Cone Penetration Testing (CPT’18), 21-22 June, 2018, 775–782.
Delft, The Netherlands, 25–44. Karlsrud, K., Jensen, T. G., Wensaas Lied, E. K.,
Briaud J.-L. and Tucker. L.M. (1988). Measured and Pre Nowacki, F., and Simonsen, A. S. (2014) Significant
dicted Axial Response of 98 Piles. Journal of Geotech ageing effects for axially loaded piles in sand and clay
nical Engineering, 114(9), 984–1001. verified by new field load tests. Proc. Offshore Technol
Bustamante, M. and Gianeselli, L. (1982). Pile Bearing ogy Conference, Houston, USA, OTC-25197-MS.
Capacity Prediction by Means of Static Penetrometer doi:10.4043/25197-MS.
CPT. Proc. 2nd European Symposium on Penetration Kolk H. and van der Velde E. (1996). A Reliable Method
Testing, Amsterdam, 493–500. to Determine Friction Capacity of Piles Driven into
Chow F. (1997). Investigations into the behaviour of dis Clays. In Proc. Offshore Technology Conference
placement piles for offshore foundations. PhD thesis, (OTC), Houston, Texas, Pub No. 413532.
Imperial College London (University of London). Kolk H.J., Baaijens A.E., and Vergobbi, P. (2005). Results
Chow, F.C., Jardine, R.J., Nauroy, J.F., and Brucy, F. of axial load tests on pipe piles in very dense sands: The
(1998). Effects of Time on Capacity of Pipe Piles in EURIPIDES JIP. Proc. of the Int, Symp. Frontiers in
Dense Marine Sand. J. Geotech. & Geoenv. Engineer Offshore Geomechanics, ISFOG. Taylor & Francis,
ing, 124, 254–264. London. pp. 661–667.
Clausen C. J. F., Aas P. M. & Karlsrud, K. (2005). Bearing Lehane B.M. (1992). Experimental investigations of pile
capacity of driven piles in sand, the NGI approach. behaviour using instrumented field piles. Imperial Col
Proc. 1st Int. Symp. on Frontiers in Offshore Geotech lege London (University of London).
nics (ISFOG). Balkema, 574–580. Lehane B.M., Chow F.C., McCabe B.M. and Jardine R.J.
De Ruiter J. and Beringen F.L. (1979). Pile foundations for (2000). Relationships between shaft capacity of driven
large North Sea structures. Marine Geotechnology, 5(3), piles and CPT end resistance. Geotech. Engrg., ICE,
267–314. 143, 93–101.
Doherty, P. and Gavin, K. (2011). Shaft Capacity of Lehane B.M., Gaudin C., and Schneider J.A. (2005). Scale
Open-Ended Piles in Clay. J. Geotech. & Geoenv. effects on tension capacity for rough piles buried in
Engrg, 137(11): 1090–1102. dense sand. Géotechnique, 55(10), 709–720.
Eslami, A. and Fellenius, B.H. (1997). Pile capacity by Lehane B.M., Jardine R.J., Bond A.J. and Frank, R. (1993).
direct CPT and CPTu methods applied to 102 case Mechanisms of shaft friction in sand from instrumented
histories. Canadian Geotechnical J., 34(6), 886–904. pile tests. Journal of Geotechnical Engineering,
Eslami A., Tajvidi I. and Karimpour-Fard M. (2014). ASCE,119(1), 19–35.
Efficiency of Methods for Determining Pile Axial Lehane B., Li L. and Bittar E. (2020b). CPT-based
Capacity – Applied to 70 Case Histories in Persian load-transfer formulations for driven piles in sand. Geo-
Gulf Northern Shore. International Journal of Civil technique Letters, 10, 568–574.
Engineering, 12(1), 45–54. Lehane B., Li L. and Bittar E. (2022). CPT-based load-
Hu Z., McVay M., Bloomquist D., Horhota D. and P. Lai. transfer formulations for driven piles in clay (in
P. (2012) New Ultimate Pile Capacity Prediction preparation).
Method Based on Cone Penetration Test (CPT). Canad Lehane, B.M., Li, Y. and Williams, R. (2013). Shaft Cap
ian Geotechnical Journal, 49(8), 961–967. acity of Displacement Piles in Clay Using the Cone
Gavin K., Jardine R., Karlsrud K., and Lehane B.M. Penetration Test. J. Geotech. & Geoenv. Engrg., ASCE,
(2015). The effects of pile ageing on the shaft capacity 139(2):253–266.
of offshore piles in sand. Proc. 3rd Int. Symp. Frontiers Lehane B.M. and Randolph M.F. (2002). Evaluation of
in Offshore Geotechnics, 1, 129–151. a Minimum Base Resistance for Driven Pipe Piles in
Gavin K.G. and Lehane B.M. (2003). The shaft capacity of Siliceous Sand. J. Geotech. and Geoenviron. Engineer
pipe piles in sand. Canadian Geotechnical Journal, 40 ing, ASCE,128(3), 198–205.
(1), 36–45. Lehane, B.M., Liu, Z., Bittar, E., Nadim, F.,
Gudavalli S.R., Safaqah O. and Seo H. (2013). Effect of Lacasse, S., Jardine, R., Carotenuto, P., Rattley, M.,
soil plugging on axial capacity of open-ended pipe piles Jeanjean, P., Gavin, K., Gilbert, R., Bergan-haavik,
in sand. Proc. 18th Int. Conf. Soil Mech. and Geotech. J. and Morgan, N. (2020a). A new CPT-based axial
Eng., Paris, 1487–1490. pile capacity design method for driven piles in
Ho T.Y.K., Jardine R.J. and Anh-Minh N. (2011). Large- sand. Proc. 5th Int. Symp. Frontiers Offshore Geo
displacement interface shear between steel and granular technics, ISFOG-21, 462–477. DFI publications,
media. Geotechnique, 61(3): 221–234. New Jersey.
Jardine R.J., Lehane B.M. and Everton S.J. (1993). Friction Lehane B.M., Liu Z., Bittar E., Nadim F., Lacasse S.,
coefficients for piles in sands and silts. In Offshore site Bozorgzadeh N., Jardine E., Ballard J-C, Carotenuto P.,
investigation and foundation behaviour. Springer. Gavin K., Gilbert R., Bergan-Haavik J., Jeanjean P. and
661–677. Morgan N. (2022). CPT-based axial capacity design
14
method for driven piles in clay. J. Geotech. & Geoenv. serviceability limit states of transportation structure
Engineering, ASCE (under review) foundations. Rep. No. FHWA/IN/JTRP- 2011/03, Indi
Lehane, B.M., Lim, J.K., Carotenuto, P., Nadim, F., ana Dept. of Transportation and Purdue Univ., West
Lacasse, S., Jardine, R.J. and van Dijk, B.F.J. (2017). Lafayette, US.
Characteristics of Unified Databases for Driven piles. Schmertmann J.H. (1978). Guidelines for cone penetration
Proc. 8th International Conf. Offshore investigation and test, performance and design. Report FHWA-TS-78
Geotechnics: Smarter solutions for offshore develop 209, Washington, 145pp.
ments, Society for Underwater Technology, 1, 162–194. Van Dijk B.F.J. and Kolk H.J. (2011). CPT-based design
Lim J.K. and Lehane B.M. (2014). Characterisation of the method for axial capacity of offshore piles in clays. In
effects of time on the shaft friction of displacement piles Proc. of the Int. Symposium on Frontiers in Offshore
in sand. Géotechnique, 64(6), 476–485. Geotechnics II. Taylor & Francis Group, London.
Liu T.F., Quinteros V.S., Jardine R.J, Carraro J. A. H., and 555–560.
Robinson J. (2019). A Unified database of ring shear Van Impe W.F. (1986). Evaluation of deformation and bear
steel-interface tests on sandy-silty soils. Proc. XVII ing capacity parameters of foundations from static
European Conf. Soil Mech. and Geotechnical Engg. Rey CPT-results. Proc. 4th Int. Geotechnical on Field Instru
kjavik, Iceland. doi: 10.32075/17ECSMGE-2019-268. mentation and in-situ measurements, Singapore, 51–70.
Nadim F., Lehane B.M., Liu Z., Bittar E., Bozorgzadeh Van Mierlo, W.C. and Koppejan, A.W. 1952, ‘Lengte en
(2021). New CPT-based method for calculation of axial draagvermogen van heipalen’, Bouw, January.
pile capacity in sand and in clay- Summary and White D., Schneider J., and Lehane B.M. (2005). The
Guidelines. Doc No. 20180089-03-R, Norwegian Geo- influence of effective area ratio on shaft friction of
technical Institute, Oslo, Norway. displacement piles in sand. Proc. 1st Int. Symp.
Niazi F. S. and Mayne P.W. (2013). Cone Penetration Test Frontiers in Offshore Geotechnics, Perth, Australia.
Based Direct Methods for Evaluating Static Axial Cap pp. 741–747.
acity of Single Piles. Geotechnical and Geological Whittle A.J. and Baligh M.M. (1988). A Model for Predict
Engineering, 31(4), 979–1009. ing the Performance of TLP Piles in Clays. Final Report
Niazi, F.S. and Mayne P.W. (2016). CPTu-based enhanced Phase III to sponsors, Dept. of Civil Engineering, Mas
UniCone method for pile capacity. Engineering Geol sachusetts Institute of Technology.
ogy, 212: 21–34. Elsevier B.V. Xu X., Schneider J.A., and Lehane B.M. (2008). Cone
Randolph M.F., Dolwin J., and Beck R. (1994). Design of penetration test (CPT) methods for end-bearing assess
driven piles in sand. Geotechnique, 44(3), 427–448. ment of open- and closed-ended driven piles in siliceous
Robertson, P.K. (2009). Interpretation of cone penetration sand. Canadian Geotechnical Journal, 45(8), 1130–1141.
tests - A unified approach. Canadian Geotechnical J., Yang Z.X., Jardine R.J., Zhu B.T., Foray P. and Tsuha C.H.
46(11):1337–1355. C. (2010). Sand grain crushing and interface shearing
Robertson P.K., and Wride C.E. (1998). Evaluating cyclic during displacement pile installation in sand. Geotechni
liquefaction potential using the cone penetration test. que, 60(6), 469–482.
Canadian Geotechnical Journal, 35(3), 442–459. Zhou J., Xie Y., Zuo Z.S., Luo M.Y. and Tang X.J. (1982),
Salgado R., Woo S. I., and Kim D. (2011). Development of Prediction of limit load of driven pile by CPT, 2nd Eur.
load and resistance factor design for ultimate and Symp. On Penetration testing, Amsterdam, 957–963.
15
© 2022 the Author(s), ISBN 978-1-032-31259-0
Material Point Method simulations of cone penetration and CPT

interpretation
R. Salgado
Purdue University, West Lafayette, USA
V. Bisht
Itasca Consulting Group, Minneapolis, USA
M. Prezzi
Purdue University, West Lafayette, USA
ABSTRACT: It is now possible to realistically simulate cone penetration in either sand or clay using the
Material Point Method (MPM). To do that, the following requirements must be met: (1) a realistic constitutive
model must be used; (2) stress integration of the constitutive model must be accurate; (3) the explicit scheme
for the solution of the governing equation must be accurate, robust and efficient; and (4) shear strain localiza
tion must be correctly captured. The paper provides an overview of how to correctly perform such simulations
and then discusses the implications for the future of CPT interpretation of the availability of accurate, reliable
cone penetration simulation methods.
1 INTRODUCTION 1.2 Methods of cone penetration simulation

To remedy the gap in the theoretical basis for CPT
1.1 The cone penetration test
interpretation, efforts started early. Initial attempts
One of the attractions of the cone penetration test relied on modeling the soil as a rigid-plastic material
(“CPT”) as originally conceived is its simplicity. with either a Tresca or Mohr-Coulomb yield surface
A penetrometer with a simple, axisymmetric shape and applying the notion of a limiting equilibrium to
is pushed into the ground, and the resistance to an axially loaded penetrometer and a surrounding soil
penetration at its tip is measured. Given the simpli mass. The Durgunoglu & Mitchell (1976a) model was
city of the test, one would expect that a theoretical the most successful of these early models, but fell
basis for its interpretation would have been in place short when people tried to use it. A major shortcom
for some time. Unfortunately, what is simple in con ing of the model was its reliance on perfect plasticity
cept is not necessarily so once implemented (Sal and an interpretation based on estimating the friction
gado 2012). The complexity appears when we try to angle of the soil. As discussed elsewhere (e.g., Sal
develop methods of analysis for realistic interpret gado 2012), the mobilized friction angle varies across
ation of CPT results. For example, when we attempt the soil domain in any boundary-value problem
to estimate fundamental state variables or direct involving real soil, so any method based on such an
functions of state variables – such as relative dens over-simplification of soil response is fatally flawed.
ity (“DR”) for sand or undrained shear strength The next main thrust in developing a framework
(“su”) for clay – we find out that there have not for CPT simulations was cavity expansion theory (e.g.,
been rigorous theories to guide us. Salgado et al. 1997; Salgado and Prezzi 2007; Salgado
The two main sources of difficulties in cone and Randolph 2001). Cavity expansion theory was an
penetration simulation are: (1) the modeling of the approximation to the problem based on the require
soil response itself and (2) the method of solving ment that the cone must expand a cylindrical cavity in
the global equation of motion. To address (1), the soil in order to advance. With the exception of Sal
researchers have developed constitutive models gado and Randolph (2001), these efforts concentrated
that are now accurate and reliable, at least in the on the use of simple elastoplastic soil models with
absence of cycling. To address (2), researchers Tresca or Mohr-Coulomb yield criteria, although with
have explored a variety of approaches, which are some modifications attemping to capture the effects of
detailed next. soil nonlinearities. It produced useful relationships
DOI: 10.1201/9781003308829-2
16
(e.g., Salgado and Prezzi 2007) that can be used in move during a computation step, they can be reposi
CPT interpretation, but it could not truly simulate tioned to their original locations before the next compu
a process that is much more involved than a simple tation time step. It may also be advatangeous to
cavity expansion process. consider parts of the grid to either translate in some dir
The strain path method (Teh and Houlsby 2009) was ection or to compress one-dimensionally. This means
proposed for analysis of cone penetration in clay, and it that, whether the grid remains static, moves in some
was a reasonable approximation to the penetration pro direction or is unidirectionally compressed in MPM, it
cess, but its impact was again limited by a simple does not get distorted as in FEM, escaping therefore the
elasto-plastic model with a Tresca yield criterion. difficulties created by mesh distortion.
Early application of the Finite Element Method The second most important implication of the use
(“FEM”) also met with difficulties. In geomechanics of material points instead of Gauss points in an ana
applications, FEM has traditionally followed lysis is that the absolute rigor of FEM with respect
a Lagrangian approach. This means that the nodes of to integration of the constitutive model is, to some
a mesh are tracked throughout the analysis, and it is extent, lost. This happens mainly because Gauss
through node displacements that solutions are integration is no longer possible. This is not fatal,
obtained. Deformation follows from relative node but a number of remedial measures must be taken to
displacements, and stress follows from deformation. limit any error from this partial loss of rigor.
After considerable deformation has occurred, mesh The next sub-section reviews the MPM formula
elements may be so distorted that accurate solutions tion used for the analyses whose results are pre
are no longer possible. In fact, a solution may not sented in this paper and the key steps in an MPM
even be possible, with the analysis crashing instead. computation time step.
The Arbitrary Lagrangian-Eulerian (“ALE”) FEM
approach (Belytschko and Kennedy 1978) and the
2.2 The MPM formulation
material point method (“MPM”) (Schreyer et al.
1994) were proposed to avoid this limitation. In The MPM formulation used in the simulations dis
ALE FEM, mesh distortion is limited by remeshing cussed later is based on the uniform generalized
or node repositioning after some threshold level of interpolation material point method (“uGIMP” or, in
deformation has happened. When nodes are reposi this paper, also “GIMP”) variant of MPM discussed
tioned or a new mesh formed, we need to map the in detail by Bisht and Salgado (2018), Bisht et al.
state varia-bles from quadrature points on the old (2021b) and Salgado and Bisht (2021).
mesh to quadrature points on the new mesh. The Figure 1 shows how space and matter can be dis
remapping can lead to stress states that lie outside cretized for a cone penetration problem. Space can
the yield surface for plasticity models with single be discretized in the form of a computational grid,
yield surface or outside the bounding surface for and the soil and cone can be discretized using mater
bounding surface mod-els. Remapping can be chal ial points. The grid can be a rectangular grid with
lenging with bounding surface models or complex elements sufficiently small around the cone pene
models with several variables to map. In contrast to trometer. Material point discretization is achieved
ALE FEM, MPM does not require remapping of derivatively: the initial number of material points per
variables because the state variables are carried by element is fixed across the grid (e.g., four material
the same material points throughout the computa points per element), and finer discretization is there
tions. The material point method is discussed next. fore achieved where the grid is itself made finer.
How fine the grid must be depends on whether
2 THE MATERIAL POINT METHOD shear strain localization is expected. The smallest elem
ent size in any zone in which shear bands are expected
to form should have a width equal to the width of the
2.1 Fundamentals of the method
shear band. We discuss this in greater detail later.
The material point method can be best understood as There are two other aspects of discretization that
a variant of the finite element method in which Gauss should be briefly discussed. First, the grid can be set up
points are replaced by so called material points. so that part of it moves with the cone penetrometer
A material point (“MP”) may be a point (as in classical during penetration and another part, at the bottom of
MPM), but it may also be a small area (for two- the grid, compresses vertically. Second, when material
dimensional analysis) or small volume (for three- points flow into a grid element for which they are too
dimensional analysis). In contrast with Gauss points, large, they are split in a way that conserves energy,
material points may move within and even across momentum and mass. What these two measures
elements. This key difference between FEM and MPM accomplish is that, when the cone moves down,
has a number of computational consequences. together with the fine grid around it, material points
The most important and most useful implication of located in the coarser grid elements below it will sub
the use of material points instead of Gauss points as divide as they enter the fine grid around the cone, main
used in ALE FEM is that the approach is a pure taining the required level of material discretization and
Lagrangian approach, with material point displacements a sufficient number of material points per element for
being the basis for the solution. Whereas grid nodes do adequate quadrature.
17
Figure 2. Application of tractions to a body in uGIMP
MPM: the tractions are first applied on the edges of the
material points (rep-resented by thick lines) and then, using
finite element-element shape functions, mapped to the
Figure 1. Spatial and material discretization for simulation nodes of the element containing the material point.
of penetration of a flat-tip and a conical-tip penetrometer in
sand (after Bisht 2021).
Displacement or velocity boundary conditions are

straightforward in MPM: the correct displacement or
velocity is directly applied on the outermost grid
nodes corresponding to that boundary. Application
of traction boundary conditions is more involved,
requiring distribution of tractions to the material
point boundaries aligned with the boundary of the
body for every material point so positioned, and then and
redistribution of these tractions to the nodes of any
element containing any part of the material point
(Bisht and Salgado 2018). See Figure 2.
To model a semi-infinite soil medium using a fi
nite grid, silent boundaries are required. Silent bound
aries let energy dissipate just as it would by radiation In Equations (2) through (4), σij is the Cauchy
in a semi-infinite medium. Kellezi (2000) proposed stress, bi is the body force per unit mass, m(P) is the
a “cone boundary” that assumes that a load applied on mass of the material point, Γ is the boundary of the
a free surface of a half-space would propagate within domain Ω of the problem and ti ¼ σij nj is the traction
a volume expanding with increasing depth, which can on an arbitrary point on Γ, with nj being the unit
be approximated as a cone. At the boundary, there is vector normal to Γ at that point. Also appearing in
both a viscosity term to simulate the energy dissipation Equations (2) through (4) are the finite element shape
through radiation and a stiffness term. Bisht and Sal function N(I) at node I, the material point shape func
gado (2018) adapted it for MPM, and it is their for tion S(IP), which maps the value of a variable at the
mulation that is used here. centroid of a material point P to node I, and its gradi
MPM formulations for solving the global equation ent S,j(IP). These last two are defined as:
of motion have typically been explicit. The govern
ing equation for a general analysis applicable to
simple models and any drained simulation is:
and
ðIÞ
The rate of the momentum pi pi(I), external force
ðIÞEXT ðIÞINT
fi and internal force fi at node I, all appear-
ing in (1), are given by (Bisht et al. 2021a):
18
where χðPÞ is the particle characteristic function, usu near incompressibility appears in Tresca solids in the
ally taken to be uniformly equal to 1 across the MP plastic range and at critical state for any soil model. It
and zero outside it: often appears as strong spatial oscillations in the
values of fundamental variables, such as stress, which
appears as a checkerboard pattern when results are
plotted as contour plots.
In the finite element method, volumetric locking
can be avoided by using higher-order elements,
which lead to a sufficient number of variables to
One reason for using an explicit solution scheme
overcome any excess equations from the incom
in MPM are the difficulties in inverting ill-
pressibility constraint. This is not an option in MPM
conditioned stiffness matrices that result when there
are only a few material points in some element or because shape functions for higher-order elements
elements (Wang et al. 2016). can take negative values for some regions in an
The Update Stress Last (USL) explicit time inte element, which in turn can lead to negative mass at
gration scheme, in which the stresses in the material the corresponding nodes from a material point
point are computed after momentum in the grid is located in such regions. The attending complications
updated (Wallstedt and Guilkey 2008), is used here. are discussed in detail in Bisht et al. (2021a).
Explicit time integration is conditionally stable, An effective way to deal with volumetric locking
which means that time increments must be less than in MPM is the use of the so-called non-linear B
some limit for the computations to converge. For method (Hughes 1980). The method is based on the
nonlinear problems, the time step Δt is usually fact that volumetric locking can be prevented by
taken as: reducing the number of constraints imposed by
incompressibility in each element. For example, for
Q4 elements, volumetric locking can be prevented if
a reduced, single-point quadrature rule is used (see
Sloan and Randolph (1982)). The use of single
Gauss point in reduced integration for Q4 elements
where κ is the Courant number (Courant et al. 1967), is sufficient to prevent volumetric locking, but this
elemin is the minimum element size, Ec is a measure strategy is not applicable in MPM.
of material stiffness (the Young’s modulus or the For MPM, Bisht et al. (2021a) used a large-
constrained modulus are common choices, depend deformation formulation of the B method with the
ing on the problem), and ρ is the material density. deformation gradient F split into a volumetric and
For dynamic problems, κ≈0:4 yields suitable results a deviatoric component, with the volumetric compo
for the USL approach, but higher values of κ may be nent calculated only at the element center. The com
used for quasi-static problems (Wallstedt and putational algorithm that results requires that the
ðIPÞ
Guilkey 2008). For quasi-static problems, such as gradient S;j of the material point shape functions
cone penetration, the time step Δt can be increased (material point mapping functions) must be calcu
by mass scaling to reduce the overall computation lated also at the element center. A modified deform
ðPÞ
time. In mass scaling, the material density ρ is artifi ation gradient F at the material point P, obtained
cially increased by a mass scaling parameter η: from combining the deviatoric component of the
deformation gradient with the volumetric component
of it at the element center, is then used in stress
integrations.
This allows the use of a larger time step Δtms : 2.4 A computation time step in an MPM simulation
During a time step in an MPM simulation, we follow
the steps outlined below.
1. Start with initialization of all variables at material
points and grid nodes and set the time t to 0.
2.3 Volumetric locking
2. Determine an appropriate time increment Δt
In numerical solutions to problems in mechanics using Eq. (10).
involving incompressible or nearly incompressible 3. For each material point P:
materials, a phenomenon known as volumetric locking (a) Find every node I around the material point
may occur. Volumetric locking is the insufficiency of P to which nonzero values of material point
variables provided by a grid discretization to allow variables
R must be transferred, i.e., nodes for
solution of all equations. This results from the add which ΩðPÞ N ðIÞ χðPÞ dΩ≠0
itional equation involving strain tensor components (b) Compute mapping functions SðIPÞ using Eq.
resulting from the imposition of the incompressibility (5) and their gradients S; j ðIPÞ using Eq. (6). If
constraint on an integration point. Incompressibility or the node is a hanging node, transfer mapping
19
functions and their gradients to surrounding problem, in which a flat plane-strain punch (a strip
nodes as explained by Bisht et al. (2021a). footing) is forced into a Tresca half-space from the sur
(c) For simulations in incompressible soil, com face, is a good test of the quality of the formulation
pute mapping function gradients S;j ðIPÞ (C) at because the solution to it is known exactly. But limit
element center C. analysis solutions also exist bracketing the bearing cap
acity of footings, strip or square, embedded to various
4. For each node I: depths D in Tresca soil (Salgado et al. 2004). The
(a) Compute nodal mass
values of limit unit bearing capacity qbL are given in
and momentum
Table 1 for strip and Table 2 for square footings. We
(b) Compute nodal external f i (Eq. (3)) and can see very good agreement between MPM and the
internal forces (Eq. (4)). benchmark results, including the value for a strip foot
(c) Compute the rate of change of momentum ing on the surface of the half-space, which is known to
ðIÞ be 2 þ π times the yield stress c of the Tresca material
p_i using Eq. (1) and update nodal momen and is well approximated by the MPM solution.
tum using t The MPM analyses were performed for very stiff
(d) If the node is on a velocity boundary, set Tresca soils. This allowed development of the limit
ðIÞ
pi ¼ mvi� , where vi� is the prescribed velocity. bearing capacity for small footing deflection, before
ðIÞCT any significant change to the geometry of the prob
(e) Compute contact forces fi using the con-
lem could occur. This is important because this is
tact algorithm proposed by Bardenhagen
a feature of the limit analysis solutions used to
et al. (2000, 2001) and update the nodal
benchmark the MPM values of qbL.
momentum using In only three of the ten cases considered in the
5. For each material point P: two tables was the MPM value outside, and only
ðPÞ ðPÞ margin-ally outside, the range defined by the lower
(a) Update material point velocity vi vi þ
and upper bounds from limit analysis.
and position
Þ
ðPÞ Table 1. Value of limit unit bearing capacity qbL of a strip
(b) Compute the velocity gradient vi;j and de footing in Tresca soil from MPM compared to values
ðPÞ
formation gradient Fij obtained using formation gradient.
(c) For simulations in incompressible soil, com
ðP Þ D/B L U MPM
pute the velocity gradient vi;j ðCÞ at the
element center and compute the modified 0.0 5.132 5.203 5.213
ðPÞ
deformation gradient F ij 0.1 5.448 5.548 5.471
(d) Using a suitable objective rate formulation, 0.2 5.696 5.806 5.761
update the stresses and internal variables using 0.4 6.029 6.133 5.977
the algorithms proposed in Nazem et al. (2009). 1.0 6.562 6.657 6.535
(e) Update the material point volume
Þ B = footing width.
6. Adjust (translate/compress/reset) the back-ground

grid. Table 2. Value of limit unit bearing capacity qbL of
7. Update time . If t4tend , go to step 2; a square footing in Tresca soil from MPM compared to
otherwise exit. values obtained using limit analysis.
An objective time rate is required in large-defor D/B L U MPM

mation, large-rotation analyses (Lubliner 2008).
0.0 5.856 6.227 6.149
There is also a specific objective rate formulation
0.1 6.491 7.140 7.063
that must be used to obtain the correct solution of
0.2 6.897 7.523 7.426
a bound-ary-value problem (Lim, J., Salgado, R. and
0.4 7.303 8.104 8.002
Prezzi, M., paper in preparation).
1.0 8.771 9.429 9.296
B = footing width.
3 VALIDATION OF MPM SIMULATIONS OF
RELATED GEOMECHANICS BOUNDARY 3.2 Footing loaded to large settlements
VALUE PROBLEMS
Rather than simply producing the correct values of
limit bearing capacity, the MPM formulation must
3.1 Limit bearing capacity
also reproduce the load response of a footing to large
If an MPM formulation is capable of modeling cone settlements. This problem has been studied using
penetration, then it must be able to also model several methods, including sequential limit analysis
a bearing capacity problem. The Prandtl punch (da Silva et al. 2011), ALE FEM (Kardani et al.
20
2015; Nazem et al. 2006, 2009), a form of ALE therefore no need to perform coupled flow-
FEM known as “remeshing and interpolation tech deformation analyses. Coupled analyses are required
nique with small strain” (RITSS) (Wang et al. 2013) for clay, as we will discuss later.
and MPM (Iaconeta et al. 2019; Sołowski and Sloan To properly simulate penetration in sand, a realistic
2015; Woo and Salgado 2018). sand constitutive model is required. The model must
In the validation simulations, a 1-m wide footing be able to accurately reproduce soil element response
was modeled as a rigid material, and the soil was mod in triaxial compression, triaxial extension and simple
eled as a weightless Tresca material with E = 100 kPa, shear loading. A model satisfying this requirement is
v ¼ 0:49, ρ ¼ 1:0g=cm3 and su = 1 kPa (the same that of Loukidis and Salgado (2009). The analyses
parameter values used by Wang et al. 2013). The prob reported here were performed using this model.
lem domain was discretized using an irregular back Model parameters and the test data required for
ground grid with a minimum element size equal to their determination are listed in Table 3. Also given
B/20 near the base of the footing and a maximum in the table are the values of model parameters for
element size equal to B/80 away from the footing. Ini Ot-tawa sand. It is for Ottawa sand that MPM simu
tially, each element contained 4 material points in a lations were performed. Although the number of
2 × 2 configuration. The size of the domain was 10B in parameters (22) may seem daunting, only 5 param
both the horizontal and vertical directions when meas eters (h1, h2, elim, μ and kh) are determined using
ured from the center of the base of the footing. The a trial and error approach. The remaining 17 can be
Courant number κ was taken as 0.3. To simulate quasi- directly determined by fitting the model equations to
static conditions, the vertical penetration velocity the data. Most parameters can be determined using
increased linearly from 0 to 0.02 m/s over 1 s and then routine tests such as drained and undrained TXC,
remained constant at 0.02 m/s after 1 s. while others can be reasonably assumed based on
Comparison of the MPM formulation of Bisht the corresponding values for similar sands.
et al. (2021a), noted in the legend as “B-bar GIMP,”
with a variety of other methods is shown in Figure 3.
Table 3. Constitutive model input parameters for Ottawa
The favorable comparison with the high-quality sand (after Loukidis and Salgado 2009).
ALE FEM simulation (labeled “RITSS” in the
figure) is an important validation result. Symbol Value Test(s)
Small-strain(“elastic”) ν 0.15 compression

parameters w/unloading
Cg 611 BE or RC
ng 0.437 BE or RC
γ1 0.00065 RC or TX
α1 0.47 TXCU
Critical state �c 0.780 TXC
λ 0.081 TXC
ξ 0.196 TXC
Mcc 1.21 TXC
Bounding surface kb 1.9 TXC
Dilatancy D0 1.31 TXC
kd 2.2 TXC
Plastic modulus h1 2.2 TXC
h2 0.240 TXC
elim 0.81 TXC
μ 1.2 TXCU
Stress-induced c1 0.71 TXE
anisotropy c2 0.78 SS
ns 0.35 SS
Figure 3. MPM GIMP simulations of a footing pushed to Inherent anisotropy α 0.31 TXE
large penetration in an elasto-plastic Tresca soil compared kh 0.39 TXE
with the solution of various other methods of analysis
Yield surface radius m 0.05
(after Bisht et al. 2021a).
TXC: triaxial compression; TXE: triaxial extension;
TXCU: undrained triaxial compression; SS: simple shear;
4 CONE PENETRATION IN SAND RC: resonant column; and IC: isotropic compression.
4.1 Constitutive model

4.2 Cone penetration simulation scheme
Simulation of cone penetration in sand using MPM
is less complex than in clay largely because drained A fixed time step size – determined using Eq. (10) –
penetration can be assumed a priori, there being was used for all cone penetration simulations. For us
21
to use Eq. (10), certain variables and parameters must The grid used to perform the simulations resem
be evaluated. As will be discussed in detail in the next bles that shown in Figure 1. It has a moving grid
sub-section, the minimum element size elemin used for region that moves down with the cone penetrometer
the simulations in this study was 5.95 mm, and the and a compressible region at the bottom (Woo and
relative density DR for all tests was ~90% (implying Salgado 2015). The nodes at the bottom of the com
a dry density ρof 1.72 ton/m3). Using a reasonably pressible grid do not move, but those above them do.
high mean stress p’ of 10 MPa and a void ratio e of To capture shear bands that develop near the cone
0.524, corresponding to a relative density DR of 90%, base and cone shaft, the smallest element size was
the small-strain shear modulus Gmax is equal to 800 chosen to be roughly equal to the shear band width
MPa using the Hardin and Richart (1963) correlation. observed in sands (e.g., Alshibli and Hasan 2008;
Assuming κ ¼ 0:7 and η ¼ 25, a scaled time step Alshibli and Sture 1999; Desrues and Viggiani 2004;
Δtms ¼ 2:0 x 10-5 s was computed. Simulations per Salgado et al. 2017; Tehrani et al. 2017).
formed using smaller κ and smaller η values produced
nearly identical results. A fixed time step
4.3 Validation using calibration chamber tests
Δtms ¼ 2:0 x 10-5 s was therefore considered accept
able and was used in the simulations. The calibration chamber with a viewing window to
Penetration was simulated by imposing on the enable digital image correlation (DIC) analysis of
cone penetrometer successive displacement incre penetration tests was used to perform two cone pene
ments. The same rate of penetration is ideally used tration tests in dense samples subjected to two differ
in computations as in reality, but to reduce computa ent levels of vertical effective stress.
tion time, a greater penetration velocity was used in The chamber, shown in Figure 4, is a half-
the simulations. This is possible so long as penetra cylindrical calibration chamber with a diameter
tion remains quasi-static, so that there is no rate Dc of 1680 mm and a height of 1200 mm. The tests
dependence in the value of qc. Table 4 details the were performed under BC3 calibration chamber
parameters for the MPM analyses reported in this boundary conditions (Ghionna and Jamiolkowski
paper. 1991; Salgado et al. 1998) by using an inflatable air-
rubber bladder at the top of the sample and rigid lat
eral walls. Three cameras with a resolution of 5
megapixels (SV5M10 complementary metal-oxide
Table 4. General numerical scheme used for the MPM semiconductor CMOS cameras, EPIX, Buffalo
simulations of cone penetration in sand (after Bisht 2021). Grove, Illinois) were used to take images during test
ing (GalvisCastro et al. 2019; Salgado et al. 2016;
Feature Implementation
TovarValencia et al. 2018).
Type of MPM uGIMP
Time integration Explicit: Δt ¼ 2 x 10-5 s
Mass scaling factor: 25
Mesh Structured irregular grid divided
into moving and compressible
regions; Q4 element with initially
four material points; element size
approximately equal to the width
of the shear band
Penetration velocity Vcone = 0–10 cm/s for t ≤ 1s;
10 cm/s for t > 1s
Initial stress field Initial vertical stress known; K0 =
0.45 assumed
Contact model Coulomb friction model with μf =
0.3 Figure 4. DIC calibration chamber at COFFEE, Purdue
Stress integration Explicit with automatic sub- University: front view, showing viewing windows.
algorithm stepping and error control
Objective stress rate Modified kinematic logarithmic Figure 5 shows a comparison of measured qc with
and strain measure rate; Hencky strain the qc values computed using MPM for two values
Volumetric locking Locking not observed of vertical effective stress at the level of the cone.
Constitutive model Loukidis and Salgado (2009) sand The MPM simulations capture the gradual rise of
model; Parameters calibrated to qc from the time the cone penetrometer enters the
Ottawa sand used
soil, which is also observed in the real experiment,
ρ: density; Δt = time step size; Q4: linear quadrilateral and also the plateau that forms in the qc versus depth
elements; K0: coefficient of lateral earth pressure at-rest; plot once boundary effects are no longer a factor.
mps.: material points; Vcone: penetration velocity of cone; The experiments have not been repeated for confirm
and μr coefficient of friction. ation, but the comparisons of the plots from the
22
MPM simulation with those obtained from the cali- 5 CONE PENETRATION IN CLAY
bration chamber test are encouraging.
5.1 Formulation
Simulation of cone penetration in clay requires
a coupled, hydro-mechanical, effective stress ana
lysis. Such an analysis allows the calculation of
effective stress and pore pressure everywhere in the
soil domain as the penetration progresses. It is
accomplished in the analysis discussed here using
a single material point to represent both the soil
matrix and water and an explicit scheme to solve the
governing equations, which are in terms of the vel
ocity of the soil matrix and the velocity of water as
the primary variables (Bisht et al. 2021b). Incom
pressibility constraints in the soil matrix are resolved
using the non-linear B-bar method, and pore pres
sures are computed at element centers.
A coupled formulation can be used to handle
penetration under partial drainage conditions.
Undrained conditions need not be assumed. If we do
assume undrained conditions, a simplification of the
Figure 5. Comparison of MPM simulations of cone pene governing equations is possible. The governing equa
tration tests in a calibration chamber with the results of tions are derived from consideration of momentum of
those tests (modified after Bisht 2021). the water, momentum of the soil matrix, and inter
action of the soil particles and water though a drag
resistance between the two due to a relative velocity.
What a DIC calibration chamber enables us to do
that a regular calibration chamber does not is the
measurement of displacement fields. Figure 6 shows 5.2 Validation
a comparison of the horizontal and vertical displace The MPM formulation has been validated using,
ment increments measured using a DIC analysis of among other benchmarks, the one-dimensional con
images collected during a small downward move of solidation theory of Terzaghi (1943). We show that
the cone penetrometer to those computed using here for illustration of the quality of the analysis
MPM. The comparison shows that MPM can effect framework, but point readers to Bisht et al. (2021b)
ively capture displacement fields around the advan for a more extended discussion of the accuracy and
cing cone penetrometers in addition to also robustness of the numerical schemes.
approximating closely the cone resistance values. Figure 7(a) shows a soil wall with height equal to
1.0 m and width equal to 0.075 m. An MPM grid was
superposed on the wall. The grid is a structured regu
lar grid with linear quadrilateral (Q4) elements with
dimensions 0.025 m × 0.025 m. Each element con
tained four material points initially. The Young’s
modulus E of the soil is 10,000 kPa, its Poisson’s
ratio is 0.3, soil particle density is 2.67 ton/m3, the ini
tial porosity is 0.4, and the hydraulic conductivity is
10−3 m/s. The density of water is 2.67 ton/m3 and its
bulk modulus is 2.2 GPa. An initial pore pressure of
10 kPa was assigned to the soil. A fixed boundary
condition was applied at the bottom, and roller bound
aries restricting horizontal movement were applied to
the sides. Impermeable boundary conditions were
applied at the bottom boundary and the two side
boundaries. The consolidation test was performed by
applying a 10 kPa traction at the top surface and
allowing the water to drain from the top. A fixed time
step size equal to 1 x 10-5 s was used in the MPM
Figure 6. Comparison of vertical and horizontal displace simulation of the compression of the soil wall.
ment increments – du and dv – obtained from DIC analysis Figure 7(b) shows that the coupled MPM formu
of the chamber test and MPM simulations of the same test lation closely matches the results obtained using the
(modified after Bisht 2021). traditional Terzaghi solution.
23
Figure 7. One-dimensional consolidation test: (a) slide of
soil undergoing 1D consolidation; and (b) a comparison of
the results obtained using the coupled MPM formulation Figure 8. Grid and material point discretization for simula
and Terzaghi’s solution after (Bisht et al. 2021b). tion of cone penetration at a depth of 9.6 m for the cone
penetration tests of Landon (2007) (after Bisht et al.
2021b).
5.3 Simulation of a real CPT in clayey soil

We use the two CPTs performed by Landon (2007), surface, but penetration just into the clay layer.
which were located approximately 10 m apart, to Accordingly, the initial position of the cone for the
test the MPM formulation for clay. The cone resist analysis was 5dc below the upper boundary of the
ances measured in these two tests at a depth of clay layer, which is at a depth of 9.6 m from the
9.6 m were approximately 580 kPa and 730 kPa. ground surface.
Soil samples were collected from several depths. The soil was discretized using a mesh of dimen
Based on testing done on these samples, the soil sions 20dc in width by 25dc in height. The mesh was
state at a depth of 9.6 m was estimated to be defined built with linear quadrilateral elements, each initially
by an initial vertical effective stress σ0 v0 ¼ 100 kPa, containing 2 x 2 material points.
initial vertical total stress σv0 ¼ 175 kPa, initial hori Figure 9 shows the results of the MPM simulation
zontal effective stress σ0 h0 ¼ 65 kPa, and overconso compared to the two values measured in each CPT
lidation ratio OCR = 2.2. The groundwater level sounding at a depth of 9.6m. The comparison is very
was at a depth of 1.7 m. The hydraulic conductivity encouraging.
varied from 10−10 to 10−9 m/s.
The constitutive model used in the simulations is
that of Chakraborty et al. (2013). The model contains
a total of 22 parameters, most of which, as in the case
of the sand model discussed earlier, can be determined
through simple inspection and curve fitting of equa
tions to data obtained from routine laboratory tests.
The analyses were performed using the calibrated
values of Boston Blue Clay (“BBC”) that are given in
Chakraborty et al. (2013), except that strain-rate inde
pendence is assumed because the problem is quasi-
static.
Figure 8 shows both space and material point dis
cretization of the simulations. The default four mater
ial points per elements were used. The same general
approach followed for the grid in the analysis of cone
penetration in sand was repeated for clay. A structured
irregular grid with both a moving and a compressible
mesh zone was used to discretize the do-main.
A standard cone with diameter dc = 35.7 mm and
cone tip height hc = 30.9 mm (apex angle = 60°) was
assumed. Axisymmetric conditions were assumed, and
thus only half the cone was discretized. Figure 9. Comparison of cone resistance qc, during
The goal of the analysis was not to produce the a penetration path 5dc long, with the values measured in
entire history of penetration from the ground two CPT soundings (after Bisht et al. 2021b).
24
6 IMPLICATIONS FOR FUTURE OF CPT can be determined. The “equations” in this case are
INTERPRETATION whatever relationships are available between the
knowns (qc and any other measurement) and the
6.1 Importance of MPM in future development of unknowns. The unknowns are, in sands, variables
CPT interpretation approaches such as the relative density or void ratio, the effect
ive stress state and soil fabric. In other words, the
This paper has shown that MPM, combined with real unknowns would be state variables. The relation
istic constitutive models, is capable of accurate predic ships would be equations developed based on MPM
tions of cone resistance in both sandy and clayey soils. simulations between cone resistance and the state
What does that mean for CPT interpretation in the variables. Parameters in these equations would be
future? It depends on how far in the future, and it the soil intrinsic variables, those related to factors
depends on what type of interpretation we are like mineralogy, particle size distribution, and the
pursuing. like, as well as critical state parameters. Intrinsic
variables would be determined from routine labora
tory tests.
6.2 Limitation of MPM in the present and in the
Perhaps the greatest impact of the availability of
near term
reliable MPM simulation capabilities will be on
At present, MPM with finite-size material points, and interpretation in clay. Relationships already exist
even, to a lesser degree, with material points that are relating qc to relative density and lateral effective
actual points, is computationally expensive if accur stress, for example, in sands (Salgado and Prezzi
ate, realistic simulations are desired. It is important 2007). However, interpretation of CPT results in
to understand that much faster simulations can be clay has remained linked to a Tresca-linked view of
done, and can be found in the literature, but not with clay, with relationships between qc and su instead of
the level of discretization required for accuracy. At between qc and void ratio, effective stress and fabric.
the time of this writing, simulation of an entire CPT In these traditional relationships, a cone factor Nk is
starting from the ground surface is cost-prohibitive in defined to relate qc to su, but how should su be
an engineering design setting. However, cloud com defined in such a relationship? Should it be that
puting is expected to ease the computational burden obtained under triaxial compression, triaxial exten
significantly in the next several years. sion or simple shear loading? Additionally, some
This means that routine application of MPM to what awkward relationships must be resorted to
simulate entire soundings and the richness of inter relating Nk to a rigidity index defined in terms of
pretation that would result from that lies in the more a modulus and a shear strength that are again not
distant future. However, it is entirely possible to defined. An interpretation method leading to the
compute cone resistance locally. For example, if the values of state variables would be preferable.
interest is in a specific layer with a certain soil exist
ing with a certain state, then that is a simulation that
6.4 Direct interpretation of CPT results
is practically possible to perform. The computations
done earlier for a clay layer illustrate this. Direct interpretation occurs in a design application
The other use of MPM is to perform simulations context, instead of a soil mechanics or soil character
ahead of time for various combinations of soil intrin ization context. To develop a direct CPT method, it
sic and state variables, and then use the results of is necessary that the soil response for the specific
these simulations to derive correlations of cone design problem can be correlated with qc and/or
resistance with these variables or to feed them to AI other CPT measurement. If the design problem can
tools. There are two types of relationships that can be itself simulated numerically, we have an entirely
be developed: direct and indirect relationships. science-based correlation that can be used in design.
Indirect relationships constitute the basic tool for Otherwise, the correlation will be semi-empirical, as
pure CPT interpretation: from qc and other measure liquefaction potential assessment methods must still
ments made during a CPT, obtain the values of the be. Table 5 contains three of the most important
soil state variables. Direct relationships allow you to design problems to which the CPT is routinely
obtain estimates of quantities of direct interest in applied. Foundation loading can already be effect
design, such as unit shaft and base resistance of ively simulated using the finite element method to
a pile or liquefaction resistance from cone resistance settlement levels of interest in design (see, e.g., Han
and any other CPT measurements. et al. 2017; Loukidis and Salgado 2011; Salgado
et al. 2017). MPM can be used for problems in
which significant mesh distortion will render FEM
6.3 Indirect interpretation of CPT results
analyses not feasible. Liquefaction cannot as yet be
In indirect CPT interpretation methods, we are inter simulated. The main obstacle is the unavailability of
ested in evaluating soil state from CPT measure a constitutive model that can properly simulate the
ments. The thinking is somewhat akin to that of soil response under repeated cycling, so methods
solving a system of equations (Salgado 1993). If will for now and in the near future remain semi-
there are enough equations, the unknown variables empirical.
25
Table 5. Direct interpretation of CPT results. soil state variables and soil type, will also be better
taken into account with the more advanced com
Design Basis of mand of the mechanics of cone penetration.
problem Target variables relationships
Shallow Limit unit bearing Computational simu REFERENCES

foundation capacity qbL; settle lations using MPM
design ment w in service or FEM Alshibli, K. A. & Hasan, A. 2008. Spatial variation of void
Deep foun- Ultimate unit base Computational simu ratio and shear band thickness in sand using X-ray com
dation resistance qb,ult; lations using MPM puted tomography. Géotechnique 58(4): 249–257.
design limit unit shaft resist or FEM Alshibli, K. A. & Sture, S. 1999. Sand Shear Band Thick
ance qsL ness Measurements by Digital Imaging Techniques.
Liquefaction Cyclic resistance Semi-empirical Journal of Computing in Civil Engineering 13(2):
potential ratio CRR because liquefaction 103–109.
assessment cannot yet be accu Bardenhagen, S. G., Brackbill, J. U., & Sulsky, D. 2000.
rately simulated The material-point method for granular materials. Com
puter Methods in Applied Mechanics and Engineering
187(3– 4): 529–541.
Bardenhagen, S. G., Guilkey, J. E., Roessig, K. M.,
Brackbill, J. U., Witzel, W. M., & Foster, J. C. 2001. An
improved contact algorithm for the material point
7 CONCLUSIONS method and application to stress propagation in granular
material. CMES - Computer Modeling in Engineering
The main purpose of this paper was to review our and Sciences 2(4): 509–522.
ability to simulate the cone penetration process. For Belytschko, T. B. & Kennedy, J. M. 1978. Computer
realistic simulation of cone penetration in both sand models for subassembly simulation. Nuclear Engineer
and clay, it is necessary to have (1) realistic, robust ing and Design 49(1–2): 17–38.
constitutive models; (2) realistic, robust constitutive Bisht, V. 2021. Cone penetration analysis using the Mater
model stress integration schemes; and (3) accurate, ial Point Method. Purdue University.
efficient, robust numerical schemes for integration of Bisht, V. & Salgado, R. 2018. Local transmitting boundar
ies for the generalized interpolation material point
the governing equation for the cone penetration pro method. International Journal for Numerical Methods
cess. The paper focused on prong 3 of the simulation in Engineering 114(11): 1228–1244.
framework: the use of the u-GIMP MPM formula Bisht, V., Salgado, R., & Prezzi, M. 2021.a. Simulating
tion to simulate the penetration process. penetration problems in incompressible materials using
MPM replaces Gauss points in Lagrangian FEM the material point method. Computers and Geotechnics
formulations by material points. The material points Elsevier Ltd 133(December 2019): 103593.
are not affixed to the finite element constituting the Bisht, V., Salgado, R., & Prezzi, M. 2021.b. Material Point
mesh or grid into which a body is discretized but Method for Cone Penetration in Clays. Journal of Geo-
rather can move through the grid, even crossing technical and Geoenvironmental Engineering 147(12):
1–16.
elements. The material points in fact are the discret Bisht, V., Salgado, R., & Prezzi, M. 2021.c. Simulating
ization of the body, whereas the grid is the discret penetration problems in incompressible materials using
ization of space. the material point method. Computers and Geotechnics
There are certain consequences from this funda 133: 103593.
mental difference between the two methods. Two of Chakraborty, T., Salgado, R., & Loukidis, D. 2013.
the most consequential are: (1) integration is not as A twosurface plasticity model for clay. Computers and
accurate as in FEM; (2) some techniques used in Geotechnics 49(765): 170–190.
FEM to avoid pathologies, such as volumetric lock Courant, R., Friedrichs, K., & Lewy, H. 1967. On the Par
ing, are not effective in MPM; (3) because the grid tial Difference Equations of Mathematical Physics. IBM
Journal of Research and Development 11(2): 215–234.
can always be reset between computation time incre Desrues, J. & Viggiani, G. 2004. Strain localization in
ments, mesh distortion does not impact calculations, sand: an overview of the experimental results obtained
and MPM simulations can progress to very large dis in Grenoble using stereophotogrammetry. International
placement levels. Journal for Numerical and Analytical Methods in Geo-
The paper illustrated some of the validation exer mechanics 28(4): 279–321.
cises that the MPM computational framework dis Galvis-Castro, A. C., Tovar-Valencia, R. D., Salgado, R., &
cussed here has been subjected to and successfully Prezzi, M. 2019. Effect of loading direction on the shaft
passed. These validations have shown that accurate resistance of jacked piles in dense sand. Geotechnique 69
cone penetration simulations are possible. The (1).
Ghionna, V. N. & Jamiolkowski, M. 1991. A critical
impact of this capability is yet to come. In the future, appraisal of calibration chamber testing of sands. Proc.,
the ability to accurately, realistically simulate cone 1st Int. Symp. on Calibration Chamber Testing Elsevier,
penetration and other processes involved in the Amsterdam, The Netherlands 13–39.
direct or indirect interpretation of CPT results will Han, F., Salgado, R., Prezzi, M., & Lim, J. 2017. Shaft and
lead to high-quality interpretation methods. Other base resistance of non-displacement piles in sand. Com
factors in interpretation, such as spatial variability of puters and Geotechnics 83: 184–197.
26
Hardin, B. O. & Richart, J. F. E. 1963. Elastic Wave Veloci Salgado, R. & Prezzi, M. 2007. Computation of cavity
ties in Granular Soils. Journal of the Soil Mechanics expansion pressure and penetration resistance in sands.
and Foundations Division 89(1): 33–65. International Journal of Geomechanics 7(4): 251–265.
Hughes, T. J. R. 1980. Generalization of selective integra Salgado, R. & Randolph, M. F. 2001. Int. J. Geomech.
tion procedures to anisotropic and nonlinear media. 2001.1:175-192. International Journal of Geomechanics
International Journal for Numerical Methods in Engin 1(2): 175–192.
eering 15(9): 1413–1418. Salgado, R., Tovar, R. D., Tehrani, F. S., Castro, A. G.,
Iaconeta, I., Larese, A., Rossi, R., & Oñate, E. 2019. Han, F., & Prezzi, M. 2016. Effect of surface roughness
A stabilized mixed implicit Material Point Method for on the shaft resistance of non-displacement piles embed
non-linear incompressible solid mechanics. Computational ded in sand. Géotechnique 66(5): 386–400.
Mechanics Springer Berlin Heidelberg 63(6): 1243–1260. Schreyer, H. L., Sulsky, D., & Chen, Z. 1994. A particle
Kardani, M., Nazem, M., Carter, J. P., & Abbo, A. J. 2015. method for history-dependent materials. Computer
Efficiency of high-order elements in large-deformation Methods in Applied Mechanics and Engineering 118:
problems of geomechanics. International Journal of 179–196.
Geomechanics 15(6): 04014101. da Silva, M. V., Krabbenhoft, K., Lyamin, A. V., &
Kellezi, L. 2000. Local transmitting boundaries for transi Sloan, S. W. 2011. Rigid-plastic large-deformation ana
ent elastic analysis. Soil Dynamics and Earthquake lysis of geotechnical penetration problems. 13th Inter
Engineering 19(7): 533–547. national Conference of the International Association for
Landon, M. M. 2007. Development of a non-destructive Computer Methods and Advances in Geomechanics
sample quality assessment method for soft clays. (IACMAG) 42–47.
Loukidis, D. & Salgado, R. 2009. Modeling sand response Sloan, S. W. & Randolph, M. F. 1982. Numerical predic
using two-surface plasticity. Computers and Geotech tion of collapse loads using finite element methods.
nics Elsevier Ltd 36(1–2): 166–186. International Journal for Numerical and Analytical
Loukidis, D. & Salgado, R. 2011. Effect of relative density Methods in Geomechanics.
and stress level on the bearing capacity of footings on Sołowski, W. T. & Sloan, S. W. 2015. Evaluation of mater
sand. Géotechnique 61(2): 107–119. ial point method for use in geotechnics. International
Lubliner, J. 2008. Plasticity Theory. Dover Publications, Journal for Numerical and Analytical Methods in Geo-
INC. mechanics 39(7): 685–701.
Nazem, M., Carter, J. P., Sheng, D., & Sloan, S. W. 2009. Teh, C. I. & Houlsby, G. T. 2009. An analytical study of
Alternative stress-integration schemes for largedeforma the cone penetration test in clay. Géotechnique 41(1):
tion problems of solid mechanics. Finite Elements in 17–34.
Analysis and Design 45(12): 934–943. Tehrani, F. S., Han, F., Salgado, R., & Prezzi, M. 2017.
Nazem, M., Sheng, D., & Carter, J. P. 2006. Stress integra Laboratory study of the effect of pile surface roughness
tion and mesh refinement for large deformation in on the response of soil and non-displacement piles. Geo-
geomechanics. International Journal for Numerical technical Frontiers 2017 American Society of Civil
Methods in Engineering 65(7): 1002–1027. Engineers, Reston, VA 256–264.
Of, N. & Yu, L. A. Y. B. H. S. 2000. of S Teady C One Terzaghi, K. 1943. Theoretical Soil Mechanics. John Wiley
P Enetration. Manager (July): 594–605. & Sons, Inc., Hoboken, NJ, USA.
Salgado, R. 1993. Analysis of Cone Penetration Resistance Tovar-Valencia, R., Galvis-Castro, A., Salgado, R., &
in Sand. University of California, Berkeley. Prezzi, M. 2018. Effect of surface roughness on the
Salgado, R. 2012. The mechanics of cone penetration: Con shaft resistance of displacement model piles in sand.
tributions from experimental and theoretical studies. Journal of Geotechnical and Geoenvironmental Engin
Geotechnical and Geophysical Site Characterization 4 eering 144(3).
CRC Press, Porto de Galinhas, Brazil 131–153. Wallstedt, P. C. & Guilkey, J. E. 2008. An evaluation of
Salgado, R. & Bisht, V. 2021. Advances in the solution of explicit time integration schemes for use with the gener
geotechnical boundary-value problems. Computers and alized interpolation material point method. Journal of
Geotechnics Elsevier Ltd 138(July): 104183. Computational Physics 227(22): 9628–9642.
Salgado, R., Han, F., & Prezzi, M. 2017. Axial resistance Wang, B., Vardon, P. J., Hicks, M. A., & Chen, Z. 2016.
of nondisplacement piles and pile groups in sand. Riv Development of an implicit material point method for
ista Italiana di Geotecnica 51(4): 35–46. geotechnical applications. Computers and Geotechnics
Salgado, R., Lyamin, A. V., Sloan, S. W., & Yu, H. S. Elsevier Ltd 71: 159–167.
2004. Twoand Three-dimensional Bearing Capacity of Wang, D., Randolph, M. F., & White, D. J. 2013.
Foundations in Clay. Géotechnique Ice Virtual Library A dynamic large deformation finite element method
54(5): 297–306. based on mesh regeneration. Computers and Geotech
Salgado, R., Mitchell, J. K., & Jamiolkowski, M. 1997. nics Elsevier Ltd 54: 192–201.
CAVITY EXPANSION AND PENETRATION Woo, S. I. & Salgado, R. 2015. Bounding surface modeling
RESISTANCE IN SAND By R. Salgado,’ Associate of sand with consideration of fabric and its evolution
Member, ASCE,. Journal of Geotechnical and Geoen during monotonic shearing. International Journal of
vironmental Engineering 123(April): 344–354. Solids and Structures Elsevier Ltd 63: 277–288.
Salgado, R., Mitchell, J. K., & Jamiolkowski, M. 1998. Woo, S. I. & Salgado, R. 2018. Simulation of penetration
Calibration chamber size effects on penetration resist of a foundation element in Tresca soil using the general
ance in sand. Journal of Geotechnical and Geoenviron ized interpolation material point method (GIMP). Com
mental Engineering 124(9): 878–888. puters and Geotechnics Elsevier Ltd 94: 106–117.
27
© 2022 the Author(s), ISBN 978-1-032-31259-0
Practical use of shear wave velocity measurements from SCPTU in clays

M. Long
University College Dublin (UCD), Ireland
ABSTRACT: The main objective of this paper is to promotes use of shear wave velocity (Vs) measurements
from SCPTU testing as a compliment to standard CPTU data. Recent developments in the technique are
described. Uncertainties in the methods have been well researched and several methods have been proposed
to deal with these uncertainties. Nonetheless Vs measurements seem reliable in isotropic soil conditions. Vs
profiles for a range of soft to firm clays world-wide are presented. Clear links between these profiles are iden
tified based on fundamental soil properties. Stiff overconsolidated clays do not follow the same patterns due
to pure stress anisotropy and their fissured nature. Correlations between Vs and CPTU data and with a variety
of soil properties can be very powerful. However these correlations should ideally be local and only applied
to other soils and areas with great caution. An exception might be for correlation between Vs and preconsoli
dation stress pc’. Use of SCPTU data to help assess sample disturbance and for the prediction of settlement of
footings is described. Some future outlooks including dealing with anisotropy are discussed.
1 INTRODUCTION considered. The meaning of numerical Vs values

will be discussed and typical Vs profiles for various
Shear wave velocity (Vs) is a fundamental measure clays will be presented and compared.
ment in all solids for example steel, concrete, wood, Use of Vs profiles for site characterisation will be
soil and rock (Mayne, 2000). Because of this broad described. Correlations between Vs and other SCPTU
range of application, Vs values are an attractive measurements and with a range of soil properties will
means of characterising a range of natural geomater be considered. An emphasis on estimating preconsoli
ials. Water is not able to tolerate shear waves. There dation stress from Vs. Finally use of Vs for sample
fore measurements in soil are not influenced by the disturbance assessment and for prediction of settle
presence of groundwater (unlike P-waves). ment of shallow foundations will be outlined.
Over the last decades, Vs measurements have gained The paper will focus on clay soils only (with
popularity in geotechnical engineering practice. Advances some comparison made with peat). It will not deal
in cost effective and efficient methods of determination of with earthquake problems such as liquefaction.
Vs focused attention on this parameter, which originally It is stressed that the objective of the paper is to pro
was mainly used for seismic hazard assessment or mote increasing use of SCPTU Vs measurements as
dynamic analyses. However, its use has be extended to a compliment, not as a replacement, to other CPTU
general site characterisation studies, ground movement measurements. The author supports the suggestion of
analyses for tunnels and excavation problems, determin Mayne et al. (2009), in their Key Note lecture to the
ation of strength and compressibility parameters by 17th ICSMGE in Alexandria, who proposed that in site
empirical correlation, prediction of the behaviour of deep characterisation studies the standard approach should be
and shallow foundations, assessment of sample disturb in the use of the SCPTU (or seismic dilatometer,
ance effects and in the quality control of ground improve SDMT) as up to five separate measurements are cap
ment schemes among other applications. tured at each depth tested.
Vs can be measured by a wide range of intrusive
and non-intrusive methods, see L’Heureux and Long
2 SOME BASIC CONSIDERATIONS
(2017) for example. In this paper the focus will be
on the use of the seismic piezocone test (SCPTU) for
Vs measurement. 2.1 Shear waves
The paper will deal with some basics on shear Engineer’s interest in Vs determination has largely been
waves including shear wave anisotropy. The meas driven by the need to obtain values of the small strain
urement technique will be briefly described together shear modulus (Gmax), as this is an important parameter
with some recent developments in the method. for a range of geotechnical design applications. This usu
Uncertainties in the measurement technique will be ally involves strains of 10-4 to 10-3 % and less.
DOI: 10.1201/9781003308829-3
28
According to elastic theory Gmax may be calculated For simplicity in this paper the SCPTU measure
from the shear wave velocity using the following ment will be denoted as Vs unless otherwise stated.
equation:
3 SCPTU MEASUREMENT TECHNIQUE
3.1 Traditional approach

where: Gmax is the shear modulus (in Pa), Vs is in m/ The SCPTU was first introduced in 1984 at the Uni
s, and ρ is the total mass density (in kg/m3). versity of British Columbia, see Rice (1984), Cam
It is important to remember that for practical use, panella et al. (1986) and Robertson et al. (1986).
the Gmax value needs to be reduced to give the strain Some schematics to illustrate the test set-up are
level relevant to a particular problem. This point will shown on Figure 2.
be addressed later.
2.2 Anisotropy of shear wave velocity

Anisotropy of shear wave velocity/stiffness may be
significant in many clays. This is particularly the case
for overconsolidated materials. Butcher and Powell
(1995) and others have suggested that to distinguish
between shear wave velocities with different propaga
tion and polarisation directions, subscripts can be used
to donate these. For example Vs-he denotes a vertically
propagating, horizontally polarised shear wave velocity
such as would be measured in a downhole or SCPTU
test. Similarly Vs-hv or Vs-hh would be measured in
cross-hole testing.
An example for the well-researched Italian clay
site at Augusta (saline area) is shown on Figure 1.
The materials below 15 m comprise a medium stiff,
overconsolidated (overconsolidation ratio, OCR = 2
to 6) Pleistocene marine clay with low to medium
plasticity (Lo Presti et al., 1998). Vs-vh values from
seismic dilatometer (SDMT) and downhole testing
are very similar but they are some 100 m/s less than
the cross-hole Vs-hv values.
Figure 2. Set-up for CPTU test (a) typical test arrangement

(Powell, 2005), (b) details of cone (Mayne, 2000) and (c)
schematic for Vs determination (Butcher et al., 2005).
Ideally the test should be carried out according to

ASTM (2019) or ISO (2014). Some detailed guid
ance for the execution of the test is given by Butcher
et al. (2005). A standard cone penetrometer is
equipped with one to three horizontally aligned seis
mic sensors. Originally one sensor was used but
most modern equipment has two sets of sensors
comprising three component geophones. Butcher
and Powell (1996) discuss some problems associated
Figure 1. Anisotropy of shear wave velocity for August with a single sensor SCPTU. Some commercial com
(saline area) research site in Italy. Data replotted from Lo panies successfully use equipment with one acceler
Presti et al. (1998) and Cavallaro (2020). ometer instead of a geophone.
29
Recordings are made during a pause in the cone Mayne (2005b) and others advocate the use of
penetration typically every 0.5 m. The seismic sig frequent interval SCPTU (or SDMT) with meas
nals are generated by striking a horizontal beam urements taken every 0.2 m. An example of such
which is coupled to the ground, usually by the a set of readings compared to conventional single
weight of the testing vehicle. Hammers of mass sensor measurements from the famous Treporti
about 10 kg are typically used. The beam should be test embankment, Venice are shown on Figure 3
aligned parallel to the axis of the receivers. Assum (McGillivray and Mayne, 2004). The site is
ing straight ray paths Vs is determined by (Figure 2): underlain by a complex mixture of silts, sands
and clays (Simonini et al., 2007). These data
clearly demonstrates that enhanced depth reso
lution can be achieved with this technique.
Mayne and McGillivray (2008), McGillivray
and Mayne (2008), Ku and Mayne (2012), Ku
where: t2 and t1 are the shear wave travel time from et al. (2013a) and Ku et al. (2013b) introduce
source to receiver the devel-opment of continuous interval seismic
Several methods exist for picking the shear wave piezocone testing (CiSCPTU). Here Vs is meas
arrival time such as manually picking the first arrival ured at continuously during cone penetration
time, the first major cross-over method or by use of using a specially developed autoseis source.
cross correlation. In a dual array seismic cone, the The system employs an electro-mechanical auto
wave traces from each of the two sensors can be hammer to deliver impact type seismic waves
used to get the “true interval” travel time between that are continuously recorded while the device
the sensors. Many authors (Lunne et al., 1997b) have is penetrating. As well as producing a detailed
stressed the advantage of the “true interval” set-up Vs profile, the significant advantage of
stating that it gives more accurate results than the this system is that there is no halting in penetra
“pseudo interval” method where only one sensor is tion during the SCPTU procedure and
used. A particular advantage is the same source therefore no effect on the resulting qt, fs, and u2
signal, rather than two successive blows, is used in readings.
the calculations. Examples from this system are also given by
Similar techniques are used in offshore SCPTU Agaiby et al. (2016) and Mayne and Woeller (2015).
testing, see for example Peuchen et al. (2002), Looi
jien et al. (2017) or Lunne et al. (2013) with the
signal being produced by different techniques such as
using a seismic cap (Campanella et al., 1986),
hydraulic underwater shear wave box (de Lange
et al., 1990) or underwater hammers (Paoletti et al.,
2013) or the twin hammers of the PROD system
(Nguyen et al., 2015).
3.2 SDMT
The seismic dilatometer (SDMT) is the combin
ation of the standard flat dilatometer (DMT) with
a similar seismic module for measuring Vs as
employed in the SCPTU (Marchetti et al., 2008,
Marchetti, 2015). The SDMT employs two geo
phones spaced 0.5 m apart and therefore allows
a more accurate “true- interval” two-receiver test
configuration.
3.3 Recent developments in techniques

Typically modern dual sensors are spaced at
between 0.5 m and 1.0 m. The system used by the
Norwegian Public Roads Administration allows the Figure 3. Frequent interval and “pseudo” interval SCPTU
sensors to be set at either 0.5 m or 1.0 m (Valsson from Treporti test embankment (McGillivray and Mayne,
et al., 2020). 2008).
Typically Vs measurements are made at 0.5 to
1 m intervals, the latter often corresponding to
rod breaks. Although the process is more time 3.4 Uncertainties in measurements
consuming and expensive, McGillivray and In engineering practice it is often commonplace to
Mayne (2004), (McGillivray and Mayne, 2008), provide a single Vs profile versus depth for
30
a specific location to the design engineer with no The length of the wave path between the
description of the uncertainty involved (Stolte and source and receiver is often assumed to be
Cox, 2020). There are two types of uncertainty a straight line (ASTM, 2019). Alternatively the
associated with SCPTU measurements, aleatory refracted wave path method has been proposed
variability and epistemic uncertainty (Hallal and (Joh and Mok, 1998, Kim et al., 2004). This
Cox, 2019). technique is based on refraction caused by stiff
Aleatory variability is connected to spatial vari ness differences between layers. Bang et al.
ability of the ground across the site area and can be (2014) propose an improvement on this system
dealt with by performing multiple tests across a site which they term the mean refracted wave path
with perhaps the “gaps” in the data infilled by method.
another geophysical technique such as electrical Much work has been done on assessing different
resis-tivity tomography (ERT). ways of obtaining shear wave travel times and con
An epistemic uncertainty refers to the deficiencies verting these data into shear wave velocity profiles,
by lack of knowledge or uncertainty. Some epistemic see for example Baziw (1993), Baziw (2002) Pidli
uncertainties associated with dual-receiver SCPTU secky and Haines (2011), Stolte and Cox (2020) or
testing are as follows (Peuchen et al., 2020, Stolte Valsson et al. (2020).
and Cox, 2020, Styler and Weemees, 2016, Parasie In engineering practice, for example in seismic
et al., 2022): response analyses, upper and lower bound estimates
of Vs versus depth are often used to account for epi
○ choice of signal source, stemic uncertainty (Stolte and Cox, 2020). However
○ uncertainties related to timing, this approach has recently been criticised especially
○ lateral source offset distance, where uncertainty in Vs is assumed to be depth inde
○ choice of testing interval, pendent (Teague and Cox, 2016). Several
○ vertical depth control, researchers, e.g. Garofalo et al. (2016) have shown
○ assumption of isotropic ground conditions, that Vs uncertainty is more pronounced nearer to the
○ assumption of straight-line ray path of the signal, surface than with depth.
○ identification of signal arrival time, Stolte and Cox (2020) provide a suggested
○ external noise (expressed as signal to noise means of quantifying the epistemic uncertainty by
ratio.SNR),
involv-ing various wave travel time and velocity
○ near field effects especially at shallow depths
analysis methods. These authors also state that,
(< 5m) e.g. from surface waves or acoustic
although the conventional single Vs profile may be
reflections,
sufficient for design engineers, any site report
○ conversion of signal arrival time to shear wave
velocity, should clearly state the assumptions that were made
○ wave interference due to presence of rods/ and should report on how the wave travel time was
penetrometer. obtained and the report should detail the actual
results so the engineer can check the resulting Vs
In offshore SCPT testing some additional uncer values.
tainties could be considered (Peuchen and Gomez Parasie et al. (2022) suggest a technique for esti
Meyer, 2021) such as: mating an “uncertainty budget” for Vs/Gmax based
on an inventory of items similar to those listed
○ sea wave interference effects, above. They give an example of two cases for SCP
○ influence of the physical seabed frame used to TU testing in overconsolidated clay and dense sand
support the measuring devices (particularly over and determine that the uncertainty budget in both
the top 5 m of a profile), cases is reasonable.
○ ambient noise.
To end this section on a positive note it has
With respect to the lateral distance between been found that, despite the uncertainties dis
the source and the receiver it would seem that if cussed above, there have been many published
the distance is too great the wave form may be accounts of Vs measurements where different
closer to that of Vs-hh rather than the conven systems were used by different operators but
tional Vs-vh (Personal communication P.W. returned very similar output. This is particularly
Mayne). ASTM (2019) suggests the lateral source the case for sites which are un-derlain by rela
offset should be in the range 1 m to 6 m. Hallal tively homogenous, isotropic materials. More
and Cox (2019) suggest that this distance is care needs to be taken in interpreting data form
limited to a maximum of 1.5 m. non-homogenous or layered or highly aniso
Hallal and Cox (2019) suggest that ideally the tropic soils. The example shown on Figure 4 for
testing interval should be less than 1 m and that the well characterised Fucino site in Italy. The
issues with data reduction may occur of there is site is underlain by a thick deposit of soft
a stiff layer close to the ground surface. cemented clay (Soccodato, 2003).
31
Table 1. Vs values for clay soil classification.
Material NIBS/EN Poulos (2021)

Vs (m/s) Vs (m/s)
Very soft clay < 100 85 - 105
Soft clay {100 to 106 - 135
Medium stiff clay 180} 136 - 185
Stiff clay 180 - 360 186 - 275
Very stiff clay 360 - 800 276 - 365
4 SHEAR WAVE VELOCITY VALUES
Before looking at Vs profiles from various countries

in detail it is worth considering the possible mean-
ing/classification of numerical Vs values. For seismic
assessments the average Vs over the top 30 m of
a site (Vs,30) is often considered.
Both the US (NIBS, 2003) and European (EN,
2004) guidelines quote the same range of Vs values
for site classification as summarised on Table 1.
Recently for general engineering classification
Poulos (2021) suggested the range of values for site
classification also given on Table 1.
Figure 4. Vs profiles for Fucino site, Italy. Data from Bur
ghignoli et al. (1991), Foti et al. (2006) and Soccodato (2003).
The range of values and their allotted meaning
is very similar in both cases. For the purposes of
this paper an upper bound Vs value for stiff to
Vs measurements by SCPTU, SDMT, cross-hole very stiff clays of 350 m/s will be considered,
and SASW (spectral analysis of surface waves) all and all plots will be drawn where possible with
give similar results, certainly for the purposes of an this as the maximum value on the x-axis. Simi
engineering type analysis. larly the value Vs = 100 m/s will be considered
Figure 5. Vs values for Norwegian clays (a) south Norway clays and (b) clays from Trondheim area. Data from L’Heureux
and Long (2017) with additional data from Paniagua et al. (2017), L’Heureux et al. (2019) and Dahl and Haugen (2022).
32
as the upper bound for very soft clays for illus
tration purposes.
These values need to be treated with caution and
can be misleading as will be described later.
5 TYPICAL VS PROFILES FOR CLAYS
5.1 Norwegian clays

A summary of all the available Vs data for Nor
wegian clays is given on Figure 5, with sites
from South-East Norway shown on Figure 5a and
those from Mid-Norway/Trondheim area on
Figure 5b. The data is mostly taken from L’Heur
eux and Long (2017) with some additional data
as indicted.
Some of the data was obtained using the
MASW method (multichannel analysis surface
waves) tech-nique. As has been shown in the
Figure 6. Vs values for Swedish clays. Data from Long
above paper and elsewhere, the results do not et al. (2017) with additional data from Kania et al. (2000),
seem to be affected by the technique used or the (Dannewitz et al., 2005), Comina et al. (2017) and Gran
directions of propagation and polarisation of the skär (2018).
waves. This is likely to be due to the largely iso
tropic nature of these materials.
All the Southern Norway sites show a very similar 5.2 Southern Swedish clays
trend between Vs and depth and differ only in the
value of Vs close to the surface. Teachavorasinskun A summary of all the available data from southern
and Lukkunaprasit (2004) found a similar pattern for Sweden is shown on Figure 6. These data are taken
soft Bangkok clay and they expressed the relation from Long et al. (2017) augmented with additional
ship in the form: data as shown.
The data are in general very similar and fall
into a relatively narrow band with Vs values
typically increasing from 50 m/s at the surface
to about 125 m/s at 20 m depth. A similar pat
where:
tern to the Norwegian data can be seen with the
Vsz = Vs at any depth z (m/s)
values gradually increasing with depth. However
Vsg = Vs close to the ground surface (m/s)
the measured values are significantly less than
m = slope of the line of Vs versus depth (units m/
those from Norway. Also the slope of the Swed
s.m)
ish profiles with depth are not as steep as those
Some exceptions are the very soft, high water of the Norwegian sites. These lower values are
content and organic, clays at Onsøy and especially compatible with the higher water content, greater
Farriseidet which show much lower values of Vs. In organic content and lower unit weight and lower
general the values increase from about 125 m/s near silt content of the Swedish clays compared to
the surface to 225 m/s at 20 m depth classifying the the Norwegian sediments as will be explored
materials as soft becoming medium stiff to stiff with later.
depth (Table 1) consistent with other measurements The range of Vs values encountered suggest that
on the soils. For example anisotropically consoli the materials are very soft becoming soft to firm
dated undrained triaxial compression (CAUC) tests (Table 1), again consistent with the results of other
on the clay from the NGTS (Norwegian GeoTest testing.
Site) at Tiller-Flotten show undrained shear strength
(su) values increasing from about 25 kPa near the
surface to about 115 kPa at 20 m depth (L’Heureux 5.3 Clays from Finland
et al., 2019). Only limited data is available for Vs profiles from
Data for the Trondheim and Mid-Norway sites sites in Finland. SCPTU data has been published for
show similar values to those from Southern Norway. the soft clay sites at Perniö, Lempäälä and Masku
The data fits in a very tight band within the limits from southern Finland These are shown on Figure 7.
for the Southern Norway clays. The fact that the Two profiles from each site are shown and in general
soils have similar values is not surprising given the they are very similar.
similar depositional and stress history and mineral The profiles shown on Figure 7 are compared
ogy of the sediments. to the range of values for Swedish clays from
33
Figure 8. Vs profiles for Eastern Canadian sensitive clays
Figure 7. Vs values for Finnish clays (Di Buò et al., 2016, Data from Bouchard et al. (2017), Lefebvre et al. (1994),
Di Buò et al., 2018, Di Buò et al., 2019). Leroueil et al. (2003), Mayne et al. (2019), Fabien-Ouellet
etal. (2014), Elbeggo et al. (2021) and Agaiby (2018).
Figure 6. Overall the range of values measured The device is very similar to the “pseudo interval”
for the sites in Finland are very like those from SCPTU described above. Vs measurements, together
Sweden and fall within the range of values with parallel water content measurements, for four peat
measured for the Swedish clays. One possible sites in four different countries are show on Figure 9.
reason leading to consistent Vs versus depth pro Water content values are often in the range 1000% to
files is to be found in the similarities between 2000%. Not surprisingly the Vs values are very low,
the basic clay properties (water content, bulk typically 15 m/s. There seems to be a clear trend of
density and plasticity) (Long and D’Ignazio, lower Vs corresponding to the highest water content.
2020).
5.4 Eastern Canadian clays

A compilation of some Eastern Canadian data is
shown on Figure 8. Note the geology of the Eastern
Canadian area is complex and a more detailed dis
cussion of SCPTU data in this area has recently been
published by (Salsabili et al., 2022). The profiles
chosen here correspond to well-known geotechnical
research sites.
Many of the profiles fall within the bounds of the
southern Sweden sites. An exception is the profile
from the Quyon Landslide site and perhaps the City
of Ottawa data.
5.5 Peats
Trafford and Long (2020) developed a small light
weight probe for measuring the shear wave velocity
in peat. The motivation was to allow testing on site Figure 9. Water content and Vs profiles for four peat sies
where it would not be easy to gain access with trad from four countries (x-axis ranged changed from earlier
itional SCPTU equipment. figures).
34
5.6 London clay
Perhaps most well characterised stiff clay in the world
is London Clay (Hight et al., 2003). Some profiles for
Vs-vh from SCPTU and other testing in London Clay
are shown on Figure 10. Due to the stiffness of the
clay and the presence of stiff “clay- stone” bands the
tests often need to be terminated at relatively shallow
depths. The SCPTU data for the Bradwell B nuclear
power station, measured by In Situ Site Investigations
Ltd., and reported by LessiCheimariou et al. (2019) are
particularly interesting as the tests were able to pene
trate to the base of the London clay due to use of
a technique of continuing the tests in grouted up bore
holes once the original tests were terminated.
On Figure 10 the Vs profiles for soft to medium
stiff southern Norwegian clays and stiff to very stiff
London clay. Intuitively one would expect the
London clay Vs values to be much higher than those
of the Norwegian soils. However, as can be seen, the
range of values recorded are more or less identical.
The reason for the low values in the London
clay is likely to be linked to the fissured nature
of these soils (Hight et al., 2003) and due to
pure stress ani-sotropy with the horizontal propa
gating waves pass-ing along layers of higher Figure 11. Vs-vh values for other UK stiff overconsolidated
stiffness (Butcher and Powell, 1996, Piriyakul clays (other than London clay). Data from Butcher and
and Haegeman, 2008). Powell (1995), Hosseini-Kamal et al. (2014), Bates and
However it is clear that correlations derived for Philips (2000) and Matthews et al. (1996).
example between Vs and su at one location or for
one specific soil type need to be applied only with
great caution to another soil type. Data for UK stiff clays other than London clay are
shown on Figure 11. A picture similar to that of
London clay emerges with surprisingly low Vs-vh
5.7 Other UK stiff overconsolidated clays values for these stiff to hard heavily overconsolidated
materials. Again the values are similar to those of the
South Norway soft to firm clays. Many of these
materials have very high Vs-hh values and it is likely
that the low Vs-vh values are due to stress anisotropy
and possibly also due to fissuring of the materials.
5.8 Comparison/link between various clay types

Given the consistence of the Vs profiles for different
areas it is worth considering the basic engineering
properties of the clays in an attempt to identify the
links between the Vs values.
According to Leroueil and Hight (2003) and
Hardin (1978) the empirical equation describing the
influence of the controlling factors on Gmax (and Vs)
′can then be written as follows:
Gmax ¼ SFðeÞðσ0 v σ0 h Þn pa ð1-2nÞ ð4Þ

Figure 10. Comparison of Vs-vh values for London Clay
and range for South Norway clays (L’Heureux and Long, where: S is a dimensionless parameter characterising
2017). London clay data from Powell and Butcher (1991),
the considered soil; F(e) is a void ratio function; σV 0
Butcher and Powell (1995), Gordon et al. (1996), Matthews
et al. (1996), Menzies and Matthews (1996), Hight et al. and σh 0 are the vertical and horizontal effective stres
(2003), Hight et al. (2007), Clayton (2011) and Lessi- ses respectively; n is a parameter indicating the influ
Cheimariou et al. (2019). ence of stress; and pa is atmospheric pressure.
35
Table 2. Summary of material properties for selected Also Tiller-Flotten has very low organic con
study sites. w = water content, Ip = plasticity index, Org. = tent compared to the other sites. The Onsøy site
organic content OCR = overconsolidation ratio, St = fall parameters are much closer to those of the Swed
cone sensitivity. Main references Gundersen et al. (2019), ish, Finnish and Canadian sites. All sites under
L’Heureux et al. (2019), Wood (2016), Di Buò et al. (2019) consideration have similar clay content and stress
and Trak et al. (1980). history.
In order to explore this further, the coefficients
Clay Ip Org.
Site w (%) (%) (%) (%) OCR St
Vsg and m in Equation 3 are plotted against aver
age water content (w) and unit weight ðγÞ for
Onsøy 3 40-80 50-70 25 2.5-4 1.1-2.0 5-8 each site (over the interval where Vs data is
50 available) on Figure 12 (L’Heureux and Long,
Tiller 30-50 45-70 8-20 Very 1.5-2.0 up to 2017, Long et al., 2017). It can be seen that both
Flotten low 350 parameters decrease with increasing w and
Göteborg 60-90 70-90 27 2-5 1.5-2.0 12-30 increase with increasing ðγÞ as would be
CS 40 expected. The trend between the parameters is
Perniö 80-110 50-90 30 1-2 1.5-2.0 40-60 reasonably good and these relationships could
40 therefore be used for first order estimates of Vs
St. Alban 60-90 45-81 5-30 0.9 2.2 14-22 or for controlling site measurements.
Peat 1000 Very n/a 90 Very n/a
2000 low 100 low
6 USE OF VS IN SOIL BEHAVIOUR TYPE
CHARTS
A summary of the key properties of the clays
from well characterized test sites in Norway, Robertson (1990) introduced the well-known soil
Sweden, Canada and Finland are given on Table 2, behaviour charts based on normalised cone resist
together with some typical properties of peat. Two ance (Qt), normalised sleeve friction (Fr) and the
sites from Norway were chosen namely the pore pressure parameter (Bq)
NGTSsites at Onsøy and Tiller-Flotten.
The Tiller-Flotten site is characteristic of many Nor
wegian clays and is significantly different from the
other sites under study. These Norwegian sites have
relatively lower water content and plasticity and higher
density (1.7 – 1.9 Mg/m3 compared to 1.6 – 1.7 Mg/
m3) than the Canadian, Finnish and Swedish sites.
where:
qt = corrected cone end resistance
fs = sleeve friction
u2 = pore pressure measured behind cone
u0 = ambient pore pressure
σv0 =σv0 0 = total/effective initial vertical stress
Later Robertson et al. (1995) proposed
a complementary SCPTU soil behaviour chart based
on Qt and normalised small strain shear modulus
(Gmax/qt). This chart was intended mostly for identi
fying “unusual” soils such as highly compressible
sands, cemented and aged soils and clays with either
high or low void ratio.
Robertson (2015) suggested that IG = Gmax/qt can
be considered to be a small strain rigidly index as it
relates small strain stiffness to a measure of soil
strength. He proposed a new chart relating Qtn to IG
which could be used to identify soils with micro-
Figure 12. Coefficients Vsg and m (Equation 3) as structure such as cementation/bonding or aging.
a function of water content and unit weight for sites in Note Qtn is an update to Qt using a variable stress
Norway and Sweden. exponent (Robertson, 2009).
36
Figure 13. Contours of normalised shear wave velocity
superimposed on the normalized Qtn - Fr soil behavior type
chart (Robertson, 2009). Figure 14. Vsn versus Δu=σv0 0 for some Norwegian and
UK soft clays. Adapted from Long and Donohue (2010).
Based on 20 years’ experience of SCPTU tests in

California combined with results from the literature
Robertson (2009) also developed a series of contours Although there appears to be some promise in
of normalised shear wave velocity (Vsn) and superim the use of the charts shown on Figure 13 and 14
posed these on a base of the Qtn – Fr chart as shown and other similar charts, it is felt that some
on Figure 13. The soils involved were mostly of Holo more work, and some more data, is needed to
cene age unbonded silica-based soils which tended to bring these into civil engineering practice.
plot in the central part of the chart. Some older Pleisto It is also stressed that these charts are designed to
cene age soils plot toward the upper right-hand part. complement the more well-known CPTU soil behav
iour charts and not to replace them.
7 CORRELATION BETWEEN VS AND OTHER

SCPTU MEASURMENTS
Robertson (2009) and others have chosen the ex
ponent n in Equation 8 = 0.25 and have used the Various researchers have studied relationships
expression Vs1 for Vsn. Ku and Mayne (2013) and between CPTU parameters and Vs in clayey soils.
Moon and Ku (2016a) and others have pointed out These studies have explored relationships between
that n = 0.25 is based on limited laboratory data on in situ Vs and various parameters such as CPTU tip
clean quartz and silica sands and suggested n is resistance (qc), corrected tip resistance (qt), cone net
chosen on site or area specific conditions. resistance (qnet), sleeve friction �(fs),
� pore pressure
Several authors have suggested that the CPTU u2 parameter (Bq), effective stress σ0v , water content
measurement is perhaps the most reliable of the (w) and void ratio (e).
normal three sets of data collected (Lunne et al., Many empirical relationships between Vs and
2018). Therefore, as Vs is also relatively easy to these CPTU parameters have been published. For
measure reliably in isotropic material, it seems example L’Heureux and Long (2017) summarise 21
logical to attempt to develop a characterisation chart such relationships for clayey soils. Since the publica
based on these two data sets. tion of the 2017 paper the author is aware of at least
On Figure 14 Vsn data are plotted against ten further newly developed relationships. These
Du=σv0 0 ð¼u2 -u0 =σv0 0 Þ for some Norwegian and UK include those by Holmsgaard et al. (2016) for silty
soft soils. soils in Denmark, Shahri and Naderi (2016) for clay
Here n was chosen to be = 0.5. A tentative div soils in south-west Sweden, Zhang and Tong (2017),
ision can be made between the lightly overconsoli Zou et al. (2017), Tong et al. (2018) and Duan et al.
dated material (OCR < 2) and the moderately (2019a) all for Jiangsu clays in China, McGann et al.
overconsolidated soils (OCR >3). (2015) for silty soils in Christchurch.
37
New Zealand and Salsabili et al. (2022) for post-
glacial sediments in Southern Québec. In addition
Ahmed (2017) and Karray and Hussien (2017)
derived generalised relationship for a variety of soils.
Robertson (2015) suggests that even though qt
and Vs are controlled by the same factors there is no
unique correlation between qt and Vs. However it
should be possible to obtain a reasonable relation
ship if the database used is limited to soils of similar
mineralogy, stress history, age and cementation.
An example from the Skatval site in Mid-Norway
is shown on Figure 15 (Paniagua et al., 2017). Data
from this site was not included in the original Nor
wegian clay database of L’Heureux and Long Figure 16. (a) Sherbrooke block sampler and (b)
(2017). However the correlation developed by these mini-block sampler. Image from (Emdal et al., 2016).
authors, as follows on Equation 9, fits very well with
the measured data for this new site. 8 CORRELATION BETWEEN VS AND SOIL
PROPERTIES
8.1 General
In contrast the relationship developed for South Much work has been done in the past in correlating
ern Swedish clays by Shahri and Naderi (2016), Vs with various soil properties derived from labora
given in Equation 10, significantly underestimates tory testing. An important consideration is that, for
the measured Vs values the correlation to be reliable, the quality of the soil
samples must be very high. To that end in this paper
soil samples extracted by high quality samplers such
as the Sherbrooke block (Lefebvre and Poulin,
1979), the mini-block sampler (Emdal et al., 2016)
The message here is that very good correlations (Figure 16) or the Laval sampler (La Rochelle et al.,
can be developed between CPTU parameters and Vs 1981) only are used.
but these should be used locally and they should It is also important that the laboratory tests and
only be used with great care for soil types other than the derivation of the resulting parameters is carried
those for which they were intended. out in a consistent manner.
Figure 15. Skatval site Norway – predicted and measured Vs.
38
wide database of soil tests. They have shown how the
resulting relationship will depend amongst other fac
tors on soil type, in situ stress and plasticity.
Different equations are suggested for different
soil types and the fit between the measured and pre
dicted data can be improved by normalisation of Vs
or by introduction of additional parameters such as
the plasticity index (Ip) into the relationships.
Although these relationships are very useful it is per
haps more worthwhile to develop local correlations.
An example of such a correlation for Norwegian
clays is shown in Figure 17.
8.3 Undrained shear strength from Vs

Similar to Vs-CPTU relationships much work has
been done on deriving empirical relationships
between Vs and undrained shear strength su. L’Heur
eux and Long (2017) summarise 16 such relation
Figure 17. Unit weight versus Vs for Norwegian clays. ships. Since the publication of the 2017 paper the
Data from L’Heureux and Long (2017). author is aware of at least four other publications on
the topic by Powell et al. (2016), Moon and Ku
(2016b), Duan et al. (2019a) and Duan et al.
8.2 Unit weight from Vs (2019b).
An important relationship between geophysical Undrained shear strength has no unique value but
parameters and soil index parameters is that between depends on the age of the sample, test type, rate of
Vs and unit weight ðγÞ. Various authors such as loading and other factors. Therefore comparisons
Mayne (2007) and Moon and Ku (2016a) have devel can only be made on results of tests carried out on
oped relationships between Vs and γ using a world exactly the same manner.
Figure 18. Norwegian clay relationship between su and Vs for (a) CAUC tests and (b) CAUE tests. Original 2017 data from
L’Heureux and Long (2017) is in black and white font. Recent data is in colour font. Recent data is from Gundersen et al.
(2019), L’Heureux et al. (2019), Paniagua et al. (2017) and Dahl and Haugen (2022).
39
Figure 19. Norwegian clay relationship between su and Vs Figure 20. Relationship between su and Vs for CAUC tests
for DSS tests. Original 2017 data from L’Heureux and on world-wide clays for testing on high quality samples.
Long (2017) is in black and white font. Recent data is in Data is from Landon (2007), Tanaka et al. (1998), Tanaka
colour font. Recent data is from Gundersen et al. (2019), et al. (2001), Di Buò et al. (2019), Lo Presti et al. (2003)
Long et al. (2019) and Paniagua et al. (2017). and Wood (2016).
In Norway it is common practice to carry out tri should only be applied to other materials/other test
axial testing after the sample has been first consoli types with great caution.
dated anisotropically to the best estimate of its in situ
stress. Shearing can subsequently be by compression
(CAUC tests) or by extension (CAUE tests).
The su values obtained from CAUC and CAUE tri
axial tests on high quality samples of Norwegian clay
are plotted against in situ shear wave velocity in
Figure 18a and 18b respectively. A similar set of data
for direct simple shear tests (DSS) are shown on
Figure 19.
The”best fit” empirical relationship derived in the
2017 study by L’Heureux and Long (2017) is also
shown on Figure 18 and 19. The data used in the
2017 study, which was used to derive the empirical
relationships, is shown in black and white font. New
data available since the 2017 study are shown in
colour. It can be seen that in all three cases the new
data fit the old relationships very well and it can be
concluded that the 2017 relationships still apply.
A similar set of CAUC data but in this case for Figure 21. Classical Janbu plot of 1D compression stiff
various world-wide clays is shown on Figure 20. All ness versus effective stress.
the samples were of high quality as described above.
In this case the 2017 relationship for Norwegian
clays does not work and in fact would underestimate 8.4 Preconsolidation pressure from Vs
the measured su. In this and the following sections Vs measurements
The clear message here is that these Vs-su correl are compared to the classical 1D compression
ations can be very powerful and useful but they parameters often used in Scandinavian practice
should only be used for the test type and materials (Janbu, 1963, Janbu, 1969). The classical Janbu plot
for which they were originally derived and they of 1D compression stiffness against stress is shown
40
Figure 23. Swedish clay relationship between p’c and Vs.
Data from Wood (2016).
Norway, Professor Guy Lefebvre of the University

of Sherbrooke suggested to the author that plots sim
ilar to that on Figure 22 and Figure 23 could be
made more universal by normalising the Vs value by
Figure 22. Norwegian clay relationship between p’c and effective stress. This seems a logical suggestion
Vs. Original 2017 data from L’Heureux and Long (2017) is given the clear differences in the basic properties of
in black and white font. Recent data is in colour font. the two soil types as demonstrated on Table 2.
Recent data is from Gundersen et al. (2019), L’Heureux Data from various countries, where high quality
et al. (2019), Paniagua et al. (2017) and Dahl and Haugen
(2022).
block sample data were available, are shown on
Figure 24a. The Vs data were normalised using
Equation 8 with a stress exponent n chosen to = 0.5.
on Figure 21. Janbu (1963) used the resistance con Var-ious other exponents, including n = 025, were
cept to interpret one dimensional consolidation in an also trailed but no difference in the output was
oedometer test. He defined the tangent modulus (or found. Full details of this process are given in
the constrained modulus), M, as the ratio of the a companion paper to this conference by Long and
change in stress ðDσ0 Þ to the change in strain ðDεÞ L’Heureux (2022).
for a particular load in increment, i.e.: The normalisation by effective stress appears to
be unsuccessful in unifying the data.
To take the stress history of the materials into ac
count the measured Vs data have been normalised by
the preconsolidation stress (p’c) on Figure 24b. As
can be seen on Figure 24b this form of normalization
As for su above high quality data preconsolidation
As for su above high quality data preconsolidation stress (p’c) for Norwegian clays published by
stress (p’c) for Norwegian clays published by L’Heureux and Long (2017) has been augmented by
L’Heureux and Long (2017) has been augmented by more recent data (Figure 22). Again, as found for su,
more recent data (Figure 22). Again, as found for su, the recent data fit the 2017 correlations very well
the recent data fit the 2017 correlations very well and support the idea that local Vs correlations based
and support the idea that local Vs correlations based on high quality data can be very powerful. was very
on high quality data can be very powerful. successful in harmonising the fifteen sets of data. All
The relationship derived for Norwegian clays is fifteen profiles are very similar and show an average
tested against data for high quality samples of Swed Vsnp value of about 96 m/s. Taking this average Vsnp
ish clay on Figure 23. It can be seen that there is value the following equation can be obtained to
a reasonable relationship between the two param relate Vs and p’c.
eters for Swedish clay but the data falls above the
line derived for Norwegian clay.
At the 2017 International Workshop on Land
slides in Sensitive Clays (IWLSC) in Trondheim,
41
Figure 24. Normalised Vs profiles (a) by vertical effective stress and (b) preconsolidation stress. Data from Wood (2016),
Gundersen et al. (2019), L’Heureux et al. (2019), (L’Heureux and Long, 2017), (Paniagua et al., 2017), Landon (2007),
Tanaka et al. (1998), Tanaka et al. (2001) and Di Buò et al. (2019).
Equation 12 is trialed against data for two Italian

clays on Figure 25. The mean relationship together
with those representing ±0:5 times the standard
deviation are shown. Italian data was not included in
the original derivation in Figure 24. The Fucino clay
data fits the proposed relationship very well. The fit
is reasonable for the complex Pancone clay found
beneath Pisa Tower.
Figure 25. Italian clay relationships between p’c and Vs
trialed against formulae derived in this paper. Data from Lo 8.5 Other one-dimensional consolidation
Presti et al. (1998), Jamiolkowski and Pepe (2001) and Soc
codato (2003).
parameters from Vs
Relationships have also been proposed between Vs
This form of power equation supports and justifies and other 1D consolidation parameters, for example
some previous similar empirical equations that have M0 and ML (Figure 21). Such data for Perniö clay
been developed. These include the general relation from Finland and Göteborg Central Station clay
ship developed by Mayne et al. (1998) as shown on from Sweden are compared to relationships for Nor
Equation 13, that derived for Norwegian marine wegian clays (L’Heureux and Long, 2017) on
clays by L’Heureux and Long (2017) (Equation 14) Figure 26. In both cases there is a clear relationship
and by Duan et al. (2019a) for Jiangsu clays in between the parameters but the previously published
China (Equation 15). Norwegian clay relationships would underestimate
42
Figure 26. Relationship between (a) M0 and (b) ML versus Vs for Swedish (Wood, 2016) and Finnish (Di Buò et al., 2019)
clays.
and overestimate respectively the values for Finnish assessing sample dis turbance and to act as a rapid
and Swedish clays. screening tool for assessing sample disturbance prior
to advanced laboratory testing. This is of particular
concern in offshore site investigations given the cost
9 USE OF VS FOR SAMPLE DISTURBANCE of these investigations and the need to maximise
ASSESSMENT
Various authors have studied the use of shear wave

velocity measurements to assess sample disturbance Table 3. Sample quality designation (SQD) from Landon
effects. The overall idea is to compare in situ Vs, as et al. (2007) and Lunne et al., (1997a).
measured by SCPTU for example, to Vs measured
Sample
on soil samples. Some examples of this work, specif Lunne et al. with Lunne et al. with quality
ically on clay soils, include that of Landon et al. SQD OCR 1-2 De=e0 OCR 2-4 De=e0 (SQ)
(2007), Ferreira et al. (2011), Ramsey (2020), Gu
et al. (2021) and Long et al. (2021). Other authors SQ1 0:6 < 0.04 < 0.03 Very good/
have used a combination of Vs and suction measure excellent
ments to assess sample disturbance, e.g. Donohue SQ2 0:6 0.04-0.07 0.03-0.05 Good to
and Long (2010) and Horng et al. (2010). fair
Landon et al. (2007) developed a series of criteria SQ3 0.35–0.6 0.07-0.14 0.05-0.10 Poor
for assessing sample disturbance based on the ratio SQ4 < 0.35 > 0.14 > 0.10 Very poor
of Vs measured on the sample (in unconfined condi
tions) to that of the in situ Vs (e.g. from seismic
piezocone, SCPTU). However these criteria were
based on tests on low sensitivity, relative high water data quality. They were not intended to replace the
content and high plasticity clays from Onsøy, more rigorous techniques.
Norway, Newbury, USA and Burswood, Australia. The sample disturbance assessment criteria of
The specific guidelines and criteria for assessing Landon et al. (2007) are tested against recently
sample disturbance from this work is summarised on derived data from three Norwegian sites on
Table 3. They are compared to the well known Figure 27. The data are plotted in the form of Vs,0/
method of Lunne et al. (1997a) which involves Vs,insitu against De=e0 obtained from CAUC or
measuring the change in void ratio relative to the ini CAUE tests or CRS tests on Figures 27a and 27b.
tial void ratio ðDe=e0 Þ when consolidating a sample The sample shear wave velocity was measured using
back to its in situ stress in the lab. bender el-ments at zero confining stress (Vs,0).
It is important to note that Landon’s criteria were For both plots it can be seen that although the
primarily intended to provide a quick technique for data under study here is classified as either SQ1 or
43
to those found here in that the low plasticity clay
from Skoppum showed a significant discrepancy be
tween the field and laboratory tests whereas the
results for Onsøy were similar for both techniques.
It is concluded that the criteria Landon et al.
(2007) are too of strict for the materials under study
here. It is suggested that a lower Vs,0/Vs,insitu limit of
0.35 (compared to 0.6) might be more appropriate
for low plasticity sensitive clays considered here.
However insufficient data exits here to give
a definitive recommendation.
10 USE OF VS FOR SETTLEMENT

PREDICTION
Several researchers have used Vs measurements for

the purpose of estimating settlement of foundations.
The technique has mostly been used for foundations
in sandy soils, e.g. Lehane et al. (2008) but has also
been used in soft clay Mayne (2000) and Omar et al.
(2011) and soft clayey silt (Mayne, 2005a). Small
strain shear modulus (Gmax) can be calculated dir
ectly from Vs (Equation 1). For the ground beneath
foundations the strains will be larger and therefore
the Gmax value needs to be adjusted downwards to
be compatible with the actual strains encountered.
Perhaps the most common technique for this pur
pose is the use of the hyperbolic stiffness degradation
formula suggested by Fahey and Carter (1993) as
follows
where τ and τmax are the applied shear stress and shear
Figure 27. Comparison of shear wave velocity ratio and strength respectively and f and g are fitting param
normalised void ratio for (a) CAUC and CAUE tests and eters. The ratio τ=τmax is analagous to the inverse of
(b) CRS tests. Data from Landon (2007), L’Heureux et al. factor of safety. Values of f = 1 and g = 0.3 appear to
(2019), Paniagua et al. (2017) and Long et al. (2021). give reasonable estimates for unstructured and unce
mented geomaterials and provide a general fit to
experimental data (Mayne et al., 2009).
SQ2 by the criteria of Lunne et al. (1997a) consistent
with the observations of the stress – strain and stress
path plots, it falls well below that of the data of
Landon (2007) and would be classified as “poor”
according to the latter criteria.
The clays at Tiller-Flotten, Skatval and Koa differ
significantly from those studies by Landon (Table 2).
These clays have lower water content, lower plasticity
and higher sensitivity compared to Landon’s clays.
Valsson et al. (2020) and Dahl and Haugen (2022)
carried out similar studies on low sensitivity but
high plasticity Onsøy clay and low plasticity sensi
tive clay from Skoppum. Their results are not dir
ectly comparable with those presented here as the
bender elementtests were carried out after the samles
had been consolidated to the best estimate of the-in Figure 28. Prediction of settlement of footing on soft silt
situ stress. Nonetheless their conclusion were similar using Vs (Mayne, 2005a).
44
Some details for the Class “A” footing response For offshore investigations a recent development
prediction on a trial structure founded on soft silt has been to produce images similar to that on
(known locally as “sleech”) in Belfast, Northern Ire Figure 29 using fibre optic/distributed acoustic sens
land are shown in Figure 28. Mayne (2005a) used ing (DAS) techniques, see for example (Trafford
the measured Vs values (Lehane, 2003) and the stiff et al., 2021).
ness degradation formula describe in Equation 16 Another option is to combine SCPTU with ERT
above to obtain a Class “A” estimate of the footing as mentioned above.
settlment. It can be seen that the predicted and meas
ured data are in good agreement, especially to an
11.2 Anisotropy
applied stress of about 80 kPa, at which point the
onset of bearing capacity failure occurs. A significant input in the design of engineering sys
tems such as retaining walls and piles subject to lat
eral loading is the horizontal ground stiffness. In
11 FUTURE OUTLOOKS AND CHALLENGES particular for large monopiles, used for example in
TO SOLVE offshore wind turbine construction, the behaviour of
the system is heavily dependent on the horizontal
Despite significant developments in the equipment, ground stiffness at small strain. A technique for
data acquisition systems and the cumulative experience obtaining such input parameters from SCPTU has
gained over the past 35 years since the introduction of been described by McAdam et al. (2020) based on
the SCPTU some significant challenges remain to be work in the PISA project. Wang et al. (2022) have
solved. Perhaps of these the most significant are (Per proposed a simple rotational spring model for the
sonal communication from Nico Parasie, Fugro): design of laterally loaded rigid piles in sand with Vs
as a key input parameter.
○ measurement uncertainty,
Strictly speaking the required input parameter in
○ spatial variability/aleatory uncertainty, these cases is Vs-hh. Unfortunately SCPTU yields
○ anisotropy. Vs-vh. In unform relatively homogenous softer soils
Issues related to measurement uncertainty have been Vs-hh and Vs-vh have the same values. However for
well researched and are dealt with in Section 3.2 above. more overconsolidated soils this is not always the
case as has been illustrated above for example for
UK stiff clays and some Italian clays.
11.1 Spatial variability
Spatial variability issues, in sites with widely vary
ing ground conditions, can only be dealt with by
carrying out a relatively large number of tests or by
introducing another technique to “fill the gaps”
between the SCPTU testing.
Figure 29. 2D stiffness profile for Onsøy 2 site in Norway.

Image courtesy of Dr. Shane Donohue, UCD. Note horizo
natal and vertical scales the same.
An example of such an approach would be to Figure 30. Vs-hh versus qt. Data from Powell and Butcher
combine SCPTU with a surface wave technique (2004) and Powell et al. (2016) augmented with additional
such as MASW. A 2D stiffness profile for the data from Lo Presti et al. (2003), Jamiolkowski and Pepe
(2001), DeGroot and Lutenegger (2003) and Ku and
highly uniform Onsøy 2 site in Norway is shown Mayne (2014).
on Figure 29. This profile was produced by “con
tinuous” MASW where a streamer of geophones Powell and Butcher (2004) and Powell et al.
mounted on wooden blocks were pulled along (2016) have studied the relationship between cor
and repeated hammer shots were imparted to pro rected once resistance (qt) and the various in situ
duce regular seismic signals. The resulting series derived shear moduli Gvh, Ghv and Ghh using
of 1D profiles were then combined to produce a database of high quality tests. These authors found
the image.
45
that qt is most strongly influenced by horizontal stiff ○ An important exception to this is data for stiff
ness and stress in the ground. They proposed that overconsolidated clays, such as the UK stiff
Ghh or Vs-hh could therefore be obtained from correl clays / London clay. For these soils the Vs-vh
ation with qt. values are unexpected low. This is due to the nat
The data presented by these authors, augmented ural stress anisotropy and also partially due to the
by some additional results, are plotted using shear fissured nature of the materials.
wave velocity rather than shear modulus on ○ Vs-CPTU correlations can be very powerful. New
Figure 30. It can be seen that the relationship data for the same soil type in the same area fit the
between the two parameters is very good and the older correlations very well. However these cor
resulting “best fit” power function has a good R2 relations may not work for other areas. Therefore
value of 0.87. As previously suggested by Powell correlations should ideally be locally developed
et al. (2016), a suggestion for future work is there and only applied to other soils and areas with
fore that Vs-vh is obtained from SCPTU and Vs-hh is great caution.
obtained from the same test by correlation with qt. ○ A similar finding was made for soil properties
such as su and M0 where once again strong correl
ations can be developed but these correlations
12 CONCLUSIONS may not apply to other areas
○ An exception might be for correlation between Vs
The main objective of this paper was to summarise and preconsolidation stress p’c where it seems
the practical and useful application of Vs measure that high quality data from several areas may fit
ments from SCPTU testing. The intention is to pro the same trend.
mote the use of SCPTU testing and to highlight the ○ SCPTU data can also be used to assess sample
application of Vs data as a compliment to standard disturbance effects by comparing in situ Vs
CPTU data in future studies and practical projects. It values to measurements made on samples. How
is not intended that Vs measurements replace stand ever it would seem that previously published
ard CPTU data. Some conclusions from this work assessment criteria derived for relatively high
are as follows: plasticity clays may not apply to low plasticity
clays.
○ Vs is a fundamental measure in all solids. There ○ Vs has been successfully used to directly predict
fore it is an attractive way of characterising settlement of footings on clays.
a wide range of geomaterials. ○ Some future outlooks include further work on
○ Use of Vs has gained popularity in recent times understanding test uncertainties and anisotropy. It
due to developments in geophysical testing sys is suggested that for future studies Vs-vh could be
tems and analytical methods. obtained from SCPTU and Vs-hh by correlation
○ Application of the SCPTU testing technique is
with qt.
well covered in guidelines and codes of practice.
○ Some descriptions of the equipment used have
been given together with some recent develop
ments in the technique including the frequent ACKNOWLEDGMENTS
interval testing and continuous interval testing.
○ Uncertainties in the methods have been well The work presented here largely comes from long
researched and understood and several methods term collaboration between the author and col
have been proposed in the literature suggesting leagues in the Norwegian Geotechnical Institute
ways of dealing with these uncertainties. (NGI), and the Norwegian University for Science
○ Despite these uncertainties Vs measurements and Technology (NTNU). In particular the author is
seem repeatable in isotropic soil conditions using grateful to Jean-Sebastien L’Heureux, Priscilla
a variety of techniques and methods. Paniagua and Tom Lunne.
○ Vs profiles for sites in southern Norway, mid John Powell of BRE/Geolabs has been very help
Norway, Sweden, Finland, Canada and peat have ful in providing ideas and data in particular with
been presented. These data often fall in tight respect stiff overconsolidated clays.
bands. From the academic world Professors Paul Mayne
○ There appears to be clear links between these pro of Georgia Tech and Barry Lehane of the University
files depending on fundamental soil properties of Western Australia (UWA) were both very helpful
such as water content, density, plasticity and also in providing ideas and guidance and in sharing data.
stress history. This has resulted in tentative soil Several colleagues in the geotechnical industry
behaviour type charts involving Vs being pro including Nick Ramsey, Joek Peuchen and Nico
posed. This approach seems promising but further Parasie of Fugro, Darren Ward and Louisa Dhimitri
work is required on the development and applica of In Situ Site Investigation Ltd. and Joe Hobbs of
tion of these charts. Lankelma Ltd. provided useful data and guidance.
46
REFERENCES Campanella, R. G., Robertson, P. K. & Gillespie, D. (1986)
Seismic cone penetration tests. Use of In-Situ Tests in
Agaiby, S. S. (2018) Advancements in the interpretation of Geotechnical Engineering, ASCE GSP 6, 116–130.
seismic piezocone tests in clays and other geomaterials. Cavallaro, A. (2020) The use of Cone Penetration Tests
PhD Thesis, School of Civil and Enironmental Engin (CPT) for the study of the dynamic characteristics of the
eering, Georgia Institute of Technology. soils. 2020 IMEKO TC-4 International Conference on
Agaiby, S. S., Cargill, E., Ku, T. & Mayne, P. W. (2016) Metrology for Archaeology and Cultural Heritage.
Continuous-interval seismic piezocone testing in Pied Trento, Italy.
mont residuum. IN Lehane, B. M., Acosta Clayton, C. R. I. (2011) Stiffness at small strain: research
martÍnez, H. E. & Kelly, R. B. (Eds.) 5th International and practice. Géotechnique, 61, 5–37.
Conference on Geotechnical and Geophysical Site Comina, C., Krawczyk, C. M., Polom, U. & Socco, L. V.
Characterisation (ISC’5) Gold Coast, Australia, Austra (2017) Integration of SH seismic reflection and Love-
lian Geomechanics Society. wave dispersion data for shear wave velocity determin
Ahmed, S. A. (2017) Correlating the shear wave velocity ation over quick clays. Geophysical Journal Inter
with the cone penetration test. 2nd World Congress on national, 210, 1922–1931.
Civil, Structural, and Environmental Engineering Dahl, M. & Haugen, E. (2022) Evaluation of selected pro
(CSEE’17). Barcelona, Spain. cedures for estimating shear wave velocity with SCPTu
ASTM (2019) Standard test methods for downhole seismic and bender element. 20th International Conference on
testing, ASTM D7400-19. West Conshohocken, PA, Soil Mechanics and Geotechnical Engineering
USA, ASTM International. (ICSMGE). Sydney, Australia, International Society for
Bang, E.-S., Cho, S.-J. & Kim, D.-S. (2014) Mean refracted Soil Mechanics and Geotechnical Engineering
ray path method for reliable downhole seismic data (ISSMGE).
interpretations. Soil Dynamics and Earthquake Engin Dannewitz, N., Eriksson, H., Mattsson, H., Larsson, R.
eering- Elsevier Press, 65. & Holm, G. (2005) Utvecklingsprojekt 11357 - sesi
Bates, C. R. & Philips, D. R. (2000) Multi-component seis misk kontrollmetod för kc-pelare - sluttrapport
mic surveying for near surface investigations: examples 2005-05–31.
from central Wyoming and southern England. Applied De lange, G., Rawlings, C. G. & Willet, N. (1990) Com
Geophysics, 22, 257–272. parison of shear moduli from offshore seismic cone tests
Baziw, E. J. (1993) Digital filtering techniques for inter and resonant column and piezoceramic bender element
preting seismic cone data. Journal of Geotechnical laboratory tests. Society of Underwater Technology
Engineering ASCE, 119, 98–108. (SUT), 16, 13–20.
Baziw, E. J. (2002) Derivation of seismic cone interval vel Degroot, D. J. & Lutenegger, A. J. (2003) Geology and
ocities utilizing forward modeling and the downhill sim engineering properties of Connecticut valley varved
plex method. Canadian Geotechnical Journal, 39, clay. IN Tan, T. S., Phoon, K. K., Hight, D. W. &
1181–1192. Leroueil, S. (Eds.) In Characterization and Engineering
Bouchard, S., Ali, H., Leboeuf, D., Leroueil, S. & Properties of Natural Soils: Proceedings of the Inter
Cascante, G. (2017) Dynamic properties of sensitive national Workshop. Singapore A.A. Balkema,
clay deposits. IN Thakur, V., L’heureux, J.-S. & Rotterdam.
Locat, A. (Eds.) 2nd International Workshop on Land DI buò, B., D’ignazio, M., Selānpāā, J., Haikola, M.,
slides in Sensitive Clays (IWLSC), Chapter 15 in Land Lānsivaara, T. & Di sante, M. (2019) Investigation and
slides in Sensitive Clays, Advances in Natural and geotechnical characterisation of Perniö clay, Finland.
Technological Hazards Research 46. Trondheim, AIMS Geosciences, 5, 591–616.
Norway, Springer International Publishing AG. DI buò, B., D’ignazio, M., selānpāā, J. & Lānsivaara, T.
Burghignoli, A., Cavalera, L., Chieppa, V., (2016) Preliminary results from a study aiming to
Jamiolkowski, M., Mancuso, C., Marchetti, S., Pane, V., improve ground investigation data. 17th Nordic Geo-
Paoliani, P., Silvestri, F., Vinale, F. & Vittori, E. (1991) technical Meeting, NGM 2016, Reykjavik. Reykjavik,
Geotechnical Characterization of Fucino Clay. 10th Iceland.
European Conferece for Soil Mechanics and Foundation DI buò, B., Selānpāā, J., Lānsivaara, T. & D’ignazio, M.
Engineering (ECSMFE). Fierenze, Italy, Balkema, (2018) Evaluation of existing CPTu-based correlations
Rotterdam. fo rthe deformation properties of Finnish soft clays. IN
Butcher, A. P., Campanella, R. G., Kaynia, A. M. & Hicks, M. A., Pisanò, F. & Peuchen, J. (Eds.) CPT18.
Massarsch, K. R. (2005) Seismic cone downhole Delft, The Netherlands.
procedure to measure shear wave velocity Donohue, S. & Long, M. (2010) Assessment of sample
a guideline. ISSMGE TC10: Geophysical Testing in quality in soft clay using shear wave velocity and suc
Geotechnical Engineering. ISSMGE Bulletin tion measurements. Géotechnique, 60, 883–889.
April 2015 issue. Duan, W., Cai, G., Liu, S. & Puppala, A. J. (2019a) Correl
Butcher, A. P. & Powell, J. J. M. (1995) The effects of geo ations between shear wave velocity and geotechnical
logical history on the dynamic stiffness in soils. XIth parameters for Jiangsu clays of China. Pure and Applied
European Conference on Soil Mechanics and Founda Geophysics, 176, 669–684.
tion Engineering (ECSMFE). Copenhagen, Danish Geo- Duan, W., Cai, G., Liu, S., Puppala, A. J. & Chen, R.
technical Society (DGF). (2019b) In-situ evaluation of undrained shear strength
Butcher, A. P. & Powell, J. J. M. (1996) Practical consider from seismic piezocone penetration tests for soft marine
ations for field geophysical techniques used to assess clay in Jiangsu, China. Transportation Geotechnics, 20,
ground stiffness Proceedings of the Conferences on 100253.
Advances in Site Investigation Practice London, U.K., Elbeggo, D., Ethier, Y., Dubé, J. S. & Karray, M.
Thomas Telford London. (2021) Critical Insights in laboratory shear wave
47
velocity correlations of clays. Canadian Geotechnical Hardin, B. O. (1978) The nature of stress – strain behaviour
Journal, Accepted manuscript. for soils. In Proceedings ASCE Speciality Conference
Emdal, A., Gylland, A., Amundsen, H. A., Kåsin, K. & on Earthquake Engineering and Soil Dynamics, Pasa
Long, M. (2016) The mini-block sampler. Canadian dena, California.
Geotechnical Journal, 53, 1235–1245. Hight, D. W., Gasparre, A., Nishimura, S., Minh, N. A.,
EN (2004) EN ISO 1998-1:2004: Eurocode 8: Design of Jardine, R. J. & Coop, M. R. (2007) Characteristics of
structures for earthquake resistance - Part 1: General the London clay from the Terminal 5 site at Heathrow
rules, seismic actions and rules for buildings European Airport. Géotechnique, 57, 3–18.
Committee for Standardisation (CEN), Brussels, Hight, D. W., Mcmillan, F., Powell, J. J. M.,
Belgium. Jardine, R. J. & Allenou, C. P. (2003) Some charac
Fabien-ouellet, G., Fortier, R. & Giroux, B. (2014) Joint teristics of London Clay. IN Tan, T. S., Phoon,K. K.,
acquisition and processing of seismic reflections and Hight, D. W. & Leroueil, S. (Eds.) Proceedings Inter
surface waves in a sensitive clay deposit in the Out national Workshop on Charaterisation and Engineer
aouais region (Québec), Canada. IN L’heureux, J.-S., ing Properties of Natural Soils. Singapore, Balkema,
Locat, A., Leroueil, S., Demers, D. & Locat, J. (Eds.) Rotterdam.
Advances in Natural and Technological Hazard Holmsgaard, R., Ibsen, L. B. & Nielsen, B. N. (2016) Inter
Research 36, From Geoscience to Risk Management, 1st pretation of seismic cone penetration testing in silty soil.
International Workshop on Landslides in Sensitve Clays Electronic Journal of Geotechnical Engineering, EJGE,
(IWLSC). Québec, Springer. 21, 4759–4479.
Fahey, M. & Carter, J. P. (1993) A finite element study of Horng, V., Tanaka, H. & Obara, T. (2010) Effects of
the pressuremeter in sand using a nonlinear elastic plas sampling tube geometry on soft clayey sample qual
tic model. Canadian Geotechnical Journal, 30, ity evaluated by nondestructive methods. Soils and
348–362. Foundations Japanese Geotechnical Society, 50,
Ferreira, C., Vianna Da Fonseca, A. & Nash, D. F.T. (2011) 93–107.
Shear wave velocities for sample quality assessment on Hosseini-Kamal, R., Coop, M. R., Jardine, R. J. &
a residual soil. Soils and Foundations - Japanese Geo- Brosse, A. (2014) The post-yield behaviour of four
technical Society, 51, 683–692. Eocene-to-Jurassic UK stiff clays. Géotechnique, 64,
Foti, S., Lancellotta, R., Marchetti, D., Monaco, P. & 620–634.
TOTANI, G. (2006) Interpretation of SDMT tests in ISO (2014) ISO 19901-8: International Organization for
a transversely isotropic medium. IN Failmezger, R. A. Standardization, Petroleum and Natural Gas Indus
& Anderson, J. B. (Eds.) Second International Confer tries -Specific Requirements for Offshore Structures –
ence on the Flat Dilatometer. Washington, D.C., USA. Part 8: Marine Soil Investiga-tions. . Geneva, ISO/
Garofalo, F., Foti, S., Hollender, F., Bard, P. Y., Cornou, C., CEN.
Cox, B. R., Dechamp, A., Ohrnberger, M., Verron, V., Jamiolkowski, M. & Pepe, M. C. (2001) Vertical yield
Sicilia, D., Teague, D. & Vergniault, C. (2016) Interpa stress of Pisa clay from piezocone tests. Journal Geo-
cific project: Comparison of invasive and non-invasive technical and Geoenvironmental Engineering, ASCE,
methods for seismic site characterization. Part II: 127, 893–897.
Intra-comparison between surface wave and borehole Janbu, N. (1963) Soil compressibility as determined by
methods. Soil Dynamics and Earthquake Engineering - oedometer and triaxial tests. Proceedings of the 3rd
Elsevier Press, 82, 241–254. European Conference on Soil Mechanics and Founda
Gordon, M. A., Clayton, C. R. I., Thomas, T. C. & tion Engineering. Wiesbaden, Germany, Deutsche
Matthews, M. C. (1996) The selection and interpretation Gesellschaft für Erd-und Graundbau e.V.
of seismic geophysical methods for site investigation. Janbu, N. (1969) The resistance concept applied to deform
Proceedings of the Conferences on Advances in Site ations of soils. Proceedings of the 7th International Soil
Investigation Practice London, U.K., Thomas Telford Mechanics and Foundation Engineering Conference.
London. Mexico City, A.A. Balkema, Rotterdam.
Granskār, J. (2018) Evaluation of SCPT-surveys as method Joh, S. H. & Mok, Y. J. (1998) Development of an inver
for accessing dynamic modulus, MSc Thesis Luleå Uni sion analysis technique for downhole testing and con
versity of Technology, Department of Civil, Environ tinuous seismic CPT. Journal of the Korean
mental and Natural Resources Engineering. Geotechnical Society, 14, 95–108.
Gu, X., Jiang, W., Qian, J., Huang, M. & Shen, P. (2021) Kania, A., Madshus, C. & Zackrisson, P. (2000) Ground
Sampling disturbance evaluation based on the shear vibration from high-speed trains: prediction and counter
wave velocity measured in laboratory and field tests. 6th measure. Journal Geotechnical and Geoenvironmental
International Conference on Geotechnical and Geo Engineering, ASCE, 126, 531–537.
physical Site Characterisation (ISC’6). Budapest, Hun Karray, M. & Hussien, M. N. (2017) Shear wave velocity
garian Geotechnical Society. as a function of cone penetration resistance and grain
Gundersen, A. S., Hansen, R. C., Lunne, T., L’heureux, J.-S. size for Holocene-age cemented soils: a new
& Strandvik, S. O. (2019) Characterization and engineer perspective. Acta Geotechnica, 12, 1129–1158.
ing properties of the NGTS Onsøy soft clay site. AIMS Kim, D. S., Bang, E. S. & Kim, W. C. (2004) Evaluation of
Geosciences, 5, 665–703. various downhole data reduction methods to obtain reli
Hallal, M. M. & Cox, B. R. (2019) Theoretical evaluation able Vs profile. ASTM Geotechnical Testing Journal, 27,
of the interval method commonly used for downhole 585–597.
seismic testing. ASCE Geo-Congress 2019: 8th Inter Ku, T. & Mayne, P. W. (2012) Frequent-interval SDMT and
national Conference on Case Histories in Geotechnical continuous SCPTu for detailed shear wave velocity pro
Engineering. Philadelphia, Penn., USA, American Soci filing in soils. Geotechnical Engineering Journal of the
ety of Civil Engineering, Reston, Va. SEAGS & AGSSEA, 43, 34–40.
48
Ku, T. & Mayne, P. W. (2013) Evaluating the in situ lateral Leroueil, S. & Hight, D. W. (2003) Behavior and properties
stress coefficient (K0) of soils via paired shear wave vel of natural soils and soft rocks. IN Tan, T. S.,
ocity modes. ASCE Journal of Geotechnical and Geoen Phoon, K. K., Hight, D. W. & Leroueil, S. (Eds. Pro
vironmental Engineering, 139, 775–787. ceedings International Workshop on Charaterisation
Ku, T. & Mayne, P. W. (2014) Stress history profiling in and Engineering Properties of Natural Soils. Singapore,
soils using OCD–G0 anisotropy relationship. Proceed Balkema, Rotterdam.)
ing Institute of Civil Engineers, Geotechnical Engineer Lessi-Cheimariou, A., Tromans, I., Hadlow, N.,
ing, 167, 476–490. Floyd, M. & Pateman, J. (2019) A novel technique
Ku, T., Mayne, P. W. & Cargill, E. (2013a) Continuous- for deep sesimic cone tests in challenging ground
interval shear wave velocity profiling by auto-source condiitons. 2019 SECED Conference - Earthquake
and seismic piezocone tests. Canadian Geotechnical risk and engineering towards a resilient world. Gren
Journal, 50, 382–390. wich, London, Society for Earthquake and Civil
Ku, T., Weemees, I., Cargill, E., Mayne, P. W. & Engineering Dynamics.
Woeller, D. (2013b) Post-processing continuous Lo Presti, D. C. F., Jamiolkowski, M. & Pepe, M. (2003)
shear wave signals taken during cone penetrometer Geotechnical characterization of the subsoil of Pisa
testing. ASTM Geotechnical Testing Journal, 36, Tower IN Tan, T. S., Phoon, K. K., Hight, D. W. &
1–11. Leroueil, S. (Eds.) In Characterization and Engineering
L’heureux, J.-S., Lindgård, A. & Emdal, A. (2019) The Properties of Natural Soils: Proceedings of the Inter
Tiller-Flotten research site: Geotechnical characterisa national Workshop. Singapore A.A. Balkema,
tion of a sensitive clay deposit. AIMS Geosciences, 5, Rotterdam.
831–867. Lo Presti, D. C. F., Pallara, O., Maugeri, M. &
L’heureux, J.-S. & Long, M. (2017) Relationship between Cavallaro, A. (1998) Shear modulus and damping of
shear wave velocity and geotechnical parameters for a stiff marine clay from in situ and laboratory tests. IN
Norwegian clays. Journal of Geotechnical and Geoen Robertson, P. K. & Mayne, P. W. (Eds.) Proceedings
vironmental Engineering ASCE, 04017013–1 – 1st. International Conference on Geotechnical Site
04017013–20. Characterisation. Atlanta, Georgia, Balkema,
La Rochelle, P., Sarrailh, J., Tavenas, F., Roy, M. & Rotterdam.
Leroueil, S. (1981) Causes of sampling disturbance and Long, M. & D’ignazio, M. (2020) Shear wave velocity as
design of a new sampler for sensitive soil. Canadian a tool for characterising undrained shear strength of
Geotechnical Journal, 18, 52–66. Nordic clays. 18th Nordic Geotechnical Meeting. Hel
Landon, M. E. (2007) Development of a non sinki, Finland (Conferece held virtually January 2021),
destructive sample quality assessment method for IOP Conf. Series: Earth and Environmental Science 710
soft clays. PhD Thesis, Department of Civil and (2021) 012008.
Environmental Engineering, University of Massachu Long, M. & Donohue, S. (2010) Characterisation of Nor
setts, Amherst. PhD Thesis, Department of Civil and wegian marine clays with combined shear wave velocity
Environmental Engineering, University of Massachu and CPTU data. Canadian Geotechnical Journal, 47,
setts, Amherst. 709–718.
Landon, M. E., De Groot, D. J. & Sheahan, T. C. (2007) Long, M. & L’heureux, J.-S. (2022) Shear wave velocity
Nondestructive sample quality assessment of a soft clay SCPTU correlations for sensitive marine clays. 5th
using shear wave velocity. ASCE Journal of Geotech International Symposium on Cone Penetration Testing
nical and Geoenvironmental Engineering, 133, (CPT’22) Bologna, Italy.
424–432. Long, M., L’heureux, J.-S., Fiskvik Bache, B. K.,
Lefebvre, G., Beloueuf, D., Rahhal, M. E., Lacroix, A., Lund, A. K., Hove, S. E., Sødal, K. G.,
Warde, J. & Stokoe, K. H. (1994) Laboratory and field Amundsen, H. A., Nordal, S. & Montafia, A. (2019)
determinations of small-strain shear modulus for Site characterisation and some examples from large
a structured Champlain clay. Canadian Geotechnical scale testing at the Klett quick clay research site. AIMS
Journal, 31, 61–70. Geosciences, 5, 344–389.
Lefebvre, G. & Poulin, C. (1979) A new method of sam Long, M., Paniagua, P. & L’heureux, J.-S. (2021)
pling in sensitive clay. Canadian Geotechnical Journal, Assessing sample disturbance in low plasticity sensi
16, 226–233. tive clays using shear wave velocity. 6th Inter
Lehane, B. M. (2003) Vertically loaded shallow foun national Conference on Geotechnical and
darions on soft clayey silt. Institution of Civil Engin Geophysical Site Characterisation (ISC’6). Budapest,
eers Journal of Geotechnical Engineering, 156, Hungarian Geotechnical Society.
17–26. Long, M., Wood, T. & L’heureux, J.-S. (2017) Relationship
Lehane, B. M., J.P., D. & Schneider, J. A. (2008) Settle between shear wave velocity and geotechnical param
ment prediction for footings on sand. IN BURNS, S. E., eters for Norwegian and Swedish sensitive clays. IN
Mayne, P. W. & Santamarina, J. C. (Eds.) Deformation Thakur, V., L’heureux, J.-S. & Locat, A. (Eds.) 2nd
Characteristics of Geomaterials. Atlanta, Georgia, IOS International Workshop on Landslides in Sensitive
Press, Amsterdam. Clays (IWLSC), Chapter 6 in Landslides in Sensitive
Leroueil, S., Hamouche, K., Tavenas, F., Boudali, M., Clays, Advances in Natural and Technological Hazards
Locat, J., VIrely, D., Roy, M., LA Rochelle, P. & Research 46. Trondheim, Norway, Springer Inter
Leblond, P. (2003) Geotechnical characterisation and national Publishing AG.
properties of a sensitive clay from Québec. IN Looijien, P., Cooper, A. G. & Doan, D. H. (2017) Seis
Tan, T. S., Phoon, K. K., Hight, D. W. & Leroueil, S. mic cone testing using seafloor drill technology. 79th
(Eds.) Proceedings International Workshop on Chara EAGE Conference and Exibition. Paris, France, Euro
terisation and Engineering Properties of Natural Soils. pean Association of Geoscientists and Engineers
Singapore, Balkema, Rotterdam. (EAGE).
49
Lunne, T., Berre, T. & Strandvik, S. (1997a) Sample dis Mayne, P. W., Robertson, P. K. & Lunne, T. (1998) Clay
turbance in soft low plasticity Norwegian clay. IN stress history evaluated from seismic piezocone tests. IN
Almeida, A. (Ed.) Symposium on Recent Developments Robertson, P. K. & Mayne, P. W. (Eds.) Proceedings
in Soil and Pavement Mechanics. Rio de Janeiro, Balk 1st. International Conference on Geotechnical Site
ema, Rotterdam. Characterisation. Atlanta, Georgia, Balkema,
Lunne, T., Kelleher, P. J., Banimahd, M., Degroot, D. J., Rotterdam.
Nguyen, H. Q. & Senders, M. M. (2013) Offshore site Mayne, P. W. & Woeller, D. J. (2015) Advances in seismic
characterisatin of small strain modulus using a seabed piezocone testing. XVIth European Conference on Soil
drilling system. Offshore Technology Conference, Hous Mechanics and Geotechnical Engineering (ECSMGE)
ton, Texas. Edinburgh, Scotland, Thomas Telford Ltd.
Lunne, T., Robertson, P. K. & Powell, J. J. M. (1997b) Mcadam, R. A., Byrne, B. W., Houlsby, G. T.,
Cone Penetration Testing in Geotechnical Practice, Beuckelaers, W. J., Burd, H. J., Gavin, K. G.,
Blackie Academic and Professional, London. Igoe, D. J. P., Jardine, R. J., Martin, C. M., Muir
Lunne, T., Strandvik, S., Kåsin, K., L’heureux, J.- S., Wood, A., Potts, D. M., Gretlund, J. S.,
Haugen, E., Uruci, E., Veldhuijzen, A., Carlson, M. & Tarborda, D. M. G. & Zdravkovic, L. (2020) Monotonic
Kassner, M. (2018) Effect of cone penetrometer type on laterally loaded pile testing in a dense marine sand at
CPTU results at a soft clay test site in Norway. IN Dunkirk. Géotechnique, 70, 986–998.
Hicks, M. A., Pisanó, F. & Peuchen, J. (Eds.) CPT18. Mcgann, C. R., Bradley, B. A., Taylor, M. L.,
Delft, The Netherlands. Wotherspoon, L. M. & Cubrinovski, M. (2015)
Marchetti, D. (2015) Applications and Recent Develop Applicability of existing empirical shear wave vel
ments of the Flat Dilatometer (DMT) and Seismic Dilat ocity correlations to seismic cone penetration test
ometer (SDMT). XVIth European Conference on Soil data in Christchurch New Zealand. Soil Dynamics
Mechanics and Geotechnical Engineering (ECSMGE) and Earthquake Engineering - Elsevier Press, 75,
Edinburgh, Scotland, Thomas Telford Ltd. 76–86.
Marchetti, S., Monaco, P., Totani, G & Marchetti, D. Mcgillivray, A. V. & Mayne, P. W. (2004) Seismic
(2008) In situ tests by seismic dilatometer (SDMT). piezocone and seismic flat dilatometer tests at
From Research to Practice in Geotechnical Engineer Treporti. IN Viana Da Fonseca, A. & MAYNE, P. W.
ing, A Geotechnical Special Publication Honoring (Eds.) 2nd International Conference on Geotechnical
John H. Schmertmann. ASCE GeoInstitute. and Geophysical Site Characterisation, ISC’2. Porto,
Matthews, M. C., Hope, V. S. & Clayton, C. R. I. (1996) Millpress.
The use of surface waves in the determination of ground Mcgillivray, A. V. & Mayne, P. W. (2008) An automated
stiffness profiles. Institution of Civil Engineers Journal seismic source for continuous-push shear wave velocity
of Geotechnical Engineering, 119, 84–95. profiling with SCPT and frequent-interval SDMT. IN
Mayne, P. W. (2000) Enhanced geotechnical site character Huang, A. B. & Mayne, P. W. (Eds.) 3rd International
ization by seismic piezocone penetration tests. Invited Conference on Geotechnical and Geophysical Site
lecture, Fourth International Geotechnical Conferece, . Characterization - ISC’3. Taipei, Taylor and Francis,
Caoiro University, Egypt. London.
Mayne, P. W. (2005a) Class “A” footing response predic Menzies, B. & Matthews, M. C. (1996) The continuous sur
tion from sesimic cone tests. IN Di Benedetto, H., face wave system: a modern technique for site
Doanh, T., Geoffroy, H. & Sauzéat, C. (Eds.) 3rd Inter investigation. Special Lecture at the Indian Geotech
national Symposium on Deformation Characteristics of nical Conference. Madras, Geotechnical Data Systems
Geomaterials. Lyon, France, Swets and Zeitlinger, (GDS), Egham, Surrey, UK.
Lisse. Moon, S.-W. & Ku, T. (2016a) Development of global
Mayne, P. W. (2005b) Keynote: Integrated ground behav correlation models between in situ stress-normalized
ior: in-situ and laboratory tests. IN Di Benedetto, H., shear wave velocity and soil unit weight for plastic
Doanh, T., Geoffroy, H. & Sauzéat, C. (Eds.) Inter soils. Canadian Geotechnical Journal, 53,
national Symposium on Deformation Characteristics of 1600–1611.
Geomaterials. Lyon, Taylor and Francis. Moon, S.-W. & Ku, T. (2016b) Empirical estimation of soil
Mayne, P. W. (2007) Cone Penetration Testing: unit weight and undrained shear strength from of shear
A Synthesis of Highway Practice, National Cooperative wave velocity measurements. IN Lehane, B. M., Acosta-
Highway Research Program Synthesis 368. Washington, Martínez, H. E. & Kelly, R. B. (Eds.) 5th International
D.C., Transport Research Board, National Cooperative Conference on Geotechnical and Geophysical Site Char
Highway Research Program. acterisation (ISC’5) Gold Coast, Australia, Australian
Mayne, P. W., Cargill, E. & Miller, B. (2019) Geotechnical Geomechanics Society.
characteristics of sensitive Leda clay at Canada test site Nguyen,H. Q., Degroot, D. J., Meidani, M.,
in Gloucester, Ontario. AIMS Geosciences, 5, 390–411. Vanneste, M. & Lunne, T. (2015) Measurement and
Mayne, P. W., Coop, M. R., Springman, S. M., interpretation of downhole sesimic probe data for
Huang, A. B. & Zornberg, J. G. (2009) Geomaterial estimating shear wave velocity in deep-water
behavior and testing - Keynote lecture. IN Hamza, M. environments. IN MAYER, V. (Ed.) International
(Ed.) 17th International Conference on Soil Mechanics Symposium on Frontiers in Offshore Geotechnics
and Geotechnical Engineering (ICSMGE). Alexandria, (ISFOG III). Oslo, Norway, CRC Press.
Egypt, Millpress/IOS Press Rotterdam. Nibs (2003) National Institute of Building Sciences/Build
Mayne, P. W. & Mcgillivray, A. V. (2008) Improved shear ing Seismic Safety Council - National Earthquake
wave measurements using autoseis sources. Deform Hazard Reduction Program - Recommended provisions
ational Characteristics of Geomaterials. Amsterdam, for seismic regulations for new buildings and other
The Netherlands, Millpress/IOS Press. structures (Federal Emergency Management Agency
50
450). Part 1: Provisions. Washington, DC, Building Conference on Geotechnical and Geophysical Site
Seismic Safety Council. Characterisation, ISC’2. Porto, Millpress.
Omar, M. N., Abbiss, C. P., Taha, M. R. & Powell, J. J. M., Dhimitri, L., Ward, D. & Butcher, A. P.
Nayan, K. A. M. (2011) Prediction of long-term (2016) Small strain stiffness assessment from in situ
settlement on soft clay using shear wave velocity and tests - revisited. IN Lehane, B. M., Acosta-
damping characteristics. Engineering Geology, 123, Martínez, H. E. & Kelly, R. B. (Eds.) 5th International
259–270. Conference on Geotechnical and Geophysical Site
Paniagua, P., L’heureux, J.-S., Carroll, R., Kåsin, K., Characterisation (ISC’5) Gold Coast, Australia, Austra
Sjursen, M. & Amundsen, H. A. (2017) Evaluation of lian Geomechanics Society.
sample disturbance of three Norwegian clays. 19th Ramsey, N. (2020) Optimising Geotechnical Engineering
International Conference on Soil Mechanics and Models (GEMs). IN S., H., S., P. & R., G. (Eds.)
Geotechnical Engineering (ICSMGE). Seoul, South Advances in Offshore Geotechnics. Lecture Notes in
Korea. Civil Engineering, Vol 92. Singapore, Springer.
Paoletti, L., Mouton, E. & Liposcak, I. (2013) Comparison Rice, A. H. (1984) The seismic cone penetrometer. Master
of underwater MASW, seismic CPT and downhole of Science, Civil Engineering, University of British Col
methods offshore Croatia IN Coutinho, R. Q. & umbia, Vancouver, B.C.
Mayne, P. W. (Eds.) Proceedings 4th. International Robertson, P. K. (1990) Soil classification using the cone
Conference on Geotechnical and Geophysical Site penetration test. Canadian Geotechnical Journal, 27,
Characterisation (ISC’4) Recife, Brazil, 151–158.
September 2012. Robertson, P. K. (2009) Interpretation of cone penetration
Parasie, N., Franken, T. & Peuchen, J. (2022) Assessment tests - a unified approach. Canadian Geotechnical Jour
of seismic cone penetration testing for small strain shear nal, 46, 1337–1355.
modulus. 5th International Symposium on Cone Pene Robertson, P. K. (2015) Comparing CPT and Vs liquefac
tration Testing (CPT’22) Bologna, Italy. tion triggering methods. ASCE Journal of Geotechnical
Peuchen, J., De Ruijter, M. R., Hospers, B. & Assen, R. L. and Geoenvironmental Engineering, 141, 04015037–1–
(2002) Shear wave velocity integrated in offshore geo 04015037–10.
technical practice. SUT-OSIG-02-379 Society of Under Robertson, P. K., Campanella, R. G., Gillespie, D. &
water Technology Conference. London, UK (2002), Rice, A. (1986) Seismic CPT to measure in situ shear
Society of Underwater Technology. wave velocity. Journal of Geotechnical Engineering
Peuchen, J. & Gomez Meyer, E. (2021) Geo-intelligence ASCE, 112, 791–803.
from databases of offshore in situ tests in public Robertson, P. K., Sasitharan, S., Cunning, J. C. &
domain. 6th International Conference on Geotechnical Sego, D. C. (1995) Shear wave velocity to evaluate flow
and Geophysical Site Characterisation (ISC’6). Buda liquefaction. ASCE Journal Geotechnical Engineering,
pest, Hungarian Geotechnical Society. 121, 262–273.
Peuchen, J., Kaltekis, K., Klein, M., Murali, M., Van Salsabili, M., Saeidi, A., Rouleau, A. & Nastev, M. (2022)
Erp, F. C. W. & Hicks, M. A. (2020) Characteristic Development of empirical CPTu-Vs correlations for
values for geotechnical design of offshore monopiles post-glacial sediments in Southern Quebec, Canada, in
in sandy soils – case study. In(ed.).: proceedings (pp. consideration of soil type and geological setting. Soil
1892–1901). . IN Westgate, Z. (Ed. 4th International Dynamics and Earthquake Engineering - Elsevier Press,
Symposium on Frontiers in Offshore Geotechnics 154.
(ISFOG 2020). Deep Foundations Institute. Shahri, A. A. & Naderi, S. (2016) Modified correlations to
Pidlisecky, A. & Haines, S. S. (2011) Bayesian approach predict the shear wave velocity using piezocone penetra
for determining velocity and uncertainty estimates from tion test data and geotechnical parameters: a case study
seismic cone penetrometer testing or vertical seismic in the southwest of Sweden. Innov. Infrastruct. Solut.,
profiling data. Canadian Geotechnical Journal, 48, Springer, 1:13.
1061–1069. Simonini, P., Ricceri, G. & Cola,S. (2007) Geotechnical
Piriyakul, K. & Haegeman, W. (2008) Anisotropic shear characterization and properties of Venice lagoon hetero
wave velocity of overconsolidated Boom clay. IN geneous silts. IN Tan, T. S., Phoon, K. K., Hight, D. W.
Huang, A. B. & Mayne, P. W. (Eds.) 3rd International & Leroueil, S. (Eds.) Proceedings 2nd International
Conference on Geotechnical and Geophysical Site Workshop on Characterisation and Engineering Proper
Characterization - ISC’3. Taipei, Taylor and Francis, ties of Natural Soils (“Natural Soils 2006”). Taylor and
London. Francis Group, London.
Poulos, H. G. (2021) Use of shear wave velocity for foun Soccodato, F. M. (2003) Geotechnical properties of Fucino
dation design. Geotechnical and Geological Engineer clayey soil. IN Tan, T. S., Phoon, K. K., Hight, D. W. &
ing, Published online October 2021. Leroueil, S. (Eds.) In Characterization and Engineering
Powell, J. J. M. (2005) Session report: Technical session Properties of Natural Soils: Proceedings of the Inter
1c: In situ testing. Proc. 16th International Conference national Workshop. Singapore A.A. Balkema,
on Soil Mechanics and Geotechnical Engineering. Rotterdam.
Osaka, Balkema, Rotterdam. Stolte, A. C. & Cox, B. R. (2020) Towards consideration of
Powell, J. J. M. & Butcher, A. P. (1991) Assessment of epistemic uncertainty in shear-wave velocity measure
ground stiffness from field and laboratory tests In ments obtained via seismic cone penetration testing
Proceedings Xth ECSMFE. Fierenze, Rotterdam, (SCPT). Canadian Geotechnical Journal, 57, 48–60.
Balkema. Styler, M. A. & Weemees, I. (2016) Evaluating and report
Powell, J. J. M. & Butcher, A. P. (2004) Small strain stiff ing uncertainty in downhole shear wave velocities. 3rd
ness assessments from in situ tests. IN Viana Da International Conference on Performance-based
Fonseca, A. & Mayne, P. W. (Eds.) 2nd International Design in Earthquake Geotechnical Engineering.
51
Vancouver, Canada, International Society for Soil Mech Trafford, A. & Long, M. (2020) Relationship between
anics and Geotechnical Engineering, London, UK. shear wave velocity and undrained shear strength of
Tanaka, H., Hamouche, K. K., Tanaka, M., Watabe, Y., peat ASCE Journal of Geotechnical and Geoenviron
Leroueil, S. & Fournier, I. (1998) Use of Japanese sam mental Engineering, 146, 04020057–1 - 04020057–10.
pler in Champlain Sea clay. IN Robertson, P. K. & Trak, B., La Rochelle, P., Tavenas, F., Leroueil, S. &
Mayne, P. W. (Eds.) Proceedings 1st. International Con Roy, M. (1980) A new appeoach to the stability analysis
ference on Geotechnical Site Characterisation. Atlanta, of embankments on sensitive clays. Canadian Geotech
Georgia, Balkema, Rotterdam. nical Journal, 17, 526–544.
Tanaka, H., Locat, J., Shibuya, S., Soon, T. T. & Valsson, S. M., Dahl, M., Haugen, E. & Degago, S. A.
Shiwakoti, D. R. (2001) Characterisation of Singapore, (2020) Estimating shear wave velocity with the SCPTU
Bangkok and Ariake clays. Canadian Geotechnical and bender element. 18th Nordic Geotechnical Meeting.
Journal, 38, 378–400. Helsinki, Finland (Conferece held virtually Jan
Teachavorasinskun, S. & Lukkunaprasit, P. (2004) uary 2021), IOP Conf. Series: Earth and Environmental
A simple correlation for shear wave velocity of soft Science 710 (2021) 012017.
Bangkok clays. Géotechnique, 54, 323–326. Wang, H., Lehane, B. M., Bransby, M. F., Askarinejad, A.,
Teague, D. P. & Cox, B. R. (2016) Site response implica Wang, L. Z. & Y., H. (2022) A simple rotational spring
tions associated with using non-unique Vs profiles from model for laterally loaded rigid piles in sand. Marine
surface wave inversion in comparison with other com and Structures, Inpress.
monly used methods of accounting for Vs uncertainty. Wood, T. (2016) On the small strain stiffness of some Scan
Soil Dynamics and Earthquake Engineering - Elsevier dinavian soft clays and impact on deep excavations,
Press, 91, 87–103. PhD thesis Department of Civil and Environmental
Tong, L., Che, H., Zhang, M. & Li, H. (2018) Review of Engineering, Chalmers University of Technology, Göte
shear-wave velocity prediction equations based on borg, Sweden. Department of Civil and Environmental
piezocone penetration test data: example from Yangtze Engineering, Chalmers University of Technology, Göte
River floodplain deposits at Nanjing, Jiangsu Province, borg, Sweden. Göteborg, Sweden, Chalmers University
China. Quarterly Journal of Engineering Geology and of Technology.
Hydrogeology, 51, 229–246. Zhang, M. & Tong, L. (2017) New statistical and graphical
Trafford, A., Donohue, S., Ellwood, R., Godfrey, A. & assessment of CPT-based empirical correlations for the
Wacquier, L. (2021) The application of distributed shear wave velocity of soils. Engineering Geology, 226,
acoustic sensing for shallow marine investigations - an 184–191.
intertidal case study. Near Surface Geoscience 2021 Zou, H., Liu, S., Cai, G., Puppala, A. J. &
(NSG’21): The 27th European Meeting of Environmen Bheemasetti, T. V. (2017) Multivariate correlation ana
tal and Engineering Geophysics. Bordeaux, France (and lysis of seismic piezocone penetration (SCPTU) param
Online), European Association of Geoscientists & eters and design properties of Jiangsu quaternary
Engineers (EAGE). cohesive soils. Engineering Geology, 228, 11–38.
52
Invited papers
© 2022 the Author(s), ISBN 978-1-032-31259-0
Uncertainties associated with CPT data acquisition

R. Soage Santos
Ørsted, London, UK
ABSTRACT: Cone penetration test (CPT) and piezocone penetration test (CPTU) are the dominant in situ
testing techniques for soil investigations. CPT/CPTU data is used directly in design or used to derive geotech
nical properties for design of infrastructure. Occasionally, acquired CPT/CPTU data exhibit anomalies which
can cause uncertainties about use in geotechnical design. Consequently, it is very important to understand,
and when possible, minimize the uncertainty associated to CPT/CPTU measurements so these can be con
sidered in planning, design, and installation of infrastructure. The practical challenges associated to the quan
tification of uncertainty and the difficulties of its standardization are discussed. A comprehensive review of
uncertainty contributors associated to CPT/CPTU measurements is provided. This document discusses and
proposes a standardization approach that combines performance-based cone classification system based on
calibration data, with a method-based approach to control and minimize the uncertainty contributions rising
from execution of CPT/CPTU.
1 INTRODUCTION uncertainty is even more important when using the

CPT/CPTU measurements directly in design. Note
The cone penetration test and piezocone penetration that additional uncertainty is introduced when infer
test (CPT/CPTU) are the most dominant in situ testing ring geotechnical parameters from CPT/CPTU data.
technique for soil investigations. CPT/CPTU provide Additional uncertainties associated to data interpret
valuable input for safe and economical design of ation is out of the scope of this document.
many structures and therefore, assessing the represen The objective of this document is to propose a
tativeness of the CPT/CPTU results is essential. way forward for maximizing the quality and repre
This document broadly reviews the challenges sentativeness of the CPT/CPTU results by control
related to the assessment of CPT/CPTU results qual ling and reducing of the uncertainties.
ity. Most often, the CPT provides reliable data with Before proposing a new approach and method
depth, particularly the independent parameters qc ology, it is necessary to understand and review
(cone resistance), fs (sleeve friction), and u2 (pore potential sources of uncertainty (uncertainty con
pressure). Occasionally, acquired CPT/CPTU data tributors) and previous standardization efforts aimed
exhibit anomalies which can cause uncertainties about to control and minimize uncertainty. The following
applicability in geotechnical design of offshore struc sections will review standardization efforts and dis
tures. Peuchen et al. 2020 provides detailed examples cusses the challenges associated to standardization.
of data anomalies and discusses how to avoid them.
CPT/CPTU are among the most fundamental
datasets used to define soil layers and infer param 2 CPT/CPTU STANDARDS
eters for safe geotechnical design of infrastructure
and is often used directly in design of foundations
2.1 Development in requirements
(for instance pile capacity calculations). Thus, it is
very important to understand the quality and there There have been several documents describing
fore the uncertainties associated to CPT/CPTU requirements and recommendations for the cone pene
results that are to be applied in design. The term tration test. Before 1980s only cone penetrometers
uncertainty of measurement intends to express the without pore pressure measurements were included.
doubt regarding the validity/representativeness of But, in later years the CPTU, or piezocone, has also
the results. Quantification of uncertainty is very been covered. In 1977 the Society of Soil Mechanics
important, because ignoring uncertainty may result and Foundation Engineering (ISSMFE) issued a docu
in the choice of non-optimal geotechnical design and ment that included recommendations regarding cone
decision making which could lead to costly and geometry and provided some requirement to preci
unsafe infrastructure development. Understanding sion: The precision to be obtained should not be
DOI: 10.1201/9781003308829-4
55
worse than the larger of the following values: 5 % of near impossible and this presents a fundamental
the measured value or 1 % of the measuring range of problem with the demonstration of compliance.
the sensor used. Later documents have left out such The ASTM, 2020, approach avoids the intrinsic
references to the measuring range. difficulties associated with expressing requirements
The Swedish Geotechnical Society Guideline in in terms of uncertainty, by expressing requirements
1993 (SGF, 1993) introduced the concept of accuracy in terms of repeatability or precision. ASTM, 2020
classes where the requirements were related to what states: “There are little direct data on the preci
the results were to be used for. ASTM released, in sion of this test method, in particular because the
1979, the first CPT standard publish by a national natural variability of the ground” and merely pro
standardization institute. ASTM did not embrace the vides informative and generic estimates of the pre
accuracy class approach but just included reference to cision of qc, fs and u2 from experience in uniform
estimates of precision of cone resistance (qc) and deposits. ASTM, 2020 clearly attempts to mitigate
sleeve friction (fs). Later updates in ASTM 5778-12 or diminish the uncertainty associated with the
(ASTM, 2012) and ASTM 5778-20 (ASTM, 2020) fol CPT/CPTU measurements by providing procedural
lows the same approach providing estimates of requirements regarding the calibration of cone
expected precision for qc, fs, and u2 and providing penetrometer in a laboratory environment and sets
requirements for the calibration errors expressed as minimum requirements for the precision of qc, fs
maximum allowable errors as percentage of full-scale and u2 in relation to a series of potential sources
output (FSO). of uncertainty, although omits some important con
In 2012, the first CPT/CPTU standard was pub tributors. The most obvious omission is the
lished by an international standardization institution. requirement for the minimum uncertainty (or pre
ISO 22476-1:2012 (ISO 2012) opted for an applica cision) of the reference source used in calibration.
tion class scheme, defining four classes for given ASTM, 2020 states: “In cases where higher level
soil profiles and the use of CPT results. Each appli of precision is required, stricter calibration
cation class specifies minimum accuracy require requirements would be required” however not fur
ments for qc, fs and u2 with an associated degree of ther advice is provided regarding the cases where
uncertainty. In the ISO standard for Marine Soil higher precision is required or how to make the
Investigations 19908-1:2014 (ISO 2014) a simplifi calibration requirements stricter. In summary
cation was made to accommodate offshore practice ASTM, 2020, focuses on providing requirements
resulting in three application classes. for the calibration of cone penetrometers and for
field methodology but does not provide require
ments for assessing performance in the field.
2.2 Present status
None of these standards provide unequivocal guide
Two principal CPT/CPTU standards widely used are lines on how to evaluate data quality. They only discuss
ISO 2012 and ASTM 2020. While the overall pro some of the factors that could influence the data quality.
cedural approach is rather aligned between these two A number of methods are used throughout the industry,
standards, there are still some fundamental differ and the approaches vary from company to company.
ences regarding the uncertainty requirements of the This makes it very difficult to compare the quality of
CPT/CPTU results. CPT/CPTU results from various systems and providers
ISO 2012 and ISO 2014 propose a performance- and creates difficulty for geotechnical practitioners,
based approach to classify CPT/CPTU results into who may often need to take conservative judgements
application classes defined by limiting values of for design. The practical challenge related to this topic
accuracy. Thus, these standards expresses the uncer is setting up robust and detailed guidelines for the
tainty requirements in terms of accuracy: “If all pos assessment of data quality of a performed test more or
sible sources of errors are added, the minimum less immediately upon completion of the CPTU, so it
accuracy of the recorded measurements should be can be repeated if necessary.
better than the largest of the values given in Table 1.
The uncertainty analysis should include internal fric
2.3 Present challenges using standards
tion, errors in the data acquisition, eccentric load
ing, temperature (ambient and transient) effects, Recent publications discussed in depth the difficul
pore pressure effects in gaps below and above the ties of using ISO 2012 and ISO 2014 to assess
friction sleeve, and dimensional errors. “ the quality of CPT/CPTU results (Peuchen and
While the statement has the right intention it is Terwindt 2014 and 2015, Lunne et al 2017). The
challenging to implement in practice. Using the term main difficulty arises from the definition of appli
accuracy implies the knowledge of the true value. It cation class and the lack of procedural detail for
is not possible, to prove and document, with 100% the classification of CPT/CPTU results in practise.
confidence, how accurate the CPTU measurements The vague nature of the recommendations opens
are, since the true values are not known. The stand up for different interpretation and can cause con
ard provides little guidance as to how errors and tractual disputes. Some of the practical questions
accuracy can be calculated. In practice, identifying that CPT/CPTU data providers and users are
and quantifying ‘all possible sources of error’ is facing are listed here:
56
• How to determine and document that a certain measurement can be expressed as the positive square
Application Class can be obtained before the start root of the sum of the uncertainty contributors. The
of field work, for instance in the tender phase? expanded standard uncertainty can then be used to
• Regarding the CPTU parameter accuracy require express the probability or level of confidence of the
ments: is it an absolute requirement that all three estimations of uncertainty by using a coverage
parameters (qc, fs and u2) shall fulfill the limiting factor.
values given in Table 1? Should the main Some of the uncertainty contributors may be
emphasis be given to the measured parameters, evaluated by “Type A” evaluation of measurement
which are most crucial for the interpretation uncertainty. Where the statistical distribution of the
methods? quantity values from series of repeated measure
• ISO 2012 Annex D, clause D2.1 refers to zero ments under the same conditions are compared
readings before and after a test: For CPTU’s, a against a reference and can be characterized by
first attempt at recommendations for limiting standard deviations. Other uncertainty contributors
values is to follow the minimum accuracy accord may be evaluated by “Type B”, which are evalu
ing to the applications given in Table 1. How reli ations based on scientific judgement considering
able are the zero readings before and after for this probability density functions based on experience,
evaluation? How does cleaning of the cones etc. theoretical modelling, manufacturers specifications
play a role? or other information.
• What is the overall best approach for the evalu It is not possible to apply “Type A” approach to
ation of the resulting Application Class after a field measurements given the difficulties associated
test has been completed? in finding a perfectly uniform site that removes soil
• If a test is evaluated to be in a lower Application variability from the uncertainty equation. In addition,
Class than required in the project specifications, the uncertainties may also depend on the soil type,
how shall the results be treated in the definition and this make this approach unpractical.
of representative parameters for design? Peuchen and Terwindt (2014, 2015) proposed a
• From a client’s point of view: Which Application model that applies a “Type B” evaluation of uncer
Classes and default values are reasonable to tainty of CPT/CPTU profiles. Their work really rep
require in the project specifications? What is resents an in-depth evaluation of the factors that can
achievable in which conditions? influence the various measurements during a CPTU.
The analyses of uncertainties are based on input
parameters found in literature, own investigations,
some general knowledge on transducers, etc. The
3 CHALLENGES ASSOCIATED WITH THE uncertainty analysis proposed becomes rather sub
QUANTIFICATION OF UNCERTAINTY jective and hazy when weighting coefficients are
applied to the uncertainty calculation model for rep
The term uncertainty of measurement intends to resenting poor operational procedures and adverse
express the doubt regarding the validity of the soil conditions. Peuchen and Terwindt concluded
results. It follows, from the discussions in previous that the approach proposed can be considerable con
sections, that a detailed and robust methodology for servative and that further studies were required.
the quantification of the uncertainty associated with In current practice the implementation of “Type
CPT/CTU profiles is desired and necessary to B” evaluations for estimating the uncertainty of
resolve the challenges associated with the assess CPTU measurements is cumbersome and inevitably
ment of quality. Unfortunately, the quantification of involves making conservative assumptions to fill the
measurement uncertainty for CPT/CPTU results is gap of knowledge regarding the uncertainty distribu
not trivial. tions of certain components. An added challenge is
The formal definition of the term uncertainty of to reach consensus between all sectors, contractors,
measurement is, in accordance with the International design engineers, consultants, certifiers, and
Vocabulary of Metrology (VIM:2012), as follows:” researchers may have different interest from enter
non-negative parameter characterizing the disper prises involve in manufacturing or those involve in
sion of the quantity values being attributed to a mea
the distribution of services.
surand, based on the information used”. ISO/IEC Modelling uncertainty using a “Type B” is diffi
Guide 98-3:2008 adopt the use of uncertainty of cult to implement without adding excessive conser
measuring in the operational sense. This is to avoid vatism. It requires deep knowledge on the behaviour
the previous issues associated to terms like accuracy of each particular measurement system and this in
and error, which implies by definition the true value turn requires substantial research from geotechnical
needs to be known in order to calculate accuracy. practitioners, equipment operators and manufactur
ISO/IEC Guide 98-3:2008 and JCGM 100:2008 ers. In particular research effort is required to gain
provides guidance for the estimation of uncertainty sufficient understanding on:
of measurement. Measurement uncertainty com
prises, in general, many components or uncertainty • Development of standard methodologies and
contributors. The combined standard uncertainty of a requirements to test and measure the behavior of
57
a particular cone penetrometer design to minim the extent of the implementation period must be con
ize or eliminate subjective judgement when sidered to ensure that the industry is able to react to
selecting uncertainty probability distributions and requirement changes. In this regard Working Groups
sensitivity coefficients. have often to consider the standard document as
• Development of an unambiguous methodology “work in progress” and consider a staged approach
for reporting uncertainty probability distribution particularly when large changes in requirements are
and sensitivity coefficients. needed.
In summary, modelling uncertainty using a “Type
B” evaluation can be a viable option for determining 5 UNCERTAINTY CONTRIBUTORS
uncertainty of CPTU measurement, further matur
ation and consensus for calculation methodologies Whether a series of equations for quantifying the
are required, before this approach can be incorpor uncertainty associated with the CPT/CPTU measure
ated into a standardization document. ments are formulated, or whether a series of effect
ive testing methodology requirements are
established to minimise the level of uncertainty, it is
4 CHALLENGES ASSOCIATED WITH fundamental to start by mapping out the potential
STANDARDISATION sources of uncertainty and classifying their relative
weight on the overall CPT/CPTU quality.
ISO Working Groups are responsible for the devel The tables presented below shows the main uncer
opment of standardisation documents. Working tainty contributors associated to CPT/CPTU results
groups are generally composed by technical experts grouped into three categories. Tables provide a sub
and delegates appointed by the national standardisa jective qualitative assessment of the impact on
tion participating in the standard. The main objective uncertainty, as function of the maximum cone resist
of a working group is delivering the minimum ance registered in the test. Additional commentary
requirements to minimise the level of uncertainty describing the contributors and the potential mitiga
associated to results and to deliver sufficient method tion measures are included. Table 1A describes con
ology descriptions to ensure consistent and repeat tributors associated with the measurement system;
able results. Table 1B lists the contributors linked to testing meth
Working Groups are generally composed by odology (including procedure, equipment and oper
diverse experts with different areas of competence ator dependency) and Table 1C presents the
and with specific experience on the local practise in contributors intimately associated the effects, caused
their countries. Generally, the most important chal by the interaction between cone penetrometer and
lenge is reaching agreement and consensus within soil, during the penetration phase.
the working group to ensure that the draft standard From the varied list of uncertainty contributors, it
document is supported by the majority of countries is clear, that the measuring system uncertainties are
during the balloting stages. Reaching consensus easier to measure with robust laboratory testing pro
often requires long discussions, preparation and revi cedures. Soage Santos et al. (2022) discuss in detail
sion of formulations, datasets and evidence to dem the quantification of calibration uncertainty.
onstrate the scientific or operational rigor of a The uncertainties associated with testing method
position or opinion. ology are described in detail in recent publications
Whenever possible, requirements should be (Peuchen and Terwindt, 2014, Lunne et al, 2017,
expressed in terms of performance criteria, leaving Kardan et al. 2016). Testing methodology related
maximum flexibility to technical development and uncertainties can be minimised or fully removed by
innovation. implementing best practise guidance and prescriptive
An important consideration for the Working procedures.
Group is to use clear and concise language, well Uncertainties associated with the effects derived
define terminology, clear unambiguous requirements from cone penetrometer-soil interaction are more
to reduce the possibility of multiple interpretations. difficult to measure and control. The most important
In this regard the provision of informative material contribution to uncertainty during testing is tempera
in the form of examples, explanations and figures is ture variations. The importance of temperature
very valuable to improve understanding and ease changes is widely discussed in literature (Boylan,
implementation. 2008, McCallum, 2010, Peuchen, 2014). Transient
Working Groups must consider the needs and the temperature gradients resulting from frictional heat
challenges from the industry and reach stakeholders during cone penetration and from thermal flux
to ensure key objectives, needs and challenges are through push rods causes heating of a cone pene
served. During preparation of the standard the Work trometer. Temperature-induced drift under in situ
ing Group must take into consideration the cost of stress and axial loading is without a doubt the larger
implementation of the standard in the industry error contributor on CPTU results. Figure 1 shows
against the value added and ensure, that the new an example of the magnitude of the drift and the
requirements are practical and cost effective. Finally, behavior of the measurement system when a cone is
58
Table 1A. Uncertainties associated with the CPT/CPTU measurement system.
Uncertainty Potential impact to

contributors uncertainty of CPT/CPTU results Comments Potential mitigation
Profiles with Profiles with Profiles with

qc<1MPa 1<qc<3MPa qc>3MPa
Reference High Medium to Medium to Refers to the measurement of uncertainty or the reference used for the calibration Introduction of a robust standard
high high of a sensor. The reference uncertainty should be several times lower than the instru- procedure for calibration of cone
ment that is under calibration. penetrometers. Definition of
Reproducibility High Medium to Medium to Refers to the extent, to which consistent results are obtained when an experiment is standard methodology for type A
high high repeated under a set of different conditions evaluation of uncertainty (in
accordance with ISO/IEC Guide
Repeatability High Medium to Medium to Refers to the closeness of successive measurements when carried out under as set 98-3:2008 and JCGM 100:2008
high high repeatability conditions recommendations) specific for the
Resolution High Medium to Medium to Refers to the ability of a measurement system to detect small changes. The resolution calibration of cone penetrometers
high high indicates the smallest change in the measurand that a measurement system can detect. under laboratory conditions
Zero drift High Medium to Medium to Refers to the difference in the measurement at zero load or pressure before and after
high high applying a load or pressure increment
Interpolation High Medium to Medium to Refers to the difference of the values given by the calibration interpolation equation
high high and the actual measurements taken during the calibration
Reversibility High Medium to Medium to Refers to the difference between two separate measurements taken at the same point,
high high the first is taken during a series of increasing measurement values, and the other
during a series of decreasing measurement values
Apparent load High Medium to Medium to Refers to the undesired influence experienced by a sensor caused by another sensor.
transfer high high For instance, when in compression type cone penetrometers sleeve friction strain
gauge perceives changes, when the cone resistance sensor is loaded or vice versa.
Normalised net High Medium to Medium to Applying a correct net area ratio for cone resistance corrections is of importance spe
area ratio high high cially in clays. Best practise recommendations suggest the determination of net area
ration in a in a pressure cell during calibration. It should be noted that the net area
ratio varies with applied pressure and there is some variance along the measurement
range.
Dimensions High Medium to Medium CPT/CPTU measurements are highly dependent on geometry and deviation from the
and geometry high standard recommended geometries will compromise the representativeness of the
results. In addition, cone resistance and sleeve friction are reported in units of pres
sure thus the uncertainty associated to the determinations of cone and sleeve friction
area will have an important impact on the CPTU reported measurements
Table 1B. Uncertainties associated with testing methodology.
Uncertainty Potential impact to uncertainty of CPT/CPTU

contributors results Comments Potential mitigation
Profiles
Profiles with with Profiles with
Temperature High Medium to Medium Changes in temperature will affect the results of the test due to sensors Introduction of robust methodology, equipment
high do not readily compensate for transient temperature changes. Prior to and reporting requirements to prevent or minim-
testing, the penetrometer should be brought to a stable temperature as ise the error caused by these potential
close as possible to the expected ground temperature. This is particu- contributors.
larly important when the penetrometer temperature is significantly
higher or lower than the ground temperature.
Zeroing High Medium to Medium Loads caused by trapped soil, or any other loads influencing the cone
high (handling, embedment in soil) can result in offsets being wrongly
introduced while zeroing the measurement system and adding
unnecessary uncertainty to CPTU datasets. In addition, unrepresenta
tive zero reference values are likely to penalise the cone performance
assessment after testing by comparing reference readings before and
after the test
Saturation High Medium to Medium Gas or air trapped in the pressure chamber will cause delays in the
high sensor response (sluggish response) resulting in unrepresentative data.
The pore pressure data will be invalid. This has a direct impact on the
representativeness of the cone resistance, as cannot be corrected for
pore pressure effects. Careful selection of filter elements, saturation
fluid and saturation method are required to minimize the adverse
effects on uncertainty.
Rate of High Medium to Medium CPT/CPTU measurements are highly dependent on the penetration
penetration high rate and deviations from the standardised rates will compromise the
representativeness of the results.
Placement of Medium-high Medium Low-medium The weight of the CPT rig can cause changes in the natural stresses in
CPT rig at the ground. Placement of the weight need to be considered to avoid
ground surface influencing the soil to be tested.
Stops in Medium Low to Low May cause adherence of soil to cone penetrometer body and thus false
penetration medium increase in cone resistance and sleeve friction. The acceleration and
deceleration required for starting and stopping a test will inevitably
cause local change in cone sensor readings
Table 1C. Uncertainties associated with effects caused by cone penetrometer-soil interaction.
Uncertainty Potential impact to uncertainty of CPT/
contributors CPTU results uncertainty Comments Potential mitigation
Profiles Profiles
Profiles with with with
Temperature High Medium to Medium Frictional heat developed during penetration can be significant and the Introduction of additional sensors to monitor tempera
high consequence is large drifts will occur during testing. This is perhaps ture and bending changes and applied sound correction
the most important source of uncertainty methods based on laboratory studies. Introduction of
Loss on High Medium to Medium It is sometimes difficult to maintain filter saturation during testing, par robust methodology, equipment and reporting require
saturation high ticularly when pushing through a desiccated crust or through dilatant ments to prevent or minimise the error caused by these
soils. Loss of saturation leads to sluggish pore pressure response and potential sources
reliable fluid pressure data may not be consistently obtained. In
instances testing approach need to be adjusted, for example by apply
ing predrilling though a partially saturated soil to avoid losing satur
ation before reaching the saturated zone.
Dimensions Medium to high Medium Low to As discussed, early variations dimensions and geometry of a cone can
and medium significantly contribute to uncertainty. Even when the dimensions and
geometry geometry are well known at the commencement of a test, the high fric
tional forces, and the abrasive nature of some soil types of my cause
small variations on cone penetrometer dimensions and therefore affect
the results. This is most likely to occur in sandy profiles and will have
a significant effect on soft and loose sediments underlaying the sand
layer.
Bending Medium Low to Low During penetration the cone can be subjected to bending moments.
medium This influence is generally small for qc and fs (error <10kPa and <1kPa
respectively) for cone penetrometers using double bridge strain gauge.
Inclination Medium Low to Low Control of inclination is of particular importance contributor to depth
medium uncertainty. In addition, larger inclination will affect the cone resist
ance and sleeve friction readings. Cone resistance can be affected sig
nificantly (10%) in anisotropic soils when inclinations exceed 30
degrees
Soil trapped Low to medium Low Low Trapped soil and dirt on gaps and seals can cause undesirable resist
or adhered ance that will impact the measurements specially when testing soft
clays. Soils can also stick to the cone penetrometer surfaces effectively
increasing the projected area and providing overestimates of qc and fs.
submerged into a water bath at different temperat 3. Technological enhancement: Incorporating new
ures in laboratory conditions. technology that facilitates deeper understanding
of the uncertainties rising from soil-equipment
interaction.
6.1 Application of uncertainty analysis to cone

penetrometer calibration results
The uncertainty contributions intrinsically associated
to the CPT/CPTU measurement system can be better
understood and quantified by conducting a series of
well design and robust procedures for the calibration
of cone penetrometers. Uncertainty analysis “Type
A” is applied to the calibration results, allowing sys
tematic quantification of the cone penetrometer abil
ity to produce repeatable results under laboratory
conditions.
Requirements for calibration of cone penetrom
eters and verification of cone penetrometers and
uncertainty calculations are incorporated in (draft)
Figure 1. Example of verification data for temperature standards published by ISO, particularly ISO/DIS
influence.
22476-1:2021 and ISO/DIS 19901-8:2022. The use
of uncertainty analysis on calibration results implies
Additional uncertainty is introduced when infer the departure from the existing application class type
ring geotechnical parameters from CPT/CPTU data. approach and the inevitable implementation chal
These additional uncertainties associated with data lenges discussed earlier. Instead, a series of “cone
interpretation is out of the scope of this document. penetrometer classes” can be established using limit
Powell and Dhimitri, 2022 provide detailed discus ing values that defines the minimum uncertainty
sion and advice for avoiding introduction of requirements to satisfy a given cone class. This pro
undesired uncertainty when deriving geotechnical vides a cone penetrometer classification framework
parameters from CPT/CPTU results. that allows transparent comparison between systems,
regardless of manufacturer or supplier. And facili
tates the selection of cone penetrometers to match
6 RECOMMENDED WAY FORWARD the project and design sensitivity requirements.
Calibration uncertainty analysis and performance
The implementation of a rigid performance-based indicators based on verification tests defines the best
approach for quantifying CPT/CPTU uncertainties cone class that a particular system can obtain under
using modelling techniques (“Type B” uncertainty laboratory conditions. Soage Santos et al. 2022
analysis) is complex and not fully developed. Model- describes this scheme in some detail and provides
ling uncertainty requires maturation and large amount example of the results.
of research effort. Peuchen and Terwindt 2014 work
has shown good potential for the application of this 6.2 Improving and controlling quality by
type of analysis, but currently it is not considered a implementation of testing methodology
feasible approach in practise as discussed in section requirements
3. In this section a recommended methodology for
standardising quality control is proposed. Most of the uncertainties associated to testing meth
There is ample opportunity to improve the overall odology (listed in Table 1B) can be greatly reduced
quality of the CPT/CPTU results by implementing by prescribing detailed methodologies and equip
standard requirements with the objective of minimis ment requirements. For instance, zeroing the cone
ing the sources of uncertainties presented in Table 1. penetrometer acquisition system is critical. If zeroing
Different approaches can be followed to target the is performed while the cone penetrometer is under
three main uncertainty groups: going temperature changes or when the cone is
embedded or covered in soil, the cone will experi
1. Performance-based approach: Uncertainty ana ence significant drifts affecting the representative
lysis (Type A) to quantify uncertainty on measur ness of the data. Thus, clear requirements regarding
ing systems. temperature stability and presence of undesired loads
2. Method-based approach. Provide unequivocal while zeroing the measurement system are
procedures and equipment specifications to necessary.
reduce uncertainty caused by operator and equip Significant uncertainty, up to 4% on cone resist
ment bias. ance, can rise from using the default areas (cone
62
design areas) in the conversion of force to pressure users and interpreters of the CPT/CPTU datasets.
measurements rather than the actual measured areas. The introduction of rigid reporting requirements for
This can be readily prevented by ensuring that cone capturing qualitative performance assessment can be
geometry is measured in a calibration laboratory and a suitable approach to deal with this challenge. Some
monitored in the field. Powell and Lunne 2005 and examples of qualitative performance observations
Powell et al. 2021 showed the impact of cone dimen are listed below:
sions and geometry on CPTU results.
• Information related to the responsiveness of
While CPT/CPTU measurement systems are gen
sensor to soil changes
erally reliable and robust, experience shows that
• Pore pressure sluggish responses
every single component of the CPT/CPTU measure
• Sleeve friction or cone resistance insensitivity
ment system may deteriorate, suffer damage, or
caused by dirt trapped in cone seals and gaps.
experience drifts during operations (transport,
• Pore water pressure cavitation effects experienced
system set up or testing). Consequently, it is impera
during testing
tive to monitor the integrity of the system during
• Qualitative indication of the soil disturbance
field operations. A diagnostic check scheme can be
caused by the seabed frame or drilling activities
introduced to classify the CPT/CPTU test results.
Reference readings and output stability are non- Draft ISO/DIS 22476-1:2021 and ISO/DIS
subjective diagnostic checks that can be measured 19901-8:2021capture quality monitoring by means
and quantified. For instance, observed differences of test categories, as discussed above, and provide
of reference readings taken before and after a test highly detailed normative procedures ranging from
will indicate if the measuring system has experi field operations preparations to test termination and
enced a drift during the penetration or retraction reporting activities.
of the cone from the ground. The absolute differ
ence of values observed can be used for classify
ing the CPTU results in accordance with
6.3 Improving and controlling uncertainty by
prescribed limits to assign a “performance cat
encouraging technological development
egory” to the test.
CPT/CPTU “performance categories” does not The uncertainty contributors resulting from the
provide quantification the uncertainty of a particular effects of the interaction between cone penetrometer
test but allows identification of results that may have and soil (see Table 1C) are especially challenging to
been affected by equipment malfunction or proced quantify, control or correct. The introduction of add
ural error. Certain CPT/CPTU applications, like the itional technical requirements targeted to reduce and
characterisation of soft clays for foundation design, understand the uncertainty caused by the develop
may require highly representative and repeatable ment of temperature variations and bending moment
results and consequently better performance is during testing will be greatly beneficial to improve
required. The CPT/CPTU category scheme can the quality of CPT/CPTU results, especially in
accommodate this by defining allowable limits for a unfavourable soil profiles (i.e., dense sands, stiff to
given soil profile or for a range of soil conditions. In hard clays and interbedded soils profiles). Tempera
order to avoid potential conflicts caused by interpret ture changes of 10 degrees are not uncommon in
ation, the soil categories can be linked to prescribed unfavorable soils (i.e., Dense sands) and can cause
ranges of cone resistance, for example profiles show drifts on qc in the order of 300kPa, which is a very
ing qc<1MPa can be use instead of using terms like large error when trying to characterize a clay unit
“Very soft to firm clays”. underlaying a dense sand unit.
It is important to highlight those changes such as Laboratory verification tests, designed to measure
ambient temperature, ambient pressure, the presence how temperature or bending moment variations
of dirt in the cone seals and adhesion of clayey soils affects the performance of qc, fs and u2 sensors can
can cause reference readings differences, that are not be introduced. Performance indicator limiting values
directly related to the changes that the measurement (expressed in terms of kPa/°C and kPa/N) can be
system may have experienced in the ground. For a included as part of the cone classification scheme
performance category scheme to work as intended, it described in section 6.2 such that cone penetrometer
is fundamental, that reference readings are taken classes are defined by the combination of the uncer
under similar conditions every time so that a mean tainty associated to the calibration and the test verifi
ingful comparison can be made. This challenge can cation results.
be addressed by the incorporation of very clear test Further, the inclusion of temperature sensors in
ing methodology requirements. cone penetrometers can additionally provide valu
Not all the uncertainty contributors associated to able information on sensor performance during field
test methodology can be easily quantified and testing. As the effect of temperature changes on a
included in a CPT/CPTU performance classification particular cone design can be measured under
scheme due to the impossibility to quantify them. laboratory conditions it is relatively straight forward
However, qualitative assessment of test performance to develop correction algorithms based on the vari
is very valuable and relevant information to the ations of temperature recorded by the temperature
63
sensor. The application of this technology can reporting activities will also improve repeatability
greatly reduce the uncertainty of CPT/CPTU results and representativeness of the results.
in unfavourable conditions where temperature Uncertainties intimately associated the effects
changes dominate the uncertainty of the results. caused by the interaction between cone penetrometer
Draft ISO/DIS 22476-1:2022 and ISO/DIS and soil are often unavoidable. The introduction of
19901-8:2022 introduces a cone penetrometer class new technology that allows measurement of these
that requires the inclusion of a temperature sensor. uncertainties and enables the development of suit
This clearly promotes technological development able corrections improving the quality and therefore
targeting the reduction of the uncertainties caused the overall certainty of the CPT/CPTU measure
unavoidable temperature changes during testing in ments even in unfavourable conditions.
unfavourable conditions. It should also be noted that
the required presentation of results of the ISO verifi
cations covers the influences of temperature and ACKNOWLEDGEMENTS
bending on the parameters qc , fs , and u.
The author would like to thank his colleagues at ISO
19901-8 work group, Tom Lunne, Joek Peuchen,
7 SUMMARY AND CONCLUSIONS Tim Carrington for sharing their knowledge and for
the challenging and fruitful technical discussions
CPT/CPTU provide fundamental input for safe and that greatly influenced the opinions expressed in this
economical design of many structures. Assessing the document. Special thanks go to Gülin Yetginer for
representativeness of the CPT/CPTU results is essen her amazing work keeping the group working as a
tial. Consequently, it is important to understand, and team with a focus on reaching consensus and
when possible, minimize the uncertainty associated delivering a much-improved draft standard docu
to CPT/CPTU measurements, so these can be con ment. Further, the author would like to acknowledge
sidered in planning, design, and installation of the valuable support and input received from all the
infrastructure. members of the ISO 22476-1 workgroup, special
The practical challenges associated to the quantifi thanks go to Håkan Garin, Kees-Jan van der Made,
cation of uncertainty and the difficulties of its stand Marius Tremblay, Kristoffer Kåsin, Diederick Bouw
ardization have been discussed in depth in this meester and Wouter Vervaele. Last but not least
document. thanks to John Powell and Tom Lunne for providing
Uncertainty contributors associated to CPT/CPTU moral and technical support and for sharing their
measurements have been categorized into three main vast experience and knowledge.
groups: (1) Uncertainties associated with the measur
ing system, (2) Uncertainties resulting from testing
methodology (including procedure, equipment and REFERENCES
operator dependency) and (3) Uncertainties intim
ately associated with the effects caused by the inter ASTM International, 2012. ASTM D5778-12 Standard Test-
Method for Electronic Friction Cone and Piezocone
action between cone penetrometer and soil. This
Penetration Testing of Soils. West Conshohocken:
classification of uncertainty contributors allows pro ASTM International.
posing a standardization approach that addresses ASTM International, 2020. ASTM D5778-20 Standard Test-
each group with a combination of performance- Method for Electronic Friction Cone and Piezocone
based and method-based requirements, aiming to Penetration Testing of Soils. West Conshohocken:
control and minimize the uncertainty contributions ASTM International.
rising from execution of CPT/CPTU. Boylan, N., Mathijssen, F., Long, M. and Molenkamp, F.
Uncertainties associated with the measurement 2008. Cone penetration testing of organic soils. Proceed
system can be reduced by the introduction of a ings, Baltic Geotechnical Conference, September.
detailed calibration and verification procedures and Gdansk, Poland.
International Organization for Standardization, 2008. ISO/
prescribed methodology for uncertainty quantifica IEC Guide 98-3:2008 Uncertainty of Measurement –
tion. This also enables the development of a per Part 3: Guide to the Expression of Uncertainty in Meas
formance-based cone penetrometer class scheme. urement (GU:1995). Geneva: ISO.
The approach allows for easy comparison of cone International Organization for Standardization. 2012. Geo-
penetrometers, regardless of manufacturer or sup technical investigation and testing – Field testing – Part
plier, and assists the appropriate selection of cone 1: Electrical cone and piezocone penetration tests,
penetrometer for specific applications. International Standard ISO 22476-1:2012. (With Tech
Uncertainties resulting from test methodology can nical Corrigendum 1, January 2013). Geneva: Inter
be addressed by implementing a quality monitoring national Organization for Standardization.
International Organization for Standardization, 2014. ISO/
scheme defining test categories based on quantitative DIS 19901-8:2021 Petroleum and Natural Gas Indus
performance indicators such as reference readings tries - Specific Requirements for Offshore Structures –
differences and output stabilities. In addition, Part 8: Marine Soil Investigations. Geneva: ISO.
detailed normative procedures ranging from field International Organization for Standardization, 2021. ISO/
operations preparations to test termination and DIS 22476-1:2021 Geotechnical Investigation and
64
Testing – Field Testing – Part 1: Electrical Cone and International Symposium on Cone Penetration Testing
Piezocone Penetration Test. Geneva: ISO. CPT10: 2010 – Huntington Beach, California.
International Organization for Standardization, 2022. ISO/ Peuchen, J., Santos, R., Yetginer, A.G., Eckart, W.S., Car
DIS 19901-8:2022 Petroleum and Natural Gas Indus rington, T.M. & Lunne, T. 2020. CPT data showing
tries - Specific Requirements for Offshore Structures – anomalies – assessment and potential postprocessing. In
Part 8: Marine Soil Investigations. Geneva: ISO. Z. Westgate (ed.). 4th International Symposium on
ISSMFE International Society for Soil Mechanics and Frontiers in Offshore Geotechnics (ISFOG 2020): pro
Foundation Engineering. 1977. Report of the subcom ceedings (pp. 1026–1035). Deep Foundations Institute.
mittee on standardization of penetration testing in Peuchen, J. & Terwindt, J., 2014. Introduction to CPT
Europe, Appendix A, recommended standard for the accuracy. In 3rd International symposium on cone pene
cone penetration test (CPT). Proceedings of the ninth tration testing CPT14: May 12-14 (2014). Las Vegas,
international conference on soil mechanics and founda Nevada.
tion engineering, Tokyo. Peuchen J and Terwindt J. (2015). Measurement uncer
JCGM. 2008 Joint Committee for Guidelines in Metrology. tainty of offshore Cone Penetration Tests. Proc. ISFOG
Evaluation of measurement data_ Guide to expression 15 Frontiers in Offshore Geotechnics III – Meyer (Ed.)
of uncertainty in measurements JCGM 100: 2008. © 2015 Taylor & Francis Group,
JCGM. 2012 Joint Committee for Guidelines in Metrology. Powell JJM and Lunne T. (2005). A comparison of different
International vocabulary of metrology – Basic and gen sized piezocones in UK clays. Proc. XVIth ICSMGE,
eral concepts and associated terms (VIM). JCGM 200: Osaka, September 2005. Vol. 2, pp729–734.
2012. Powell JJM and Dhimitri L. and Ward, D. (2021). The fric
Kardan C., Viking k., Nik L., Larsson S., 2016. Influence of tion sleeve measurement in CPTU-Does size matters? –
operator performance on quality of CPTu results. 17th A new study. 6th International Conference on geotech
Nordic Geotechnical Meeting. Challenges in Nordic nical and geophysical site characterization. Budapest.
Geotechnics. Reykjavik. Powell JJM and Dhimitri L. (2022). Watch out for the use
Lunne, T., Santos, R. Brink Clausen, J. & Powell, J.J.M. of global correlations and “blackbox” interpretation of
2017. Guidelines for Use of CPTU Application Classes CPTU data. 5th International Symposium on Cone Pene
According to ISO 19901-8: (2014). In: Society For tration Testing CPT’22. Bologna.
Underwater Technology Offshore Site Investigation and Soage Santos R., Gomez Meyer E, Peuchen J., Yetginer
Geotechnics Committee, ed. Offshore Site Investigation G., Lunne T., Carrington T., 2022. Calibration of
and Geotechnics: Smarter Solutions for Future Offshore cone penetrometers according to International Organ
Developments: Proceedings of the 8th International ization for Standardization requirements. 5th Inter
Conference 12-14 September 2017, Royal Geographical national Symposium on Cone Penetration Testing
Society, London, UK: Volume 2. London: Society for CPT’22. Bologna.
Underwater Technology, pp. 300–307. Swedish Geotechnical Society (1993) “Recommended
McCallum, A. Barwise, A. Soage Santos, R. (2010). CPT Standard for Cone Penetration testing”. SGF Report
in polar snow-equipment and procedures, 2rd 1:93, June 15 1992.
65
© 2022 the Author(s), ISBN 978-1-032-31259-0
CPT equipment: Recent advances and future perspectives

D.J. White
University of Southampton, UK
ABSTRACT: The basic configuration of the cone penetrometer test (CPT) has remained constant for half
a century since the electric piezocone became widespread in the 1970s and the dimensions and test protocols
were standardized soon after. The results of a conventional CPT performed today might be indistinguishable
from a test conducted at the same location in 1972, aside from improved data logging and presentation. This
discussion reflects on advances in CPT and other penetrometer equipment presented at this conference and in
the recent literature, and presents discussion points to be considered during the event. Two general observa
tions emerge: (i) advances in instrumentation, sensing and mechanical engineering have created a vast range of
variations on the conventional CPT, providing excitement for technologists, researchers and a small subset of
users; (ii) the conventional CPT is heavily standardized, commoditized and automated, and is progressively
replacing the SPT in onshore practice. The result is a dispersion of practice: the majority of civil engineers
base their ground characterization on a durable 1970s technology, updated with automation and improved data
quality, while a minority are enjoying exploration of non-standard innovations – with mixed levels of impact.
This conference provides a chance for us to (i) reflect where our activities sit in this spectrum, (ii) enjoy learn
ing about recent advances, and (iii) debate how we will characterize the ground in another 50 years.
1 INTRODUCTION but useful improvements to traditional CPT equip

ment, which are suited to widespread adoption.
1.1 Purpose and scope of this discussion paper
I was invited by the organisers of this conference to 1.2 Elements of penetrometer equipment
prepare a presentation for the start of the discussion
To review advances in CPT equipment it is useful to
session on the subject of CPT and other penetrometer
define a baseline starting point, which is taken by
equipment. This written contribution is to provide
defining a ‘conventional’ CPT. Also, it is useful to
a record of the presentation in advance, and the presen
divide CPT equipment into the various functional
tation is intended to set the scene for a productive
elements of the test and examine the advances in each.
discussion.
Table 1 lists the adopted set of functional elements.
The paper draws on the manuscripts presented at
this event related to penetrometer equipment, as well
as recent literature on the subject, and I have also Table 1. Functional elements of CPT equipment.
tried to draw in wider context from the civil engineer
ing industry, and other relevant industries, to give con Element Description
text to the drivers of CPT equipment developments.
The discussion that follows sets current develop Delivery How the CPT equipment reaches the site
location (at the soil surface)
ments in the context of the history of the CPT. This
Deployment How the CPT advances into the subsoil, to
short contribution cannot comprehensively cover the the test location
full history and range of this technology, which is Sensing The instrument types and measurements
properly described by other keynote and review taken during the test
papers such as Broms & Flodin (1988), Lunne et al. Motion The applied movement during the test
(1997), Lunne (2012) and Massarch (2014). Scale The size of the instrument
Many of the advances described in the papers of
CPT’22 are non-mainstream – relating to specialized
modifications or applications of the CPT – so this For the baseline case of the conventional CPT,
review is a journey mainly through the exotic. How these functional elements are shown in Figure 1.
ever, there are contributions that provide incremental They are essentially unchanged since the 1970s,
DOI: 10.1201/9781003308829-5
66
when standard dimensions and penetration rates for sowing, weeding and harvesting, and these carry
emerged from early standardization work. soil condition sensors. It is a simple extension to add
a penetrometer system to these vehicles. Interplanet
ary landers have also conducted penetrometry.
For delivery to offshore underwater sites, the
remote-controlled CPT vehicle is not as esoteric as
onshore. Trials of ROV-mounted CPT systems were
reported in the early 1980s by Geise & Kolk (1983)
and the GeoROV CPT system (Edmunds 2014) has
found widespread use as a lightweight seabed pene
trometer system over the past decade (Figure 3d).
During an ROV-based CPT test, the crew is above
the apparatus, on the vessel that the unit is tethered
to. A logical extension, aligned with trends in other
offshore survey technologies, is for the crew to
remain on land, and the vessel to be uncrewed.
Autonomous (or uncrewed) surface vessels are now
used for geophysical surveys, and an uncrewed geo
technical survey is a realistic prospect.
Compared to sampling operations, penetrometer
testing is easier to perform in an uncrewed situation
due to the complexity of sample handling and stor
age. Penetrometer testing will therefore be a central
element of future uncrewed geotechnical surveys,
Figure 1. Conventional CPT systems – onshore and both offshore and onshore. This likely increase in
offshore. reliance on penetrometry is a driver of more
advanced test types, as discussed in Section 2.5, to
replace the reduction in data if sampling is elimin
2 CPT EQUIPMENT ADVANCES ated or reduced.
2.1 Introduction
2.3 Deployment
The following sub-sections discuss advances related
to each functional element. The advances that feature 2.3.1 Introduction
in this review are summarized in Figure 2, which can For deployment of the cone penetrometer into the sub
be compared to the baseline of Figure 1. However, soil, a conventional onshore CPT rig usually has
Figure 2 is not an exhaustive record of CPT equip a hydraulic pushing system, with rods of 2.5-
ment innovations, and many more can be found in 3.0 m length, that are connected after each push by the
the earlier CPT and ISOPT/ESOPT conferences. operator. Some offshore systems use coiled rods (e.g.
Power & Geise 1994, Meunier et al. 2004), or
2.2 Delivery a robotic drilling system (e.g. Kelleher & Randolph
2005). The vertical reaction force is provided by the
For delivery to onshore sites, early CPT rigs were vehicle weight, sometimes supplemented by ground
mounted on rugged vehicles and today’s fleets of com anchors, or by the skirts of a suction-type foundation
mercial CPT rigs includes large and small wheeled or on the base of the unit (e.g. Boggess & Robertson
tracked vehicles, driven by people sat in a driving seat. 2010).
Recent onshore delivery advances have eliminated Advances in deployment of the penetrometer fall
the wheels and tracks and in some cases the driver, to into two primary categories: (i) techniques to speed
allow CPT testing in situations where these may not up the deployment process and (ii) techniques to
be used. This may, for example, be due to (i) reduce the vertical required reaction force and there
extremely soft ground (e.g. tailings or marsh), (ii) hos fore increase the potential depth of penetration.
tile locals (e.g. for military trafficability assessments)
or (iii) the lack of oxygen (e.g. on Mars) (Figure 3). 2.3.2 Speeding up deployment
Uncrewed and remote-controlled onshore CPT In this conference, Storteboom et al. (2022a)
delivery systems that have been trialled include aerial describe a new trademarked ‘SingleTwist’ rod con
drones (Baez et al. 2020) and uncrewed wheeled or nection system, which allows a long CPT string to
tracked vehicles (e.g. Scholz et al. 2014, Olmedo & be compactly stored and quickly assembled by
Lipsett 2020). These technologies allow soil consist robotics (Figure 4a). The rods are held together
ency assessments in wetlands, agricultural land and with a flexible ball joint, for storage on a reel, but
tailings ponds (Figure 3a-c). Robotic vehicles with require only 1/6th of a turn to become rigidly con
advanced manipulators are widely used in agriculture nected for deployment. The ball joint connection
67
Figure 2. CPT equipment advances: selected from this conference and recent literature – as discussed in Section 2.
features a protected data cable running through the

joint.
An alternative speed-up technique, which also
removes the need for any vertical reaction force, is to
use a free-falling penetrometer (FFP). These systems
are gaining wide use in research for shallow seafloor
characterization (e.g. Stark et al. 2009, Steiner et al.
2014, Stark 2016) and their interpretation methods are
evolving in parallel with free falling anchor research
(Chow et al. 2017). Commercial FFP systems are also
used for deepwater soft soils, sometimes with com
bined free-falling and static push test stages (e.g.
Young et al. 2011).
In this conference, Kang et al. (2022) present sea
trials in a relatively lightweight FFP, weighing 38 kg
(Figure 4c). Studies such as this demonstrate how
a self-contained low cost FFP data system (e.g. Ran
dolph et al. 2018) allows multiple rapid CPTs to be
performed.
Figure 3. Advances in delivery of CPT equipment:
However, these studies, and others (e.g. Stephan
uncrewed vehicles, (a) aerial (Baez et al. 2020), (b) land et al. 2015, Peuchen et al. 2017) show the challenge of
(Olmedo & Lipsett 2016), (c) land/water (HELIX25: www. achieving sufficient penetration in sands to reach the
copperstonetech.com) and (d) underwater (ROVcone www. depth required for typical designs – such as assessing
bluefieldgeo.com). trenchability for cable route surveys. FFP penetration
68
approximate doubling of the depth achieved, includ
ing penetration through sands with qc > 100 MPa.
2.3.4 Bypassing reaction force: Self-burrowing

An alternative to eliminating rod friction is to elimin
ate the rods themselves, and transform the CPT unit
into a burrowing robot. Setting aside questions of
navigation and power, this concept relies on the
sleeve friction to provide a reaction force to overcome
the cone tip resistance. Natural burrowers such as
razor clams, earthworms and roots have been used as
bio-inspiration for robots that could evolve into a rod-
free CPT (e.g. Winter et al. 2014, Tao et al. 2020,
Naclerio et al. 2021). Martinez et al. (2020) show that
a roughened sleeve with a length-diameter ratio of L/
D = 4-6.5 is required in dense sands – the most diffi
cult conditions – to provide the required reaction.
An additional strategy to ease burrowing is to
reduce the tip resistance by fluidizing the soil ahead
of the robot, or in dry sand by aerating it. The
approach has been successfully demonstrated in the
Figure 4. Advances in deployment of CPT equipment: (a) laboratory by a steerable robot using an externally-
SingleTwistTM connectors (Storteboom et al. 2022a), (b) provided air supply to fluidise the soil ahead
water lubrication (Yetginer-Tjelta et al. 2022), (c) free fall (Naclerio et al. 2021).
ing system (Kang et al. 2022) and (d) self-driving system The rise of robotics has led to many proposals for
(HP3 Spohn et al. 2018). burrowing robots to be used across civil engineering
applications. Perhaps self-burrowing CPTs will fea
ture in future versions of this conference.
of up to 15-20 m is reported in soft clays by the ‘CPT
stinger’ (Jeanjean et al. 2012). However, the reported 2.3.5 Bypassing reaction force: Self-driving
trials in sandy or stiff clay soils have achieved only 1- An alternative to self-burrowing is self-driving, in
3 m penetration in some situations, even with a heavy which the CPT is advanced by a driving mechanism
250-500 kg FFP unit. The challenge of achieving suffi that is incorporated within the machine itself. This
cient penetration may be exacerbated in dilatant sands concept was adopted by the Heat Flow and Physical
by the undrained response. This effect also complicates Properties Package (HP3) of the Insight Mission that
the data interpretation, due to the change in drainage took a lander to Mars during 2018.
conditions relative to a standard CPT as well as the vis The HP3 system featured a CPT-like ‘Mole’
cous rate effect (White et al. 2018). The depth range device, 400 mm long, 27 mm in diameter, with
can be extended by a static push unit in the FFP (e.g. a mass of 860 g. The Mole was designed to pull
Jeanjean et al. 2012, Peuchen et al. 2017). This a 5m-long Science Tether into the subsoil to a depth
approach requires the embedment from the initial free of 3 – 5m (Spohn et al. 2018, Kromer et al 2019)
fall stage to provide sufficient reaction. (Figure 4d).
The driving mechanism was designed to achieve
2.3.3 Reducing reaction force, raising depth range 0.1-1mm of set per blow, at a rate of 0.28 blows/sec.
Twin advances of a lighter CPT system and an Laboratory trials show this probe self-driving to
increased depth range can be achieved through 5 m depth in a sandy Mars-simulant soil of crushed
methods that reduce or eliminate the external force basalt sand. Comparisons with conventional CPTs in
required to push the rods through the ground. the same bed showed that penetration at around
A simple modification to reduce the pushing force 20 mm/min was possible in soil with qc ~ 1 MPa
that can form an addition to a conventional CPT (Spohn et al. 2018, Wippermann et al. 2020).
system is a lubrication system to inject water or drill However, attempts to penetrate the surface of
ing mud around the rods (e.g. Jefferies & Funegard Mars were unsuccessful, despite the Mole operating
1983. correctly. The probe failed to overcome the initial
In this conference, Yetginer-Tjelta et al. (2022) penetration resistance of the shallowest few centi
report a systematic investigation into different metres and created a wide hole around the probe.
methods of water injection at the tip and rods of This prevented the shaft resistance on the Mole from
a seafloor CPT system (Figure 4b). This was part of providing resistance against the hammer recoil, con
a long term campaign on the Dogger Bank in the tributing to the failure to penetrate. Despite various
North Sea, with more than 500 CPTs. These attempts over a period of two years, the Mole
improvement to the equipment have led to an remained stuck in a shallow open hole on the surface
69
of Mars (Spohn et al. 2021). Despite these difficul sources: (i) errors in measuring the applied shear
ties on Mars, the project demonstrates the feasibility stress due to load leakage through the seals and/or
of self-burial by a CPT-like device using an internal low load cell sensitivity in subtraction cones and (ii)
hammer system. variations in the applied shear stress due to the
sleeve roughness or minor differences in diameter or
2.4 Sensing alignment of the sleeve. A review of these difficul
ties is presented in this conference by McConnell &
2.4.1 Introduction Wassenaar (2022), who describe a new cone with
The sensing on a conventional cone comprises pri improved load cell design to overcome these limita
marily the three CPT parameters, tip and sleeve tions and provide more reliable friction sleeve data
resistance, and pore pressure (i.e. qc, fs and u2) with in soft soils.
additional measurements of depth and tilt. The seis
mic piezocone test also provides measurements of 2.4.3 Advances in hydraulic sensing
shear wave velocity, Vs, at discrete depths. Multiple contributions to this conference focus on
A wide range of additional sensors have been the reliable measurement of hydraulic properties.
incorporated within the cone, some of which are Lunne et al. (2022) report on a revival of multiple
illustrated in Figure 2, based on contributions to this pore pressure measurements – a review of early data
conference and other recent publications. These are that used the u1 (tip and face) and u3 (top of sleeve)
discussed under the following headings of (i) mech sensors, alongside the conventional u2 (shoulder).
anical, (ii) hydraulic and (iii) thermal. They show that normalized parameters using the dif
ference or ratio between u1 and u2 can provide
2.4.2 Advances in mechanical sensing a reasonable correlation with OCR on a site-by-site
It is well recognized that the fundamental strength basis, but the correlation with the conventional Qt
and stiffness properties of a soil, in whatever con remains stronger.
stitutive framework is considered, do not uniquely One of the difficulties with CPT pore pressure
correlate with the three basic cone parameters. This data is the potential for unreliability due to poor sat
uncertainty can be reduced by taking one or more uration of the sensor filter element, air trapped else
further mechanical readings, such as normal or where in the tip and cavitation. In this conference,
shear stress values, by testing with a modified Rocchi et al (2022) describe a new device that can
instrument equipped with additional sensors. These be used to rapidly assure the saturation of
additional sensors are commonly mounted in fur a piezocone tip (Figure 5b). This device is used to
ther sleeves behind the conventional single friction test the saturation level by applying an impulse of
sleeve. pressure to the sensor, and then checking Skemp
For example, Isaev et al. (2022) describe in this ton’s B-value for the response.
conference a lateral stress cone, which is equipped Saturation is especially important when perform
with three pairs of lateral total stress and pore pres ing CPTs in unsaturated soils, where negative pore
sure sensors in successively larger diameter sleeves pressures are to be measured. In this conference,
(Figure 5a). This type of ‘lateral stress cone’ pro Tarantino et al. (2022) describe the development of
vides data ‘on-the-fly’ (i.e. during steady penetra a Tensiocone – a CPT attachment incorporating
tion) that is analogous to a dilatometer or three a high capacity tensiometer (HCD). They describe
points on a pressuremeter expansion curve. Correl the laboratory validation of a new adapter that sits
ations are explored to link these measured stresses, between the cone tip and friction sleeve, equipped
via their ratios or differences, to the in situ stress with a surface-mounted tensiometer. Field trials did
state and OCR. not show agreement between in situ tensiometers
A similar theme of multiple sleeves was proposed and the tensiocone measurements. This was traced to
by DeJong & Frost (2002), using different values of a cross-sensitivity between CPT axial load and tensi
roughness. In this case the multi-sleeve approach ometer reading, which will be rectified in a new
allows rough and smooth interface strengths to be design of the sensor.
measured, and the CPT sleeve to be tailored to the A further development of hydraulic measurement
interface for which a design friction is required. By using the CPT is to impose a flow of water from the
measuring sleeve friction at two positions, a new tip or shaft of the cone, and measure the associated
normalized friction parameter related to the degrad pressure, in order to determine the local permeabil
ation of friction can be calculated, and used as ity. This adaptation is analogous to the conventional
a fourth non-dimensional CPT parameter, to aug constant head borehole flow test (Hvorslev 1951).
ment Qt, Fr and Bq. This multi-friction parameter, A similar adaptation has been applied to the pres
MFP, allows the conventional range of CPT soil clas suremeter, by withdrawing the device to create a test
sification charts (e.g. Robertson 2009) to be supple ‘well’ and applying controlled flow (Ratnam et al.
mented by an extra chart using a derivative of MFP 2001).
(Hebeler et al. 2018). Gundersen et al. (2022), in this conference,
Friction sleeve data is sometimes treated with describe trials using a ‘flow cone’ equipped to
caution, due to potential unreliability from two release water through a porous sleeve located
70
These advances in the use of CPTs for thermal
measurements reflect new markets related to
ground heat energy systems and buried cables for
offshore wind. Both of these applications relate
to the energy transition and are rapidly growing
markets.
2.5 Motion
2.5.1 Introduction
The motion in a conventional CPT is continuous
penetration, usually at a rate of 2 cm/s, with intermit
tent pauses for rod changes or dissipation phases.
There have been many attempts to expand this
motion through different speeds, patterns and direc
tions of movement, some of which are the subject of
papers in this conference. Changes in the pattern of
vertical movement are made possible by advanced
drive systems, which allow specific patterns of
movement or load to be applied to the top of the
cone rods. Meanwhile, modern miniature robotic
Figure 5. Advances in sensing on CPT equipment: (a) Ten
siometer (Tarantino et al. 2022), (b) saturation verification technology has allowed miniature actuators to be
(Rocchi et al. 2022), (c) the ‘flow cone’ (Gundersen et al. incorporated in sleeve elements located behind the
2022), (d) Tensiometer adaptor (Tarantino et al. 2022) and cone tip. This allows elements to be displaced rela
(e) a parallel thermal probe element (www.datem.co.uk). tive to the main cone rod, with the corresponding
soil response measured. The extensions can be sum
1.23 m above the shoulder of the cone, which is also marized as follows:
equipped with pressure measurement (Figure 5c).
Tests are reported from NGI’s Halden test site, 1. Novel vertical movement: (i) different speeds,
which comprises of silt. In these conditions, the (ii) large amplitude cycles, (iii) small vibratory
sleeve measurement indicates an excess pore pres cycles;
sure remaining from the cone penetration process, 2. Separate elements for other movements: (iv)
and the dissipation response at the porous sleeve is expansion – i.e. the cone-pressuremeter, (v) rota
used to infer the soil permeability. Prior studies tion or (vi) lateral translation
using the same device at a sand site reported by Gun Recent contributions to each of these test vari
dersen et al. (2019), show accurate direct measure ations are briefly discussed below.
ment of permeability through constant head or
falling head tests at the sleeve. 2.5.2 Variations in the vertical movement
Studies of large-amplitude cyclic CPTs – with verti
2.4.4 Advances in thermal sensing cal movement by several cone diameters – were
CPTs are usually equipped with a temperature sensor prompted by similar testing of flow-round pene
within the tip, which can be used to check for tem trometers (such as the T-bar or ball). For these
perature-induced drift of any sensors, and to also devices, the cyclic penetration process is symmet
indicate the ground temperature – which may feature ric, with soil flowing in reverse around the protru
a step change at the depth of the water table. sion. Cyclic T-bar or ball penetrometer tests
More advanced use of the temperature data is provide a measure of soil sensitivity (e.g. Randolph
described by Vardon et al. (2018, 2019). They describe 2004).
the use of passive temperature responses to evaluate For a CPT, cyclic vertical movement may lead
the thermal conductivity of the soil surrounding the to a hole opening beneath the CPT, but if this
CPT. The passive approach relies on the dissipation of does not occur the process can be similarly sym
heat generated by the penetrometer-soil friction during metric. Regardless of the flow at the cone tip, the
the installation process. This is a simplification of the cyclic movement on the friction sleeve also pro
active approach, which was developed earlier for geo vides data that can support estimates of sensitiv
physical research, which relies on an internal heating ity or the degradation of friction on piles (e.g.
element to provide a pulse of heat that is recorded at Hebeler et al. 2005, Diambra et al. 2014, Shon
other locations along the device. This type of ‘thermal berg et al. 2019).
needle probe’ can involve a thin extension from the Small amplitude high-frequency vertical cycles
front of a conventional CPT, or a parallel thinner elem can also be applied concurrent with the standard
ent behind the friction sleeve, such as the example installation push. This type of test and device is
shown in Figure 5e produced by Datem. known as a Vibro-CPT (VCPT) or Vibropiezocone
71
(e.g. McGillivary et al. 2000, Jorat et al. 2014, Al- system designed particularly to verify the results
Sammarraie et al. 2018). These vibrations can be of ground improvement activities (Figure 6a). For
imposed by an oscillating mass attached to the rods, this application the system must be small, port
or by control of the hydraulic actuator. VibroCPT able and suited to use within existing buildings,
installations typically feature 10-15 Hz vibrations on and requires only a low penetration force and
top of the standard 2 cm/s penetration, creating depth capability. This system uses water as the
a vibratory amplitude on the order of 1 mm at the control fluid, in common with another recently-
cone tip. Due to rod compression, accelerometer developed pressuremeter described by Lopes dos
data from the tip is double integrated to find the tip Santos et al. (2019).
motion. These new systems could support wider adoption
The simplest interpretation of the VCPT is as the of the cone-pressuremeter, taking advantage of the
reduction ratio between qc in a VCPT and an adja use of unload-reload loops and the unloading phase
cent conventional CPT. This has been used as an to gather strength and stress-strain or stiffness deg
indicator of liquefaction potential (e.g. McGillavary radation curves from the measured data. It has also
et al. 2000). In this conference, Al-Sammarraie et al. been demonstrated that pressuremeter tests can
(2022) describe recent VCPT studies that have inves include large numbers of constant pressure-
tigated more detailed interpretation of the qc amplitude cycles, which may allow insights into the
response within individual VCPT cycles. The aim is cyclic response, that can be translated into lateral
to link the measured response to pile-soil interaction p-y response models (e.g. Briaud et al. 1984, Lopes
models used for vibratory pile driving. dos Santos 2020).
In softer soils, a further variation in vertical
movement is to apply varying rates of penetration, 2.5.4 Torsional or lateral motion elements
altering the rate in steps – a so-called ‘twitch’ test Miniaturization of electromechanical systems
(House et al. 2001, Randolph & Hope 2004, Silva allows actuators to be located within the cone
et al. 2006). This test provides a basis to estimate the sleeve. This allows concepts such as a torsional
consolidation coefficient, albeit taking approxi (t-θ) sleeve element (Frost & Martinez 2013,
mately the same time as a conventional dissipation Figure 6b) and a laterally-moving p-y-type element
test. However, a further benefit is that twitch results (Diambra et al. 2022, in this conference, Figure 6c).
provide a ratio between the slow (drained) and fast These concepts offer the potential to provide more
(undrained) resistance, which can be linked to detailed soil characterization, by measuring load-
strength, stiffness or state (e.g. OCR) parameters
(Schneider et al. 2007, Suzuki & Lehane 2015,
Dienstmann et al. 2017, White et al. 2018, Mo et al.
2020).
Penetrometer tests that span a range of motion
rates and also involve periods of cyclic movement
and dissipation periods allow a wide range of soil
strengths to be mapped out (e.g. Chow et al. 2019).
These tests can mimic the loading history applied by
infrastructure to the ground. The results may there
fore be suitable for scaling to the changing capacity
of a foundation system under ‘whole life’ loading
(e.g. Zhou et al. 2020, Gourvenec 2020).
2.5.3 Pressuremeter elements

The earliest example of an additional actuated elem
ent combined with the CPT is the cone pressure-
meter. Early studies using a pressuremeter
incorporated above the cone led to interpretation
solutions that focused on the cyclic and unloading
stages, acknowledging that the initial penetration of
the cone will have disturbed the in situ stress state
(e.g. Houlsby & Withers 1988, Withers et al. 1989,
Bellotti et al. 1989, Ferreira & Robertson 1992,
Bolton & Whittle 1999, Whittle & Liu 2013).
Recently the cone pressuremeter has found new Figure 6. Advances in motion during CPT-type testing: (a)
usage in relation to offshore wind, for the determin pressuremeter module (Vinco & Sachetto 2022), (b) Tor
ation of soil stiffness for monopile foundations (e.g. sional sleeve (Frost & Martinez 2013), (c) lateral (p-y)
Whittle et al. 2017). In this conference Vinco & sleeve (Diambra et al. 2022) and (d) rotational shallow
Sachetto (2022) describe a new cone pressuremeter penetrometer tests (White et al. 2022).
72
Figure 7. Shallow penetrometer system – a combination of CPT, T-bar and ball strength profiling tools and the shallow
ring, toroid and hemiball devices for near-surface soil strength, interface strength and consolidation property measurement
(White et al. 2022).
displacement responses for modes of soil deform several commercial projects (White et al. 2022).
ation that are different to the cylindrical expansion The system features electromechanical drives
of the pressuremeter. with feedback control, so a vertical load can be
For example, laboratory torsional shear tests indi maintained while the device rotates and settles.
cate differences in the rate of pore pressure gener Drained and undrained interface strength param
ation compared to axial shearing. Meanwhile, lateral eters, as well as consolidation properties, are
deformation is more closely linked to pile derived from the tests.
p-y loading than cylindrical expansion. These
locally-actuated sleeve concepts remain to be proven 3 HOW DOES CPT EQUIPMENT ADVANCE?
beyond trial prototype devices, but such robotic
cones (or ROBOCONEs, to use the term coined by
3.1 Are equipment advances led by new ideas?
Diambra et al. 2022) could create exciting possibil
ities for complex in-situ tests. This discussion and the wider CPT’22 conference
The introduction of more complex dual- feature numerous advances in CPT equipment, some
directional actuation in a penetrometer test has been of which will influence future practice, and many of
successful for shallow devices, such as the ring, which will not. So this closing part of the discussion
toroid and hemiball penetrometers. These tests offers some views on how the CPT equipment used
involve penetration of the device by a fraction of in practice becomes more advanced. Hopefully this
a diameter, followed by phases of dissipation (on will prompt the sharing of wider views during the
clay soils) and rotation (Figure 6d). conference, and I look forward to learning from
In this conference, we present test execution these.
and interpretation procedures for a rotational The last equipment example discussed above –
penetrometry system that has been used for more rotational testing using shallow penetrometers
than 300 tests over the past 6 years, including for (Figure 7) – is an example that I have been heavily
73
Figure 8. The bevameter – the original shallow penetrometer? (a) Bevameter principles (Bekker 1969), (b) Laboratory
Bevameter system (Apfelbeck et al. 2011), (c) Shear plates for torsional Bevameter tests (top: Apfelbeck et al. 2011,
bottom: Kim et al. 2021).
involved in, so I will use this as a living case study create an advance but (ii) there may be plenty of old
of an advance that might or might not have a future ideas that we are yet to develop into a suitable form
life. that can be brought into the geotechnical market.
These shallow penetrometers emerged from
research at UWA at a time when we were heavily 3.2 Are equipment advances led by new markets?
focused on seabed pipelines, and their friction on the
seafloor. Centrifuge model testing of short pipeline 3.2.1 Introduction
elements was being used to assess design friction coef The shallow penetrometer story leads to the question
ficients (White & Gaudin 2008). This led to the idea of of markets for advances in penetrometer testing.
a ‘pipe-like’ in situ test, but using rotation to allow Many of the advances reviewed in this paper target
longer distances of interface sliding. Prof. Mark Ran a particular market that requires additional informa
dolph and I proposed this concept as Dr Yue Yan’s tion from a penetrometer test – through an additional
PhD topic and she developed the shallow penetrometer sensor or motion stage. Or, the advances relate to
concept (Yan 2013). Dr Sam Stanier took the topic fur a new market where conventional delivery and
ther in the RIGSS Joint Industry Project, with close deployment methods are unsuited.
support from industry, and co-supervising Dr Mark The offshore engineering examples discussed
Schneider’s PhD on this topic. The devices worked below illustrate three market pulls that have sup
well, we devised rigorous interpretation methods, and ported recent penetrometer equipment advances.
the devices appeared to meet an industry need.
But was this type of test a new idea? It turns out 3.2.2 High temperature high pressure pipelines
not. After the PhD studies and the RIGSS JIP were The emergence of high temperature high pressure
completed, we discovered parallel research in the (HTHP) seabed pipelines around 20 years ago led to
world of terramechanics and agriculture that an increased focus on near-surface seabed strength.
involved a similar type of test – the bevameter Engineered lateral buckling schemes for HTHP pipe
(Bekker 1969, Wong 2010). This device was devel lines require reliable estimates of the embedment
oped in the 1950s and resembles the ring penetrom and the subsequent axial and lateral soil resistance
eter and the torsional plate penetrometer (Watson & (Bruton et al. 2007).
Randolph 1998). This market led to innovations such as miniature
A bevameter test involves penetration and rota box-core-based CPT, T-bar and ball penetrometer
tion of a ring-shaped device, with the pressure systems (Low & Randolph 2010, Puech et al. 2010)
sinkage and torque-rotation results being used to and the shallow penetrometer systems (Yan et al.
characterize the soil for assessment of trafficability 2010, Boscardin & DeGroot 2015, White et al.
(Figure 8). Bevameters have been developed for 2017). These devices are designed to characterize
attachment to lunar landers and Mars rovers only the shallowest half metre or so of the seafloor,
(Bekker 1969, Apfelbeck et al. 2011, Edwards et al. by conducting tests in box core samples. There is
2017). Their interpretation is rarely rooted in the a particular emphasis on remoulded properties,
same soil mechanics principles as are familiar for which control pipeline embedment, and interface
CPTs and our shallow penetrometers, but the test strength, which controls axial pipeline friction.
concept is similar.
So – shallow penetrometers might deserve to be 3.2.3 Subsea power cables
counted as an advance, but they are not a new idea. The offshore wind industry has created a major
The message is perhaps that (i) a new idea may market for cable route surveys, which has
74
similarities with the on-bottom pipeline survey samples to reach the testing laboratories, be tested,
market. However, a particular feature of subsea and an interpreted set of design parameters be
cable design is the importance of thermal properties. generated.
The thermal insulation created by the trench backfill If in situ testing takes a larger role in determining
can have a strong influence on the cable rating and design parameters, then there is a strong market pull
operating temperature (Dix et al. 2017). for new types of in situ test, or extensions of the
A response to this market has been the develop CPT, which can explore the detailed aspects of soil
ment of new in situ thermal tests, that can be inte response that are conventionally determined from
grated within a conventional CPT-based route laboratory testing – such as cyclic behaviour, and
survey. Some of these devices (e.g. Figure 5e) use stiffness parameters.
the violin bow-type design of thermal probe that has This market pull has created the driver for cyclic
been used by the earth science community for study and small strain components of CPT-type tests such
ing heat flow from the deep earth into the oceans as the seismic CPT, the VCPT, the cone
(e.g. Lister 1979). Once again – the idea is not new, pressuremeter and the locally-actuated lateral and
but the concept has recently entered geotechnical torsional sleeves. The ability to determine thermal
survey practice in response to a market pull. properties in situ, rather than relying on laboratory
Another approach to a thermal measurement is tests, also allows compression of the project
the passive thermal CPT (T-CPT) system, which is timeline.
easily added to a standard cone test. In this test the
decay of the heat developed by the penetration pro
3.3 Are equipment advances led by new
cess is used to infer the thermal conductivity. This
technology?
simplicity has allowed the T-CPT to rapidly enter
geotechnical survey practice. A final driver of equipment advances is the emer
A further driver for adoption of the T-CPT is the gence of new technologies across mechanical and
evidence that thermal conductivity measurements electronic engineering, as well as new types of
from samples are biased compared to in situ data. vehicle. Over the past twenty years, the emergence
A review of hundreds of in situ and ex situ thermal of low cost sensors, data acquisition systems, actu
conductivity measurements by Vardon et al. (2019) ators and communication technologies have permit
showed that the thermal conductivity measured in ted new types of sensor and testing protocol to be
laboratory tests is consistently lower than in situ devised. Miniaturization has allowed additional
data. This is attributed to slight de-saturation and equipment to be packed into the shaft of the CPT.
a reduction in density of the samples (Vardon et al. Also, the low cost and wide availability of mecha
2019). tronic devices has opened up and democratized the
process of equipment development. Small companies
3.2.4 Project acceleration: Reliance on in situ data or individuals can now prototype sensors and
A further example of a market pull that is motivating devices that could previously only be produced and
advances in CPT equipment, is the need to accelerate trialled by major organisations.
the maturation of projects, from planning to oper Massarch (2014), in his comprehensive review
ation. This is linked to the energy transition, with of the CPT, past and present, discussed a future
wind projects aiming for faster progression from in which sensors are so cheap that the device can
planning or final investment decision (FID) to the be left in situ after the penetration test. This idea
beginning of energy production, compared to oil and opens up the possibility that penetration tests are
gas projects. not a snapshot of the ground condition at
As an illustration, offshore wind projects in a moment in time, but implant a sensor in the
Europe typically take 7-10 years to progress from ground, that continues to operate, relaying back
initial planning to operation and in China the process the condition of the ground through the life of
is compressed into 2-4 years. In the past, major oil the surrounding infrastructure. The boundary
and gas projects often had a longer timeline than between in situ testing and field monitoring thus
European offshore wind, and also a smaller footprint becomes blurred.
to survey. Automation technology also means that the
The rest of the world needs to approach Chinese execution of a CPT is becoming an ‘office job’.
rates of project maturation if net zero targets are to In this conference, Storteboom et al. (2022b)
be met. DNV (2021) predict a global wind capacity describe a new hybrid CPT truck that operates
of 5.9 TW by 2050, which will require around using battery when on site. A carousel system
a quarter of a million turbines to be installed in the automatically connects the CPT rods so the oper
next 30 years, with the annual rate of installation ator has minimal mechanical activities to per
increasing by a factor of 10 compared to now. form. Storteboom et al. (2022b) report that “the
One method to compress the project timeline is to workplace of the CPT operator is turned into an
increase the use of in situ geotechnical test data for office environment”. They describe an improve
determining design parameters. This can eliminate ment in working conditions, reduced absenteeism
or reduce the importance of the period required for and higher productivity.
75
A final example of emerging technology is the which creates difficulties when clients wish to
digital environment used to hold and process CPT assure data quality or set contract specifications and
data, and the integration of this into the geotechnical develop a competitive supplier market.
design workflow and the digital model of the result Two key challenges are therefore faced by devel
ing construction. The geotechnical survey industry opers of any new equipment that deviates from the
in the UK was an early adopter of standardized data standards:
formats, through the AGS standard, which is named
1. The need for a specification: a test specification
after the Association of Geotechnical and Geoenvir
is required to define and allow assurance that any
onmental Specialists, who are custodians of the
testing is conducted according to an appropriate
standard. Version 1 of this data format was released
procedure, agreed between client and contractor.
in 1992, and version 4.1.1 is the current edition
A hierarchy of specification types, listed by the
(AGS 2022).
strength of their basis, might be: (i) a publication
The AGS format is widely used in the UK and
by the system developers, (ii) a guideline emer
worldwide, and improves the workflow and assur
ging from a Joint Industry Project, or developed
ance of geotechnical data, as well as simplifying the
with a verification agency, (iii) a guideline devel
sharing and public release of geotechnical data. For
oped more openly by an industry technical com
example, the Crown Estate (England and Wales)
mittee or association (e.g. CFMS, CIRIA and the
requires wind farm operators to release survey data
ISSMFE have supported guidelines on in situ
to the Marine Data Exchange portal. This data com
testing) (iv) a recommended practice or standard
monly takes the form of reports, but for geotechnical
produced by an international standards body (e.g.
in situ testing the data is often held in the AGS
ISO, CEN, API).
format, where it can be rapidly assimilated into other
In practice, procedures generally evolve upwards
software for wider use.
through this hierarchy over a period of perhaps
A common software platform for storing and
a decade – as seen by the procedures adopted for
interpreting in situ data is gINT, which is owned by
flow-round penetrometers that have now reached
the Bentley software house. Bentley also own
ISO 19901-8.
PLAXIS, and the two programs have growing inter
2. The need for market competition. A second chal
face capabilities. A consequence of this connectivity
lenge facing proponents of an advance in pene
from raw CPT data to geotechnical numerical mod
trometer equipment is that new test types will
elling software is that there is a single digital work-
rarely be specified by clients if there is only one
flow from site measurements to design output.
This connectivity can be used to allow more rapid supplier. This places developers of new equip
progression from survey to construction. It can also ment in the difficult position of requiring
eliminate the human attention given to geotechnical a competitor to come into existence in order for
demand to be created. This situation does not
data that is often held dear in our discipline. For
example, the progression of data through factual apply to equipment advances that are ‘within
reporting, interpretive reporting and ground model standard’, since the specification is unaffected.
development involves detailed evaluation and However, using the example of the shallow pene
thought processes that support geotechnical risk ana trometers for a final time, research funding for
lysis, as well as design parameter choices. the RIGSS JIP from owners/operators was condi
On the other hand, we could take the counter tional on more than one contractor also being
view, which is that a smooth processing pipeline part of the project, in order to ensure that
from in situ data towards design will liberate geo a competitive supply chain could emerge.
technical specialists. We can spend more time The heading for this final sub-section was per
assessing the information created by the data, rather haps a little frivolous. Standards are invariably
than passing data between spreadsheets and different beneficial for industry and should not hinder innov
types of software. ation. However, innovators of new ‘non-standard’
in situ test equipment need to be prepared to
3.4 Equipment advances: Does standardisation address the two challenges set out above. Develop
help or hinder? ing reliable equipment that offers a technical benefit
over standard test methods will not necessarily lead
Standardisation of the CPT emerged in the 1970s to industry adoption. Time and effort must be spent
and has more recently been set out in various inter ‘socialising’ the technology among the community
national standards produced by the ISO, CEN and of practice, for example at events like this confer
others (e.g. ISO 2012, 2014). ence. It is then necessary to develop open guidance
Many advances in CPT equipment remain and specifications for execution and interpretation
‘within standard’, such as methods for efficient of non-standard tests. Finally, it might even be
deployment of the rod string or improvements to necessary to support open competition in the
load cells or data transfer. However, many advances marketplace to avoid the technology being from
in penetrometer equipment are ‘non-standard’, a sole source.
76
4 CONCLUDING COMMENTS Al-Sammarraie D., Kreiter S., Stahler M.O., Goodarzi M.
& Morz T. (2018) New vibratory cone penetration
The basic configuration of the cone penetrometer device for in-situ measurement of cyclic softening.
test (CPT) has remained constant for around 50 International Symposium on Cone Penetration Testing,
years. This discussion has reviewed recent advances 2018.
in penetrometer equipment presented in this confer Apfelbeck M., Kuß S., Rebele B., Schäfer B. (2011)
ence and elsewhere, considering separately the (i) A systematic approach to reliably characterize soils
based on Bevameter testing. Journal of Terramechanics,
delivery, (ii) deployment, (iii) sensing and (iv) 48(5):360–371
motion aspects of the penetrometer test. Baez V.M., Shah A., Akinwande S., Jafari N.H. &
It is evident that new innovations in instrumentation, Becker A.T. (2020) Assessment of soil strength using
electronics and mechanical engineering have created a robotically deployed and retrieved penetrometer.
a vast range of variations on the conventional CPT IEEE/RSJ International Conference on Intelligent
equipment. In some cases these advances provide Robots and Systems (IROS) pp. 7324–7329, doi:
a small but widespread improvement in penetrometer 10.1109/IROS45743.2020.9341424.
testing, such as improved rod connection systems, Bekker, M.G. (1969) Introduction to Terrain-Vehicle Sys
tems. University of Michigan Press, Ann Arbor, Mich.
hybrid-powered CPT trucks and better saturation and
Bellotti, R., Ghionna, V., Jamiolkowski, M., Robertson, P.
calibration systems. In other cases the advances are K. & Peterson R.W. (1989) Interpretation of moduli
more ambitious, aiming to incorporate new sensors and from self-boring pressuremeter tests in sand. Géotechni
motion systems within a penetrometer, to allow new que 39(2):269–292.
types of measurement to be incorporated in the testing Boggess, R. and Robertson, P.K. (2010) CPT for soft sedi
process. ments and deepwater investigations. International Sym
The result is a dispersion of practice: the majority posium on Cone Penetration Testing, CPT’10.
of civil engineers base their ground characterization 2:127–136.
on a durable 1970s technology, updated with auto Bolton, M. D. & Whittle, R. W. (1999) A non-linear elas
mation and improved data quality, while a minority tic/perfectly plastic analysis for plane strain undrained
expansion tests. Géotechnique, 49(1):133–141.
are enjoying exploration of non-standard innov Boscardin A.G. & Degroot D.J. (2015) Evaluation of
ations – with mixed levels of impact. a toroid for model pipeline testing of very soft offshore
This discussion, and the wider conference, is an box core samples Frontiers in Offshore Geotechnics III
opportunity to reflect, learn and debate how we will – Proc. 3rd Int. Symp. on Frontiers in Offshore Geo-
characterize the ground in another 50 years. tech., ISFOG 2015. 363–368.
Briaud, J. L., Riner, K. B. & Ohya, S. (1984) Cyclic pres
suremeter tests for cyclic lateral loads. Offshore Tech
ACKNOWLEDGEMENTS nology Conference, Houston. doi:10.4043/4678-MS
Broms, B.B. & Flodin, N. (1988) History of soil penetra
I am grateful for the kind invitation by the CPT’22 tion testing. International symposium on penetration
organising committee to present this discussion testing, ISOPT 1, 1:157–220.
Bruton D., Carr M. & White D.J. (2007) The influence of
paper. In relation to the shallow penetrometer pipe-soil interaction on lateral buckling and walking of
research, I acknowledge the support for that work pipelines: the SAFEBUCK JIP. Proc. 6th Int. Conf. on
from the RIGSS JIP (Fugro, Total, Woodside and Offshore Site Investigation and Geotechnics, London.
Shell), as well as the collaboration with Dr Sam 133–150.
Stanier, Dr Henning Mohr, Prof. Mark Randolph and Chow S.H., O’Loughlin C.D., White D.J. & Randolph M.F.
Dr Mark Schneider. (2017) An extended interpretation of the free-fall piezo
This review has also been supported by the cone test in clay. Géotechnique. 67(12):1090–1103
EPSRC Offshore Renewable Energy Supergen Hub Chow S.H., O’Loughlin C.D., Zhou Z., White D.J. &
Randolph M.F. (2020). Penetrometer testing in
(grant EP/S000747/1) and by the EPSRC grant EP/
a carbonate silt to explore changes in soil strength. Géo
W006235/1 (ROBOCONE – intelligent robotics for technique 70(12):1160–1173
next generation ground investigation and design – PI DeJong, J.T. and Frost, J.D. (2002). A multi-friction sleeve
Dr Andrea Diambra). All opinions and any mistakes attachment for the cone penetrometer. ASTM Geotech
or misrepresentations are my own. nical Testing Journal, 25(2):111–127
Diambra A., Ciavaglia F., Harman A., Dimelow C.,
Carey J. & Nash D. F. T. (2014). Performance of cyclic
REFERENCES cone penetration tests in chalk. Geotechnique Letters, 4
(3):230–237
AGS (2022) Electronic transfer of geotechnical and geoen Diambra A., Creasey J., Leonet J., Conn A., Ibraim E.,
vironmental data, AGS4 Edition 4.1.1 – March 2022 Mylonakis G., White D.J., Cerfontaine B.,
Publ. Association of Geotechnical and Geoenvironmen Gourvenec S.M. & Igoe D. (2022) Concept design
tal Specialists of a new CPT module for direct in situ measurement
Al-Sammarraie D., Kreiter S., Kluger M.O., Goodarzi M. of p-y soil responses. Proc. 5th Int. Symposium on
& Morz T. (2022) VCPT: an in situ soil investigation Cone Penetration Testing (CPT’22). Bologna. (this
method to validate the vibratory pile-soil interaction conference).
models. Proc. 5th Int. Symposium on Cone Penetration Dienstmann, G.; Schnaid, F.; Maghous, S.; Dejong, J.T.
Testing (CPT’22). Bologna. (this conference). (2017). Piezocone penetration rate effects in transient
77
gold tailings. Journal of Geotechnical and Geoenviron ISO (2012). ISO 22476-1:2012: Geotechnical investigation
mental Engineering, ASCE. 04017116-1 and testing - Field testing - Part 1: Electrical cone and
Dix, J.K., Hughes, T.J., Emeana, C.J., Pilgrim, J.A., piezocone penetration test. International Standardisation
Henstock, T.J., Gernon T.M., Thompson, C.E.L. & Organisation
Vardy, M.E., (2017) Substrate controls on the life-time ISO (2014). ISO 19901-8:2014: Petroleum and natural gas
performance of marine HV cables. Proc. 8th Int. Conf. industries - Specific requirements for offshore structures
Offshore Site Investigations and Geotechnics. Society - Part 8: Marine soil investigations. International Stand
for Underwater Technology. 88–108. ardisation Organisation [a 2022 edition is currently
DNV (2021) Energy transition outlook. DNV, Hovik. under international ballot]
Edmunds J. (2014) Gaining traction. Offshore Engineer Jeanjean, P., Spikula, D. & Young, A. (2012). Technical
Magazine. 3 February 2014. vetting of free-fall cone penetrometer. Proc. Int. Conf.
Edwards M.B., Dewoolkar M.M., Huston D.R. & on Offshore Site Investigation and Geotechnics,
Creager C. (2017) Bevameter testing on simulant Fillite London, UK. 179–186
for planetary rover mobility applications. Jefferies M.G. & Funegard E. (1983). Cone penetration
J. Terramechanics, 70:13–26 testing in the Beaufort Sea. Proc Conference on Geo-
Ferreira, R.S. & Robertson, P.K. (1992) Interpretation technical Practice in Offshore Eng., Austin, Texas,
of undrained self-boring pressuremeter test results ASCE, 220–243
incorporating unloading. Can. Geotechnical J., Jorat, M. E., Mörz, T., Schunn, W., Kreiter, S., Moon, V., &
29:918–928. de Lange, W. (2014). Geotechnical Offshore Seabed
Frost J.D. & Martinez A. (2013) Multi-sleeve Tool (GOST): a new cone penetrometer. Proc. 3rd Inter
axial-torsional-piezo friction penetration system for sub national Symposium on Cone Penetration testing.
surface characterization. Proc. 18th Int. Conf. on Soil 207–215
Mech. and Geotech. Eng. (18th ICSMGE). Paris, Kang H., Kwon O., Shin C., Seo J. & Jang I. (2022).
France, 1:527–530. Development of free fall cone penetration testing
Geise, J.M. & Kolk, H.J. (1983). The use of submersibles system. Proc. 5th Int. Symposium on Cone Penetration
for geotechnical investigations. Proc. Submarine Tech Testing (CPT’22). Bologna. (this conference).
nology, The Design and Operation of Underwater Kelleher, P. J. & Randolph, M. F. (2005) Seabed geotech
Vehicles, Paper 7.3, London, Society for Underwater nical characterisation with a ball penetrometer deployed
Technology. from the Portable Remotely Operated Drill. Int. Symp.
Gourvenec S.M. (2020) Whole-life geotechnical design: on Frontiers in Offshore Geotechnics I, Perth, Australia.
What is it? What’s it for? So what? And what next?: 365–371
Keynote. In Proc. 4th Int. Symp. on Frontiers in Off Kim, J.T., Im, D.U., Choi, H.J., Oh, J.W., Park, Y.J., (2021)
shore Geotechnics. Deep Foundations Institute. Development and performance evaluation of a bevameter
206–246. for measuring soil strength. Sensors 21, 1541
Gundersen A.S., Carotenuto P., Lunne T., Walta A. & Krömer O., Scharringhausen M., Fittock M., Tsakyridis G.,
Sparrevik P.M. (2019) Field verification tests of the Wippermann T., Witte L., Grott M., Knollenberg J.,
newly developed flow cone tool—In-situ measurements Spohn T., Krause C., Hudson T.L., Lichtenheldt R.,
of hydraulic soil properties. AIMS Geosciences, 5 Grygorczuk J., & Wisniewski L. (2019) Design details
(4):784–803. of the HP3 mole onboard the InSight mission. Acta
Gundersen A.S., Lunne T., Stelzer R., Blaker O, Tucker G. Astronautica. 164:152–167
W. & Krogh L. (2022) Flow cone – new CPTU add-on Lister, C.R.B., (1979) The pulse-probe method of conduct
module trialed in Halden silt. Proc. 5th Int. Symposium ivity measurement,Geophys. J. Royal Astr. Soc.,
on Cone Penetration Testing (CPT’22). Bologna. (this 57:451–461
conference). Lopes dos Santos, A. & Dupla, J-C & Canou, J &
Hebeler, G.L., Frost, D.J., Schneider, J.A. & Lehane, B.M. Droniuc, N. (2019). Laboratory validation of an innova
(2005). Cyclic friction piezocone tests for offshore tive mono-cell pressuremeter probe: test procedures and
applications, Int. Symp. on Frontiers in Offshore Geo- first results. Proc. XVII European Conference on Soil
tech. 967–972. Mechanics and Geotechnical Engineering. Reykjavik,
Hebeler, G.L., Martinez A. & Frost J.D. (2018) Interface Iceland. 8pp.
response-based soil classification system. Canadian Lopes dos Santos A. (2020). Determination of soil shear
Geotech. J., 55(12):1795–1811. modulus at low strain level using an innovative pres
Houlsby, G.T. & Withers, N.J. (1988) Analysis of the cone suremeter probe: Application to cyclic pile design. Dis
pressuremeter test in clay. Géotechnique, 38 sertation presented for the degree of Doctor of
(4):575–587. Université Paris-Est.
House, A.R., Oliveira, J.R.M. and Randolph, M.F. (2001) Low, H.E. & Randolph, M.F. (2010). Strength measure
Evaluating the coefficient of consolidation using pene ment for near-seabed surface soft soil using manually
tration tests. Int. J. Physical Modelling in Geotechnics, 1 operated miniature full-flow penetrometer. ASCE Jour
(3): 17–26. nal of Geotechnical and Geoenvironmental Engineering
Hvorslev, M. J. (1951) Time lag and soil permeability in 136:1565–1573.
ground-water observations. US Army Corps of Engin Lunne, T., Robertson, P.K. & Powell, J.J.M. (1997). Cone
eers, Waterways Experiment Station Bulletin No. 36, Penetration Testing in Geotechnical Practice, Blackie
Mississippi, USA. Academic/London, Routledge, New York: 312 p.
Isaev O.N., Sharafutdinov R.F. & Zakatov D.S. (2022) Lunne, T. (2012). The CPT in offshore investigations –
Determination of natural stress state parameters for clay a historic perspective. Geomechanics and Geoengineer
soils by using 3LSU-CPTU penetrometer. Proc. 5th Int. ing: An International Journal. 2(2):75–101.
Symposium on Cone Penetration Testing (CPT’22). Bol Lunne T., Ghanekar R., Tucker G.W., Santos R. &
ogna. (this conference). Krogh L. (2022) The revival of multiple pore pressure
78
measurements in the Cone Penetration Test. Proc. 5th of Soil Properties and Case Histories, Bali, Indonesia.
Int. Symposium on Cone Penetration Testing (CPT’22). 667–672.
Bologna. (this conference). Rocchi I., Tonni L., Gottardi G. & Marcolongo M. (2022)
Martinez A., DeJong, J.T., Jaeger, R. & Khosravi A. Experimental procedure for checking the saturation
(2020). Evaluation of self-penetration potential of a degree of piezocone tips. Proc. 5th Int. Symposium on
bio-inspired site characterization probe by cavity expan Cone Penetration Testing (CPT’22). Bologna. (this
sion analysis. Canadian Geotech. J., 57(5):706–716. conference).
Massarsch K.R. (2014) Cone Penetration Testing – Schneider J.A., Lehane B.M. & Schnaid F. (2007) Velocity
A Historic Perspective. Proc. 3rd International Sympo effects on piezocone measurements in normally and
sium on Cone Penetration Testing. 97–134 over consolidated clays. International Journal of Phys
McConnell A. & Wassenaar E. (2022) An innovative new 3 ical Modelling in Geotechnics. 7:23–34
MPa CPT – to detect and measure very small fs values. Scholz C., Göttinger M., Hinck S., Möller K. &
Proc. 5th Int. Symposium on Cone Penetration Testing Ruckelshausen A. (2014) Automatic soil penetrometer
(CPT’22). Bologna. (this conference). measurements and GIS-based documentation with the
McGillivary, A., Casey, T., Mayne, P. W. & autonomous field robot platform BoniRob. 12th Int.
Schneider, J. A. (2000) An electro-vibrocone for Conf. of Precision Agriculture.
site-specific evaluation of soil liquefaction potential. Shonberg A., Harte M., Tucker G. Rattley M. &
Proc. Geo-Denver 2000, Innovations and Applications Gibbs P. (2019) Innovative use of the CPT to assess
in Geotechnical Site Characterization. the in situ degradation of an offshore overconsoli
Meunier, J., Sultan, N., Jegou, P. & Harmegnies, F. (2004) dated clay. Proc. Offshore Technology Conference.
First tests of Penfeld: a new seabed penetrometer. Proc. Paper OTC29414.
14th Int. Offshore & Polar Eng. Conf., Toulon, France, Silva, M. F., White, D. J., and Bolton, M. D. (2006) An
338–345 analytical study of the effect of penetration rate on
Mo, P-Q, Gao, X-W, Yang, W, Yu, H-S. (2020) A cavity piezocone tests in clays. Int. J. Numerical & Analytical
expansion–based solution for interpretation of CPTu Methods in Geomechanics 30(6):501–527.
data in soils under partially drained conditions. Spohn T., Grott M., Smrekar S.E., Knollenberg J.,
Int J Numer Anal Methods Geomech. 44:1053–1076. Hudson T. L., Krause C., Müller N., Jänchen J.,
Naclerio, N.D., Karsai, A., Murray-Cooper M., Ozkan- Börner A., Wippermann T., Krömer O., Lichtenheldt R.,
Aydin Y., Aydin E., Goldman D.I. & Hawkes E.W. Wisniewski L., Grygorczuk J., Fittock M.,
(2021) Controlling subterranean forces enables a fast, Rheershemius S., Spröwitz T., Kopp, E., Walter I.,
steerable, burrowing soft robot. Science Robotics 6:55 Plesa A. C., Breuer D. & Morgan, P. (2018) The Heat
eabe2922. Flow and Physical Properties Package (HP3) for the
Olmedo, N.A. & Lipsett, M.G. (2016) Design and field InSight mission. Space. Sci. Rev., 214, 96
experimentation of a robotic system for tailings Spohn T., Hudson T.L., Witte L., Wippermann T.,
characterization. J. Unmanned Vehicle Systems 4 Wisniewski L., Kediziora B., Vrettos C., Lorenz R.D.,
(3):169–192. Golombek M., Lichtenfeld R., Grott M., Knollenberg J.,
Peuchen J., Looijens P. & Stark N. (2017) Offshore charac Krause C., Fantinati C., Nagihara S., & Grygorczuk J.
terisation of extremely soft sediments by free fall (2021) The InSight HP3 mole on Mars: Lessons learned
penetrometers. Proc. Int. Conf. on Offshore Site Investi from attempts to penetrate to depth in the Martian soil.
gation and Geotechnics. Society for Underwater Tech Submitted to Advances in Space Research. On arXiv,
nology. 370–377 ref. arXiv:2112.03234
Power P. & Geise J. (1994) Offshore soil investigation Stark N., Hanff H., & Kopf A. (2009) Nimrod: a tool for
techniques and equipment for the next century. Proc. of rapid geotechnical characterization of surface
the Int. Conf. on Behaviour of Offshore Structures, sediments. Sea Technology. 50(4):2230–2235.
BOSS’94, Boston, 1:97–109 Stark N. (2016) Geotechnical site investigation in energetic
Puech, A., Orcozo-Calderón, M. and Foray, P. (2010). Mini nearshore zones: opportunities & challenges. Keynote
T-bar testing at shallow penetration. Int. Symp. on Fron lecture. Int. Conference on Geotechnical and Geophys
tiers in Offshore Geotechnics II, Perth, Australia. ical Site Characterisation 5. 69–80
305–310. Steiner A., Kopf A.J., L’Hereux J-S., Kreiter S.
Randolph, M.F. (2004). Characterisation of soft sediments Stegmann S., Haflidason H. & Moerz T. (2014). In situ
for offshore applications. 2nd Int. Conf. on Site Charac dynamic piezocone tests in natural clayey soils –
terization, Rotterdam, The Netherlands, 1:209–232. a reappraisal of strain-rate corrections. Canadian Geo-
Randolph, M.F., Stanier, S.A., O’Loughlin, C.D., Chow, S. technical Journal. 51(3): 272–288.
H., Bienen, B., Doherty, J.P., Mohr, H., Ragni, R., Stephan S., Kaul N., & Villinger, H. (2015). Validation of
Schneider M.A., White D.J., Schneider J.A. (2018) impact penetrometer data by cone penetration testing
Penetrometer equipment and testing techniques for off and shallow seismic data within the regional geology of
shore design of foundations, anchors and pipelines. the southern north sea. Geo-Marine Letters. 35(3):
Invited keynote Lecture. Proc. 4th Int. Symp. on Cone 203–219.
Penetration Testing, CPT ’18. Delft. 3–24. Storteboom O., Woollard M. & Allen B. (2022a) Increased
Randolph, M. F., and Hope, S. N. (2004) Effect of cone safety and productivity for seabed Cone Penetration
velocity on cone resistance and excess pore pressure. Test using the SingleTwist™ with ROSON. Proc. 5th
Proc. Int. Symp. On Engineering Practice and Perform Int. Symposium on Cone Penetration Testing (CPT’22).
ance of Soft Deposits, Yodogawa Kogisha Co. Ltd., Bologna. (this conference).
Osaka, Japan, 147–152. Storteboom O., Woollard M. & Ooms B. (2022b) Sustain
Ratnam S., Soga K., Mair R.J., Whittle R.W. & Tedd P. ability in CPT practice: hybrid CPT track-truck. Proc.
(2001) Permeability measurement using the self-boring 5th Int. Symposium on Cone Penetration Testing
pressuremeter. Proc. Int. Conf. on In Situ Measurement (CPT’22). Bologna. (this conference).
79
Suzuki Y., & Lehane B.M. (2015) Cone penetration at vari Symposium on Cone Penetration Testing (CPT’22). Bol
able rates in kaolin–sand mixtures. International Journal ogna. (this conference).
of Physical Modelling in Geotechnics 15(4): 209–219. Whittle R., & Liu L. (2013). A method for describing the
Tao J., Huang S., Tang Y. (2020) SBOR: A minimalistic stress and strain dependency of stiffness in sand. Proc.
soft self-burrowing-out robot inspired by razor clams. 18th Int. Conf. on Soil Mechanics and Geotechnical
Bioinspiration and Biomimetics. 15, 055003. Eng. Paris.
Tarantino A., Capotosto, Bottaro F., Bellio M. & Whittle R., Palix E., & Donaghy D. (2017). The influence
Gallipoli D. (2022) Tensiocone: a cone penetrometer of insertion process on determining the stiffness charac
with the facility to measure negative pore-water teristics of chalk, using pre-bored, self-bored and
pressure. Proc. 5th Int. Symposium on Cone Penetration pushed pressuremeters. Proc. 8th Int. Conf. on Offshore
Testing (CPT’22). Bologna. (this conference). Site Investigation & Geotech. London Soc. for Under
Vardon P.J., Baltoukas D. & Peuchen, J. (2018) Thermal water Tech. 308–315.
Cone Penetration Test (T-CPT). In Proc. 4th Inter Winter A.G., Deits R.L.H., Dorsch D.S., Slocum A.H. &
national Symposium on Cone Penetration Testing. Hosoi A.E. (2014) Razor clam to roboclam: Burrowing
649–655. drag reduction mechanisms and their robotic adaptation.
Vardon P.J., Baltoukas D. & Peuchen J. (2019) Interpreting Bioinspiration and Biomimetics. 9, 036009.
and validating the thermal cone penetration test Withers N.J., Howie J., Hughes J.M.O. & Robertson P.K.
(T-CPT). Géotechnique, 69(7): 580–592. (1989) Performance and analysis of cone pressuremeter
Vinco G. & Sachetto M. (2022) New portable pressiocone tests in sands. Géotechnique 39(3): 433–454
system for carrying out CPT & FDP tests. Proc. 5th Int. Wippermann T., Hudson T.L., Spohn T., Witte L.,
Symposium on Cone Penetration Testing (CPT’22). Bol Scharringhausen M., Tsakyridis G., Fittock M.,
ogna. (this conference). Krömer O., Hense S., Grott M., Knollenberg J. &
Watson, P.G. & Randolph, M.F. (1998). Torsional vane and Lichtenheldt R. (2020) Penetration and performance
plate load tests. Proc. International Conference Centri testing of the HP3 Mole for the InSight Mars mission.
fuge ‘98, Rotterdam, Netherlands. 1: 167–173. Planetary Space Science. 181, 104870
White D.J. & Gaudin C. (2008) Simulation of seabed Wong J.Y. (2010) Theory of ground vehicles. 4th Edition.
pipe-soil interaction using geotechnical centrifuge Wiley, New York.
modelling. Proc. 1st Asia-Pacific Deep Offshore Tech Yan Y., White D.J. & Randolph M.F. (2010) Investigations
nology Conference, Perth, Dec 2008. 28pp. into novel shallow penetrometers for fine-grained soils.
White D.J., Stanier S.A., Schneider M., O’Loughlin C.D., Proc. 2nd Int. Symp. on Frontiers in Offshore Geotech
Chow S.H., Randolph M.F., Draper S.D., Mohr H., nics. Perth. 321–326
Morton J.P., Peuchen, J., Chow F.C., Fearon R. & Yan Y. (2013) Novel methods for characterising pipe-soil
Roux A. (2017) Remote intelligent geotechnical seabed interaction forces in situ in deep water. PhD thesis, Uni
surveys – technology emerging from the RIGSS JIP. versity of Western Australia
Proc. Int. Conf. Offshore Site Investigation & Geotech Yetginer T.I., Bodtker S., Rose M., Lunne T., Meyer V. &
nics. SUT, London. 1214–1222 Duffy C. (2022) Development of an enhanced CPT
White D.J., O’Loughlin C.D., Chow S.H. & Stark N. system for Dogger Bank. Proc. 5th Int. Symposium on
(2018) Interpretation of free fall penetrometer tests in Cone Penetration Testing (CPT’22). Bologna. (this
sands: An approach to determining the equivalent static conference).
resistance. Proc. 4th Int. Symp. on Cone Penetration Zhou Z., O’Loughlin C.D. & White D.J. (2020) The chan
Testing. 695–701. ging strength of carbonate silt: Parallel penetrometer
White D.J., Stanier S.A. & Mohr H. (2022) Characterisa and foundation tests with cyclic loading and reconsoli
tion of near-surface sediments using a blend of vertical dation periods. Canadian Geotechnical Journal 57
and shallow rotational penetrometers. Proc. 5th Int. (11):1664–1683.
80
© 2022 the Author(s), ISBN 978-1-032-31259-0
Non-deterministic interpretation and applications of CPT testing data

M. Uzielli
Department of Civil and Environmental Engineering, University of Florence, Italy
Georisk Engineering S.r.l., Florence, Italy
ABSTRACT: The evolution of geotechnical design codes towards reliability-based and probabilistic con
cepts, the growing global awareness towards risk management for engineered and natural systems, and the
global digital transition are steering geotechnical research and practice towards non-deterministic and data-
centric approaches to geotechnical characterization and design. The quasi-continuity and high repeatability of
CPT measurements inherently endow this testing method with a central role in the momentous shift in para
digm. This paper contributes principles and selected examples which exemplify the nature and use of non
deterministic CPT-based methods for geotechnical site characterization, design, and geohazards risk manage
ment. Current cautions and limitations to the development, dissemination, and implementation of these methods
are arguably surmountable and provide stimulating opportunities for a collective effort on the part of the CPT
community to meet the evolving requirements and trends in geotechnical practice, research, and education.
...ἔοικα γοῦν τούτου γε σμικρῷ τινι αὐτῷ τούτῳ groundwater level, etc.) and to imperfect measurement
σοφώτερος εἶναι, ὅτι ἃ μὴ οἶδα οὐδὲ οἴομαι εἰδέναι. capability. Some degree of inaccuracy and imprecision
is always present in testing operations.
''... I seem, then, in just this little thing to be wiser The variability in measured values thus stems
than this man at any rate, that what I do not know from inherent soil variability and measurement uncer
I do not think I know either.'' tainty. When transposed in quantitative engineering
- Plato, Apology analyses, these variabilities result in uncertainty.
When representative values are calculated from
measured data (e.g., the mean value over a depth
1 LEVERAGING THE UNAVOIDABLE interval), additional statistical estimation uncertainty
arises from the limited number of values used to cal
“As far as laws of mathematics refer to reality, they culate these values. Moreover, when testing data are
are not certain; and as far as they are certain, they used to calculated engineering design parameters,
do not refer to reality.” a further source of uncertainty is introduced due to
- Albert Einstein, “Geometry and experience” (1921) transformation models which are always approximate
The variability in testing data and derived design and imperfect models of the physical world.
parameters is a distinctive trait of the geotechnical dis Given the evident existence of considerable
cipline with respect to other engineering disciplines uncertainties in geotechnical engineering, the ques
which deal with artificial materials. Measured proper tion arises as to “whether” and “how” to address
ties effectively vary due to the natural heterogeneity them. The complete neglection of uncertainty is
and complexity of soils, which are generated and con unacceptable from both the ethical and technical
tinuously modified by natural geologic and geomor viewpoints. Failing to acknowledge the existence
phologic processes. In-situ effects due to stress state and relevance of uncertainty would imply that basic
and stress history lead to spatially variable measure geotechnical notions and geotechnical code prescrip
ments even for compositionally homogeneous tions are not adhered to, since these do address
deposits. The spatial and temporal variability of geo uncertainty (either implicitly or explicitly, as dis
technical properties can be especially appreciated in cussed later). This paper focuses on how the geo
the results of geophysical testing and geotechnical in- technical discipline has evolved from the primordial
situ testing, which involve larger volumes of soils than “deterministic” paradigm, in which the modeling of
laboratory tests. Moreover, the same test conducted at uncertainty is implicit, to the currently trending
the same spatial location may produce different results “non-deterministic” paradigm, in which uncertainties
due to the temporal variability of the natural environ are modelled, processed, and reported explicitly, and
ment (e.g., seasonal fluctuations in water content, is continuing to evolve towards a “data-centric”
DOI: 10.1201/9781003308829-6
81
mode which fosters the synergy between engineering data-centric paradigm is provided below, readers are
judgment and the increased availability of data, referred to Baecher & Christian (2003), Phoon
facilitating the treatment of uncertainty in character (2020) and Tang & Phoon (2021) for further insights
ization and design. into the philosophical, conceptual, and operational
Through a specific focus on cone penetration evolution of the geotechnical discipline.
(CPT) testing, this paper attempts to provide
examples from research and case-study applications
2.1 The past: The illusion of certainty
of the pivotal role of CPT testing in accompanying
the geotechnical discipline towards an emerging Geotechnical engineering, as all engineering discip
non-deterministic and data-centric paradigm. Far lines, originated and developed speaking the deter
from aiming to provide a structured theoretical treat ministic language, in which uncertainties are neither
ment of non-deterministic and data-centric modelled, processed, nor reported explicitly, but are
approaches and from contributing an exhaustive lumped together and represented implicitly in design
review of available examples, this contribution methods, for instance through factors of safety.
wishes to spark both an interest and the willingness, Numerous research contributions and real-world
on the part of researchers and practitioners, to con cases continue to provide new evidence that deter
tribute proactively to the continued development, ministic approaches are prone to overestimating or
refinement, promotion, and dissemination of uncer underestimating the performance of geotechnical
tainty-based and data-centric methods relying on systems because they are unable to parameterize the
CPT. Lastly, it aims to stimulate a constructive dis “real” complexity in soil-structure systems and the
cussion focusing on how some current aspects inevitable approximation in the models used to
requiring attention actually represent opportunities obtain design parameters from testing data. For
for the further refinement of methods for interpret instance, neglecting to model the spatial variability
ation and use of CPT results in characterization and of design parameters prevents the identification of
design. relevant failure modes (e.g., non-symmetric failure
While this paper advocates the increased use of modes foreseeable even in presence of symmetric
non-deterministic, data-centric methods, it wishes to foundations and axial loading conditions, see
convey the awareness that such methods largely rest Figure 1).
on the broad shoulders of the ingenious, pioneering Overconservatism and unconservatism are both to
deterministic methods which have laid the founda be avoided: while the first entails the excessive use
tions of the geotechnical discipline, and which have of resources to attain a target level of performance,
nurtured its growth and evolution. the second, more seriously, fails to achieve perform
ance and can result in failures which bring human
and economic losses.
2 A HISTORY OF CHANGE
''The quest for certainty blocks the search for mean 2.2 The present: Embracing uncertainty
ing. Uncertainty is the very condition to impel man In the context of engineering disciplines, the adjec
to unfold his powers.'' tive “non-deterministic” refers to a broad range of
- Erich Fromm methods and techniques relying most frequently –
The concepts of “determinism” and “non but not exclusively – on statistical and probability
determinism” stem from the fundamental question theory.
regarding the cause behind the occurrence of events, The development of ingenious approaches based
which has engaged philosophers in the East and on concepts such as reliability-based and perform
West since ancient times. The deterministic approach ance-based analysis and design, along with increased
prevailed over the stochastic approach in the tech computational power and efficiency, have allowed
nical community for many centuries. In the past dec civil engineering to embark on the evolution from
ades, led by the high-energy physics community and the primordial deterministic paradigm towards the
facilitated by advancements in statistical-probability non-deterministic one. The latter is characterized by
theory, enhanced data analysis capabilities, and the explicit consideration of uncertainty. The adop
increased computational power, the non tion of the non-deterministic approach allows to
deterministic approach has replaced the deterministic exploit powerful mathematical tools such as statistics
one in many scientific and technical disciplines, and and probability to model, process, and report uncer
is progressively gaining momentum in others, tainties quantitatively.
including geotechnical engineering. Moreover, While all engineering disciplines originated
improved data collection technologies and computa within the deterministic framework, geotechnical
tional power increasingly advocate the adoption of engineering has been delayed in initiating its transi
approaches making use of big data, which can fur tion, probably due to the greater inherent difficulty
ther boost the performance of non-deterministic in quantifying and modelling its uncertainties. The
approaches. While a brief account of the transition geologic environment is made up of two- or even
from the deterministic to the non-deterministic and three-phase natural materials and is thus inherently
82
total uncertainty. Prior to the introduction of the
data-centric paradigm, soil property statistics deter
mined from total uncertainty analyses could only be
applied reliably to the specific set of site conditions,
measurement techniques and procedures, and correl
ation models for which the design soil properties
were derived. The site-specificity of variability
parameters had thus constituted a long-standing limi
tation and an obstacle to the diffusion of non
deterministic approaches. A main strength of the
data-centric paradigm thus lies in the provision of
quantitative probabilistic methods to allow the inte
Figure 1. Comparison of failure modes for a surficial foot gration between site-specific data collected for
ing on uniform and spatially variable soils (from Li et al. a given project and “big indirect data”, i.e., existing
2014). data of any type collected from past stages of the
same project or past projects at the same site, neigh
more complex and less easily characterized and boring sites, or beyond.
modelled than structural materials such as steel, con
crete, or timber. Nonetheless, in the past decades The progressive surge of the data-centric paradigm
seminal research contributions and increased compu is arguably entwined with the increasing centrality of
tational power have fostered and accelerated the the Bayesian approach in geotechnical engineering.
development of non-deterministic methods. Bayesian statistical analyses, differently from “fre
The progressive shift towards non-deterministic quentist” statistical analyses, provide analysts with
methods is perhaps the most distinctive trait behind the tools to update a “prior” probability model
the development of evolutionary geotechnical design through the availability of new evidence (e.g., data
codes which are based on statistical, probabilistic, observations) to obtain a “posterior” model. This
and performance-based concepts and which are grad approach is particularly suited for the geotechnical
ually replacing the traditional deterministic design engineering discipline, in which it is often necessary
codes. Notable examples are Eurocode 7, Eurocode to supplement site- and case-specific data (or aprioris
8, the AASHTO (American Association of State tic engineering judgment) with data from other sites,
Highway and Transportation Officials) LRFD Bridge or to update characterization and design values as
Design Specifications, CHDBC (Canadian Highway new data become available. The use of Bayesian
Bridges Design Code), and JSHB (Japanese Specifi methods in the data-centric approach promises to be
cations for Highway Bridges). The adoption of the useful in circumventing the limitations of site-
non-deterministic paradigm enables the conceptual specificity of geotechnical variability estimates, and
and formal harmonization between geotechnical and in actually leveraging the availability of “indirect”
structural design, thus contributing to enhanced data from other sites (see, e.g., the outcomes of Pro
homogeneity and effectiveness in the design of soil- ject DeepGeo at http://yo-1.ct.ntust.edu.tw/jge/files/
structure systems. Embracing non-determinism for articlefiles/v16i2202106031836319463.pdf) The data-
geotechnical practice is no longer a choice in many centric paradigm arguably provides the natural evolu
parts of the world and is becoming less and less so at tion of the geotechnical discipline, provided that
a global level. Despite the above, the transition to researchers and practitioners are provided with the
the non-deterministic paradigm is far from complete. necessary simple knowledge required to correctly
interact with data (Tang & Phoon 2021).
“Data-centric” and “non-deterministic” are not
2.3 The future: Data meets knowledge
synonymous. While the data-centric approach does
The momentous global process of digital transition not imply per se resorting to non-deterministic
which characterizes our time brings unprecedented methods, the former clearly fosters the latter.
possibilities. A new “data-centric” paradigm is emer Increased data numerosity increases the applicability,
ging involving the synergy between the increased meaningfulness, and reliability of non-deterministic
availability of data, evolutionary approaches such as techniques. Many statistical and probabilistic ana
machine learning, and expert engineering judgment lyses, for instance, rely on the availability of suffi
(e.g., Phoon 2020). Such paradigm is extremely cient quantities of data, and the level of confidence
well-suited to synergize with non-deterministic of their outputs increases with increasing numerosity
approaches in defining the new identity of the geo of data samples. As the quantity and quality of data
technical discipline. are expected to continue to increase over time, fail
Comprehensive compendia of literature estimates ing to proactively confer further centrality to the
of uncertainties in soil parameters are available in data-centric approach in both geotechnical research
the geotechnical literature (e.g., Uzielli et al. 2007; and practice would be nothing short of a wasted
ISSMGE 2021). However, most estimates refer to huge opportunity.
83
2.4 In medio stat virtus The concepts of testing repeatability and accuracy
is related to measurement uncertainty. The latter is
In their illuminating treatment of geotechnical
due to equipment, procedural-operator, and random
uncertainty, Baecher & Christian (2003) opined
testing effects. Previous research (e.g., Kulhawy &
that a hybrid approach involving both determinis
Trautmann 1996) indicates that CPT testing repeat
tic and non-deterministic methods is “proper to the
ability is higher in comparison to that pertaining to
geotechnical engineering discipline”. This state
other testing methods. Peuchen & Terwindt (2014)
ment is increasingly and fully supported by
provided an extensive insight into CPT measurement
a conspicuous, growing corpus of research and
uncertainty and affirmed that piezocone testing is
real-world applications, in which some aspects of
closest to an ideal penetration tool than any other in-
a specific geotechnical analysis can be sufficiently
situ testing method. In principle, high accuracy is
well-described deterministically for practical pur
desirable in non-deterministic analyses because low
poses, while for others it is convenient to describe
epistemic measurement uncertainty contributes to
and model phenomena and variables as if they
a lower total uncertainty. High standardization
behaved, at least to some degree, in a random
allows more stringent control on testing data quality.
manner. A typical example is found in the quanti
Peuchen & Parasie (2019) highlighted the ongoing
tative modeling of spatial variability of soil prop
technical and normative activities aimed at the con
erties, which is often pursued through the
tinued pursual of high CPT testing standardization.
decomposition of data into a deterministic trend
The innumerable possible ramifications stemming
and a fluctuating, supposedly random component.
from these virtuous attributes of CPT testing do not
The hybrid approach is fully compatible with the
allow their exhaustive enumeration. Selected examples
data-centric paradigm since the availability of
are provided in the following among the many ingeni
more and better data is beneficial to both determin
ous available in the geotechnical literature.
istic and non-deterministic analyses.
4 NON-DETERMINISTIC CPT-BASED SITE

3 THE COMPETITIVE ADVANTAGES OF CPT CHARACTERIZATION
TESTING
''Exploring the unknown requires tolerating
CPT testing has long been known to surpass other uncertainty.''
in-situ testing methods with respect to the high - Brian Greene
numerosity, quasi-continuity, and accuracy of its
measurements. Moreover, CPT testing is highly Non-deterministic approaches have been proposed
standardized. Such features endow this testing for fundamental activities related to CPT-based geo
method with several relevant “competitive advan technical site characterization including soil behav
tages” in the light of the emerging data-centric, non ior classification, stratigraphic profiling, and spatial
deterministic approach to geotechnical site charac variability modeling. Selected examples are pre
terization, geotechnical design, and geohazards risk sented in the following, and readers are referred to
analysis. Cao et al. (2017) for a more comprehensive treat
The small measurement interval which is unique ment of such topic.
to CPT testing among in-situ testing methods results
in significantly higher data numerosity, i.e., in 4.1 Soil behavior classification
a higher number of measurements obtained for any
given depth interval. Sample numerosity plays CPT testing is frequently used for soil behavior clas
a decisive role in the applicability and significance sification through approaches which link the quantita
of non-deterministic analyses. In most general terms, tive measurements of the mechanical resistance to
higher data numerosity is beneficial at least in terms cone penetration to likely corresponding soil types.
of: (a) reduced statistical uncertainty; (b) improved The association between mechanical resistance and
modelling of inherent spatial variability and estima soil type is pervaded by significant uncertainties
tion of spatial averaging effects; and (c) improved stemming from in-situ state effects (e.g., cementation,
estimation of transformation uncertainty. Ultimately, overconsolidation), by the inter-site and intra-site
increased data numerosity contributes to the opti variability in compositional and mechanical proper
mization of cost-performance optimization in geo ties of soils, and by the complexity of soil behavior.
technical characterization and design by reducing the CPT-based soil behavior classification methods rou
likelihood of unconservatism and the degree of over- tinely used in geotechnical practice are overwhelm
conservatism. Data-centric approaches relying on ingly deterministic. While their use worldwide and in
machine learning also enhance their performance many different soil and site conditions attests to their
(when not outright requiring) large data sets. Phoon high performance, these methods are inherently unable
et al. (2021) opined that CPT data is the only prac to quantify explicitly the degree of uncertainty in their
tical source of data for data-driven site characteriza outputs. Methods for the non-deterministic CPT-based
tion (DDSC). soil classification have nonetheless been made
84
available in the geotechnical literature since many behavior type and stratigraphic profiling from CPT
years. Zhang & Tumay (2003) proposed the probabil testing data.
istic region estimation method. Such approach pro
vides a depth-wise profile of the probability that a set
4.3 Spatial variability modeling
of CPT measurements (more specifically, cone resist
ance and friction ratio) correspond to cohesive-, inter In-situ testing typically reveals that engineering proper
mediate-, or cohesionless-behavior soil type (see ties of geomaterials vary spatially both in the horizon
Figure 2). tal and vertical directions (most often with higher
variability in the latter case due to layering- and stress-
related effects). In absence of geological, geomorpho
logical, or geotechnical macro-discontinuities (e.g.,
layer interfaces) which should be accounted for by sep
arating data by homogeneous soil units through strati
graphic profiling and soil behavior classification, soil
properties (as parameterized, for instance, in CPT data)
can be expected to vary gradually in conformity with
the conceptual principle stated by Leibniz (1704) by
which “tout va par degrés dans la nature, et rien par
saut” (“everything occurs gradually in nature, and
nothing by jumps”). Spatial variability can never be
estimated with full precision and accuracy at any scale,
from the micro-scale to the macro-scale, due to the
finite number of tests and measurement uncertainty.
The importance of the quantitative modeling of
spatial variability is increasingly acknowledged in
research and practice as: (1) it is necessary for the
conduction of simulation-based analyses based on
Figure 2. Non-deterministic depth-wise soil behavior clas techniques such as random field modeling; (2) it
sification (Zhang & Tumay 2003). allows the spatial interpolation of in-situ testing meas
urements and of derived parameters and the quantifi
cation of uncertainties in interpolation estimates at
unsounded locations, thereby providing support in the
4.2 Stratigraphic profiling planning and integration of geotechnical investigation
campaigns and optimizing their cost-performance
Stratigraphic profiling is one of the main applications
ratio; (3) allows the quantitative estimation of the spa
of CPT testing. Such process involves the identifica
tial averaging effect, by which the effects of variabil
tion of interfaces between soil units which can be con
ity on geotechnical design can be rationally calibrated
sidered sufficiently homogeneous and/or significant for
with respect to the spatial extension of geotechnical
specific purposes. Reliable stratigraphic profiling is
structures (typically, pile foundations).
also relevant to geotechnical design. For example, in
Multiple approaches and techniques have been
CPT-based pile design, the identification of
used to the purpose of geotechnical spatial variabil
a stratigraphic interface between a layer with low bear
ity modeling. In many of these techniques, variabil
ing capacity and one with high bearing capacity and
ity is often modeled most efficiently through the
stiffness is fundamental in the definition of design pile
integration between deterministic and non
length. Cone resistance in layered soil is affected not
deterministic approaches, i.e., through the decom
only by the layer currently being traversed but also by
position of data into a trend which is expressed
the underlying and, under certain conditions, overlying
deterministically (for instance through a polynomial
layers, depending on the position of the cone relative
function) and a fluctuating component, which is
to such layers.
assumed to be a random variable, and which is
Several non-deterministic approaches for CPT-
investigated using statistical, probabilistic, geostatis
based stratigraphic profiling have been proposed.
tical, and random field theory among others.
Wang et al. (2019), for example, developed a Bayesian
Regardless of the specific technique used, CPT is
unsupervised learning approach for automatic layer
particularly suited for the modeling of spatial variabil
detection and soil classification with an explicit quanti
ity because of its small measurement interval, which
fication of uncertainty, using CPT-based parameters
allows, jointly with its high testing repeatability, more
(i.e., normalized sleeve friction FR and normalized
reliable estimation of small-scale variability in the ver
cone resistance Qt ) commonly used in the well-
tical direction with respect to other in-situ testing
established soil behavior classification system by
methods. Moreover, higher data numerosity reduces
Robertson (1990). Figure 3 provides a visual represen
statistical uncertainty in non-deterministic analyses
tation of the approach, highlighting the capability of
such as the modeling of semivariograms for the model
the method to simultaneously provide both soil
ing of spatial correlation.
85
Table 1. Perceived applicability of CPTU for deriving
soil parameters (from Robertson & Cabal 2015).
Parameter Coarse Fine
Relative density 2-3 -

State parameter 2-3 -
Young’s and shear moduli 2-3 2-4
Overconsolidation ratio 5 1
Undrained strength - 1-2
Coefficient of consolidation 3-4 2-3
Peak friction angle 2-3 4
In-situ stress ratio 5 2
Small-strain shear modulus 2-3 2-4
1-D Compressibility 2-3 2-3
Sensitivity - 2
Permeability 3-4 2-3
Note: 1=high; 2=high to moderate; 3=moderate; 4=moder

ate to low; 5=low.
Figure 3. Statistical clustering for non-deterministic strati

graphic profiling (from Wang et al. 2021).
5 NON-DETERMINISTIC CPT-BASED
GEOTECHNICAL DESIGN
Non-deterministic approaches are available for the

utilization of CPT data in geotechnical design, both
in the derivation of design parameters and in design
methods. As shown in the following selected
examples, such approaches allow improved and
more rational calibration and assessment of the per
formance of geotechnical design. Readers are
referred to Tang & Phoon (2021) for a more compre
hensive insight into geotechnical transformation
uncertainty.
Engineering models which relate in-situ and
laboratory testing measurements to geotechnical
design parameters, whether theoretical, empirical,
experimental, or a combination thereof, are always Figure 4. Source data and deterministic correlation for esti
approximate simplifications of physical soil- mating friction angle from CPT (Kulhawy & Mayne 1990).
structure systems. Geotechnical models are thus
invariably affected by bias (which relates to the In the light of the non-deterministic insight into
model’s precision) and/or dispersion (which refers CPT testing, the “degree of perceived applicability”
to the model’s accuracy). Empirical correlations could be related conceptually to the reliability of
have been made available to estimate numerous the transformation models, i.e., to the magnitude of
geotechnical design parameters from CPT testing transformation uncertainty. However, this qualita
results. Lunne et al. (1997) tabulated comparatively tive assessment can hardly be translated into quanti
the degree of applicability of a wide range of in-situ tative probabilistic formulations which can be
testing methods to the estimation of geotechnical seamlessly used in new geotechnical design codes.
parameters, assessing that CPT testing overall out Available CPT-based correlations are overwhelm
performed other methods. Robertson & Cabal ingly deterministic. They are typically obtained as
(2015) tabulated the perceived applicability of “best-fit”, central tendency models from scattered
piezocone testing to the estimation of soil param data, and are thus amenable to an “intermediate”
eters for coarse-grained and fine-grained soils degree of conservatism.
(Table 1), once again assessing the overall utility of An example of the transition to non-deterministic
CPT for geotechnical parameter estimation. CPT-based estimation of geotechnical parameters is
86
related to the estimation of effective friction angle of
sands from stress-normalized cone resistance.
Figure 5 shows the data used by Kulhawy & Mayne
(1990) to develop their well-known deterministic
model
While goodness-of-fit statistics (typically, the coeffi

cient of determination R2 ) or descriptive second-
moment dispersion statistics (e.g., standard deviation,
variance) are provided together with the regression
model in best-practice cases such as this one, these
are in principle not sufficient to allow the model-
based estimation of properties for different levels of
conservatism. Moreover, the statistical distribution of
data around the best-fit model cannot be character
ized statistically beyond second-moment description.
Uzielli & Mayne (2019) proposed a probabilistic
transformation model using a high-quality database
comprising CPT and triaxial tests. The logarithmic Figure 5. Comparison of probabilistic (Uzielli & Mayne
formulation 2019) and deterministic (Kulhawy & Mayne 1990) models
for the estimation of effective friction angle from stress-
normalized cone resistance.
founded on data from multiple sites. While geotech

has the same ease of practical applicability as the nical engineers are trained to operate the transition
previous deterministic model while allowing the esti between “indirect” data (i.e., from other sites) to
mation of 0 for varying levels of conservatism (i.e., “direct” (i.e., local) data by applying transformation
for different probabilities of exceedance Pexc ) by models and assessing the results critically using
varying model coefficients p1 and p2 as shown in engineering judgment, non-deterministic
Table 2. Figure 5 plots the curves corresponding to approaches based on the data-centric paradigm
probabilities of exceedance 0.05, 0.50, 0.75, and allow an improved synergy between indirect and
0.95 of effective friction angle in comparison with direct data. Through Bayesian approaches relying
the deterministic model by Kulhawy & Mayne on the data-centric paradigm, it is possible to quan
(1990). While the performance of the deterministic tify the reliability of indirect data with respect to
methods is overall comparable to the probabilistic direct data and to provide a “hybrid” transformation
method for a “median” probability of exceedance, it model which relies on both datasets, and in which
is apparent that the latter allows the fuller exploit the weight of the “indirect” model is determined on
ation of the experimental dataset. It is also apparent the basis of objective criteria such as the “similar
that the deterministic model does not refer uniformly ity” between indirect sites and the target site, e.g.,
to a probability of exceedance for increasing values in terms of soil types. For example, Ching et al.
of qt1 . Thus, the degree of conservatism in the deter (2021) used a hierarchical Bayesian modelling
ministic model is not established univocally. framework to apply hybridization to the estimation
The transition to a data-centric paradigm can fur of undrained strength of clays from CPT testing.
ther enhance the quality and utility of new- Figure 6 illustrates indirect site-specific models as
generation geotechnical transformation models. Site- well as the hybrid model for the target site-specific
specific data are typically sparse and incomplete. model. Ellipses represent the uncertainties associ
The geotechnical discipline has adapted to this fact ated with the respective models, which are quanti
by providing transformation models which are fied explicitly.
An increasing set of non-deterministic CPT-based
design methods are also available in the geotechnical
Table 2. Coefficients of the logarithmic CPT-based model literature. Cai et al. (2021) provided a framework for
for the estimation of friction angle for various probabilities the CPT-based calculation of axial pile capacity in
of exceedance (Uzielli & Mayne 2019).
spatially variable soil. The study relied on the simu
Pexc 0.95 0.75 0.50 0.05 lation of CPT profiles through random field model
ing and on the statistical-based calculation of pile
p1 4.27 4.00 3.81 1.70 capacity for varying target levels of conservatism.
p2 18.1 20.3 22.2 35.1 Figure 7 provides a comparative visual representa
tion of simulated profiles of cone resistance at preset
87
spatial separation distances within the design site for
smaller and larger horizontal correlation distances.
Random field modeling also allows to refer expli
citly to the scale of fluctuation, thus allowing the
quantification of the spatial averaging effect by
which the influence of the variability in soil proper
ties is smoothed out along the shaft. The rigorous
inclusion of the spatial averaging effect contributes
to calibrating the effect of inherent soil variability
and to reconduct rationally the degree of conserva
tism in design to its target level. CPT testing data are
particularly suited for simulation purposes because
of the higher reliability in the estimation of vertical
spatial variability parameters stemming from the
quasi-continuity of measurements in comparison to
other in-situ testing metods.
Figure 7. Simulation of horizontal site-scale variability of

cone tip resistance for: (a) lower; and (b) higher horizontal
correlation distance (adapted from Cai et al. 2021).
6 NON-DETERMINISTIC CPT-BASED
GEORISK ANALYSIS
In-situ testing results are frequently used in the

increasingly relevant contributions of geotechnical
engineering to geohazards risk analysis. Methods
have been made available for various hazards and
related phenomena including the prediction of
triggering of seismic liquefaction, seismically
induced soil densification, lateral spreading, and
slope stability, among others. The prediction of
seismic liquefaction triggering has arguably
received most attention. Numerous simplified
liquefaction evaluation approaches based on the
results of in-situ tests (SPT, CPT, DMT, seismic
testing) have been made available over the past
decades. Among these, CPT-based methods have
been shown to provide significant advantages
over others due to: (a) the capability to detect
thin layers; (b) the repeatability of measurements;
(c) the relative speed and economy.
As in all ramifications of the geotechnical discip
line, deterministic methods for liquefaction assess
ment have preceded non-deterministic methods
chronologically and are still today predominant
even though literature findings increasingly indicate
that non-deterministic methods provide multiple
significant advantages. First, they allow users to
define liquefaction boundary curves for target
levels of conservatism, thereby contributing to cost-
performance optimization with respect to the vul
nerability and value of elements at risk in a specific
area. Second, they explicitly include uncertainties
in the estimation of the parameters and models used
to estimate seismic demand and soil resistance
Figure 6. Hybridization of direct and indirect data-driven parameters. Third, they allow a more realistic and
models for undrained strength from CPT: (a) target site and physics-based modelling and treatment of geotech
generic data; (b) resulting site-specific and target site nical phenomena. The latter point can be well
models (Ching et al. 2021). exemplified by the following.
88
A vast bulk of research and experimental evi
dence indicates that seismic liquefaction typic
ally occurs in loose or medium-density, saturated
cohesionless soils, while high-plasticity, fine
grained soils are inherently “non-liquefiable” due
to the relevance of the cohesive component of
shear strength. A fundamental step in the imple
mentation of simplified CPT-based methods lies
in the assessment of inherent soil liquefiability.
Such methods are developed from databases per
taining to soils which are expected to be inher
ently liquefiable based on prior knowledge, and
are thus applicable solely to such soils, i.e., the
calculation of the cyclic resistance ratio CRR,
which parameterizes the cyclic resistance of
soils to liquefaction, is pursued solely at depths
corresponding to soils preliminarily assessed to
be liquefiable. The applicability of simplified Figure 8. Probability of liquefaction susceptibility as
methods must be assessed depth-wise, as strati a function of the soil behavior type index (from Maurer
graphic profiles can be very heterogeneous in et al. 2017).
terms of soil types and can include both liquefi
able and non-liquefiable soil.
All simplified CPT-based liquefaction evaluation
methods include quantitative approaches to the
depth-wise assessment of the liquefiability or non
liquefiability of soils. Such assessment typically
relies on the definition of deterministic threshold
values of the soil behavior classification index Ic
(typically varying between 2.4 to 2.6 among
methods) which provide a strict boundary between
liquefiable and non-liquefiable layers. The assess
ment resulting from the adoption of deterministic
thresholds is invariably binary, i.e., a set of CPT
measurements corresponds to either a “fully liquefi
able” or “fully non-liquefiable” measured soil
volume interval. The hypothesis of an abrupt discon
tinuity between non-liquefiability and full liquefia
bility is hardly compatible with well-established
geotechnical knowledge and evidence by which
liquefiability can be expected to decrease gradually
with increasing percentage of fines in a soil. Inciden
tally, the non-binary effect of varying fines content is
already included in simplified methods through the
calculation of fine-sand equivalent stress-normalized Figure 9. Probabilistic liquefaction boundaries (Schmidt &
cone resistance (usually denoted by qc1N;cs ), which Moss 2021).
takes continuous (albeit usually inferiorly and super
iorly bounded), values, but does not intervene in the
preliminary assessment of liquefiability. Moreover, methods, could provide a useful means to quantify
calculated values of Ic are pervaded by uncertainty liquefiability as a continuous (as opposed to binary)
stemming from measurement depth offset effects, variable and to partially circumvent the limitations
layering effects, and uncertainty in measurements of which are inherent to deterministic approaches.
qc and fs . The progressive variation of the degree of “Fully” non-deterministic CPT-based approaches,
liquefiability should be accounted for quantitatively. mostly relying on Bayesian methods for the predic
Non-deterministic approaches to susceptibility tion of liquefaction triggering, are available in the
evaluation can contribute to improving deterministic geotechnical literature and are being increasingly
methods. Maurer et al. (2017) proposed deterministic used in practice. In comparison with deterministic
and probabilistic correlations relating Ic to liquefac approaches which provide a single liquefaction
tion susceptibility as defined by four well-established boundary curve, these allow the selection of lique
published criteria. Figure 8 shows the respective faction boundaries for varying degrees of conserva
liquefaction susceptibility probability curves. Such tism while accounting explicitly for uncertainties.
curves, if duly integrated into existing deterministic Figure 9 provides an example of probabilistic
89
liquefaction boundaries for varying probability possible to assess where estimation uncertainty
levels as calculated from normalized cone resistance exceeds thresholds of tolerability/acceptability, thus
using the approach proposed by Schmidt & Moss providing direct indications for the conduction of
(2021). This approach addresses and processes supplementary tests. The visual representation of the
uncertainties in the parameters used in the estimation spatial variability of the interpolation estimates of
of cyclic resistance and cyclic stress, as well as the factor of safety allows rational decision-making
uncertainties stemming from intra-event correlation regarding possible supplementary investigations.
in loading variables and differences in event sample
sizes.
Non-deterministic methods can also be used to 7 TURNING CAUTIONS INTO
spatialize the outputs of deterministic simplified OPPORTUNITIES
CPT-based methods. For example, Guan & Wang
(2022) developed a probabilistic method for charac The adoption and implementation of methods such
terizing the horizontal spatial distribution of the as those presented in the previous sections should
deterministically calculated cyclic resistance ratio, never occur in absence of engineering judgment and
cyclic stress ratio, and factor of safety against lique geotechnical knowledge. Cautions should be adopted
faction in data from CPT soundings using Bayesian at all stages of CPT-based analyses to ensure the
compressive sampling and Monte Carlo simulation. geotechnical significance and formal correctness of
The method relies on the generation of random field the quantitative procedures relying on uncertainty
samples and allows the modeling and quantification modeling. This paper argues that current cautions
of spatial variability, interpolation uncertainty, and can be successfully addressed through synergies
model uncertainty. Example outputs of the methods between research and practice.
include spatial distributions of statistics of the factor
of safety resulting from the interpolation of deter
7.1 Interpretation in layered soils.
ministic values obtained at 6 CPT sounding verticals.
Figure 10a refers to the mean values of the factor of Previous studies explored the problem of cone pene
safety calculated from 10,000 realizations of random tration in layered soils analytically (e.g., Yu 2006) or
fields, while Figure 10b plots the spatial distribution experimentally, either in calibration chambers (e.g.,
of the coefficient of variation, which provides Tehrani et al. 2017) or centrifuges (e.g., Mo et al.
a quantitative measure of the level of uncertainty in 2017). Some of these studies define “sensing dis
the spatialized values of the factor of safety. tance” as the distance from a layer interface at which
The availability of quantitative measures of uncer the cone resistance first starts changing because the
tainty associated with interpolated values provides cone is approaching it, and “development distance”
a fundamental advantage over deterministic charac the distance from a layer interface at which the cone
terization methods, at least because: (1) the param resistance ceases to be affected by the interface as it
eter is defined probabilistically for direct use in moves away from it.
reliability-based methods; and (2) it becomes A further complication is given by the fact that
the three fundamental CPT parameters pertain to
different depths and to different volumes of soil.
This fact stems from the physical geometry of the
cone and by the inherent differences between the
volumes of soils and the physical phenomena
affecting cone resistance, sleeve friction, and pore
pressure. These three measurements, involving
different volumes of soil during penetration. The
oretical modeling of cone penetration has shown
that such extent may depend on cone size, geom
etry, and soil type and stiffness. Thus, the vol
umes of soil involved in cone tip resistance and
sleeve friction measurements are variable. The
existence of a measurement depth offset between
cone resistance, sleeve friction, and pore pressure
is well-known. Approaches relying on statistical
techniques such as cross-correlation have been
proposed and are implemented in widely used
commercial CPT interpretation software.
While CPT measurements ultimately reflect the
Figure 10. Example statistical outputs of the horizontal true mechanical behavior of the in-situ soil volume
spatialization of the factor of safety against liquefaction: in terms of its resistance to cone penetration, they do
(a) mean; (b) coefficient of variation (from Guan & Yang not provide fully reliable information if data are not
2022). corrected for layering effects, and measurement
90
equipment and procedural control in testing data.
Geotechnical uncertainty models are typically addi
tive (e.g., Phoon & Kulhawy 1999). The magnitude
of individual uncertainty components can thus be
estimated in reverse-mode, i.e., by subtracting
known uncertainty terms from the total uncertainty
which is observed. Consistent approaches to the
quantification of measurement uncertainty would
result in the availability of more reliable parameter
ization of measurement uncertainty, leading to more
confident subtractive-mode implementation of uncer
tainty models to estimate aleatory and other epi
stemic uncertainties from total uncertainty. To this
purpose, Peuchen & Terwindt (2014) advocated the
Figure 11. Schematic of cone penetration in multilayered development of a shared, less ambiguous glossary
soils: strong soil embedded in weak soils (Mo et al. 2017). (e.g., “repeatability” is strictly not synonymous to
“accuracy”) and the compilation of a reference
measurement uncertainty dataset from real cases (as
offset prior to being used for stratigraphic profiling, opposed to “ideal” cases referring to unrealistically
soil classification, and in the calculation of geotech controlled conditions).
nical design parameters. This is because depth-wise
measured values do not correspond to the true values
at the same depth. CPT-based approaches relying on 7.3 Upgrading of geotechnical correlations
the depth-wise processing of measurements should While new, non-deterministic transformation models
in principle be preceded by a pre-processing of data yielding design properties from CPT data such as the
to account for layering effects and for the inherent ones presented in this paper are progressively being
depth offset between measurements. If such pre made available, the vast majority of geotechnical
processing is not conducted, these approaches will correlations (including CPT-based correlations)
be pervaded by additional epistemic uncertainty, remain deterministic. The full migration of the geo
however sophisticated they may be. The magnitude technical discipline to the non-deterministic para
of the epistemic uncertainty related to layering digm cannot prescind from the upgrading of
effects and depth offset tend to increase with strati geotechnical correlations to formats which are com
graphic complexity. patible with new paradigms. This could be aided by,
Despite the inherent complexity in physical for instance, through: (a) specific and systematic
phenomena involving cone penetration in layered focus of new research on the development of non
soils, studies such as those mentioned above have deterministic correlations; and (b) a collective effort
provided ingenious analytical models for the esti on the part geotechnical researchers and practitioners
mation of sensing distance and development dis to contribute existing data for the non-deterministic
tance and for the correction of raw measurement reformulation of existing literature correlations.
data. The development of efficient algorithms for
the translation of the results of these studies into
software applications provides a stimulating chal 8 CONCLUDING REMARKS
lenge and an opportunity for considerable
improvement in the significance of CPT data This paper has hopefully provided a meaningful
interpretation for routine implementation in non insight into the possible benefits of non-deterministic
deterministic methods for geotechnical site char methods for the interpretation of CPT data and its
acterization and design. Operating in the light of utilization in many and diverse aspects of geotech
the non-deterministic paradigm provides an ideal nical engineering. It has also aimed to highlight the
scenario as uncertainties in the “true” values of inherent competitive advantage of CPT over other
CPT measurements in layered profiles can be in-situ testing methods in meeting the demands of
expressed explicitly. the new non-deterministic, data-centric paradigm of
the geotechnical discipline.
7.2 Improved quantification of measurement The progressive migration of geotechnical design
uncertainty codes towards non-deterministic formats and the
ever-strengthening global digital transition indicates
The separation of the components of total uncer that embracing and implementing such paradigm in
tainty may prove to be a complex process depending practice is no longer a choice for the geotechnical
on the quality and quantity of data from a specific community at large, but rather a beneficial, conveni
case. The relative contribution of total uncertainty is ent obligation. Providing students, practitioners, and
also highly case-specific, and depends on site condi researchers with sufficient theoretical knowledge
tions and inherent spatial variability, and degree of and practical tools to understand and implement
91
non-deterministic methods and approaches is REFERENCES
a necessity. ISSMGE TC304 “Engineering Practice
of Risk Assessment & Management” recently Baecher, G.B. & Christian, J.T. 2003. Reliability and statis
exerted a formidable effort resulting in a state-of-the tics in geotechnical engineering. New York: John Wiley
-art report containing a conspicuous bulk of quantita & Sons.
tive estimates of inherent variability parameters as Box, G. E. P., Luceño, A. & del Carmen Paniagua-
Quiñones, M. 2009. Statistical Control By Monitoring
well as statistics for geotechnical design model fac
and Adjustment. John Wiley & Sons.
tors, random field parameters of geotechnical proper Cai, Y., Bransby, F., Gaudin, C. & Uzielli, M. 2021.
ties, statistics for transformation uncertainties, and A framework for the design of vertically loaded piles in
characteristic values of geotechnical parameters for spatially variable soil. Computers and Geotechnics 134.
use in non-deterministic design methods. DOI: 10.1016/j.compgeo.2021.104140
To further exploit the opportunities offered by the Cao, Z., Wang, Y. & Li, D. 2017. Probabilistic approaches
global digital revolution and improved measurement for geotechnical site characterization and slope stability
technologies, and to meet the requirements of evolu analysis. Springer. ISBN: 978-3-662-52914-0.
tionary design codes, the geotechnical community Ching, J., Wu, S. & Phoon, K.K. 2021. Constructing
quasi-site-specific multivariate probability distribution
must persist in the transition towards the data-centric
using hierarchical Bayesian model. Journal of Engineer
approach, which is fully compatible with the non ing Mechanics, ASCE. DOI: 10.1061/(ASCE)EM.1943
deterministic paradigm. To facilitate and stimulate 7889.0001964.
this transition, new geotechnical databases providing Einstein, A. 1921. “Geometry and experience”. Address to
high-quality, validated information are made publicly the Prussian Academy of Sciences in Berlin, January
available. An example database including CPT testing 27th, 1921.
outputs is the ISSMGE TC304 database – 304dB at Guan, Z. & Wang, Y. 2022. CPT-based probabilistic lique
http://140.112.12.21/issmge/tc304.htm. This paper faction assessment considering soil spatial variability,
wishes to encourage the CPT community to actively interpolation uncertainty and model uncertainty. Com
puters & Geotechnics 141. DOI: 10.1016/j.
contribute to this database.
compgeo.2021.104504.
The surge of data-driven approaches does not
ISSMGE – Technical Committee TC304 “Engineering
undermine the central role of geotechnical expertise Practice. 2021. State-of-the-art review of inherent vari
and engineering judgment. Rather, new engineering ability and uncertainty in geotechnical properties and
paradigms and technological evolutions now require models.
that geotechnical engineering judgment also rely on Kulhawy, F.H. & Mayne, P.W. 1990. Manual on estimating
skills allowing the comprehension of uncertainties and soil properties for foundation design. EPRI Report EL
the enhanced analysis of the data which geotechnical 6800, Electric Power Research Institute.
testing provides. A modern geotechnical engineering Kulhawy, F.H. & Trautmann, C.H. 1996. Estimation of in-
education would thus greatly benefit from structured situ test uncertainty. In C.D. Shackleford, P.P. Nelson
training in statistical and probabilistic methods, specif and M.J.S. Roth (eds.), Uncertainty in the Geologic
Environment: From Theory to Practice, ASCE Geotech
ically including Bayesian approaches, along with data
nical Special Publication No. 58: 269–286.
analysis and machine learning methods. Current practi Leibniz, G.W. 1704. Nouveaux éssais IV, 16, 12.
tioners could receive these contents through high- Li, J., Tian, Y. & Cassidy, M.J. 2014. Failure mechanism
quality professional development initiatives. On the and bearing capacity of footings buried at various
academic side, geotechnical researchers focusing on depths in spatially random soil J. Geotech. Geoenviron.
non-deterministic and data-centric methods should Eng., ASCE. DOI: 10.1061/(ASCE)GT.1943
continue striving to make their promising findings 5606.0001219.
increasingly accessible through open-source sharing, Maurer, B.W., Green, R.A., van Ballegooy, S. &
free software, and on an improved communication Wotherspoon, L. 2017. Assessing liquefaction susceptibil
ity using the CPT soil behavior type index. Proceedings
skills to allow a more immediate comprehension of
of the Third International Conference on Performance-
new findings on the part of the geotechnical commu Based Design in Earthquake Geotechnical Engineering –
nity at large. ISSMGE-related initiatives focusing on PBDIII, Vancouver, July 16-19.
these topics could be disseminated through structured Mo, P.-Q., Marshall, A.M. & Yu, H.-S. 2017. Interpretation
protocols with academia and professional bodies. of cone penetration test data in layered soils using
The translation of non-deterministic mathematical cavity expansion analysis. J. Geotech. Geoenviron.
frameworks of varying complexity and sophistica Eng., ASCE. DOI: 10.1061/(ASCE)GT.1943
tion into practically utilizable executables for the 5606.0001577.
interpretation of CPT data may further foster the Peuchen, J. & Terwindt, J. 2014. Introduction to CPT
accuracy. Proceedings of the 3rd International Sympo
transition of the geotechnical community towards
sium on Cone Penetration Testing, Las Vegas.
cost-performance optimization in geotechnical site
Peuchen, J. & Parasie, N. 2019. Challenges for CPT
characterization, design, and in the analysis and accuracy classes. Proceedings of the XVII ECSMGE
management of georisks, in full accordance with cur 2019, Geotechnical Engineering foundation of the
rent global trends aimed at sustainability, circularity, future, Reykjavik. ISBN 978-9935-9436-1-3. DOI:
and community participation. 10.32075/17ECSMGE-2019-0049
92
Phoon, K.-K. & Kulhawy, F.W. 1999. Characterisation of J. Geotech. Geoenviron. Eng., ASCE. DOI: 10.1061/
geotechnical variability. Canadian Geotechnical Journal (ASCE)GT.1943-5606.0001809.
36: 612–624. Uzielli, M., Lacasse, S., Nadim, F. & Phoon, K.-K.
Phoon, K.-K. 2020. The Story of Statistics in Geotechnical 2007. Soil variability analysis for geotechnical prac
Engineering. Georisk: Assessment and Management of tice. Keynote paper, In T.S. Tan, K.K. Phoon, D.
Risk for Engineered Systems and Geohazards. DOI: W. Hight & S. Leroueil (eds.), Proceedings of the
10.1080/17499518.2019.1700423. 2nd International Workshop on Characterisation and
Phoon, K.-K., Ching, J. & Shuku, T. 2021. Challenges in Engineering Properties of Natural Soils. Singapore,
data-driven site characterization. Georisk: Assessment November 29 – December 1, 2006. The Netherlands:
and Management of Risk for Engineered Systems and Taylor & Francis.
Geohazards. DOI: 10.1080/17499518.2021.1896005. Uzielli, M., and P. W. Mayne. 2019. “Probabilistic assign
Robertson, P.K. 1990. Soil classification using the cone ment of effective friction angles of sands and silty sands
penetration test. Canadian Geotechnical Journal from CPT using quantile regression.” Georisk: Assess
27:151–158. ment and Management of Risk for Engineered Systems
Robertson, P.K. & Cabal, K.L. 2015. Guide to Cone Pene and Geohazards 13(4): 271–275. DOI:10.1080/
tration Testing for Geotechnical Engineering – 6th Edi 17499518.2019. 1663388
tion. Gregg Drilling & Testing, Inc. Yu, H.-S. 2006. The first James K. Mitchell lecture in situ
Schmidt, J. & Moss, R. 2021. Bayesian hierarchical and soil testing: From mechanics to interpretation. Geomech.
measurement uncertainty model building for liquefac Geoeng.: An Int. J., 1(3),165–195.
tion triggering assessment. Computers & Geotechnics Zhang, Z. & Tumay, M.T. 2003. Non-traditional
132, 103963. DOI: 10.1016/j.compgeo.2020.103963. approaches in soil classification derived from the
Tang, C. & Phoon, K.-K. 2021. Model uncertainties in cone penetration test. In G.A. Fenton & E.
foundation design. CRC Press, Boca Raton. DOI: H. Vanmarcke (eds.), Probabilistic Site Characterisa
10.1201/9780429024993. tion at the National Geotechnical Experimentation
Tehrani, F.S., Arshad, M.I., Prezzi, M. & Salgado, R. 2017. Sites, ASCE Geotechnical Special Publication
Physical modeling of cone penetration in layered sand. No. 121: 101–149.
93
© 2022 the Author(s), ISBN 978-1-032-31259-0
Combined use of CPT & DMT: background, current trends and ongoing
developments
P. Monaco
University of L’Aquila, L’Aquila, Italy
ABSTRACT: This paper aims to introduce and stimulate the discussion on the use of CPT & DMT and on
the potential of their mutual integration for enhanced site characterization and geotechnical design. It provides
a brief overview of background information, current trends and ongoing developments on the combined use
of CPT & DMT testing. Specific issues addressed in the paper include: (a) CPT-DMT comparisons and inter
relationships; (b) derivation of soil parameters from combined CPT & DMT data; (c) applications based on
combined CPT & DMT data.
1 INTRODUCTION advantages of full-displacement penetrometer probes

with downhole geophysics (seismic piezocone
The use of in-situ testing for geotechnical site charac SCPTu, seismic dilatometer SDMT), capable to pro
terization has continued to expand over the past few vide information on soil behavior from the small- to
decades, gaining an increasing role over the traditional the large-strain range.
practice based on drilling and sampling for laboratory A single SCPTu sounding provides up to five
testing. Often today in-situ tests represent the major independent measurements: the basic CPTu meas
part of a site investigation, particularly in geomaterials urements (cone resistance qt, sleeve friction resist
that are difficult to sample and test using conventional ance fs, pore pressure u2), the shear wave velocity
methods (e.g., sands, tailings, semi-liquid soils). VS, and information on soil consolidation / perme
Mayne et al. (2009) summarized the key advantages of ability parameters if dissipation tests are performed
in-situ tests: (a) they can be done relatively quickly as by monitoring the u2 decay with time (e.g., time to
compared with laboratory tests, (b) results are available reach 50% degree of consolidation, t50). A single
immediately, (c) large numbers of data are obtained, SDMT sounding provides a comparable number of
and (d) vertical and lateral variability can be assessed independent measurements: the basic DMT pressure
over the site. measurements (A, B and optional C corrected to p0,
In this context, the cone/piezocone penetration test p1, p2 respectively), the shear wave velocity VS, as
(CPT/CPTu) and the flat dilatometer test (DMT) have well as the compression wave velocity VP when
proven to be particularly efficient and expedient for using the SPDMT configuration equipped with two
routine site investigations. In fact: (a) both are direct- pairs of receivers for S- and P-wave (Amoroso et al.
push probes, which do not need a borehole for 2020), and information on soil consolidation / per
advancement, (b) multiple measurements are collected meability parameters if dissipation tests by monitor
from a single sounding, (c) the data are obtained digi ing the A-pressure decay with time are performed
tally and can be accessed for use immediately after (e.g., reference time in the A-decay curve, tflex).
completion of the sounding or even in real-time during For these reasons, Mayne et al. (2009) suggested
advancement of the probe, (d) their instrumental accur that the SCPTu and SDMT direct-push tests should
acy is high, unlike “older” techniques such as the serve as the basis for the minimum required level of
standard penetration test (SPT), (e) technological effort in routine and daily site investigation practices
upgrades, such as additional sensors or full automation by the profession, in order that adequate amounts
(e.g., Medusa DMT, Marchetti et al. 2019), can be and sufficiently different types of data are collected
easily implemented. for a given project. These data are complemented
The use of direct-push in-situ tests providing with sampling and laboratory testing.
multi-measurements was emphasized by Mayne et al. The CPT and the DMT have in common the same
(2009) as a more efficient approach to geotechnical equipment for insertion (push rig, rods) and a similar
site characterization. Particularly beneficial and versa range of soil type applications (clays, silts, sands).
tile is their “hybrid” configuration that combines the Both tests are utilized in site investigations to obtain
DOI: 10.1201/9781003308829-7
94
information on stratigraphy, soil types and param normalized friction ratio Fr and Soil Behavior Type
eters to be used with general or dedicated design (SBT) Index Ic, defined as follows:
methods. However, each test has advantages and
limitations and may prove more adequate in specific
applications. E.g., the CPT is commonly used for
pile design and liquefaction assessment, the DMT
generally provides good predictions of settlements of
shallow foundations.
In common practice, the CPT is often the pre
ferred primary in-situ test, supported by a much
larger consolidated use. On the other hand, the DMT
can provide distinctive contributions in a routine site
investigation: (1) information on stress history,
which has a dominant influence on soil behavior; (2)
being an in-situ pressure-displacement test, the DMT
results are more closely related to “working strain”
soil stiffness than penetration tests.
Some questions that may arise are then: are the
two in-situ tests equivalent? Is one test (CPT or
DMT alone) self-standing and sufficient for exhaust
ive geotechnical site characterization? Are there any
advantages in the combined use of CPT & DMT?
One notable emerging trend (Marchetti 2015, where p0 = corrected first DMT pressure reading, p1
Marchetti & Monaco 2018) is the increasing diffu = corrected second DMT pressure reading, qt = cor
sion in site investigation practice of a “multi rected cone resistance, fs = sleeve friction resistance,
parameter / multi-test” approach, based on the com u0 = pre-insertion in situ equilibrium pore pressure,
bination of CPT & DMT. This approach benefits, σv0 = in situ total vertical stress, σ’v0 = in situ effect
besides from the multiple measurements provided by ive vertical stress.
a single CPT or DMT sounding, also from the com Robertson (2009a) explained that in most of the
bination of the different measurements obtained available published records the DMT KD is normalized
from nearby CPT and DMT soundings. Examples of by σ’v0 as in the original formulation (Eq. 2) proposed
application of a multi-parameter / multi-test approach by Marchetti (1980). Therefore, the CPT cone resist
based on the combined use of CPT & DMT are illus ance Qt1 was normalized by σ’v0 in a consistent
trated in this paper, aiming to stimulate the discussion manner, i.e., assuming the exponent for stress normal
on the potential of the mutual integration of these two ization n = 1, as originally proposed by Robertson
in-situ tests for enhanced site characterization and (1990), although currently used updated formulations
geotechnical design. The terms ‘CPT’ and ‘DMT’ are (e.g., Robertson 2009b) refer to a normalized cone
herein intended in a general sense, referring to any resistance Qtn that uses a variable n:
equipment configurations (i.e., CPT(u), SCPT(u);
DMT, SDMT, Medusa (S)DMT).
2 CPT-DMT CORRELATIONS
Interrelationships between CPT & DMT parameters where pa = atmospheric pressure in same units as qt
can be helpful to expand and improve correlations and σv, n = stress exponent that varies with SBT Ic
and applications by using existing experience and (for n = 1, Qtn = Qt1).
databases from one test and extrapolating to the In addition, stress normalization for KD is a debated
other (Robertson 2012). issue. Robertson (2009a, 2015) suggested that a more
Robertson (2009a) reviewed published records of complex normalization for KD, similar to Qtn, would
data from nearby CPT and DMT soundings, as well likely be more appropriate, especially in sands, and
as existing correlations for geotechnical parameters, future CPT-DMT correlations could use more suitable
in an effort to identify possible intercorrelations. The normalized parameters. However, for typical stress
key in this approach was the recognition that, since levels in geotechnical engineering of about 65-200
the main DMT interpreted parameters are normal kPa (i.e., ≈ 4-20 m) the normalization method has
ized, they should be correlated with normalized CPT little influence on the normalized parameters, hence
parameters. Correlations were tentatively established updated CPT-DMT correlations using a different nor
between the three “intermediate” DMT parameters, malization may not change significantly. On the other
i.e., the material index ID, the horizontal stress index hand, Marchetti (2015) remarked that a linear increase
KD and the dilatometer modulus ED (Marchetti with depth of the DMT p0 (i.e., a nearly constant KD
1980), and the CPT normalized cone resistance Qt1, calculated using n = 1) is observed to large depths at
95
various test sites. In contrast, the increase of the CPT
qt with depth is generally less than linear, correspond
ing to n < 1. Such a different trend was attributed by
Marchetti (2015) to the fact that the DMT blade, The (smoother) contours of DMT KD correspond
having a rectangular cross section with a width/thick ing to Eq. 11 are also plotted in Figure 1 (red dashed
ness ratio ≈ 6, is less affected by arching than lines) on the CPT normalized SBT Qt1 – Fr chart.
a conical probe, resulting in a more linear trend of p0
with depth which justifies n = 1 for KD.
The preliminary set of average CPT-DMT correl
ations proposed by Robertson (2009a) is:
When Ic ≤ 2.60 (i.e., for sand-like soils):
When Ic > 2.60 (i.e., for clay-like soils):
Figure 1. Contours of DMT KD, ID on the CPT normalized

SBT Qt1 – Fr chart for young uncemented soils (from
Robertson 2017).
where the constant β varies with soil sensitivity

(0.30 < β < 0.7), with an average value β = 0.3. To stimulate a discussion on this topic, it is worth
The correlations for ID, KD (Eqs 8, 9a, 9b) recalling what was observed by Robertson (2009a,
depend on soil type, reflected in the CPT index Ic, 2015), i.e., the proposed CPT-DMT correlations are
which is a function of both Qt1 and Fr. These correl approximate and influenced by variations in in-situ
ations are shown in Figure 1 in the form of contours stress state, soil density, stress and strain history, age,
of DMT ID, KD on the CPT normalized SBT chart cementation, soil sensitivity. These correlations are
Qt1 – Fr. unlikely to be unique for all soils, but they may form
The relationship between DMT ID and CPT Ic a framework for possible future refinements. More
(Eq. 8) was established considering that both param over, they may provide further insight into possible
eters are used to identify soil type. It is noted that ID, future correlations for the DMT with other geotech
likewise Ic, is a parameter reflecting the mechanical nical parameters and design applications, since the
soil behavior, not a soil classification index based on CPT has a somewhat more extensive theoretical back
real grain size distribution (Marchetti 1980). Robertson ground compared to the DMT, as well as a larger
& Wride (1998) suggested that Ic = 2.60 can be database of documented case histories for certain
assumed as an approximate boundary between sand- applications (e.g., liquefaction assessment). Marchetti
like and clay-like soils. Based on Eq. 8, Ic = 2.60 cor (2011) recognized that procedures helpful for extract
responds roughly to ID ≈ 1. In a general sense, CPT ing as much information as possible from the field
and DMT results are drained in sand-like soils (Ic ≤ data represent a precious contribution since soil infor
2.60, ID > 1) and undrained in clay-like soils (Ic > mation is vital for the determination of reliable soil
2.60, ID < 1). The correlation for KD (Eqs 9a, 9b) can parameters for design, but costly to obtain. Correl
be sensitive to the cut-off Ic = 2.60 when CPT data fall ations interconnecting CPT & DMT are useful in that
close to the boundary between clay-like and sand-like translation formulas permit: (1) the use of interpret
soils. DMT results in the transition region of silt- ation methods or charts developed for one test with
mixture soils may also be influenced by partial the results of the other test, (2) converting a database
drainage. available for one test to a database for the other test.
Robertson (2015), based on additional pairs of However, Marchetti (2011) stressed that intrinsic
CPT & DMT (e.g., Togliani et al. 2015), suggested limits exist to the accuracy of the CPT-DMT transla
a more simplified link between CPT Qt1 and DMT tions, which, despite further refinements, are bound to
KD, ID for young uncemented soils (i.e., soils with remain of an approximate nature. These limits are
little or no microstructure): mostly due to the different sensitivity to stress history
96
of the representative normalized CPT and DMT the CPTu can be utilized to duplicate equivalent
parameters, as will be discussed in the following. DMT pressures p0 and p1, and vice versa, according
A different theoretical approach for establishing to the relationships:
interrelationships between the CPT and the DMT in
soft clays has been explored by Mayne (2016),
Ouyang & Mayne (2017, 2018) and earlier studies,
based on a CPT-DMT compiled database in clays
that progressively expanded over the years.
Eq. 14 is in agreement with earlier studies

(Mayne & Bachus 1989, Mayne 2006) that found the
p0 from DMT very similar to the pore pressures u2
measured by CPTu in intact clays and clayey silts.
Figure 2 shows a good agreement between the pres
sures p0, p1 measured by DMT and their CPTu
equivalent p0, p1 estimated by Eqs 14 and 15, based
on 49 paired CPTu-DMT data sets in a variety of
natural soft to firm clays and silts worldwide, having
an overconsolidation ratio OCR between 1 to 2.5.
The approach proposed by Ouyang & Mayne
(2017, 2018) is valid only for undrained penetration
conditions, which are generally accepted to occur for
both CPTu and DMT in low permeability clays at
a standard penetration rate of 20 mm/s. The
approach could be invalidated in intermediate per
meability soils, such as silts, where drainage condi
tions may be different. The CPTu readings qt and u2
are taken during penetration (at time t = 0), whereas
the DMT readings p0 and p1 are obtained at approxi
mately t = 15 s and t = 30 s after penetration respect
ively, hence some pore pressure dissipation may
Figure 2. Comparison of DMT and CPTu-equivalent pres have occurred prior to obtaining the DMT readings.
sures p0 (a) and p1 (b) for 49 clays (adapted from Ouyang
& Mayne 2018).
3 CPT VS. DMT DATA COMPARISONS
Mayne (2016) and Ouyang & Mayne (2017)
established a theoretical nexus between CPTu read 3.1 Sensitivity to stress history of DMT vs. CPT
ings (qt and u2) and DMT pressures (p0 and p1)
based on spherical cavity expansion solutions for The horizontal stress index KD is a key parameter
undrained penetration of both probes in soft to firm obtained from DMT interpretation. KD reflects
clays. This nexus would permit an exchange of inter cumulatively various stress history effects (overcon
pretations between the two tests, offering solidation, aging, in-situ horizontal earth pressure,
a complementary extension of methodologies. Data K0), as summarized by Marchetti (2010).
from paired sets of CPTu-DMT results in a variety Numerous researchers have observed that the DMT
of clays were used to support and validate the theor KD is considerably more sensitive than the CPT cone
etical links. The net cone resistance, qnet, defined as: resistance qc in monitoring compaction in the field.
The higher sensitivity of the DMT to stress history is
confirmed by comparisons CPT-DMT in controlled
calibration chamber (CC) testing conditions. Jamiolk
owski & Lo Presti (1998), in CC tests in Ticino sand,
found the DMT KD much more sensitive to stress-
was linked to the DMT pressure readings p0, p1 by strain history (including aging-like effects) than the
the following relationship: CPT qc. Lee et al. (2011) investigated the effects of
stress history on CPT and DMT in CC testing on 40
large specimens of Busan sand having different relative
density Dr and OCR in the range 1 to 8. The overcon
solidation produced an almost negligible increase in
Ouyang & Mayne (2018) further expanded this the normalized cone resistance qc /(σ’v)0.5 (Figure 3a),
concept and suggested that qt and u2 measured from but a substantial increase in KD (Figure 3b). It appears
97
that, while the normalized qc reflects essentially Dr and depth profiles of M estimated from DMT and CPTu
only to a minor extent OCR, before/after ground improvement with Rammed
Aggregate Piers (RAP) in silty and sandy soils at the
Bondeno test site, Italy. In natural soils they found
a rather good agreement between M estimated from
DMT (Marchetti 1980) and from CPTu using the cor
relations proposed by Robertson (2009b), as shown in
Figure 4, while other CPT-based correlations (Lunne
& Christophersen 1983, Senneset et al. 1988) pro
vided M equal to about half the values of M obtained
from DMT and from Robertson (2009b). The agree
ment between the latter two methods is not surprising.
Figure 3. Effect of stress history on (a) normalized qc from In fact, the development of the CPT correlations for
CPT, and (b) KD from DMT in CC testing on Busan sand M by Robertson (2009b) was partly “guided” by the
(modified from Lee et al. 2011). author’s correlations between normalized DMT and
CPT parameters (Robertson 2009a), based on the rec
KD reflects not only Dr, but also stress history ognition that M estimated from DMT has often been
effects. The higher reactivity of KD to OCR indicated shown to provide excellent estimates of settlement
by Figure 3, in agreement with other available (e.g., Monaco et al. 2006). However, Figure 4 also
experiences, implies that to the same normalized qc shows that in treated soils, using the same correl
may correspond different values of KD (Marchetti ations, the increase in M from DMT was found much
2016). In Figure 3a the coefficient of determination more evident than the increase in M from CPTu, sug
r2 close to 1 for data points of any OCR suggests gesting higher ability of MDMT to reflect the benefit of
poor ability of the normalized qc to distinguish OC soil treatment.
from NC sands, hence estimating OCR from CPT As to the undrained shear strength (su) in clay,
alone appears problematic. This example supports available comparisons of su from DMT vs. other
the necessity of a multi-parameter/multi-test tests at well-documented sites have often found that
approach based on both DMT & CPT for estimating the depth profiles of su estimated from DMT
OCR in sand, as will be discussed in the following. (Marchetti 1980) plot in an intermediate position,
in particular in between the su profiles obtained
from CPT(u) assuming variable values of the cone
3.2 Soil parameters estimated from DMT vs. CPT bearing capacity factor (Nk or Nkt). An example is
Many indirect correlations exist between DMT and shown in Figure 5 (Marchetti et al. 2022), where
CPT results, since both tests are used to estimate the su values obtained in Fucino clay by Medusa
various geotechnical parameters. DMT are compared with the su from CPT(u) and
The DMT interpretation is commonly based on the from other tests.
original correlations developed by Marchetti (1980) Comparisons CPT-DMT are also available in terms
for “textbook” soils, although several updates, of pore pressure index and consolidation/drainage
reviewed by Marchetti et al. (2001), have been pro parameters inferred from dissipation tests. The DMT,
posed. Among the various parameters obtained from though non provided with a pore pressure sensor as the
DMT, the 1-D constrained modulus M (MDMT) is the CPTu, permits to determine the pre-insertion in-situ
most generally appreciated. According to Marchetti equilibrium pore pressure u0 in sand and to discern
(1980) MDMT is obtained from ED (Eq. 3), which links layers of different permeability based on the closing
soil stiffness to the p0 and p1 measured during mem pressure p2. As described by Marchetti et al. (2001), in
brane expansion (a “mini load test”), corrected as sand p2 ≈ u0, while in clay p2 > u0 due to some excess
a function of ID (soil type) and KD (OCR). Several pore pressure induced by penetration persisting at the
favorable comparisons in terms of both MDMT vs. ref time of the p2 measurement. The pore pressure index
erence M and DMT-predicted vs. measured settlements UD (Lutenegger & Kabir 1988) is defined as:
have shown that, in general, MDMT is a reasonable esti
mate of the “operative” or “working strain” modulus.
Numerous correlations have been proposed to
estimate the constrained modulus M from CPT in
various soil types. Typically, M is obtained by multi
plying the cone resistance (qc or qt), or the net cone The parameter UD is the DMT equivalent of the
resistance (qt – σv0), by an empirical coefficient α pore pressure ratio Bq from CPTu:
which depends on soil type and other soil properties,
and may vary in a rather large range (e.g., 1 to 10).
Comparisons of moduli M estimated from DMT
vs. CPT have been published by several authors. As
an example, Amoroso et al. (2022a) compared the
98
Figure 6. Comparison of UD from DMT and Bq from CPTu
(Benoit 1989, from Marchetti et al. 2001).
pressurization and penetration rates in intermediate

Figure 4. Comparison of constrained moduli M estimated
soils in the Po river valley, Italy, combined with
from CPTu (Robertson 2009b) and SDMT (Marchetti CPTus at variable penetration rates. In these soils, the
1980) in natural and treated soils at the Bondeno test site highly accurate and repeatable time-for-reading facil
(modified from Amoroso et al. 2022a). ity of the Medusa DMT permitted to identify some
trends in the variation of DMT pressure measure
ments in response to variable-rate testing conditions,
i.e., a slower penetration/pressurization rate ‘shifts’
the interpretation towards drained behavior, while
a faster penetration/pressurization rate ‘shifts’ the
interpretation towards undrained behavior. In sand,
a fully drained response was found for any
penetration/pressurization rate. UD appears as
a suitable independent parameter to discern between
drained, undrained or partially drained soil behavior
and may be usefully combined with ID to obtain some
‘soil behavior type’. Ongoing research investigates
the potential of combining variable-rate Medusa
DMT and CPTu tests as an innovative approach for
characterizing the in-situ behavior of intermediate
soils.
The DMT is also used to estimate in-situ
consolidation/permeability parameters by means of
dissipation tests. Differently from CPTu, the decay
with time monitored by the DMT is not that of the
pore pressure, but of the total contact A-pressure.
This involves lack of worry over desaturation or
poor saturation of a porous element, which may
Figure 5. Comparison of profiles of the undrained shear affect the u2 measurement in CPTu dissipation
strength su from standard/Medusa DMT, CPT/CPTu and tests. The Medusa DMT technology, which permits
other in-situ and laboratory tests (data from Burghignoli to acquire up to 3 repeated A-pressure readings
et al. 1991) at Fucino-Telespazio (Marchetti et al. 2022). per second, has extended the range of soil types in
which A-dissipations are feasible from low perme
ability soils (clays, silty clays) to intermediate per
Similar to Bq, UD may help delineate soil stratig meability soils (silts, silty sands).
raphy, particularly in highly stratified deposits. In
free-draining soils UD ≈ 0, in non free-draining soils
UD > 0. The example in Figure 6 (Benoit 1989) 4 DERIVATION OF SOIL PARAMETERS BY
shows a good agreement between DMT UD and COMBINED USE OF CPT & DMT
CPTu Bq.
The use of the DMT in intermediate soils (silty
4.1 Estimating OCR in sand from DMT & CPT
sands, silts, sandy silts) is being increasingly investi
gated (Schnaid et al. 2016, 2018). Monaco et al. The combined use of DMT & CPT may provide esti
(2021) presented preliminary results of tests carried mates of the OCR in sand. Marchetti et al. (2001)
out using the Medusa DMT adopting variable suggested to use the ratio between the constrained
99
modulus from DMT MDMT and the cone resistance the times when reference “imparted OCR” were
qc from CPT as an indicator OCR in sand: known, i.e. at end-of-construction and post-
removal, Monaco et al. (2014) derived the correl
ation (Figure 7):
The above preliminary indications were based on

available CPT-DMT data before/after compaction of
sand fills (e.g., Jendeby 1992), where MDMT/qc was
found to increase with the overconsolidation achieved Using the same data set, Monaco et al. (2014) also
by compaction, combined with data from CC research constructed a correlation OCR-KD based only on
and from other tests. DMT. However the OCR-KD correlation is not
unique, but likely depending also on the relative dens
ity Dr. In fact, going back to the CC results shown in
Figure 3b (Lee et al. 2011), it can be deduced that for
a given Dr there is a one-to-one correspondence
between KD and OCR values (i.e., a vertical line
having constant Dr in the diagram intersects each
OCR curve for a unique KD value), but if Dr is not
constant the same OCR may result from a different
combination of KD and Dr. Hence a given KD may be
due either to a low Dr and a high OCR, or to a high
Dr and a low OCR. In order to separate the Dr effect
from the OCR effect, i.e., to pinpoint the right OCR,
Dr pair and then to estimate OCR, the normalized qc
(Figure 3a) is also necessary to provide an indication
of Dr on the horizontal axis. Hence both qc and KD
would be needed to estimate OCR in sand, i.e., CPT
Figure 7. Correlation OCR-MDMT/qt in sands obtained alone or DMT alone are insufficient. The OCR-KD
from paired DMT-CPTu data at the Treporti trial embank correlation by Monaco et al. (2014) works well at the
ment test site (modified from Monaco et al. 2014). Treporti test site where it was calibrated because in
these sands Dr is almost uniform, but it may not work
at sites where Dr is variable. Based on CC tests on
Busan sand samples having different Dr and OCR,
Choo et al. (2015) proposed a correlation to estimate
OCR in sand based on KD and Dr, with Dr also esti
mated from DMT, hence using only DMT data. This
OCR-KD correlation needs to be calibrated for each
specific sand.
The OCR-MDMT/qt correlation (Eq. 19) by
Monaco et al. (2014), based on both DMT and CPTu
data, appears to have more general validity and may
provide broad OCR estimates in different sands.
Additional research is encouraged to investigate the
Figure 8. Chart for evaluating K0 = f(KD, qc/σ’v) proposed dependency of the correlation OCR-MDMT/qt on Dr,
by Marchetti (1985). stress level and possibly sand type.
The possibility to estimate OCR in sand by the 4.2 Estimating K0 in sand from DMT & CPT
combined use of DMT and CPT was investigated by Similar considerations may be extended to the esti
Monaco et al. (2014) as part of an extensive experi mation of K0 in sand. The K0 – qc – KD chart in
mental study carried out at the Treporti Test Site, Figure 8, elaborated by Marchetti (1985), is prob
Venice, Italy. At this site, a full-scale cylindrical ably the first example of combined use of CPT &
trial embankment (40 m diameter, 6.7 m height, DMT. The chart permits to estimate K0 once qc and
applied pressure 106 kPa) was built and monitored KD are known. Figure 8 shows, besides the continu
from the beginning of its construction until its com ous curves obtained from CC tests, an additional
plete removal, four years later. Using same-depth scale for qc/σ’v based on 25 data points in Po river
values of MDMT and qt obtained in sand layers from sand, where an average value K0 = 0.55 was
paired DMT and CPTu soundings carried out at estimated.
100
Baldi et al. (1986) enriched such K0 – qc – KD (Figure 9b). Amoroso et al. (2022a, b), using paired
chart with additional CC work and converted it into DMT and CPTu data before/after ground improve
the following equations: ment with Rammed Aggregate Piers (RAP) at the
Bondeno test site, found a similar trend of the ratio
MDMT/qt (Figure 10), consistent with the observed
increase in MDMT higher than M from CPTu in trea
ted soils (Figure 4). This evidence suggests that the
DMT is more sensitive to stiffness variations as
a consequence of the increase in horizontal stress
produced by soil treatment, as reflected by the
Eq. 20a was determined as the best fit of CC data, increase in K0 (and OCR), also shown in Figure 10.
obtained on pluviated artificial sand, while Eq. 20b Similar pre-post comparisons were reported by
was obtained by modifying the last coefficient to Amoroso et al. (2018). The finding that compaction
predict “correctly” K0 for the natural Po river sand. (a sort of “imparted overconsolidation”) increases
Marchetti et al. (2001) recommended using Eqs
both MDMT and qc, but MDMT at a faster rate, inspired
20a and 20b (Baldi et al. 1986) with the following
the use of the ratio MDMT/qc as a “proxy” of OCR, as
values of the last coefficient: -0.002 in “freshly depos previously discussed.
ited” sand, -0.005 in “seasoned” sand. In this way, the
influence of OCR on K0 is incorporated by the choice
of the last coefficient, which involves appreciable sub 5.2 Liquefaction assessment based on DMT &
jectivity (e.g., a sand could be assumed as “fresh” if CPT
KD = 1-2, “seasoned” if KD = 5-6). The use of the DMT for liquefaction assessment is
Choo et al. (2015), based on CC testing on Busan receiving increasing attention. Much of the interest on
sand, proposed a correlation to estimate K0 in sand the development of simplified methods based on the
from DMT only, but similarly to OCR this correlation DMT KD derives from its high sensitivity to stress his
requires specific calibration for different sands. tory, besides to other factors (e.g., Dr, K0, aging) that
Based on the CC data set by Baldi et al. (1986), influence liquefaction resistance. Numerous
Hossain & Andrus (2016) proposed to estimate K0 in researchers (e.g., Pyke 2003, Leon et al. 2006 and
sand by use of the following correlation: many others) have pointed out that past stress-strain
history (overconsolidation, aging) is likely to have
a much greater effect on increasing liquefaction resist
ance than penetration resistance.
Simplified methods for estimating the cyclic
resistance ratio (CRR) based on DMT KD have
in which the influence of OCR on K0 is explicitly
been proposed over the years, including the most
taken into account. Eq. 21 could be used in combin
recent by Monaco et al. (2005), Tsai et al. (2009),
ation with OCR estimated by Eq. 19.
Robertson (2012), Marchetti (2016), Chiaradonna
All the above considerations imply that the chal
& Monaco (2022) shown in Figure 11 (valid for
lenging task of estimating OCR and K0 in sands
magnitude M = 7.5 and clean uncemented sand).
requires a multi-parameter/multi-test approach,
All these methods have in common a strong link
based on the combined use of DMT and CPT.
with CPT-based methods in their origin, in an effort
to relate in some way to the field performance data
base that provides a vast experimental validation for
5 APPLICATIONS BASED ON COMBINED
current methods based on CPT, but is currently
USE OF CPT & DMT DATA
limited for DMT-based methods.
5.1 Use of DMT & CPT in monitoring ground
improvement
Several comparisons of pre-post CPT & DMT exe
cuted for monitoring ground improvement are avail
able. Various studies have found the increase in
MDMT after soil treatment approximately twice the
increase in qc. For instance, Jendeby (1992) found
an increase of the ratio MDMT/qc from ≈ 5-12 pre
compaction to ≈ 12-24 post-compaction of a loose
sand fill (Figure 9a). Bałachowski & Kurek (2015)
found the mean increase in MDMT after vibroflotation
of sand about 2.3 times higher than the correspond Figure 9. Ratio MDMT/qc before-after compaction of sand
ing increase in qc and an increase of MDMT/qc from ≈ fills/deposits: data from (a) Jendeby (1992), (b) Bała
2-10 pre-compaction to ≈ 10-24 post-compaction chowski & Kurek (2015).
101
soundings at five saturated loose sand sites in Taiwan
that experienced liquefaction.
Using a similar approach, independently, Robert
son (2012) re-elaborated the CPT-DMT data set col
lected by Tsai et al. (2009) and proposed, for young
uncemented sands (ID > 1.2 and 2 < KD < 6), an
average relationship between the normalized clean
sand equivalent cone resistance Qtn,cs, defined as:
where Kc = correction factor for fines content esti

mated using Ic, and the DMT KD:
Based on Eq. 23, Robertson (2012) converted the

Robertson & Wride (1998) CRR-Qtn,cs curve into
a CRR-KD curve. Later Marchetti (2016) obtained
a new CRR-KD curve by combining the same Eq. 23
Figure 10. Profiles of MDMT/qt, OCR and K0 obtained from
with the CPT curve by Idriss & Boulanger (2006).
combined DMT-CPTu data in natural and treated soils at the Chiaradonna & Monaco (2022) proposed an
Bondeno test site (modified from Amoroso et al. 2022b). update of the CRR-KD curve adopting the CPT-based
framework by Boulanger & Idriss (2014). Using the
CPT-DMT data set by Tsai et al. (2009), considering
only CPT data having Ic between 1.5 and 2.6,
a direct correlation was established between KD and
the clean sand equivalent cone resistance qc1Ncs
defined according to Boulanger & Idriss (2014):
where qc1N is the normalized cone resistance and

Δqc1N accounts for the effect of fines content.
Despite the dispersion, the data set (Figure 12) is
well described by a linear trend with a slope of 20.
Figure 12 also shows a second data set from the
Scortichino site, Italy, where both CPT and SDMT
data were available (Tonni et al. 2015), which is
better interpreted by a linear trend with a slope of
about 30. Based on this limited data, Chiaradonna &
Monaco (2022) adopted an average coefficient of 25:
Figure 11. Recent CRR-KD correlations for clean sand.
The CRR-KD curve by Monaco et al. (2005) was

obtained by translating “consensus” CPT (and SPT) which is compatible with Eq. 23 by Robertson
curves, namely the CPT curve by Robertson & Wride (2012), also used by Marchetti (2016). By substitut
(1998), using the relative density Dr as intermediate ing Eq. 25 into the CRR-qc1Ncs curve by Boulanger
parameter. This approach is inevitably affected by & Idriss (2014), a new CRR-KD curve was obtained.
some uncertainty in estimating Dr from the in-situ Average CPT-DMT correlations such as Eq. 23
tests. Differently, Tsai et al. (2009) obtained a CRR-KD (Robertson 2012) and Eq. 25 (Chiaradonna & Monaco
curve by translating “consensus” CPT (and SPT) 2022) may be useful to convert CPT- into DMT-based
curves, again Robertson & Wride (1998) for CPT, by liquefaction triggering curves. On the other hand, the
using direct correlations between the CPT normalized example in Figure 12 suggests that the qc1Ncs – KD rela
clean sand equivalent cone resistance and the DMT tionship varies from one site to another, presumably as
KD, established using data from paired CPT-DMT a consequence of the higher sensitivity of KD to stress
102
history. Two sites may have similar qc1Ncs, but different Besides the CRR-KD curve, Marchetti (2016) also
KD depending on different stress history (Marchetti proposed a method to estimate CRR based on the com
2016). This implies that qc1Ncs and KD do not contain bined use of DMT & CPT (Figure 13), aiming to com
equivalent information, and may also explain in part plement the higher sensitivity of KD to stress history
the high dispersion of the CPT-DMT correlations with the much larger field performance support of
(Marchetti 2015). Robertson (2015), recognizing the CPT-based methods. The conceptual framework is the
higher sensitivity of KD to factors that cause soil micro- following: (1) estimate CRR using a “consensus” CPT
structure (stress and strain history, age, cementation/ liquefaction curve; (2) increase (shift upwards) the
bonding), suggested using the Qtn,cs – KD relationship CRR predicted by CPT if KD is high, reduce it (shift
to assess microstructure, noting that in young uncemen downwards) if KD is low. As an example, Figure 13 for
ted sands generally Qtn,cs > 16 KD, while soils with sig a normalized cone resistance Qcn = 100 and KD = 3
nificant microstructure tend to have Qtn,cs < 16 KD. provides CRR = 0.125 (i.e., lower than predicted by the
In perspective, the DMT could offer a valuable dashed CPT curve), while for the same Qcn = 100 and
integration to current methods for liquefaction KD = 5 the chart provides CRR = 0.17 (i.e., higher than
assessment based on CPT, SPT or VS, also given the predicted by CPT). The method proposed by Marchetti
general recommendation towards the use of “redun (2016) for estimating CRR from both DMT & CPT is
dant” correlations based on different in- another example of multi-parameter / multi-test
situ techniques / parameters. The main drawbacks of approach. It is expectable that an estimate based on
current DMT-based methods are: (1) lack of any cor two measured independent parameters could be more
rection factor for the fines content, and (2) limited accurate than an estimate based on just one parameter.
experimental validation based on field performance
data from real earthquakes. Ongoing research is
attempting to fill these gaps. The implementation of 6 PROPER MATCHING OF CPT & DMT DATA
an adequate case-history database for validation of
the DMT-based approach could support the introduc One specific issue that arises when using any interpret
tion of more consistent liquefaction triggering ation or design approach that requires the combination
curves, considering also the fines content influence. of CPT & DMT results is the proper matching of cor
responding data. In fact, the CPT measurements are
taken every 10-20 to 50 mm, whereas the DMT pres
sure readings are taken every 200 mm. Data from
paired CPT & DMT soundings shall be matched at
respective elevations. Care must be adopted even
when nearby soundings are carried out at the same
ground elevation, particularly in non uniform, highly
stratified deposits.
Comparisons between individual same-depth values
from nearby CPT & DMT profiles often show consid
erable scatter due to variations in soil stratigraphy and
consistency, hence adjacent data from the same depth
may not always represent the same soil. Sand deposits
Figure 12. Relationship qc1Ncs – KD from published CPT are often characterized by high variability in grain size
DMT data records at two test sites (Chiaradonna & Monaco distribution and Dr, then plots of individual data points
2022). from nearby CPT & DMT may show large scatter. The
profiles of qt may also differ slightly from those of
adjacent DMT since the CPT senses soil slightly ahead
and behind the cone tip (up to 15 cone diameters,
depending on soil strength / stiffness and in-situ effect
ive stresses, Ahmadi & Robertson 2005). The DMT
appears to be less influenced by soil layers ahead and
behind since the probe is stopped and the membrane
expanded in horizontal direction. Hence, in interbedded
soils the CPT may be influenced by adjacent soil layers
somewhat more than the DMT (Robertson 2009a).
Any comparison between adjacent CPT & DMT
should be done in terms of the near continuous pro
files with depth, so that any variation in soil stratig
raphy can be identified from the profiles. It is also
common to compare values obtained at the same
Figure 13. Correlation for estimating CRR based on both depth within relatively uniform sections of a deposit.
DMT & CPT for clean sand (modified from Marchetti As an example, at the highly stratified Treporti test
2016). site (Monaco et al. 2014) the paired CPTu-DMT
103
data, including the data pairs MDMT – qt shown in should be planned. The increase in time and cost of
Figure 7, were carefully selected by retaining only a site investigation program including paired CPT &
pairs from uniform soil layers of significant thick DMT soundings can be efficiently optimized consid
ness to avoid any possible mismatching of data. ering that both tests can share the same insertion rig
In general, the following matching criteria may be and rods, and even the same cable when using the
adopted: (1) match paired CPT & DMT profiles Medusa DMT (i.e., only the probe connected to the
accounting for respective ground elevations, if not bottom rod needs to be changed). Such “redun
coincident; (2) correct both CPT & DMT test depths dancy” would be largely compensated by substantial
for inclination; (3) select soil type (e.g., sand-like, Ic increase in soil information.
≤ 2.60 and ID > 1, or clay-like, Ic > 2.60 and ID < 1) Interrelationships between CPT & DMT can be
and possibly depth range of interest; (4) identify used to expand and improve correlations and appli
“visually” or by cross-correlation of representative cations by using experience and databases from one
parameters (e.g., qt and MDMT) corresponding soil test and extrapolating to the other test. The accumu
layers in CPT & DMT profiles, to detect any depth lation and sharing of paired data sets from compan
misalignment offset; (5) identify soil layers having ion series of CPT & DMT soundings at well-
a reasonable thickness (e.g., in which at least two documented and benchmark test sites is a desirable
DMT readings are present) and relatively uniform address for future research and practice.
CPT & DMT parameters; (6) use average values of
CPT & DMT parameters in each layer. It would be
desirable that the matching algorithm is implemented ACKNOWLEDGEMENTS
in software tools, although some visual inspection
and judgment may be still needed. Special thanks to Peter K. Robertson for sharing his
updates on CPT-DMT correlations. Thanks to Diego
Marchetti, Sara Amoroso, Anna Chiaradonna for
7 CONCLUDING REMARKS their helpful suggestions, to Gianfranco Totani and
Giovanni Bosco for the shared experience on in-situ
Direct-push in-situ tests such as the CPT and the testing. Grateful thanks to Silvano Marchetti for his
DMT, particularly in their powerful seismic config constant and bright guidance over many years.
urations, have proven to be remarkably expedient in
geotechnical site investigations. Recent experience
highlights an increasing awareness of researchers REFERENCES
and practitioners that an adequate site investigation
for a given project cannot rely solely on one isolated Ahmadi, M.M. & Robertson, P.K. 2005. Thin layer effects
testing technique, delineating the trend of expanding on the CPT qc measurement. Can. Geotech. J. 42(9):
diffusion of a multi-parameter / multi-test approach 1302–1317.
Amoroso, S., Comina, C. & Marchetti, D. 2020. Combined
in site investigation practice. P- and S-Wave Measurements by Seismic Dilatometer
The CPT and the DMT can be effectively com Test (SPDMT): A Case History in Bondeno (Emilia
bined to obtain more accurate information on stratig Romagna, Italy). Geotech. Testing J. 43(2): 383–393.
raphy and stress history of soil deposits, more Amoroso, S., Garcia Martinez, M.F., Monaco, P., Tonni, L.,
reliable estimates of soil parameters (e.g., OCR and Gottardi, G., Rollins, K.M., Minarelli, L., Marchetti, D.
K0 in sand) and for various design applications (e.g., & Wissmann, K.J. 2022a. Comparative study of CPTU
monitoring ground improvement, liquefaction and SDMT in liquefaction prone silty sands with ground
assessment). Moving towards an in-situ multi- improvement. J. Geotech. Geoenviron. Eng. (accepted)
parameter / multi-test approach appears a logical http://dx.doi.org/10.1061/(ASCE)GT.1943-5606.0002801
Amoroso, S., Martínez, M.F., Tonni, L., Gottardi, G.,
trend. In this respect, the availability of the DMT
Monaco, P., Rollins, K.M., Minarelli, L., Marchetti, D.
stress history parameter KD is important not only & Wissmann, K.J. 2022b. Monitoring ground improve
“per se”, but also in combination with parameters ment by Rammed Aggregate Piers using a combined
obtained from CPT. CPTU and SDMT approach at a silty sand
Each test has advantages and limitations. The liquefaction-prone site in Emilia-Romagna. Proc. 5th
CPT is often used more than the DMT, especially for Int. Symp. Cone Penetration Testing CPT’22, Bologna,
smaller projects, being faster and less expensive (no 8-10 June 2022.
pauses of penetration are required for obtaining the Amoroso, S., Rollins, K.M., Monaco, P., Holtrigter, M. &
measurements, as for the DMT). However, the DMT Thorp, A. 2018. Monitoring Ground Improvement
Using the Seismic Dilatometer in Christchurch, New
has been shown to be more sensitive to stress history,
Zealand. Geotech. Testing J. 41(5): 946–966.
which has a dominant influence on soil behavior. Bałachowski, L. & Kurek, N. 2015. Vibroflotation Control
Hence, their mutual integration could helpfully of Sandy Soils. In S. Marchetti, P. Monaco & A. Viana
merge the respective advantages and counterbalance da Fonseca (eds), Proc. 3rd Int. Conf. Flat Dilatometer
their limitations. For instance, the CPT could be DMT’15, Rome, 14-16 June 2015, 185–190.
used for a “screening” of the site on a large scale, Baldi, G., Bellotti, R., Ghionna, V., Jamiolkowski, M.,
the DMT at locations of particular relevance for the Marchetti, S. & Pasqualini, E. 1986. Flat Dilatometer
project. Adjacent CPT & DMT at selected locations Tests in Calibration Chambers. Proc. Spec. Conf. Use of
104
In Situ Tests in Geotech. Eng. In Situ ’86, Blacksburg, Marchetti, S. 2011. Discussion of “CPT-DMT Correl
431–446. ations” by P.K. Robertson. J. Geotech. Geoenviron.
Benoit, J. 1989. Personal communication to S. Marchetti. Eng. 137(4): 441–442.
Boulanger, R.W. & Idriss, I.M. 2014. CPT and SPT lique Marchetti, S. 2015. Some 2015 Updates to the TC16 DMT
faction triggering procedures. Report No UCD/GCM Report 2001. In S. Marchetti, P. Monaco & A. Viana da
14/01, University of California at Davis. Fonseca (eds), Proc. 3rd Int. Conf. Flat Dilatometer
Burghignoli, A., Cavalera, L., Chieppa, V., DMT’15, Rome, 14-16 June 2015, 43–65.
Jamiolkowski, M., Mancuso, C., Marchetti, S., Pane, V., Marchetti, S. 2016. Incorporating the Stress History Param
Paoliani, P., Silvestri, F., Vinale, F. & Vittori, E. 1991. eter KD of DMT into the Liquefaction Correlations in
Geotechnical characterization of Fucino clay. Proc. 10th Clean Uncemented Sands. J. Geotech. Geoenviron. Eng.
Eur. Conf. Soil Mech. Found. Eng., Florence, 26-30 142(2): 04015072.
May 1991, 1: 27–40. Rotterdam: Balkema. Marchetti, S. & Monaco, P. 2018. Recent Improvements in
Chiaradonna, A. & Monaco, P. 2022. Assessment of lique the Use, Interpretation, and Applications of DMT and
faction triggering by seismic dilatometer tests: compari SDMT in Practice. Geotech. Testing J. 41(5): 837–850.
son between semi-empirical approaches and non-linear Marchetti, S., Monaco, P., Totani, G. & Calabrese, M. 2001.
dynamic analyses. Proc. 20th Int. Conf. Soil Mech. Geo- The Flat Dilatometer Test (DMT) in Soil Investigations –
tech. Eng., Sydney, 1-5 May 2022. A Report by the ISSMGE Committee TC16. Official ver
Choo, H., Lee, W., Lee, C., Kim, Y.-S. & Lee, M.-J. 2015. sion approved by ISSMGE TC16 reprinted in R.
Perception of Overconsolidated States of A. Failmezger & J.B Anderson (eds), Proc. 2nd Int. Conf.
Coarse-Grained Soils Using DMT Results. In Flat Dilatometer, Washington D.C., 2-5 April 2006, 7–48.
S. Marchetti, P. Monaco & A. Viana da Fonseca (eds), Mayne, P.W. 2006. Interrelationships of DMT and CPT
Proc. 3rd Int. Conf. Flat Dilatometer DMT’15, Rome, Readings in Soft Clays. In R.A. Failmezger & J.
14-16 June 2015, 205–212. B Anderson (eds), Proc. 2nd Int. Conf. Flat Dilatometer,
Hossain, M.A. & Andrus, R.D. 2016. At-Rest Lateral Stress Washington D.C., 2-5 April 2006, 231–236.
Coefficient in Sands from Common Field Methods. Mayne, P.W. 2016. Evaluating effective stress parameters
J. Geotech. Geoenviron. Eng. 142(12): 06016016. and undrained shear strengths of soft-firm clays from
Idriss, I.M. & Boulanger, R.W. 2006. Semi-empirical proced CPTu and DMT. Australian Geomech. J. 51(4): 27–55.
ures for evaluating liquefaction potential during Mayne, P.W. & Bachus, R.C. 1989. Penetration Pore Pres
earthquakes. Soil Dyn. Earthquake Eng. 26(2-4): 115–130. sures in Clay by CPTU, DMT, and SBP. Proc. 12th Int.
Jamiolkowski, M. & Lo Presti, D.C.F. 1998. DMT research Conf. Soil Mech. Geotech. Eng., Rio de Janeiro,
in sand. What can be learned from calibration chamber 291–294. Rotterdam: Balkema.
tests. 1st Int. Conf. Site Characterization ISC ’98, Mayne, P.W., Coop, M.R., Springman, S.M., Huang, A.B.
Atlanta, 19-22 April 1998 (oral presentation). & Zornberg, J.G. 2009. Geomaterial behavior and
Jendeby, L. 1992. Deep Compaction by Vibrowing. Proc. testing. Proc. 17th Int. Conf. Soil Mech. Geotech. Eng.,
Nordic Geotech. Meeting NGM-92, Aalborg, 1: 19–24. Alexandria, 5-9 October 2009, 4: 2777–2872. Amster
Lee, M.J., Choi, S.K., Kim, M.T. & Lee, W. 2011. Effect of dam: IOS Press.
stress history on CPT and DMT results in sand. Eng. Monaco, P., Amoroso, S., Marchetti, S., Marchetti, D.,
Geol. 117(3-4): 259–265. Totani, G., Cola, S. & Simonini, P. 2014. Overconsoli
Leon, E., Gassman, S.L. & Talwani, P. 2006. Accounting dation and Stiffness of Venice Lagoon Sands and Silts
for Soil Aging When Assessing Liquefaction Potential. from SDMT and CPTU. J. Geotech. Geoenviron. Eng.
J. Geotech. Geoenviron. Eng. 132(3): 363–377. 140(1): 215–227.
Lunne, T. & Christophersen, H.P. 1983. Interpretation of Monaco, P., Marchetti, S., Totani, G. & Calabrese, M. 2005.
cone penetrometer data for offshore sands. Proc. 15th Sand Liquefiability Assessment by Flat Dilatometer Test
Annual Offshore Technology Conf., 181–188. (DMT). Proc. 16th Int. Conf. Soil Mech. Geotech. Eng.,
Lutenegger, A.J. & Kabir, M.G. 1988. Dilatometer Osaka, 4: 2693–2697. Rotterdam: Millpress.
C-reading to help determine stratigraphy. Proc. Int. Monaco, P., Tonni, L., Amoroso, S., Garcia Martinez, M.F.,
Symp. Penetration Testing ISOPT-1, Orlando, 1: Gottardi, G., Marchetti, D. & Minarelli, L. 2021. Use of
549–554. Rotterdam: Balkema. Medusa DMT in alluvial silty sediments of the Po river
Marchetti, D., Monaco, P., Amoroso, S. & Minarelli, L. valley. Proc. 6th Int. Conf. Geotech. Geophys. Site Char
2019. In situ tests by Medusa DMT. Proc. XVII Eur. acterisation ISC’6, Budapest, 26-29 September 2021.
Conf. Soil Mech. Geotech. Eng., Reykjavik, 1-6 Septem Monaco, P., Totani, G. & Calabrese, M. 2006. DMT-
ber 2019. predicted vs observed settlements: a review of the avail
Marchetti, D., Monaco, P., Totani, G., Totani, F. & able experience. In R.A. Failmezger & J.B Anderson
Amoroso, S. (2022). Comparisons CPT-DMT in soft (eds), Proc. 2nd Int. Conf. Flat Dilatometer, Washington
clay at Fucino-Telespazio GeoTest site. Proc. 5th Int. D.C., 2-5 April 2006, 244–252.
Symp. Cone Penetration Testing CPT’22, Bologna, 8-10 Ouyang, Z. & Mayne, P.W. 2017. Spherical cavity expan
June 2022. sion nexus between CPTu and DMT in soft-firm clays.
Marchetti, S. 1980. In Situ Tests by Flat Dilatometer. Proc. 19th Int. Conf. Soil Mech. Geotech. Eng., Seoul,
J. Geotech. Eng. Div. 106(GT3): 299–321. 17-22 September 2017, 631–634.
Marchetti, S. 1985. On the Field Determination of K0 in Ouyang, Z. & Mayne, P.W. 2018. Effective stress param
Sand. Proc. 11th Int. Conf. Soil Mech. Found. Eng., San eters of clays from DMT. Geotech. Testing J. 41(5):
Francisco, 5: 2667–2673. Rotterdam: Balkema. 851–867.
Marchetti, S. 2010. Sensitivity of CPT and DMT to stress Pyke, R. 2003. Discussion of “Liquefaction Resistance of
history and aging in sands for liquefaction assessment. Soils: Summary Report from the 1996 NCEER and
Proc. 2nd Int. Symp. Cone Penetration Testing CPT’10, 1998 NCEER/NSF” by T.L. Youd et al., J. Geotech.
Huntington Beach, 9-11 May 2010. Geoenviron. Eng. 129(3): 283–284.
105
Robertson, P.K. 1990. Soil classification using the cone Schnaid, F., Odebrecht, E., Sosnoski, J. & Robertson, P.K.
penetration test. Can. Geotech. J. 27(1): 151–158. 2016. Effects of test procedure on flat dilatometer test
Robertson, P.K. 2009a. CPT-DMT Correlations. (DMT) results in intermediate soils. Can. Geotech. J. 53
J. Geotech. Geoenviron. Eng. 135(11): 1762–1771. (8): 1270–1280.
Robertson, P.K. 2009b. Interpretation of cone penetration Senneset, K., Sandven, R., Lunne, T., By, T. &
tests – a unified approach. Can. Geotech. J., 46(11): Amundsen, T. 1988. Piezocone Tests in Silty Soils.
1337–1355. Proc. Int. Symp. Penetration Testing ISOPT-1, Orlando,
Robertson, P.K. 2012. The James K. Mitchell Lecture: 2: 955–966. Rotterdam: Balkema.
Interpretation of in-situ tests – some insights. In Togliani, G., Calzolari, L. & Menghini, A. 2015. Governolo
R. Coutinho & P.W. Mayne (eds), Proc. 4th Int. Conf. & Gonzaga (Mantua, Italy) Experimental Sites: In situ
Geotech. Geophys. Site Characterization, Porto de Test comparison. In S. Marchetti, P. Monaco & A. Viana
Galinhas, 1: 3–24. da Fonseca (eds), Proc. 3rd Int. Conf. Flat Dilatometer
Robertson, P.K. 2015. Soil Behavior Type using the DMT. DMT’15, Rome, 14-16 June 2015, 259–270.
In S. Marchetti, P. Monaco & A. Viana da Fonseca Tonni, L., Gottardi, G., Amoroso, S., Bardotti, R.,
(eds), Proc. 3rd Int. Conf. Flat Dilatometer DMT’15, Bonzi, L., Chiaradonna, A., d’Onofrio, A.,
Rome, 14-16 June 2015, 243–250. Fioravante, V., Ghinelli, A., Giretti, D., Lanzo, G.,
Robertson, P.K. 2017. DMT-CPT Correlations. Prof. Sil Madiai, C., Marchi, M., Martelli, L., Monaco, P.,
vano Marchetti Memorial Session on In-Situ Testing, Porcino, D., Razzano, R., Rosselli, S., Severi, P.,
19th Int. Conf. Soil Mech. Geotech. Eng., Seoul, 17 Silvestri, F., Simeoni, L., Vannucchi, G. & Aversa, S.
22 September 2017 (oral presentation). 2015. Interpreting the deformation phenomena triggered
Robertson, P.K. & Wride, C.E. 1998. Evaluating cyclic by the 2012 Emilia seismic sequence on the Canale
liquefaction potential using the cone penetration test. Diversivo di Burana banks. Rivista Italiana di Geotec
Can. Geotech. J. 35(3): 442–459. nica 49 (2):28–58 (in Italian).
Schnaid, F., Belloli, M.V.A., Odebrecht, E. & Marchetti, D. Tsai, P.-H., Lee, D.-H., Kung, G.T.-C. & Juang, C.H. 2009.
2018. Interpretation of the DMT in silts. Geotech. Test Simplified DMT-Based Methods for Evaluating Lique
ing J. 41(5): 868–876. faction Resistance of Soils. Eng. Geol. 103(1-2): 13–22.
106
Technical papers
Session 1: Equipment and procedures
© 2022 the Author(s), ISBN 978-1-032-31259-0
Incorporation of SH source wave parameter “SH Polarization” within DST

seismic trace characterization
Erick Baziw & Gerald Verbeek
Baziw Consulting Engineers Ltd., Vancouver, Canada
ABSTRACT: Downhole Seismic Testing (DST) is an important geotechnical testing technique for site char
acterization that provides low strain in-situ interval shear wave velocity estimates, which are fundamental
design parameters for static and dynamic soil analysis. A challenging problem in DST is to obtain an accurate
assessment or characterization of the quality of the acquired seismic data, which is then used to guide the
analysis process to obtain the most accurate interval velocity values. The characterization process is referred
to as Seismic Trace Characterization (STC). STC derives various independent parameters of the acquired seis
mic data at a particular depth, which are then fused together into a single classification. To date Baziw Con
sulting Engineers has identified five STC independent parameters. These five parameters are the linearity
estimates from the polarization analysis, the cross correlation coefficient of the full waveforms at the particu
lar depth and the preceding depth, a uniquely developed parameter referred to as the signal shape parameter,
the signal-noise-ratio and the peak symmetry differential, which provides insight into the skewing or time
shifting of the peak source wave response. This paper outlines a newly identified seismic trace feature that is
independent of the parameters listed above. This new parameter is SH Polarization (SHP), which quantifies
the desired polarization of the generated source on the horizontal plane compared with particle motion gener
ated on the vertical plane.
1 INTRODUCTION the seismic data? And 3) What is the appropriate

confidence level in the calculated interval velocities
The fundamental goal of Downhole Seismic Testing estimates? Over time work in addressing these three
(DST) is to obtain accurate estimates of low str concerns has resulted in the standardization of
ain (<10-5) shear (Vs) and compression (Vp) wave vel a DST post data processing methodology, which has
ocities. These velocities are directly related to the vari proven highly accurate and reliable. Currently qual
ous soil elastic constants, such as the Poisson’s ratio, ity assessment through Seismic Trace Characteriza
shear modulus, bulk modulus and Young’s modulus. tion (STC) relies upon five independent parameters
These parameters form the core of mathematical the (Baziw and Verbeek 2016, 2017 and 2018):
orems to describe the elasticity/plasticity of soils and
they are used to predict the soil response (settlement, • Parameter 1: the linearity estimates (LIN) the lin
liquefaction or failure) to imposed loads. Accuracy in earity or rectilinearity from polarization analysis.
the estimation of these two in-situ velocities is of para The LIN trace metric quantifies the correlation
mount importance because their values are squared between X, Y and Z axis responses. The linearity
during the calculation of the soil elastic constants. In approaches unity when the rectilinearity is high and
DST a seismic source wave is generated at the ground approaches zero when the rectilinearity is low. Lin
surface, and one or more downhole seismic receivers earity values nearing 1.0 identify seismic recordings
are used to record this wave at predefined depths. From that have highly correlated responses and strong
these recorded seismic traces arrival times are esti directionality, the quality of the data set with a high
mated and corresponding interval velocities calculated. linearity value can be considered good. Lower lin
Baziw Consulting Engineers (BCE) has invested earity values on the other hand indicate lower qual
considerable resources into the characterization of ity traces.
DST traces (Baziw and Verbeek 2016, 2017, and • Parameter 2: the Cross Correlation Coefficient
2018) to address three fundamental concerns. 1) (CCC) of the full waveforms at the particular
What is the quality of the acquired seismic data depth and the preceding depth. The CCC trace
sets? 2) What signal processing techniques can be metric gives an indication of the similarity
applied to improve the signal-to-noise ratio (SNR) of between the two waves being correlated when
DOI: 10.1201/9781003308829-8
109
deriving relative arrival times. CCC values
approaching 1.0 indicate that the two waveforms
are highly correlated. CCC values approaching 0
indicate very poor correlation.
• Parameter 3: the Signal Shape Parameter (SSP).
The SSP trace metric quantifies the deviation of
the shape of the frequency spectrum from an
ideal bell shape. SHP values approaching 1.0
indicate that the frequency has a desirable bells
shape. SHP values approaching 0 indicates that
the frequency spectrum deviates significantly
from the desired bell shape.
• Parameter 4: the Peak Symmetry Differential
(PSD) trace metric facilitates the identification of
traces whose peak source wave responses have
been significantly skewed due to measurement
noise or source wave reflection interference. The
“peak symmetry” error assessment is also carried
out on the adjacent peaks and/or troughs if the
amplitude exceeds 70 % of that for the peak
response. Traces with low PSD value are of
a lesser quality and require more attention during
analysis.
• Parameter 5: Signal to Noise Ratio (SNR).
The SNR trace metric is solely provided to
quantify what portion of the spectral content
of the recorded seismogram resides within the
desired source frequency spectrum irrespect
ive of source wave distortions such as near- Figure 1. DST data processing flow chart incorporating
field effects, reflections, refractions, and STC parameters.
“dirty sources”.
In this paper, the preliminary implementation and
As part of the post analysis of the seismic traces mathematical details of a new STC parameter is out
these parameters are then converted into a STC lined. This new parameter, the so called SH Polariza
grade ranging from A to F where A is highly desir tion (SHP), quantifies the desired polarization of the
able and F is unacceptable without corrective action. generated source on the horizontal plane compared
Next they are used as a guide for the data analysis with particle motion generated on the vertical plane.
and seismic signal processing (Baziw and Verbeek,
2018). A central part of this processing is the source
wave Signature Feature Isolation (SFI) to clearly 2 STC PARAMETER SH POLARIZATION
identify the source wave in the seismic trace by (SHP)
applying an exponential decay function to the
remainder of the trace. This can be performed in Seismic sources are designed to generate either com
three different ways: pression (P),vertically polarized shear (SV) waves or
• ASD or Automatic Signal Decay, where the horizontally polarized(SH) shear waves. Figure 2
program identifies the absolute maximum amp illustrates the compression and shear source waves
litude on all trace under analysis and then impacting upon a triaxial seismic sensor package. As
applies the decay function on either side of it is shown in Figure 2, the P-wave’s particle motion
that feature. is in the same direction as the raypath, the SH waves
• GSD or Guided Signal Decay, where the user particle motion is perpendicular to the raypath and is
identifies a specific feature for the traces under parallel to the horizontal ground surface, while the SV
analysis and then applies the decay function on wave’s particle motion is also perpendicular to the
either side of that feature. raypath, but along the vertical normal to the raypath.
• ISD or Individual Signal Decay, where the user P or SV waves generate four outgoing waves
identifies a specific feature for each trace under when impacting an interface (reflected SV and
analysis and then applies the decay function on P waves and transmitted SV and P waves). In con
either side of that feature on that particular trace. trast, SH waves have the desirable property of only
generating one reflected and one transmitted SH at an
The analysis process can then be visualized as interface. This results in considerably simplified seis
outlined in Figure 1. mic data sets. A popular SH source is the well-known
110
hammer beam (ASTM, 2017). The hammer-beam The SHP rankings outlined in Table 1 are prelimin
consists of applying a hammer blow laterally to the ary values. These values will be adjusted and refined
sides of special designed plates fixed at the surface. as a greater number DST seismic traces are processed.
3 IMPLEMENTATION OF SHP STC

PARAMETER ON REAL DATA SETS
The first and second real data sets outlined in this

section were acquired by Perry Geotech Limited
located in Tauranga, New Zealand. Figure 3 illus
trates a SH source wave DST vertical seismic profile
(x, y and z axis responses) of SH DST seismic data
acquired on the Left Side (LS) of the seismic probe,
while Figure 4 shows the VSP of data acquired on
the Right Side (RS) of the seismic probe. The STC
parameters for the LS and RS are outlined in
Tables 2 and 3, respectively.
Figure 2. Source P, SV, and SH body waves impacting

upon a triaxial sensor package.
The new STC parameter SHP quantifies and

applies a grade [A to F] to how closely the measured
SH wave seismic traces adhere to the desired polar
ization of the shear wave on the horizontal plane.
The SHP algorithm can be summarized as follows,
assuming that a triaxial (x, y and z axes) seismic
trace has been recorded.
1. Apply a minimal filter (200 Hz low pass) to the
acquired SH source wave. Figure 3. Data Set 1 – real data set. VSP of data acquired
2. Determine time index, t*, where the maximum on LS of seismic probe (200Hz low pass filter applied).
absolutepffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
amplitude ffiffiffiffiffiffiffiffiffiffiffiffiffiffiof theffi full waveform
(ρðtÞ ¼ x2 ðtÞ þ y2 ðtÞ þ z2 ðtÞ) occurs.
3. Establish a time window T which is defined as t*
Δt to t*+Δt where Δt = 30ms.
4. Over the established time T calculate the energy
of the full waveform, Eρ, and the energy of the
particle p ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
motion on ffi the X-Y plane Exy
(xyðtÞ ¼ x2 ðtÞ þ y2 ðtÞ).
5. Calculate energy ratio ER = Exy/Eρ.
6. Assign a SHP rank based upon calculated ER
value as outlined in Table 1.
Table 1. SHP rank and description.
ER Numeric Value SHP Rank Figure 4. Data Set 1 – real data set. VSP of data acquired
[0-1] [A-F] STC Description on right side of seismic probe (200Hz low pass filter
applied).
0.8 to 1.0 A very good to good
0.65 to 0.8 B good to acceptable As is evident from Tables 2 and 3 the data sets
acceptable to from the LS and RS have low STC values of ‘D’s and
0.5 to 0.65 C
questionable
‘F’s (predominantly due to low LIN values), but with
questionable to
0.3 to 0.5 D
unacceptable
high SHP rankings. This suggests that the acquired
< 0.3 F unacceptable
seismic traces have the desirable polarization on the
horizontal plane, but will require preferable axis
111
Table 2. STC parameters for LS for Data Set 1. probe. The STC parameters for the LS and RS are
outlined in Tables 5 and 6, respectively.
Depth LIN SSP CCC PSD SNR STC SHP The SHP rankings suggest that the acquired seismic
[m] [0-1] [0-1] [0-1] [0-1] [0-1] [A-F] [A-F] traces have desirable polarization on the horizontal
plane, but will require signal feature isolations based
3 0.68 0.54 0 0.86 0.71 N/A A upon the LIN and SSP (LS) and PSD (RS) values.
4 0.71 0.77 0.69 0.81 0.86 D A
5 0.59 0.71 0.96 0.73 0.88 D A
6 0.44 0.70 0.99 0.74 0.91 D A
7 0.60 0.75 0.96 0.57 0.97 D A
8 0.88 0.80 0.96 0.59 0.92 D B
Table 3. STC parameters for RS for Data Set 1.
Depth LIN SSP CCC PSD SNR STC SHP

[m] [0-1] [0-1] [0-1] [0-1] [0-1] [A-F] [A-F]
3 0.77 0.76 0 0.89 0.54 N/A A

4 0.30 0.77 0.77 0.94 0.88 D A
5 0.21 0.63 0.80 0.82 0.84 F A
6 0.46 0.65 0.87 0.77 0.88 D A
7 0.53 0.69 0.93 0.93 0.98 D A
Figure 5. Data Set 2 – real data set. VSP of data acquired
8 0.34 0.73 0.85 0.97 0.96 D B on LS of seismic probe (200Hz low pass filter applied).
selection (in this case the X axis) given the LIN

values. The SHP rankings give added confidence in
the ability to isolate SH responses and derive accurate
interval velocities after SFI implementation.
Table 4 outlines the estimated LS and RS interval vel
ocities and corresponding spread. As is shown in
Table 4 there is overall high correlation between the LS
and RS results (desired values should be ≤ 10% spread),
with only the estimated values for the depth interval
3.0m to 4.0m showing a spread slightly above 10%.
The second data set is another example where the
LS and RS have again low STC values of ‘D’s and
‘F’s (but now due to low PSD values) and high SHP
rankings.
Figure 6. Data Set 2 – real data set. VSP of data acquired

Table 4. Estimated LS and RS interval velocities. on RS of seismic probe (200Hz low pass filter applied).
Depth Interval Velocity
Depth LS RS Avg. Spread1
[m] [m/s] [m/s] [m/s] [%] Table 5. STC parameters for LS for Data Set 2.
0.0-3.0 182.4 204.1 193.3 5.6 Depth LIN SSP CCC PSD SNR STC SHP
3.0-4.0 150.4 122.2 136.3 10.3 [m] [0-1] [0-1] [0-1] [0-1] [0-1] [A-F] [A-F]
4.0-5.0 129.4 141.5 135.5 4.5
5.0-6.0 171.3 164.6 168 2.0 2 0.83 0.65 0.62 0.6 0.95 N/A A
6.0-7.0 151.4 141.4 146.4 3.4 3 0.8 0.79 0.71 0.89 0.97 D A
7.0-8.0 214.5 239.5 227 5.5 4 0.81 0.67 0.88 0.7 0.98 B A
5 0.77 0.7 0.95 0.58 0.93 B A
1 The spread is defined as ½ x (LS Interval Velocity – RS 6 0.76 0.63 0.96 0.8 0.97 D A
Interval Velocity)/Avg. Interval Velocity 7 0.71 0.58 0.97 0.98 0.98 D B
8 0.75 0.55 0.99 0.96 0.98 D B
For this data set Figure 5 illustrates a vertical seis- 9 0.76 0.5 0.98 0.89 0.98 D B
mic profile (x, y and z axis responses) of SH DST 10 0.79 0.53 0.98 0.86 0.98 D A
seismic data acquired on the Left Side (LS) of the 11 0.78 0.54 0.98 0.86 0.98 D B
seismic probe, while Figure 6 shows the VSP of data 12 0.84 0.55 0.98 0.84 0.98 D B
acquired on the Right Side (RS) of the seismic
112
Table 6. STC parameters for RS for Data Set 2.

[m] [0-1] [0-1] [0-1] [0-1] [0-1] [A-F] [A-F]
2 0.79 0.58 0.58 0.77 0.98 N/A A

3 0.88 0.66 0.81 0.77 0.98 B A
4 0.83 0.69 0.88 0.92 0.98 B A
5 0.81 0.63 0.92 0.01 0.98 F A
6 0.8 0.63 0.97 0.01 0.98 F B
7 0.81 0.55 0.96 0.01 0.98 F B
8 0.82 0.54 0.98 0.01 0.98 F B
9 0.84 0.48 0.98 0.01 0.98 F C Figure 8. Data Set 3 – real data set. VSP of data acquired
on LS of seismic probe (200Hz low pass filter applied).
10 0.83 0.58 0.98 0.2 0.98 F B
11 0.84 0.51 0.99 0.51 0.98 D B
12 0.84 0.54 0.98 0.65 0.98 D B
Table 8. STC parameters for LS for Data Set 3.

Table 7. Estimated LS and RS interval velocities for Data
Set 2. [m] [0-1] [0-1] [0-1] [0-1] [0-1] [A-F] [A-F]
Depth Interval Velocity 44.2 0.87 0.33 0.9 0.89 0.89 N/A C
Depth LS RS Avg. Spread 1 45.9 0.85 0.27 0.97 0.34 0.94 D C
[m] [m/s] [m/s] [m/s] [%] 47.9 0.68 0.41 0.95 0.36 0.8 D D
49.9 0.76 0.46 0.88 0.94 0.81 D D
0.0-2.0 103.1 106.5 119.1 0.8 51.9 0.51 0.46 0.62 0.01 0.83 F F
2.0-3.0 120.8 117.4 135.85 0.7 53.2 0.82 0.53 0.47 0.01 0.79 F F
3.0-4.0 134,0 137.7 171.45 0.7 54.4 0.82 0.55 0.84 0.01 0.81 F D
4.0-5.0 179.1 163.8 173.85 2.2 55.7 0.63 0.59 0.75 0.64 0.82 F F
5.0-6.0 160.7 187 158.55 3.8 56.9 0.39 0.54 0.97 0.01 0.77 F F
6.0-7.0 157.7 159.4 171.2 0.3 58.9 0.57 0.51 0.65 0.49 0.79 F F
7.0-8.0 173.9 168.5 176.35 0.8 59.8 0.87 0.53 0.26 0.85 0.76 F F
8.0-9.0 174 178.7 175.6 0.7
9.0-10.0 180.4 170.8 172.8 1.4
10.0-11.0 166.2 179.4 148.1 1.9
11.0-12.0 143.8 152.4 119.1 1.5, Table 9. STC parameters for RS for Data Set 3.
1 The spread is defined as ½ x (LS Interval Velocity – RS Depth LIN SSP CCC PSD SNR STC SHP
Interval Velocity)/Avg. Interval Velocity [m] [0-1] [0-1] [0-1] [0-1] [0-1] [A-F] [A-F]
44.2 0.72 0.37 0.92 0.63 0.88 N/A C

Table 7 outlines the estimated LS and RS interval
45.9 0.78 0.31 0.96 0.83 0.91 D C
velocities and corresponding spread for Data Set 2.
47.9 0.62 0.56 0.95 0.25 0.79 D D
As is shown Table 7 there is overall very high correl
49.9 0.43 0.5 0.87 0.69 0.86 D F
ation between the LS and RS results.
51.9 0.73 0.48 0.73 0.01 0.89 F F
53.2 0.71 0.51 0.89 0.45 0.73 F F
54.4 0.7 0.52 0.72 0.01 0.8 F F
55.7 0.63 0.5 0.46 0.94 0.7 F F
56.9 0.75 0.52 0.53 0.28 0.8 F F
58.9 0.71 0.59 0.89 0.52 0.51 D F
59.8 0.73 0.14 0.92 0.42 0.86 D D
The third data set demonstrates that for a data set

with poor STC and SHP values it is (generally) not
possible to obtain accurate SH interval velocity
estimates. Figures7 and 8 illustrate the LS and RS
traces for a real data set acquired during an offshore
Figure 7. Data Set 3 – real data set. VSP of data acquired DST investigation. The STC parameters are out-
on LS of seismic probe (200Hz low pass filter applied). lined in Tables 8 and 9, respectively.
113
The combination of very poor STC and SHP As illustrated in this paper, the initial assessment of
values strongly suggests that it is not possible to iso the SHP STC parameter when processing real data
late source wave response by implementing SFI. sets is very promising, although it is fully expected
And therefore this data set was indeed dropped and that the SHP ranking will be refined once the param
not analyzed. This shows the importance of having eter has been applied to additional data sets.
these parameters available in real-time during data In addition the authors intend to develop an
acquisition to ensure that the collected data is useful approach to incorporate the SHP ranking into the
and can be used to derive interval velocities. post-processing methodology for seismic data
4 CONCLUSIONS REFERENCES
ASTM (American Standards and Testing Methods). 2017.
Downhole Seismic Testing (DST) is an important D7400: Standard Test Methods for Downhole Seismic Test
geotechnical testing technique which provides esti ing. ASTM Vol. 4.09 Soil and Rock (II): D5877–latest.
mates of low strain (<10-5) shear and compression Baziw, E. 2002. Derivation of Seismic Cone Interval Vel
wave velocities, but there is a need for a widely ocities Utilizing Forward Modeling and the Downhill
accepted seismic trace characterization (STC) meth Simplex Method. Can. Geotech. J., 39(5),1181–1192.
odology of the acquired data. Baziw, E., and Verbeek, G. 2016. Frequency spectrum
In the past BCE had identified five independent “bell-curve” fitting as a component of SCPT interval
STC parameters (linearity estimates from the polar velocity accuracy assessment. 5th International Confer
ence on Geotechnical Site Characterization (ISC-5).
ization analysis, the cross correlation coefficient of
September 5- 9, 2016. Queensland – Australia. Austra
the full waveforms at the particular depth and the lian Geomechanics Society, 1431–1436.
preceding depth, the signal shape parameter, the Baziw, E., and Verbeek, G. 2017. Quality Assessment of
signal-noise-ratio and the peak symmetry differen Seismic Data Sets and the Impact on Interval Velocity
tial). In this paper, the mathematical and implemen Estimates in DST. Presented and published in the pro
tation details of a new STC parameter have been ceedings of the DFI 42nd Annual Conference on Deep
outlined. This new parameter, denoted as SHP, quan Foundations. October 24-27, 2017 - New Orleans, USA.
tifies the desired polarization of a horizontally gener Baziw, E. and Verbeek, G. 2018. The use of seismic trace
ated shear wave on the horizontal plane compared characterization to guide the analysis of DST results to
obtain more accurate soil parameters. presented and
with particle motion generated on the vertical plane.
published in the DFI 43rd Annual Conference on Deep
SH source waves should have negligible particle Foundations conference proceedings. October 24-27,
motion along the vertical axis. 2018 - Anaheim, CA.
114
© 2022 the Author(s), ISBN 978-1-032-31259-0
Methodology for obtaining true cone bearing estimates from blurred

and noisy measurements
Erick Baziw & Gerald Verbeek
Baziw Consulting Engineers Ltd., Vancouver, Canada
ABSTRACT: Cone penetration testing (CPT) is an important and widely used geotechnical in-situ test for
assessing soil properties and mapping soil profiles. CPT consists of pushing at a constant rate an electronic
cone into penetrable soils and recording the resistance to the cone tip or cone bearing (qm). These values
(after correction for the pore water pressure to get qt) are utilized to characterize the soil profile along with
measured sleeve friction and pore pressure. The qm measurements can have significant fluctuations when
penetrating sandy, silty gravelly soils, resulting in “high” peaks due to interbedded gravels and stones and
“low” peaks due to softer materials or local pore pressure build-up. Furthermore, the qm values are blurred
and/or averaged which results in the inability to identify thin layers and the distortion of the soil profile char
acterization. Baziw Consulting Engineers has invested considerable resources in addressing these two qm
measurement challenges. The qmKF algorithm was developed to address the additive measurement noise. In
this case the dynamics of qm are modeled within a state-space mathematical formulation and a Kalman filter
is then utilized to obtain optimal estimates of qm. The qmHMM algorithm implements a hidden Markov model
smoother so that true cone bearing are obtained from the averaged/blurred qm values. This paper outlines the
integration of the qmKF and qmHMM algorithms and demonstrates the performance first with test bed simula
tions (to show the functionality of the algorithm) and then through the analysis of various actual qm data sets.
1 INTRODUCTION relatively small diameter cone tip penetrating sandy,

silty and gravelly soils (Baziw and Verbeek, 2021b).
1.1 Cone bearing measurements
The Cone Penetration Test (CPT) is a geotechnical in- 1.2 Cone bearing smoothing/averaging
situ tool which is utilized to identify and characterize
The measured cone resistance at a particular depth is
sub-surface soils (Lunne et al., 1997; Robertson,
an averaged/smoothed measurement of the true
1990; ASTM D6067, 2017). In CPT a cone penetra
values qv above and below the depth of interest
tion rig pushes the steel cone vertically into the
(Boulanger and DeJong, 2018; Robertson, 1990;
ground at a standard rate and data are recorded at
Baziw and Verbeek, 2021a, 2021b). Mathematically
regular intervals during penetration. The cone has
the measured cone tip resistance qm is described as
electronic sensors to measure penetration resistance at
(Baziw and Verbeek, 2021a)
the tip (qm) and friction in the shaft (friction sleeve)
during penetration. A CPT probe equipped with
a pore-water pressure sensor is called a piezo-cone
(CPTU cones). For piezo-cones with the filter element
right behind the cone tip (the so-called u2 position) it
is standard practice to correct the recorded tip resist
ance for the impact of the pore pressure on the back
of the cone tip. This corrected cone tip resistance is
normally referred to as qt. The distortions which
effect the cone tip measurements are two-fold: 1) the where
cone tip resistance are smoothed/averaged (Boulanger d the cone depth
and DeJong, 2018; Baziw and Verbeek, 2021a) where dc the cone tip diameter
cone tip values measured at a particular depth are Δqm the qm sampling rate
affected by values above and below the depth of inter qm(d) the measured cone tip resistance
est, and 2) the cone bearing measurements are suscep qv(d) the true cone tip resistance (prior to pore
tible to anomalous peaks and troughs due to the water pressure correction)
DOI: 10.1201/9781003308829-9
115
wc(d) the qv(d) averaging function result from the penetration of interbedded gravels
v(d) additive noise, generally taken to be white and stones and the “low” peaks results from the
with a Gaussian pdf penetration of softer materials or local pore pressure
build-up. Figure 3 illustrates an example of a cone
In equation (1) it assumed that wc averages qv bearing profile which contains significant anomal
over 60 cone diameters centered at the cone tip. Bou ous/spurious qm data from approximate depths 10m
langer and DeJong (Boulanger and DeJong, 2018) to 18m and 22m to 30m. There is also significant
outline how to calculate wc below (after correcting pore pressure variability at these depths.
the equation for w1 (Baziw and Verbeek, 2021a)):
where:
z0 the depth relative to the cone tip normalized by
the cone diameter
z050 the normalized depth relative to the cone tip
where w1 = 0.5 C1
The cone penetration averaging function wc for
varying qt;z0 =qt;z0 ¼0 ratios is illustrated in Figure 1.
Figure 2. Schematic of anomalous/spurious cone bearing

data (after Mortensen and Sorensen, 1991).
Figure 1. Schematic of thin layer effect for a sand layer

embedded in a clay layer (Boulanger and DeJong, 2018).
1.3 Cone bearing measurement noise

The smoothed/averaged cone bearing measurement
qm given by eq. (1) can also contain sharp anomalous
and spurious peaks and troughs (Lunne, Robertson
and Powell, 1997)
These anomalous and spurious cone bearing
measurements are due to the relatively small diam
eter cone tip penetrating sandy, silty and gravelly Figure 3. Combined results of piezocone test and nuclear
soils. As illustrated in Figure 2, the “high” peaks density test at Gullfaks C in the North Sea (Tjelta et al. 1985).
116
2 FILTER FORMULATION maximum, minimum and resolution values are
specified. The HMM governing equations are
2.1 qmHMM algorithm formulation outlined in Table 1.
The qmHMM algorithm implements a BRE
The initial algorithm developed by Baziw and Verbeek smoother. BRE smoothing uses all measurements
(2021a) (the so called qmHMM-IFM) to address the available to estimate the state of a system at a certain
smoothing/averaging of cone bearing measurements time or depth in the qv estimation case (Arulampa
(eq. (1)) combined a Bayesian recursive estimation lam et al., 2002; Baziw and Verbeek, 2021a; Gelb,
(BRE) Hidden Markov Model (HMM) filter with Itera 1974). This requires both a forward and backward
tive Forward Modelling (IFM) parameter estimation in filter formulation. The forward HMM filter (q ^Fk ) pro
a smoother formulation. The qmHMM-IFM provided cesses measurement data (q ) above the cone tip
estimates of the true qv values from the measured � � m
(j ¼ 1 to 30 x dΔc ) in (1)). Next the backward
blurred values. In recent modifications and enhance- HMM filter (^ qBk ) is implemented, where the filter
ments of the qmHMM-IFM it was possible to drop the recurses through the� data below the cone tip
IFM portion of the algorithm. This was predominantly � � �
(j ¼ 30 x dΔc to 60 x dΔc in (1)) starting at the final
accomplished by refining the HMM filter parameters. qm value. The optimal estimate for qv is then
The HMM filter (also termed a grid-based filter) defined as
has a discrete state-space representation and has
a finite number of states (Arulampalam et al., 2002).
In the HMM filter the posterior PDF is represented Þ
by the delta function approximation as follows:
where the index k represents each qm measurement.
In both the forward and backward HMM filter for
mulation a bank of discrete qvvalues (i = 1 to N)
varying from low (qtL) to high (qtH) and
where xik-1 and wk-1jk-1
i
, i = 1,…,Ns, represent a corresponding qt resolution qtR is specified. The
the fixed discrete states and associated condi required number of fixed grid HMM states is given
tional probabilities, respectively, at time index as NS = (qtH - qtL)/qtR. In Table 1 the notation of the
k-1, and Ns the number of particles utilized. In states xi is mapped to qi to reflect the bank of
the case of the qmHMM algorithm the HMM dis qtvalues.
crete states represent possible qv values where In the qmHMM forward and backwardffilter fornmu
lation the transitional probabilities (i.e., p xik jxkj -1 or
f n
Table 1. HMM filtering algorithm. p qik jqjk-1 for each HMM state (i.e., discrete cone
Step Description Mathematical Representation tip, q,i) is set equal due to the fact that there is
equal probability of moving from a current cone
1 Initialization e:g:;wik � 1=Ns ; i ¼ 1; . . . ; Ns ð4Þ tip value to any other value between the range
(k=0) – initial qtL to qtH. The likelihood PDF pðzk jqik Þ in the
ize particle HMM filter outlined in Table 1 is calculated
weights. f n based upon an assumed Gaussian measurement
PNs
2 Prediction - wiknk-1 ¼ wjk-1nk-1 p xik jxk-1
j
ð5Þ
j¼1 error as follows:
predict the
weights.
wi pðzk jxik Þ
3 Update - wiknk ¼ PNsknk-1j ð6Þ
pðzk jxk Þ
j
w
update the j¼1 knk-1
weights.
P s i i where σ2is the variance of the measurement noise.
4 Obtain opti- ^xk ≈ Ni¼1 wkjk xk & (7)
mal minimum
Baziw and Verbeek (2021a) outline the details of the
PNs � � qmHMM algorithm forward and backward filter
variance esti- P^xk ≈ wikjk xki - ^xk ðxki - ^xk Þ T
mate of the i¼1 formulation.

state vector
and corres
ponding error
2.2 qmKF algorithm formulation

covariance.
The Kalman Filter (Gelb, 1974) is an optimal (least
5 Let k = k+1 squares) recursive filter which is based on state-
& iterate to space formulations of physical problems. Applica
step 2. tion of this filter requires that the physical problem
In the above equations it is required that the likelihood pdf
be modified by a set of first order differential equa
tions which, with initial conditions, uniquely define
pðzk jxi k Þ and the transitional probabilities pðxik jxjk-1 Þ be
known and specified.
the system behaviour. The filter utilizes knowledge
117
Table 2. KF governing equations.
Description Mathematical Representation
System equation xk ¼ Fk-1 xk-1 þ Gk-1 uk-1 (10)
Measurement equation zk ¼ Hk xk þ vk (11)

State estimate ^xkjk-1 ¼ Fk-1 ^xk-1jk-1 (12) The discrete system equation (eq. (10)) is given as
extrapolation
Error cov. extrapolation Pkjk-1 ¼ Fk-1 Pk-1jk-1 Fk-1 T þ
Gk-1 Qk-1jk-1 Gk-1 T (13)
Measurement ^zk ¼ Hk-1 ^xkjk-1 (14)
extrapolation
Innovation Δk ¼ zk - ^
zk (15)
Variance of innovation Sk ¼ Hk Pkjk-1 H Tk þ Rk (16)
Kalman gain matrix Kk ¼ Pkjk-1 Hk ðSk Þ-1 (17) where Δ is the qm sampling rate and aw, and
State estimate update ^xkjk ¼ ^
x�kjk-1 þ Kk �Δk (18) bw are the defining parameters of a first order

Error covariance update Pknk ¼ I - Kk Hk Pkjk-1 (19) Gauss-Markov (GM) process, The GM process
describes the cone bearing acceleration and wk is
In (10) and (11) vk and uk are i.i.d Gaussian zero mean white Gaussian noise with zero mean and unit
white noise processes with variances of Qk and Rk, respect variance.
ively (i.e.,vk � Nð0; Rk Þ and uk � Nð0; Qk Þ ).
2.4 Measurement model

of system and measurement dynamics, assumed
statistics of system noises and measurement errors Currently two synthesized measurements are incorp
and statistical information about the initial orated into the qmKF algorithm: 1) The best fit sev
conditions. enth degree polynomial to the qm profile and 2)
Table 2 outlines the KF governing equations. Output after applying a fourth order low pass fre
In Table 2 xk denotes the state to be estimated, quency filter to the qm profile. At a later date it is
Fk-1 denotes the state transition matrix which envisioned that additional measurements could be
describes the system dynamics, uk-1 the process incorporated into the qmKF algorithm as previously
or system noise (model uncertainty), Gk-1 described.
describes the relationship between xk and uk-1,
and Hk the relationship between the state and the
available measurement (measured cone resistance
qm). The KF can be applied to problems with
linear time-varying systems and with non-station
ary system and measurement statistics. The KF
can be implemented for estimation, smoothing
and prediction. A best fit 7thdegree polynomial is made to the qm
The motivation of implementing the KF for the measurements every 1m to 1.4 m depth increment
optimal removal of spurious cone bearing measure (allowed to be refined by investigator based upon data
ments is that it can use any number, combination under analysis) so that the anomalous and spurious
and sequence of external measurements. For peaks and troughs are minimized. This polynomial is
example, it is envisioned measurements from the then fed into the qmKF algorithm as a measurement.
vane shear test undrained strength could be incorp The order of the polynomial and depth increment were
orated within qmKF algorithm based upon empirical selected due to the averaging/blurring of the qm meas
correlations. Furthermore, it also fits into our goal urements where it would be highly unlikely that there
of implementation of data fusion techniques into would be greater than 6 turnings1 in a 1m to 1.4m
CPTU and SCPT. depth increment. This assumption was tested with
Baziw and Verbeek (2021B) present a thorough out extensive test bed simulation.
line of the qmKF algorithm. For completeness, the KF A 4th order 250Hz to 300 Hz (allowed to be
state and measurement equations are described below. refined by investigator based upon data under ana
lysis) Butterworth low pass frequency filter is
applied to the qm measurements measurement so that
2.3 System model the anomalous and spurious peaks and troughs are
To specify the qmKF systems equations in the stand minimized even further. This 250Hz low passed fre
ard KF state-space form, the following states need to quency filtered trace is then fed into the qmKF algo
be defined rithm as a measurement.
1 The maximum number of turnings of a polynomial function is always one less than the degree
118
3 IMPLEMENTATION OF QMHMM AND QMKF
ALGORITHMS
3.1 Test bed simulation

The performance of the qmHMM and qmKF algo
rithms were evaluated by carrying out challenging
test bed simulations. This section outlines one of
those simulations.
Figure 4 illustrates a simulation of thin bed layering
(0.2m) where there is alternating true qv values of
30MPa and 90MPa (light grey trace) interbedded
within a 50 MPa background layer. As is shown in
Figure 3 there is a resulting oscillation averaged/
smooth qm trace (black trace) with no sharp peaks or
troughs. The output (black dotted trace) of the qmHMM
algorithm is also illustrated in Figure 4. It shall be
obvious that the qmHMM algorithm performed well as
the derived qv/ values closely matched the originally Figure 5. Simulated cone bearing averaged/blurred qm
specified qv values. (black trace) of Figure 3, spurious qm trace (red trace) feed
into the qmKF algorithm, and the qmKF algorithm output
(blue trace).
Figure 4. Simulated true cone bearing measurements qv

(light grey trace) and corresponding averaged/blurred qm
(black trace) measurements. The qmHMM estimated qv/ Figure 6. Simulated true cone bearing measurements qv
trace (black dotted trace) is superimposed upon the true (light grey trace) and corresponding averaged/blurred qm
cone bearing values. (black trace) measurements. The qmHMM estimated qv/
trace (black dotted trace) is superimposed upon the true
cone bearing values.
Figure 5 illustrates the simulated qm data of Figure 4
(black) with additive noise to represent anomalous/
spurious qm data (red trace). The spurious data was algorithms results in obtaining impressive estimates
simulated with Gauss-Markov process noise (Baziw of true qv values from challenging qm data sets.
and Verbeek, 2021b) with a variance of 60 and time
constant of 0.1. The simulated Gauss-Markov noise
then had a 250Hz high pass filter applied. 3.2 Real data examples
Figure 6 illustrates the estimated qv (black dotted) After extensive test best analysis, the qmKF and
trace from the qmHMM algorithm after processing qmHMM algorithms were evaluated implemented on
the output of the qmKF algorithm (blue trace) of real data sets. Figures 7a, 7b and 7c show qm profiles
Figure 5. Superimposed on these traces is the true qv acquired by Perry Geotech Limited located at Tau
(light grey) trace of Figure 4. As is evident from ranga New Zealand. It is clear from the results pre
Figure 6, the combination of the qmKF and qtHMM sented in these figures that the effect of averaging/
119
smoothing of the true qv values (eq. (1)) can results 4 CONCLUSIONS
in a significant reduction in the recorded peaks of qv
values, which may very well impact the design The qmKF and qmHMM algorithms outlined in this
based on the CPT data. The qmKF and qmHMM algo paper can effectively minimize the anomalous and
rithms significantly minimize or undo this effect. spurious peaks and troughs to provide a more accur
ate depth profile of the cone tip resistance.
By applying these algorithms CPT will provide
a more realistic soil behavior profile and also
allow for more accurate identification of thin
layers. In turn it will provide more accurate input
data for any design process that uses the CPT
results as direct input.
REFERENCES
ASTM D6067/D6067M – 17 (2017), “Standard Practice
for Using the Electronic Piezocone Penetrometer Tests
for Environmental Site Characterization and Estimation
of Hydraulic Conductivity”, ASTM Vol. 4.09 Soil and
Rock (II): D5877-latest.
Arulampalam, M.S., Maskell, S., and Clapp, T. 2002.
A tutorial on particle filters for online nonlinear/non-
Gaussian Bayesian tracking. IEEE Transactions on
Signal Processing, vol. 50, no. 2, 174–188.
Baziw, E. and Verbeek, G. 2021a. Cone Bearing Estimation
Utilizing a Hybrid HMM and IFM Smoother Filter For
mulation. Accepted for publication within the Inter
national Journal of Geosciences (IJG) Special Issue on
Geoscientific Instrumentation, Methods and Data
Systems.
Baziw, E. and Verbeek, G. 2021b. Implementation of
Kalman Filtering Techniques for Filtering CPT Cone
Bearing Measurements. Accepted published in the DFI
46th Annual Conference on Deep Foundations confer
ence proceedings. October 12-15, 2021 - Las Vegas,
NV.
Boulanger, R.W. and DeJong, T.J. 2018. Inverse filtering
procedure to correct cone penetration data for
thin-layer and transition effects. Proc., Cone Penetra
tion Testing 2018, Hicks, Pisano, and Peuchen, eds.,
Delft University of Technology, The Netherlands,
25–44.
Gelb, A. (1974). Applied Optimal Estimation (4th Edition).
Cambridge, Mass: MIT Press.
Lunne, T., Robertson, P.K., and Powell, J.J.M. 1997. Cone
penetrating testing: in geotechnical practice. Taylor &
Francis.
Robertson, P.K. 1990. Soil classification using the cone
penetration test. Canadian Geotechnical Journal 27 (1),
151–158.
penetration test. Canadian Geotechnical Journal 27 (1),
151–158.
Tjelta, T. I., Tieges, A.W.W., Smits, F.P., Geise, J.M., and
Lunne, T. 1985. In-Situ Density Measurements by
Figure 7. Real data sets. qm (red trace), output from qmKF Nuclear Backscatter for an Offshore Soil Investigation.
(blue trace) and qmHMM estimated qv/ trace (black dotted Proc. Offshore Technology Conference, Richardson
trace). Texas, Paper No 40917
120
© 2022 the Author(s), ISBN 978-1-032-31259-0
Evaluation of parameters inducing desaturation of a piezocone: Saturation

liquid viscosity and exposure to dry sand
G. De Backer, R.D. Verastegui-Flores, W. Vervaele, L. Vincke & K. Haelterman
Department of Mobility and Public Works, Flemish Government, Ghent, Belgium
ABSTRACT: Several CPTU tests were performed in both ideal and adverse conditions at a test site in Zwij
naarde. The influence of exposure to unsaturated soils was examined by holding well-saturated piezocones in dry
sand for several minutes without a protective rubber membrane. After the exposure to dry sand, penetration was
performed from the ground level through the unsaturated and saturated subsoil. These tests show suboptimal pore
pressure measurements, compared to the tests under ideal circumstances, regardless of the duration of the exposure
to dry sand. Results reveal that, in adverse conditions, the densely packed sandy top layer, rather than the exposure
to dry sand, has contributed to the desaturation of the piezocones. The impact of the viscosity of the saturation
liquid on pore pressure quality was studied by performing tests with a 50 and 100 cSt silicone oil. Under ideal
circumstances, the viscosity does not play a role in terms of quality of the pore pressure measurements. However,
in adverse conditions, results showed a better performance for a silicone oil with 100 cSt viscosity.
1 INTRODUCTION In this paper, the influence of the viscosity of the

silicone oil on the quality of the pore water pressure
The repeatability of CPTU parameters has been evalu- measurements is evaluated. Also the influence of
ated in many research studies. Generally a reasonably exposure to unsaturated soils is evaluated by hold-
good repeatability is found for pore water pressure ing the penetrometer several minutes in dry sand.
measurements, provided good saturation is achieved
and maintained, even when performed with equipment
from different manufacturers, as described by e.g. 2 TEST SITE IN ZWIJNAARDE
Paniagua et al. (2021), Lunne et al. (2018), Powell
et al. (2005). Although dynamic pore water pressure The test site in Zwijnaarde is located in the south of
measurements might be a reliable parameter, it is Ghent, Belgium. The subsoil consists of Quaternary
known that many aspects can influence these measure- sand layers, with small sublayers of sandy silt to silt,
ments, such as the element location, design and volume as can be seen in Figure 1. The Quaternary layers
of ports, the type and degree of saturation of the fluids, have a thickness of approximately 17 m. With
cavitation of the element fluid system, resaturation lag a depth of 15 m, all CPTUs in this testing campaign
time, depth and saturation of soil during testing were performed in the Quaternary layers.
(ASTM 2012). The importance of properly saturating
a piezocone sensor has been elaborately documented in
literature (e.g. Lunne et al. 1997, ISO 22476-1:2012,
ASTM 2012). The need for de-airing the saturation
liquid in a vacuum chamber has often been
emphasized.
In our daily practice, we experience that even well-
saturated piezocones may produce poor pore water
pressure measurements, if penetration is started above
the groundwater level. This is generally avoided by
executing a pre-drilling. However, the problem still
occurs in cases where the groundwater level is not
exactly known or where the groundwater is to be found
at considerable depth. In these situations in presence of
unsaturated soils we noticed that the penetrometers
seem to be quite prone to desaturation. Figure 1. CPTU-data and soil classification.
DOI: 10.1201/9781003308829-10
121
In the sandy layers, the measured pore water pres sand for a certain time interval before executing
sures equal the hydrostatic pressure, since the sand is a CPTU test. The piezocones were protected with
permeable. When passing through the silty layers a rubber membrane which was removed right before
a drop in the pore water pressure can be noticed in entering a bucket of dry sand. The dry sand is
Figure 1. This reduction in pore water pressure is known as ‘Mol Sand’, characterized by a median
noticed for all pore pressure measurements and is particle diameter D50 of 0.2 mm and a uniformity
caused by dilative behavior of the dense silt layers, coefficient UC = D60/D10 of 1.6. Four different time
when being sheared during penetration. Although the intervals were considered: 5 min, 15 min, 30 min
tests are performed in a small area, there are some and 60 min.
spatial variations, both in terms of level and magni Immediately after exposure to dry sand, CPTU
tude of the peak cone resistance in the sand layers as tests were performed over a depth of 15 m. The
well as in the occurrence of the small silt layers, caus results of these pore pressure measurements in
ing pressure drops in the pore pressure measurements. adverse conditions are evaluated and compared
with CPTU tests performed in the same test field,
however, this time starting from the saturated sub
3 TEST CAMPAIGN AND CONE
soil (ideal conditions). To achieve this, these
PENETROMETER DATA
CPTUs were initiated from a pre-drilled borehole
of a depth of 1.5 m. PVC tubes along the shaft pre
The test campaign took place from March to
vented collapse of the borehole. The piezocone fil
April 2020. All tests were carried out within an area
ters were covered with a rubber membrane directly
of 10 by 25 m with a distance of approx. 1 m between
after saturation. Water was added to the drilling
two neighboring test points. Four identical piezocones
hole, while descending the piezocone in the bore-
from the same manufacturer were used. The piezo
hole. The additional water dissipated entirely in the
cones are of the compression type.
sandy top layers before penetration through the sat
The pore pressure is measured at the u2 location,
urated subsoil started. The CPTUs performed
which is just above the conical part. A silicone oil,
according to best practices are referred to as tests
with varying viscosity, was used as saturation fluid.
under ideal circumstances. Adverse conditions in
An overview of the piezometer characteristics is
the present research refers to exposure to dry sand,
given in Table 1.
absence of the protective rubber membrane and no
pre-drilling.
Table 1. Cone penetrometer data.
Cone type Compression 4.2 Viscosity

Ac 10 cm² The standard saturation liquid, as recommended
Filter type HDPE - 10 micron by the piezocone manufacturer, is a silicone oil
Capacity* qc 75 MPa with a viscosity of 50 cSt, combined with a filter
fs 1 MPa with a 10 μm pore diameter. In order to decrease
u2 2 MPa desaturation, a more viscous silicone oil of 100
Saturation fluid Silicone oil cSt was also used to saturate the piezocone.
Similar CPTU tests were carried out as with the
* Nominal values – penetrometers calibrated according to 50 cSt liquid.
EN ISO 22476-1 2012 class 2 Table 2 gives an overview of the parameters that
were varied in relation to the number of tests
The same saturation procedure has been used for all performed.
tests, according to EN ISO 22476-1 (2012). The sili
cone oil was de-aired in a vacuum chamber together
with the filters for a duration of at least 24 hours. In
a next step, the cone tip was filled with de-aired sili Table 2. Test program.
cone oil using a syringe. Thereafter, the cone was
assembled with a pre-saturated filter and placed back in Parameter Tested values Number of tests
the vacuum chamber for at least 2 hours at vacuum fol Viscosity 50 cSt 12
lowed by 20 minutes at atmospheric pressure before 100 cSt 12
removing and covering it with a protective rubber Duration in dry sand 0 min* 9
membrane. 5 min** 4
15 min 4
4 TEST PROGRAM 30 min 3
60 min 4
4.1 Exposure to dry sand
* Zero minutes in dry sand corresponds to ‘ideal
To examine the impact on desaturation, well- circumstances’.
saturated piezocones were kept in a bucket with dry ** For one test the duration was 7 min instead of 5 min.
122
4.3 Evaluation criteria For the ‘poor’ measurements, in Figure 2 indi
cated with a light blue solid line, a clear sluggish
All pore pressure measurements were evaluated and div
res-ponse is noticed and it is impossible to deter
ided into 4 quality categories: good, fair, mediocre and
mine the hydrostatic line from the pore pressure
poor. Evaluation criteria were: similarity to the hydro
measurements.
static pressure line and speed of response of the sensor.
Based on a borehole measurement, the water table
is at about 7.3 m above sea level, which corresponds
5 TEST RESULTS
to 1.5 to 1.7 m depth. At a depth of 15 m a short dis
sipation test was conducted for all soundings, resulting
in a slightly lower water table. Pore pressure measure 5.1 Comparison of measured data
ments show this slight shift in hydrostatic line too at Figure 3 shows the measured cone resistance qc,
approximately 0 to 1 m reference level, where a small sleeve friction fs and pore water pressure u2 for all
silty layer occurs. For this reason, an upper and lower tests, grouped according to their pore pressure quality
hydrostatic line is drawn in all pore water pressure category. The reference level is based on the depth
figures, indicated by u0-u and u0-l respectively. measurements corrected for inclination and elevation.
Figure 2 shows an example of pore pressure meas As expected, all tests performed under ideal
urements for each category. The blue dashed line rep circumstances show good pore pressure measure
resents a good measurement as it coincides well with ments. Remarkably in this group, one outlier has
the hydrostatic pressure lines and shows a quick been 60 minutes in dry sand and still shows no
response after a pressure drop, as can be seen at e.g. clear signs of desaturation. This is not the case
a level of 6.5 m. Some small peaks above the hydro for the other tests that have been a certain time
static pressure can be noticed between level -3 and in dry sand without the protection of a rubber
-5 m. Unfortunately, there are no cohesive soils at the membrane.
test site that would generate large and sustained posi Although not the purpose of this study, it is worth
tive pore pressures which would facilitate the evalu while to have a look at the qc and fs measurements
ation of the sensor response. The pore pressure drops as well. For the top layers the repeatability of qc is
in Figure 2 are not the result of a drop in penetration generally quite good. The values of fs show more
speed. A constant speed of 2 cm/s was achieved, scattering. Larger variations for fs compared to qc
using a continuous sounding technique. However, the are also found in literature (Paniagua et al. 2021,
drops agree well with the occurrence of small silty Lunne et al. 2018). Some anomalies also drew our
layers in the corresponding CPT-profile. attention: S112 (cone 2) and T59 (cone 3) show
The green dotted line represents a ‘fair’ pore pres lower qc and fs values, whereas T31 and S94 (both
sure measurement. Slightly slow response is noticed cone 4) give higher fs values. It is unclear what
after a pressure drop, for instance at around 5 m and might have caused these anomalies. Since the pore
-0.5 m. However, it still rejoins the good measure water pressure measurements don’t seem to be
ment line. The red dotted line shows the ‘mediocre’ affected, they were not excluded from the analysis.
results, for which the sensor shows even slower res Note that the less reliable pore pressure measure
ponse and a larger deviation from the hydrostatic line. ments cannot be attributed to a single underperform
ing piezocone, since all 4 piezocones showed very
similar pore pressure measurements in ideal circum
stances and because no correlation was observed
between any pore pressure quality category and
a particular piezocone.
5.2 Viscosity
Since all tests with zero exposure to dry sand
produced good pore pressure measurements,
regardless of the viscosity of the silicone oil, it
can be stated that the viscosity has no influence
on the results, if they are performed under ideal
circumstances.
However, when the piezocone is exposed to dry
sand and subjected to relatively high suction pressures
during shearing, it seems that the viscosity does play
a role. Based on the results depicted in Figure 4, it is
likely that a 100 cSt viscosity might withstand better
to desaturation than a 50 cSt viscosity. However, the
Figure 2. Examples of 4 quality categories of pore pressure results should be interpreted with caution given the
measurements versus reference level. relatively small number of tests. Further research is to
123
Figure 3. Measured qc, fs and u grouped by quality category based on the pore pressure measurements (a) good, (b) fair,
(c) mediocre and (d) poor.
124
sand, still produced good pore pressure measure
ments. During penetration this test experienced
rather small qc values (max. 6.5 MPa) just above
the water table and consequently small suction
pressures were measured (up to -15 kPa), which
is in the same order of magnitude as for the
other tests producing good results.
Figure 4. Percentage of tests under adverse conditions as

a function of pore pressure quality categories for 100 cSt
and 50 cSt viscosity of silicone oil.
be carried out to better understand the advantages and

possible drawbacks of using a more viscous sili
cone oil.
5.3 Exposure to dry sand

The impact on the quality of the pore pressure
measurements as a function of the time of contact
with dry sand is presented in the bubble plots of
Figure 5 for both silicone oil viscosities, 50 cSt
and 100 cSt.
Each bubble shows the number of tests for
a certain combination of pore pressure quality cat
egory and number of minutes in dry sand. One
would expect a descending trendline showing
decreasing quality in the pore pressure measure
ments for increasing minutes in dry sand. However,
this is not the case, especially not for the plot with
100cSt. There is no clear relation between pore pres
sure quality and contact duration in dry sand. Only
zero minutes in dry sand results in an obvious link
with good quality measurements.
During pore water pressure observation tests in
dry sand, no or very small suction pressures
smaller than the measurement accuracy- were
measured. On the other hand, significant negative
pore water pressures were observed going from
-30 to -50 kPa during penetration for the tests in
adverse conditions. For this reason, it seems
likely that the negative pressures, occurring
shortly after penetration, have been more detri
mental to the saturation of the piezocone than the
static contact with dry sand. At small depth fairly
high qc values are registered (up to approx. 10 Figure 5. Pore pressure quality category in relation to the
MPa), indicating a high relative density of the exposure to dry sand for 50 cSt and 100 cSt silicone oil.
sand layer. This sand dilates during penetration
and it mobilizes negative pore pressures. This
might explain why the contact time in dry sand 6 CONCLUSIONS
seems not to be decisive on the quality of the
pore pressure measurement and moreover, it Twenty-four CPTUs were performed at the test site in
might explain why the outlier (CPTU S93), Zwijnaarde in both ‘ideal’ and ‘adverse’ circumstances
which remained 60 minutes in contact with dry in order to compare pore pressure measurements.
125
The tests have been performed for two differ ACKNOWLEDGEMENTS
ent values of viscosity of the saturation liquid,
i.e. 50 and 100 cSt. The results showed that, The authors would like to thank David Fraeyman
under optimal conditions, good pore pressure and Andy Fraeyman for preparing and performing
measurements are observed, regardless of the all field tests and for their valuable contribution in
viscosity. However, in adverse conditions, the optimizing the execution process.
results suggest that cone penetrometers saturated
with 100 cSt silicone oil might withstand better REFERENCES
to desaturation compared to those saturated with
50 cSt silicone oil. Further research is to be car ASTM D5778–12 2012. Standard Test Method for Perform
ried out to evaluate the impact of the viscosity ing Electronic Friction Cone and Piezocone Penetration
of the silicone oil on desaturation. Testing of Soils, ASTM International. www.astm.org.
The influence of exposure to unsaturated soils ISO 2012. Geotechnical investigation and testing – Field
was simulated by holding initially well-saturated testing – Part 1: Electrical cone and piezocone penetra
piezocones in contact with dry sand for several tion tests, International Standard ISO 22476-1.
Lunne T., Robertson P.K. & Powell J.J.M. 1997. Cone
minutes. During their exposure to dry sand the
Penetration Testing in geotechnical practice. Taylor &
piezocones were not protected by a rubber mem Francis Ltd. ISBN 0 419 23750 X.
brane. All but one piezocones showed clear signs of Lunne T., Strandvik S., Kåsin K., L’Heureux J.-S., Haugen E.,
desaturation during penetration, even when exposed Uruci E. & Kassner M. 2018. Effect of cone penetrometer
briefly to dry sand. Since no or very small suction type on CPTU results at a soft clay test site in Norway.
pressures were generated during the pore pressure Proceedings of the 4th International Symposium on Cone
observation tests in dry sand, it is likely that mainly Penetration Testing CPT 2018: 417–422. Delft.
the negative pressures, associated with dilation Paniagua P., Lunne T., Gundersen A., L’Heureux J.-S. &
during penetration of the sandy top layer, have Kåsin K. 2021. CPTU results at a silt test site in
Norway: effect of cone penetrometer type. IOP Conf.
adversely impacted the saturation rather than the
Ser.: Earth Environ. Sci. 710 012010.
exposure to dry sand. In a subsequent study it Peuchen, J. & Terwindt J. 2014. Introduction to CPT
would be interesting to separate the effect of expos accuracy. Proceedings of the 3rd International Symposium
ure to dry sand and initial penetration through the on Cone Penetration Testing, Las Vegas, Nevada, USA.
unsaturated dense sandy layer, by also performing Powell, J.J.M. & Lunne, T. 2005. A comparison of different
a pre-drilling for the penetrometers exposed to dry piezocones in UK clays. Proceedings of ISSMGE Con
sand. ference, Osaka.
126
© 2022 the Author(s), ISBN 978-1-032-31259-0
Large diameter cone penetrometers: What is an appropriate location for the

transition to the rod diameter?
D.A. de Lange, T.A. van Duinen & D.J. Peters
Deltares, Delft, The Netherlands
ABSTRACT: 15 cm2 cone penetrometers are typically pushed to depth by means of 10 cm2 rods. Therefore,
a transition in diameter will be present at a certain distance from the cone shoulder. Although the standards
give a requirement for the position of this transition, cone penetrometer configurations that don’t meet this
requirement are often preferred by CPT sounding companies and are being used in practice. This raises
a dilemma. In order to fuel discussion, a first step was made by performing two series of 9 CPTs with different
penetrometer geometries. The effects of the penetrometer designs on the measured quantities were investi
gated. From a preliminary analysis of the results no systematical differences between the different penetrom
eters were observed. When these preliminary findings are better understood and validated, the standards can
be updated.
1 INTRODUCTION 2. What was the basis for the requirements in the

testing standards?
From perspective of a CPT sounding company it is
beneficial to have a system that has enough resist Some companies refer to Powell & Lunne (2005)
ance against breakage of the pushing rods and and claim that the measured quantities are hardly
allows deep penetration without the need of an affected by the shorter distance from shoulder to
excessive load. For Dutch practice, most commonly transition. Powell and Lunne discuss tests in UK
utilized cone penetrometers have a projected base clays, utilizing 15 cm2 piezocones with sleeve areas
area of 10 cm2 (diameter, D = 3.6 cm) or 15 cm2 of 200 and 300 cm2. They conclude there is little
(D = 4.4 cm). However, in general similar pushing difference in results between the different sleeve
rods are used for both cone types, which implies the areas. However, the distances from shoulder to tran
sition are not mentioned. Heijnen (1972) presents
need for a transition from the wider penetrometer to
the smaller pushing rods in case of a 15 cm2 cone extensive field research on the shape of the cone
penetrometer. According to NEN-EN-ISO 22476, penetrometer (10 cm2), as different results in qc
the lowest point of the transition should be at least were observed from a “Dutch mantle cone” and
11.2D (= 49 cm) above the cone shoulder. In prac a “straight electrical cone” in sands. Heijnen con
tice, however, we notice that companies do prefer cluded that a reduction in diameter behind the cone
and use configurations in which the transition is might influence qc significantly, but that such
much closer to the cone shoulder (±23 cm or about effects also rapidly decrease with distance from the
5D). The main reason for this deviation from the cone shoulder. Interestingly, no significant differ
standards is reduction in required force to reach ences were observed between a “straight” cone
depth, as the configuration acts like a friction (constant diameter over full length) and a cone with
a reduced diameter from 10 cm above the cone
reducer. Further, smaller inclination deviations are
observed with a 15 cm2 cone compared to a 10 cm2 “tip” (2.8D). However, we believe that such conclu
cone. Therefore, it is understandable that CPT sions should treated with caution due to natural
sounding companies prefer such a system over sys variability of the subsoil and measurement
tems that meet the standards. uncertainty.
These observations lead to the following Most likely, the basis for the requirements in
questions: NEN-EN-ISO 22476 are the classical analytical
solutions for the bearing capacity of footings, like
1. What are the effects of the penetrometer configur the Prandtl-wedge. The size of the failure wedge is
ation on cone resistance, qc, sleeve friction, fs, typically a function of the friction angle of the soil
and/or pore water pressure? and can extend to great distances from the cone
DOI: 10.1201/9781003308829-11
127
shoulder, e.g. Figure 1. However, this seems to
contradict with the pile base shape factors from
NEN 9997-1+C2, see Figure 2. For instance, in case
of a 15 cm2 cone with 10 cm2 rods (Deq/deq = 1.5),
only H ≥ 2D is required to become comparable with
a fully straight pile concerning the tip resistance in
sands.
From above findings it might be expected that the
different configurations of the 15 cm2 cone as we
come across in the field will not significantly influ
ence qc. However, it is important to validate this
with empirical evidence. Further, the potential influ
ence on the friction readings is less clear so far, as
the penetrometer geometry also might affect the
horizontal soil stresses around the sleeve. Deltares
proposed a series of CPTs using 15 cm2 cone pene Figure 2. Pile base shape factor β for piles in sand (NEN
trometers having the transition at different distances 9997-1+C2).
from the shoulder. This paper discusses the test
details, results, and the next steps required to solve
this issue.
2 MATERIALS AND METHODS
Two test series were performed in April and

December 2021. In total 18 CPTs were executed at
a green field, that has been empty for years, adja
cent to the Deltares campus, Delft, The Nether
lands. Four different 15 cm2 penetrometer
configurations have been used. Configurations
A and B do meet the requirements of NEN-EN
ISO 22476 with respect to the location of the tran
sition, while configurations C and D don’t. The
cone dimensions are given in Figure 3 and Table 1.
As can be observed, the area and the location of
the friction sleeve was near identical for all pene
trometers. All penetrometers were manufactured
by the same company and all tests were executed
by the same company. The locations and the order
of the CPTs are given in Figure 4. E.g. “A2” refers
to penetrometer type A and the second CPT of the
test series. The target penetration level was
25 m below surface level for all tests, passing
through Holocene clay and peat deposits and Pleis
tocene sand. Pore water pressures were not
measured.
Table 1. Penetrometer characteristics.
A B C D
Penetrometer cm cm cm cm
Shoulder – transition distance 52 52.5* 31.5 19
Sleeve – transition distance 34 34.5* 13.5 1
Transition length 1 - 5 8
Figure 1. Logarithmic spirals (Van Mierlo & Koppejan
1952). * Distance to friction reducers is given.
128
Figure 3. Schematization of utilized 15 cm2 penetrometers.
Figure 5. Results from A6 with respect to other type

A results.
Figure 4. CPT locations and order.
3 PRELIMINARY RESULTS
The results are presented in Figures 5-14. Total cone

inclinations of about 1° to 3° were measured at final
depth. Therefore, interference between the tests are
excluded. Figure 5 shows the qc-profiles from all type-
A tests to indicate the inherent variability. The results
of the central CPT are emphasized. Figure 6 also pre
sents the results of this central CPT, but now with
respect to the results of all type-C tests. Similar plots
are given for fs, see Figures 7-8. qc- and fs-profiles as Figure 6. Results from A6 with respect to type C results.
129
Figure 7. Results from A6 with respect to other Figure 8. Results from A6 with respect to type-C results.
type-A results.
Figure 9. Results from D5 with respect to other type

Figure 10. Results from D5 with respect to type B results.
D results.
130
Figure 11. Results from D5 with respect to other type Figure 12. Results from D5 with respect to type B results.
D results.
Figure 13. Results from D5 with respect to type B results Figure 14. Results from D5 with respect to type B results
(zoomed in on the soft Holocene layers). (zoomed in on the soft Holocene layers).
131
being measured during the second test series are given ACKNOWLEDGEMENTS
in Figures 9-14. From the results, no systematical dif
ferences were observed, neither for qc nor fs, neither in The authors would like to acknowledge Gouda
clay nor peat nor sand. Geo-Equipment for making the different cone
penetrometers available, Inpijn Blokpoel Inge
nieurs for execution of the CPTs and Delft Uni
4 CONCLUDING REMARKS versity of Technology for allowing us to use
their land for this first investigation.
In practice, different penetrometers with large cone
diameter are used, several of which don’t meet the
requirements of the standards. A range has been tested
in the presented first investigation. The location of the REFERENCES
diameter transition ranged from 4.4 to 12 times D from Heijnen, W.J. (1972). De vorm van de elektrisch
the cone shoulder. From a preliminary analysis of the sondeerconus. Verhandelingen Fugro Sondeersympo
results no systematical differences were observed. This sium 1972: 17–27 (in Dutch).
is an important finding. As a first next step, the results NEN-EN-ISO 22476-1:2012 en Geotechnical
can be analyzed in more detail. Further, we foresee the investigation and testing - Field testing
need to consult archives and involved people to under Part 1: Electrical cone and piezocone penetration test.
stand the background and motivation of the adopted NEN 9997-1+C2:2017 Geotechnical design of structures
standards. In order to achieve a broader basis and have Part 1: General rules (in Dutch).
Powell, J.J.M. & Lunne, T. (2005). A comparison
a better statistical support of present results, tests should
of different sized piezocones in UK clays. Proceed
be repeated at other locations, in other soil types. For ings of the 16th International Conference on
better understanding of the mechanisms, a combination Soil Mechanics and Geotechnical Engineering:
of numerical (e.g. MPM) and physical modelling can be 729–734.
applied. Lastly, should you want to contribute, any rele Van Mierlo, W.C. & Koppejan, A.W. (1952). Lengte en
vant information from your side is more than welcome! draagvermogen van heipalen. Bouw 19-1-1952, no. 3:
Please contact the corresponding author. 1–11 (in Dutch).
132
© 2022 the Author(s), ISBN 978-1-032-31259-0
Issues related to piezocone sleeve friction measurement accuracy in soft

sensitive clays
B. Di Buò & M. D’Ignazio
Tampere University / Ramboll Finland Oy, Tampere, Finland
T. Länsivaara & M. Haikola

Tampere University, Tampere, Finland
ABSTRACT: Over the past decades, the piezocone testing has been increasingly adopted for field investiga
tion as it offers a quick and cost-effective methodology for subsoil profiling and geotechnical parameters esti
mation. Although the piezocone testing has revealed good applicability worldwide, difficulties are
encountered in complex soil conditions, such as soft sensitive clays. One of the key issues in such soils is
represented by the poor quality of the sleeve friction (fs) measurement due to inaccuracy and poor resolution
of the sleeve sensor. This paper investigates the influence of fs data quality on soil parameters determination
with particular emphasis on the soil behavior type (SBT) chart classification. The field investigation was con
ducted in a soft sensitive clay site located in Finland using two different penetrometers: a standard piezocone
and an advanced piezocone characterized by enhance accuracy sleeve friction sensor. Results show that the
use of high-resolution piezocone plays a key role in soft clays to avoid misleading soil type classification.
1 INTRODUCTION
1.1 Cone penetration testing (CPT)

Cone Penetration Testing (CPT) is a fast and reliable
means of conducting subsurface investigation for soil
profiling, site characterization and geotechnical
parameters evaluation. Since the first introduction of
the mechanical probe in 1930s, the equipment has
been significantly improved by adding porous filters
and transducers, obtaining the modern electronic
piezocones (CPTu). Nowadays, the traditional piezo Figure 1. Accuracy and precision definition (source
cone testing provides three independent readings with Wikipedia).
depth: the cone tip resistance (qc), the sleeve friction
(fs), and the excess pore water pressure (u2) measured
behind the cone tip (u2). The qc value should be cor This is crucial in soft sensitive clays as the value of
rected (qt) to consider the pore pressure acting behind fs is generally very low (<1 kPa).
the shoulder. This correction is significant in soft to The equipment available on the market are charac
stiff clayey soils (Jamiolkowski et al. 1985, Robert terized by different features such as dimensions,
son and Campanella 1988, Lunne et al. 1997). tolerances, and sensor accuracy. However, all the
The data provided by piezocone test is generally specifications, technical requirements, and test
characterized by high precision and accuracy. The procedures are outlined in the European Standards
sensor accuracy is the difference between the target (EN-ISO 22476-1) and American Standards (ASTM
and the measured value while the precision refers to D5778 - 20). All these aspects are extensively dis
the degree of reproducibility of a measurement. cussed by (Lunne at al. 1997). Although several piezo
These aspects are summarized in Figure 1. cones can be employed for CPTu testing, it is
Another aspect that plays a key role in data qual fundamental to choose the appropriate equipment to
ity is the sensor resolution which is the smallest obtain high-quality and reliable data, especially in soft
detectable incremental change that can be measured. sensitive clays. As an example, studies conducted by
DOI: 10.1201/9781003308829-12
133
Tampere University (Di Buò et al. 2016, Di Buò 2020) Table 1. Technical features of the piezocones.
have revealed that standard penetrometers are not suit
able to correctly measure fs in Finnish soft soil, which Standard penetrometer
negatively affects the data interpretation and the accur
acy of soil profiling as the small variability of fs qc fs u2 Inclination
throughout the entire deposit cannot be detected. Maximum 7.5 MPa 0.15 MPa 2 MPa 20°
Moreover, the fs is fundamental for the soil classifica capacity
tion based on the Soil Behavior Type (SBT) chart pro Accuracy 0.2% 0.7% 0.25% 0.5°
posed by Robertson (1990) or for the assessment of (15 kPa) (1 kPa) (5 kPa)
a number of geotechnical parameters, such as the soil
sensitivity (Mayne 2014). Despite the importance of Advanced penetrometer
the fs data, its correct measurement is still considered
one of the main challenges in soft sensitive clays. qc fs u2 Inclination
This paper presents the CPTu test results obtained Maximum 50 MPa 1.6 MPa 2.5 MPa 20°
from a soft sensitive clay test site located in Pohja, capacity
Southern Finland. The soundings have been per Accuracy 0.01% 0.0025% 0.0015% 0.5°
formed using a “standard” probe and a penetrometer (5 kPa) (0.04 kPa) (0.04 kPa)
characterized by enhanced sensor accuracy and reso
lution, herein after referred to as “advanced”. The
results are analyzed and compared, pointing out the
influence of fs measurement on the SBT classifica
tion and the improvements that can be obtained measurement. Similarly, the u2 sensor consists of
using high accuracy sensors. The main goal of the a silicon piezoresistive load cell in Wheatstone bridge
study is to investigate the influence of the piezocone configuration.
sensor accuracy on the soil interpretation, rather than
a comparison between cone manufacturers. 1.3 Test site
The Pohja site locates on the southern coast of Finland
1.2 Equipment along the railway line connecting the cities of Helsinki
The two penetrometers adopted in this study were pro and Turku. The ground elevation is around 9 m above
vided by two different manufacturers. Both cones are current sea level and the overall topography is flat.
characterized by 60° apex angle, 10 cm2 base and However, on a wider scale, the site locates in a valley
150 cm2 sleeve area. The first one, referred to as surrounded by shallow hills with bedrock outcrops.
“standard”, has been largely used for site investigation The Pohja site subsoil consists of about 1.5 m thick dry
in Finnish soft clays as detailed by Di Buò et al. crust layer overlaying a soft slightly consolidated clay
(2020): it consists of an electronic instrumented probe layer. Below. silty and sandy layers are encountered.
with a nominal range of 7.5 MPa, which is particularly The bedrock is located at depth of 14 m. The water
suitable for soft soils investigation. The second, content decreases with depth, from 130% between 2m
referred herein after as “advanced”, is characterized by and 4m till reach 60% at about 7.5 m depth. The nat
a nominal range of 50 MPa and higher accuracy com ural water content exceeds the liquid limit throughout
pared to the previous cone. This cone has been used at the entire deposit. Plasticity index is 30–60 and sensi
the Pohja site as an attempt to overcome the issues tivity, defined as the ratio between intact (su) and
related to the fs measurement previously discussed. remoulded undrained shear strength (su,rem), varies
The technical features of both piezocones are summar between 40 and 60 below the dry crust layer. The
ized in Table 1. The enhanced accuracy of the undrained shear strength evaluated by the fall cone
advanced penetrometer is the result of the embedded (FC) test is 7 kPa under the dry crust layer and
sensors type and their configurations. In particular, the increases linearly with depth reaching 12 kPa at the
qc and fs load cells are characterized by four strain depth of 7.5 m. The geotechnical properties of Pohja
gauges wired into a Wheatstone bridge configuration. clay are shown in Figure 2.
The bridge is compensated with four modules: two for
the material elastic modulus temperature compensation,
one for the zero-offset correction and the last one for
zero-offset temperature compensation. The Wheatstone
Bridge circuit consists of two simple series-parallel
arrangements of resistances connected between
a voltage supply terminal and ground producing
zero voltage difference between the two parallel
branches when balanced. It has two input terminals and
two output terminals consisting of four resistors
configured in a diamond-like arrangement. This config
uration allows for high accuracy in the parameter Figure 2. Index properties of Pohja clay.
134
2 CPTU DATA ANALYSIS 2.2 Soil Behavior Type (SBT) chart
Since its first introduction, the CPT has been widely
2.1 CPTu soundings used for soil profiling and classification. Several
A total of three CPTu soundings are performed, two authors have proposed classification charts that link
of them obtained from the advanced piezocone and the CPT parameters (qt and fs) to the soil type (Bege
one representative vertical from the standard cone. mann 1965, Robertson et al. 1986, Robertson 1990).
The main purpose herein is to investigate the impact Among them, the soil behavior type (SBT) chart pro
of the sensor accuracy in the measurement readings. posed by Robertson et al. 1986 has become quite
The same testing procedure has been adopted for popular. It identifies 12 types of soil based on the qt
both penetrometers. An initial pre-drilling is made to and fs values. This approach is mainly based on
avoid pushing the cone into the dry crust layer which classifying the soil based on the in-situ behavior
may cause the desaturation of the porous stone. Prior which depends on the strength, stiffness and com
to CPTu sounding, the cone is placed into the hole pressibility. In contrast, the unified soil classification
filled with water for temperature balancing. Then, system (USCS) does not provide any information on
the apparatus is pushed into the soil at a standard the mechanical behavior as it is based on the grain-
rate of penetration of 20 mm/s till reaching the size distribution and plasticity. However, in most
coarse layer (≈ 7 m). The measured qc has been fur cases, both approaches agree fairly well as detailed
ther corrected to account for the pore water pressure by Molle (2005). The SBT chart has been further
acting behind the cone as follows: improved introducing the normalized parameters:
where a is the net cone area ratio provided by the

manufacturer after the calibration process. The
soundings are conducted using a ceramic filter elem
ent replaced after each test. The saturation is ensured
by submerging the cone tip into a silicon oil bath in
a vacuum device.
Results illustrated in Figure 3 indicate the presence
of a homogeneous clay layer from 1 m to 7 m depth where Qt is the normalized cone tip resistance, Fr is
followed by interlayers of silts and sands. It is worth the normalized friction ratio, σv0 is the total overbur
observing that both piezocones provide nearly identical den vertical stress, σ’v0 is the vertical effective
response in terms of qt and u2 while fs measurements stress, u0 is the hydrostatic pore water pressure. The
made using the standard cone are characterized by normalized SBT (SBTn) chart is more reliable as
poor accuracy and low resolution. In particular, higher it takes into account the influence of the in-situ
fs values are measured by the standard cone (>4 kPa) stress in the soil classification which is particularly
while the advanced cone indicates fs values lower than important when the sounding is performed at
1 kPa. Even though this difference may seem neglect- great depths. The charts are illustrated in Figure 4.
able, the total error is significant, and it has a great
impact in the SBT classification as soil parameter esti
mation as discussed later.
Figure 4. Normalized SBT chart proposed by Robertson

1990.
More recently, Robertson et al. (1998) introduced

a normalized cone parameter with a variable stress
Figure 3. CPTu soundings at Pohja site. exponent (n), defined as:
135
where pa is the atmospheric pressure (≈100 kPa) and
Ic is the SBT index first introduced by Jeffries and
Davies (1993) and further modified by Robertson
(1990) as:
Figure 6. Proposed updated SBTn chart based on Qtn–Fr

(Robertson 2016).
The normalized SBT chart is shown in Figure 5

with the indication of the Ic values for the different
soil type regions.
Figure 7. Soil behavior type (SBTn) chart based on stand

ard CPTu cone data.
Figure 5. Contours of Ic on a normalized soil behavior type

(SBTn) chart (Robertson 1990).
Figure 8. Soil behavior type (SBTn) chart based on the S-1
sounding (advanced cone).
Recently, Robertson (2016) proposed an updated

version of the SBT chart to capture the contractive-
dilative soil behavior (Figure 6).
In this study, the original SBT chart proposed by
Robertson (1990) is taken into account as the main pur
pose is to evaluate the soil sensitivity instead of the
contractive-dilative soil behavior. It is worth observing
that soft sensitive clay region is located at the bottom
left of the Qtn and Fr chart, or in the bottom right of the
Qt – Bq chart (Figures 7, 8, and 9). Clearly, the correct
evaluation of the fs is fundamental for a correct evalu
ation of the SBT. To investigate this aspect, the CPTu
soundings performed with the two penetrometers have Figure 9. Soil behavior type (SBTn) chart based on the S-2
been assessed separately for soil classification by using sounding (advanced cone).
136
the SBTn charts. Results are presented in Figures 7,8, 4 DISCUSSION AND CONCLUSIONS
and 9. As expected, in the Qtn- Fr chart, the data
obtained using the advanced cone fall almost entirely The presented study investigated the influence of the
in the sensitive clay region while the data points pro sleeve friction fs measurement accuracy and reso
vided by the standard cone fall between the clay and lution on the interpretation of piezocone data in soft
organic clay regions. The reason for this lies obviously sensitive clays. The study is limited to a single test
in the overestimation of fs from the standard cone site located in Pohja, Southern Finland. It has been
which negatively affected the SBT evaluation. In con shown that in these soils fs values are significantly
trast, the Qtn – Bq chart provides more reliable SBT low (<1 kPa) and, therefore, the sleeve sensor should
classification in both cases, even though most of the have sufficient resolution to perform accurate meas
standard cone data points appear to fall in the clay urements. As shown for the Pohja site, traditional
region. piezocones with fs resolution >1 kPa may provide
misleading results, thus inducing to an incorrect clas
sification of the soil type based on the SBT charts.
3 SENSITIVITY EVALUATION BASED ON Moreover, the fs appears to be a key parameter for
CPTU DATA a reliable estimation of the soil remoulded strength
and sensitivity. Therefore, when performing CPTu
In addition to the SBT classification, the fs is used in soundings in soft sensitive clays, it is highly sug
a number of correlations for soil parameters estima gested to adopt suitable piezocone equipment with
tion. Several authors (Schmertmann 1978, Robertson accurate and precise sleeve friction sensor for
and Campanella 1988, Lunne et al. 1997, Robertson a reliable and robust data interpretation
2006) tried to obtain the soil sensitivity based on fs The use of high-accuracy site investigation equip
or the normalized friction ratio (FR). From ment is a key aspect for geotechnical risk assessment
a theoretical point of view, it is reasonable to correl in soft sensitive clay areas. This aspect is relevant in
ate the fs data with the remoulded shear strength relation to detection of sensitive clay layers that may
(su,rem) obtained from the fall cone test. As shown in trigger progressive failure during e.g. excavation
Figure 10, the fs measurement provided by the works, or, as often observed in Norway or Canada,
advanced cone fits fairly well with the su,rem data may induce large landslides because of human activ
(Figure 10) while the standard cone does not catch ity in the area or by other natural phenomena.
the trend.
REFERENCES
ASTM, D. 3441, 1986. Standard test method for deep qua
sistatic, cone and friction-cone penetration tests of soils:
414–419.
Lunne, T., Robertson, P.K. & Powell, J.J.M. 1997. Cone
penetration testing in geotechnical practice. London,
Spon Press. 312p
Di Buò, B. 2020. Evaluation of the Preconsolidation Stress
and Deformation Characteristics of Finnish Clays based
on Piezocone Testing. PhD Thesis, Tampere University,
Tampere, Finland. ISBN 978-952-03-1468-2
Di Buò, B., D’Ignazio, M., Selänpää, J. & Länsivaara, T.
2016. Preliminary results from a study aiming to
improve ground investigation data. Proceedings of the
17th Nordic Geotechnical Meeting: 187–197.
Di Buò, B., Selänpää, J., Länsivaara, T., & D’Ignazio, M.
2018. Evaluation of existing CPTu-based correlations
for the deformation properties of Finnish soft clays. In
Cone Penetration Testing 2018 (pp. 185–191). CRC
Press.
Jefferies, M.G., & Davies, M.P. 1993. Use of CPTU to esti
mate equivalent SPTN 60. Geotechnical Testing Jour
nal, 16(4), 458–468.
ISO, E. 22476-1, 2009. Geotechnical investigation and test
ing. Field testing. Part, 1.
Jamiolkowski, M., Ladd, C.C., Germaine, J.T. &
Lancellotta, R. 1985. New developments in field and
Figure 10. Comparison between the remoulded shear laboratory testing of soils. Proceedings of the 11th Int.
strength (su,rem) measured from the fall cone test and the Conf. on Soil Mech. and Found. Engineering, San Fran
sleeve friction from CPTu soundings. cisco, Vol. 1, pp. 57–153.
137
Robertson, P.K. 1990. Soil classification using the cone Robertson, P.K., & Wride, C.E. 1998. Evaluating cyclic
penetration test. Canadian Geotechnical Journal, 27(1), liquefaction potential using the cone penetration test.
151–158. Canadian geotechnical journal, 35(3), 442–459.
Robertson, P.K. 2009. Interpretation of cone penetration Schmertmann, J. H. 1978. Guidelines for cone penetration
tests—a unified approach. Canadian Geotechnical Jour test: performance and design (No. FHWA-TS-78-209).
nal, 46(11), 1337–1355. United States. Federal Highway Administration.
Robertson, P.K. 2016. Cone penetration test (CPT)-based Selänpää, J., Di Buò, B., Haikola, M., Länsivaara, T., &
soil behaviour type (SBT) classification system - an D’Ignazio, M. 2018. Evaluation of existing CPTu-based
update. Canadian Geotechnical Journal, 53(12), correlations for the undrained shear strength of soft
1910–1927. Finnish clays. In Cone Penetration Testing 2018 (pp.
Robertson, P.K., and Campanella, R.G. 1988. Guide xxx–xxx). CRC Press.
lines for geotechnical design using CPT and CPTU. Zhang, G., Robertson, P. K., & Brachman, R. W. 2002.
University of British Columbia, Vancouver, Depart Estimating liquefaction-induced ground settlements
ment of Civil Engineering, Soil Mechanics Series, from CPT for level ground. Canadian Geotechnical
120. Journal, 39(5), 1168–1180.
138
© 2022 the Author(s), ISBN 978-1-032-31259-0
Flow cone – new CPTU add-on module trialled in Halden silt

A.S. Gundersen, T. Lunne, R. Stelzer & Ø. Blaker
Norwegian Geotechnical Institute
G.W. Tucker & L. Krogh

Ørsted
ABSTRACT: Hydraulic conductivity is an important soil parameter for design of shallow foundation con
cepts for offshore wind, and there is a need for a new tool that can measure this parameter in a reliable way.
A new module has been developed that can be mounted behind a standard piezocone test (CPTU) probe. Water
can be injected in a controlled manner into the CPTU equipment and flow out into the surrounding soil through
a filter an offset behind the friction sleeve, while water pressure (uf) is measured by a transducer mounted in
the filter itself. During penetration of the CPTU probe, water flows out at a constant rate, while uf is measured
in addition to CPTU parameters qc, fs and u2. At desired depth intervals penetration can be stopped and either
a dissipation test or constant flow rate test can be carried out to determine hydraulic conductivity.
Hydraulic fracture tests can be performed in low permeability soils, where water flow is used to induce
a vertical crack in the soil. The vertical crack is then allowed to close while pore pressure is monitored, from
which the closing pressure can be determined and subsequently used to determine the in-situ K0-condition.
This paper describes a series of hydraulic conductivity tests and a hydraulic fracture test carried out at one
of Norway’s recently established geotechnical test sites, a silt dominated site in Halden. Several tests were
successfully carried out and the results were benchmarked against hydraulic conductivity as measured by fall
ing head tests in standpipes and laboratory tests. In general, the results compared well to credible benchmark
ing tests, showing a promising potential for this tool.
1 INTRODUCTION The normally consolidated clayey silt layers below

(Units II and III) extend down to about 15
Hydraulic conductivity (k) and coefficient of earth 16 m below ground level (bgl) in the southwest
pressure at rest (K0) are required parameters for corner of the site. The silt is uniform and structure
a wide range of geotechnical engineering problems, less to mottled, with primary bedding and lamin
including design of shallow foundation solutions for ations almost absent due to bioturbation. Units II
offshore wind. However, these parameters are chal and III contain similar amounts of quartz (40%),
lenging to measure accurately, both in-situ and in the plagioclase (30%), feldspar (12%), clay minerals
laboratory. The purpose of the flow cone add-on and mafic minerals (amphibole). Classification and
module is to measure hydraulic conductivity in sands in-situ tests suggest that the silt becomes sandier
but trial testing at the Halden silt site suggests closer to the deeper soil unit, which consists of
a wider range of application for the tool. This paper a low to medium strength clay (Unit IV). The clay
presents results from cone penetration testing, dissipa unit has a slightly laminated structure, with occa
tion testing and hydraulic fracture testing in Halden sional shell fragments and drop stones (Blaker
silt, including current interpretation methodology for et al., 2019).
flow cone dissipation data in low permeability soils.
3 OVERVIEW OF TESTS
2 HALDEN SITE
This paper presents a range of laboratory and in-situ
The site is located close to the city Halden in tests as part of the assessment of flow cone results. To
a recreational park, Rødsparken. A silty, clayey facilitate easy reading, Table 1 is included providing
sand constitutes the topsoil and extends down to about a general overview of benchmark tests and flow cone
4.5-5.0 m below ground level (Unit I), being generally tests, number of tests carried out, nomenclature and
loose to medium dense, with some organic material. depth ranges.
DOI: 10.1201/9781003308829-13
139
Table 1. Overview of benchmark tests and flow cone tests.
Number
Equipment Test type of tests Test abbreviation Location IDs Test IDs Depth range Comments
Oedometer Constant head* 22 Oedometer - - 4.5-18.7 m bgl Benchmark tests

Triaxial Constant head 13 Triaxial - - 5.3-14.6 m bgl Benchmark tests
Standpipe Falling head 10 HA-SLT 1 to 5 1 to 2 6-15 m bgl Benchmark tests
CPTU Cone penetration 1 HA-CPTU - - 2-18 m bgl Benchmark test
Flow cone Cone penetration 4 HA-FCPTU 1 to 4 - 2-18 m bgl
Flow cone Dissipation 13 HA-FCPTU 1 to 4 1 to 4 6-15 m bgl
Flow cone Hydraulic fracture 1 HA-FCPTU 4 5 16.41 m bgl
* Hydraulic conductivities at zero axial strain (back-extrapolated along the linear e-log(k) line)
4 BENCHMARK TESTS
4.1 Oedometer and triaxial tests

Constant-head hydraulic conductivity tests were
conducted at different stress levels during
a selected number of oedometer tests and during
the consolidation stage of several triaxial tests
(Sandbækken et al., 1986, Berre, 1982).
Hydraulic conductivity was determined by flow
ing de-aired water through the specimens. Values
from oedometer tests represent hydraulic conduct
ivity at zero axial strain (i.e., at a void ratio com
parable to in-situ conditions). Values from triaxial
test specimens represent the hydraulic conductiv
ity at the in-situ effective stress state (i.e., after
consolidation and some subsequent change in
void ratio).
4.2 In-situ falling head tests

Ten in-situ falling head tests (slug tests) were
conducted in parallel with flow cone testing at
five corresponding depths. The excess pore pres
sure from standpipe installation could dissipate
for 24 hours before falling head tests were initi
ated by pouring clean tap water into the stand
pipes. Figure 1 illustrates pressure heads with
velocity and best fit linear regression lines, Figure 1. Pressure head with velocity from in-situ falling
head tests. Measured results and best fit linear regression
showing good repeatability for all tests except lines.
the tests 15 m bgl, which may be explained by
a small gap between the standpipe and surround
ing soil for test HA-SLT-5-1. 5 EQUIPMENT AND PROCEDURES
For interpretation of hydraulic conductivity, the
velocity method was preferred over time lag The flow cone is a standard cone penetrometer
method due to its simplicity and independence of paired with a custom-built hydraulic module, see
piezometric level (Chapuis, 2012). From the Figure 2 (Gundersen et al., 2019). The hydraulic
slope, mv, of the linear regression lines in module consists of a bronze filter offset behind the
Figure 1, the hydraulic conductivity, k, was deter cone sleeve and a control system at ground surface.
mined using Equation (1), refer to Daniel (1989) The control system handles data acquisition and
for further details, where A is the internal cross- provides flow rate control by means of a linear step
sectional area of the standpipe and L and D are motor driving a piston. Parameters such as ambient
the length and diameter of the well screen, pressure, system pressure, water pressure inside filter
respectively. (uf) and flow rates are recorded. During cone
140
A hydraulic fracture test is performed by inducing
a crack in the soil by water flow for subsequent
monitoring of pore pressure decay. The test is
intended for low permeability soils with K0 < 1,
meaning that the in-situ horizontal stress is lower
than the vertical, and thus a vertical crack is
expected to initiate first. To split the soil,
a sufficiently high flow rate must be applied. At
Halden, 30 ml/min was selected based on previous
experience from testing in soft clays. The vertical
crack closes when the pore pressure equals the soil
pressure normal to the crack, and the basis for K0
derivation is that this soil pressure is equal to the in-
situ total horizontal stress, σh0 (Bjerrum and Ander
sen, 1972).
All tests at Halden were carried out using
a standard 10 cm2 Geotech Nova cone with acoustic
data transmission from probe to user interface, with
standard penetration rate of 20 mm/s ± 5 mm/s. In
total, 13 dissipation tests were carried out to a target
75 % dissipation of initial excess pore pressure where
all CPTU parameters were logged including uf.
Figure 2. Equipment and setup of the flow cone (Gunder 6 MEASURED AND DERIVED PARAMETERS
sen et al., 2019). The offset between u2 and uf measure
ments is 123 cm.
6.1 Cone penetration tests
penetration, water flows into the surrounding soil Figure 3 illustrates the corrected cone resistance (qt),
through the filter at a constant rate while uf is meas sleeve friction (fs) and pore pressures (u2 & uf) with
ured in addition to cone resistance (qc), sleeve fric depth below ground level corrected for inclination. The
tion (fs) and pore pressure (u2). The purpose of water figure includes results from a standard CPTU (HA
flow during cone penetration in low permeability CPTU), suggesting that a constant flow rate of 5 ml/
soils is to maintain filter saturation and prevent filter min during cone penetration, which was selected based
clogging. At desired depths penetration can be on previous experience, has negligible effect on the
stopped and either a dissipation test, constant flow standard cone penetration measurements (i.e., qc, fs,
rate test or hydraulic fracture test can be conducted. and u2).
Figure 3. Corrected cone resistance, qt, sleeve friction, fs, pore pressure behind cone shoulder, u2, and pore pressure, uf,
with depth.
141
6.2 Pore pressure dissipation tests Best estimate values of Gmax, σv0, σ’v0 and
M found in Blaker et al. (2019) were used as
Figures 4 and 5 illustrate the measured pore pres
basis for interpretation of hydraulic conductivity.
sures u2 and uf with square root of time. Most of the
Input of corrected cone resistance in Equation 2
tests in Figure 4 exhibit dilative behavior (non
was derived from linear interpolation of the
monotonic), which is consistent with previous dissi
results presented in Figure 4.
pation tests conducted at the Halden site (Carroll and
As part of proposing an interpretation method
Paniagua, 2018). The pore pressure, uf, generally
ology for uf-dissipation, triple element CPTU
exhibits a monotonic decrease with time but a small
results from Halden were reviewed. This review
delay in pore pressure decay is evident in Figure 5.
suggested that the initial excess pore pressure,
The square root method proposed by Sully et al.
uf,i, is primarily caused by the cone penetration
(1999) was used to determine hydraulic conductivity
itself, not the input water flow of 5 ml/min. On
from u2-dissipation data, back-calculating the initial
this basis, it was concluded that the Teh and
pore pressure, assuming an initial linear relationship
Houlsby (1991) framework for location 10 radii
between pore pressure and square root of time.
behind the cone shoulder could be used to esti
Rigidity index (Ir = G/su) and constrained modulus
mate the hydraulic conductivity. Figures 6 and 7
(M) are required for estimating hydraulic conductivity
show the head ratio with dimensionless time
from flow cone dissipation tests. Teh and Houlsby
factor for u2-dissipation and uf-dissipation using
(1991) showed that Ir influences the plastic failure
the framework by Teh and Houlsby (1991). The
zone around the cone tip during penetration and hence
figures show a good fit to the shape of the theor
the associated stresses and pore pressures. Carroll and
etical solutions, which is discussed further in
Paniagua (2018) interpreted u2-dissipation results from
Section 7.
Halden using Ir based on advanced laboratory testing
and conservative undrained shear strength. In addition,
two methods proposed by Krage et al. (2014) were 6.3 Hydraulic fracture test
examined, Method-A and Method-B. The latter com
One hydraulic fracture test was conducted
pared the best with advanced laboratory data and has
16.41 m below ground level, the result of which
thus been applied. Method-B rigidity index is calcu
is illustrated in Figure 8. The figure shows meas
lated using Equation (2), where Gmax is the small
ured pore pressure with velocity (i.e., how
strain shear modulus and σv0 and σ’v0 are the in-situ
quickly the pore pressure decays). The pressure
total and effective vertical stress, respectively.
at which the crack closes was estimated to
2240 cm. Based on piezometric level from stand
pipe falling head tests and an average soil unit
weight of 19.2 kN/m3, this yields a K0 value of
0.44, which falls within the expected K0-range
(0.4 to 0.65) for the Halden site.
Figure 4. Measured pore pressure u2 with square root of Figure 5. Measured pore pressure uf with square root of
time. time.
142
Figure 6. Pore pressure, u2, head ratio with dimensionless Figure 8. Pore pressure, uf, with rate of change in pore pres
time factor proposed by Teh and Houlsby (1991). sure, velocity, including discontinuous interpretation line.
Figure 7. Pore pressure, uf, head ratio with dimensionless Figure 9. Hydraulic conductivity from benchmark tests
time factor proposed by Teh and Houlsby (1991). and flow cone tests with depth.
7 DISCUSSION AND COMPARISON consolidation, variations in piezometric level, shape

factor, etc.
Figure 9 illustrates the hydraulic conductivities from Hydraulic conductivities from u2-dissipation and
benchmark tests (oedometer, triaxial and in-situ uf-dissipation show good agreement with triaxial test
falling head tests) and flow cone u2-dissipation and results, especially from 10 m to 15 m bgl. It is
uf-dissipation tests with depth. Oedometer and in- evident from Figure 9 that u2-dissipation and
situ falling head tests generally present the highest uf-dissipation tests can identify the main soil layer
and lowest hydraulic conductivities, respectively. ing i.e., the clay below 16 m bgl and the sand
The difference may be explained by the oedometer 4 m bgl presenting significantly lower and higher
values being picked at zero axial strain with presum hydraulic conductivities, respectively, as expected.
ably higher void ratio than in-situ conditions. For the Except for falling head and u2-dissipation results
in-situ falling head tests, a concern is filter clogging below 14.5 m bgl, in-situ tests show good repeatabil
during installation, causing reduced water injection ity. In contrast, oedometer results show some scatter
area and lower calculated hydraulic conductivities. which may be due to disturbance during sampling
Other aspects of the in-situ falling head tests involve and sample handling having significant influence on
soil disturbance, stress changes, transient the results due to small sample size.
143
Distribution of excess pore pressures around the the triaxial test specimen, the hydraulic conductivity
cone tip and shoulder are generally more complex than measurements made on these specimens are generally
alongside friction sleeve and CPTU rods. Based on Teh expected to be more reliable. Results from flow cone
and Houlsby (1991) it is expected that flow patterns for dissipation and constant head triaxial tests compare well.
uf are predominantly radial and more repeatable com These observations present a confident potential of the
pared to u2. With pure radial flow, the influence of filter proposed methodology for uf dissipation, which is based
size is expected to be negligible, however this should on classical uncoupled solution for undrained cone pene
be investigated for confirmation. tration and subsequent pore pressure dissipation. How
The laboratory and in-situ tests differ in that for ever, further studies should be carried out to fully verify
laboratory tests, de-aired water was used for hydraulic the proposed interpretation methodology, also consider
conductivity testing, while clean tap water was used for ing the influence of rigidity index, compressibility, filter
in-situ tests. For future in-situ testing water with prop size, soil disturbance and general soil behavior.
erties as in the field should be used. In addition, the One hydraulic fracture test was conducted in the
laboratory tests gave vertical hydraulic conductivity clay unit at Halden resulting in a coefficient of earth
while horizontal consolidation properties dominate in- pressure at rest that compares well with the general
situ dissipation, suggesting hydraulic conductivity trend from literature.
anisotropy ratio close to one at Halden.
The proposed interpretation methodology for flow
cone hydraulic conductivity requires input of rigidity ACKNOWLEDGEMENTS
index and constrained modulus, which may be
unknown or difficult to determine. The impact of This work was funded by Ørsted and in part the Norwe
these parameters on the estimated hydraulic conduct gian Research Council. The authors are grateful to many
ivity should be further investigated. of their colleagues for valuable discussions and greatly
From the testing at Halden it is evident that the appreciate the high-quality work by NGI field investiga
effect of water flow during cone penetration is pore tion group with Kristoffer Kåsin, Stig Bjørnsvik and
pressure build-up. However, the influence of 5 ml/min Don Terje Christiansen as main contributors.
of water flow on the excess pore pressure was negli
gible compared to that from cone penetration itself. It REFERENCES
is suggested that flow rate be determined based on Berre, T. (1982) Triaxial Testing at the Norwegian Geotech
experience and equations from Gundersen et al. (2019) nical Institute. Geotechnical Testing Journal, 5, 3–17.
to produce project specific hydraulic heads. It is con Bjerrum, L. & Andersen, K. H. (1972) In-situ measure
sidered most important to avoid flow rates yielding ments of lateral pressures in clay. European Conference
excessive hydraulic heads and thereby zero effective on Soil Mechanics and Foundation Engineering, 5.
stresses in the soil surrounding the uf-filter also leading Madrid 1972. Proceedings. Madrid, Sociedad Española
to significant soil disturbance and/or erosion. de Mecánica del Suelo y Cimentaciones.
The determined K0-value from hydraulic fracture test Blaker, Ø., Carroll, R., Paniagua Lopez, A. P., Degroot, D. J.
compares well with the general trend presented in & L Heureux, J.-S. (2019) Halden research site: geotech
nical characterization of a post glacial silt.
Blaker et al. (2019). However, more tests should be car Carroll, R. & Paniagua, P. (2018) Variable rate of penetra
ried out before conclusions can be drawn on the appro tion and dissipation test results in a natural silty soil.
priateness of K0-determination using flow cone module. Cone Penetration Testing 2018. CRC Press.
Chapuis, R. P. (2012) Predicting the saturated hydraulic
conductivity of soils: a review. Bulletin of engineering
8 SUMMARY AND CONCLUSIONS geology and the environment, 71, 401–434.
Daniel, D. E. (1989) In situ hydraulic conductivity tests for
The flow cone is a standard cone penetrometer paired compacted clay. Journal of Geotechnical Engineering,
115, 1205–1226.
with a custom-built hydraulic module including
Gundersen, A. S., Carotenuto, P., Lunne, T., Walta, A. &
a pressure transducer inside a porous cylindrical filter Sparrevik, P. (2019) Field verification tests of the newly
located an offset behind the friction sleeve and was developed flow cone tool—In-situ measurements of
trialled at the Halden silt site. Pore pressure development hydraulic soil properties.
with time was measured at two locations, behind cone Krage, C., Broussard, N. & Dejong, J. (2014) Estimating
shoulder, u2, and 1.23 m behind cone shoulder, uf. The rigidity index (IR) based on CPT measurements. Pro
majority of u2 dissipation plots suggest dilative behavior ceedings of the 3rd International Symposium on Cone
and a square root of time method was used to correct the Penetration Testing, Las Vegas, Nev.
initial pore pressure. Sandbækken, G., Berre, T. & Lacasse, S. (1986) Oedometer
Hydraulic conductivity from flow cone dissipation testing at the Norwegian Geotechnical Institute. Consolida
tion of soils: Testing and evaluation. ASTM International.
(u2, uf), in-situ falling head and constant head oedometer Sully, J. P., Robertson, P. K., Campanella, R. G. &
and triaxial tests were compared. uf dissipation presents Woeller, D. J. (1999) An approach to evaluation of field
best repeatability, whereas in-situ falling head tests and CPTU dissipation data in overconsolidated fine-grained
constant head oedometer tests yielded the lowest and soils. Canadian geotechnical journal, 36, 369–381.
highest values of hydraulic conductivity respectively. Teh, C. & Houlsby, G. (1991) An analytical study of the cone
Due to the larger volume of soil and greater height of penetration test in clay. Geotechnique, 41, 17–34.
144
© 2022 the Author(s), ISBN 978-1-032-31259-0
Estimating in-situ frozen loamy soil viscosity from CPT

O.N. Isaev
Gersevanov Research Institute of Bases and Underground Structures (NIIOSP), “Research Center of Construction”
JSC, Moscow, Russia
ABSTRACT: Viscosity is used for calculation of settlements and deformations of frozen loamy soils
and ice caused by viscous flow at constant speed under action of long-term loads. A method for deter
mining the viscosity coefficient of frozen loamy soil by CPT was developed. Experimental studies by
CPT were carried out in a quasi-static mode (a series of short tests at low well-controlled speeds) and
in a step-increasing relaxation-creeping mode (a series of penetrometer stabilizations in which the pene
trometer and the speed of its penetration gradually decreased). Comparative data «unit cone resistance
qc – velocity of penetration vc» were analyzed using mathematical statistics methods. For frozen loamy
soil, a generalized regression equation qc = M(vc)m was established. Parameters M and m are calculated
based on empiric equations, depending on the soil temperature. Viscosity values obtained by uniaxial
compression CPT were compared. As a result of statistical analysis, an empirical transition coefficient
between them was established.
1 INTRODUCTION frozen soils, their original temperature can be dis

turbed and soils may thaw.
Permafrost soils (Roman 2002, Khrustalev 2005) The paper proposes a method for determining
occupy 25% of the world dry land: 65% in Russia, the viscosity coefficient of frozen loamy soils
50% in Canada, 60% in Alaska, 65% in Mongolia by CPT.
and 22% in China. They possess special physical
and mechanical properties due to ice. If permafrost
soils are frozen during construction and operation, 2 EXPERIMENTAL AND THEORETICAL
settlements of foundations must be determined PREREQUISITES FOR DETERMINING THE
taking into account the developing plastic deform VISCOSITY COEFFICIENT OF FROZEN
ations (flow) of ice. SOILS BY CPT
The flow rate is calculated based on the viscosity
coefficient. It is determined by the laboratory test for When testing plastic-frozen loamy soils with
uniaxial compression with an increasing stepwise a temperature of -0.3 … -0.1 °C, Ladanyi (1976)
load in accordance with the Standard (GOST revealed that cone resistance qc versus the velocity
12248.9-2020). There exist other laboratory methods of penetration vc is perfectly described by the
for the viscosity coefficient to be determined: in- power law
plane shear method (Pekarskaya 1963), “ball”
method (Maslov 1984), multi-plane shear device
(Recommendations 1989), ball stamp (Roman &
Kotov 2013, Roman & Kotov 2016) and viscometer
(Ter-Martirosyan & Ermoshina 2019). They have
not been widely used yet. In-situ methods have not where q0c and v0c = coordinates
� �m of a point on the
been used until recently. cross-plot qc ðvÞ; M ¼ q0c = v0c = empirical para
Laboratory methods have some disadvantages: meter;m51 – nonlinearity parameter.
� � If v0c ¼ 1
0 m
soil under natural stress cannot be tested; soil sam (M ¼ qc ) and qc ¼ 1 (M ¼ 1= v�c � ) the formula
0 0
m
ples are often damaged when drilling out. As (2.1) takes a simpler form qc ¼ q0c vc .
a result, reconsolidation is required before testing; Let us assume that the thickness of the frozen soil
extra work is also required (drilling to select sam compression zone under the cone is approximately
ples; transportation, storage and preparation of (generally, compression zone has more complex
samples for testing). When drilling out samples of dependence on penetrometer diameter) equal to one
DOI: 10.1201/9781003308829-14
145
tenth of its diameter. Then the relative velocity of
penetration ε_ c is equal to
Let us take the uniaxial viscosity coefficient ηuc

as a reference value. The condition of equality of the
where dc = cone diameter.
viscosity coefficients for uniaxial compression ηuc
The relationships τ(γ_ ) and σ(ε_ ) are mostly non
and CPT ηCPT is
linear in frozen soils and ice. Viscosity coeffi
cients ητ and ησ depend on the applied load during
creeping. The rheological equations for shear and
compression stresses are τ ¼ ητ ðτÞγ_ and
σ ¼ ησ ðσÞε_ . Such viscosity is characteristic for
bodies possessing structural strength. This viscos where ψ - empirical transition coefficient (further
ity is sometimes referred to as structural viscosity. transition coefficient).
Such bodies include frozen soil and ice. Thus, in order to determine the viscosity coeffi
Let us consider cone penetration into a nonlinear cient from CPTηCPT, it is enough to determine the
viscous body with a variable viscosity coefficient η. transition coefficient ψ. The following problem is
The rheological equation is as follows discussed further.
3 GENERAL DESCRIPTION OF SOILS.

EQUIPMENT AND METHODS FOR
EXPERIMENTAL STUDIES
where ηc = conditional viscosity coefficient
� of soil�1=min
CPT. Using (2.1) and (2.2) – ε_ c =ε_ 0c ¼ qc =qc0 . CPT was performed in-situ and in the laboratory
Let us change this equation to (2.3) (refrigeration chamber). Permafrost soils were tested
in-situ: loams, sandy loams and clays – less fre
quently; soil temperature T = -0.2 …- 2.1 °C; mois
ture content w = 12 … 38 %; massive and thin-
layered cryogenic texture; ice inclusions up to
1-4 mm. Artificially frozen loams were tested in the
laboratory, T = -1.1 and -2.2 °C, w = 21%.
Thus, the conditional viscosity coefficient in CPT A penetrometer with a friction sleeve, Ø35.6 mm
will be (Ryzhkov & Isaev 2016) and a temperature sensor in
a cone (to measure soil temperature T) was used. In-
situ tests were carried out in a quasi-static mode and
a stepwise stabilization mode.
The quasi-static mode involved the procedure of
penetrometer stopping and holding until it froze into
soil. The penetrometer is then tested by performing
a series of short pushes of the penetrometer was
�m account that ε_ c ¼ vc =ð0:1dc Þ and
0 0
Taking� into performed at different velocities vc ¼ 3 · 10-4 … 2 m/
M ¼ qc0 = vc0 , the equation (2.5) is as follows: min. To prevent the effect of soil heterogeneity on the
test results, high and low velocities were alternated.
Stepwise stabilization (relaxation-creep) mode –
the test is also performed after stopping and holding
until it froze into soil. However, it differs from the
quasi-static in that the penetrometer is tested by per
forming a series of stepwise increasing stabiliza
tions. The stabilization of the penetrometer at the
Using (1.1) and (2.2) - (2.5) can be ηc ðvc Þ step was carried out by supplying oil to the hydraulic
or ηc ðε_ c Þ jacks which push the penetrometer. The given value
qc having been obtained, the supply of oil was
stopped and the stabilization mode started. The load
acting on the penetrometer and its velocity (as
a result of relaxation-creep of the surrounding soil)
gradually decreased and the intensity decreased as
well, vc ¼ 2 · 10-5 . . . 0:6 m/min.
146
The quasi-static mode was used in the refriger (2.1) without an inflection point on the plot
ation chamber. The velocity of penetration was set ln vc » if T ¼ -0:2 . . . - 2:2 � C
«ln qc � and
by regulating the speed of the screw press on which vc ¼ 10-6 . . . 4 · 10-2 m/min.
the lever was dropped, vc ¼ 10-6 . . . 0:08 m/min. Figure 3.2 makes it possible to conclude that
The study procedure was as follows. M possesses great variability and depends on the test
Stage 1: The test results were divided into statis mode and soil temperature. Its value is slightly
tical samples of data pairs «qc –vc » depending on higher under stepwise stabilization than when tested
their mode and soil temperature. under the quasi-static mode. The difference in M
Stage 2: An approximation (1.1) was used for values during in-situ and laboratory tests is due to
each statistical sample; M and m were determined; the fact that the soil is permafrost in the first case,
sampled values of the correlation coefficient ρ, deter and it is artificially frozen in the second case. This
mination coefficient R2 and theoretical correlation obviously led to its decompaction and decrease in
ratio R were determined as well; the degree of linear strength. In addition, the difference in soil moisture
ization of the experimental data in the coordinates and dispersion also affected. A linear relationship
«ln qc � ln vc » was visually evaluated. has been established
Stage 3: The effect of the test mode and soil tem
perature on M and m was assessed. Statistical ana
lysis was performed; the regression equations
applicable for M and m to be determined by tempera
ture T were specified; M and m, ρ, R2 and R were where aM and bM = empirical coefficients, for in-situ
also determined. tests under the quasi-static mode aM ¼ -10:07
Stage 4: The conditional viscosity coefficient in and bM ¼ 5:060.
CPT ηc and the viscosity coefficient obtained during Due to soil strength increase, M increases when
uniaxial compression tests ηuc were compared; the temperature decreases. Extrapolation of the straight
transition coefficient ψ ¼ ηuc =ηc was determined. lines M (T) for all test modes illustrates that the
The values of ηuc were taken from the books on straight lines intersect approximately in one point
laboratory data for uniaxial compression tests. The which shows the initial temperature of soil freezing.
values of ηc were taken based on the generalized When loamy soils are in a thawed state - M =
results of CPT under the quasi-static mode – in 6.57 … 7.58 for the temperature range of the onset
accordance with the empirical equations (see of freezing (-0.15 … -0.25 °C).
stage 3). The compared pairs of values ηuc and ηc The nonlinearity coefficient m is less variable,
had the same values of T and ε_ c . m<1. With the temperature decrease m decreases, i.e.
Stage 5: The effect of various factors on the the nonlinearity of the equation (1.1) increases since
transition coefficient ψ was studied (cone resist ice-cement content and rheology of soil increase. The
ance qc , velocity of penetrationvc , relative speed value of m hardly depends on the test mode and soil
of soil deformation ε_ c , soil temperature modulus | type. Temperature T has the greatest effect on it.
T|, compressive stresses under uniaxial Figure 4.2.b illustrates that the experimental values
compressionσ, soil type). Regression equations of m (T) obtained for various modes and test condi
were calculated. tions can be described by the general linear
Stage 6: Two-stage verification of the proposed relationship
method for determining ηuc from CPT data was carried
out. The first stage was based on the data used in the
previous stages. The second stage was based on the
new comparative data used on the experimental site.
where am and bm = empirical coefficients, for the
4 RESULTS AND ANALYSIS OF general linear relationship am ¼ 0:0235 and
EXPERIMENTAL STUDIES bm ¼ 0:1280, for in-situ tests in the quasi-static
mode am ¼ 0:0184 and bm ¼ 0:1217.
The total number of data pairs «qc –vc » – over 400. The correlation coefficient ρ between m and T is
Typical cross-plots are shown in Figure 4.1. 0.59 for the total statistical sample. This can be
Statistical analysis has shown that qc versusðvc Þ is considered acceptable, since it takes into account
described by the power equation (2.1); a high degree the large heterogeneity of the data included in it.
of data linearization in the coordinates «ln qc �
ln vc » Substituting (4.1) and (4.2) in (2.7) and (2.8), the
is revealed (Figure 4.1). The values of R have been as conditional viscosity coefficient ηc can be as
follows: 0.76 … 0.98 for in-situ tests under the quasi- follows:
static mode, 0.89 … 0.98 under stepwise stabilization,
0.92 … 0.94 under laboratory conditions. The cross-
plot of qc versusðvc Þ is described by a single curve
147
According to (2.7), the conditional viscosity coef
ficient of thawed loamy soils ηc is as follows for the
temperature range of the onset of freezing (thawed
state):
The viscosity coefficients of frozen loamy soils

obtained after uniaxial compression tests ηuc (data
taken from the books, Table 4.1) and ηc evaluated by
the formula (4.4.a) depending on ε_ c and T have been
compared.
The transition coefficient ψ ¼ ηuc =ηc was used for
correlation. The influence of various factors on ψ was
studied: cone resistance qc , velocity of penetrationvc ,
relative speed of soil deformation ε_ c , soil temperature
modulus |T|, stresses under uniaxial compressionσ,
loamy soil type. The values of jTj, ε_ c and σ were
taken after uniaxial compression tests; vc was evalu
ated by the formula (1.2); qc – by the formula (1.1);
M and m – by the formulas (3.1) and (3.2).
Figure 4.1. “vc – qc ” diagrams: in-situ tests under a quasi- Analysis of the results presented in Table 4.2 and
static mode, T ¼ -0:2 � C (a); in-situ tests under stepwise in Figure 4.3 makes it possible to conclude the
stabilization, T ¼ -1:0� C (b); laboratory tests in following.
a refrigeration chamber under the quasi-static mode, The greatest influence on ψ is exerted by qc and
T ¼ -1:1� C (c).
T, the relationship is nonlinear. The linkage is strong
or very strong. The influence of vc , ε_ c and σ on the
coefficient ψ is very weak. The influence of the soil
type (loam, clay loam, sandy loam) is insignificant –
the points referring to experimental data for different
soils do not meet any definite patterns.
Empirical equations for the viscosity coefficient
For in-situ tests in the quasi-static mode (4.3) and of frozen loamy soil under uniaxial compression
(4.4) are as follows: based on CPT data were proposed
Table 4.1. Comparison of viscosity coefficients (ηuc ) and (ηc ) obtained after uniaxial compression tests (data taken from
the books) and calculated by the formula (4.4.a).
Uniaxial compression
Soil type T, � C σ, MPa ε_ , min-1 ψ ¼ ηuc =ηc n tests after
Loam W=31…38 % -7.9…-1.7 0.25…0.50 2:5 · 10-5 . . . 2:4 · 10-4 0.002…0.129 4 Tsytovich 1973
Sandy loam W=12…13 % -6.0…-0.9 0.25…0.50 1:5 · 10-5 . . . 1:3 · 10-4 0.015…0.105 5 Tsytovich 1973
Clay loam -3.0 1.20 1:0 · 10-3 0.017 1 Roman 2002
W=23
Clay loam -3.0 0.40…0.80 1:3 · 10-8 . . . 6:6 · 10-10 0.045…0.071 5 Grechishchev 1963
Notes: T – soil temperature; σ – stress under uniaxial compression; ε_ – speed of flow under uniaxial compression; n –
number of comparative data pairs
148
Table 4.2. Statistical analysis of various factors influencing the coefficient ψ for frozen loamy soils.
Statistical indicators
Parameter ρ R2 R Regression equation Linkage
0.61 - - ψ = -0.0004qc + 0.0632 Average

-:0:018qc
- 0.81 0.90 ψ ¼ 0:0618e Strong
qc - 0.86 0.93 ψ = 0.3299qc -°.86° Very strong
0.12 - - Linear
- 0.0093 0.096 Exponential Very weak
vc - 0.058 0.24 Power
0.12 - - Linear
ε_ c - 0.058 0.24 Power
0.77 - - ψ = -0.01jT j + 0.10 Strong
-0:531jT j
- 0.85 0.92 ψ ¼ 0:1945e Very strong
jT j - 0.68 0.82 ψ = 0.1653jT j-1.552 Strong
0.14 - - Linear
σ - 0.00005 - Power
Notes: ρ – sample correlation coefficient; R2 – sample determination coefficient; R – theoretical correlation ratio
Experimental verification of (4.5) and (4.6) was

carried out in two stages by calculations and statis
tical analysis of the ratios ηCPT ðqÞ =ηuc
and ηCPT ðT Þ =ηuc .
The data from Table 4.1 were used at the first
stage. The value of ηc in (4.5) and (4.6) was
evaluated by (4.3.a). Figure 4.4 shows the histo
grams of these ratios. The analysis showed that
the statistical data were close for ηCPT ðqÞ and
ηCPT ðTÞ . Taking into account the influence of
a large number of factors, they can be considered
quite acceptable.
At the second stage, verification was carried out
by direct comparative tests on the experimental site.
The viscosity coefficients of plastic-frozen clay loam
calculated from CPT data were compared with the
coefficients ψq and ψT obtained after direct tests for
uniaxial compression ηcu . The CPT procedure was
done next to a deep pit where uniaxial compression
tests were carried out. T = -0.6 …- 0.7 °C, w =
17.9 … 18.2 %, ice content 0.1.vc ¼ 3:1 · 10-4 …2
Figure 4.2. Diagrams of М (Т) and m(Т) after in-situ and m/min, qc ¼ 4:1…8:5 MPa. Figure 4.5 shows that
laboratory CPT data. the coefficient ψT is more accurate – ηCPT ðT Þ =ηuc =
0.75 and ψq - less accurate – ηCPT ðqÞ =ηuc =0.44. The
medians Me of the ratios ηCPT ðqÞ =ηuc and ηCPT ðT Þ =ηuc
actually coincide with their mean values, i.e. the
values are significant. Experimental verification has
shown that the transition coefficient ψT evaluated by
(4.6) makes it possible to determine the viscosity
coefficient of frozen loamy soil under uniaxial com
where ψq and ψT = empirical coefficients; pression by CPT data with sufficient accuracy for
ψq ¼ 0:330qc -0:860 , ψT ¼ 0:195e-0:531jTj . practical use.
149
Figure 4.3. The effect of various factors on ψ (cone resistance qc (a), velocity of penetration vc (b), relative speed of soil
deformation ε_ c (c), soil temperature modulus |T| (d), compressive stresses under uniaxial compression σ (e).
Figure 4.4. Histograms of ratios ηCPT ðqÞ =ηuc and ηCPT ðT Þ =ηuc (ηuc is taken from Table 4.1).
5 CONCLUSIONS regression equations to be obtained. The effects

of the CPT mode and soil temperature on
At present, laboratory methods are the only ones soil parameters have been studied. Empirical
used to determine the viscosity coefficient of relationships have been established to
frozen soil. The uniaxial compression test is the determine the transition coefficient from the con
one which is preferred. Proposed is the method ditional viscosity coefficient under CPT to
to determine the viscosity coefficient of frozen the viscosity coefficient under uniaxial compres
loamy soils from CPT data. CPT can be per sion. Thus, soil resistance to CPT and soil tem
formed in the quasi-static or relaxation-creeping peratures exert the greatest influence on the
modes. The method is based on the power equa transition coefficient. Experimental verification
tion of cone resistance versus velocity of pene of the method has shown its applicability in
tration. The studies have resulted in power-law practice.
150
Figure 4.5. Histograms of ratios ηCPT ðqÞ =ηuc and ηCPT ðT Þ =ηuc (direct comparative tests).
Notes: ηucðqÞ and ηucðT Þ - viscosity coefficient evaluated by formulas (4.5) and (4.6); n - number of statistical sample data;
B mean values; Me - median; σ - standard of deviation; D - dispersion; As - asymmetry; Ek - kurtosis;Rvar - range of vari
B-
ation; min - minimum value; max - maximum value.
REFERENCES 7. Recommendations for Soil Creep and Consolidation to

be determined by Laboratory Tests/PNIIIS - M.:
1. GOST 12248.9-2020. Soils. Strength and Deformabil Stroyizdat, 1989, 64 pp. (rus).
ity Parameters of Frozen Soils Determined by Uniaxial 8. Roman, L.T. 2002. Mechanics of Frozen Soils. - M.:
Compression. (rus). MAIK “Nauka/Interperiodica” - 426 pp. (rus).
2. Grechishchev, S. E. 1963. Creeping of Frozen Soils in 9. Roman, L.T. & Kotov, P. I. 2013. Viscosity of Frozen
a Difficult Stress State. Collection of Papers “Strength Soils determined with a Ball Stamp. Earth Cryosphere,
and Creep of Frozen Soils”. Sib. Dept. of Academy of vol. XVII, №4, P. 30–35. (rus).
Sciences of the USSR/ (rus) Publ. House of the Academy 10. Roman, L.T. & Kotov, P. I. 2016. Viscosity of
of Sciences of the USSR, 1963. (rus). Frozen and Thawing Soils. “OFMG”. №1. P.16–19.
3. Khrustalev, L. N. 2005. Fundamentals of Geotechnics (rus)
in Permafrost: Textbook. - M.: Publ. House of Moscow 11. Ryzhkov, I.B. & Isaev, O.N. 2016. Cone Penetration
State University, – 542 pp (rus). Testing of Soils in Geotechnics. Stockholm, Sweden:
4. Ladanyi, B. 1976. Use of static penetration test in Bokforlaget Efron & Dotter AB, 408 p.
frozen soils. Canad. Geotech. J., Vol.13 p.p. 95–110. 12. Ter-Martirosyan, A.Z. & Ermoshina, L.S. 2019.
5. Maslov, N. N. 1984. Physical and Technical Theory of Experience in determining viscosity of soil on the basis
Loamy Soils Creep in Construction Practice. –M.: of experimental studies. IOP Сonference Series: Mater
Stroyizdat – 176 pp. (rus). ials Science and Engineering, 687 (4), DOI: 10.1088/
6. Pekarskaya, N. K. 1963. Shear Strength of Frozen Soils 1757-899X987/4/044039.
with its Dependence on Temperature. M. Publ. House of 13. Tsytovich, N. A. Mechanics of Frozen Soils. –M.:
the Academy of Sciences of the USSR – 108 pp. (rus) Vysschaja Shkola, 1973. – 448 pp. (rus).
151
© 2022 the Author(s), ISBN 978-1-032-31259-0
Determination of natural stress state parameters for clay soils by using

3LSU-CPTU penetrometer
O.N. Isaev, R.F. Sharafutdinov & D.S. Zakatov
Gersevanov Research Institute of Bases and Underground Structures (NIIOSP), “Research Center of Construction”
JSC, Moscow, Russia
ABSTRACT: The article presents the results of experimental studies on the use of lateral stress measure
ments by using the LS-module (includes full and pore pressure sensors) of 3LSU-CPTU penetrometer to deter
mine the parameters of the natural stress state (NSS) of clay soils (the coefficient of lateral pressure at rest Ko,
the effective pre-compaction stress POP, the over-compaction stress σc' , the overconsolidation ratio OCR, the
natural total σho and effective σ'ho horizontal stress in the soil). In experiments, the LS-module with three pairs
of sensors located in areas with different diameters (Ø37 mm, Ø41 mm and Ø45 mm) was used. The results
of a statistical analysis of the relationships between LS-module (twenty-six direct and derived types were con
sidered) and NSS parameters of clay soils are presented. It is shown that Ko, OCR, σ'ho in the soil is determined
with the greatest accuracy.
1 INTRODUCTION 2 INVESTIGATED SOILS, EQUIPMENT AND

TEST PROCEDURE
The efficiency of using CPT is higher when using
penetrometers with additional sensors and devices The main physical and mechanical properties of clay
(Burns & Mayne 1998, Lunne et al. 2004, Ryzhkov I. soils are given in Table 1. Stiff clay and firm-stiff
B. & Isaev 2016). For example, a penetrometer with loam were studied. Soils lay at a depth of
a lateral stress module (LS-module), includes full and 3.5-41.7 m. Tests were carried out at experimental
pore pressure sensors. These penetrometers have been sites in Moscow.
developed and investigated since the 1980s (Campa
nella et al. 1990, Howie et al. 2014, Huntsman et al.
1986, Bayne & Tjelta 1987, Sully & Campanella Table 1. Properties of clay soils.
1990, Sully & Campanella 1991, Takesue & Isano
2001). However, according to the Lunne et al. (2004), ρ, е, Ip, d. Il, C, ’, Е,
�
«the lateral stress cone is still not a tool that has found Soil g/cm3 d.q. q. d.q. kPa MPa
practical use». In the USSR and Russia, until recently,
they have not been developed or investigated. Stiff clay 1.77 1.17 0.45 0.18 57 20 25
Modern powerful software geotechnical platforms Firm-stiff
2.00 0.65 0.14 0.36 10 22 10
loam
(PLAXIS, MIDAS GTS, Z-Soil, etc.) are using
numerical methods. Knowledge of natural stress
state parameters of soils is required as initial data of
numerical calculations. The following parameters were calculated and
The paper presents the results of experimental defined: natural (hydrostatical) pore0 pressure in the
studies on the use of lateral stress measurements by soil u0 , over-compaction stress 0
σ , effective pre
0 c
using the LS-module of 3LSU-CPTU penetrometer compaction stress 0POP 0
¼ σ c - σ vo , overconsolida
to determine the parameters of the natural stress tion ratio OCR ¼ σc =σ ,
0 vo 0
coefficient of lateral pres-
state (NSS) of clay soils (the coefficient of lateral sure at rest Ko ¼ σho =σvo , natural total σvo and
0
pressure at rest Ko , the effective pre-compaction effective σvo ¼ σvo - u0 vertical 0
stress in the soil,
stress POP, the over-compaction stress σ0c , the over- natural total σho and effective σho horizontal stress in
consolidation ratio OCR, the natural total σho and the soil. 0
0
effective σho horizontal stress in the soil). Studies to The σc , POP, OCR parameters were determined
determine the natural pore pressure in soil u0 are by the methods of Casagrande (1936) and Becker
described in detail in Isaev et al. (2020a) paper. et al. (1987) by laboratory tests in compression
DOI: 10.1201/9781003308829-15
152
Figure 1. General schematic of LS-module. Figure 2. 3LSU-CPTU penetrometer general assembly.
Figure 3. Example of σLS, uLS, σ'LS profiles with a layer of stiff clay at Moscow (Isaev et al. 2020b)
153
instruments. The results of both methods were aver nonlinear (exponential, logarithmic, polynomial,
aged. Coefficient Ko was calculated by Meyerhof power-law) function. Selection of dependencies with
(1976) formula, coefficient – by Jaky (1944): maximum correlation coefficient ρ, theoretical correl
ation relation R, coefficient of determination R2 . The
strength of the statistical relationship (for ρ and R) was
estimated in accordance with Chaddock-Snedecor
scale.
where φ – angle of internal friction; α = 0.5 - degree Fourth stage. Comparison of NSS parameters
index. 0 obtained by standard (direct) methods and calculated
The σho value was calculated using the formula using LS-module parameters (based on regression
equations). Estimation of the accuracy of determining
the parameters of the NSS by empirical dependencies.
Statistical analysis was performed using IBM
SPSS Statistics and MS Excel software.
Tests were carried out using 3LSU-CPTU probe.
It was developed and manufactured jointly by
3 RESULTS AND ANALYSIS OF
NIIOSP and USM-Engineering (Isaev et al. 2020b).
EXPERIMENTAL STUDIES
It included two modules (Figure 1, 2) – the CPTU
main (lower) module with a cone and friction
Tables 2 and 3 summarize the experimental data for
sleeve and 3LSU additional (upper) module with
LS-module and NSS parameters of clay soils. Table 4
lateral sensors of soil pressure and pore pressure.
give the correlation matrix of ρ for statistical relation
Above the modules was a unit with electronics,
ship between LS-module and NSS parameters of clay
above it there was a push rod. The main module
soils.
had Ø35.7 mm. The 3LSU-module included three
pairs of sensors located in cylindrical sections of
Ø37 mm, Ø41 mm and Ø45 mm. When the probe Table 2. Generalized LS-module measurement results of
was pushed, total (σLS1 , σ0 LS2 ,σLS3 ), pore (uLS1 , 0 uLS2 , clay soils.
uLS3 ) and effective
0
(σLS1 ¼ σLS1 - uLS1 ; σLS2 ¼ 0 0 0
σLS2 - uLS2 ,σLS3 ¼ σLS3 - uLS3 ) pressures were σLS1 =σLS1 =uLS1 , σLS2 =σLS2 =uLS2 , σLS3 =σLS3 =uLS3 ,
measured by pairs of sensors (pore pressure and lat kPa kPa kPa
eral stress) №1, №2, №3 on the lateral module
254 … 1657 84 … 1879 67 … 2090
surface.
42 … 1576 88 … 1409 42 … 1706
Studies were carried out in several stages.
-17 … +215 -20 … +670 -15 … +383
First stage. Listing and NSS parameterization.
0
Six
parameters
0
were selected: K o , P OP , σ c , OCR, σ ho ,
σho . Listing and LS-module parameterization. Nine
direct
0
parameters
0 0
(σLS1 , σLS2 ,σLS3 ; uLS1 , uLS2 , uLS3 ;
σLS1 , σLS2 ,σLS3 ) and seventeen combined parameters Table 3. Generalized NSS parameters of clay soils.
(σLS3 =σLS1 , σLS2 =σLS1 , σLS3 =σLS2 ; uLS3 =uLS1 , 0 0
uLS2 =uLS1 ; ID13 ¼ ΔσLS13 =ΔσLS1U ; ID12 ¼ ΔσLS12 σc , kPa OCR, d.q. P OP ,kPa Ko , d.q. σho , MPa
=ΔσLS1U ; KD1 ¼ ΔσLS1U =σ‘vo , KD2 ¼ ΔσLS2U =σ‘vo ,
140 … 2.9 … 0.93 …
KLS1 ¼ 0 σLS1 =σvo0
, KLS2 ¼0 σLS2 =σvo0 ; KLS3 ¼ σ0 LS3 2400 12.7
77 … 1938
2.23
0.06 … 0.72
=σvo0 , KLS1 ¼ σLS1 =σ‘vo , KLS2 ¼ σLS2 =σ‘vo ; KLS3
¼ σLS3 =σ‘vo ; ED13 ¼ R1 ΔσLS13 =ΔR13 ; ED12 ¼ R1 Δ
σLS12 =ΔR12 ) were investigated.
Where ΔσLS12 ¼ σLS2 - σLS1 ; ΔσLS13 ¼ σLS3 Regression data analysis showed that the empir
-σLS1 , ΔσLS1U ¼ σLS1 - u0 ; ΔσLS2U ¼ σLS2 - u0 ; ical data can be approximated with approximately
ΔR12 ¼ R2 - R1 , ΔR13 ¼ R3 - R1 ; R1 , R2 , R3 = equal accuracy by linear and nonlinear regression
radius of LS-module in the areas where the sensors equations (the difference between ρ and R in almost
are located 1, 2, 3. all cases is less than 5%).
The NSS and LS-module parameters for each Analysis of Table 4 allows to note the following.
0
comparative test were combined into one data group. 1. For σc the highest degree of correlation
19 comparative data groups were compiled. (ρ0 ¼ 0:61 . . . 0:68)
0
was revealed
0
with σLS1 ,
Second stage. Сorrelation data mining, determin σLS1 , σLS2 , σLS2 , σLS3 , σLS3 (Figure 4). In
ation of the strength of statistical linkage between NSS other cases, it is less. For factors of the com
and LS-module parameters of the clay soils. The value bined type, it is mainly statistically insignifi
0
and statistical significance of the sample correlation cant. With
0
an increase in σc , the values of σLS
index ρ was determined. Assessment of the presence of and σLS also increase, but with little change in
0
nonlinear significant statistical relationships. intensity
0 0
(Δσc =ΔσLS ¼ 0:5 . . . 0:8 and
Third stage. Regression data analysis for strength Δσc =ΔσLS ¼ 0:7 . . . 0:8), which weakly
correlations. Approximation of data by linear and depends on the diameter of the module.
154
2. For OCR the highest degree of correlation The relationship can be approximated by
strength (ρ ¼ -0:56 . . . - 0:60) was revealed linear equations of the form y ¼ kx or
with the parameters of the combined type y ¼ kx þ b. For σLS1, their plots actually over
σLS2 =σLS1 , ID13 , KD2 , ED12 (Figure 5). As the lap and pass through the origin of coordinates.
OCR increases, the parameter values decrease. For other lateral pressures, the line plots are
The relationship with most of the other param also quite close to each other. With an increase
eters is statistically insignificant. in σho , the values of these LS-module
3. For Ko the highest degree of correlation strength parameters increase with little change in
(ρ ¼ -0:58 . . . - 0:64) was revealed with the intensity 0 (Δσh0 =ΔσLS ¼ 0:4 . . . 0:5 and
parameters
0
of the combined type KD2 , σLS2 =σLS1 Δσh0 =ΔσLS ¼ 0:5 . . . 0:6).
and KLS2 (Figure 6). With their increase, the A weaker relationship with the results of
value of Ko decreases. In other cases, the relation measurements by pore pressure sensors and
ship is insignificant. the combined parameters obtained on their
4. The correlation between σho and LS-module basis. The exception is uLS1 , for which it
parameters is the strongest. The maximum cor is not statistically significant. This is prob
relation
0
(ρ=0.91…0.95)
0 0
was found for σLS1, ably due to the fact that, in comparison
σLS1 , σLS2 , σLS2 , σLS3 , σLS3 (Figure 7). with the soil adjacent to the probe, the
change in pore pressure with time has less
inertia – as the friction sleeve passes along
Table 4. Correlation matrix of ρ for statistical relationship
between LS-module and NSS parameters.
the soil layer, before the uLS1 sensor
approaches it, the pore pressure has time
to change significantly. Additional effects
on uLS1 do not occur, because the cone
with the friction clutch and the uLS1 sec
tion have approximately the same diameter.
Figure 4. Regression analysis of relationship between

0 0
over-compaction stress σc , and LS-module parameters σLS1
(a) and σLS1 (b).
With other LS-module parameters of the com

bined type, the relationship is much weaker. For
most of them it is statistically insignificant.
5. The nature
0
and strength of the statistical relation
ship σho
with LS-module parameters are gener
ally similar to that for σho.
6. The correlation between POP and LS-module

parameters is not statistically significant, except
for σLS1 (ρ=0,48).
For strong correlations, regression equations were
calculated. The verification of their applicability was
155
Figure 5. Regression analysis of relationship between
overconsolidation ratio OCR and LS-module parameters
KD2 (a) and σLS2 =σLS1 (b).
Figure 7. Regression analysis of relationship between nat
ural total horizontal stress in the soil σho and LS-module
parameters σLS1 (a) and σLS2 (b).
Table 5. Verification the accuracy of regression

relationships.
Ratio statistics
Ratio Regression equation B

B Me σ Vσ
0 0
σce ¼ 0:834σLS1 þ 428:984 1.56 1.12 1.32 0.85
0
σce
¼ 0:762σLS1 þ 443:625 1.60 1.09 1.39 0.86
OCRe OCRe ¼ 15:403e-0:394�KD2
1.11 1.04 0.50 0.45
OCR
OCRe ¼ -8:341 σσLS2 LS1
þ 13:033 1.29 0.98 0.68 0.53
Koe Koe ¼ -0:276KD2 þ 2:280 1.05 1.06 0.24 0.23
Ko
Koe ¼ -0:967 σσLS2
LS1
þ 2:293 1.07 0.99 0.28 0.26
σhoe σhoe ¼ 0:000556σ LS1 - 0:0269 1.18 0.99 0.63 0.54
σ ho
σhoe ¼ 0:000439σ LS2 þ 0:0723 1.23 0.98 0.74 0.60
0
σhoe ¼ 0:000334σ LS1 þ 0:0472 1.18 0.92 0.61 0.51
0 0
σhoe ¼ 0:000354σ LS1 þ 0:0480 1.16 0.92 0.58 0.50
BB = mean; Me = median; σ = standard deviation;

Figure 6. Regression analysis of relationship between coef Vσ ¼ σ=B B = variability index; e = subscript, means
ficient of lateral pressure at rest Ko and LS-module param that parameter is calculated according to the regression
eters KD2 (a) and σLS2 =σLS1 (b). equation
carried out by analyzing the ratios of the calculated

(according to the regression equations) to the refer variability (Vσ ¼ 0:45 . . . 0:60); Ko – give good
ence values of NSS parameters (Table 5): mean (B B=1.05…1.07) and satisfactory variabil
OCRe =OCR, Koe =Ko , σhoe =σho , . ity (Vσ ¼ 0:23 . . . 0:26).
Regression equations for determining: – give In order to improve the accuracy of determining
unsatisfactory mean (B B=1.56…1.60) and variability OCR and , the possibility of their estimation
0
(Vσ ¼ 0:85 . . . 0:86); OCR, σho and σho – give rela based on the regression equations obtained for Ko
B
tively satisfactory mean (B=1.11…1.29) and great (Table 5) was investigated.
156
Formula (1) can be transformed to the form Equations (3) - (5) have features.
OCR ¼ ½Ko =ð1 - sin ’Þ]1=/ . Substituting into it
from the Table 5 regression equation Ko ðKD2 Þ, you a. They are valid only if the coefficient of lateral
can obtain semiempirical equation in form pressure at rest Ko is determined by formula (1).
OCRðKD2 ; ’Þ. If we substitute the regression equa b. Calculations by equations require knowledge of
tion for Ko from the Table 5 into the expression the soil pressure on the lateral surface, both of
the first and second sections of the LS-module
�¼ Ko vo , we obtain a semiempirical equation
0 � (σLS1 and σLS2 ). In this case, the diameters of the
KD2 ; σvo . Both semiempirical equations are
first and second sections should be Ø37 mm and
shown in the Table 6.
Verification and assessment of the applicability of Ø41 mm.
semiempirical equations were performed (Table 6). c. All equations contain not direct measurements of
For this purpose, the statistics of the ratios pressures σLS1 and σLS2 , but their normalized
OCRe =OCR and (combined) values (KD2 , σLS2 =σLS1 ).
were calculated for the
new equations and compared with the previous stat
istics given in Table 5.
Comparative analysis showed that semiempirical 4 CONCLUSIONS
equations, compared to regression ones, have less
variability (coefficient of variation Vσ is 2 times The paper shows that the parameters of the natural
less). Therefore, they allow more precise definition stress state of clay soils can be determined through
of OCR and . the use of tests by penetrometer with cylindrical lat
eral stress module (LS-module).
It is recommended to use LS-module with two
Table 6. Verification the accuracy of semiempirical
sections with different diameters. For section diam
relationships.
eters Ø37 mm and Ø41 mm, empirical dependences
are proposed to determine coefficient of lateral pres
Ratio statistics
sure at rest Ko , overconsolidation ratio OCR and
Ratio Regression equation B
B Me σ Vσ effective horizontal stress in the soil .
1.12 0.88 0.57 0.21

REFERENCES
1.05 1.07 0.25 0.23
1 . Bayne J.M. & Tjelta T.I. 1987. Advanced cone pene
BB = mean; Me = median; σ = standard deviation; trometer development for in-situ testing at Gulfaks C.
Vσ ¼ σ=B B = variability index; e = subscript, means that In Proceedings of the 19th Offshore Technology Con
parameter is calculated according to the semiempirical ference, Richardson, Texas, USA, pp. 531–540.
equation 2 . Becker, D.B., Crooks, J.H.A., Been K. & Jefferis, M.G.
1987. Work as criterion for determining in-situ & yield
stresses clays // Can Geotech. J. №24, pp. 549–594.
Thus, for preliminary geotechnical calculations, 3 . Burns S.E. & Mayne P.W. 1998. Penetrometers for Soil
parameters of the natural stress state of stiff clay and Permeability and Chemical Detection. Funding pro
firm-stiff loam according to the LS-module measure vided by NSF and ARO issued by Georgia Institute of
ments are recommended to be determined using the Technology Report No GIT-CEEGEO-98-1, July 1998.
following equations: Georgia Institute of Technology. 144 p.
4 . Campanella R.G., Sully J.P., Greig J.W. & Jolly G.
1990. Research and development of a lateral stress
piezocone. Transportation Research Record. No. 1278,
pp. 215–224.
5 . Casagrande, A. 1936. The determination of the precon
solidation load and its practical significance. Proc. First
Intern. Conf. on Soil Mech. & Found. Eng. Cambridg.
pp. 60–64.
6 . Howie J.A., Campanella R.G. & Rivera Cruz I. 2014.
Evaluation of the UBC lateral stress module. Proceed
ings of the 3rd International Symposium on Cone
Penetration Testing, Las Vegas, USA. pp. 497–505.
7 . Huntsman S.R., Mitchell J.K. & Klejbuk L.W.Jr. 1986.
where = normalized total Shinde S.B. Lateral stress measurement during cone
soil pressure on lateral surface of LS-module second penetration. Proceedings of the Conference of Use of
section; σLS1 and σLS2 = total soil pressures on lat In-Situ Tests in Geotechnical Engineering, Blacksburg,
eral surface of LS-module first and second sections; VA, USA. pp. 617–634.
0 8 . Isaev O.N., Sharafutdinov R.F. & Zakatov D.S. 2020a.
σvo ¼ σvo - u0 = effective vertical stress in the soil, 3LSU-CPTU dissipation soil test of long duration //
u0 = natural pore pressure in soil; ’ = angle of «JSC Research Center of Construction» Gazette. № 3
internal friction. (26), pp. 50–62.
157
9 . Isaev O.N., Sharafutdinov R.F., Zakatov D.S., in-situ test methods. Transportation Research Record.
Baukov A.Y., Pavlov S.V. & Kamenev A.A. 2020b. 1278, pp. 164–171.
3LSU-CPTU electronic lateral stress cone 14 .Sully J.P. & Campanella R.G. 1991. Effect of lateral stress
penetrometer. Geotechnics, Vol. XII, No. 1, pp. 60–72. on CPT penetration pore pressures // Journal of Geotech
10 . Lunne T., Robertson P.K. & Powell J.J.M. 2004. Cone nical Engineering, ASCE. No. 117(7), pp. 1082–1088.
penetration testing in geotechnical practice. Spon 15 . Takesue K., Isano T. 2001. Development and applica
Press, London and New York. 312 p. tion of a lateral stress cone. Proceedings of the Inter
11 . Meyerhof G.G. 1976. Bearing capacity and settlement national Conference on In-situ Measurement of Soils
of pile foundations // Journal of Geotechnical Engin Properties and Case Histories, Bali, India. pp.
eering, ASCE. V. 102. GT3, pp. 197–228. 623–629.
12 . Ryzhkov I.B. & Isaev O.N. 2016. Cone penetration 16 . Jaky J. 1944. Anyugalmi nyomas tenyezoje (The coef
testing of soils in geotechnics. Stockholm, Sweden: ficient of earth pressure at rest) // Magyar Mernok es
Bokforlaget Efron & Dotter AB. 385 p. Epitesz Egylet Kozlonye (Journal for Society of Hun
13 . Sully J.P. & Campanella R.G. 1990. Measurement of garian Architects and Engineers). October, pp.
lateral stress in cohesive soils by full-displacement 355–358.
158
© 2022 the Author(s), ISBN 978-1-032-31259-0
Towards correlating seabed penetrometer and chirp sonar measurements
R. Jaber & N. Stark

Virginia Tech, Blacksburg, VA, USA
ABSTRACT: Cone Penetration Testing (CPT) and seabed coring represent key methods to determine geo
technical properties offshore. Physical testing is typically combined with geophysical surveying, i.e., chirp
sonar sub-bottom profilers, and results are qualitatively related to extrapolate geotechnical properties between
physical test locations. Literature suggests that more quantitative relationships can be established between geo
technical properties and geoacoustic responses. However, specific correlations are still rare, and correlating
frameworks are often complex and in need of additional parameters. This study attempts to explore correlations
between geotechnical properties measured using a portable free fall penetrometer (PFFP) and determined from
soil samples and chirp sonar at a frequency of 8-12 kHz. Preliminary results focusing on fine-grained sediments
using data collected in the York River, Virginia, USA suggest relations between sediment properties using
PFFP measurements and geoacoustic properties using chirp measurements and laboratory testing.
1 INTRODUCTION (Harris et al. 2008) to geoacoustic models that investi

gate the theory of sound propagation using direct or
The development of coastal infrastructure and off indirect methods (Ballard 2017, Hamilton 1980).
shore renewable energy, scour and erosion prediction Geoacoustic parameters such as acoustic impedance,
and assessment, as well as naval applications require reflection coefficient, backscatter intensity, index of
a thorough understanding of seabed sediment proper impedance, and others have been related to geotech
ties (Harris et al. 2008). In-situ geotechnical testing nical soil properties such as porosity, particle size and
such as Cone Penetration Testing (CPT), soil sam bulk density. However, specific geoacoustic frame
pling, and core extraction are typically relied on to works readily available and including geotechnical
determine geotechnical seabed properties (Saleh & soil properties particularly from in-situ testing are still
Rabah 2016). Geophysical methods provide high- rare, and often hampered by the need for and sensitiv
resolution bathymetry, sub-surface soil strata mapping, ity to multiple input parameters which may be difficult
seabed morphology, and even estimates of sediment to obtain (Pecknold & Osler 2011). Challenges also
properties (Ballard 2017). The ability of geophysical include the spatial variability of seabed stratification
methods to survey large areas of the seabed efficiently on the order of centimeters which obscures field cali
have contributed significantly to the knowledge and bration and validation (Jackson & Richardson 2007).
understanding of the seabed (Saleh & Rabah 2016, This study makes an attempt to relate portable
Jackson & Richardson 2007). The development of free fall penetrometer measurements (PFFP) and
more robust frameworks that utilize penetrometer test chirp sonar sub-bottom profiling at 8-12 kHz
ing for the calibration of geophysical tools to derive towards a direct correlation between penetrometer
geotechnical properties with high accuracy and reli tests and geoacoustic response for seabed site inves
ability from geophysical surveying while reducing the tigation. With regards to geoacoustic properties, this
need for physical testing and coring has the potential study focuses on acoustic impedance and the reflec
to increase efficiency and decrease costs in offshore tion coefficient.
site investigation, as well as provide more geotech
nical data during scour and erosion monitoring.
Several studies examined the integration of acoustic 2 METHODS & SURVEY LOCATIONS
and geotechnical testing, towards better soil character
ization, scour assessment, and direct correlations Data was collected in the York River estuary, Vir
between geotechnical properties and acoustic measure ginia, USA, with water depths ranging from ~ 2 m to
ments (Wang & Stewart 2015, Pecknold & Osler ~ 7.5 m. For this article, a cross-transect was chosen
2011, Harris et al. 2008, Osler et al. 2006, Bull et al. that started at 37°20′23.15′′N 76°37′52.35′′W and
1998). Most correlations between geotechnical proper ends at 37°20′44.75′′N 76°37′0.16′′W (Figure 1),
ties and acoustics evolved from empirical approaches featuring predominately fine-grained sediments.
DOI: 10.1201/9781003308829-16
159
the chirp sonar surveying. The data processing from
the measured PFFP deceleration to estimate qsbc and
su followed the concept presented by, e.g., Aubeny
and Shi (2006) and Albatal and Stark (2017) and as
discussed by Stark and Ziotopoulou (2017) for the
PFFP used. A detailed description can be found in
those publications, amongst others. In this paper, the
qsbc calculations utilizes logarithmic and power law
expressions for strain rate correction factors (Steiner
et al. 2014; Stark and Ziotopoulou 2017).
Box core samples were tested in the geotechnical
laboratory for grain size analysis, laboratory miniature
vane shear, water content, and bulk density using
ASTM D1140, D4648, D2216, and D7263 respect
ively. Porosity was measured from the core samples
and deduced from phase relationships using water
content.
Figure 1. The investigated cross-shore transect in the York
River estuary indicated by the red arrow with the five sta
tionary stations (S1-S5) along the station indicated by the 2.1 Acoustic impedance from sediment samples
different colors (Map data: Google, SIO, NOAA, U.S
Navy, NGA, GEBCO). The results of porosity and bulk density were used to
calculate the expected acoustic impedance Z=ρbcb,
and reflection coefficient RF based on LeBlanc et al.’s
The chirp sonar was installed at the side of the (1992) approach summarized in the equations below:
boat with its head approximately 30 cm under water.
The entire transect was measured with the chirp
sonar with longer stationary recordings at five spe
cific locations (S1-S5) along the transect (Figure 1).
The PFFP was deployed from the same vessel simul
taneously to chirp sonar surveying at locations S1
S5. Five PFFP deployments were conducted per
location. Box core samples of ~ 30 cm length below
seabed surface were recovered at the same locations.
Chirp sonar data was collected using a firmware where cb is the compressional wave velocity, βb is
data acquisition software. The chirp sonar was oper the bulk compressibility of water, ρb is the bulk
ated at an 8-12 kHz sweep in low energy mode to density of sediment, μ is the bulk modulus of rigid
focus on sediments in the upper < 10 m of the estu ity, βg is the grain compressibility, βg =βR βw taken
ary bed (Stratabox HD manual 2016). The chirp as 4.85E-11 m2, n is the soil porosity, βw is the com
sonar sends sound signals into the water column, pressibility of bottom water, and βR is the grain com
which then penetrate the seabed. The intensity of the pressibility relative to bottom water. The bulk
signal reflected at different soil layers varies based on modulus of rigidity can be estimated using:
the different soil layer properties and interfaces with
different geoacoustic properties. It should be noted
that chirp sonars are typically deployed for deeper
soil stratigraphy investigations, while the work pre
sented here explores the deployment of chirp sonar to
investigate top-surface sediments at a sediment depth where μo is the rigidity constant taken as 5E-7 N/m2,
< 1 m in line with the depth of interest for sediment ρw is the density of water, and η is an arbitrary expo
dynamics as well as with the PFFP penetration depth. nential power constant, taken as 1.
Portable free fall penetrometers (PFFP) are a rapid The reflection coefficient can then be calculated:
in-situ investigation tool of the uppermost seabed
layers (Randolph 2016). PFFP results have evolved
from rapid seabed characterization to detection of soil
layering, characterization of geotechnical sediments,
mapping of sediment type, and more quantitative site
characterization (Stark & Wever 2009, Stoll & Akal
1999). The PFFP was used in this study to estimate where Zw is the impedance of water (ρw/βw)1/2, taken
quasi-static bearing capacity (qsbc) and undrained as 1.53E06 kg/m2s and Z is the acoustic impedance
shear strength (su) of seabed surface sediments. The of seabed sediments.
specific lightweight instrument enabled multiple tests At each of the stationary sites, the chirp sonar
per location in a rapid manner and simultaneously to recorded the same location for a few minutes.
160
The recordings were saved in segy format files and of 1.69 g/cm3 at sites S3 and S5, respectively.
were read and processed in a MATLAB code that Hence, following LeBlanc et al. (1992) (Eqs. 1-4),
displays the variation in the signal amplitude enve the lowest acoustic impedance Z and reflection coef
lope (without phase information) with seabed depth. ficient RF based on sediment properties predicted
The multiple reflection approach suggested by Bull from the core samples are 1.9E+06 kg/m2s and 0.11 at
et al. (1998) was then tested to calculate the reflec S3 respectively, and the highest are 2.3E+06 kg/m2s
tion coefficient of the seafloor using the amplitude and 0.2 at S5, respectively (Table 1).
and the two-way travel time of the primary seabed
reflector and the first multiple of the seabed reflector
as shown in the formula below: Table 1. Sediment geotechnical and acoustic properties
for sites S1-S5.
ρb Z
Site n (g/cm3) (kg/m2s) RF
S1 0.78 1.28 2.0E+06 0.13

where TWTm is the two-way travel time of the first S2 0.74 1.54 2.2E+06 0.18
multiple of the seabed reflector, Am is the amplitude S3 0.81 1.21 1.9E+06 0.11
of the first multiple seabed reflector, TWTp is S4 0.83 1.46 2.0E+06 0.15
the two-way travel time of the primary seabed S5 0.75 1.69 2.3E+06 0.20
reflector, Ap is the amplitude of the primary seabed
reflector. The acoustic impedance of the sediments
can then be calculated from the reflection coeffi
cient results using Eq. 4. The reflection coefficient Table 2 shows the Z and RF derived from the
was calculated after applying a moving average of chirp sonar following the approach by Bull et al.
the trace range amplitudes to account for local (1998) (Eq. 5) for sites S1-S5. The presented ranges
noise variations. account for variability observed during the stationary
measurements. It should be noted that “stationary”
still included some boat motion within the range of
3 RESULTS a single anchor line.
The chirp sonar measurements along the measured

transect as derived from the manufacturer software is Table 2. Acoustic impedance and reflection coefficient
shown in Figure 2 with S1-S5 highlighted. In the fol results from the chirp sonar.
lowing sections, results from the stationary chirp
recordings, PFFP deployments, and soil testing of core Z
samples collected at sites S1 through S5 are presented. Site (kg/m2.s) RF
S1 3.6-4.6E+06 0.40-0.50
S2 3.7-8.5E+06 0.42-0.70
S3 3.2-4.6E+06 0.35-0.50
S4 1.9-6.0E+07 0.85-0.95
S5 4.6-9.8E+06 0.50-0.73
The reflection coefficients estimated ranged from

0.35 up to 0.73 for all sites, except for S4, where the
results seem to be drastically higher than the rest of
the sites. The acoustic impedance varied between 3.2E
+06 kg/m2s and 9.8E+06 kg/m2s, except for S4 also
Figure 2. Depiction of chirp sonar backscatter intensity (as where the results reached up to 6.0E+07 kg/m2s. No
relative digital number) along the cross-transect at York clear reason for the observations at S4 was found. It
River, VA. The five different locations surveyed are may be hypothesized that gas was present in the sedi
marked by the arrows with annotations S1-S5. ment from biogenic processes, but due to a lack of
data to confirm or reject the hypothesis, chirp results
at S4 will be rejected from further analysis.
3.1 Chirp sonar
The porosity varied slightly with a minimum and
3.2 PFFP results
a maximum of 0.74 and 0.83 at sites S2 and S4,
respectively (Table 1). The bulk density ranged A representative deceleration profile for each of the
between a minimum of 1.21 g/cm3 and a maximum five sites (S1-S5) along the transect is shown in
161
Figure 3. A maximum deceleration (dec) of ~ 3.5 the multiple approach adopted by Bull et al.
g was recorded for sites S1, S4 and S5 and a higher (1998) might not be applicable. A reason for
maximum deceleration of ~ 5.7 and 7.5 g was recorded this may be the local water depth. Warner
for sites S2, and S3, respectively. The increase in (1990) initially introduced this approach for
deceleration with depth at sites S1, S4, and S5 seems deep water investigations, and Bull et al. (1998)
gradually linear compared to a steeper increase at S2 applied it to water depths of ~ 9 m. The water
with some irregularities observed in the profile depth along the tested transect reached only ≤
between 20-30 cm. The deceleration profile at S3 can 7.5 m. Also, multiples are only clearly visible at
be divided into two layers, an upper softer layer some of the target locations when revisiting the
(above ~ 40 cm), and a stiffer lower layer (below ~ initial output (Figure 2). This necessitates the
40 cm). need to explore different processing methods to
The maximum deceleration was then used to calculate the reflection coefficient, likely utiliz
calculate the maximum qsbc and su using strain ing phase information that were not collected
rate parameters k and β ranging from 0.1 to 0.15, here for simplification, or to investigate the
and 0.035 to 0.085, respectively (Chow et al. application of an inverse model as suggested by
2017, Steiner et al. 2014). Cone factor (Nkt) Wang & Stewart (2015). However, it should be
values typically reported for CPT in fine-grained noted that despite the mismatch in Z and RF
sediments range between 4 and 20 (Robertson & between the geoacoustic properties estimated
Cabal 2015). Mayne & Peuchen (2018) reported using the chirp measurements and those esti
Nkt ranging from 8 to 25 based on a database mated from sample testing, both seem to follow
composed of 62 clays. Therefore, cone factors of a similar trend with respect to the different loca
12.3 and 10 were chosen for this study based on tions. For instance, sites S2 and S5 record the
the values reported by Mayne & Peuchen (2018) highest reflection coefficients while sites S1 and
for offshore normally consolidated clay and sensi S3 record the lowest, as reported by both meas
tive clays, respectively. urement methods. This supports further evidence
for the expected response of the chirp sonar
backscatter intensity based on differences in
sediment properties; however, also highlights
limitations of the Bull et al. (1998) approach to
accurately derive Z and RF for the shallow
water – shallow penetration conditions encoun
tered here.
Table 3 summarizes the main results of the
PFFP, represented by the maximum dec, qsbc,
and su along with the chirp and laboratory testing
results represented by one of the geoacoustic
properties, RF. The multiple qsbc and su results
reported per site correspond to the different strain
rate correction methods, parameters, and cone
factors used.
The variation in the qsbc values at each site
based on the different k and β reaches up to ~
10 kPa, with the variations becoming more sig
nificant at large deceleration values (e.g. S2 and
S3). The power-law equation for the strain rate
Figure 3. Deceleration and qsbc-depth profiles for sites correction with β of 0.085 always seem to give
S1-S5. Different colors represent the different sites. the lowest qsbc result. Similarly, the undrained
shear strength values seem to vary between 0.3
up to 0.9 kPa with the largest variations reported
4 DISCUSSION at sites S2 and S3. Sediment strength was rela
tively consistent towards the estuary banks (S1,
Based on LeBlanc et al. (1992) and others, these S4, and S5) and higher and more variable in the
values suggest sediments coarser than sand, and center of the estuary (S2 and S3). The signifi
thus, represents not only a mismatch to the esti cant difference of S4 observed in the chirp
mates obtained from sediment samples, but also sonar was not reflected in the PFFP. Acoustic
with expectations based on literature. The differ methods are highly sensitive to gas, while gas
ence between both reflection coefficients varied content would be expected to decrease sediment
between a factor of 2 and 4. However, no con strength, the threshold for detection may be
sistent factor was obvious, likely indicating that higher.
162
Table 3. PFFP results (dec, qsbc, su), chirp and sample a higher reflection coefficient RF was suggested
testing results (RF) at the five sites (S1-S5) for different from the sediment samples and confirmed by the
strain rate parameters and cone factors. chirp sonar (Table 3).
Despite the limited variability in the seabed sedi
dec qsbc su su RF RF- ments, results seem to suggest certain trends between
Site k/β (g) (kPa) Nkt-12.3 Nkt-10 chirp sample
sediment strength and reflection coefficient, as meas
S1 0.1 3.4 24.1 2.0 2.4 0.45 0.13
ured by the PFFP and the chirp sonar. This encour
0.15 22.1 1.8 2.2 ages further work that would follow-up on current
0.035 24.6 2.0 2.5 findings and address limitations and shortcomings
0.085 19.5 1.6 2.2 mentioned here. This mainly includes considering dif
S2 0.1 5.7 40.2 3.3 4.0 0.56 0.18 ferent soil types, layered systems, and comparisons at
0.15 36.9 3.0 3.7 different sediment depths. Additionally, it includes
0.035 41.1 3.3 4.1 exploring reasons behind the mismatches observed
0.085 32.1 2.6 3.2 between the RF results based on the chirp and sample
S3 0.1 7.5 53.3 4.4 5.3 0.43 0.11 testing and exploring different processing methods
0.15 49.3 4.0 4.9 that are better suited for shallow water depths.
0.035 54.5 4.4 5.5
0.085 43.7 3.6 4.4
S4 0.1 3.4 24.2 2.0 2.4 - 0.15 5 CONCLUDING REMARKS
0.15 22.3 1.8 2.2
0.035 24.7 2.0 2.5 This paper investigates the relations between
0.085 19.5 1.6 2.0 reflection coefficients and acoustic impedance pre
S5 0.1 3.4 24.3 2.0 2.4 0.62 0.20 dicted from sediment samples and observed from
0.15 22.4 1.8 2.2 chirp sonar measurements with portable free fall
0.035 24.8 2.0 2.5 penetrometer (PFFP) testing along a transect cross
0.085 19.6 1.6 2.0 ing the York River estuary characterized by fine
sediments. The presented data set is a small
excerpt of a larger data set that includes many dif
ferent seabed sediment characteristics. A limitation
Focusing on S1 and S5 showing similar PFFP of the presented data set is the limited variability
profiles with soft sediments (Figure 3 and Table 3), of the seabed sediments. Nevertheless, the data
significant differences were noted in the chirp sonar suggests that locations of increased sediment
(Figure 1 and Table 2), and similarly, differences strength (S2 with an undrained shear strength esti
were noted in the sediment cores’ bulk density and mated of up to 4.1 kPa within the uppermost
porosity (Table 1). This may suggest that those slight meters of the seabed surface) also featured lower
variations in porosity (n = 0.75 and n = 0.78) affect porosity suggesting a higher reflection coefficient
the geoacoustic response noticeably, particularly which was confirmed by the chirp sonar measure
when considering the possibility of abundant gas ments. It should be noted that this trend was vis
content), but may not affect the sediment strength, ible even with the limited sediment variability of
and thus, the PFFP to the same degree. lower undrained shear strength sites reaching up to
Location S3 suggested an overall higher sedi 2.5 kPa within the upper meter of the seabed sur
ment strength based on the maximum values face. However, the data also revealed the need for
throughout the penetration; however, matched the further considerations such as: i) Which sediment
observations at S1 and S5 up to a penetration depth depths or sediment depth ranges should results be
of 40 cm. As sediment cores were only obtained to compared at? ii) Related to i), how should layering
a penetration depth of 30 cm and the method be addressed then? iii) The effects of gas content
applied to estimate Z and RF from the chirp sonar on geotechnical properties versus geoacoustic prop
bases on surface layer reflection, a discussion of erties. iv) Possible limitations regarding water
PFFP vs. chirp sonar for S3 may be included in the depth in which chirp sonar is deployed in. v)
discussion of S1 and S5. Then, S3 matches well What is the sensitivity of the PFFP and the chirp
with S1, and S5 strikes as an outlier (Table 3). sonar to different sediment properties? These ques
Revisiting Figure 3 reveals that S5 shows stronger tions will be addressed further in follow-on work.
reflections and variability with multiple stronger
reflectors within the top meters of the seabed. Simi
lar to S4, the presence of gas may be hypothesized. ACKNOWLEDGEMENTS
Additionally, the shallower water depth approaches
the systems minimum, possibly also affecting the The authors of the paper would like to thank the
chirp response. National Science Foundation (NSF) through grant
Finally, location S2 suggested an increased sedi CMMI-1751463 and the Naval Research Lab
ment strength for the entire PFFP penetration depth through grant N00173-19-1-G018 for funding this
over all other deployment locations, and indeed, work. The authors would also like to thank Grace
163
Massey, Cristin Wright and Liz Smith for their help Mayne, P. W. & Peuchen, J. 2018. CPTu bearing factor Nkt
in data collection, and Joe Calantoni and Jesse for undrained strength evaluation in clays. Proc. of Cone
McNinch for valuable discussions of the topic. Penetration Testing 2018 (CPT’18). Delft, Netherlands,
June 21-22, 2018.
Osler, J, Furlong, A. & Christian, H. 2006. Acoustic sens
REFERENCES ing techniques for the shallow water environment: inver
sion methods and experiments. 2nd Workshop on
Albatal, A. & Stark, N. 2017. Rapid sediment mapping and Experiments Acoustic Inversion Techniques for Assess
in situ geotechnical characterization in challenging ment of the Shallow Water Environment, Ischia, ITALY.
aquatic areas. Limnology and Oceanography: Methods, Pecknold, S. & Osler, J. C. 2011. Sensitivity of acoustic
15(8), 690–705. propagation to uncertainties in the marine environment
ASTM D4648M. 2016. Standard Test Methods for Labora as characterized by various rapid environmental assess
tory Miniature Vane Shear Test for Saturated Fine- ment methods. Ocean Dynamics, 62(2), 265–281.
Grained Clayey Soil. West Conshohocken, PA; ASTM Randolph, M.F. 2016. New tools and directions in offshore
International. site investigation. Australian Geomechanics, 51(4),
ASTM D1140. 2017. Standard Test Methods for Determin 81–92.
ing the Amount of Material Finer than 75-μm (No. 200) Robertson, P.K. & Cabal, K.L. 2015. Guide to cone pene
Sieve in Soils by Washing. West Conshohocken, PA; tration testing-6th Edition. Gregg Drilling and Testing
ASTM International Inc, Signal Hill.
ASTM D2216. 2019. Standard Test Methods for Labora Saleh, M. & Rabah, M. 2016. Seabed sub-bottom sediment
tory Determination of Water (Moisture) Content of Soil classification using parametric sub-bottom profiler.
and Rock by Mass. West Conshohocken, PA; ASTM NRIAG Journal of Astronomy and Geophysics, 5(1),
International. 87–95.
ASTM D7263-2021. Standard Test Methods for Laboratory Stark, N. & Wever, T.F. 2009. Unraveling subtle details of
Determination of Density and Unit Weight of Soil Spe expendable bottom penetrometer (XBP) deceleration
cimens. West Conshohocken, PA; ASTM International. profiles. Geo-Marine Letters, 29(1), 39–45.
Ballard, M. S. & Lee, K. M. 2017. The acoustics of marine Stark, N., & Ziotopoulou, K. 2017. Undrained shear
sediments. Acoust. Today, 13(3), 11–18. strength of offshore sediments from portable free fall
Bull, J. M., Quinn, R., & Dix, J. K. 1998. Reflection coeffi penetrometers: theory, field observations and numer
cient calculation from marine high resolution seismic ical simulations. Proc. of Offshore Site Investigation
reflection (Chirp) data and application to an archaeo Geotechnics 8th International Conference, Society
logical case study. Marine Geophysical Researches, 20(1), for Underwater Technology, South Kensington,
1–11. 391–399.
Chow, S. H., O’Loughlin, C. D., White, D. J., & Steiner, A., Kopf A. J., L’Heureux J. S., Kreiter S.,
Randolph, M. F. 2017. “An extended interpretation of Stegmann S., Haflidason H. & Moerz T. 2014. In situ
the free-fall piezocone test in clay.” Géotechnique, 67 dynamic piezocone penetrometer tests in natural clayey
(12), 1090–1103. soils-a reappraisal of strain-rate corrections. Canadian
Hamilton, E. L. 1980. Geoacoustic modeling of the sea Geotechnical Journal. 51: 272–288.
floor. The Journal of the Acoustical Society of America, Stoll, R. D. & Akal, T. 1999. XBP - tool for rapid assessment
68(5), 1313–1340. of seabed sediment properties. Sea Technology, 40(2),
Harris, M. M., Avera, W. E., Abelev, A., Bentrem, F. W., & 47–51.
Bibee, L. D. 2008. Sensing shallow seafloor and sedi SyQwest 2016. StrataBox HD Chirp Operations and Main
ment properties, recent history. Proc. of Oceans 2008, tenance Manual. http://www.syqwestinc.com/products/
Quebec City, September 15-18, 1-11. sub-bottom-profilers/stratabox (June 5, 2019).
Jackson, D. R., & Richardson, M. D. 2007. High-frequency Wang, J. & Stewart, R. 2015. Inferring marine sediment
seafloor acoustics, The Underwater Acoustics Series, type using chirp sonar data: Atlantis field, Gulf of
Springer, New York. Mexico. Society of Exploration Geophysicists,
LeBlanc, L. R., Mayer, L., Rufino, M., Schock, S. G., & 2385–2390.
King, J. 1992. Marine sediment classification using the Warner, M. 1990. Absolute reflection coefficients from
chirp sonar. The Journal of the Acoustical Society of deep seismic reflections. Tectonophysics, 173(1-4),
America, 91(1), 107–115. 15–23.
164
© 2022 the Author(s), ISBN 978-1-032-31259-0
Development of free fall cone penetration testing system
Hyoun Kang, Osoon Kwon, Changjoo Shin, Jungmin Seo, Insung Jang & Man Dong-Woo
Korea Institute of Ocean Science & Technology, Busan, Korea
ABSTRACT: For the rigorous design of offshore structures, the geotechnical characteristics of the seabed
ground should be rigorously investigated, in the deep water, seabed ground surveys are complex equipment
and costly. Free fall type ground survey equipment that can acquire a simple and comparatively accurate
ground is developed when only the characteristic investigation of the shallow depth ground such as the sub
marine pipeline, the submarine cable burial is the main concern. In this study, FFCPT was developed. Based
on behavior results of numerical model simulation for FFCPT, it was manufactured. Water depth was obtained
using pressure sensor and penetrated depth was estimated by the double integration of acceleration signal. In
addition, we can verify vertical penetration at real field test of FFCPT using 2axis of inclinometer. So, the
results of the seabed type offshore CPTu experiment and the developed model were compared and the per
formance was evaluated.
Keywords: FFCPT (cone penetration testing system), tip resistance, sleeve friction, pore pressure
1 INTRODUCTION sediment interface which is particularly helpful

on high porosity fluid-mud seabed.
Recently, various types of equipment that are operated CPT(Cone Penetration Test) is the most accurate
underwater have been developed. Among them, it is field survey method to measure the strength and
very important to characterize the seabed ground for characteristics of the ground in the field. In the
the movement of equipment. However, there is not ocean, CPT(Cone Penetration Test) are conducted in
much technology of the equipment to understand the a variety of ways.
characteristics of the seabed ground, size of the equip In this study, compared to the equipment used on
ment is large, and It takes a very complicated process land, it is relatively easy to carry in the marine field,
to identify the characteristics of the seabed ground. A Free Fall Cone Penetration Testing System
The Free-Fall cone penetration testing system (FFCPT) was developed to obtain accurate ground
(FFCPT) has been developed as a tool for the rapid data(tip resistance, friction resistance). and the char
environmental assessment of seabed and water column acteristics of the seabed ground were identified by
properties for anti-submarine warfare and mine counter the sea trials.
measure operations In addition, the depth of subsea ground and
The FFCPT makes direct measurements of ground data were finally compared with the results
geotechnical properties of the seabed. It has the of the seabed type offshore cone penetration test and
same scaling factors as a conventional pushed Free Fall Cone Penetration Testing System in the
cone penetrometer, but with a larger diameter to same area to assess performance.
house the instrumentation and power supply. It is
designed to free-fall into the seabed and to sur
vive impacts with rock, if and when that happens. 2 SITE CONDITIONS AND INVESTIGATION
It consists of a nose cone instrumented with geo
technical sensors, power supply, electronics, and In this study, experiments were conducted in
tail pressure sensor. It measures acceleration and Gunsan, located on the west coast of South Korea, as
dynamic sediment pore pressure as a function of shown in Figure 1.
depth of penetration into the seafloor. The FFCPT As shown in Figure 2, the feasibility of the experi
also records hydrostatic pressure, to monitor its mental location was evaluated by the seabed type
descent velocity during free-fall, and optical offshore cone penetration test subsea survey results,
backscatter for the detection of the water- and the subsea ground survey results were shown.
DOI: 10.1201/9781003308829-17
165
Figure 1. Test site location map.
Figure 2. Result of test site location.
The water depth is about 1.0 meter at low tide

and about 8.0 m at high tide. The prevailing cur
rent velocity is 1.59 m/sec at maximum and
0.80 m/sec at minimum. The seafloor at the test
site consists of roughly 8.0 meter of clayey silt
or sandy clay underlain by medium silty sand. As
shown in Figure 2, the characteristics of the
ground as a distribution map.
3 DEVELOPMENT OF FFCPT
FFCPT should not be disturbed by power or commu

nication line, in order to increase penetration force
during freely falling. Therefore, all parts such as Figure 3. Configuration of FFCPT.
power source, electronic board, sensors, etc. must be
installed in body of the FFCPT except recovery wire.
Figure 3 is the canister for electronic part is made Several sensors and electronic parts such as DAQ,
of aluminum alloy to have a light weight and buoy battery, etc. are embedded as shown in Figure 4(a).
ancy compared to the rod and cone probe which are To obtain these signals, NI crio for DAQ was
made of iron. Since the center of buoyancy is set to applied. Estimation of the penetration depth of
be 0.6m upper than the center of mass, the flip-over FFCPT using acceleration is the fundamental data
motion of FFCPT can be prevented during free fall. for ground characteristics analysis, so dedicated
166
equipment that can accurately measure it is required. performance of the developed FFCPT, free fall experi
In addition, the acceleration signal and other sensor ments were repeated five times and data were obtained.
signals require time synchronization. For this pur
pose, a dedicated module for vibration measurement
and a DC-type acceleration were used, and when
other sensor signals were input, time synchronized
measurements were performed based on the acceler
ation signal. A 2-axis inclinometer is installed for
tracking the underwater free fall motion. The pres
sure sensor was applied to check the water depth at
the test point. It is manufactured so that it can oper
ate continuously for more than 8 hours after power
on using large-capacity battery. According to New
ton’s Law, the underwater terminal speed just before
impact is calculated by integration of the obtained
acceleration.
Figure 6. Drop test of FFCPT_1.
Figure 4. Detailed design of electronic part in the FFCPT.
4 TESTING METHODS AND RESULTS
4.1 Testing methods

Figure 5 is the an experiment was conducted to
obtain subsea ground data (cone tip resistance, fric
tion resistance) using the developed FFCPT Figure 7. Drop test of FFCPT_2.
4.2 Result of test

In this study, subsea ground characteristics data(cone
tip resistance, friction resistance) were obtained
using the developed FFCPT(Free Fall Cone Penetra
tion Testing System).
As shown in Table 1, the final penetration depth
was 0.21 m, 0.24 m, 0.26 m, 0.22 m and 0.49 m, and
the average was 0.28 m.
Table 1. Result of test.
Final pene
tration Cone friction
depth(m) resistance(MPa) resistance(MPa)
FFCPT-1 0.21 0.0048~5.2755 0.0~0.0153

Figure 5. FFCPT.
FFCPT-2 0.24 0.0006~4.8709 0.0~0.0118
FFCPT-3 0.26 0.0141~5.3009 0.0~0.0114
Figures 6 and 7 are the FFCPT with a length of 4m FFCPT-4 0.22 0.0131~4.2026 0.0~0.0116
and weight of 38kg was placed on a moving crane and FFCPT-5 0.49 0.0006~2.5260 0.0~0.0115
dropped freely in sea 20m deep. In order to verify the
167
4.2.1 Result of FFCPT-1 0.1m to 0.18m, it is judged as Clay silt to silty clay,
Figure 8 is the final penetration depth of FFCPT-1 and up to 0.25m, it is judged as Silty sand to sandy
was 0.21m and ground data (cone tip resistance, fric silt.
tion resistance) values were measured. The results of
the experiment were shown as follows 4.2.3 Result of FFCPT-3
Figure 10 is the final penetration depth of FFCPT-1
was 0.26m and ground data (cone tip resistance, fric
the experiment were shown as follows.
Figure 8. Result of FFCPT-1.
Figure 8 is the distribution range of cone tip Figure 10. Result of FFCPT-3.
resistance was measured in the range of 0.0048 to
5.2755MPa, and the distribution range of the friction Figure 10 is the distribution range of cone tip
resistance was measured in the range of 0.0000 to resistance was measured in the range of 0.0141 to
0.0153MPa. As a result of the experiment, from 0m 5.3009MPa, and the distribution range of the friction
to 0.1m, it is judged as Clay silt to silty clay, and up resistance was measured in the range of 0.0000 to
to 0.2m, it is judged as Silty sand to sandy silt. 0.0114MPa. As a result of the experiment, from 0m
to 0.2m, it is judged as Silty sand to sandy silt, and
4.2.2 Result of FFCPT-2 up to 0.27m, it is judged as Clay silt to silty clay.
Figure 9 is the final penetration depth of FFCPT-1
was 0.24m and ground data (cone tip resistance, fric 4.2.4 Result of FFCPT-4
tion resistance) values were measured. The results of Figure 11 is the final penetration depth of FFCPT-1
the experiment were shown as follows. was 0.22m and ground data (cone tip resistance, fric
the experiment were shown as follows

Figure 9 is the distribution range of cone tip
resistance was measured in the range of 0.0006 to
4.8709MPa, and the distribution range of the friction Figure 11 is the distribution range of cone tip
resistance was measured in the range of 0.0000 to resistance was measured in the range of 0.0131 to
0.0118MPa. As a result of the experiment, from 0m 4.2026MPa, and the distribution range of the friction
to 0.1m, it is judged as Silty sand to sandy silt, from resistance was measured in the range of 0.0000 to
168
0.0116MPa. As a result of the experiment, from 0m locations of the test sites because it was judged that
to 0.17m, it is judged as Clays-clay to silty clay, and there was a gap of up to 5 to 10 meters due to the
up to 0.22m, it is judged as Silty sand to sandy silt. movement of the ship.
Due to the following conclusions, it is deemed
4.2.5 Result of FFCPT-5 necessary to estimate the appropriate depth of water
Figure 12 is the final penetration depth of FFCPT-1 in future field experiments. In addition, it is deemed
was 0.49m and ground data (cone tip resistance, fric necessary to increase the weight of the FFCPT even
tion resistance) values were measured. The results of if the depth of the water is not deep enough to
the experiment were shown as follows. improve the final penetration depth.
Through this study, the basis of underwater behav
ior analysis of the FFCPT for seabed penetration was
secured. If this equipment is improved and apply to
obtain the ground characteristics in deep seabed, it is
expected to be used very conveniently at low cost.
ACKNOWLEDGMENT
This research was supported by the project “Devel

opment of the support vessel and systems for the off
shore field test and evaluation of offshore
equipments (PM62241)” funded by Korea Institute
of Ocean Science & Technology(KIOST).
Figure 12. Result of FFCPT-5. REFERENCES

[1] Changjoo Shin, Sung Gyu Won, OSoon Kwon, In
Figure 12 is the distribution range of cone tip Sung Jang, Jung-min Seo, Hyoun Kang, 2019. Estima
resistance was measured in the range of 0.0006 to tion of Ground Penetration Depth for Underwater Free
2.5260MPa, and the distribution range of the friction Fall Cone Penetration Tester. Proceeding of Korean
resistance was measured in the range of 0.0000 to Marine Robot Technology Society, Seoul, South Korea,
0.0115MPa. As a result of the experiment, from 0m 62–64.
to 0.3m, it is judged as Clays-clay to silty clay. [2] By John Osier, Arnold Furlong, Harold Christian “The
Integration of the Free Fall Cone Penetrometer(FFCPT)
with the Moving Vessel Profiler (MVP) for the Rapid
5 CONCLUSIONS Assessment of Seabed Characteristics”, “International
hydrographic review”, Vol. 7 No. 3, 2006, 46–47
[3] Evans and Ashley D., “Hydrodynamics of mine impact
FFCPT was developed to obtain seabed geotechnical burial,” Naval Postgraduate School, 2002.
properties conveniently. After the FFCPT falls under [4] John Osler, Arnold Furlong, Harold Christian and
water and impacts on seabed ground, the device Mike Lamplugh, “The integration of the free fall cone
acquires the geotechnical properties during penetra penetrometer(FFCPT) with the moving vessel profiler
tion. It is very important to secure the penetration (MVP) for the rapid assessment of seabed
depth because the equipment measures the properties characteristics,” The International Hydrographic
according to the penetration depth. Sea-trial free-fall Review, vol. 7(3), pp. 45–53, 2006.
test was performed. [5] Koumoto, T et. al, “Theory and practice of the fall
Analysis of this study, subsea ground characteris cone test”, Vol. 51, No. 8, 2001, pp.701–712
[6] Kurt Seifert and Oscar Camacho, “Implementing pos
tics data were obtained using the developed FFCPT. ition algorithms using accelerometers,” Freescale
The analysis results are as follows. Semiconductor Application Note, AN3397, 2007
The final penetration depth was 0.21 m, 0.24 m, [7] McCarty JL, Carden HD, 1962. Impact characteristics
0.26 m, 0.22 m and 0.49 m, and the average was of various material obtained by an acceleration-time
0.28 m. This is judged not to penetrate more than history technique applicable to evaluating remote
0.49m because it is not deep water and it is light targets. NASA technical report(NASA-TN-D-1269),
weight of ffcpt. University of Texas at Austin.
FFCPT1-5 are the distribution range of cone tip [8] Stegmann, Sylvia et. al, “Design of a modular, marine
resistance was measured in the range of 0.0006 to free-fall cone penetrometer”, Sea Technology, Vol. 47,
No. 2, 2006, pp.1–27.
5.3009MPa, and the distribution range of the friction [9] Stegmann, Sylvia et. al, “Initial Results of a new Free
resistance was measured in the range of 0.0000 to Fall-Cone Penelromeler (FF-CPT) for geotechnical
0.0153MPa. in situ characterisation of soft marine sediments”, Vol.
According to the comparative analysis of the 86, No. 3, 2006, pp.1–199.
seabed type offshore cone penetration test and [10] Sung-Chul Lee, 2008. A study on the Estimation on
FFCPT, it was difficult to accurately compare the Undrained Shear Strength and Preconsolidation Pressure
169
Using N-Value Obtained from Standard Penetration envelopes. Journal of Measurements in Engineering,
Test, Graduate School of Korea Maritime University. 4(2),117–121.
[11] Sylvia Stegmann, Tobias Morz and Achim Kopf, “Initial [13] Youngshin Consultant, [Online] (Updated 2010)
results of a now free fall-cone penetrometer for geotech Available at:<http://www.piletest.net/product_3_2_4.
nical in-situ characterization of soft marine sediments,” html> [Accessed July 2019].
Norwegian Journal of Geology,. 86(3), pp. 199–208,
2006. The Integration of the Free Fall Cone Penetrometer
[12] Yanli Yang, Yanfei Zhao, Dali Kang, 2016. Integra (FFCPT) with the Moving Vessel Profiler (MVP) for the
tion on acceleration signals by adjusting with Rapid Assessment of Seabed Characteristics
170
© 2022 the Author(s), ISBN 978-1-032-31259-0
Evaluation of statistical fluctuation of measured data from nuclear density

cone penetrometer
Muthusamy Karthikeyan
Surbana Jurong Consultants Pte Ltd, Singapore
ABSTRACT: This paper describes research conducted to investigate the statistical fluctuation and interpret
the density profiles obtained by a nuclear-density cone penetrometer (ND-CP). The laboratory results show
that the amount of fluctuations present in the ND-CP measurements can be approximated by Normal or Gauss
ian distribution. The coefficient of variation for RI count and BG count data’s is 2.4% and 15.1% respectively.
The moving-average technique has been employed for the measured raw data at different time intervals in
order to reduce the statistical fluctuation. It was found that the error due to the statistical fluctuation in ND-CP
measurement is less than 1% if an averaging span of 10 cm is used. The raw data obtained from the field
results when averaged over 10cm depth shows that the cone profile count distribution trend is not distorted or
deformed compared to the original raw data, which means the loss of information about the density through
the filtering is negligible. This averaging is necessary to smooth the measured wet density profile. Therefore,
the ND-CP measurements are averaged over a span of 10cm depth to minimize the statistical fluctuations.
1 INTRODUCTION thickness of each layer change, the ND-CP measure

ments will produce a pattern that changes correspond
The in-situ wet density of soil can be measured reli ingly. An ability to interpret this signature is essential
ably using Nuclear-Density Cone Penetrometer (ND to the profiling of such soil, which is a major chal
CP) (Shibata et al., 1993; Karthikeyan et al., 2001; lenge. Any measurement based on observing the radi
2004; Tan et al., 2004; Dasari et al., 2006 and Karthi ation emitted in nuclear decay is subject to some
keyan et al., 2007; Karthikeyan and Tan, 2008; Rui degree of statistical fluctuation. These inherent fluctu
et al. 2013). Most of the earlier researchers men ations represent an unavoidable source of uncertainty
tioned above have focused on the effectiveness of the in all nuclear measurements and often can be the pre
ND-CP in determining the soil properties for settle dominant source of imprecision or error. To account
ment analysis, evaluation of liquefaction potential of for uncertainty in cone measurement, knowledge of
sandy deposits and determination of the strength of the uncertainties should be understood. In this paper,
sub-soil after ground improvement. Recently, the results from an experimental investigation will be pre
ND-CP was used to characterize a reclaimed land sented to investigate the statistical fluctuation and
(Karthikeyan et al., 2004, Dasari et al., 2006 and interpret the density profiles obtained by a nuclear-
Karthikeyan et al., 2007). density cone penetrometer (ND-CP) so as to improve
Natural deposits laid down under water are often the interpretation of the actual soil profile.
layered. Sometimes, these layers can be identified by
changes in the density of soil. The ND-CP measure
ment is affected by the measuring volume and if this 2 DESCRIPTON OF NUCLEAR-DENSITY
volume is occupied by layers of different densities, CONE PENETROMETER (ND-CP)
then the relationship between the observed count rate
and true wet densities become correspondingly more Figure 1 shows major components of a ND-CP. The
complex (Karthikeyan and Tan, 2008). As the meas lower part of the cone houses various sensors to
uring volume of the ND-CP traverses through this measure the usual cone parameters, namely, cone
soil with changing density, the count rate measured resistance (qc), sleeve friction (fs) and pore pressure
corresponds to a measurement of some composite (u2). The size of the lower part conforms to the
material. If the ND-CP is used to characterize soil standards recommended by the International Society
with greater heterogeneity, for example, a fill made of for Soil Mechanics and Foundation Engineering
lumps, the interpretation of the exact nature of such (ISSMFE, 1989) for cone penetration testing and
lumpy fill is more complex. As the density and ASTM Standard D 5778. The diameter of the cone is
DOI: 10.1201/9781003308829-18
171
35.6 mm and the apex angle is 60°. The base area of detailed description and working procedure of ND
the cone is 10 cm2 and the area ratio, a, is equal to CP have been discussed by Karthikeyan (2005),
0.75 which is determined from calibration measure Dasari et al. (2006), Karthikeyan et al. (2007) and
ment at the laboratory (the effort to eliminate the Karthikeyan & Cisy (2019).
adverse effects of friction sleeve and unequal cone
tip). A porous ceramic filter is located just behind
2.1 Procedure for processing the ND-CP results
the cone tip. The total length of the shaft housing the
sensors is 258 mm. After this, the shaft tapers out Current design of the ND-CP requires two prob
wardly at an angle of 15°. The tapered portion of the ings for every single measuring point, one prob
shaft is 49 mm long, above which the shaft has ing to obtain the background (BG) count of
a constant diameter of 48.6 mm and extends for naturally occurring gamma photons and another
a total length of 896 mm. This upper part houses the probing to measure the actual nuclear density (RI)
radioisotope source, the detector, and a preamplifier count. The background count is measured using
The ND-CP uses gamma-ray as the source. a dummy cone, in which only the detector is
Gamma rays interact with soil predominantly placed to measure the naturally occurring gamma
depending on the level of the energy. The Compton photons. This natural radioactive (background)
scattering is predominant within an energy range of count is a type of noise which must be subtracted
600 keV and 1.2 MeV, and is a function of the mater from the count measured to give the actual
ial density. If the detector is designed to measures nuclear density measurement. To determine the
only the incoming photons within the range wet density and water content of soil, the meas
described, then the incoming photons are a function ured raw data needs to be processed for the final
of the density of the material only. The gamma ray profiles through the following steps: (i) depth cor
source used in the construction of the ND-CP is rections (ii) noise suppression or elimination (iii)
a Cesium (Cs137) isotope with a half-life of 37.6 statistical fluctuations and averaging of measured
years, and the detector is sodium iodide activated RI cone data. Figure 2 also shows the flow chart
with thallium (NaI (TI)) scintillator mounted on for processing of measured RI cone data. The
a photomultiplier tube. The length of the NaI scintil detailed procedures for processing of measured RI
lation detector is 10.2 mm. The separation distance cone data will be discussed subsequently in the
between the source and center of gamma detector is following section.
255 mm. The ND-CP is pushed into the soil layer at
a rate of approximately 1 - 2 cm/sec and cone resist 2.1.1 Depth corrections
ance (qc), sleeve friction (fs), pore pressure (u2), As stated earlier, the upper part of the RI cone
radioisotope (RI) count are recorded continuously. houses the radioisotope source, the detector and
The intensity of the radioisotope sources used in a preamplifier. The lower part of the cone houses
these cones is very low compared to the radioiso various sensors to measure the usual cone param
topes used in medical equipment that pose problems eters, namely cone resistance, pore pressure and the
to human health. However, it is still necessary to sleeve friction. Therefore, it is important to correct
take precautionary measures to avoid radiation the measurement of the different sensors reading
exposure in humans by wearing a lead apron and into the same depth level as the other parameters to
also by frequent monitoring of radiation exposure ensure proper characterization of soil layers. There
using a radiation meter (Dasari et al., 2006). fore, the following depth correction needs to be car
In using the ND-CP, it is the net above the back ried out for raw data for further interpretations. If the
ground (BG) radioisotope count that is used. Thus, measurement center for a RI cone is considered at
a separate reading of this BG count has to be taken the cone resistance sensor (load cell), then the other
either immediately prior or after the test. The sensors readings are corrected accordingly, for
example, the depth of sleeve friction, pore pressure,
RI count, BG count are raised about 0.11 m, 0.04 m,
0.60 m and 0.36 m, respectively. The distances are
based on the function of the cone design.
2.1.2 Noise suppression or elimination

In several practical cases, it is possible that the depth-
meter reading could be affected by either delayed
response or disturbing inertial effects, caused, for
example, by velocity fluctuations (acceleration, vibra
tions, and so forth). Hence, it will affect the accuracy
of RI cone measurement. To avoid this discrepancy in
the RI cone results, it is important to measure the RI
Figure 1. Diagram of Nuclear-Density Cone Penetrometer Count and BG Count in both penetration and uplift
(ND-CP). (after Shibata et al., 1993 and Dasari et al., stage so that the retardation error can be pointed easily.
2006). The “analog recording” of the density–depth profile
172
then will show “retardation hysteresis” loops, which 2.1.3 Averaging of measured RI cone data
enclose the representative density profile. Figure 3 Radioactive decay is a random process. Consequently,
shows the typical BG count profile measured for both any measurement based on observing the radiation
penetration stage and uplift stage profile. From the emitted in nuclear decay is subject to some degree of
figure, it can be seen that there is an abnormal behavior statistical fluctuation. These inherent fluctuations rep
observed in the penetration stage measurement due to resent an unavoidable source of uncertainty in all
undesirable shock. These undesirable shocks can be nuclear measurements and often can be the predomin
removed by comparing with the uplift stage measure ant source of imprecision or error (Knoll, 2000). The
ment. Figure 3 illustrates the typical BG count profile source-originated fluctuations affect the precision of
after elimination of the undesirable noise. the RI cone. The error in density/water content due to
these fluctuations can be minimized by designing
equipment such that the radioactive count is high.
Many sources of uncertainties can be reduced with
increased sampling and good practices, but due to the
nature of geomaterials, inherent uncertainties cannot
be eliminated. Inherent uncertainties can be quantified
by a statistical analysis. The error can also be minim
ized by decreasing the statistical fluctuation through
filtering data over a short depth. Nobuyama (2000)
reported that the RI cone measurement needs to be
averaged over a depth in order to reduce the statistical
fluctuation. To re-evaluate the statistical fluctuation of
RI cone measurement, a laboratory counting experi
ments is carried out and discussed in the following
paragraphs. The understanding of these statistical
fluctuations is also important in the interpretation of
the ND-CP results for highly heterogeneous soils.
Figure 3. An example of Noise Suppression or Elimination

(after Nobuyama, 2000).
3 EXPERIMENTAL SETUP
An experimental setup was designed for studying the

statistical fluctuation of wet profile measured by ND
CP. A series of laboratory counting experiment was
carried out using the ND-CP under identical condi
Figure 2. Flow chart for processing the RI cone results. tions. One of the controlling factors is the size of the
173
laboratory calibration chamber. The density is meas because the mean and standard deviation tend to
ured using a source and detector. The signals from the change together, so typically the coefficient of vari
source interact with soil and the modified signals are ation remains fairly stable (Snedecor and Cochran,
captured by the detector. The degree of modification of 1989).
the signals is a function of material property. RI cone
measurements (density) are evaluated in an extended
volume around the central point of the radioactive
source and detector configuration, which is called the
“measuring volume”.
Based on the theory of gamma scattering and neu
tron methods, the “measuring volume” around the
source was found to be about 30 cm radius. Karthike
yan and Tan (2008) reported that the maximum radius
of the influence zone for ND-CP is 23.6 cm when
used in water and it decreases with increasing wet
density of the material. It was also established that the
ND-CP measurement provides the average wet density
of the composite soil within the measuring volume.
Therefore, a stainless-steel chamber with diameter
of 700 mm and height of 1000 mm was used for the
experiments. This calibration chamber was filled with
water and the ND-CP was kept inside at the centre of
the calibration chamber to measure the statistical fluc
tuations of RI Count and BG Count under identical
condition.
4 STATISTICAL FLUCTUATION OF ND-CP
Figure 4 shows the typical variations of RI count and

BG count data measured in water under ideal condi
tion. There are 121 separate measurements, each Figure 4. Typical variations of RI Count and BG Count
taken at a 1 sec interval. When the measurements data measured in water under ideal condition.
were being made during the 121 secs, the source of
the radiation was ‘steady’. The source of the radiation
did not change in its nature. Nevertheless, the number Figure 5 illustrates the form of normal distribution
of counts recorded per sec is clearly not constant. along with its mean and standard deviation of meas
This is the statistical nature of ND-CP measurement. ured RI count and BG count data. This figure show
The frequency of occurrence of an error of any that the amount of fluctuations present in the ND-CP
given magnitude can be calculated for a given deter measurements can be approximated by Normal or
mination by the application of the laws of probability. Gaussian distribution. It is also noted that the ND-CP
Various distribution laws are used to relate the magni measurements fluctuate within the range of ± (twice
tude of a deviation from true average and the number of the standard deviation, σ) from its mean value. The
of events which experiences that particular deviation. coefficient of variation for RI count and BG count
The normal distribution is useful for describing the data’s is 2.4% and 15.1% respectively. Nobuyama
spread in data occurring in various aspects of nuclear- (2000) reported that the error could be minimized
radiation detection (Price, 1964; and Tsoulfanidis, through filtering the data over a short depth. For this,
1995). The extent of the statistical fluctuations about the moving-average technique has been employed for
the true mean may be expressed in terms σ, the stand the measured raw data at different time intervals.
ard deviation and is approximated by normal distribu Figure 6 shows the changes in COV for different
tion (Nobuyama, 2000). The standard deviation is moving-averaging interval for RI Count and BG count
defined as the square root of the scatter data about the data. From the figures, it can be seen that the COV
mean value. The standard deviation is used commonly decreases with increasing moving-average time inter
to indicate the accuracy with which a sample of val. It is also inferred from Figure 6 that the COV is
a given size can estimate the mean value of test results. fairly constant after a moving-average interval of 6
The coefficient of variation, COV, is another measure secs to 7 secs. This further confirms that the averaging
of variability of a sample and is defined as the ratio or filtering of wet density profiles measured by ND
between the standard deviation and mean value. The CP is necessary to reduce the statistical fluctuations.
COV expresses the magnitude of variability as In field test, the ND-CP was pushed into the
a fraction or as a percentage of the mean value. This ground at a rate of approximately 1.5 cm/sec, and
facilitates comparison of data from different samples thus it will cover a distance of 10 cm within 6 – 7
174
Figure 5. Illustrating the statistical fluctuations of RI
Count and BG Count data in the form of normal distribu Figure 6. Changes in COV with different moving aver
tion along with its mean and standard deviation (σ) (a) RI aging intervals (a) RI Count (b) BG Count.
Count (b) BG Count.
secs. While comparing the BG count with RI count

in water, the BG count is too small, and it is almost
equal to the minimum detectable count of ND-CP.
Therefore, if the effect of the BG count is neglected,
as can be seen in Figure 6 (a) the error due to the
statistical fluctuation in ND-CP measurement is less
than 1% if an averaging span of 10 cm is used. This
can be also demonstrated using the field results
shown in Figure 7 (a) and (b) which were obtained
by averaging over 5 cm and 10 cm depths respect
ively. The raw data are also shown for comparison
purpose. As can be seen, when averaged over 10 cm
depth, the RI cone profile count distribution trend is
not distorted or deformed compared to the original
raw data, which means the loss of information about Figure 7. Comparison of measured ND-CP raw data and
the density and moisture through the filtering is negli averaged data (a) 5 cm average span (b) 10 cm average
gible. This filtering is necessary to smooth the density span.
profile. Therefore, the RI cone measurements that are
reported subsequently are averaged over a span of
10 cm depth to minimize the statistical fluctuations. interpretation of the actual soil profile. The labora
tory results show that the amount of fluctuations pre
sent in the ND-CP measurements can be
5 CONCLUSIONS approximated by Normal or Gaussian distribution. It
was also found that the averaging or filtering of wet
This paper describes research conducted to investi density profiles measured by ND-CP is necessary to
gate the statistical fluctuation and interpret the dens reduce the statistical fluctuations. It was found that
ity profiles obtained by a nuclear-density cone the error due to the statistical fluctuation in ND-CP
penetrometer (ND-CP) so as to improve the measurement is less than 1% if an averaging span of
175
10 cm is used. The raw data obtained from the field Karthikeyan M., Tan T.S., Mimura, M., Yoshimura, M. and
results when averaged over 10 cm depth shows that Tee C.P. (2007). “Improvements in Nuclear-Density
the cone profile count distribution trend is not dis Cone Penetrometer for Non-Homogeneous Soils”, Soils
torted or deformed compared to the original raw Found., Vol. 47 No. 1, pp. 109–117.
data, which means the loss of information about the Karthikeyan, M. and Tan, T.S (2008). Profiling of Hetero
density through the filtering is negligible. This aver geneous Soil Using Nuclear-Density Cone
aging is necessary to smooth the measured wet dens Penetrometer. Geotechnical Testing Journal, American
Society of Testing Materials, Vol.31, No. 6, pp.
ity profile. Therefore, the ND-CP measurements are 513–525.
averaged over a span of 10 cm depth to minimize the Knoll, G.F. (2000). Radiation detection and measurement,
statistical fluctuations. Third Edition, John Wiley and Sons, 801p.
Nobuyama, M. (2000). The result of the RI cone penetra
tion tests. NUS internal report no 6004998/264 sub
REFERENCES mitted by Soil and Rock Engineering Co. Ltd., Japan.
2000.
Dasari, G. R., Karthikeyan, M., Tan, T. S., Mimura, M. and
Price, W. J. (1964) Nuclear radiation detection. Second
Phoon, K.K., (2006). “In situ evaluation of radioisotope
Edition, McGraw-Hill Book Company, 430p.
cone penetrometers in clays”, Geotech. Test. J., Vol. 29,
Rui J, Hino T, Chai J, Hamada, T, Yoshimura, M. (2013).
No. 1, pp. 45–53.
Interpretation of density profile os seabed sediment
ISSMFE, (1989). “International reference test procedure
from Nuclear desnity cone penetration test results, Soils
for cone penetration test (CPT)”, In Report of the
and Foundation, Volume 53, No.5, pp.671–679.
ISSMFE Technical Committee on Penetration testing of
Shibata, T., Mimura, M. and Shrivastava, A.K., (1993). “RI
soils –TC 16, with references to test procedures, Swed
cone penetrometer experience in marine clays in Japan”,
ish Geotechnical Institute (Linkoping), Information,
Proceedings of the 4th Canadian Conference on Marine
Vol. 7, pp. 6–16.
Geotechnical Engineering, St. John’s Nfld., 27-29 Sep
Karthikeyan M. (2005). Application of Radioisotope Cone
tember 1993. Vol. 3, pp.1024–1033.
Penetrometer to characterize a lumpy fill, PhD Disserta
Snedecor, G.W and Cochran, W. G. (1989) Statistical
tion, National University of Singapore, 219p.
Methods, 8 ed. Ames: Iowa State Univ. Press.
Karthikeyan, M, Dasari, G.R., Tan, T.S., Lam, P.W.,
Tan, T. S., Karthikeyan, M., Phoon, K. K., Dasari, G. R.,
Loh, Y.H. Wei, J. and Mimura, M. (2001). “Character
and Mimura, M., (2004), “Use of a Radioisotope Cone
ization of a reclaimed land site in Singapore”, Proceed
to characterize a lumpy fill,” Proceedings of the Second
ings of 3rd International Conference on Soft Soil
International Conference on Geotechnical and Geophys
Engineering, Hong Kong, 6 – 8 December 2001, pp.
ical Site Characterization, ISC-2, Porto-Portugal, 19–
587–592.
22 September 2004, Vol. 1, pp. 801–808.
Karthikeyan, M., Dasari, G.R. and Tan, T.S., (2004). “In
Tsoulfanidis, N. (1995). Measurement and detection of
situ characterization of land reclaimed using big clay
radiation, Second Edition Taylor and Francis,
lumps”, Canadian Geotech. J., Vol. 41, No. 2. pp.
614p. 1995.
242–256.
176
© 2022 the Author(s), ISBN 978-1-032-31259-0
On the accuracy and precision of the seismic cone penetration test – a field
test study on the seismic source
O. Koreta, A.H. Augustesen & L. Krogh
Ørsted
K. Lundvig & S. Bøtker-Rasmussen

Geo
ABSTRACT: The seismic cone penetration test (SCPT) is one of the most frequently used in situ testing
methods for deriving the small strain shear modulus. Offshore SCPT results often present large scatter, in
particular when data are obtained with different operational setups. No standard or industry consensus exist in
terms of requirements to system key characteristics, operational testing procedures, interpretation and the
numerical quality control of the traces obtained. Seismic sources currently deployed as part of the offshore
industry practice have been developed individually by the various suppliers. This paper presents the findings
of an onshore experimental study with the objective to increase the understanding of the influence of the off
shore seismic source setup on SCPT results under controlled operational procedures. Accuracy and precision
of the SCPT are investigated as part of the study.
1 INTRODUCTION Most of the details concerning the listed system

parameters are not standardized in ISO (2014).
With the expansion and development of the offshore This implies that the acquisition of SCPT data off
wind industry, the small strain stiffness (Gmax) has shore suffers from a) a significant degree of scatter
become a key parameter for offshore wind facility (Gibbs et al. 2018, Masters et al. 2019), b) challenges
designs, especially for the design of monopile founda to acquire reliable data at large scale c) challenges to
tions. Gmax can be derived from laboratory and in situ replicate results in a consistent manner, d) lack of
tests. numerical control and acceptance criteria, e) inefficient
The seismic cone penetration test (SCPT) is a cost- operational procedures. Furthermore, a key concern
effective in situ technique used to derive Gmax, remains the quantification of the measurement uncer
through the measurement of the velocity (Vs) of tainty, i.e. accuracy and precision of the measurements.
propagating shear waves. The SCPT was initially An onshore field test study was performed with
developed for onshore use with simple beam sources the key objective to increase the understanding of the
equipped with sledge hammers. In contrast, the off influence of the offshore seismic source setup on
shore seismic sources are complex electro-mechanical SCPT results under controlled operational procedures.
systems and the SCPT offshore is associated with add Furthermore, accuracy and precision were investi
itional challenges compared to the onshore source; gated as part of the study. The high-level scope of
less control of the source orientation and contact with work compares performance of adjacent SCPTs using
the seabed (detached sources) and difficulties in isolat an offshore and an onshore seismic source, each with
ing the background noise originating from the vessel, a different frequency content, recorded with the same
umbilical or drill string (source attached to the frame). receivers at the same SCPT location.
The offshore seismic sources are designed and built As no standard reference test exists for the SCPT,
individually by the suppliers. The main differences the accuracy cannot be known exactly. However,
between the various offshore applications involve a) a series of laboratory (bender element and resonant
method of deployment, b) source characteristics (fre column testing) and alternative in situ (seismic dilat
quency, energy, geometry, horizontal distance between ometer, SDMT and multichannel analysis of surface
source and SCPT, orientation, method of contact), c) waves, MASW) stiffness measurements were per
type of triggers (contact, sensor), d) type and distance formed to validate and benchmark the SCPT results.
between receivers (geophone, accelerometers) e) This paper presents the background, the test pro
acquisition system and f) method of interpretation. gram, and relevant results from the field test study.
DOI: 10.1201/9781003308829-19
177
2 BACKGROUND Table 1. Summary of soil parameters.
Campanella et al. (1986) presented the SCPT for off Undrained Peak
shore applications. However, there are no compre Soil Fines Plasticity shear Friction
hensive studies on the evaluation of accuracy and type Depth content index strength angle
precision of modern offshore seismic sources. In
contrast, extensive studies have been performed in m % % kPa °
relation to the repeatability of the onshore source Sand 0.0-11.5 1-5 - - 33-39
(beam) setup (e.g. Rice, 1984, Campanella & Clay 11.5-20.0 - 18-28 200-270 -
Robertson, 1984, Robertson et al.,1986).
Rice (1984) developed a beam source (sledge
hammer), with a length between 2-3 m, loaded using
a weighted plank such as a CPT rig vehicle. High (Figures 8-9). A summary of classification and
quality and polarized signals could be achieved to strength parameters is provided in Table 1.
a depth of 30 m without stacking the signals.
Laing (1985) confirmed the results using the source
proposed by Rice (1984) and concluded that the beam 3.2 Equipment and test setup
source produced strong, directional, and repeatable The SCPTs were conducted using Geo’s CPT crawler
shear waves. The beam source is used nowadays as rig with a maximum thrust capacity of 20 tonnes
a common form to generate shear waves (ASTM (Figure 1). The 10 cm2 seismic module consists of
2019). As a consequence of preceding literature stud a dual receiver array spaced at 0.5 m distance and
ies, the beam source is used as benchmark for the equipped with triaxial accelerometers. The SDMT con
results from the offshore source. sists of a dual accelerometer dilatometer.
Gillespie et al. (1985) and Gillespie (1990) investi The onshore source is a timber beam with
gated the repeatability of Vs from SCPT through a length of 3.5 m equipped with steel plates at both
a series of onshore testing at three Norwegian research ends. The two steel hammers (left and right) have
sites, Onsøy and Drammen (clay sites) and Holmen each a length of 1 m and a weight of 12.8 kg. The
(sand site). The tests were performed using a beam offshore seismic source is a dual-hammer source
source, with the same characteristics as described by designed and built by Geo. The total weight is
Rice (1984). 1300 kg and the sampling frequency is 5000 Hz.
Three adjacent SCPT tests at the clay sites
showed agreement within ±5% of the measured Vs.
In sand, the Vs was more scattered, visually exceed
ing a variation of ±10%, which was attributed to the
variation of cone resistances between adjacent tests.
Gillespie (1990) concluded that ensuring a close
contact between source and ground will signifi
cantly improve the quality of the signal and subse
quently the identification of the arrival times. The
error on the Vs was found to be as low as ±2% for
a Vs of 200 m/s. The depth control of the test was
shown to induce an additional error of up to ±5%.
3 TESTING PROGRAM
Figure 1. Photo from the field test setup (main configuration).
3.1 Site conditions
The test site is located in a sand pit near Kongen
sbro west of Aarhus, Denmark. The soil conditions To obtain comparable testing conditions, both
were initially confirmed by a CPT (CPT2) and an seismic sources were placed with their centers at
adjacent borehole (BH1) to depths of 20 m and a distance of 2.6 m from the SCPT (Figure 2). The
17 m, respectively (Figure 3). The ground water respective orientations of both sources and hammer
table (GWT) was constantly monitored at a depth directions (left and right) were kept consistent for all
of 0.6 m during the test period. The soil comprises tests.
of a 0.3 m thick layer of sand fill overlying
a medium to coarse medium dense sand to a depth
of 11.5 m. The sand is uniformly graded. A layer 3.3 Test program
of marine clay continues to the end of the BH/ The test program comprised three clusters of tests,
CPTUs. The CPTUs show uniform soil conditions equally distributed in a 3 m radius around the center
with some local variations in the cone resistances (Figure 3). Three SCPT tests were performed for
178
Figure 4. Test program for Cluster 2 illustrating the
approach for assessing repeatability and reproducibility.
Figure 2. Seismic test setup (main configuration).

The repeatability was investigated by comparing
results from multiple adjacent SCPT tests from the
same source (being either the offshore or onshore
beam source). The reproducibility was investigated
through variations in the results as a consequence of
using two different sources for generating the shear
waves (each with a different frequency content)
recorded with the same receivers at the same SCPT
location.
Two SCPTs in Cluster 3 aimed to investigate the
effect of source distance by changing the horizontal
distance between the offshore source and the SCPT.
Distances of 5 m (C3-SCPT1) and 10 m (C3-SCPT2)
were adopted. The remaining test (C3-SCPT3) aimed
to investigate the effect of cone size on the SCPT
results by using a 15cm2 cone. This test was unfortu
nately aborted after cone refusal on a stone.
Laboratory and other in situ (SDMT and MASW)
tests were performed for benchmarking the SCPT
results. The SDMT was performed using a different set
of receivers. A total of 15 bender element (BE) tests
(vertically propagating, horizontally polarized waves)
were performed in conjuction with isotropically con
solidated triaxial tests (CID). The shear wave velocity
(Vs) was measured at five different frequencies at the
target consolidation stress to estimate the optimal fre
Figure 3. Detailed field test program.
quency and the peak-to-peak and zero crossing
methods were adopted to interpret the traces. In add
ition, two resonant column (RC) tests were performed.
each cluster and the signals were recorded in
sequence from both seismic sources. One SDMT test
4 RESULTS
was performed in Cluster 3 and four MASW lines
(M1 to M4) were run in a triangular setup within the
same area. The target penetration depth for SCPT 4.1 Quality assessment
tests was 15 m. The SCPTs at Clusters 1 and 2 were The interval Vs has been calculated from left and
scheduled for assessing repeatability and reproduci right shots using traces that met certain quality
bility of the measured Vs. With reference to the tests requirements only. Figure 5 shows an example of
performed in cluster 2, Figure 4 shows the a representable signal of acceptable quality. The Seis
approaches used for clusters 1 and 2. mic Trace Characterization (STC), as proposed by
179
Baziw & Verbeek (2017), has been performed on all
traces to evaluate the quality of the signals. The data
quality varies from acceptable (Class C) to very
good (Class A). The quality of traces at shallow
depth (less than 2 m) has been problematic to
unacceptable (Class D, F) for both sources. However,
in general a higher signal quality is observed from
the beam source compared to the offshore source.
Figure 6. Induced error on Vs as function of sampling fre

quency, distance between receivers and soil Vs (maximum
measured Vs is 300 meter per second, m/s).
Bandpass filters have not been applied on the ac

quired data. However, the application of filters close
to the dominant frequency (60 Hz to 80 Hz for the
offshore source) can cause a phase shift and induce
an additional error to Vs. The time shift caused by
the application of a low pass filter can be estimated
based on Rice (1984). For the project presented in
this paper, an application of a low pass filter with
Figure 5. Seismic signals for C2-SCPT3 from both sources. a cut-off frequency of 60 Hz is expected to induce
an error larger than 10% for a Vs equal to 200 m/s.
The depth accuracy of the CPT rig is estimated to
be better than 1%. In general, the uncertainty in the
4.2 Uncertainties depth penetration for push systems is determined
The key input to the determination of Vs includes with high accuracy.
travel time and travel distance. They are susceptible The uncertainties on Vs related to the interpret
to uncertainty originating from the specific system ation method depend primarily on the data quality
characteristics, including sampling frequency, trigger and on the assumed travel path. As described by
repeatability, data filtering, depth uncertainty, etc. Lunne et.al (1997), correction of the travel path due
For a given sampling frequency, the error induced to refraction can have a considerable impact on Vs.
on Vs can theoretically be expressed as a function of Vs According to Laing (1985), the straight-line approxi
and the distance between receivers as presented and mation is satisfactory for calculation of Vs when the
demonstrated by Rice (1984). Adopting Rice’s angle of travel path is higher than 45°. Hence, for
approach to the SCPT data in this study, Figure 6a depth shallower than the horizontal source distance,
shows that for a sampling frequency of 5 kHz and refraction of shear waves along the soil boundaries
a distance between receivers of 0.5 m, the error on Vs is likely to occur.
varies from 4% to 12 %. Figure 6b shows that in order
to reduce the Vs error to 10% (accuracy requirements
according ISO 2014), the sampling frequency must
be increased to 6 kHz. Alternatively, the accuracy
can be increased by increasing the distance between
receivers for the same sampling frequency.
The error induced on Vs due to trigger repeatabil
ity has been assessed by comparing the arrival times
between stacked and multiple unstacked signals at
the same depth. The offset between relative arrival
time should be less than 1%, as recommended by
ASTM (2019). For the range of measured Vs at
the site, the error due to trigger repeatability is
assessed to a maximum of ± 2%. This error is
reduced to zero when Vs is calculated from traces Figure 7. Illustration of seismic traces at Cluster 3, for
generated by the same shot in a dual array setup. three considered offshore source distances.
180
Figure 8. Comparison of the SCPT measured Vs at Cluster 2 between Beam (onshore source) and Offshore source.
Figure 9. Comparison of the measured Vs at Cluster 2 between SCPT, seismic dilatometer (SDMT), multichannel analyses
of surface waves (MASW), bender elements (BE), resonant column (RC), correlation from CPT data using Robertson
(2009) correlation.
The impact of the horizontal distance between the in Vs of up to ±10 %. Locally, the variation of certain
seismic source and the SCPT was investigated by Vs measurements exceeds ±20 % (8–10 m depth).
placing the offshore source at different distances of This difference cannot immediately be explained by
2.6 m, 5.0 m and 10.0 m to the SCPT. Figure 7 the data. However, the larger variation in Vs is
shows the signal quality (signal-to-noise-ratio) reflected by correspondingly larger variations in
decreasing with increasing distance. The results are cone resistances (qc), indicating that differences in
in line with the findings by Rice (1984). Vs from the two sources are likely also reflecting
local changes in geology. No obvious differences
were observed in the variation in Vs between the
4.3 Repeatability and reproducibility
sand and the clay. The repeatability of Vs for both
Figures 8a-f present the comparison of the derived clusters is similar and is summarised in Table 2.
Vs from the different sources for Cluster 2. The reproducibility of Vs is found reasonable for
Figures 8a and b indicates a good repeatibility for Cluster 2, with a variation from ±6 % to ±11 %
both sources, with an average coefficient of variation (Figures 8c-e). The differences observed for the two
181
sources cannot be attributed to the geology as the position-specific wind turbine foundation designs, Gmax
tests were performed at the same position. However, is traditionally determined from SCPTs. However, the
the individual travel paths of the shear waves may be performance of SCPTs are affected by a number of fac
subject to local soil variations. Another factor con tors associated with the seismic source and with the
tributing to the reproducibility is the trigger timing. operational setup in general. Therefore, the repeatibility
It appeared challenging in the field to run the trigger and reproducibility of offshore SCPTs have been
signal from both sources simultaneously, using the observed to relatively low compared to an onshore
same software. Table 3 presents similar reasonable setup. The main objective of this study was to obtain
reproducibility of Vs for both clusters. In general, an understanding of the impact of the offshore seismic
both repeatibility and reproducibility of the sources source setup on SCPT results and to understand the
show an average variation of about ±10 %. challenges with the existing setups and hence shed
Figures 9a-f compares SCPT results for Cluster light on the scatter observed offshore. Therefore, two
2 with SDMT, MASW, BE and RC test results as different sources were used in a sand pit in Denmark.
well as Vs based on Robertson (2009). The SCPT results using an offshore source were bench
marked against the SCPT results using an onshore
source. The results show that under reasonably con
Table 2. Source repeatability. trolled operational conditions and when potential
sources of errors are taken into account, the SCPTs can
Beam Offshore source
be satisfactorily repeated and reproduced with
Cluster % % a variation of ±10%, using both onshore and offshore
sources. Careful preparation and execution of the
Cluster 1 ±9 ±8 SCPTs are essential. Especially it is deemed necessary
(SCPT1 vs SCPT2) to a) obtain a good contact between the ground and the
Cluster 2 ±10 ±7 source, b) orientate the source towards the receivers to
(SCPT1 vs SCPT2 vs SCPT3) maximise the signal amplitude in one direction, and c)
Average all tests ±10 ±8 position the source horizontally. Furthermore,
key system characteristics (sampling frequency,
trigger repeatibility) play an important role in the
accuracy of the measured shear wave velocity.
Table 3. Source reproducibility.
Cluster 1 SCPT Cluster 2 SCPT Average

ACKNOWLEDGMENT
1 2 1 2 3 All
tests The research presented herein was performed as
a Joint Industry Project (JIP) carried out between
Sources % % % % % % Ørsted and Geo. The care and effort exerted by both
parties during testing are highly appreciated.
Beam vs ±11 ±9 ±11 ±6 ±6 ±10
Off.
source
REFERENCES
ASTM D7400 2019. Standard test methods for downhole
Figure 9b illustrates the results of using a set of seismic testing.
different receivers on the SDMT and confirms the Baziw, E.J. & Verbeek, G. 2017. Quality assessment of
seismic data sets and the impact on interval velocity
assumption that Vs is independent of the recording
estimates. In Proceedings of DFI 42nd Annual Confer
equipment (receivers), which is in accordance with ence on Deep Foundations. Article 2821.
similar findings from McGillivray & Mayne (2004). Campanella, R.G. & Robertson, P.K., 1984. A seismic cone
MASW data is uniform and overlaps the SCPTs penetrometer to measure engineering properties of soil.
in the depth range of 1.5 m to 4.5 m. Vs measured In SEG Technical Program Expanded Abstracts 1984:
from BE test is comparable with SCPT measure 138–141.
ments for the sand, whereas the RC tests return Campanella, R.G., Robertson, P.K. & Gillespie, D. 1986.
higher values. The range of Vs from all measurement A seismic cone penetrometer for offshore applications.
are similar and representative of the soil conditions. In Oceanology: 479–486.
Gibbs, P., Pedersen, R.B., Krogh, L., Christopher, N.,
Sampurno, B. & Nielsen, S.W. 2018. Challenges in
marine seismic cone penetration testing. In Cone Pene
5 CONCLUSIONS tration Testing 2018: 303–308.
Gillespie, D., Lunne, T. & Campanella, R.G. 1985. Tests
The small strain stiffness (Gmax) has become a key par with UBC seismic cone at three Norwegian research
ameter for offshore wind facility designs, especially for sites. Report No. 59040, 1. NGI.
the design of monopile foundations. Hence, accurate Gillespie, D.G., 1990. Evaluating shear wave velocity and
and precise measurements of Gmax is important. For pore pressure data from the seismic cone penetration
182
test (Doctoral dissertation, University of British seismic CPT and PS logging. In Offshore Technology
Columbia). Conference. Article 29485.
ISO 19901-8 2014. Petroleum and natural gas industries McGillivray, A. & Mayne, P.W. 2004. Seismic piezocone
Specific requirements for offshore structures - Part 8: and seismic flat dilatometer tests at Treporti. In Proc.
Marine soil investigations. 2nd Int. Conf. on Site Characterization 2: 1695–1700.
Laing, N.L. 1985. Sources and receivers with the seismic Rice, A.H. 1984. The seismic cone penetrometer (Doctoral
cone test, MASC (Doctoral dissertation, Thesis, Dept. of dissertation, University of British Columbia).
Civil Eng., Univ. of British Columbia). Robertson, P.K. 2009. Interpretation of cone penetration
Lunne T., Robertson P. & Powell J. 1997. tests—a unified approach. Canadian geotechnical jour
CPT in geotechnical practice. New York: Blackie nal, 46(11): 1337–1355.
Academic. Robertson, P.K., Campanella, R.G., Gillespie, D. &
Masters, T.A., Juszkiewicz, P., Mandolini, A. & Rice, A. 1986. Seismic CPT to measure in situ shear
Christian, H. 2019. A critical appraisal of the benefits of wave velocity. Journal of Geotechnical Engineering,
and obstacles to gaining quality data with offshore 112(8): 791–803.
183
© 2022 the Author(s), ISBN 978-1-032-31259-0
The revival of multiple pore pressure measurements in the cone

penetration test
T. Lunne
Norwegian Geotechnical Institute, Oslo, Norway
R.K. Ghanekar
Consultant, Panvel, India
G.W. Tucker, R. Santos & L. Krogh

Ørsted, Denmark
ABSTRACT: In common practice the most usual positioning of the filter location for the piezocone (CPTU) is
at the cone shoulder (u2) which is recommended by ISO 19901-8:2014 and ISO 22476-1:2012. However, these
ISO standards allow additionally for pore pressure measurements to be taken at the cone face (u1) or behind the
friction sleeve (u3). The triple element piezocone (CPTU3) offers a solution whereby measurements of pore pres
sure can be taken simultaneously at all three locations. By taking these three pore pressure measurements the soil
behaviour classification may be enhanced by interpretation in terms of ch from u1, u2 and u3 and by correlations
to overconsolidation ratio (OCR) and lateral stress ratio (K0). In this paper the correlations to OCR are explored
by establishing a database from historical data as presented in international geotechnical literature.
1 INTRODUCTION • Enhanced correlations to stress history and in situ

horizontal stress
The pore pressure measurement taken using the
piezocone (CPTU) can be recorded with depth or This paper explores the last aspect of correlations
as dissipation over time when penetration is to stress history by developing a database from avail
paused. The most common pore pressure measure able CPTU3 tests published mainly in the 1980s and
ment is behind the cone shoulder (u2), as preferred 1990s. First a review of the earlier reported use of
pore pressures in connection with CPT is made.
by ISO 19901-8:2014 and ISO 22476-1:2012.
However, these ISO standards allow additionally
for penetration pore pressure measurements to be
2 HISTORY OF PORE PRESSURE
taken on the cone face (u1) or behind the cone
MEASUREMENT IN THE CPT/CPTU
sleeve (u3) as shown in Figure 1. The so-called
triple element piezocone (CPTU3) was used by
Pore pressure measurement during penetration of
several researchers in the 1980s and 1990s but has
a probe was introduced in the 1970s (Janbu & Senne
seen very little use in the last 25 years. The avail
set, 1974, Torstensson, 1975, Wissa et al., 1975). The
able literature indicates that CPTU3 has consider
pore pressure probe used by Janbu & Senneset (1974)
able potential relative to the standard CPTU
resembled an electrical cone, with the same area at the
including:
base of the cone. The other probes had different geom
• Correction of sleeve friction (fs) for pore pressure etries and dimensions, but all had cylindrical porous
effects elements. From the beginning, the large potential of
• Increased reliability of assessment of horizontal such probes was recognised. At that time, CPTs and
coefficient of consolidation (ch) from three sets of pore pressure probe tests were performed in adjacent
dissipation tests deployments, for instance Janbu & Senneset (1974)
• Evaluation of drainage conditions around the used a CPT and in parallel a standard NGI piezometer.
cone – especially for intermediate soils where Although Parez et al. (1976) had already devel
partial drainage occurs oped and used equipment able to measure cone
• Improved layering detection using u1 compared resistance (qc) and pore pressure simultaneously, it
to u2 was only in the early 1980s that many researchers
DOI: 10.1201/9781003308829-20
184
around the world started to use a pore pressure meas suggested that they will vary with soil type as shown
urement element incorporated in the electrical cone in Figure 3, this was later confirmed by Powell and
(e.g. Roy et al.,1980, Campanella & Robertson, Lunne (2005).
1981). Battaglio et al., (1986) presented results from
CPTU3 tests carried out in a medium to stiff clay at
the Pontida site (Italy). Fugro and McClelland both 3 LEARNINGS FROM DATABASE
developed CPTU3 probes in the mid1980s, see
Bayne & Tjelta (1987) and Zuidberg et al. (1988). Based on the potential use of the CPTU3 seen from the
Based on CPTU data from several clays with literature review, it was decided to collect available
a range of overconsolidation ratio (OCR) values, data from previously published field test results to
Robertson et al. (1986) presented a conceptual pore
pressure distribution along the length of a penetrating
Figure 1. Pore pressure measurement positions.
cone penetrometer for clays, ranging from normally

consolidated to heavily overconsolidated clays
(Figure 2). The pore pressure distribution is normal
ised by in-situ hydrostatic pore pressure (u0).
Figure 3. PPD vs OCR for OCR > 10 (Sully et al., 1988).
create a database of pore pressure measurements, index

properties and stress history values. The objective of
the database was to evaluate reported correlations
between the measured pore pressures and reported cor
relations between geotechnical engineering parameters
and parameters derived from pore pressure measure
ments, referred to hereafter as pore pressure
Figure 2. Pore pressure distribution in saturated clays
around a penetrating cone based on field measurements parameters.
(Robertson et al., 1986). The data used to create the database represent
a range of clays from Europe and the USA, taken
mainly from Chen & Mayne (1994), Larsson &
Figure 2 indicates that the pore pressure measured Mulabdic (1991), Sandven (1991), Powell & Lunne
at the tip or face of the cone (u1t or u1f, designated uuc (2005), and from unpublished data from offshore soil
and u1 respectively on inset of Figure 2) is higher investigations in the North Sea where NGI was the
than that measured at u2, which again is higher than consultant and some recent NGI projects. When pos
that measured at u3. Figure 2 also illustrates the poten sible, datasets were checked by studying the original
tial to use relative values of u1t or u1f, u2 and u3 to source. Uncertain or poor-quality data were removed
estimate OCR and possibly lateral stress ratio (K0). from the originally prepared database where
Early studies based on limited data, reported by appropriate.
Sully et al. (1988), Sully & Campanella (1991) and Altogether, the final database contained 546 data
others showed the potential of correlations among Pore sets from 63 sites (both offshore and onshore). OCR
Pressure Difference (PPD) = (u1-u2)/u0 vs OCR and values, based on oedometer tests, are available for
185
all 63 sites, whereas K0 values are available from 5
UK and 2 Norwegian clays, mainly based on results
of self-boring pressuremeter tests.
Table 1. Details of relevant datasets in final database.
Parameter u1t u1f u1f* u2 u3
Data points 56 381 36 512 413

Parameter OCR K0
Data points 313 112
* Calculated from u1t (as explained later in the paper)
In addition to PPD (defined above) other derived

pore pressure parameters utilising u1 and u2, are
found in literature:
Figure 5. PPDa vs OCR for global data base.
1. Pore pressure ratio (PPR) = u1/u2 (Sully et al.,
1988)
2. Excess pore pressure ratio (PPR1) = (u1-u0)/(u2
-u0) (Sully et al., 1988)
3. Normalized Pore Pressure Difference (PPSV) =
(u1-u2)/σ’v0
Since the present database contained a reasonably
good number of datasets for u3 also, similar param
eters involving u1 and u3, and u2 and u3 were also
derived. Parameters derived using u1 and u2 were
designated with suffix ‘a’, using u1 and u3 with
suffix ‘b’ and using u2 and u3 with suffix ‘c’.
Excellent correlation was found between u1t
and u1f (Figure 4). Using this relationship u1f
was calculated for 36 datasets where only u1t
values were measured. For subsequent analyses
Figure 6. PPSVb vs OCR for global data base.
only u1f was used and for simplicity called u1.

However, ratios involving u1 and u2, u1 and u3 or
u2 and u3 did not show good correlations when
all data were plotted. With reference to Figure 2
this is as expected.
The database contains OCR ranging from 1 to
80, evaluated mainly from oedometer test results.
OCR data plotted against PPRa, PPRb and PPRc,
and against PPR1a, PPR1b and PPR1c did not
show any clear trends. PPDa, PPDb, PPDc,
PPSVa, PPSVb and PPSVc showed only a weak
trend of increasing with increasing OCR but no
Figure 4. Relationship between u1t and u1f for global data
base. clear relationship could be established due to
186
Figure 7. Relationship between Qt vs OCR for global data Figure 8. Correlation of PPDb vs OCR for Cowden glacial
base. till.
very large scatter, as shown by the examples in

Figure 5 and 6. Figure 7 shows that the most
frequently used correlation between OCR and Qt
(= (qc – σv0)/ σv0ʹ), also shows a very large scat
ter where it was not meaningful to fit
a trendline.
The findings from the database indicate that it
may not be possible to establish a strong correlation
between pore pressure parameters and OCR that are
valid for all cohesive soils. This status is similar to
the correlation factor Nkt, the ratio of net cone
resistance to undrained shear strength, where after
decades of attempts by researchers and industry
reported Nkt values are still wide ranging. With this
example in mind, it will therefore be necessary to
establish local correlations between pore pressure
parameters and OCR as is commonly the practice
for Nkt.
4 ATTEMPTS AT LOCAL CORRELATIONS

Figure 9. Correlation of PPSVa vs OCR for Cowden gla
4.1 General cial till.
Two data sets that are believed to be of good and

reliable quality have been selected for an illustration
of local correlations. pressures measured suffered from unsatisfactory
saturation and historic measurements given by
– The glacial till at Cowden, UK, thoroughly Powell and Lunne (2005) were used, since these
tested in the 1990s as documented by Powell were considered to be of good quality. Correlations
and Butcher (2003). to PPD parameters showed equally good results
– Two Norwegian moderately overconsolidated when based on u1 – u2 (PPDa) and u1 – u3 (PPDb).
stiff marine clays tested by Sandven (1991). An example of the latter is shown in Figure 8. The
same conclusion can be drawn regarding OCR to
4.2 Glacial till at Cowden PPSV where, as an example, PPSVa is shown in
Part of the programme was to carry out new Figure 9. All four PPD and PPVS correlations had
CPTU3 tests at Cowden. However, the pore R2 values in the range 0.83 -to 0.87.
187
Figure 10. Correlation of Qt vs OCR for Cowden glacial Figure 12. Correlation of PPDb to OCR for Bakklandet
till. and Glava clays.
Figure 11. Correlation of PPDa vs OCR for Bakklandet

and Glava clays.
Figure 13. PPSVa vs OCR for Bakklandet and Glava clays.
Figure 10 shows that the most frequently as moderately overconsolidated stiff marine
used correlation between OCR and Qt is clays. Figure 11 show that for PPDa
actually showing less scatter, with an R2 value of a reasonably good correlation can be found with
0.92, compared to the PPD and PPSV the Bakklandet data generally plot somewhat
correlations. lower compared to Glava data. However, when
plotting OCR vs PPDb as shown in Figure 12
the Bakklandet data plot significantly lower
4.3 Trondheim area clays compared to Glava data, but both the sites show
Sandven reported results of CPTU3 tests from local and reasonably good linear trends.
two sites in the Trondheim area of Norway: In Figure 13 PPSVa show a reasonably good cor
Glava and Bakklandet. Both sites are described relation to OCR, whereas Figure 14 shows that for
188
CPTU3 OCR correlations compared to correlations
with a standard CPTU.
5 DISCUSSION ON CORRELATIONS
The correlation attempts described in this paper are

relatively simple.
Since the data base includes soil classification
parameters like water content liquid and plastic
limits, future work could explore multiparameter
regression analyses to see if this can result in any
better correlations. Since K0 estimates are also avail
able at 7 (out of 63) sites, correlations to this param
eter could also be explored.
6 SUMMARY AND CONCLUSIONS
Based on geotechnical literature and unpublished

Figure 14. PPSVb vs OCR for Bakklandet and Glava material a data base has been established with
clays. CPTU3 parameters (qc, fs, u2, u1 and u3) as well as
laboratory data giving index parameters and stress
history values in terms of overconsolidation param
eters. After removing some uncertain data in quality
assessment process, the data contain data sets from
63 sites representing a range of clays from Europe
and the USA.
The simple correlation study reported in this
paper indicates that using all the data available it is
not really possible to show any reliable correlation
between pore pressure parameters and OCR.
When choosing two sets of local data from the
data base (Cowden glacial till and Trondeim area
moderately over consolidated marine clays) reason
ably good correlations have been found with pore
pressure parameters PPD and PPSV. However, for
the two sub-data sets even better correlations are
obtained when using correlations to Qt which can be
obtained from standards CPTU tests. This indicates
that regarding correlations to OCR the inclusion of
u1 and/or u3 in the CPTU may not improve the
accuracy/reliability of the OCR values over the
values obtained from standard CPTU.
Other advantages of the CPTU3 relative to the
standard CPTU, including results of dissipation tests,
Figure 15. Qt vs OCR for Bakklandet and Glava clays.
are not evaluated in this paper.
PPSVb Bakklandet data plot somewhat lower than ACKNOWLEDGEMENTS

the Glava data.
The authors are grateful to the support of Ørsted to
Figures 12 and 14 really show that the difference
give permission to publish this work. The authors
between u2 and u3 is much lower for the tests at Bakk
would like to thank NGI for the development of the
landet compared to Glava. Sandven (1991) does not
historical database.
make any comments about this difference.
Figure 15 shows that the correlation that is mostly
used for standard CPTUs based on Qt (e.g. by Lunne REFERENCES
et al., 1997) actually gives a better correlation than
PPD or PPSV illustrating with this example that Battaglio, M., Bruzzi, D., Jamiolkowski, M. &
there may not be any additional benefit using Lancellotta, R. 1986. Interpretation of CPTs and CPTUs
189
- 1st part: undrained penetration of saturated clays. Powell, J.J.M. & Lunne, T. 2005. Use of CPTU data in
Field Instrumentation and In-Situ Measurements: Pro clays/fine grained soils. Studia Geotechnica. XXVII
ceedings of the 4th International Geotechnical Seminar, (3 – 4): 29–65.
Singapore: 129–143. Robertson, P.K., Campanella, R.G., Gillespie, D. &
Bayne, J.M. & Tjelta, T.I. 1987. Advanced cone Greig, J. 1986. Use of piezometer cone data. Proceed
penetrometer development for in-situ testing at ings of In-Situ 86 Specialty Conference, American Soci
Gullfaks C. Proceedings of the 19th Offshore Tech ety of Civil Engineers, Blacksburg: 1263–1280.
nology Conference, Houston, OTC paper 5420(1): Roy, M., Tremblay, M., Tavenas, F. & La Rochelle, P. 1980.
531–540. Induced pore pressures in static penetration tests in sensi
Chen, S.C & Mayne, P.W. 1994. Profiling overconsolida tive clay. Proceedings of the 44th Canadian Geotechnical
tion ratio by piezocone tests. National Science Founda Conference, Calgary, Preprint Volume, 11.3.1 to 11. 3.13.
tion, Report no. GIT-CEEGEO-94-1. Sandven, R. 1990. Strength and deformation properties of
Campanella, R.G. & Robertson, P.K. 1981. Applied cone fine-grained soils obtained from piezocone tests. PhD
research. Procs Symp on Cone Penetration Testing and thesis, Norwegian Institute of Technology, Trondheim,
Experience, ASCE, St. Louis: 343–362. Norway.
Janbu, N. & Senneset, K. 1974. Effective stress interpret Sully, J.P., Campanella, R.G. & Robertson, P.K. 1988.
ation of in situ static penetration test. Procs of the Euro Overconsolidation ratio of clays from penetration pore
pean Symposium on Penetration Testing, Stockholm, pressures. Journal of Geotechnical Engineering, Ameri
2:2: 181–193. can Society of Civil Engineers, 114(2): 209–216.
Larsson, R. & Mulabdic, M. 1991. Piezocone tests in clay. Sully, J.P. & Campanella, R.G. 1991. Effect of lateral stress
Swedish Geotechnical Institute. Report no. 42. on CPT penetration pore pressures. Journal of Geotech
Lunne, T., Robertson, P.K. & Powell, J. (1997) CPT nical Engineering, American Society of Civil Engineers,
and piezo-cone testing in geotechnical practice. Text 117(7): 1082–1088.
book: Spon Press, Taylor and Francis Company, Torstensson, B.A. 1975. Pore pressure sounding
London. instrument. Procs Conference on In Situ Measurement
Parez, L., Bachelier, M. & Sechet, B. 1976. Pression inter of Soil Properties, American Society of Civil Engineers,
stitieIle developpee au foncage des penetrometres. Pro Raleigh, 2: 48–54.
ceedings of the 6th European Conference on Soil Wissa, A.Z.E., Martin, R.T. & Garlanger, J.E. 1975. The
Mechanics and Foundation Engineering, Vienna, 1.2: piezometer probe. Procs Conf on In Situ Measurement
533–538. of Soil Properties, ASCE, Raleigh, 1: 536–545.
Powell, J.J.M. & Butcher, A.P. 2003. Characterisation of Zuidberg, H.M., ten Hoope, J. & Geise, J.M. 1988.
a glacial till at Cowden, Humberside. Workshop, Char Advances in in-situ measurements. Procs 2nd Inter
acterisation and Engineering Properties of Natural national Symposium on Field Measurements in Geome
Soils: 983–1020. Singapore, 2002. chanics, Kobe, 1: 279–291.
190
© 2022 the Author(s), ISBN 978-1-032-31259-0
Comparisons CPT-DMT in soft clay at Fucino-Telespazio GeoTest site
D. Marchetti
Studio Prof. Marchetti, Rome, Italy
P. Monaco & G. Totani

F. Totani
Consultant Engineer, L’Aquila, Italy
S. Amoroso
University of Chieti-Pescara, Pescara, Italy
Istituto Nazionale di Geofisica e Vulcanologia, L’Aquila, Italy
ABSTRACT: This paper presents the comparison of results obtained from CPT/CPTU tests carried out in
past investigations and from recent DMT tests carried out using the automated Medusa DMT, adopting differ
ent test procedures (standard, repeated A-readings, A-reading while penetrating), at the benchmark soft clay
test site of Fucino-Telespazio, Italy. In particular, the depth profiles of the undrained shear strength su of the
Fucino clay obtained from CPT/CPTU interpretation are compared with the su profiles obtained from standard
DMT and from Medusa DMT using different test procedures, as well as with the su values obtained from field
vane test (FVT), self boring pressuremeter test (SBPT) and laboratory tests available from past investigations,
resulting generally in good agreement.
1 INTRODUCTION in 2020 the same site was selected for experimenta

tion with the automated Medusa DMT, ideally link
Benchmarking is of significant importance in geo ing past experience and recent technological
technical engineering for testing and validating both developments.
innovative soil investigation methods and foundation This paper presents the comparison of results
solutions. The growing interest in this topic is docu obtained at Fucino-Telespazio from cone/piezo
mented by the 1st International GeoTest Site Sympo cone penetration tests (CPT/CPTU) carried out in
sium ISGTS, Oslo, 2019, aiming to promote an past investigations and from recent DMT tests car
increasing use of benchmark test sites, selected to ried out using the Medusa DMT equipment (briefly
ensure simplicity of geotechnical conditions (i.e., described in the next section) by adopting different
relatively homogenous soil deposit) and ease of test procedures. In particular, the focus in this
interpretation (i.e., pre-existing laboratory and field paper is on the evaluation of the undrained shear
data). strength su of the Fucino clay by different test
Fucino-Telespazio is a well known benchmark methods.
‘GeoTest site’ in Italy, constituted by a thick
deposit of cemented, homogeneous soft lacustrine
clay of high plasticity. The site was extensively 2 MEDUSA DILATOMETER TEST
investigated at the end of the 1980s by means of
several in situ and laboratory tests carried out by The Medusa DMT (Figure 1) is the last-generation,
various international research groups (Burghignoli fully-automated version of the DMT. It is a self-
et al. 1991). Earlier experimentation with the flat contained probe able to autonomously perform
dilatometer (DMT) at this site was carried out in dilatometer tests using a blade of standard dimen
the 1970s, as reported by Marchetti (1980). Subse sions without the pneumatic cable, the control unit
quently, in 2004-2005 the site was selected for val and the gas tank required in the traditional pneu
idation of the seismic dilatometer (SDMT, Foti matic DMT configuration. A motorized syringe,
et al. 2006, Marchetti et al. 2008). More recently, driven by an electronic board powered with
DOI: 10.1201/9781003308829-21
191
rechargeable batteries, hydraulically expands the 3 THE FUCINO-TELESPAZIO BENCHMARK
membrane to obtain the DMT A, B, C pressure TEST SITE
readings, which are acquired and stored automatic
ally at each test depth (typically every 0.20 m). The Fucino-Telespazio test site is located in the
The automatic (volume controlled) hydraulic pres Fucino basin, central Italy, about 80 km east of
surization of the membrane is highly repeatable Rome. In 1986 the site was selected as a national
and permits to impose a programmable timing (i.e., benchmark test site and investigated by means of
the recommended standard timing, or different a joint effort of a number of Italian and inter
timing corresponding to variable pressurization national research groups. The primary aim of the
rates) to obtain the pressure readings. The probe investigation was to carry out comparative in situ
may operate in cableless mode, which is and laboratory tests in a suitable soft clay
a significant practical advantage in the offshore deposit. The choice fell on the Fucino-Telespazio
industry and for deep investigations. An optional site due to its marked spatial homogeneity and
electric cable may be used to obtain real-time data apparently simple geological history. The site is
during test execution. constituted by a thick deposit of soft, homoge
neous, highly structured CaCO3 cemented clay of
lacustrine origin (Fucino clay). The experimental
activity, which lasted more than two years (1986
88), was documented in several research reports
and papers. In particular, the data shown in this
paper are taken from Burghignoli et al. (1991),
who presented a comprehensive description of
the in situ and laboratory testing program and
a detailed geotechnical characterization of the
Fucino clay.
The in situ testing program included boreholes
with undisturbed sampling, cone/piezocone pene
tration tests (CPT/CPTU), self boring pressure-
meter tests (SBPT), flat dilatometer tests (DMT),
field vane tests (FVT), seismic cone tests (SCPT),
cross-hole tests (CH), down-hole tests (DH),
surface wave tests (SASW) and piezometer
measurements.
Laboratory tests were carried out on a large
number of undisturbed samples, including deter
mination of index properties, incremental loading
(IL) and constant rate of strain (CRS) oedometer
tests, unconsolidated undrained (UU), isotropically
consolidated drained (CID) and undrained (CIU)
Figure 1. Main components of the Medusa DMT.
triaxial compression tests, direct simple shear tests
(DSS-CK0U), laboratory vane tests (VT), resonant
column tests (RC) and cyclic torsional shear
The Medusa DMT has several advantages over tests (CTS).
the traditional DMT equipment in terms of test auto To minimize the influence of spatial variability
mation, field productivity and increased accuracy of the clay properties, the field investigations were
(Marchetti 2018, Marchetti et al. 2019). It also pro concentrated in an area of 10 × 40 m2, to a depth of
vides the possibility of performing additional meas about 40 m. The superimposed profiles of the cone
urements, not feasible with the traditional DMT, resistance qc measured from four CPTs performed
including continuous measurement of the total hori at the corners of the area (Figure 2) show a fair
zontal pressure against the membrane with time at homogeneity of the site, both in vertical and hori
a stationary test depth, to obtain information on soil zontal directions: the soil stratigraphy is virtually
response in terms of fully drained/partially drained/ identical at all test locations, with thin sandy layers
undrained behavior, or during probe penetration, to identified at the same depths by all tests. Similar
obtain information on the in-situ stress state. Due to results were found in three CPTUs carried out in
the increased accuracy of pressure measurements the same area.
and controlled pressurization rate, the Medusa DMT The Fucino clay is characterized by high plas
is particularly useful for testing soils which are usu ticity (plasticity index PI mostly between 40 and
ally difficult to characterize using common in-situ 70%, natural water content w between 60 and
techniques, such as very soft or even nearly liquid 120%). Despite its relatively recent deposition,
soils (Marchetti et al. 2021), mine tailings and inter the clay is highly structured and cemented. The
mediate soils (Monaco et al. 2021). calcium carbonate
192
test depth is reached, the penetration is stopped
and the DMT test cycle starts. The activated
motorized syringe gradually increases the hydraulic
pressure to the membrane. When the internal oil
pressure equals the external soil pressure, the mem
brane lifts-off from its seat and starts to expand lat
erally. When the membrane has expanded of
0.05 mm at its centre, the A-pressure is recorded.
After the A-reading, the motorized syringe con
tinues to increase the pressure until the membrane
displacement at the centre equals 1.10 mm. At this
instant the second pressure reading B is recorded.
As soon as the B-reading is obtained, the motor
ized syringe starts decreasing the oil pressure. If
the C-pressure reading is requested, the motorized
syringe applies a gradual and controlled depressur
ization after the B-reading and the membrane
slowly returns to its initial position against the
sensing disc. At the instant in which the contact
reactivates, the corresponding pressure is recorded
as the C-pressure reading. In the STD procedure
the pressurization rate is regulated by the motor
ized syringe so that the A-pressure reading is
obtained 15 s after start of pressurization and the
B-pressure reading 15 s after the A-pressure read
Figure 2. CPT results from past investigations at ing, in accordance with existing standards of the
Fucino-Telespazio (Burghignoli et al. 1991). pneumatic DMT (ASTM D6635-15, ISO 22476
11:2017(E)). The C-reading is typically obtained
30 s after start of depressurization following the
CaCO3 content was found to range between 10 B-reading.
and 30% in the upper 25 m, increasing to an average
value of about 60% below this depth. The clay
deposit is geologically normally consolidated. How 4.2 DMT repeated A-readings procedure
ever both laboratory and in situ tests indicated a light (DMT-RA)
overconsolidation, which was attributed mostly to The DMT-RA procedure differs from the STD pro
diagenetic interparticle bonding due to CaCO3 cedure only in the first part of the measurement
cementation. Burghignoli et al. (1991) pointed out sequence, before membrane expansion, while the B
that structure and cementation have a dominant influ and C-pressure readings are taken exactly in the
ence on stress history, compressibility, consolidation same way. As previously described, in the STD pro
and shear strength properties of the Fucino clay. cedure the A-reading is taken when the membrane
centre has expanded horizontally 0.05 mm against
the soil, replicating exactly the same procedure
4 MEDUSA DMT TESTS AT THE implemented in the traditional pneumatic DMT test.
FUCINO-TELESPAZIO BENCHMARK TEST The motorized syringe of the Medusa DMT, driven
SITE by the electronic board, is also able to maintain the
membrane in equilibrium with negligible horizontal
An experimental program at the Fucino-Telespazio displacement of the membrane. The DMT-RA pro
test site, including three Medusa DMT soundings cedure makes use of this capability. The initial stage
and one traditional pneumatic DMT sounding carried of the DMT test cycle with the DMT-RA procedure,
out using the SDMT equipment, was completed in before penetrating to the next test depth, consists in
September 2020. All soundings were performed at maintaining the membrane in equilibrium with the
close mutual distance, to a depth of 30 m. soil pressure. This state is obtained with very rapid
The Medusa DMT soundings were carried out pressure corrections operated by the motorized syr
using three different test procedures (see Monaco inge, with negligible membrane displacement. In this
et al. 2022 for details), which differ essentially for situation the membrane is in equilibrium at 0.05 mm
the technique adopted for measuring the A-pressure. distance from the sensing disc. When the new test
depth is reached, the test cycle starts (t = 0) and
4.1 Standard DMT procedure (STD) repeated sequential A-readings are taken with time
during the rapid pressure corrections of the motor
The STD procedure is the same procedure of the ized syringe, monitoring the total horizontal soil
traditional pneumatic DMT test. As soon as the pressure against the membrane with time. All the
193
sequential A-readings are obtained without any dis results obtained from Medusa DMT were processed
placement of the soil, because the blade is advanced using the same data reduction and interpretation for
to the test depth with the membrane already at mulae used for the traditional DMT test (ISSMGE
0.05 mm displacement from the sensing disc. Such TC16 Report, Marchetti et al. 2001). In particular,
DMT-RA procedure is characterized by the duration Figure 3 shows the depth profiles of the corrected
(Tdiss) of the sequential A-readings taken with time pressure readings p0, p1, p2 (A, B, C corrected with
(dissipation), before concluding the DMT test cycle the calibration offsets ΔA, ΔB to account for mem
with the standard B and C readings. The parameter brane stiffness), as well as of the derived parameters
Tdiss is selected and pre-programmed before starting material index ID and horizontal stress index KD. For
the test cycle, so that the membrane expansion will the DMT-RA sounding, the A-pressure reading used
be activated after the time Tdiss has elapsed. At in data processing is the last value obtained from the
Fucino-Telespazio Tdiss was set equal to 15 s, to A-dissipation series, i.e. the A-pressure recorded 15
comply with the standard timing adopted in the STD s after start of pressurization. In the processing of
procedure. The procedure and timing for taking the data from all soundings the groundwater table was
B and C readings are the same as in the STD assumed at a depth of 0.60 m below the ground sur
procedure. face, as indicated by the p2 values observed in the
very few thin sand layers.
The profiles of p0 obtained by Medusa DMT
4.3 DMT A-reading while penetrating procedure
using the three different test procedures are very
(DMTA-WP)
similar, despite the different techniques adopted
The capability of the Medusa DMT to maintain the for measuring the A-pressure, and in good agree
membrane in equilibrium with negligible horizontal ment with the profile of p0 obtained by traditional
displacement enables to obtain continuous measure DMT.
ments of the total horizontal pressure of the soil The profiles of p1 and p2 obtained by Medusa
against the membrane during penetration of the DMT (all test procedures) and traditional DMT are
probe. The DMTA-WP procedure consists in per nearly coincident. The values of p1 and p2 obtained
forming repeated A-pressure measurements (equiva by the DMTA-WP procedure are discontinuous,
lent to A-pressure reading at t = 0 instead of the because in this case the B and C pressure readings
standard time of t = 15 s) recorded during penetra are performed at depth intervals of 1 m, instead of
tion of the Medusa DMT at a constant rate. The 0.20 m as in the STD and DMT-RA procedures.
sequence of A-readings is generally taken over depth The profile of the material index ID, which
intervals of 1 m, corresponding to the typical length depends on the difference (p1 – p0), shows some
of push rods. Almost all penetrometers require to inconsistency between the values obtained by differ
stop penetration every meter to add a push rod, ent test procedures. In particular, the values of ID
during which B and C pressure readings may be calculated from p0 and p1 data acquired by the
taken without employing additional time. A constant DMTA-WP procedure appear significantly lower
penetration rate of 20 mm/s, as in the standard test than the ID values provided by the STD and the
procedure, is generally adopted. The current Medusa DMT-RA procedures. This discrepancy could be due
DMT equipment does not include instrumentation to to the fact that in the DMTA-WP procedure the
measure the penetration depth during the readings. A-pressure is measured at t = 0 instead of t = 15 s,
Most penetrometers include an encoder (for CPT resulting in lower values of the difference (p1 – p0),
measurements) and may output a time versus depth and for low ID values such incongruity is amplified
file, helpful for accurately associating the A-readings by the logarithmic scale. The values of the horizontal
to the corresponding depth at which they were taken. stress index KD, which depends only on p0, do not
When such information is missing, the time-depth seem influenced by the adopted test procedure.
relation may be estimated assuming an average
speed of penetration, estimated by measuring the
time for penetrating a 1-meter rod. The average 5 EVALUATION OF THE UNDRAINED SHEAR
speed and the time from the start of penetration STRENGTH FROM CPT/CPTU AND DMT
enables to estimate the depth of each A-reading.
Although not as accurate as with an encoder, the As described by Burghignoli et al. (1991) and Socco
error is reasonably limited in terms of % error, since dato (2003), the shear strength characteristics of the
each measuring interval is maximum 1 m long. Fucino clay were determined by means of a variety of
in situ and laboratory tests carried out in the 1987
investigation. In particular, the undrained shear
4.4 Comparison of results obtained by different
strength su was obtained from CPT and CPTU inter
test procedures
pretation with the usual relationships, respectively:
Figure 3 shows the comparison of the results
obtained by Medusa DMT using the three different
test procedures (STD, DMT-RA, DMTA-WP) and
the results obtained by traditional DMT. The test
194
Figure 3. Results obtained by Medusa DMT using different test procedures and by traditional DMT at Fucino-Telespazio.
observed that, below an upper desiccated crust, there

is a fair agreement between the su values resulting
from FVT and DMT. The SBPT provided values of
su significantly higher than those resulting from
where qc = cone resistance (CPT), qt = corrected other in situ and laboratory tests. Soccodato (2003)
cone resistance (CPTU), σv0 = total overburden suggested that low values of Nk for CPT are needed
stress, Nk and Nkt = cone bearing capacity factors for in order to match the data, while Nkt = 15 seems to
CPT and CPTU respectively, which were assumed to
range between 14 and 22.
The undrained shear strength su was estimated
from DMT results, both obtained by traditional
pneumatic DMT and by Medusa DMT in the 2020
investigation, using the original correlation proposed
by Marchetti (1980):
Figure 4 shows the comparison of the depth profiles

of the undrained shear strength su obtained at Fucino-
Telespazio from different in situ and laboratory tests in
past investigations and in the 2020 Medusa DMT cam
paign. In particular, the su profiles from the 1987 inves
tigation shown in Figure 4 (data from Burghignoli et al.
1991) were obtained in situ from several CPT and
CPTU tests, field vane tests (FVT), self boring pressure-
meter tests (SBPT), DMT tests, and in the laboratory
from direct simple shear tests (DSS-CK0U), unconsoli
dated undrained triaxial compression tests (UU) and
laboratory vane tests. For CPT and CPTU only the
lower- and upper-bound trend lines, corresponding to
the average su estimated from Eqs. 1 and 2 by assuming
Nk and Nkt equal to 22 and 14 respectively, are plotted
in Figure 4. The values of su from FVT in Figure 4 are
not corrected for strain rate and anisotropy effects.
Figure 4 shows an overall agreement between the Figure 4. Undrained shear strength su at Fucino-Telespazio
su values obtained from different in situ and labora from recent DMT tests vs. su from CPT/CPTU and other
tory tests. With reference to the various field tests in situ and laboratory tests obtained in the 1987 investiga
results obtained in 1987, Burghignoli et al. (1991) tion (modified from Burghignoli et al. 1991).
195
be appropriate for CPTU. As to laboratory tests, the comparison of the results obtained at Fucino-
values of su resulting from UU and from DSS-CK0 Telespazio indicate a substantial consistency of
U tests show a marked increase below a depth of measurements provided by Medusa DMT adopting
about 25 m, reflecting the increase of CaCO3 content different test procedures and by traditional pneu
with depth. This feature is not reflected by the results matic DMT.
of the in situ tests with the exception of the SBPT, The interpretation of Medusa DMT test results in
which provided a different trend of su below 24 terms of soil parameters takes advantage of the wide
25 m of depth. Burghignoli et al. (1991) commented experience available for the traditional pneumatic
that in highly structured soft clays, like the Fucino DMT test, and essentially shares the same set of
clay, the results of push-in in situ testing techniques, established soil property correlations available in lit
such as CPT/CPTU and DMT, can be hampered by erature. The profiles of su obtained from both
the destructuration of the tested soil induced by the Medusa DMT and traditional DMT in Fucino clay
insertion of the probe, which may partially obliterate are found in an intermediate position between the su
the effect of diagenetic bonds on the inferred su. profiles determined by other in situ and laboratory
In Figure 4 the profiles of su estimated by Eq. 3 tests, in particular by CPT/CPTU tests. This finding
from the results of the three Medusa DMT tests car is in agreement with other comparisons of su esti
ried out adopting different test procedures (STD, mated from DMT and from other tests, available
DMT-RA and DMTA-WP) and from the traditional from several soft clay test sites.
DMT carried out in the 2020 campaign are superim
posed to the profiles obtained by all test methods in
past investigations. As already noted with reference ACKNOWLEDGEMENTS
to the KD profiles (Figure 3), which are used to esti
mate su by Eq. 3, the su values are not practically This study was part of an activity cofunded by the
influenced by the adopted test procedure. The su Start-up & SME Booster Programme from the EIT
values obtained from Medusa DMT and the trad RawMaterials, funded by the EIT, a body of the
itional pneumatic DMT are almost coincident, and European Union supported under the Horizon 2020
also in very good agreement with the su obtained research and innovation program.
from DMT tests carried out in 1987. In general, the Telespazio – Fucino Space Centre is gratefully
profiles of su obtained from all DMTs plot in acknowledged for permitting to access the field test
between the su profiles determined by other in situ ing area, as well as for the continuous and friendly
and laboratory tests. This finding is in line with pre support during the Medusa DMT testing program in
vious experience. In fact, as described in the TC16 September 2020.
DMT Report (Marchetti et al. 2001), the su estimated
from DMT by Eq. 3 has generally been found to be
in an intermediate position between the su estimated REFERENCES
from other tests, as presented by various researchers
ASTM D6635-15. 2015. Standard Test Method for Per
in different clays (e.g., Nash et al. 1992 at the forming the Flat Plate Dilatometer. ASTM International,
National Research Site of Bothkennar, UK). West Conshohocken, PA, USA.
Burghignoli, A., Cavalera, L., Chieppa, V.,
Jamiolkowski, M., Mancuso, C., Marchetti, S., Pane, V.,
6 CONCLUSIONS Paoliani, P., Silvestri, F., Vinale, F. & Vittori, E. 1991.
Geotechnical characterization of Fucino clay. Proc. 10th
Benchmark GeoTest sites, such as the Fucino- European Conf. on Soil Mech. and Foundation Eng.,
Telespazio soft clay test site, prove to be of para Florence, Italy, 26-30 May 1991, 1: 27–40. Rotterdam:
Balkema.
mount importance for testing and validating
Foti, S., Lancellotta, R., Marchetti, D., Monaco, P. &
innovative soil investigation methods. In this Totani, G. 2006. Interpretation of SDMT tests in
respect, the recent experimental program at a transversely isotropic medium. Proc. 2nd Int. Conf. on
Fucino-Telespazio with Medusa DMT (the last- the Flat Dilatometer, Washington D.C., USA,
generation, fully-automated version of the DMT) 2-5 April 2006, 275–280.
could uniquely benefit of the availability of an ISO 22476-11:2017(E). 2017. Geotechnical Investigation
existing large and consistent data set obtained in and Testing – Field Testing – Part 11: Flat Dilatometer
past investigations from a variety of high-quality Test. International Organization for Standardization,
in situ and laboratory tests. Geneva, Switzerland.
Marchetti, D. 2018. Dilatometer and Seismic Dilatometer
Due to the increased accuracy of pressure meas
Testing Offshore: Available Experience and New
urements and controlled pressurization rate, the Developments. Geotech. Testing J. 41(5): 967–977.
Medusa DMT is particularly useful for testing very Marchetti, D., Danziger, F. & Jannuzzi, G.M.F. 2021. Com
soft soils, in which the measured pressures are typ parison of DMT results using traditional pneumatic
ically very small. Moreover, its technical features equipment and the Medusa DMT in the Sarapuí II soft
permit to implement alternative test procedures clay deposit in Brazil. Proc. 6th Int. Conf. on Geotech
(repeated A-readings, A-reading while penetrating), nical and Geophysical Site Characterisation ISC’6,
besides the ‘standard’ DMT procedure. The Budapest, Hungary, 26-29 September 2021.
196
Marchetti, D., Monaco, P., Amoroso, S. & Minarelli, L. Monaco, P., Marchetti, D., Totani, G., Totani, F. &
2019. In situ tests by Medusa DMT. Proc. 17th Euro Amoroso, S. 2022. Validation of Medusa DMT test pro
pean Conf. on Soil Mech. and Geotech. Eng. ECSMGE cedures in Fucino clay. Proc. 20th Int. Conf. on Soil
2019, Reykjavik, Iceland, 1-6 September 2019. Mech. and Geotech. Eng., Sydney, Australia,
Marchetti, S. 1980. In Situ Tests by Flat Dilatometer. 1-5 May 2022.
J. Geotech. Eng. Div. 106 (GT3): 299–321. Monaco, P., Tonni, L., Amoroso, S., Garcia Martinez, M.F.,
Marchetti, S., Monaco, P., Totani G. & Calabrese M. 2001. Gottardi, G., Marchetti, D. & Minarelli, L. 2021. Use of
The Flat Dilatometer Test (DMT) in Soil Investigations – Medusa DMT in alluvial silty sediments of the Po river
A Report by the ISSMGE Committee TC16. Proc. Int. valley. Proc. 6th Int. Conf. on Geotechnical and Geo
Conf. on Insitu Measurement of Soil Properties and physical Site Characterisation ISC’6, Budapest, Hun
Case Histories, Bali, Indonesia, 95–131. Official ver gary, 26-29 September 2021.
sion approved by ISSMGE TC16 reprinted in Proc. 2nd Nash, D.F.Y., Powell, J.J.M. & Lloyd, I.M. 1992. Initial
Int. Conf. on the Flat Dilatometer, Washington D.C., investigations of the soft clay test site at Bothkennar.
USA, 2-5 April 2006, 7-48. Geotéchnique 42(2): 163–181.
Marchetti, S., Monaco, P., Totani, G. & Marchetti, D. 2008. Soccodato, F.M. 2003. Geotechnical Properties of Fucino
In Situ Tests by Seismic Dilatometer (SDMT). Geotech Clayey Soil. Characterisation and Engineering Proper
nical Special Publication (GSP 180), From Research to ties of Natural Soils, 1: 791–807. Lisse: Swets &
Practice in Geotechnical Engineering, 292–311. Zeitlinger.
197
© 2022 the Author(s), ISBN 978-1-032-31259-0
An innovative new 3MPa CPT – to detect and measure very small fs values
A.J. McConnell
Founder of Insitu Geotech Services Pty Ltd (IGS)
E.J.C. Wassenaar
Regional Manager Asia-Pacific at Geomil Equipment B.V.
ABSTRACT: CPT testing of extremely soft soils and tailings materials requires the ability to detect and
measure very low qc values and extremely low fs values. The former of these can and has been solved by use
of low capacity cones with high quality well-calibrated load cells. The latter, detection and measurement of
extremely low fs values is an industry-wide problem, often treated as an “elephant in the room”. This paper
describes development of an innovative new CPT cone that the authors believe has largely solved this problem.
1 INTRODUCTION c) And, they take before-vs-after zero-load-drift

comparisons, as part of each test’s management.
Insitu Geotech Services (IGS) undertakes much CPT
testing in soils and sediments that can be described It’s rigorous, it works, and it’s business as usual,
in everyday terms as “extremely soft”, or even as built into the company culture and cost structures.
“ooze”; for clients who are seeking data that permits Following the above approach IGS has been suc
them to make confident designs. cessful in regard to the ability to defensibly and
As a consequence they are almost every day work repeatably measure very low qc values, and to be as
ing in an arena where one would aspire to better than good as reasonably achievable at measuring fs values.
Application Class 1 testing quality, if being described
according to ISO 22476-1-2012. 2 LIMITATION RE MEASURING VERY
To achieve the highest quality that they can in test SMALL SLEEVE FRICTIONS
ing these conditions and in satisfying clients they:
In regard to fs though, they have run up against the
– Use known good quality CPT cones. In soft soils “industry normal” limitation in measuring very small
these have been previously usually of 25MPa or sleeve frictions; a limitation that is associated with
10MPa tip capacity. the design of normal Compression Cones.
– Maintain/manage all CPTs meticulously. To help discussion of this, refer to Figure 1 below.
– Undertake in-house calibration-checking
recalibration under a very stringent program, as
described below.
The IGS calibration program is run in-house
because, as you can see from the explanation below,
it would be unworkable to contract it out:
a) Every cone is calibrated before every job, then

recalibrated after the job, each time using the
cone’s actual dimensions, not nominal values. If the
job runs more than one week, then the cones are
changed over with freshly calibrated cones on an
approximate 7-day service cycle. All of this is care Figure 1. Design of Cone Penetrometers (Lunne et al 1997).
fully recorded and a calibration-drift-performance
type of history is developed for each cone. In this paper type (a) is designated as a Compression
b) Recalibration data is then always compared to the Cone and type (c) as a Subtraction Cone.
data from the previous calibration, to cross-check For this discussion, the relevant difference between
for any significant change. a Compression Cone and a Subtraction Cone is that:
DOI: 10.1201/9781003308829-22
198
– A Compression Cone has separate load cells for the If one is testing stronger materials then it’s
tip and the sleeve. Hence these load cells can be not a significant issue at all, but if one is testing
sized for the purpose; typically a larger cell for the extremely soft material it becomes a problem.
tip and a smaller load cell for the sleeve. Until recently this limitation was typically han
– A Subtraction Cone has two load cells that must dled as an “elephant in the room”, not talked about
both be of the larger variety. Typically they would much, but rather agonisingly hanging around in the
be identical or nearly identical. One of these, that shadows in the background of CPT testing.
just behind the tip, measures the tip load only and But it’s become a pretty big deal for organisations
the other, above the screwed on connection to the like IGS that nowadays do much testing in extremely
sleeve, measures the combined tip-plus-sleeve load. soft or ooze-like soils; and for their clients.
– Software subtracts one value from the other to
determine the sleeve friction load. Hence a rela 4.2 References to this problem by others
tively big number is subtracted from another rela
This issue was discussed at CPT’14 in a paper
tively big number to get a smaller number. If
(Santos et al 2014) where the authors ascribed the
calibration of either or both load cells is not pre
problem to the friction in sleeve seals and pre
cise, or they drift differently during a test or after
sented an example sleeve calibration showing the
calibration, one can reasonably expect significant
difference in load cell readings compared to
errors in the sleeve values determined this way.
applied load (Figure 2a) without friction sleeve
Compression Cones (these authors perceive) were seals, and (Figure 2b) with seals.
developed in the first instance ostensibly to improve
a CPT’s ability to measure sleeve friction.
3 THE CURRENT PARADIGM
So this is the current industry-wide paradigm:

– If you want to most accurately measure sleeve
friction, you must use a Compression Cone.
In support of this statement that this is the current
paradigm, refer (Robertson et al 2015), as follows:
For accurate sleeve resistance measurements in
soft sediments, it is recommended that cones have
separate (compression) load cells.
And in support of something that follows in this
paper, quoting from the same document as above, the
following paragraph, referring now to Compression
Cones:
fs measurements, in general, will be less accurate Figure 2(a). Sleeve calibration with no seals (Santos et al
than tip resistance, qc, in most soft fine-grained soils. 2014).
The problem is that, in extremely soft or ooze-like
soils, this “in general expected lower accuracy” in fs
typically ends up meaning no fs measurement at all,
or something too low to be credible.
4 EXPLANATION – WHY VERY SMALL

SLEEVE FRICTIONS ARE ELUSIVE
4.1 A simple explanation – referring to Figure 1
– In a Compression Cone, the sleeve has to move

slightly to permit it to apply load to the friction
sleeve load cell.
– Dirt seals behind the friction sleeve resist this
slight movement and use up some of the force
applied by soil friction to the friction sleeve.
– Hence a noticeable error occurs if this applied
force is very low (as it will be in extremely soft
materials). Figure 2(b). Sleeve calibration with seals (Santos et al 2014).
199
Those authors discussed an innovative spring- only 3MPa; this being adequate for the testing of
loaded (preloaded) sleeve seal design to attempt to extremely soft materials that were/are the target.
solve the problem. This has/had been taken up by And, in accordance with (e) above they opted for
A.P. van den Berg for their cone design. a special alloy base for the load cells, giving
The issue was again mentioned in a paper at the a physical strain gauge response approximately
2021 Mine Waste and Tailings Conference, (Entezari 300% greater than it would be for a conventional
et al 2021). In that paper, the authors were discussing steel base.
limitations in the use of CPT data to determine the Of course, also in accordance with (c) above,
fines content of extremely soft oil sands tailings. IGS has adopted the same calibration regime dis
In that paper, for the analysis they were undertak cussed in Section 1, with the enhancement of
ing, many data points were “screened out in order to using dead weights for the application of load to
remove data where the soil-sleeve friction was less the tip and sleeve for (at the moment) the bottom
than internal o-ring (sic) friction”. approximate 6% of loading; ie up to ~0.17MPa on
This meant discarding a good deal of their data. the tip and ~11.5kPa on the sleeve. Planning is
that in the future IGS will go full-range with the
dead weights. At present they are using a very sen
5 IGS WANTED A SOLUTION – NOT AN sitive calibrated load cell of just 5kN capacity for
EXCUSE OR ANOTHER COMPROMISE the remainder of the range.
IGS wanted a solution, so talked at length to their
supplier/partner Geomil as to what this might entail. 6 HIGH HOPES AND EXPECTATIONS
They then jointly conceived and funded an atypical
design as a trial. Because of high hopes for these cones, IGS has set
The conversation and thinking went as follows: a policy in place that, to the extent possible, they will
a) As described in Section 4, Compression Cones calibrate these cones to achieve the accuracy out
have friction sleeves that must move a small comes proposed by the new draft international stand
amount to be able to register an fs reading. ard ISO/DIS 22476-1 criteria for Class 1+ cones.
b) The sleeves of Subtraction Cones do not have to To date that objective is looking good. A full IGS
move more than a miniscule amount to register calibration for one of these cones is shown on the
friction load. But they have the historically- last page of this paper.
deemed problem of having to subtract one big In other words, these cones, under these rigorous
number from another big number to get a small calibration processes, are pretty remarkable by cur
one, as discussed in Section 3. rent measures.
c) As one solution, would it be possible to develop Of course calibration is just one aspect of a CPT.
a Subtraction Cone with unusually high quality How have these cones behaved in the field?
and sensitive load cells, and to calibrate these
very rigorously to help overcome the problem in
subtracting a big number from a big number? 7 FIELD TRIALS
d) And would it be possible to design for much
lower cone capacity overall, hence making the IGS has had several 3MPa cones in operation now for
two big numbers smaller, further reducing the a few months, in one place testing very soft natural
problem? soils and in other places testing ooze-like tailings
e) And would it be possible to make the load cells sediments. The results are satisfying, demonstrating:
more responsive than normal, by adopting differ a) Ability to repeatably detect, measure and record
ent construction materials for the elastic bases on very low qc values – say 10kPa and even below.
which the cells’ strain gauges would be fixed? b) Ability to repeatably detect, measure and record
Of course this would all comprise a significant para very low fs values that make sense when com
digm shift, compared to the existing one. Reiterating pared to the qc values being measured – say 1kPa
here: If you want to most accurately measure sleeve and even below.
friction, you must use a Compression Cone. Figure 3 below, shows the improved response to
Some heart was taking in the knowledge that sleeve friction compared to what was previously
Fugro’s very sensitive Fibre Optic Cone, announced at possible.
the prototype stage via a paper (Looijen et al 2018) at Laboratory tests on samples taken from this coal
CPT’18 in Delft, also designed for testing soft mater tailings dam site showed the materials tested (in the
ials, is a Subtraction Cone with unusually sensitive test plot “troughs”) to have:
load cells; in principle the same idea.
So IGS and Geomil talked more and eventually – 37% fine sand, 51% silt size, 12% clay size.
Geomil made a first small run of special cones to the – Liquid Limit 34%, Plasticity Index 9%, Moisture
agreed design. In accord with (c) and (d) above, they Content 40.5% (ie in liquid phase).
opted for a cone size of 15cm2 and tip capacity of – Soil Particle Density (“SG”) 1.52t/m3.
200
Figure 3. Qc and fs plots – new 3MPa cone vs 10MPa Figure 4(b). SBT to 3.2m from the new 3MPa Cone.
Compression Cone – tests by IGS – plotted to 4.0m depth
for relevance.
CCS: Clay-like – Contractive – Sensitive
CC: Clay-like – Contractive
All the tests undertaken with these new cones CD: Clay-like – Dilative
show the ability to measure very low qc and a much TC: Transitional – Contractive
improved response in regard to sleeve friction. TD: Transitional – Dilative
SC: Sand-like – Contractive
SD: Sand-like - Dilative
8 CPT-BASED SOIL BEHAVIOUR TYPE (SBT)

The data from the new 3MPa cones interpret SBT
Data from the tests shown in Section 7 have been that is significantly different to that from the 10MPa
processed using the computer software CPeT-IT Compression Cone.
v.3.0 – CPT Interpretation Software (Ioannides
2007) to plot CPT-based Soil Behaviour Type using 9 FIELD TRIALS - ZERO-LOAD-OFFSETS
the classification system proposed by Robertson
(2016). Plots are shown below as Figures 4(a) The CPT industry/profession is passionate on the
and (4(b). issue of comparing “before-vs-after” zero-load
offset values as a method of determining/monitoring:
– the quality and actual condition of a cone itself, and
– the quality of a test that has been undertaken.
IGS’s cone history recording system will one day
allow a report on the reliability of this dependence
on zero-load-offsets; a quantified study to support
the passion (or otherwise).
In the meantime however it is acknowledged that
this is an accepted field indicator of the quality of a test
undertaken, so monitoring and recording these zero-
load-offsets is part of IGS’s every-test practice.
The new draft ISO/DIS 22476-1 relies heavily on
this parameter in determining test quality, called “Test
Category”, with different Categories/qualities num
bered A to D in the draft; A being the highest quality.
Under the draft standard, one decides the Test
Category by:
a) the Class of cone used and
b) the drift/difference in the before-vs-after zero
Figure 4(a). SBT to 3.2m from 10MPa Compression Cone. load-offset values.
201
In the draft standard, a before-vs-after drift/differ that instance had not been undertaken cautiously –
ence of less than 15kPa for the tip, combined with transient temperature shift was a possible cause of that
less than 2kPa drift/difference for the sleeve, categor aberration – remember that these are all very small
ises the test as Class A, the highest test quality numbers.
classification. By the time of the CPT’22 Conference IGS’s
This typically involves measuring the after-test database will be more comprehensive.
offsets once the cone has been cleaned and reassem
bled after testing. It cannot be a value extracted from
the completed test’s data file, as this immediately- 10 TRANSIENT TEMPERATURE EFFECTS
after value is likely to be influenced by dirt in seals
and gaps after cone extraction. Transient temperature effect is a matter that IGS
Figures 5(a) and 5(b) below show sets of before takes seriously in all of their testing. Their default
vs-after drift/difference data from the first three IGS testing procedures do what is reasonably possible to
jobs, using two of these new 3MPa cones. minimise these effects during all stages of a test.
The new 3MPa cones are each fitted with a tem
perature sensor and IGS is currently developing a plot
format for these cones that will include temperature
along with the usual parameters of qc, fs and u.
The work has not yet been done; however we
anticipate that these new cones may stabilise more
quickly than conventional cones due to the use of the
special alloy instead of steel in the load cell bases.
By the time of the CPT’22 Conference we authors
expect to be able to report on this aspect of these
new 3MPa cones.
Measurement of extremely low sleeve friction (fs)

values during CPT testing is an industry-wide prob
Figure 5(a). Before-vs-After Tip Zero-Load-Offsets -23 lem, often treated as an “elephant in the room”.
tests by IGS (grey shaded band is +/- 15kPa). This paper describes the development of an
innovative new CPT cone that the authors believe
has largely solved this problem.
The solution has involved shifting of the para
digm, that “if you want to most accurately measure
sleeve friction, you must use a Compression Cone”.
This solution involved the use of a Subtraction Cone
design.
The solution also involved development of more
responsive-than-conventional load cells using a spe
cial alloy for the load cell base, rather than steel.
So far the new CPT, calibrated and managed as
described in this paper, is meeting or exceeding the
authors’ expectations.
REFERENCES
Figure 5(b). Before-vs-After Sleeve Zero-Load-Offsets -23 Entezari, I., T. Boulter, S. McGregor, & J. Sharp (2021).
tests by IGS (grey shaded band is +/- 2kPa). Machine Learning to Estimate Fines Content of Tailings
Using Gamma Cone Penetration Test. Australia: Proc
The data shows that, except for one test, all of these AusIMM Mine Waste And Tailings Conference.
tests undertaken using the 3MPa cones have experi Ioannides, (2007) CPeT-IT CPeT-IT v.3.0 – CPT Interpret
enced drift/difference of less than the designated ation Software. (Online). Available: http://geologismiki.
gr/products/cpet-it
15kPa for the tip, and all achieved better than 2kPa for
Looijen, P., N. Parasie, D. Karabacak & J. Peuchen (2018)
the sleeve. Fibre optic cone penetrometer. The Netherlands: Proc
The single test that fell outside 15kPa for the tip CPT’18.
was the second test ever made using one of these Lunne, T., P.K Robertson & J.J.M Powell (1997). Cone
cones and the before-vs-after monitoring process in Penetration Testing In Geotechnical Practice.
202
Robertson P. K. & K.L. Cabal (2015) 6th Edition Guide update. Canadian Geotechnical Journal 53, 12,
to Cone Penetration Testing for Geotechnical December 2016.
Engineering. Santos R.S., A. Barwise & M. Alexander (2014). Improved
Robertson P. K. (2016) Cone penetration test (CPT)-based CPT sleeve friction sensitivity in soft soils. Las Vegas, USA,
soil behaviour type (SBT) classification system – an 3rd International Symposium on Cone Penetration Testing.
203
© 2022 the Author(s), ISBN 978-1-032-31259-0
Assessment of seismic cone penetration testing for small strain shear

modulus
N. Parasie, T. Franken & J. Peuchen
Fugro, Nootdorp, The Netherlands
ABSTRACT: Reliable values of in situ shear modulus at small strain (Gmax) are important for foundation
design, particularly for large investments in offshore wind energy. The seismic cone penetration test (SCPT)
method is commonly considered as primary reference for providing Gmax of soil. The SCPT data acquisition pro
cess covers a combination of cone penetration testing and seismic downhole testing. Values of shear wave velocity
are correlated with Gmax. This paper presents a complete inventory of SCPT standard uncertainty components for
Gmax. These components are ranked as Low, Medium or High, indicating relative impact on a combined standard
uncertainty budget for derived values of Gmax. Supplementary discussion is provided based on theory and practice,
with suggestions on uncertainty reduction. The focus of the assessment is on conventional SCPT systems, with
a hammer impact source at ground surface and a dual-receiver seismic cone penetrometer.
1 INTRODUCTION Examples include Gillespie (1990), Howie & Amini

(2005), Stolte & Cox (2020), and Styler et al.
The in situ shear modulus at small strain (Gmax ) is (2016). This paper provides a complete overview of
one of the important soil parameters for geotechnical standard uncertainty (ui ) components for Gmax (Sec
design situations for which relatively small displace tions 3 and 4) as basis for a combined standard
ments are critical to structure performance, for uncertainty (uc ) budget (Section 5). Supplementary
example monopile response to horizontal load is quantification and discussion are provided based on
governed by soil stiffness and monopile stiffness. By theory and practice, with suggestions on uncertainty
reducing the uncertainty in soil stiffness, the mono- reduction.
pile foundation design can be further optimized. A clear and complete insight in Gmax uncertainties
At this low strain level, the soil response is too is a first step into improving SCPT measurements
small for direct measurement of Gmax . Therefore, and consequently providing more accurate Gmax
Gmax is typically derived indirectly from shear wave values for foundation design.
velocity (vs ) and bulk (mass) soil density (ρ),
using Gmax ¼ ρ · v2s .
Due to good cost-quality ratio, the seismic cone 2 PREMISES
penetration test (SCPT) is one of the most commonly
applied acoustic methods for providing Gmax of soil. This paper uses a Type B evaluation of standard
SCPT data acquisition covers a combination of cone uncertainty (ui ) as defined by ISO (2008). Where
penetration testing and seismic downhole testing. a Type A evaluation is obtained from repeated
SCPT methodology is well documented; examples are observations, a Type B evaluation is obtained by
ISO 19901-8 (2014), ASTM D7400 (2019) and Cam scientific judgement based on all of the available
panella & Davies (1994). Styler et al. (2016) provide information on the possible variability of input
an excellent overview of 35 years of SCPT experience, quantities.
including equipment, procedures and interpretation of The presented uncertainty assessments consider
seismic signals. a conventional SCPT system and two settings for
SCPT data acquisition provides derived values of soil conditions (Table 1). Procedure, operator and
‘vs;vh ’, where v and h refer to (approximately) verti system design errors are not considered.
cally downward propagating (v), horizontally polarized The selected SCPT system and procedures
(h) shear waves. Hence, Gmax discussed in this paper comply with ISO 22476-1 (2012), ISO 19901-8
is ‘Gmax;SCPT ’. (2014) and ASTM D7400 (2019). It has
Various publications cover individual uncertainty a polarized hammer impact source at ground sur
components for Gmax derived from SCPT data. face, with the centre point of the source at
DOI: 10.1201/9781003308829-23
204
a fixed horizontal offset of 1 m from the initial components, which makes them more difficult to
vertical axis of the push rods of a CPT thrust quantify and mitigate.
machine (refer to Section 4.2 for discussion on The 4th column ‘Dep.’ Refers to dependencies of
other source offset distances). A dual-receiver ui components, i.e. primary and secondary uncertain
seismic piezocone penetrometer applies, with two ties shown by the IDs in the left column of each
sets of 3-component geophones with a centre-to table.
centre spacing of 1 m (refer to Section 4.3 for The 5th column ‘Notes’ includes items such as
discussion on other receiver spacings). The cone brief clarifications of the ui components, references
penetrometer is stationary for the selected test to literature and/or references to Section 4 for more
depth interval and the uncertainty assessments detailed discussion.
only consider data for the dual-receiver interval.
The selected signal sample rate is 3 kHz, with
incorporation of pre-trigger data recording. Stack 4 SUPPLEMENTARY NOTES
ing of 5 to 10 seismic traces is assumed to be
applied as standard practice. 4.1 Ray path between source and receivers
This section provides notes on Uncertainty ID 5 of
Table 3. The uncertainties assessed for cases C1 and
Table 1. Soil conditions – two cases. C2 assume straight-line ray paths between source
and receivers. The uncertainties are ranked as High
Depth Gmax qc vs ρ impact.
Case m MPa MPa m/s kg/m3
A straight-line assumption is according to
common practice, e.g. ASTM D7400 (2019). This
C1 OC clay 4-5 17 ~0.65 100 1700 practice should be adequate for relatively homoge
C2 Dense sand 25-26 180 ~36 300 2000 neous soil conditions within the immediate zone of
the depth interval selected for deriving Gmax . Soil
OC = overconsolidated qc = CPT cone resistance homogeneity (or otherwise) can be readily inter
preted from the location-specific CPT profile
acquired as part of the SCPT activities.
3 INVENTORY OF UNCERTAINTIES Actual ray paths between source and receivers are
largely controlled by (1) source offset and depth of
Tables 2 to 4 show ui components for Gmax, derived receivers discussed in Section 4.2 and (2) soil het
for Cases C1 and C2. The uncertainties are grouped erogeneity, particularly layer boundaries with high
by time/timing components, spatial/ distance compo contrast of acoustic impedance. Travel time errors
nents and equipment/setting components. Quantifica from interval methods generally diminish with
tion of the uncertainties is based on values published depth, as the angle of incidence of ray paths between
in literature when available, calculated estimates source and receivers becomes increasingly vertical
and/or best-estimate expert assessments (Type and the assumption of a straight line becomes more
B evaluation of ui ). applicable (Butcher & Powell 1996).
The 1st column ‘Uncertainty ID’ in the table lists Kim et al. (2004) and Hallal & Cox (2019) illus
the Gmax uncertainty components. trated high impact on vs (and thus Gmax ) with
The 2nd column ‘Impact’ refers to the influence of a linear ray path assumption where soil layer bound
the ui component on Gmax , expressed as percentage aries with high impedance contrast are in-between
of standard uncertainty (ui ) normalised to the corres the two sets of receivers. For this situation ASTM
ponding derived value for Gmax given in Table 1, i.e. D7400 (2019) suggests accounting for refracted ray
ui =Gmax . Tables 2 to 4 use the following categories: paths by applying Snell’s law using ray tracing
L = Low impact with ui =Gmax in the range of 0 to 1 methods (RT), or estimation of shear wave velocities
%, M = Medium impact with ui =Gmax in the range of from linear trends per layer with support from CPT
1 % to 5 %, and H = High impact with ui =Gmax 45 profiles using slope-based methods (SM). Kim et al.
%. Indicators with a ‘?’ are for a best guess uncer (2004) provide details for these methods.
tainty impact, i.e. more research may be required.
The selected ranges for ui =Gmax allow focus on
understanding and reducing those uncertainties that 4.2 Source offset
have M or H impact. About 2/3 of the uncertainly This section provides notes on Uncertainty ID 7 of
components fall in the L range, 1/6 in the M range Table 3. Cases C1 and C2 consider a lateral source
and 1/6 in the H range. offset of 1 m. The uncertainties are ranked as
The 3rd column ‘Comp.’ stands for ‘Complexity’ Medium impact and Low impact.
and refers to challenges in quantification of an ui ASTM D7400 (2019) suggests a lateral source
component, rated from Low (L), Medium (M) to offset in the range of 1 m to 6 m. In practice, both
High (H) complexity. High complexity items often small and large source offsets have advantages and
depend on specific situations or on other ui disadvantages.
205
Table 2. Uncertainties related to time/timing components.
Impact
Uncertainty ID C1 C2 Comp. Dep. Notes
1. Trigger L M L 4, 21 ISO 19901-8 (2014) and ASTM D7400 (2019) consider trigger timing.
timing Accurate trigger timing can be obtained in several ways, such as trigger-
receiver clock synchronization, or by using a very high signal sample rate.
Calculations indicate that trigger timing variability accounts for approxi
mately u1 =Gmax ¼ 1 % for C1 and u1 =Gmax ¼ 3 % for C2.
2. Signal L L L - Signal latency has no influence on interval velocity, as it is a constant
latency within a particular data acquisition system.
3. Arrival time L L L 1, 6, 14 Stolte and Cox (2020) indicated that arrival time selection methods (first
selection cross-over, maximum peak, cross-correlation) vary by approximately 0.1
ms if any influence from ground conditions is removed. A measurement
error of 0.1 ms implies u3 =Gmax ¼ 0:4 % for C1 and C2. Presentation of
a velocity range for several methods can be useful for insight in sensitivity
of arrival time selection.
4. Signal sam- L L L 10 ISO 19901-8 (2014) and ASTM D7400 (2019) consider signal sampling
pling rate rate. For recording of traces the Nyquist sampling theorem must be met, i.e.
a sampling rate that is at least twice the maximum source signal frequency.
Calculations indicate u4 =Gmax ¼ 0:1 % for C1 and u4 =Gmax ¼ 0:2 % for
C2. For accurate arrival time selection, the signal should be up-sampled to at
least 40 kHz.
Table 3. Uncertainties related to spatial/distance components.
Impact
5. Ray path H H H 4, 7, 8, Refer to Section 4.1.

between source 9, 11, 20
and receivers
6. Signal M M M 3, 4, 10, Howie and Amini (2005) indicated effects of waveform shape, near-field
damping and 11, 14, influence, attenuation, frequency content, peak frequency shift and signal
signal distor- 15, 21 widening on arrival time selection methods (first cross-over, maximum peak
tion along ray and cross correlation). For a depth range of 4 m to 10 m, the (collective)
path uncertainties were estimated to be within 0.2% to 2% of theoretical vs . This
implies u6 =Gmax ¼ 0:4 % to 4% for C1 and C2.
7. Source M L M 5, 8, 11, Refer to Section 4.2.
offset 20
8. Spacing of M M H 4,5,9, Refer to Section 4.3.
receivers 11,16
9. Spatial pos- L L H 3,5,7 ISO 22476-1 (2012), ISO 19901-8 (2014) and ASTM D7400 (2019) require
ition of cone determination of vertical position of a cone penetrometer by measurement of
penetrometer penetration length and inclination. Calculations indicate u9 =Gmax ¼ 0:7 %
relative to for C1 and u9 =Gmax ¼ 0:1 % for C2, based on considerations given by
source Peuchenand Wemmenhove (2020). Polarization analysis can be applied to
estimate relative orientation of the receivers.
For a linear ray path assumption, a source offset in a 1 m source offset to 20 % of vs for source offsets of
the range 1 m to 4 m will give a relatively small angle > 4 m. These values correspond to u7 =Gmax ¼ 2 %
of incidence between source and horizontal soil layer and u7 =Gmax ¼ 44 %.
ing. For heterogenous conditions, a source offset For source offsets > 4 m the assumption of linear
closer to the axis of the SCPT push rods is preferred, ray paths should not be applied and refracted ray
because the effects of refraction decrease with dimin paths should be accounted for (refer to Section 4.1).
ishing source offset (Kim et al. 2004). For penetration The use of larger source offsets can limit near-field
depths of less than 5 m, Butcher & Powell (1996) effects of refracted waves and improve estimation of
indicate u7 ranging from approximately 1 % of vs for interval velocities (Baziw & Verbeek 2014).
206
Table 4. Uncertainties related to equipment/ setting components.
Impact
10. Frequency spectra L L M - Low impact for C1 and C2 considers that the receivers are capable of recording
of source and the source frequency spectrum. Records of the source signature (signal) can be
receivers useful for evaluation of frequency spectra.
11. Soil stratigraphy M M H - Refer to Section 4.4.
and soil anisotropy
12. Imperfect source L L H 15 A hammer source is theoretically seen as a point source creating a perfect pulse
signal. In practice, the source has inertia and has a specific geometry; the pulse
signal will be imperfect. Gillespie (1990) concludes that source-soil coupling and
stacking significantly reduce impact of variations in the source impulse. Isolation
of the first arrival peak may be considered during data processing to remove, for
example, ringing in the recorded signal.
13. Inertia of push rod L? L? H - The receivers are fixed in a stiff metal rod instead of fully decoupled points in
and seismic cone soil. Resulting uncertainties appear unknown.
penetrometer
14. External noise H H H - Cases C1 and C2 consider common levels of external noise. External noise is typic
ally expressed as signal to noise ratio (SNR). Examples for SNR improvement are
(1) a more powerful source, (2) SCPT system engines switched-off and (3) selection
of day/ time of data acquisition with low external noise conditions related to traffic
and/or weather. SNR can be determined by evaluating the magnitude spectrum of
external noise recorded from e.g. pre-trigger of the receivers. Filtering for external
noise can be applied outside the frequency range of the seismic source signal.
15. Source-soil coup- L L M 11, 12 Depending on source characteristics and source-soil interface, soil degradation
ling and decoupling can progressively develop with number of hammer strikes. The use of dampers
from hold-down mass between hold dawn mass and source is common. Examples for potential improve
ment include (1) increase of the hold down mass (or vertical force), (2) imple
ment geometry/ material for high shear coefficients at source-soil interface and
(3) locally increase ground resistance by surface levelling and/or ground
improvement.
16. Receiver L L L 10 Detrending and demeaning of recorded data can be applied to remove system
characteristics drift. A systematic error of system drift between receivers is sometimes observed.
The cause of this error is unclear, although Styler et.al. (2013) rule out
a systematic orientation bias and time-bias between receiver pairs.
17. Receiver-soil L L M 8, 9, 11 SCPTs generally provide good receiver-soil coupling. Soil containing e.g. gravel,
coupling cobbles or cemented zones may be unfavourable. ISO 19901-8 (2014) and ASTM
D7400 (2019) allow a seismic receiver module with a slightly larger diameter com
pared to the push rod section below and above the seismic receiver module.
18. A/D conversion of L L L 4, 10 Cases C1 and C2 assume appropriate design for instrument response, analogue to
receiver data digital (A/D) conversion, jitter, etc.
19. Near field effects H L H 3,7, Especially at shallow depths (<5 m), seismic signals can be affected by interfer
8,9, 10, ence with surfaces waves, acoustic reflections from ground surface at shallow
12,14 depth and limited recordable source frequency spectrum.
20. Wave propagation L L H 5 Acoustic waves through steel push rod are much faster than vs for soil and there-
through push rods fore pose no problem in interpretation of S-wave arrivals. Decoupling of push
rods from the thrust machine (unclamping) may reduce unwanted wave propaga
tion through the push rods.
21. Source signal L L L 1, 12, Gillespie (1990) mentions time measurement repeatability within 0.05 ms due to
repeatability 15 varying frequency content of the source. For C1 and C2, this would imply
u21 =Gmax ¼ 0:2 %. Monitoring of non-normalized traces acquired during SCPT
data acquisition can allow removal of any ‘bad’ traces from the dataset.
22. CPT thrust L? L M - Uncertainty for Gmax for C1 depends on soil stratigraphy for depth zone of 0 to
machine imposes 5 m. In general, uncertainty is assessed as Low impact below about 5 m below
stresses and strains to ground surface.
soil
23. Push rods and L? L? H - The soil influence zone is in the order of 0.15 m around the push rod, potentially
cone penetrometer creating a higher velocity zone. The geometry of this zone typically shows minor
impose stresses and dependence on soil conditions. Therefore, uncertainty is assessed as Low impact.
strains to soil Actual uncertainties appear unknown.
24. Velocity to shear H H M 11 Contributing uncertainties include (1) applicability of theoretical model
modulus conversion Gmax ¼ ρ:v2s , (2) input of soil density ρ and (3) error for vs (squared). Soil density
derived from measurements of (nearby) high quality soil samples can reduce
uncertainty compared to CPT/ SCPT based correlations for soil density.
207
4.3 Spacing of receivers (1997) highlight cases for clays with azimuthal
stress variations where Gvh ≈1:3Ghv , i.e. the assump
This section provides notes on Uncertainty ID 8 of
tion of cross-anisotropy does not apply.
Table 3. The uncertainties for Cases C1 and C2 are
The SCPT method records signals of wave propa
ranked as Medium impact for a dual-receiver inter
gation in approximately vertical direction with par
val of 1 m.
ticle motion in an approximately horizontal plane.
A choice for spacing of receivers typically repre
This means that Gmax values derived for SCPTs are
sents a trade-off for considerations such as practic
largely representative of the Gvh component for
ability, depth resolution and accuracy of Gmax values
a simple cross-anisotropic case. Soil heterogeneity,
(e.g. Gibbs et al. 2018).
stress conditions and more complex anisotropy can
Variation in vs of approximately 2% has been
require further attention.
demonstrated by Styler et. al (2013) between
Note that cross-hole seismic testing (e.g. ASTM
receiver pairs spaced 0.5 m and 1 m at a depth of
D4428 2014) can approach the Ghh component.
5 m, corresponding to u8 =Gmax ¼ 4 % for Case C1.
Ghose (2012) presented results of two field
experiments conducted with a seismic cone pene
5 COMBINED STANDARD UNCERTAINTY
trometer with 7 tri-component accelerometers at
0.25 m spacing. An advantage of multi arrays is that
The uncertainty inventory of Tables 2 to 4 is com
it allows for evaluation of measurements over more
plete, thus allowing approximate estimates to be
multiple spacings within a single CPT push. No evi
made for a combined standard uncertainty (uc )
dence for reduction of data quality was observed
budget for Gmax.
with 0.25 m receiver spacing compared to larger spa
A simplified approach by assuming sensitivity
cings within the same CPT push. These findings sug
coefficients ci equalP n to 1, would be to calculate uc
gest that high accuracy Gmax data can be obtained
according to u2c ¼ ðci ui Þ2 (ISO 2008), where i
with relatively small receiver spacing. For heteroge i¼1
refers to the uncertainty IDs of Tables 2 to 4. Approxi
neous soil, a small depth interval improves Gmax
mately average values of the selected uncertainty inter
resolution with depth. On the other hand Gibbs
vals can be assumed for an initial estimate of uc :
et. al. (2018) mention that an improved resolution
ui =Gm ¼ 0:2 % for Low impact, ui =Gmax ¼ 2 % for
with 0.5 m receiver spacing may not compensate for
Medium impact and ui =Gmax ¼ 7% for High impact.
the difficulties in accurately determining the shorter
This calculation approach gives uc =Gmax ¼ 14:6 %
time delay between the traces compared to a system
corresponding with uc ¼ 2:5 MPa for Case C1
with 1 m spacing of receivers.
(Gmax ¼ 17 MPa) and gives uc =Gmax ¼ 12:8 % cor
Selection of a large depth interval allows ‘aver
responding with uc ¼ 23:0 MPa for Case C2
aged’ results for Gmax, with due consideration of ray
(Gmax ¼ 180 MPa).
paths affected by soil layer boundaries with high
The estimated uc values seem reasonable for the
impedance contrast (see Section 4.1). SCPT results
test method (SCPT) and cases considered.
for a combination of small depth intervals and over
lapping large depth intervals within one soil unit can
provide insight in Gmax sensitivity for receiver ACKNOWLEDGEMENTS
spacing.
The authors are indebted to many dedicated col
4.4 Soil stratigraphy and soil anisotropy leagues and gratefully acknowledge Fugro’s persist
ent commitment to excellence in seismic cone
This section provides notes on Uncertainty ID 11 of penetration testing. The opinions expressed in this
Table 4. The uncertainties for Cases C1 and C2 are paper are those of the authors. They are not necessar
ranked as Medium impact. ily shared by Fugro.
Comments on soil stratigraphy are included in
Section 4.1.
Soil anisotropy effects on Gmax are usually not REFERENCES
considered or well understood.
Soil can be represented by an elastic stiffness ASTM International, 2014. ASTM D4428/D4428M-14:
matrix with three shear components Gvh , Ghv and Standard test methods for crosshole seismic testing.
Ghh , where the subscript h denotes horizontal direc West Conshohocken: ASTM International.
tion and the subscript v denotes vertical direction. ASTM International, 2019. ASTM D7400/D7400M-19:
Soil is anisotropic, i.e. the isotropic condition where Standard test methods for downhole seismic testing.
Gvh ¼ Ghv ¼ Ghh is a theoretical simplification. West Conshohocken: ASTM International.
Baziw E. & Verbeek G., 2014. Identifying critical layers
Cross-anisotropy (axi-symmetric conditions where
using SCPT and seismic source moveout. In Cone Pene
Gvh ¼ Ghv ≠Ghh ) is a more realistic representation of tration Testing 2014: Proceedings of the 3th Inter
in situ soil conditions. Pennington et al. (1997) pro national Symposium on Cone Penetration Testing
vide an example for cross-anisotropy affecting Gmax (CPT14): Las Vegas, Nevada, USA, 12-14 May 2014
where Ghh ≈1:5Gvh . Furthermore Pennington et al. (pp. 357–364). CRC Press.
208
Butcher, A.P., & Powell, J.J.M., 1996. Practical consider International Organization for Standardization, 2014. ISO
ations for field geophysical techniques used to assess 19901-8:2014 Petroleum and Natural Gas Industries
ground stiffness. In Advances in Site Investigation Prac Specific Requirements for Offshore Structures – Part 8:
tice: Proceedings of the International Conference Held Marine Soil Investigations. Geneva: ISO.
in London on 30-31 March 1995. International Organization for Standardization, 2012. ISO
Campanella, R.G. & Davies, M.P., 1994. The seismic 22476-1:2012 Geotechnical investigation and testing —
piezocone: a practical site investigation tool. In: Geo Field testing — Part 1: Electrical cone and piezocone
physical Characterization of Sites, ISSMFE, Tech penetration test. Geneva: ISO.
nical Committee No 10 for XIII ICSMFE, New Delhi, Kim, D.S., Bang, E.S. & Kim, W.C., 2004. Evaluation of
India. various downhole data reduction methods for obtaining
Gibbs, P., Pedersen, R.B., Krogh, L., Christopher, N., reliable Vs profiles. Geotechnical Testing Journal, 27(6),
Sampurno, B. & Nielsen, S.W., 2018. Challenges in 585–597.
marine seismic cone penetration testing. In M.A. Hicks, Pennington, D.S., Nash, D.F.T. & Lings, M.L., 1997.
F. Pisanò & J. Peuchen (eds.). Cone Penetration Testing Anisotropy of G0 shear stiffness in Gault clay. Géotech
2018: Proceedings of the 4th International Symposium nique, 47(3), 391–398.
on Cone Penetration Testing (CPT18): Delft, The Neth Peuchen, J. & Wemmenhove, R., 2020. Depth accuracy of
erlands, 21-22 June 2018 (pp. 303–308). CRC Press. data points in marine soil investigation. In Z. Westgate
Gillespie, D.G., 1990. Evaluating shear wave velocity and (ed.). 4th International Symposium on Frontiers in Off
pore pressure data from the seismic cone penetration test shore Geotechnics (ISFOG 2020): proceedings
(Doctoral dissertation, University of British Columbia). (pp. 1036-1045). Deep Foundations Institute.
Ghose, R., 2012. A microelectromechanical system Stolte, A.C. & Cox, B.R., 2020. Towards consideration
digital 3C array seismic cone penetrometer. Geophysics, of epistemic uncertainty in shear-wave velocity meas
77(3), WA99–WA107. urements obtained via seismic cone penetration test
Hallal, M.M. & Cox, B.R., 2019. Theoretical evaluation of ing (SCPT). Canadian Geotechnical Journal, 57(1),
the interval method commonly used for downhole seis 48–60.
mic testing. In Geo-Congress 2019: Engineering Geol Styler, M.A., Weemees, I. & Mayne, P.W., 2016. Experi
ogy, Site Characterization, and Geophysics (pp. 376 ence and observations from 35 years of seismic cone
386). Reston, VA: American Society of Civil Engineers. penetration testing (SCPTu). Proceedings GeoVancou
Howie, J.A. & Amini, A., 2005. Numerical simulation of ver 2016 (Proc. 69th Canadian Geotechnical Confer
seismic cone signals. Canadian geotechnical journal, 42 ence, Westin).
(2), 574–586. Styler, M.A., Woeller, D, & Howie, J.A., 2013, Measuring
International Organization for Standardization, 2008. ISO/ down-hole shear waves from a vibrating perpetual
IEC Guide 98-3:2008 Uncertainty of measurement source during Cone Penetration Testing. In Canadian
Part 3: Guide to the expression of uncertainty in meas Geotechnical Conference: GeoMontreal, No. 442, Mon
urement (GUM:1995). Geneva: ISO. treal, Canada.
209
© 2022 the Author(s), ISBN 978-1-032-31259-0
Experimental procedure for checking the saturation degree of piezocone tips

I. Rocchi
Department Environmental Engineering, Technical University of Denmark, Denmark
L. Tonni, G. Gottardi & M. Marcolongo

Department DICAM, University of Bologna, Italy
ABSTRACT: A correct measurement of the water pore pressure is essential to obtain reliable CPT results.
Although this has always been a major concern, it still remains a possible relevant source of error in common
engineering practice. In addition to usual reasons, such as malfunctioning of the equipment, its poor calibration
and/or maintenance, a lack of complete saturation of the piezocone tip can play a major role. In fact, a correct
and timely pore pressure measurement relies entirely on the full saturation of the pore pressure system. The var
iety of saturation fluids and methods that is encountered in practice is a clear evidence of the still existing uncer
tainties on the most suitable procedure. Furthermore, unless the saturation process is explicitly detailed, its choice
is typically left to the contractor and the measurement quality can only be established a posteriori, after the test
has been completed. This paper describes a novel experimental equipment that enables quantifying the degree of
saturation of the piezocone tip prior to testing. The methodology proposed is inspired to the B-value check,
which is routinely used in geotechnical laboratory testing to assess the degree of saturation, e.g. prior to triaxial
testing. In addition, it could also be applied after a test, so that it would be possible to assess whether saturation
was retained while testing. This promising approach, when extended to different saturation procedures and valid
ated by engineering practice, has the potential to become a relevant benchmark for reliable CPT testing.
1 INTRODUCTION Furthermore, reliable pore pressure measurements are

also important because the correct cone resistance (qt)
Piezocone testing in its simplest version, i.e. including is determined as qt = qc + (1 - a)u2, where a is the area
independent measurements of cone resistance (qc), ratio of the cone, typically between 0.8-0.9. In add
sleeve friction (fs) and pore pressure (u2), was estab ition, thin lenses are better detected by sudden vari
lished in the eighties and has since gained a leading ations in the pore pressure readings, as layers ahead of
role in field classification and characterization of soils. the tip might influence the resistance recorded due to
Piezocone measurements can be affected by a number plugging or punch through phenomena (Lunne et al.
of imprecisions, as accurately explained by Sandven 1997). Furthermore, high quality measurements are
(2010). These include, but are not limited to malfunc fundamental to provide reliable input parameters. For
tioning of the equipment, poor calibration and/or example, compliance in the measuring system can pro
maintenance, which are shared with other experimen vide misleading results when assessing consolidation
tal techniques. However, a lack of complete saturation parameters by means of dissipation tests.
plays a key role and is specific to CPTU tests. Variables such as the (porous) filter element loca
Although this has long since been acknowledged and tion and design, the volume of the chamber connect
has been a major concern since the advent of CPTU, ing the filter to the pore pressure transducer, the
it still remains a large source of error in common saturation fluid and its degree of saturation, the air
engineering practice, possibly the most important. entry (i.e. cavitation) value of the system and lag
A correct measurement of the pore pressure is time required for “resaturation”, as well as the depth
essential to obtain reliable CPTU results in terms of and saturation of soil during testing, all affect the
both soil profiling and mechanical properties evalu pore-water pressure measurement during testing
ation. In terms of profiling, high quality measurements (dynamic measurement) and dissipation, upon stop
of pore pressure are particularly important for fine ping the advancement of the tip. However, the qual
grained materials and intermediate soils, where pore ity can only be established a posteriori, i.e. after the
pressure response is crucial to an accurate soil classifi test has been completed.
cation, for example using normalized pore pressure This paper describes an innovative approach that
parameters as suggested by Schneider et al. (2008). allows quantifying the degree of saturation of the
DOI: 10.1201/9781003308829-24
210
piezocone prior to testing. The methodology pro stress (i.e. cell pressure) in undrained conditions and
posed uses the pore-pressure parameters proposed by monitoring the response in pore pressure. The ratio
Skempton (1954) and is routinely used in laboratory between the increment observed in pore pressure and
testing to assess saturation, e.g. prior to triaxial test the increment applied in cell pressure (B) is approxi
ing. Preliminary laboratory measurements on a given mately 1 for saturated soils. Since the smaller the ratio,
piezocone, filter and saturation fluid under different the larger the error in the measurement, standards
saturation conditions of the filter are presented to require to proceed with testing only after having
establish whether there is a measurable correlation achieved B = 0.95-0.97 for sands and B = 0.97-0.99 for
between the pore pressure coefficient and the satur clays. Because correct pore pressure and volumetric
ation degree of a piezocone tip. strain measurements rely entirely on the full saturation
of the pore pressure system, if the B value is not satis
factory, the pressure of the hydraulic system (i.e. back
2 SATURATION IN LABORATORY TESTING pressure) is increased maintaining a constant effective
stress, so that air bubbles are dissolved and the com
Saturation of soil pores is of paramount importance in pressibility of the fluid reduced.
laboratory testing, not simply because the strength and
deformability properties may differ considerably
between saturated and unsaturated states, but also 3 SATURATION IN PIEZOCONE TESTING
because any effective stress analysis is developed
based on tests results obtained in experimental set ups Table 1 provides an overview of the porous materials
that rely heavily on this assumption. For example, the and pore size suggested in different standards and
consolidation theory formulated by Terzaghi & Fröh used in practice, based on an international survey by
lich (1936) is based on the assumption that volumetric DeJong et al. (2007). The great variety of saturation
strain is solely a result of change in the void ratio, or methods and media encountered in engineering prac
water content, of the soil, i.e. that the soil grains and tice is a symptom of the uncertainties that still exist on
water are incompressible and the soil is saturated. the subject. The most commonly used materials are
While it is possible to calculate the volumetric strain in relatively few. Polypropylene filters have the advan
an oedometer without relying on this assumption, by tage to be inexpensive and disposable, but have com
measuring the change in the height of specimen, this is parably higher compressibility than sintered steel or
not in other cases, such as triaxial tests. In this case, the ceramic filters, which are respectively subject to wear
volumetric strain calculated is based on the volume of or brittleness on the other hand. The range of the filter
water entering or leaving a closed hydraulic system pore size is rather wide. In fact, there is one order of
connected to the soil specimen. Therefore, any element magnitude difference in the ranges allowed by ASTM
in this hydraulic system having compressibility higher (D5778-12) and EN ISO 22476-1-2012 (i.e. the stand
than the soil specimen (i.e. air bubbles) affects the ard recommended by also the Eurocode7), while sev
strain measured as a result of an increment in stress. eral sizes are encountered in practice, where the
On the other hand, in a triaxial undrained test the largest size used is beyond the prescribed ranges.
effective stress is calculated as the difference be-tween
the applied total stress and the pore pressure measured
in a closed system connected to the soil specimen Table 1. Filter characteristics in piezocones.
Based on the pore-pressure coefficients (Skempton
Pore size
1954)
Setting Material μm
ASTM Standard Sintered steel or 20-200 μm

D5778-12 bronze
Where Δσ1 and Δσ3 are the changes in the principal Ceramic
stresses, while A and B are the pore-pressure coeffi Plastic
cients. Assuming a closed hydraulic system discon EN ISO (EC7) Sintered steel or 2-20 μm
Standard bronze
nected from the specimen, if the system is not
Ceramics
compliant (i.e. infinite stiffness) an isotropic increment
Porous PVC and
of stress, must result in an equal increment in the fluid
HDPE
pressure, which corresponds to the pore pressure in the
Carborundum
soil specimen. Compliance may arise from deformabil
Current practice* Sintered steel or 10, 50, 80, 120,
ity of the hydraulic system, soil grains or the fluid, but brass 150, 250 μm
typically, it is assumed that the two increments are the Ceramic
same for a saturated soil. In other words, the only rele Plastic
vant source of compliance is assumed to arise from air Sintered glass
in the soil pores or elsewhere in the system. beds
Saturation in triaxial tests is checked based on this
principle, by applying an isotropic increment of total *Data from DeJong et al. (2007).
211
With regards to the saturation fluid for the filter,
both ASTM and Eurocode7 suggest deaired water
only if tests are carried out throughout saturated non-
dilative (e.g. dense sands) soils. Otherwise, glycerin
or similar high viscous fluids are suggested, such as
silicon oil. However, according to DeJong et al.
(2007) a much wider variety of fluids is used, includ
ing hydraulic oil, vegetable oil and water-glycerin
mix. While higher viscosity diminishes the risks of
cavitation, it is counterproductive with respect to the
ability to deair the fluid. Similarly, risks of cavitation
reduce with pore size, but saturation is more difficult
to achieve due to lower permeability.
Finally, with regards to saturation of the filters,
ASTM suggests either applying a 90 kPa vacuum
while submerging the filters or boiling them in
water, both for at least 4 h. Eurocode7 suggests
either applying the filters in a vacuum for 24 h, or
boiling them in water for at least 15 min. DeJong
et al. (2007) investigated the minimum time Figure 1. Piezocone tip: (a) disassembled, (b) schematic
requirement and vacuum needed to ensure satur cross section showing likely location of air bubbles and (c)
ation based on a qualitative assessment of 15-45 assembled tip. 1 piezocone tip, 2 porous filter, 3 seals, 4
μm polypropylene filters. Their cross sections was chamber connecting porous filter to pore pressure trans
examined after saturation using different fluids ducer and 5 pore pressure transducer.
mixed with food coloring. Based on this study,
saturation by vacuum requires at least 85 kPa.
Glycerin takes at least 4 h to saturate the filters, 4 PROCEDURE PROPOSED TO CHECK
while water and silicon oil take just above and PIEZOCONE SATURATION
below 1 h, respectively. In particular, the best per
formance was obtained with low to intermediate Figure 2a shows a schematic diagram of the
viscosity silicon oil (100-1000 cS). device proposed to check the saturation of the
While a correct functioning of the pore pressure pressure sensing system of a piezocone directly
measuring system requires saturation of the filters, on site, whereas the photo of a prototype can be
this is not sufficient in itself to guarantee good results. seen in Figure 2b. In essence, the device applies
Air bubbles may be trapped within the tip threaded a pressure impulse on the assembled tip through
hollow (1) as showed in Figure 1, or in the chamber an annular chamber that is placed around the
connecting the pore pressure transducer and the filter piezocone filter (2 in Figure 2a). The invention is
(2). For this reason, the code of practice suggests spe patented by the University of Bologna (Rocchi
cific methods to assemble the tip to ensure saturation. et al. 2017).
In particular, ASTM suggests assembling the filter The ratio between the increment recorded by
onto the piezocone while submerged in the deaired the pore pressure sensor placed inside the piezo
medium used to prepare the elements, flushing all cone and an independent pressure transducer
confined areas with fluid to remove air bubbles, monitoring the impulse applied, is used to calcu
before tightening the tip on the piezocone. Eurocode7 late the pore pressure coefficient B. The degree
also suggests assembling the piezocone while sub of saturation of the piezocone pore pressure
merged in the de-aired saturation fluid. However, it measuring system can then be assessed as
suggests that saturation of the disassembled piezocone a whole, by associating the measured B ahead of
can be carried out either by injection of fluid in the testing to the quality of the CPTU test performed.
chamber connecting the pressure transducer and the Once the thresholds values for B that guarantee
filter or by placing the disassembled piezocone under satisfactory results in terms of pore pressure
vacuum for 15-30 min. Alternatively, filter-less piezo measurements in CPTU tests are established, test
cones can be employed, where the pressure transducer ing should proceed only if these values are
is in direct contact with the soil by means of a grease- achieved. Otherwise, the tip would be saturated
filled slot. This so called “slot-filtered” piezocone and assembled again, replacing the filter if neces
does not require a specific saturation procedure. How sary, until obtaining a satisfactory result (see the
ever, it must be ensured that no air is trapped while flow chart in Figure 3). The test could also be
assembling the tip. Furthermore, because of the performed at the end of the CPTU so as to
higher viscosity of the “saturation fluid” it might be assess whether saturation was retained during the
more difficult to ensure that it is deaired. test.
212
5 METHODOLOGY
Three different saturation conditions were investi

gated on a given piezocone, filter type and saturation
fluid. The filter position in the piezocone used was u2.
Furthermore, the filter was made of 5 μm sintered
stainless steel purchased already saturated with a 20
cS silicon oil. The tests were performed at the labora
tory floor, i) under ideal saturation conditions as
recommended by the standards; ii) partial desaturation
induced by pressurizing the filter with air at 85 kPa
for 24 h; iii) desaturation by pressurizing the filter
with air at 600 kPa for 1 h. Based on the weight loss
recorded, which is due to the oil expelled from the
filter, the saturation degree was estimated to be 100%,
91% and 3%, respectively. These conditions are there
fore subsequently referred to as S, for ideal saturation
conditions and LD and HD, for low and high desatur
ation, respectively. While they clearly do not corres
pond to the conditions encountered as a result of
different saturation procedures they were selected to
investigate the widest extent by which the measured
B can vary with relevant saturation degrees.
In all cases, ahead of mounting the filter and
assembling the cone, the piezocone and the seals
were placed under an 85 kPa vacuum in a silicon oil
chamber. After 15 min, the vacuum was released and
a syringe was used to inject silicon oil in the cham
ber connecting the pore pressure transducer to the
filter (4 in Figure 1), while the piezocone remained
submersed. After an additional 15 min under
Figure 2. (a) Schematic diagram of the device seen in vacuum, the tip was assembled while immersed in
a cross section, while fitted on the piezocone (side view). 1 silicon oil and tightened. For the tests designated
Piezocone tip and 2 porous filter. (b) Photograph of the with S only, the assembled cone was subsequently
prototype while mounted on a piezocone. placed under vacuum in the same oil chamber for
further 15 min.
Afterwards, the device was positioned onto the
cone while the assembled cone was still in the oil
chamber so as to ensure hydraulic contact, while
applying a small pressure (i.e. <5 kPa). The cone and
the device were then remove from the oil chamber
and a 50 kPa was applied in the device chamber. Sub
sequently, the device was used to apply a series of
pressure impulses ranging from 50 to 300 kPa. Each
impulse of nominally 50 kPa, was applied within a 1
s interval and kept for 30 s before increasing the pres
sure in the device further. The pore pressure measure
ments recorded by the piezocone and the device were
datalogged continuously during the process.
Figure 4 shows a set of preliminary results

obtained, where it can clearly be seen that the meas
ured pore pressure coefficient B differs for the 3
saturation conditions employed. For each of the
Figure 3. Flow chart of testing procedure including assess conditions investigated, 3 filters were tested. The
ment of the pore pressure measuring system saturation. subsequent results present the average values
213
measured under each condition, while the error bars a posteriori. In addition, this measurement can also
indicate the minimum and maximum values be applied after carrying out a test so that it would
recorded. Under each condition, there is a relatively be possible to assess whether saturation was
good degree of repeatability, independently of the retained.
pressure applied for fully saturated conditions. On The preliminary results presented herein dem
the contrary, B increases with the pressure applied onstrate that there is a strict correlation between
for LD because the compliance of the system is the pore pressure coefficient and the saturation
reduced, as the pressure increases and air becomes conditions of the filter. Based on this proof of
less compressible. As expected, the B values meas concept, the device invented should be validated
ured for HD are almost zero because of the almost on other tips and with different saturation degrees
fully desaturated conditions of the filter. However, of the entire hydraulic system and could be then
there is a considerable difference measured at subsequently employed to investigate the most
50kPa for S and LD conditions, even though the suitable saturation processes and media. Further
saturation degree of the filter is still above 90%. In more, comparison of pore pressure measurements
summary, a strict correlation between the saturation for CPTU performed on a same unit under differ
degree of the filters and the pore pressure coeffi ent conditions as attested by the proposed meth
cient B was found. odology are recommended to identify threshold
values suitable for testing.
REFERENCES
ASTM D5778-12, 2012. Standard method for electronic
friction cone and piezocone penetration testing of soils.
ASTM International, West Conshohocken, PA.
DeJong, J.T., Yafrate, N.,J., DeGroot, D.J. 2007. Design of
a miniature piezoprobe for high resolution stratigraphic
profiling. Geotechnical Testing Journal, 30(4),1–11.
EN ISO 22476-1-2012, 2012, Geotechnical investigation
and testing.– Field testing – Part 1: Electrical cone and
piezocone penetration testing. Geneva, Switzerland: ISO.
Lunne, T., Robertson, P.K., Powell, J.J.M. 1997. Cone penetra
tion testing in geotechnical practice. Chapman & Hall, UK.
Rocchi, I., Tonni. L., Gottardi, G.; University of Bologna.
Device for checking the degree of saturation of a pressure
sensor unit of a piezocone and method for performing the
Figure 4. Comparison of pore pressure coefficient meas check. Italian patent PCT/IB2017/053104, filed 26/05/2017.
ured on piezocone assembled with filters having different Sandven, R. 2010. Influence of test equipment and proced
degrees of saturation. ures on obtained accuracy in CPTU. In M.A.J. Williams
& H. Faure (eds), Proceedings 2nd international sympo
sium on Cone Penetration Testing, Huntington Beach,
7 CONCLUSIONS CA, USA. Volume 1: Keynote Lectures, Paper No. KN1
Schneider, J.A., Randolph, M.F., Mayne, P.W., Ramsey, N.
The article discusses the still open issues related to R. 2008. Analysis of factors influencing soil classification
using normalized piezocone tip resistance and pore pres
poor saturation in piezocone testing and presents an sure parameters. Journal of Geotechnical and Geoenvir
innovative methodology and device, based on onmental Engineering, ASCE, 134(11): 1569–1586.
a pore-pressure coefficient, that can be used to quan Skempton, A. 1954. The pore-pressure coefficients A and
titatively assess saturation before proceeding to test B. Geotechnique, 4, 143–147.
ing. This would provide a great advantage as Terzaghi, K., Fröhlich, O.K. 1936. Theorie der Setzung von
typically the quality of tests can only be assessed Tonschichte, Leipzig/Wien Deuticke.
214
© 2022 the Author(s), ISBN 978-1-032-31259-0
Using penetrometer in situ and in box-core testing to obtain design

information for lazy wave riser-soil interaction
O. Safaqah
ExxonMobil, Houston, Texas, USA
H.E. Low, S. Pant & S. Ingarfield

Fugro
M.F. Bransby & M.F. Randolph

Centre for Offshore Foundation Systems, Oceans Graduate School, The University of Western Australia, Australia
Z.J. Westgate
Department of Civil and Environmental Engineering, University of Massachusetts Amherst, USA
Formerly Fugro
ABSTRACT: Geotechnical assessment of vertical seabed stiffness during small vertical cycles of riser
movement is important for riser fatigue analysis. These stiffness values change both with the amplitude of the
cycles (because of non-linear stiffness) but also as the seabed consolidates under the small amplitude loading
during long-term operation. To encapsulate both effects in design, a series of site-specific small displacement
ball penetrometer testing was conducted both at seabed and in box-core samples on deck during
a geotechnical survey for a recent deep water project. The paper describes the cyclic load and displacement
controlled in situ testing, the results obtained, and how this data aided design. Of particular note were the long
duration cyclic tests which were carried out for sufficiently long so as to not only capture secant stiffness
reduction due to remoulding, but also subsequent stiffness increases due to consolidation. This novel test pro
gram performed on intact soil (as opposed to reconstituted samples in the laboratory or geotechnical centri
fuge) confirms that consolidation hardening can occurs in naturally structured near-seabed soils.
1 INTRODUCTION investigated the hysteretic seabed resistance devel

oped during relatively small riser movements and
Subsea risers experience cyclic and axial strain due how the normalized stiffness (Ksec = ksec/Ncsu,
to environmental and operational loads. These are where ksec = ΔP/Δz, Nc is a bearing factor, and su is
generally concentrated within the touchdown zone the undrained shear strength) reduces with increasing
(TDZ), or the zone of the riser which first comes displacement amplitude.
into contact with the seabed. Since operational riser As compared to the SCR, TDZ movements of
movements are predominantly vertical in the TDZ, a lazy-wave riser (LWR) are several orders of magni
the riser response depends on the vertical seabed tude smaller, in the range of 0.1 to 1% of the riser
stiffness. Previous investigations of vertical seabed diameter. These movements may lead to initial seabed
stiffness concentrated on steel catenary risers (SCRs) softening but are unlikely to lead to significant tren
which experience large amplitude movements within ching. Instead, the focus is on the non-linear relation
both the vertical and lateral planes, resulting in ship between seabed stiffness and riser displacement
seabed softening and trenching (Bridge and Howells, amplitude, with the structural fatigue assessment using
2007). This has led to the development of non-linear equivalent linear-elastic springs (ksec).
soil models that describe the change in vertical Recent data obtained from centrifuge tests in nor
seabed reaction (V, or P) during vertical penetration mally consolidated clay (e.g. Hodder et al., 2009; Yuan
(z), extraction and re-penetration (e.g. Bridge et al. et al., 2016) suggest that the riser-seabed stiffness
2004; Randolph and Quiggin, 2009; Aubeny and could increase or ‘harden’ with time due to soil con
Biscontin, 2009; Clukey et al. 2017; see Figure 1). solidation. For overconsolidated clay this consolidation
Although there has been significant focus on large hardening effect on riser seabed stiffness remains
amplitude movements, some models have uncertain (Clukey et al. 2017). Long-term
DOI: 10.1201/9781003308829-25
215
consolidation hardening has only been observed in 3 FIELD TESTING
model tests conducted with reconstituted clays or
kaolin clay (e.g. Clukey et al., 2005; Hodder et al., 3.1 Overview
2009, Yuan et al., 2016; Hou et al., 2018) and to the
authors’ knowledge there is no published model test The field testing comprised in situ ball penetrometer
data on intact soil samples nor is there any existing testing at seabed and on-deck ball penetrometer test
design framework to incorporate this hardening behav ing in box core samples. The ball penetrometer tests
ior in a site-specific manner. comprised ‘standard’ tests (such as monotonic pene
This paper reports a series of in situ tests con tration/extraction and large amplitude displacement-
ducted offshore on the seabed and in box cores to controlled cyclic events to measure su profiles and
support a deep water oil and gas development. The sensitivity (St), respectively). These were supple
tests were able to take site-specific measurements of mented with novel small amplitude load and dis
cyclic softening and hardening and were used to placement-controlled cyclic tests designed to
inform riser design. Recommendations for future replicate the expected field riser movement. The
testing are made based on the findings. advantage in the box core and in situ tests is that the
‘crust’ material is undisturbed, avoiding the chal
lenge of having to recreate field conditions in
a laboratory model test set-up. This allowed the
in situ soil response to be measured, capturing the
effect of soil structure of an overconsolidated clay
on changes in stiffness. Another advantage of this
approach is that test results were available during the
geotechnical campaign, rather than having to wait
for sample transport, testing and interpretation as
would be required during conventional laboratory
ex-situ testing.
Figure 1. Schematic of cyclic vertical load-penetration

response for pipelines and risers.
2 SITE AND DESIGN CONDITIONS
The soil conditions across the project site com

prised a thin (5 to 15 cm) drape of very soft clay
with undrained shear strength (su) values less than 5
kPa, below which is a soft clay layer with a slight
‘crust’ with su up to 12 kPa, extending down to
0.5 m below the seabed. The strength profile then
gradually reduces to normally consolidated condi
tions (Figure 2).
A conventional geotechnical site investigation
campaign was conducted including sampling by
piston coring and box coring and in situ testing by
T-bar penetrometer. The long term cyclic ball pene
trometer tests described here are novel and were per
formed specifically to support geotechnical design
advice regarding operational riser-soil interaction.
During operation, the lazy-wave risers (LWRs) Figure 2. Undrained shear strength profile at site.
were expected to cycle vertically under small ampli
tudes ranging from about 0.1% to 1% of the riser 3.2 In situ ball penetrometer testing
diameter. Since high seabed stiffness reduces fatigue
life, accurate quantification of the stiffness in the In situ testing was performed using the Fugro Sea-
‘crust’ layer was key to the riser design and the vari Calf seabed frame system deployed from the back of
ation in stiffness throughout their design life was a geotechnical drilling vessel (Figure 3). A ball
important to quantify. penetrometer of diameter 78 mm (hence projected
216
base area of 4778 mm2) was pushed with a rod of
25 mm diameter and measurements of penetration
resistance, sleeve friction and pore pressure were
recorded during the tests.
The cyclic tests involved displacement-controlled
cycles of fixed amplitude (typically ranges from
±0.02 m to ±0.15 m) conducted at a displacement rate
of 0.02 m/s. The cyclic tests started with 30 cycles of
large amplitude cyclic displacement (± 0.1 m to
±0.15 m) to remold the soil, followed by long term (up
to 35 hours) small amplitude cyclic displacement (±
0.02 m) to measure the long term cyclic load response.
Figure 3. Fugro SEACALF system.
3.3 In box core ball penetrometer testing

A ball penetrometer of 33 mm diameter was actuated
using Fugro’s DeckScout system mounted on the top
of 0.5 m x 0.5 m x 0.5 m box core on the deck of the
drilling vessel (Figure 4).
The tests were started by penetrating the ball
monotonically to a target depth at a displacement
rate of 200 mm/min, measuring the plastic penetra
tion. The ball was then unloaded until the bearing
pressure reached approximately 1 kPa and ready for
the start of the test.
The long term cyclic test was then conducted using
load control, with load limits set at +11 kPa (max
imum value) and approximately 0 kPa (minimum
value) for each cycle. The ball penetration rate was
about 0.06 mm/min while the extraction rate was
about 0.12 mm/min. This difference in displacement
rate was an artefact of the testing system, but is not
expected to have affected the results. The cycling was
continued until no obvious changes in secant loading
stiffness were observed, which required up to 8 hours
(2600 cycles) of continuous cyclic loading.
4 RESULTS: IN SITU TESTING
A typical result for the in situ cyclic ball penetrom

eter test is shown on Figure 5. For each cycle of the Figure 4. Fugro DeckScout system showing (a) box core
test, the secant stiffness (ksec) was calculated using sampler and (b) close up of actuator with miniature ball
the change in bearing pressure (Δq) from peak to penetrometer.
217
trough (and vice versa) and the associated change in
penetration (Δz). Figure 6 presents the resulting
trend of increasing stiffness (in the loading phase)
with cycles, with an approximately 40% increase in
stiffness by the end of the test compared to the initial
value. The final level of consolidation in the soil for
this test was only about 50% based on the normal
ized time factor (T = tcv/D2) calculated to be 0.16,
where t is the test duration, cv is the coefficient of
consolidation, and D is the ball penetrometer diam
eter. Further cycling would therefore have likely
increased the secant stiffness further, but was not
feasible because of vessel time availability.
Figure 6. Typical result from a long-term cyclic in situ ball

penetrometer test (Δz/D ≈ 80%).
Figure 5. Overall load-penetration response during a

long-term cyclic in situ ball penetrometer test.
5 RESULTS: IN BOX CORE TESTING
A typical box core ball penetrometer test result is Figure 7. Overall load-penetration response during a
shown on Figure 7. Because the cycles were load- long-term cyclic box core ball penetrometer test.
controlled and the soil stiffness changed during the
testing, the cyclic displacement amplitude reduced
during the testing as shown in Figure 8a from about 5.1 Measurement quality
1.2% of ball diameter at the start of the test to about Both the in situ SeaCalf and on-deck box core test
0.6% of diameter at the end. ing were limited by the control mechanism used,
As for the in situ testing, for each cycle of the in which resulted in some variation in cyclic amplitude.
box core testing the secant stiffness (ksec) was calcu For the SeaCalf testing, the displacement-controlled
lated using the change of bearing pressure (Δq) from condition resulted in soil consolidating away from
peak to trough (and vice versa) and the associated the maximum depth range of the ball penetrometer,
change in penetration (Δz). Results are shown in thus likely not mobilizing the full soil resistance
Figure 8b, demonstrating the cyclic hardening behavior during later downward movements. For the box core
resulting in an increase of stiffness of approximately testing, the very small amplitude ball movements per
80%, slightly higher in the loading phase compared to cycle, e.g. 0.2 mm at the end of the test, challenged
unloading. The normalized time factor T = 0.72 for the precision of the displacement measurement
this test is higher than the in situ SeaCalf tests, result system. In both cases, there is significant ‘noise’ in
ing in increased hardening. However, it replicates the the acquired data. Consequently, long term moving
results of similar testing performed in centrifuge model averages were used to smooth out the data to allow
tests (e.g. Hodder et al., 2009, Yuan et al., 2016). the assessment of the evolution of soil stiffness as
218
results in fresh seabed soil being subject to this
cyclic process and therefore limiting the time avail
able for soil to harden. Because of these larger
motions and displacement of the seabed soils, SCRs
may generally be considered to be displacement-
controlled in the TDZ region most subject to fatigue
concerns.
For LWRs, however, the cyclic amplitudes in the
TDZ are much smaller than for SCRs due the lazy-
wave configuration, and the TDZ is therefore less
likely to relocate during the design life. The motions
of a LWR near the front of the TDZ may be displace
ment-controlled since they are driven by the transla
tion of movements in the riser initiating at the
floating facility. However, further along the TDZ,
the upward movements are by definition displace
ment-controlled, but the downward movement is
gravity-controlled, i.e. load controlled, and the
resulting cyclic settlement of the riser into the
seabed is governed by its on-bottom weight. This is
likely to be similar for SCRs towards the ‘back’ (sta
tionary end) of the TDZ where the catenary-induced
stress concentration is negligible.
With further understanding of these differences in
load and displacement-controlled behavior between
SCRs and LWRs, designers can more accurately
model the real response, or at minimum consider
a conservative approach to fatigue design comprising
a high estimate of seabed stiffness. The long-term
vertical seabed stiffness of the seabed under loading
conditions representative of a LWR, as illustrated
through the data presented in this paper, can increase
by up to 80% through this consolidation hardening
process. Additional cycles of movement and time
may have resulted in further increases in secant
stiffness.
Figure 8. Typical results from a long-term cyclic box core

ball penetrometer test 7 CONCLUSIONS
A geotechnical site investigation performed at a deep

the tests progressed. Improvements to the testing water site in soft clay comprised a novel program of
devices have since been made to reduce this meas cyclic ball penetrometer testing at the seabed (in situ)
urement uncertainty for future projects. and in box core samples on deck of the vessel. The
tests were performed under displacement-controlled
(in situ) and load-controlled (in box core) conditions.
6 IMPLICATIONS FOR DESIGN The objective of the tests was to assess the potential
for consolidation-induced increases in soil strength
There remains debate within the subsea riser design and resulting vertical seabed stiffness under small
community on the appropriateness of incorporating amplitude movements and loads which could detri
consolidation hardening in riser design (Clukey mentally affect the fatigue life of the risers. The data
et al., 2017). The debate currently hinges upon presented confirm that consolidation hardening can
whether cyclic riser response at the seabed is a load- occur in undisturbed overconsolidated clay, extending
controlled or a displacement-controlled process. For the observations by others for centrifuge and single-
SCRs, the large amplitude vertical (and often lateral) gravity model tests in reconstituted and normally
motions of a riser under operational conditions limit consolidated clays.
the potential for consolidation hardening to occur. The results suggest that consolidation hardening
This is for two reasons: first, the seabed is continu may be relevant for lazy-wave riser fatigue analysis,
ously being softened and displaced due to the larger given the potential for long-term small amplitude
riser motions; second, translation of the TDZ both cyclic movements which occur at a common touch
longitudinally and transverse to the original TDZ down zone throughout the riser design lifetime.
219
ACKNOWLEDGEMENTS ing soil-pipe interaction for SCR fatigue assessments. In
Offshore Technology Conference. Paper Number: OTC
The authors would like to express their appreciation 27564-MS, Houston, Texas, USA.
to Fugro and ExxonMobil for allowing publication Einav, I. and Randolph, M.F. (2005). Combining
of this data. The fifth author holds the Fugro Chair, upper bound and strain path methods for evaluating
whose support is gratefully acknowledged. Discus penetration resistance. International journal for
sions with Professor David White of the University numerical methods in engineering, 63(14),
of Southampton (formerly UWA) on consolidation pp.1991–2016.
Fahey, M. and Carter, J.P. (1993). A finite element study of
hardening are also acknowledged. the pressuremeter test in sand using a nonlinear elastic
plastic model. Canadian Geotechnical Journal, 30(2),
pp.348–362.
REFERENCES Hodder, M., White, D. and Cassidy, M.J. (2009). Effect of
remolding and reconsolidation on the touchdown stiff
Aubeny, C.P. and Biscontin, G., (2009). Seafloor-riser inter ness of a steel catenary riser: guidance from centrifuge
action model. Int. Journal of Geomechanics, ASCE, 9(3) modelling. In Offshore Technology Conference, OTC
133–141. 19871-MS, Houston, Texas, USA.
Bridge, C., Laver, K., Clukey, E.C., and Evans, T.R. Hou, Z., Sahdi, F., Gaudin, C., and Randolph, M. (2018).
(2004). Steel catenary riser touchdown point interaction Evolution of riser-soil stiffness in a soil crust layer. In
model. Proceedings Offshore Technology Conf., Vietnam Symposium on Advances in Offshore Engineer
OTC16628, Houston, Texas, USA. ing (pp. 130–136). Springer, Singapore.
Bridge, C.D. and Howells, H.A., (2007). Observations and Randolph, M.F. and Quiggin, P., (2009). Non-linear hys
modeling of steel catenary riser trenches, Proceedings teretic seabed model for catenary pipeline contact,
of the 17th International Offshore and Polar Engineer Proc. 28th Int. Conf. Ocean, Offshore and Arctic
ing Conference, Lisbon, Portugal, July 1-6, pp. Engineering, May 31 – June 5, Honolulu, Hawaii,
803–813. USA.
Clukey, E.C., Haustermans, L. and Dyvik, R. (2005). Viggiani, G. and Atkinson, J.H. (1995). Stiffness of
Model tests to simulate riser-soil interaction effects in fine-grained soil at very small strains. Geotechnique, 45
touchdown point region. In nternational Symposium on (2), 249–265.
Frontiers in Offshore Geotechnics (pp. 651–658). Yuan, F., White, D. J., and O’Loughlin, C. D. (2016). The
Clukey, E.C., Aubeny, C.P., Zakeri, A., Randolph, M.F., evolution of seabed stiffness during cyclic movement in
Sharma, P.P., White, D.J., Sancio, R. and Cerkovnik, M. a riser TDZ on soft clay. Geotechnique, 67(2), 127–137.
(2017). A perspective on the state of knowledge regard July 10.1680/jgeot.15.P.161.
220
© 2022 the Author(s), ISBN 978-1-032-31259-0
Calibration of cone penetrometers according to International Organization

for Standardization requirements
R. Soage Santos
Ørsted, London, UK
E. Gómez Meyer & J. Peuchen

G. Yetginer
Equinor, Stavanger, Norway
T. Lunne
Norwegian Geotechnical Institute, Oslo, Norway
T. Carrington
Fugro, Wallingford, UK
ABSTRACT: Detailed requirements for calibration of piezocone penetrometers are incorporated in standards
published by International Organization for Standardization (ISO). The use of these standards can provide
input for comparisons of cone penetration test (CPT) systems deployed in practice by means of cone pene
trometer classes. It is important that parties specifying or supplying CPT data take note of the implications of
the new requirements and the opportunities and benefits of appropriate selection of cone penetrometer classes.
In addition, the information available from calibration and verification of a particular cone penetrometer can
provide input into estimation of uncertainties of data points in CPT profiles used for design of structures.
This paper focusses on background information about topics considered for development of the ISO
requirements. These topics included (1) practical and economical test methods achievable in
a calibration laboratory, (2) assessment of differences in exposure conditions applied in the calibration
process and site conditions likely to be encountered during actual cone penetration testing, and (3)
cone penetrometers that incorporate ancillary sensors and algorithms for reducing the influence of tem
perature on CPT results.
1 INTRODUCTION Earlier versions of ISO 22476 and ISO 19901

(ISO 22476:2012 and ISO 19901-8:2014) considered
Detailed requirements for calibration of cone pene- performance specifications by ‘application classes’,
trometers are incorporated in (draft) standards pub- whereby requirements were given for accuracy of in
lished by ISO, International Organization of situ CPT results without detailed step-by-step pro-
Standardization, particularly ISO/DIS 22476-1:2021 cedures or method specifications. Application of
for cone penetration tests (CPT) conducted onshore these performance specifications proved difficult in
and nearshore. The same calibration requirements are practice (Lunne et al. 2017, Peuchen & Parasie
incorporated in ISO/DIS 19901-8:2021 for CPTs in 2019). For this reason, ISO changed to a method
offshore settings. The use of these standards (hereafter specification, particularly providing detailed require-
abbreviated to ISO 22476 and ISO 19901) can provide ments for calibration and verification of cone pene
input for comparisons of CPT systems deployed in trometers in a calibration laboratory. The calibration
practice. In addition, the information available from and verification results provide the required informa
calibration and verification of a particular cone pene- tion for assignment of a cone penetrometer to one of
trometer can provide input in estimation of uncertain- multiple ‘cone penetrometer classes’ specified in
ties of data points in CPT profiles used for design of ISO 22476 and ISO 19901. A further step in method
structures. specification is required: monitoring and logging of
DOI: 10.1201/9781003308829-26
221
acquired test data, followed by assignment of test The primary differences between ISO and ASTM
results in ‘test categories’. are related to (1) detailed requirements for
This paper focusses on background information a calibration laboratory, (2) metrological calculation
about topics considered for development of the ISO of calibration uncertainties and (3) normative (manda
requirements for calibration and verification of cone tory) versus informative (recommended) text. It can
penetrometers. These topics included (1) practical be noted that a calibration laboratory that invested in
and economical test methods achievable in calibration and verification apparatus according to
a calibration laboratory, (2) assessment of differ ISO 22476 and/or ISO 19901 should be able to pro
ences in exposure conditions applied in the calibra vide the normative calibration and verification infor
tion process and site conditions likely to be mation according to both ISO and ASTM. The
encountered during actual cone penetration testing, reverse can require additional investment.
and (3) cone penetrometers that incorporate ancillary
sensors and algorithms for reducing the influence of
2.2 Calibration laboratory
temperature on CPT results.
ISO 22476 is a draft international standard (DIS). It A normative reference to ISO/IEC 17025 provides
was published by ISO in June 2021. Where applicable, the basic requirements for the calibration laboratory.
this paper considers country feedback received for this ISO/IEC 17025 covers laboratory quality manage
DIS. Final published version of ISO 22476 and ISO ment, including detailed reporting requirements for
19901 can differ from the information presented here. calibration certificates.
2.3 Measuring intervals for calibration

2 CALIBRATION AND VERIFICATION
REQUIREMENTS ISO 22476 and ISO 19901 include normative text
with respect to measuring intervals for calibration.
2.1 Overview For example, a minimum range is specified for
inclination. Recommendations (informative) are
Table 1 presents an overview of calibration and veri given for selection of measuring intervals for cali
fication requirements for cone penetrometers. The bration of qc , fs and u.
ISO column refers to both ISO 22476 and ISO
19901; ASTM refers to ASTM D5778-20.
Note that the required presentation of results of the 2.4 Cone resistance and sleeve friction
ISO verifications covers the influences of temperature Calibration for qc and fs requires application of
and bending on the parameters qc , fs , and u. Here, a series of axial loading and unloading series to the
temperature influence is about the internal tempera cone penetrometer, similar to ISO 376:2011.
ture in the cone penetrometer possibly affecting A notable requirement is the logging and supplemen
sensor performance. tary presentation of output values of the penetrom
The ISO reporting requirements for calibration eter other than those for qc or fs, for the purpose of
and verification include assignment of the cone pene checking for any unwanted effects of axial loading
trometer to one of the cone penetrometer classes. on output of sensors other than the one selected for
calibration.
Estimation of calibration uncertainty is prescribed
Table 1. Overview of calibration and verification require in detail. Uncertainties cover those related to: (1) ref
ments for cone penetrometers. erence force, (2) geometry of the cone penetrometer
and (3) uncertainties related to force sensor in the
Item ISO ASTM cone penetrometer, particularly reproducibility,
Calibration laboratory* N I repeatability, resolution, zero drift, interpolation,
Measuring intervals for calibration N N reversibility and apparent load transfer from cone to
Penetrometer geometry* N, C, U N, V friction sleeve (and vice versa). Calculation equa
Cone resistance*, qc N, C, U N, C tions consider standard uncertainties, combined
Sleeve friction*, fs N, C, U N, C standard uncertainties and expanded measurement
Pore pressure*, u N, C, U N, C uncertainties defined according to ISO/IEC Guide
Net area ratios*, a and b N, C I, C 99:2007.
Temperature, T - I, C Table 2 presents a selection of example output
Inclination, i N, C, U I, C according to ISO requirements.
Influence of ambient temperature* N, V N, V
Influence of transient temperature* N, V - 2.5 Pore pressure and net area ratios
Influence of penetrometer bending N, V -
Calibration for pore pressure u takes place with the
C = calibration; I = informative; N = normative; U = uncer cone penetrometer in a pressure vessel. A series of
tainty calculation; V = verification; * = details in sections increasing and decreasing pressure series are
following; - not covered. applied.
222
Table 2. Example summary of laboratory calibration uncertainties for qc , according to ISO requirements.
Fr u1 u2 u3 u4 u5 u6 u7 u8 uc uc udim uc,dim Uqc

[kN] [kN] [kN] [kN] [kN] [kN] [kN] [kN] [kN] [kN] [kPa] [mm2] [kPa] [kPa]
0 0.0150 0.0007 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0150 14.65 0.3024 14.65 29.31
8 0.0390 0.0006 0.0000 0.0000 0.0004 0.0154 0.0344 0.0000 0.0543 52.96 0.3024 53.01 106.02
16 0.0630 0.0012 0.0012 0.0000 0.0008 0.0064 0.0419 0.0000 0.0759 74.11 0.3024 74.25 148.51
24 0.0870 0.0007 0.0007 0.0000 0.0012 0.0026 0.0453 0.0000 0.0982 95.80 0.3024 96.05 192.10
32 0.1110 0.0004 0.0004 0.0000 0.0016 0.0119 0.0468 0.0000 0.1210 118.13 0.3024 118.48 236.97
40 0.1350 0.0013 0.0013 0.0000 0.0020 0.0182 0.0473 0.0000 0.1442 140.76 0.3024 141.23 282.46
48 0.1590 0.0009 0.0009 0.0000 0.0024 0.0195 0.0482 0.0000 0.1673 163.27 0.3024 163.85 327.70
56 0.1830 0.0027 0.0027 0.0000 0.0028 0.0135 0.0488 0.0000 0.1899 185.34 0.3024 186.04 372.07
64 0.2070 0.0032 0.0032 0.0000 0.0032 0.0028 0.0476 0.0000 0.2125 207.35 0.3024 208.17 416.34
72 0.2310 0.0049 0.0049 0.0000 0.0036 0.0159 0.0474 0.0000 0.2364 230.73 0.3024 231.66 463.31
80 0.2550 0.0045 0.0045 0.0000 0.0040 0.0409 0.0000 0.0000 0.2583 252.10 0.3024 253.15 506.31
Fr: reference force; u1: standard uncertainty, reference force; u2: standard uncertainty, reproducibility; u3: standard uncertainty,
repeatability; u4: standard uncertainty, resolution; u5: standard uncertainty, zero drift; u6: standard uncertainty, interpolation;
u7: standard uncertainty, reversibility; u8: standard uncertainty, load transfer; uc: combined standard uncertainty, calibration;;
udim: combined standard uncertainty, cross-sectional area of cone tip; uc,dim: combined standard uncertainty, uc and udim; Uqc:
expanded measurement uncertainty, calibration uc and udim; cross-sectional area of cone tip used for calculation of
uncertainties: 1024.74 mm2
The procedure and presentation of results are gen

erally similar to those for qc and fs . The estimation
of calibration uncertainty considers a reduced
number of uncertainties.
Net area ratios a and b for the cone and the fric
tion sleeve are also determined with the cone pene
trometer in the pressure vessel. Values of a and b
typically show slight pressure dependence. ISO
determines these values at u ¼ 2 MPa.
2.6 Temperature influence

Verification of a cone penetrometer for tempera
ture influence requires two water-filled baths, one
at a temperature of 0 °C and one at 30 °C. The
baths are at atmospheric conditions. A prescribed
alternating sequence applies for immersion of the
cone penetrometer in the two thermostat baths.
The cone penetrometer is thus subjected to
induced change in ambient temperature as well as
transient temperature cycling, see Figure 1. The
‘measured temperature’ shown in Figure 1 refers
to temperature measured by a sensor in the cone
penetrometer.
The verification procedure specifically allows for
optional, explicit correction of force (qc and fs ) and
pressure (u) data for temperature influence. Uncor
rected and corrected results must be reported for this
case. For example, cone penetrometer class 1+ of ISO
22476 includes requirements for incorporation of
a temperature sensor in the cone of the cone penetrom
eter. The temperature (T) data can then be used for
correction of temperature influence on qc , fs and u.
Reporting of results is mainly by key performance
indicators that relate variation of the parameter of
interest (for example qc ) to the temperatures of the Figure 1. Example of verification data for temperature
water baths and to time. influence.
223
3 DISCUSSION the premises is that the standard uncertainty of the
reference, for example the measurement unit for force
3.1 Use of results reference, has much better uncertainty characteristics
than the force sensor of the cone penetrometer. If this
ISO 22476 and ISO 19901 distinguish between (1) is not the case, then the results of uncertainty calcula
records of calibrations and verifications and (2) test tions can be dominated by the uncertainty of the ref
report or calibration certificate. erence and will not necessarily reflect the actual
The records include substantial data files, particu laboratory performance of the cone penetrometer.
larly as they include time-based logging files for This dominating influence can apply to the top end of
a logging frequency of ≥ 1 Hz. The logging applies the cone penetrometer classes, where requirements
to output of each of the primary sensors of the cone for cone penetrometers can approach performance of
penetrometer, for the durations of the various cali commonly available reference measurement units.
bration and verification activities. The records are
retained by the calibration laboratory. Inspection of
the records can provide valuable information for 3.4 Temperature stability of primary sensors
quality management, particularly if the records are Figure 1 includes an example of temperature correc
tracked for multiple calibrations of a single cone tion of qc by post-processing. The approach for fs
penetrometer and if the records are compared for and u would be as for cone resistance.
multiple cone penetrometers. The case of Figure 1 is for a subtraction-type cone
A test report or calibration certificate covers penetrometer equipped with strain-gauge load cells
a summary of the records. The summary is adequate with conventional temperature compensation for ambi
for use in practice, i.e. understanding the performance ent temperature influence. This particular cone pene
of the cone penetrometer at the time of calibration. trometer also includes a temperature sensor within the
cone penetrometer. The acquired records of tempera
3.2 Exposure conditions ture (T) data versus time (t) are additional to the pri
mary CPT parameters and at the same frequency. The
Simulated exposure conditions for a cone penetrom correction method uses a temperature model that math
eter in the calibration laboratory will, inevitably, ematically increases (or reduces) values of qc . This
differ from in situ exposure conditions. Particularly, model is penetrometer-specific and parameter-specific
laboratory calibration and verification consider par (in this case qc ). The temperature model relies on
ticular characteristics in isolation, see Table 1. a polynomial best-fit of qc and T (and their derivatives
Laboratory checks would be challenging for in time), derived from the data recorded in the calibra
assessing the potential influence of in situ exposure tion laboratory.
conditions, such as (1) ambient and induced stress During cone penetration testing, T and qc are
conditions imposed by soil and water, (2) soil dis recorded versus t. The temperature model is subse
placement relative to the cone penetrometer, (3) tem quently applied by post-processing of the complete
perature exposure varying from freezing to, say, 60 ° CPT dataset, such that both uncorrected (raw data)
C, (4) combined and variable axial (compressive and and corrected qc data are retained.
tensile), torsional and moment loading imposed on It can be seen from Figure 1 that significant
the cone penetrometer. Common combinations of reduction of temperature influence can be achieved
these influences cannot be readily quantified in in the laboratory. Robust design of a cone penetrom
a laboratory setting. Robust design and quality moni eter and tested algorithms should also achieve sig
toring of cone penetrometers remains important (e.g. nificant reduction of temperature influence during
Peuchen & Terwindt 2014; Peuchen et al. 2020). actual cone penetration testing under conditions dif
ISO 22476 and ISO 19901 capture quality moni fering from those in the laboratory.
toring by means of test categories, as discussed
above. This is normative. Additional (informative)
guidance and recommendations are also provided. 4 CONCLUSION AND RECOMMENDATIONS
The International Organization of Standardization

3.3 Uncertainty calculations
published (draft) standards ISO/DIS 22476-1:2021
The calibration requirements include prescriptive and ISO/DIS 19901-8:2021. These documents
uncertainty calculations, i.e. a combination of include detailed requirements for calibration and veri
detailed calibration procedures and detailed require fication of piezocone penetrometers. These require
ments for calculation of uncertainties. This approach ments are believed to be practical and economical,
allows easy comparison of cone penetrometers, nevertheless exceed the extent of calibration activities
regardless of manufacturer/ supplier. that represents current (2022) industry practice.
The uncertainty calculation approach presented in It is recommended that parties supplying CPT
the ISO standards includes equations that (1) follow data take note of the implications of the new require
metrological principles (ISO, 2008) and premises and ments. Furthermore, it is recommended that parties
(2) apply to the specified laboratory setting. One of involved in specifying CPTs take note of the
224
opportunities and benefits, notably by means of International Organization for Standardization, 2017. ISO/
appropriate selection of cone penetrometer classes. IEC 17025:2017 General Requirements for the Compe
Standards tend to follow, not lead, technology tence of Testing and Calibration Laboratories. Geneva:
developments or widespread application of ISO.
a particular technology. A technology example for International Organization for Standardization, 2021a. ISO/
CPTs would be the incorporation of ancillary sensors DIS 19901-8:2021 Petroleum and Natural Gas Indus
and algorithms for reducing the influence of tempera tries - Specific Requirements for Offshore Structures –
Part 8: Marine Soil Investigations. Geneva: ISO.
ture on CPT results, now covered by the ISO stand International Organization for Standardization, 2021b. ISO/
ards. The resulting standardisation has influenced the DIS 22476-1:2021 Geotechnical Investigation and Test
cone penetrometer classes introduced in the ISO ing – Field Testing – Part 1: Electrical Cone and Piezo
standards. In turn, the cone penetrometer classes cone Penetration Test. Geneva: ISO.
allow easy comparison of cone penetrometers for use Lunne, T., Santos, R. & Brink Clausen, 2017. Guidelines
in practice, regardless of manufacturer/ supplier. for Use of CPTU Application Classes According to ISO
19901-8: (2014). In: Society For Underwater Technol
ogy Offshore Site Investigation and Geotechnics Com
REFERENCES mittee, ed. Offshore Site Investigation and Geotechnics:
Smarter Solutions for Future Offshore Developments:
ASTM International, 2020. ASTM D5778-20 Standard Test Proceedings of the 8th International Conference 12
Method for Electronic Friction Cone and Piezocone 14 September 2017, Royal Geographical Society,
Penetration Testing of Soils. West Conshohocken: London, UK: Volume 2. London: Society for Under
ASTM International. water Technology, pp. 300–307.
International Organization for Standardization, 2007. ISO/ Peuchen, J., Santos, R., Yetginer, A.G., Eckart, W.S.,
IEC Guide 99:2007 International Vocabulary of Metrol Carrington, T.M. & Lunne, T. 2020. CPT data showing
ogy – Basic and General Concepts and Associated anomalies – assessment and potential postprocessing. In
Terms (VIM). Geneva: ISO. Z. Westgate (ed.). 4th International Symposium on
International Organization for Standardization, 2008. ISO/ Frontiers in Offshore Geotechnics (ISFOG 2020): pro
IEC Guide 98-3:2008 Uncertainty of Measurement – ceedings (pp. 1026–1035). Deep Foundations Institute.
Part 3: Guide to the Expression of Uncertainty in Meas Peuchen, J. & Parasie, N., 2019. Challenges for CPT accur
urement (GU:1995). Geneva: ISO. acy classes. In Proceedings of the XVII European Confer
International Organization for Standardization, 2011. ISO ence on Soil Mechanics and Geotechnical Engineering
376:2011 Metallic Materials – Calibration of Force- ECSMGE 2019: geotechnical engineering foundation of
proving Instruments used for the Verification of Uniaxial the future. https://www.ecsmge-2019.com/uploads/2/1/7/
Testing Machines. Geneva: ISO. 9/21790806/0049-ecsmge-2019_peuchen.pdf
International Organization for Standardization, 2014. ISO/ Peuchen, J. & Terwindt, J., 2014. Introduction to CPT
DIS 19901-8:2021 Petroleum and Natural Gas Indus accuracy. In 3rd International symposium on cone pene
tries - Specific Requirements for Offshore Structures – tration testing CPT14: May 12- 14 (2014). Las Vegas,
Part 8: Marine Soil Investigations. Geneva: ISO. Nevada.
225
© 2022 the Author(s), ISBN 978-1-032-31259-0
Long-term strength determination of frozen soils by CPT

Ivan Sokolov & N.G. Volkov
Fugro, Russian Federation
ABSTRACT: Frozen soils show rheological behavior which results in changing mechanical properties in
time under applied load. It is highly complicated to determine the long-term strength of a frozen soil. Rheo
logical behavior of frozen soils appears in creep or stress relaxation. This paper presents a new method to
measure long-term strength of frozen soils using CPT equipment named Stress Relaxation Test (SRT). Con
ventional tests are focused on the creep behavior by creating constant loads and measuring strains. The SRT is
based on stress relaxation approach by creating constant deformations and measuring stress by a cone. Cone
resistance (qc) would give compression long-term strength, as for sleeve friction (fs) – shear long-term
strength. SRT long-term strength results can be applied directly to pile bearing capacity estimation.
A comparison between such estimation and results from a full scale static pile load test is provided in the
paper.
1 INTRODUCTION happens due to redistribution of elastic and plas

tic deformations in time. The relaxation period is
1.1 Frozen soil rheological behavior one of the most important rheological parameters
of frozen soils. It is important to highlight that
The frozen soil shows a rheological behavior
relaxation period is much shorter (by several
under an excess load from structures. In order to
orders) compared to after-effect period. This is
describe the degree of rheological behavior, the
the main advantage of the proposed new method.
long-term strength is used which is the resistance
of а soil to failure in response to а long-term
load application (Tsytovich, 1975). In other 1.2 S. Vyalov’s logarithmic equation application
words, long-term soil strength corresponds to for frozen soil behavior
a stress at and below which no failure takes
The strength of frozen soils decreases over time as
place within practically observable period of load
a result of the absence of a long-term strength limit
application. Long-term strength of frozen soils is
for ice, which is a permanent and important compo
5 to 15 times less than instantaneous strength, or
nent of frozen soil. An accelerated method of testing
resistance to rapid destruction. The key long-term
soils for long-term strength, using a “dynamometric”
strength characteristics of frozen soil are long
apparatus shortening the test period was invented by
term compressive strength and long-term shear
Vyalov. The design of the apparatus provides the
strength over the freezing surface (Sayles, 1968).
application of the load to a soil sample through
There are two ways to evaluate long-term soil
a dynamometer with fixation of its position and
strength. The first way is to measure deformation
recording of decreasing stress value on it. The loga
under applied constant load which are high enough
rithmic relaxation equation is used to process the
to induce non-attenuating creep which results in
obtained data.
failure with time or, in another words “deformation
Logarithmic equation of long-term strength:
vs time”. When several tests are conducted with
various loads, the curve “load vs time of failure” is
plotted. An asymptote of the curve is interpreted as
the ultimate long-term strength.
The second way is to measure the excess
stress caused by applied constant deformation.
The definition of stress relaxation is given by where τ — measured cone resistance or sleeve
Vyalov (Vyalov, 1986). It is a process of decrease friction in MPa; tp— measurement time, s; β, Т – the
in stress over time, which is necessary to main parameters are temperature and strain rate dependent
tain constant deformation. Stress relaxation (Vyalov, 1986).
DOI: 10.1201/9781003308829-27
226
2 CPT APPLICATION FOR LONG-TERM 2.2 SRT data processing
STRENGTH EVALUATION
To process the SRT results, measurements of stress
relaxation at each cone deployment depth are taken
2.1 CPT in Stress Relaxation Test (SRT) mode into account and processed using the long-term
The principle scheme of the dynamometric method is strength equation (1), reduced to the form y = kx+b.
implemented in the Stress Relaxation Test (SRT) of
frozen soil by CPT with stabilization. When the cone
is deployed on a test depth, the load is kept applied
to the soil, exceeding the value of its instantaneous
strength, but when the cone penetration is paused
like for a dissipation test, the cone is kept loaded where y ¼ σ1 ; x ¼ · lnðtp þ 1Þ; k ¼ 1β ; b ¼ - 1β · ln T
(the rods are clamped) and the stress relaxation is Figure 2 illustrates processing linearized cone
recorded. For each individual SRT, a single value of resistance data for a SRT. The value of the reliability
long-term strength of frozen soil is obtained. To be of the linear approximation of the data in this case is
more specific, for data obtained using a cone resist- 0.997.
ance sensor is the long-term compressive strength of
the frozen soil, and for data obtained using a sleeve
friction sensor, it is the value of the long-term shear
strength. A result of the measurement is the values of
the long-term strength of frozen soil were obtained at
each depth of the cone SRT deployment.
An example of SRT relaxation curve provided
below on Figure 1.
Figure 2. Example of SRT data linearization.
Based on the obtained coefficients k = 0.0125 and

b = 0.025, the coefficients β = 80, Т = 0.139 and the
equation of long-term compressive strength is
defined:
where σс — interpreted as the compressive long

Figure 1. Stress relaxation curve. term strength of the soil.
Equation (3) allows to calculate a long-term
strength value for any period of time. For instance, for
During the stress relaxation test, an increase in 10 days σс = 5.1 MPa and for 100 years σс = 3.4 MPa.
plastic deformation occurs due to a decrease in the Similarly, based on the sleeve friction sensor data,
part of elastic deformation keeping the constant value the obtained value is interpreted as the shear long
of the total deformation. The proportion change term strength of frozen soil.
causes the stress relaxation. To evaluate the value of It is important to set the frame of applicability of
the long-term strength of the soil, the SRT should be Equation (2). The stress relaxation measurements
performed until a certain point of stress reduction in under the tip of the cone were obtained. Successful
the sensor, which presents in the stage II of the relax completion of the test is reached upon condition of
ation curve, Figure 1. In stage I of the relaxation the stage II of the relaxation curve is met. The same
curve, the cone freezes into the soil after penetration situation was observed for the sleeve friction sensor.
pause starts. At this stage besides relaxation some It is assumed that, on the stage I, the cone freezes
creep happens and contributes some inconsistency to into the ground, and only after that rheological
the measured curve. Then the stage II starts when behavior can be clearly observed on the data, which
creep contribution is negligible and the curve starts to represents in data curves (qc and fs) following the
follow the equation (1) (Sokolov, 2020). Equation (2).
227
The field data quality assessment is required to Figure 3 illustrates just 7 iterations of the field
check the applicability of the field data, i.e. if the stage QA approach, but the data acquisition software adds
2 was reached during the test. The field data acquisi a new long-term soil strength value every new
tion software was developed with continuous analysis record of the stress in the sensor (Sokolov, 2020).
array of the obtained data using Equation (2). After the
stress relaxation test starts, Equation (2) is applied for
an interval of the relaxation curve. In the example on 3 FIELD DATA APPLICATION
Figure 3a, an interval of 100 seconds is used for calcu
lation. The data of 600-700 seconds interval has auto 3.1 Pile bearing capacity calculation
matically processed to calculate β, Т factors
(Figure 3b). The pile bearing capacity (PBC) calculation is based
on the long-term soil strength data. The calculation
is based on a similar conventional approach, the
sequential summation of the values of end bearing
and side friction. The difference is that the devel
oped calculation does not use any empirical factors.
The calculation is made according to the formula:
where σс – average compression long-term soil

strength; А — pile area; σs — average shear long
term soil strength; Аaf,i — unit side area of the pile.
The calculation is performed for the recorded
values of σc and σs for each measurement of stress
relaxation test. The test-time interval tp is set as 10
days, in the case of a full-scale pile test under
a static load, and the life-time interval tp is 100 years
to evaluate the ultimate long-term pile bearing
capacity.
Construction site nearby Salekhard was selected
to confirm the SRT based results and PBC calcula
tion in frozen soils. On this site, full-scale static pile
load tests (SPLT) were carried out. SRT was per
formed near the tested pile within 2 meter distance
Figure 3. Field iteration processing of the SRT data. (Volkov et al, 2019).
SPLT was conducted in accordance with GOST
5686-2012. The load was applied in 10 steps, each
Next, the equation of the long-term strength is step lasted 1 day and was equal 5 or 10 tons. The
derived, as shown above for the formula (3), and the last step lasted for 24 hours was FH = 55 tons, the
soil long-term strength (σ) is calculated for lifetime next step FH = 60 tons caused pile failure. GOST
period tp = 100 years in seconds. Figure 3c shows 20522-2012 provides equation (5) to calculation the
values for 100 years (red asterisk), also shows values ultimate long-term pile bearing capacity:
for 10 days (red square) at the end part of the relax
ation curve 600-700 seconds.
This algorithm can be applied to calculate the
long-term strength for any part of the collected data.
Six previous iterations were performed correspond
ingly for the sections of the curve for the 1-100 where FHH – ultimate long-term pile bearing cap
seconds (the First iteration), for the 100-200 seconds acity; FH – step load at which the pile failure
(the Second iteration), etc. and displayed in blue, occurred.
respectively (Figure 3c). Figure 3c shows that, the Comparison of PBC calculation results based on
calculated long-term soil strength values are consist SRT with the results on SPLT in frozen ground pro
ent started from period 400 seconds, and they are vided in the Table 1.
equal to 3.4 MPa for lifetime period tp = 100 years The bearing capacity of a driven pile with length
and 5.1 MPa for the lifetime period tp = 10 days. of 10.6 m in frozen soil with a time interval of 10
Figure 3a shows slow decreasing of the stress days is about 66 tons (663 kN), and the pile failure
value in the sensor, we suggest that the stress relax occurred at load step of 60 tons. The SRT value dif
ation curve had to reach the stage II starting from fers by 10% from SPLT value. These are very close
400 second of the test, and follows Equation (3). results. In turn, the calculated SPLT value for 100
228
Table 1. Comparison of PBC calculation results based on The results clearly show the influence of the ice
SRT and SPLT. content. In lean clay and silty clay σс and σs decrease
if the ice content increases, so as the recommended
Time 10 days 100 years R and Raf. This tendency is confirmed by theoretical
period SRT SPLT SRT SPLT
studies of the ice content influence on the mechan
End bearing 446* — 292 — ical properties of the frozen soils.
Side friction 217 — 117 — The influence of the grain size if other conditions
Ultimate 663 550-600 409 357-390 being equal, can also be observed for silty, fine and
medium sands. An increase of coarse particles gives
* All values are in kN. more resistance to both compression and shear.
4 CONCLUSIONS
years is equal to 39 tons and the calculated SRT
value is about 41 tons (409 kN). The obtained results Stress Relaxation Test (SRT) of frozen soil by CPT
for 100 years correlate to each other relatively good with stabilization was developed based on an acceler
as well. ated laboratory method of testing soils for long-term
strength, using a principle of the “dynamometric”
3.2 Long-term strength for various soils apparatus invented by Vyalov. The logarithmic relax
ation equation is used to process the field data.
The results of σс and σs measurements were col SRT provides individual values of the long-term
lected from the various sites in Russia where perma strength for frozen ground at a certain depth. Data
frost ground was encountered (Sokolov, 2020). The obtained using a cone resistance sensor is interpreted
results were classified with different soil types and as long-term compressive strength of the frozen soil
averaged to a single value. This allowed to compare (σc), and data from sleeve friction sensor – the long
them with the recommended values which are com term shear strength (σs).
monly used to verify the design values, in particular, The results on σc and σs are applicable for calcu
R – pile unit end bearing, Raf – pile unit side friction lation of pile bearing capacity in the frozen soil with
(Aksenov, 2001). The comparison of the values is out empirical factors.
presented in Table 2. The results of σc and σs on the various sites
showed consistency and good relation with the
recommended values for pile unit end bearing and
pile unit side friction.
Table 2. Comparison between long-term soil strength
measured by SRT (σс and σs) and recommended values for
pile unit end bearing (R) and pile unit side friction (Raf). REFERENCES
σс R σs Raf Aksenov, V. I. 2001. Recommendations for determine the
Ice frozen soils strength with a marine type of salinization.
Soil type content kPa kPa kPa kPa FGUP PNIIIS, Moscow. – 42 pages (in Russian).
Sayles, F.H. 1968. Creep of frozen sands, United States
Lean clay None 2430 900-1100 69 40-100 Army Corps of Engineers CRREL.
Low 1335 800-950 35 40-100 Sokolov I. 2020. Determination method for strength prop
Medium 707 400-550 19 40-60 erties of frozen soils by cone penetration testing. PhD
Rich 517 400-550 21 40-60 Thesis – Moscow State University, Moscow, 2020 – 149
pages (in Russian).
Tsytovich, N. A. 1975. The Mechanics of Frozen Ground,
Silty clay None 2228 750-1050 114 40-60
New York, McGraw-Hill, 448 p. ISBN-10: 0070654107
Low 2179 850-1050 53 40-60 Volkov N.G., Sokolov I.S., 2019. Estimation of pile bearing
capacity in permafrost based on stress relaxation meas
Silty sand Low 1963 1000-1800 23 50-130 ured by cone penetration testing. Geotechnics, Vol. XI,
No. 1, pp. 68–78, http://dx.doi.org/10.25296/2221-5514
Fine sand Low 4584 1000-1700 63 50-80 2019-11-1-68-78.
Vyalov, S.S. 1986. Rheological Fundamentals of Soil
Medium sand Low 6977 1500 171 50 Mechanics, Volume 36, 1st Edition. Publisher: Elsevier.
ISBN: 0444600566. 564 pages.
229
© 2022 the Author(s), ISBN 978-1-032-31259-0
Efficiency examined of hands-free Cone Penetration Testing using the

SingleTwist™ with COSON
O. Storteboom & M. Woollard
A.P. van den Berg, Heerenveen, The Netherlands
J. Verhagen
GWR Ingenieursbureau, Rotterdam, The Netherlands
ABSTRACT: The Cone Penetration Test (CPT) cabin as working environment, built on trucks, crawlers or
Track-Trucks®, will more and more develop from a workshop into an office. Before, operators were mainly
dealing with manual operations to keep the production going. Nowadays and in the future, it can be increas
ingly expected that time “on board” will be spend on design or other office-related work next to performing
CPTs. This paper describes a system that has been designed to support this market development. By integrat
ing the patented SingleTwist™ technology in a CPT cabin with the COSON continuous pushing system, an
automatic and hands-free CPT machine is created. The assembly and disassembly of the CPT string takes
place fully automatically. This paper describes the efficiency of the SingleTwist™ technology. The City of
Rotterdam, the first owner of a Track-Truck with COSON-ST, deployed their system in different projects.
One project is chosen to further elaborate on its advantages and experience in practice.
1 INTRODUCTION new technology is preceded by an overview of push

ing systems developed over time and is concluded
Changing requirements and regulations are spurring with a description of a practical application. This
continuous development of CPT systems. The need new insight will ensure a changing working environ
for increasing operational efficiency requires sys ment inside the CPT cabin: a shift from operator to
tems that can start up more quickly and run without engineer or from manual to hands-free operations
further intervention or control. These systems are and from manual labor to data processing, assess
expected to have a simple human-machine-interface ments and design work.
to monitor the process and receive system alerts in
the event of imminent quality loss or maintenance
requirements on the equipment or data acquisition 2 DEVELOPMENT OF CPT PUSHING
system. These expectations focus on obtaining better SYSTEMS
CPT data, performed in a shorter time and with less
effort. Smart and dedicated technology is required to To performing a CPT soil investigation, it is necessary
meet these needs. A combination of digital data pro to push a measuring cone into the soil over time. We
cessing, an accurate system feedback and a simple can see that the development of pushing systems has
and robust design delivering excellent quality, safety, kept pace with that of measuring cones. The first sys
long service life and easy maintenance is needed to tems used in the 1930s were suitable for testing with
face the increasing demands. mechanical cones. The power source used was often
The answer is “back-to-basics”. The CPT rod, the the operator’s muscle strength, a jack was used for
most simple and robust part of the CPT system, has power transmission and the maximum pushing force
become the core of a new development. The CPT was limited to 50 kN. With the arrival of the electrical
rod is the connecting link between the measuring cone in the 1950s, higher demands were also placed on
cone and the pushing system. More than 90 percent the pushing device. Soil investigation at larger depths
of the time required for a cone penetration test con required higher pushing forces. A continuous speed
sists of handling the CPT rods. Far reaching effi also became important and an electrical connection to
ciency improvements can be found in a creative the cone was required to show measurement data in
approach towards these rods. This paper presents the real time. Due to increasing demands for comfort, the
SingleTwist (ST)-technology as the result of such pushing device had to be set up in a closed and condi
a creative design process. The introduction of this tioned cabin.
DOI: 10.1201/9781003308829-28
230
The following subsections describe the pushing 2.1.3 Heavy duty continuous systems
systems developed at A.P. van den Berg for onshore The continuous CPT system (COSON, Figure 2) is
and offshore soil investigation. The specific A.P. van a further development of the above-described sys
den Berg trade names are indicated in brackets. tems with intermittent stroke. In fact, it is a dual
pushing system, built on top of each other. This
2.1 Onshore CPT pushing systems pushing system is equipped with four cooperating
hydraulic cylinders, provided with two independ
The range of pushing systems varies from a single ently moving hydraulic clamps. This allows
lightweight pushing cylinder for hand carried CPT a continuous movement of the CPT string when
solutions to heavy duty pushing devices built on pushed into the soil. The maximum pushing capacity
trucks and on large crawlers. is often limited to 200 kN.
2.1.1 Light weight systems
Lightweight CPT systems (HYSON LW) are very 2.2 Heavy duty automatic systems
suitable for installation in hand carried applications
and on light vehicles such as mini crawlers. The The continuous CPT system as described above was
pushing system has a single hydraulic cylinder with very suitable for a further automation of the hand
a maximum pushing capacity of 100 kN. Due to its ling of CPT rods (AUTOCOSON, Figure 2). This
low weight and small size, it can easily be applied in system consists of a continuous CPT system supple
basements, back yards and on dikes. mented with a carrousel system that automatically
feeds the CPT rods. The cable for data transfer was
replaced by light conductors inside the CPT rods
for wireless data transfer (Optocone). The auto
matic pushing system can run a full cone penetra
tion test without intervention of the operator. As
a result, the tasks of the operator shift to monitoring
and administrative work. In fact, a part of the office
work can be prepared or even performed in the CPT
cabin.
Figure 1. HYSON 100 kN in Mini CPT Crawler (left) and
2.3 Offshore CPT pushing systems

Track-Truck with HYSON 200 kN inside the cabin (right).
Unlike onshore pushing systems which are used on
land or above the water surface, offshore pushing
2.1.2 Heavy duty systems systems are intended to operate under the water
Heavy-duty CPT pushing systems (HYSON) are very level. This imposes special demands on the equip
suitable for installation on crawlers (Figure 1) or inside ment towards reliability and durability and use of
the CPT cabin of a truck or a truck with tracks (Track- material that are suitable for marine conditions. In
Truck®). Stand-alone CPT systems mounted on offshore CPT systems we can distinguish between
a frame can be used as a skid system onshore or on wireline systems for use in drill pipes, and seabed
a barge or jack-up rig for near-shore applications. The systems for performing a CPT from the seabed.
pushing system has two cooperating hydraulic cylin Both systems require specific facilities on board of
ders, which are interconnected by an upper and a lower a ship.
beam. A maximum pushing capacity of 300 kN is feas
ible, but it is often limited to 200 kN.
3 PATENTED SINGLE TWIST™ (ST)
TECHNOLOGY
The fact that being involved with onshore as well as

offshore developments can foster cross-fertilization
is proven by the development of the SingleTwist
(ST) technology. Initially, the ST-technology was
developed for the ROSON seabed CPT system. By
incorporating the patented folding ST-rods in the
seabed system, a compact, safe and easy to handle
ROSON-ST is created. Later it turned out that this
technology is also suitable for onshore CPT, which
has resulted in the development of the COSON-ST.
By integrating the patented ST-technology in a CPT
cabin with the COSON continuous pushing system,
Figure 2. COSON (left) and AUTOCOSON (right) 200 kN an automatic and hands-free CPT machine is
inside cabin in CPT vehicle. created.
231
aided design, strength calculations according to the
finite element method and computer aided manufac
turing. The theoretical specifications have been con
firmed with exhaustive laboratory and field tests.
3.2.1 ST-rod data transfer

A CPT rod not only ensures the transfer of the pushing
and pulling force, but also the transfer of data. Most
common is a connection between cone and data logger
by means of a cable. A cable is very reliable, has
a large bandwidth for data transfer, can be applied both
onshore and offshore, and requires no further support
ing equipment. However, the cable is vulnerable espe
cially at rod connections, which must be taken into
account particularly when used in automated systems.
The light conductor is a proven technology in CPT
Figure 3. ROSON-ST seabed CPT system. and a good option for data transfer. There is no con
tinuous cable running through the rods, which makes
3.1 ST-rods: The foldable CPT string handling a lot easier. This is a useful feature espe
cially in automated systems. However, data transfer
The ST-rod is the smart element that makes up the with light conduction has a limited bandwidth and is
CPT string. In the next subsections the interconnec not very suitable for an offshore environment.
tion of the ST-rods, the transfer of measurement data For data transfer in the ST-system it was decided
through the ST-rods and the storage of ST-rods on to use a 4-core flexible cable with a small diameter.
the Folder are described more in detail (Figure 4). Because only digital cones are used, the number of
cores can be limited to four. The requirement for
applying ST-technology offshore was an important
consideration for choosing this cable. To protect the
cable it is fully integrated into the ST-rods and the
intermediate ball joints at the rod connections. This
minimizes the risk of damage. Furthermore, the
cable must be able to compensate for the change in
length during assembly and release of the ST-rods.
This has been solved by using a flexible spiral cable.
3.2.2 ST-rod storage and maintenance

The foldable ST-string is stored on a reel called the
ST-Folder as shown in Figure 4. The Folder has
a diameter of 2 m and is provided with spacers at the
circumference. This allows the ST-rods to be pre
Figure 4. Folder (left), storage of the ST-rods, Twister cisely positioned in 13 rows. The total storage ca
(middle), to connect the ST-rods.and data transfer (right). pacity of the Folder is an ST string of maximum
70 m length. To easily reel up the string, the ST-rods
3.2 ST-rod connection have a length of 350 mm. The Folder is powered by
an electric motor. This motor is electronically coupled
By using a multiple bayonet thread, a strong and reli to the Sprocket wheel, which feeds the ST-string to
able connection has been invented that can be quickly the pushing system. In this way, the Folder keeps
fastened and loosened. The most striking external pace with the movements of the pusher. To perform
characteristic of the ST-rod connection are three sep mainte-nance, the foldable string can be replaced
arate threaded areas. Each of these areas covers 1/6th completely or par-tially. The foldable string is made
of the circumference of the rod, so that the inner up of 2.8 m sections that contain eight ST-rods. Con
thread of the female part can be directly aligned with nectors on both ends of the cable inside a section
the outer thread of the male part. This important fea allow for the electrical connection between sections.
ture of a bayonet coupling makes it possible to fix the
connection with a very short 1/6 turn. The seven
thread runs in combination with the high quality hard 3.3 Twister on continuous pushing system
ened SS material ensure a strong connection with The Twister (Figure 4) realizes the transition of the
a maximum holding force of 420 kN. At the front foldable ST-string into a solid CPT string. The
end of both the male and female threads a slope is Twister is mounted on the upper beam of the con
provided, so that the rods slide into each other easily. tinuous pusher system (COSON). Both are described
The optimal design has been obtained with computer- in following subsections.
232
3.3.1 Twister 3.4.1 Correct positioning, constant speed and
The purpose of the Twister is to connect and dis smooth twisting of ST-rods
connect the ST-rods. Therefore, the Twister is The electrically driven Sprocket is mounted on the
arranged on top of the upper beam of the pusher. static part of the pusher. There are six guiding blocks
The rod to be tightened is hydraulically clamped on the sprocket that precisely position the rods.
by the Twister. The other rod that is already part Because the Sprocket exerts a constant push or
of the CPT string is held by the COSON clamp pull load on the string, the average Sprocket rotation
of the upper beam. As soon as the rods are cor speed is exactly the same as the pusher speed. During
rectly positioned, an electrically driven rotation of a CPT the pusher moves at a continuous speed of 15
the Twister follows, providing a maximum torque to 25 mm/sec. During retraction, this movement is
of 350 Nm. To ensure a good connection between non-continuous with an average speed of 70 mm/sec.
two tightened rods, the Twister checks both the For tightening the bayonet coupling of the ST-rods,
rotation angle and the torque. At least an angle of it is necessary that the seven thread runs are exactly
50° and a torque of 250 Nm must have been opposite to each other. This is achieved by the constant
reached. The angle and torque are also an indica push load of the Sprocket. The rods to be tightened are
tion of the thread quality. If the angle becomes pushed together, so a connection is established
larger in combination with a lower torque, this properly.
may indicate wear of the thread.
3.5 Operation and HMI
The ST-system enables a fully automatic and hands-
free CPT cycle. The operator just has to bring the ST-
system to the correct condition to start or continue
a cycle. This is done with the clear push/pull and speed
symbols on the left side of the screen (Figure 6). After
positioning, the cycle is started with just one push on
the start button. The effect of an action is made visible
within this image of the system on the right side of the
screen, by changing colours or moving component
parts. All movements of the different clamps, the
pusher, Twister, Sprocket and Folder can be operated
Figure 5. Continuous pushing system (left) and Sprocket,
wheel for connection (right). separately. Important process variables are continu
ously displayed, such as the CPT speed, the angle and
torque of tightening rods, the angular position of the
sprocket wheel and the oil pressure and temperature.
3.3.2 Continuous pushing system Furthermore, it is possible to set alarm values for the
The continuous pushing system (COSON, Figure 5) maximum reachable total push load and inclination of
is used in the ST-concept, because by using the con the cone.
tinuous pusher the Folder can also move continu
ously, which is an easier process to control adding
and removing rods. Furthermore, the small installa
tion height provides sufficient room for a Sprocket
wheel on top, without special roof requirements.
Finally, the rod length of 350 mm is chosen to match
the stroke of the continuous pusher, so each rod is
clamped and (un)tightened at the same position by
the Twister. The COSON continuous pushing system
is a proven technology with a track record of more
than 20 years.
3.4 Sprocket wheel for connection

The Sprocket (Figure 5) is the connecting link
between the Folder and the Twister. It ensures the cor Figure 6. Touch screen for operation and real time presen
rect position of the ST-rods in relation to the Twister. tation CPT data.
It also ensures the correct feeding speed, so that
a smooth cooperation between Folder and Twister is 4 PRACTICAL APPLICATION
guaranteed. Furthermore, the Sprocket ensures that the
rods are pushed together in order to twist them prop The VLG, the Engineering Department of the
erly. In addition, vice versa that the rods are pulled City of Rotterdam procured the first COSON-ST
apart directly after they have been loosened. pushing system built on a Track Truck. Since
233
early 2019 the company in the Netherlands has
used this CPT system successfully in various pro
jects. For example the company executed more
than 300 CPTs on projects in Amsterdam and 32
CPTs and 30 magneto measurements with the
Icone Magneto click-on module on a project in
Rozenburg (the Netherlands). Their typical work
day is filled with two to eight CPTs at multiple
locations. One project was selected to share as
typical example in the following sections.
Compared with more traditional CPT systems,
the VLG confirms that the COSON-ST automated Figure 7. Track-Truck with COSON-ST in the cabin.
system has removed the causes for physical strain.
The operators do not have to lift, guide and screw working environment is ergonomically optimized
CPT rods anymore. Secondly, the ST-system is and the production rate is increased by at least
timesaving. Preparation takes less time, as there is 10%. In order to build new bridges (part of the
no need to raise a hatch and to place an end set Polder path in Schiebroekse polder) in Rotterdam,
including inner casing. The operator just has to pre the CPTs were performed with the Track-Truck
pare and install the Icone. The timesaving in the with COSON-ST (20 ton CPT truck with tracks) as
CPT cycle is mainly achieved during pulling. It is well. A Track-Truck can run on the highway as an
proven that a complete 50 m CPT can be performed ordinary truck with its tracks retracted. After lower
approximately ten minutes faster with the COSON ing the tracks, the Track-Truck does not get stuck in
ST with two hydraulic clamps than with the soft and hilly terrain such as the location where the
HYSON with one clamp. On top of that, after one new bridges will be built. Inside the cabin of the
push on the button, the operators ‘hands are free to Track-Truck the COSON-ST pushing system is
prepare the next test or perform other (engineering) integrated. This matches with the safety, physical
activities. With the application of thicker CPT rods, load and efficiency requirements of the VLG, the
40 mm instead of 36 mm, the risk of buckling and Engineering practice of the City of Rotterdam.
breaking is much smaller and that provides the A digital 15 cm² piezocone (Icone) was used as the
VLG with a lot of confidence. The company has measuring instrument. The modules Seismic, Con
regularly pushed with a total force of more than 15 ductivity, Magneto and Vane could be used along
tons without any problems, whereas with previous with the Icone, but additional parameters were not
CPT systems the VLG was already on guard with requested in this project.
respect to breakage at lower pushing forces. Due to
these rods, the time-consuming use of casing for
4.3 Test method
lateral support is also required less often. This all
results in an average production of 10 CPTs of The bearing capacity of the piles has been deter
40 m in a working day at one project location. On mined with the help of a D-foundation computer pro
longer days with favorable conditions, the company gram. The pile point level must be chosen in such
has even reached 17 of these CPTs. The maximum a way that the requirements of the NEN EN 9997-1
CPT depth of 70 m with VLG‘s CPT string size has +C2:2017 Geotechnical design are met. The piles
also been realized. are calculated for pressure load.
The selected project concerned the geotechnical
investigation required for the design of new bridges
4.4 Test results
that are part of the Polder path in Schiebroekse
polder in Rotterdam, The Netherlands. This is The CPTs performed with the COSON-ST and
a cycle link between the city of Rotterdam and the 15 cm² Icone for the design of new bridges part of
Schie nature reserve. Eight CPTs were performed. the Polder path in Schiebroekse polder in Rotter
dam, The Netherlands, gave a good image of the
soil for the foundation advice. An aquifer was
4.1 Purpose investigation bridges polder path
found at circa 12 m depth with the 6th and 7th
The main purpose of performing Cone Penetration Cone Penetration Test (Figure 8). In this respect,
Tests was to calculate the load capacity for prefab the risk for the development of wells, when digging
concrete piles, for the determination of the pile in the (unfavorable) area at two bridges, has been
lengths and to indicate the preconditions required to checked. This check indicates that there is no burst
prevent the potential bursting. ing risk of the soil when excavating for the abut
ments of these two bridges. The CPTs are used to
calculate the bearing capacity of precast concrete
4.2 Equipment used
piles for a pile level at a depth of 20 m. Of these
The VLG uses the Track-Truck with COSON-ST as indicative calculations is the D-Foundations output
a standard in order to execute CPTs, because the included in Figure 9.
234
system. With a single push of the button, the auto
matic CPT cycle is completed. The CPT string, with
the measuring instrument (cone) fitted at the tip, is
pushed into the soil in one continuous movement.
This string is built up automatically from a reel with
separate, but interconnected, ST-rods.
The so-called COSON-ST has proven to be
a robust CPT pushing system providing reliable data
about the soil and ensuring the operational effi
ciency, which is increasingly expected. The
COSON-ST has more than proven itself and the
VLG feels confident to let the system do its work.
VLG experienced a production rate increase of at
Figure 8. Left: CPT nr. 6 including pore water pressure: least 10%. The system works faster than standard
showing an aquifer between clay layers at 12 m depth. CPT pushers, as it can start up more quickly and
Right: CPT nr. 7 without pore water pressure: showing an runs without further intervention or control. It does
aquifer between clay layers at 12 m depth. not require any manpower during the push/pull cycle
of a CPT. The operator only has to provide the start/
stop signals. Another advantage, which is proven
according to VLG, is that the working environment
is ergonomically optimized, because manual actions
near the moving system are almost unnecessary.
Indeed time “on board” is spent on design or other
office-related work next to performing CPTs. An
additional advantage of the COSON-ST is that the
depth range is increased because of the continuous
CPT push. CPTs are realized faster and in addition
to the already shorter preparation time, this continu
ous movement results in a faster pushing and pulling
process with a higher production rate as
a consequence. Furthermore, the ST-rod is designed
in such a way that the risk of breakage and the asso
ciated downtime and costs are limited.
Figure 9. Output D-Foundations indicative calculations Based on the input from VLG we can conclude
piles bearing capacity.
that the COSON-ST is a welcome contribution to the
expected change of the working environment inside
5 CONCLUSIONS the CPT cabin. A shift from operator to engineer and
from manual operation to data processing and
Cone Penetration Testing (CPT) is a recognized and assessment will occur.
widespread method for efficiently performing in-situ In addition to the onshore application described in
soil surveys, for e.g. foundation advice. Over time, this paper, the ST-technology can also be used off
CPT has been improved by integrating digital elec shore. The ROSON-ST is immediately operational
tronics inside the measuring cone. In addition, the and can perform a CPT fully automatically from the
pushing system needed upgrades, because soil investi seabed in a very safe and timesaving way.
gation is carried out at increasing depths, which
requires higher pushing forces. In addition, develop
ments have also focused on operational efficiency and REFERENCES
an ergonomically optimized work environment. The
CPT cabin as working environment, built on trucks, Brouwer, J.J.M. 2007. In-situ soil testing. Bracknell: IHS
BRE Press.
crawlers or Track-Trucks, will more and more Lunne, T, Robertson, P.K., Powell, J.J.M. 1997. Cone
develop from a workshop into an office. The most Penetration Testing in Geotechnical Practice. London:
prominent development is the ability to perform auto Blackie Academic & Professional.
matic and hands-free CPTs using the patented Single Schnaid, F. 2009. In situ testing in Geomechanics, the main
Twist™- technology with the COSON pushing tests. London, New York: Taylor & Francis.
235
© 2022 the Author(s), ISBN 978-1-032-31259-0
Sustainability in CPT practice: Hybrid CPT Track-Truck

B. Ooms
BAM Infra Nederland, Gouda, The Netherlands
ABSTRACT: Sustainability in construction is of great importance in government projects, but also of

growing concern in the market in general. In this way, earth can be passed on to future generations in
a healthy way. A sustainable realization of construction projects, within strict environmental standards, is
the reason for companies to invest in electrification of equipment, so without carbon and nitrogen emis
sions. In terms of CPT, BAM Infra Nederland has taken the world's first hybrid truck with tracks into use
early 2021. The heavy 4x4 truck is, while at the project location, electrically powered without emitting
harmful substances such as carbon and nitrogen emissions. Both the CPT soil investigation and driving
on the tracks are executed completely electrically. On the road, the truck is powered by the usual diesel
engine. BAM Infra Nederland has now used the hybrid CPT Track-Truck for various projects. In the
paper, experiences from practice are further elaborated and explained.
1 INTRODUCTION next paragraphs the development of onshore CPT

pushing systems are described briefly. In paragraph
Rising prices of fossil fuels, increasing scarcity of 3 the hybrid Track-Truck is explained and in para
crude oil, and government regulations and policies graph 4 experiences from practice are further
against diesel and gasoline powered vehicles owing to elaborated.
surging pollution are major factors changing the pref
erence toward alternative fuel powered vehicles. This
is a factor boosting the hybrid truck market during the 2 DEVELOPMENT OF ONSHORE CPT
coming years. Moreover, evidence suggests that road PUSHING SYSTEMS
transport accounts for more than 50% of the total
health impact of air pollution caused due to emissions Nowadays, the range of onshore CPT pushing
of harmful gases from vehicles. This is prompting systems varies from a single lightweight pushing
governments and car manufacturers to design and cylinder for hand carried CPT solutions to heavy
develop alternative solutions to curb transportation duty pushing devices built on trucks and on
emissions. This has led to the development of hybrid large crawlers. With regard to the heavy duty
technology for automotive, which unlike conventional pushing devices the importance of sustainability
vehicles, does not consume a large amount of fuel and is increasing
emits less carbon. Furthermore, increasing automation
and usage of machinery to simplify manual work in
2.1 Lightweight CPT systems
various industries such as construction, agriculture
transportation, mining, and infrastructure, have fuelled Lightweight CPT systems (Figure 1) are very suitable
the application of hybrid trucks in these sectors. How for installation in hand carried applications and on
ever, the high initial and maintenance costs of hybrid light vehicles such as mini crawlers. The lightweight
trucks is anticipated to restrain the hybrid truck market pushing system has a compact size (165 × 310 mm
(Automotive Team, Hybrid Truck Market, Forecast without anchors) and low weight (only 85 kg) in com
2019 -2027). Besides this development, companies bination with sufficient pushing capacity to perform
more and more want to invest in sustainability. a reliable CPT. The 100 kN lightweight pusher is
Towards the CPT market there is more and designed to perform in-situ geotechnical soil investiga
more a need for sustainable equipment. The hybrid tions at locations inaccessible to other equipment, such
truck with tracks meets this need by performing as basements, back yards, swampy and overgrown
CPT soil investigation in a sustainable way. In the land.
DOI: 10.1201/9781003308829-29
236
The CPT Crawlers (Figure 2) have tracks for better
manoeuvrability in rough or soft terrain. The crawler is
the only option at sites where the maximum allowed
height is limited. Separate carriage is required for
transport of the crawler. The truck with tracks
(Figure 3) is suitable for soft and hilly terrain where
other heavy vehicles get stuck (for more information,
see paragraph 3.1).
Stand-alone CPT systems mounted on a frame can
be used as a skid system onshore or on a barge or
jack-up rig for near-shore applications. The pushing
system has two cooperating hydraulic cylinders,
which are interconnected by an upper and a lower
beam. A maximum pushing capacity of 300 kN is
feasible, but in many cases for reasons of maximum
vehicle weights, this is limited to 200 kN.
The heavy duty pushing system is powered by the
truck PTO or a separate power pack. Hybrid systems
Figure 1. Lightweight CPT system (100 kN pusher). with electrical mode are available since the begin
ning of 2021.
2.3 Heavy duty continuous systems

The continuous CPT system (Figure 8) is a further
development of the above-described systems with
intermittent stroke. In fact, it is a dual pushing
system, built on top of each other. This pushing
system is equipped with four cooperating hydraulic
cylinders, provided with two independently moving
hydraulic clamps. This allows a continuous move
ment of the CPT string when pushed into the soil.
The maximum pushing capacity is often limited to
200 kN.
The continuous pushing system is able to handle
1.000 mm CPT rods. Because of the continuous
movement of the CPT string, the cohesion with the
Figure 2. 100 kN pusher in mini CPT crawler.
surrounding soil is less, so a greater penetration
depth can be achieved with the same pushing force.
2.4 Heavy duty automatic systems

The continuous CPT system as described above was
very suitable for a further automation of the handling
of CPT rods. This system consists of a continuous
CPT system supplemented with a carrousel system
that automatically feeds the CPT rods. The cable for
data transfer was replaced by light conductors inside
the CPT rods for wireless data transfer.
By integrating the patented ST-technology into
a CPT cabin with the continuous pushing system,
a fast CPT machine for hands-free operations is cre
ated (Storteboom O. et al, 2022). On the vehicle, the
workplace of the CPT operator is turning into an
Figure 3. Truck-Truck with 200 kN pusher inside the cabin. office environment. The assembly and disassembly
of the CPT string takes place fully automatically
with a continuous CPT pusher. The working environ
2.2 Heavy duty systems
ment is ergonomically optimized, because less
Heavy-duty CPT pushing systems are suitable for machine work needs to be done. On average the pro
installation on crawlers or inside the CPT cabin of duction rate is increased by 20% compared to
a truck or a truck with tracks. a standard CPT pusher.
237
engine system along with fuel tanks (Figure 5) and
an electric propulsion system, which constitutes of
an electric motor and battery pack. The 4×4 truck
serves as a basis for the truck to which the hybrid
part was added.
4 PRACTICAL APPLICATION
BAM Infra Nederland, part of Royal BAM Group, is

a leading player in the Netherlands in the field of
infrastructure. As a leading company, BAM has for
mulated ambitious goals in the field of sustainability.
These objectives are concentrated around minimiz
ing emissions (up to zero emissions), reducing the
use of primary raw materials, the use of renewable
Figure 4. Automatic CPT System. energy and encouraging circular construction.
BAM wants to build emission-free. In the field of
equipment, BAM is looking for emission-free alter
natives. BAM already introduced the world’s first
electric roller. Beginning of 2021 the hybrid CPT
truck on tracks was added to the sustainable fleet
(Figure 6).
4.1 Minimizing emissions

Conducting soil investigation by a diesel-powered
truck on tracks means that the diesel engine must be
running to provide power to all drive systems and
the CPT measuring system. Although these systems
require little power, the diesel engine continuously
emits combustion gasses and noise. In urban areas,
this is more and more intolerable and limited by
strict environmental requirements.
These restrictive conditions open up an opportun
Figure 5. Small fuel (diesel) tank in the hybrid truck. ity by making soil investigation possible with tar
geted adjustments. To silence the diesel engine,
another energy source must be available. An energy
3 HYBRID TRACK-TRUCK source that is clean at the point of use and produces
virtually no noise emission. The most obvious
choice is the use of an electro motor capable of driv
3.1 CPT Truck with tracks (Track-Truck)
ing the hydraulic pumps. Together with a battery
The Track-Truck was invented by A.P. van den Berg pack with sufficient storage capacity for a full day of
and has a record of accomplishment of more than 30 soil probing, completes this solution.
years. With its tracks retracted, the Track-Truck can
run on the highway as an ordinary truck. After
4.2 New generation CPT Track-Truck
lowering the tracks, the Track-Truck is in its element
in soft and hilly terrain where other heavy vehicles A.P. van den Berg was approached by BAM Infra
are stuck. The track system reduces the wheel load Nederland to develop a CPT Track-Truck as sustain
by 60 to 70%. Manoeuvring the Track-Truck is per able as possible. The solution is a hybrid variant.
formed from the driver’s cabin or with a remote con The heavy 4×4 truck moves over the road with
trol from a safe distance. The penetrometer is easily a clean diesel engine that meets the highest emission
operated via the touch screen, the clear Human- requirements. On location, it is electrically powered
Machine Interface (HMI). A maximum pushing cap to carry out soil investigations without emitting
acity of 200 kN, enables any soil investigation using harmful substances such as carbon and nitrogen. In
the cone penetration method. addition, the hybrid Track-Truck does its heavy
work in a quiet way.
The hybrid system offers the possibility to con
3.2 Hybrid system
duct CPT soil investigations completely emission-
The hybrid truck with tracks integrates free. The Track-Truck is equipped with a 90 kW
a combination of a traditional internal combustion electric motor that drives a second pump set.
238
Figure 6. Hybrid CPT Track-Truck.
The pump set provides hydraulic pressure to drive

the tracks and to perform the CPT soil investigation.
The electric motor is powered by a battery pack with
a capacity of 70 kWh. This power is sufficient to
perform 9 medium CPTs with a CPT depth of 25
meters and to drive the Track-Truck on the tracks for
about 60 minutes. On site it is possible to change the
power source for the drive from diesel to electric.
With electric drive, the diesel engine can be switched
off. In some cases the electric drive is stopped, then Figure 8. 200 kN continuous pushing system inside cabin
the diesel engine will start and automatically switch of hybrid CPT vehicle.
from electric to diesel. This transition is so smooth
that the driven process is not affected by this. If no
power is drawn during electric drive, the electric 4.3 Equipment used
motor switches to an energy-saving mode, where the
speed is reduced to 500 rpm. As soon as power is The digital cone (Figure 7) measures the four
required, the electric motor revs up and energy is standard parameters: cone tip resistance (qc),
almost immediately available. If no power is drawn sleeve friction (fs), pore water pressure (u) and
for a longer period of time, the electric motor inclination (Ix/y). The digital cone is mechanic
switches off manually to save energy. The batteries ally 40% stronger than its analogue predecessor
for the hybrid system (see Figure 11) can be charged and is at the same time more accurate, more
at a charging station or with a 3-phase connection reliable and easier to maintain. Calibration data
with a maximum power of 22 kW. In order to fully is stored in the cone itself, so separate data
charge the battery four to five hours is required. In holders such as USB sticks are no longer
addition, a small fuel tank is installed in order to necessary. Digital cones come with a pre
drive from and to locations. pressure and seal assembly that reduces cone
maintenance.
The continuous pushing system is an advanced
CPT system for performing CPTs for soil
research. The continuous pushing system is
able to push a CPT probe into the soil with
a continuous movement and with a maximum cap
acity of 200 kN (Figure 8). The truck on tracks is
a standard 4x4 truck with tracks between the front
and rear axle. This provides the Track-Truck with
a large load capacity for moving on soft subsoil.
The Track-Truck is equipped with a hybrid drive.
This means that all functions, except truck driv
ing, can be powered by the diesel engine of the
truck or by an electric motor. The electric motor
is powered by a battery pack with a capacity of
70 kWh. The hybrid Track-Truck is capable of
emission-free CPT and track driving, meeting
Figure 7. Digital cone. a zero emissions requirement on site.
239
4.4 User experience
The hybrid CPT Track-Truck has been used for more
than six months by BAM Infra Nederland (Figure 9).
Figure 9. Example of a project in which the hybrid

Track-Truck is used.
4.4.1 Test results

The CPT in Figure 10 is an examples of the many
CPTs which the BAM has performed in the centre of
Amsterdam (The Netherlands). In this specific project
13 CPTs were performed up to a depth of 68m below
ground level. The hybrid Track-Truck can enter the
centre of Amsterdam without a special environmental
exemption. Furthermore it was experienced as much
cleaner and quieter when performing the CPTs.
4.4.2 Savings
In the first six months the hybrid Track-Truck has
used 3,000 liter of diesel. The traditional Track-
Truck used 9,000 liter of diesel, so in total 6.000
liter (67%) is saved. The Hybrid Track-Truck saves
on average fifty liter per day, or saves approximately
32,000 kilograms of CO2 per year.
4.4.3 Sound level

In practice the engine of the Track-Truck is only used
on the trips to and from the location. The fact that the
engine no longer has to run during CPTs and when Figure 10. CPT with the hybrid Track-Truck in the centre
of Amsterdam.
moving the tracks lowers the sound level considerably.
When moving on tracks and during CPTs it
hardly makes any noise when you are outside the to the traditional CPTs. There is no difference in
cabin. Inside the cabin a lower noise level is experi pushing capacity, CPT velocity and fast pulling the
enced of less than 67 dB(A). This could be lowered probe out of the soil. In general, no quality loss is
even more when the insulation is upgraded and spe experienced when performing the CPT soil investi
cial damping techniques are used. gation fully electrical.
4.4.4 CPT soil investigation 4.4.5 Hybrid battery

The experience regarding the CPT investigation is The hybrid battery (Figure 11) and drive are charac
that the CPT soil investigation results are similar terized by high operational reliability, ample drive
240
power and a long service life of 25 years. The 5 CONCLUSIONS
safety for users is increased due to less noise and
no emission of harmful gases. During times when A sustainable realization of construction pro
no power is required from the system, significant jects, within strict environmental standards, is
savings in fuel consumption are achieved. The the reason for companies to invest in electrifica
electrical system has a modular structure of com tion of equipment, so without carbon and nitro
ponents, which means that faults can be solved gen emissions.
quickly and service can be provided easily. The In terms of CPT, the hybrid Track-Truck is,
batteries are charged with a max. capacity of 22 while at the project location, electrically
kW via a wall socket or a public charging sta powered without emitting harmful substances
tions. Experience from the practice show that you such as carbon and nitrogen emissions. Both the
can work 1.5 day on a full battery charge. CPT soil investigation and driving on the tracks
are executed completely electrically. On the
road, the truck is powered by the usual diesel
engine.
The hybrid Track-Truck was launched early 2021.
Experiences from practice show that the fuel con
sumption is lowered by 67%. On daily basis fifty
liter is saved and approximately 32,000 kilograms of
CO2 per yearly basis.
The tracks are electrically driven and therefore the
sound level is lower compared to fuel driven Track-
Trucks. With a full battery charge CPTs can be per
formed during 1.5 day before the battery needs to be
recharged. The experience with regard to CPT soil
investigations are more or less similar compared to
traditional CPTs.
REFERENCES
Automotive Team Hybrid Truck Market - Global Industry
Analysis, Size, Share, Growth, Trends, and Forecast
2019 – 2027
Brouwer, J.J.M. 2007. In-situ soil testing. Bracknell: IHS
BRE Press.
Lunne, T, Robertson, P.K., Powell, J.J.M. 1997. Cone
Penetration Testing in Geotechnical Practice. London:
Blackie Academic & Professional.
Storteboom, O, Woollard, M. 2022. Efficiency examined of
hands-free Cone Penetration Testing using the Single-
Figure 11. Batteries for the hybrid system. Twist with COSON’.
241
© 2022 the Author(s), ISBN 978-1-032-31259-0
Elastic soil properties investigated using seismic tests to complement the CPT
J.L. Rangel-Núñez
UAM-Azc and Ingeum Ingeniería, Mexico
ABSTRACT: Seismic tests are performed to investigate the elastic properties of the soil. Elastic soil proper
ties are essential input for the prediction of soil-surface motions related to earthquake excitations and for the
design of foundations for vibrating equipment, for the assessment of offshore structure behavior during wave
loading and to predict deformations around excavations. To investigate these elastic properties of the soil,
seismic tests can be performed. The digital Icone is easily extendable by click-on modules to measure add
itional CPT parameters. The Icone Seismic module is the instrument to investigate the elastic soil properties.
This click-on module is automatically recognized, creating a true plug & play system. It contains three accel
erometers to receive left and right shear waves as well as compression waves. The Icone Seismic system can
be applied both onshore and offshore and has a 1,000 m water depth rating. In this paper the Icone Seismic
system is described. Feedback from onshore fieldwork with the Icone Seismic will highlight the user experi
ence with this approach.
1 INTRODUCTION 2 ICONE AND CLICK-ON MODULES
Elastic soil properties are essential input for the pre Since Cone Penetration Testing (CPT) is used for soil
diction of ground-surface motions related to earth investigation, a tremendous development has occurred
quake excitation and for the design of foundations in the techniques to measure soil parameters. After
for vibrating equipment, for the assessment of off CPT with the mechanical cone had proven to be very
shore structure behavior during wave loading and to useful, the development of the electrical cone brought
predict deformations around excavations. To investi a big step forward in ease of use and accuracy. Now
gate these elastic properties of the soil, seismic tests adays the advantages of digital technology are avail
can be performed by means of the Icone Seismic able for further improvements.
system.
The seismic piezocone penetration test (SCPTu)
2.1 Digital cone
provides multipoint simultaneous measurement of
tip resistance (qT), sleeve friction (fs), pore pressure The digital cone or Icone measures the four standard
(u2), and compressional and shear wave velocities parameters: cone tip resistance (qc), sleeve friction
(Vp and Vs, respectively) following the down-hole (fs), pore water pressure (u) and inclination (Ix/y).
geophysical testing strategy, but without using pre The Icone is mechanically 40% stronger than its ana
drilling. This type of test has been successfully logue predecessor and is at the same time more
applied in geotechnical exploration for more than accurate, more reliable and easier to maintain. Cali
three decades (Lunne et al., 1997; Mayne & Campa bration data is stored in the cone itself, so separate
nella, 2005). Improvements have now been achieved data holders such as USB sticks are no longer neces
such as using dual seismic sensors to detail the stra sary. Icones come with a pre-pressure and seal
tigraphy and to perform continuous-interval seismic assembly that reduces cone maintenance.
piezocone tests by semi-continuous wave generation
using the autoseis device at the rate of 1 strike/
2.2 Icone and click-on modules
second (Mayne & Woeller, 2015).
In the next paragraphs, the various Icone click-on In the past years, several click-on modules for the
modules are described with a focus on the seismic Icone were developed and the seismic module is part
module. of it. In this chapter, the following three are
DOI: 10.1201/9781003308829-30
242
reinforced foundation piles. Power supply cables and
above structures have in common that they affect the
earth’s magnetic field. Using the Magneto module,
metal objects in the underground can be detected by
interpreting anomalies of the earth’s magnetic field.
Figure 1. Icone data acquisition concept.
described: the conductivity module (Figure 2), the

magnetometer module (Figure 3) and the vane
module (Figure 4). The seismic module is described Figure 3. Magneto module with 10 cm² Icone.
in paragraph 3. All modules except the Icone Vane
can be used with a 10 cm² and a 15 cm² Icone. When 2.5 Vane module
CPT-data is not required, the click-on modules can
also be used with a dummy tip instead. In soft fine-grained soils, the field vane test (FVT) is
a popular and effective complement to the CPT. The
vane test allows in-situ measurement of the peak,
2.3 Conductivity module residual and remolded undrained shear strength of
The measurement of electrical conductivity in the soft fine-grained materials, such as clay, silt and
subsoil is a function of both the conductivity of the mine tailings. The Icone Vane is a digital vane with
pore water and the soil particles, the first being the a downhole motor and torque cell that is used for in-
dominant factor. With the Conductivity module situ measurement and evaluation of shear strength
changes in the concentration of (dissolved) electro that can be deployed in soft soils where a prediction
lytes are determined without specifying the exact of the undrained shear strength is required.
nature of these electrolytes. Therefore, the module
facilitates separation of zones with differentiated
water content, including determining the water table
depth and the thickness of the capillary zone or sep
aration of fresh and salt water carrying soil layers.
Another very important application of the conductiv
ity module is detection of (the degree of) contamin
ation in a soil body. Further soil investigation should
provide details on the actual contaminants.
Figure 4. Icone Vane (without protection tube).
3 SEISMIC MODULE
Figure 2. Conductivity module with 10 cm² Icone.
2.4 Magneto module

Unknown structures and obstacles, like unexploded
ordnance (UXO), are a risk factor in the execution of
earthworks. To avoid risks of damage and interruptions
of work, these underground elements must be identi
fied and mapped. Most underground structures contain
metal such as sheet-piles, ground anchors and pipelines
or a combination of metal and concrete, such as Figure 5. Seismic module with 10 cm² Icone.
243
Seismic tests are performed to investigate the elastic is based on digital data transfer. The system consists
properties of the soil. For this purpose a shear wave (S) of a digital data logger, called Icontrol, and the Ifield
or a compression wave (P) is guided into the soil. Elas software for real-time data presentation. Due to
tic soil properties are essential input for prediction of application of digital technology, multiple param
ground-surface motions related to earthquake excitation eters can be retrieved by several modules. Upon con
and for assessment of: foundation design for vibrating nection, the Icontrol data logger and Ifield software
equipment, offshore structure behavior during wave will automatically recognize the specific cone and/or
loading and deformations around excavations. module, so it is a true plug & play system.
The output signals from the seismic sensors are
being digitized inside the seismic module and from
3.1 Principles here transferred to the Icontrol data logger at surface
Elastic soil parameters are determined by measuring the level. After all tests have been performed, the data
propagation speed of an applied sound wave between obtained are then analyzed offline by processing
two known depths. Mostly this is done by pushing the software, determining the propagation speed and
seismic module into the soil and stopping at 1-meter corresponding elastic soil parameters for all investi
intervals. During the pause in penetration, a shear or gated depth ranges.
compression wave is generated at surface level and the
time required for the wave to reach the seismic sensors
is recorded. The time difference between two consecu 4 PRACTICAL EXPERIENCES
tive seismic tests performed is a measure of the elastic
properties of the soil. The following can be calculated as Determination of shear wave velocity profiles has
a function of the seismic data (shear and compressional been applied to define the stratigraphic profile and
wave velocities, Vs and Vp, respectively) and the soil to evaluate site effects, i.e. dynamic amplification
density (ρ, usually already known), eqs 1 to 6: small or liquefaction, however, its field of application
strain shear modulus (Gmax), constrained modulus (M), has now been extended. In fact, it is also applied
Poisson’s ratio (), elasticity modulus (E), bulk modulus in arability studies, for the design and control of
(κ) and Lame’s constant (λ). soil treatment, and lately in the determination of
the parameters of advanced constitutive equations.
Therefore, it is common in an exploration cam
paign to determine such a profile, but this is not
trivial since there are different methods and, in
principle, it is necessary to define which one to
use. The definition of the method depends on sev
eral factors, namely, experience and mastery of the
proper execution and interpretation of the method,
the alteration and deformation induced in the sub
soil, the degree of heterogeneity, to mention the
most important ones. For example, although the
seismic refraction method is easy to perform and
is a non-invasive technique, its interpretation is
not easy, mainly when velocity inversion is pre
sent in the subsoil; unfortunately, in geotechnical
problems this condition is frequent. Regarding the
down-hole, up-hole and cross-hole techniques, it is
necessary to take care not to induce an excessive
alteration in the subsoil when the borehole is
drilled. With respect to the techniques based on
An even faster and more accurate way is to use two the measurement of surface waves, the problem is
seismic modules, which are mounted, at a fixed dis that the solution is not unique, so the method is
tance of exactly 0.5 or 1 meter. Since the time differ very dependent on the experience and the geotech
ence between two consecutive measurements is nical information available.
approximately 2 ms, a very consistent measurement With respect to the seismic piezocone test
of the trigger signal is required. This requirement is (sCPTu), it has the advantage that the alteration gen
met by using the same high sensitive sensors for the erated in the subsurface is less than that produced
trigger module and by placing this module in the with the conventional down-hole technique or with
immediate vicinity of the hammer. the suspension-logging test (SLT), since in the latter
it is necessary to drill a borehole, where a relaxation
3.2 Data processing and visualizing is generated and a cake is created around the
borehole.
Both the Icone Seismic Module as well as the Icone The following are five projects where the shear
are part of our modular data acquisition concept that wave profile was determined. Two of them were
244
carried out in Mexico City (CDMX) for the design a basement. The foundation solution proposed for
of building foundations, a third for the design of the the tower consists of piles at a depth of 31 m.
liquefiable soil deposit improvement system and the Purpose investigation. The main purposes in
remaining two were for the study of tailings dams. using the seismic piezocone at this site, was to:
HYSON 100kN installed on a 14t Mercedes-
Unimog truck was used in all projects (Figure 6). • Determine the soil stratification and dynamic
A 10 cm2 piezocone was used with the seismic properties.
module (Figure 5). The wave generation system was • Estimate the capacity of piles.
• Determine site spectrum.
by means of the lateral strike of a beam.
Figure 6. Equipment used for secondary wave profile

determination.
4.1 Building design in CDMX

The urban area of Mexico valley can be divided in Figure 7. Primary parameters determined with the sCPTu.
three main geotechnical zones: Foothills (Zone I),
Transition (Zone II) and Lake (Zone III). In the foot
hills, or hills, very compact but heterogeneous vol
canic soils (i.e. tuffs) and lava are found. These
materials contrast with the highly compressible soft
soils of the Lake Zone. Generally, in between,
a Transition Zone is found where clayey layers of
lacustrine origin alternate with sandy alluvial
deposits erratically distributed. Mexico City is
mainly located in the lake zone.
Very soft soil deposits of the lake area, whose
average thickness is 30 m, are mainly clayey with
alternating hard lenses/layers (a superficial crust and
ash, silty sand lenses or thin layers), and they are
characterized by an high humidity (100<wn<900%),
high plasticity index, (100<PI<500), large one-
dimensional compressibility (Cc, max≃10) and low
shear wave velocity (Vs<100m/s). Those unusual
values are produced mainly by the clay microstruc
ture and composition, and the high diatoms content
(Rangel et al. 2014).
There are two important factors that affect the
design of the foundation of a building in the lake and
transition zones of Mexico City, where almost 80% of
the buildings are located: earthquakes and regional sub
sidence. Regarding the first aspect, Mexico City has Figure 8. S-wave velocity profile determined with different
a high dynamic amplification coefficient, max 25 techniques: sCPTu, sDMT, SLT, MASW and Ovando’s
(Pérez Rocha et al, 2009), therefore, it is necessary to correlation.
know the velocity law of the subsoil, at least in the
strata that amplify the seismic wave, that is, up to 30 Results. Because this building is located in an
and 40 m depth. area with significant seismic amplification, part of
the exploration campaign was directed to determine
4.1.1 Building project 1 the velocity profile of the secondary wave in order to
The construction of a hotel in the lake area of design its foundation. Figure 7 shows a detail of the
Mexico City is contemplated, which consists of stratigraphy detected with the sCPTu sounding,
a 14-story tower and low-rise buildings, both with which is typical of the Mexico City Lake area, and
245
Figure 8 shows the secondary wave velocity profile on the soil characterization and the design earth
obtained with different techniques: seismic piezocone quake, an initial liquefaction analysis was carried out
(sCPTu), seismic dilatometer (sDMT), suspension log to corroborate the liquefaction potential and based on
ging Test (SLT), multi-channel analysis of surface the S-wave velocity profile, the target for the design
waves (MASW) and an empirical correlation applicable of the improvement was established, in other words,
to CDMX deposits (Ovando et. al., 2007). It is observed that profile where liquefaction no longer occurs.
that all the techniques produce similar results, although
the values determined with MASW are almost double
those determined with borehole techniques. Likewise, it
can be seen that the values obtained with SLT were
lower, possibly attributable to the greater alteration that
occurs in the soil when drilling, and in general the
values determined with sCPTu, sDMT and using Ovan
do’s correlation were very similar.
4.1.2 Building project 2

The project considers the construction of a shopping
center and four towers (14, 18, 20 and 22 levels),
three for residential use and one for office use, with
shared parking basements on a large basement,
whose maximum general excavation level is
-16.00 m. The project is located in the lake area. Figure 9. Geotechnical model for design of the site 2.
Purpose investigation. The main purposes in using
the seismic piezocone at this site, was to:
• Determine the soil stratification and dynamic
properties.
• Estimate the capacity of piles.
• Determine site spectrum.
Results. An extensive exploration campaign was car
ried out including the following techniques: sCPTu,
sDMT, Vane Shear Test (VST), Pressuremeter
(PMT), Phicometer (PHI), undisturbed sampling and
laboratory tests. The geotechnical design model is
presented in Figure 9. For the short-term shear
Figure 10. Stratigraphic profile for the subsoil improve
strength (su), the results of the VST and TX-UU ment zone.
tests were used for the soft soils and the PMT and
PHI for the hard soils; for the shaft strength of the
piles, the fs (sCPTu test) and the residual strength 4.3 Stratigraphic conditions in tailings dams
determined with VST were used; the piezometric
conditions were evaluated from CPTu–Δu; the com In the case of tailings dams, where there are very
pressibility moduli to determine the soil long-term young and heterogeneous deposits, exploration cam
behavior were obtained from piezocone correlations paigns have been carried out to determine the strati
and oedometer tests; and finally, the dynamic graphic profile and mechanical parameters with
response was studied from the velocity profile in situ tests, since obtaining unaltered samples is dif
obtained with the sCPTu. ficult in most cases. It highlights the importance of
shear wave profiles in conjunction with vane and
CPTu_Δu tests, in order to carry out dam stability
4.2 Subsoil improvement design analyses and identify potentially unstable strata
An oil refinery will be built in southern Mexico. Due (determination of Go, su and k). The results obtained
to the high seismicity of the area and the existence of in two mines in Mexico are presented below (Fig
deposits with alluvial and fluvial soil intercalations ures 11 and 12). It is observed that in all cases the
below the groundwater level, the liquefaction poten values of the tip resistance qT oscillate strongly and
tial is high, for which an improvement of the subsoil this gives an idea of the heterogeneity that exists in
is pro-posed, so an exploration campaign has been this type of deposits. This behavior is also observed
proposed which includes the determination of in the shear velocity profiles, so it is necessary to try
S-wave velocity profiles in order to design such to decrease the distance at which the determination
improvement. Figure 10 shows a sounding showing of velocity S-wave is made. Also, it is observed that
typical conditions in the project, where intercalations sometimes the value of u2 is negative, which implies
between soft and frictional soils are observed. Based the development of suction. When observing the
results of the dissipation tests, it is obtained that the
246
curves differ from those obtained in natural soils, 5 CONCLUSIONS
where it is difficult to obtain the initial value of the
pore pressure. All these aspects make it difficult to Cone Penetration Testing, as a technique for in-situ soil
define the stratigraphic profile and to determine the investigation, is a recognized and widespread method
properties in the field, which is why conducting an for efficiently performing of soil surveys. CPT is in the
exploration campaign in tailings dams is currently course of time continuously improved like developments
a challenge. concerning the application of digital electronics inside
An alternative to determining the velocity profile is the cone. The most prominent of these is the ability to
through correlations. Figure 13 shows the velocity easily extend the digital Icone by click-on modules to
profiles obtained from the sCPTu measurement and measure additional parameters. Any module is automat
the one determined by means of qT vs Vs correlations, ically recognized by the Icontrol data logger, creating
although the correlation provides adequate values, it a true plug & play system.
does not capture the peaks. To investigate elastic properties of the soil, seismic
tests can be performed by means of the Icone Seismic
system. Four case studies are described where the
s-wave velocity profile is determined with the seismic
piezocone for the following purposes: foundation
design, ground improvement design and tailings dam
studies. In the case of foundation design, the velocity
profile is used to evaluate the amplification of clay
deposits. This is determined with several techniques and
it is observed that the procedure with seismic piezocone
produces satisfactory results due to the low level of
deformation induced during its execution. It is also
shown that by combining the sCPTu–Δu techniques with
Figure 11. Results of the sCPTu–Δu tests for the case of VST it is possible to determine parameters of the geo
tailings dam 1. technical model for foundation design.
In the case of soil improvement design, based on
the velocity profile determined with the seismic
piezocone, the liquefaction potential of a soil deposit
is evaluated, and then a target velocity profile is pro
posed for the improvement that no longer generates
liquefaction. Finally, examples are presented where
the velocity profile is determined for the case of tail
ings dams in Mexico using correlations and those
determined by seismic piezocone, and it is observed
that the profiles obtained with correlations are not
able to capture the maximum points of the profile.
The seismic module has proven to be an accurate
Figure 12. Results of sCPTu tests for the case of tailings and reliable measuring device, which can be used for
dam 2.
different measuring techniques for soil analysis.
REFERENCES
Lunne, T, Robertson, P.K., Powell, J.J.M. 1997. Cone
Penetration Testing in Geotechnical Practice. London:
Blackie Academic & Professional.
Mayne, P.W & G. Campanella. 2005. “Versatile Site Character
ization by Seismic Piezocone”. Proc. 16th Int. Conf. on Soil
Mechnc. & Geo. tech Engrg., Osaka, Japan, 2, pp. 721–724.
Mayne P.W. & D.J. Woeller. 2015. “Advances in seismic
piezocone testing”, Geotechnical Engineering for Infra
structure and Development: XVI European Conference
on Soil Mechanics and Geotechnical Engineering, ICE,
ISBN 978-0-7277-6067-8.
Ovando-Shelley E., Ossa A, & Romo M. 2007. “The sink
ing of Mexico City: Its effects on soil properties and
seismic response”, Soil Dynamics and Earthquake
Engineering 27: 333–343
Figure 13. Comparison between the s-wave velocity pro Pérez Rocha L.E., M. Ordaz, E. Reinoso. 2009. “Interpola
files obtained with the sCPTu and by correlations with qT, ción espacial de las amplificaciones dinámicas del ter
for the sCPTu-2 borehole in tailings dam 1. reno”, Ingeniería Sísmica, 80, pp. 85–112.
247
© 2022 the Author(s), ISBN 978-1-032-31259-0
Tensiocone: A cone penetrometer with the facility to measure negative

pore-water pressure
A. Tarantino & A. Capotosto
University of Strathclyde, Scotland, UK
F. Bottaro & M. Bellio

Tecnopenta S.r.l., Teolo (PD), Italy
D. Gallipoli
Università degli Studi di Genova, Italy
ABSTRACT: CPT sounding initially passes through the vadose zone, that is the zone above the phreatic sur
face where pore-water pressure is negative and degree of saturation is usually lower than unity. Negative pore-
water pressure (suction) significantly affects tip resistance and sleeve friction and lack of knowledge of soil
suction in this zone makes CPT data difficult if not impossible to interpret. For the case of Piezocone Cone
Penetration Test (CPTU), crossing the vadose zone also exposes the cone tip porous filter to desaturation jeop
ardising the measurement of positive pore-water pressure once the cone penetrates the saturated zone below the
phreatic surface. This paper presents the concept of the Tensiocone, a cone penetrometer with the facility to
measure pore-water pressure in both negative and positive range. A first prototype was tested in the laboratory
and in the field to investigate one of the major challenges in Tensiocone measurement, that is whether adequate
contact can be established between the ground and the tensiometer porous filter during penetration.
1 INTRODUCTION in calibration chambers on unsaturated soils (by con

trolling or monitoring suction). This has led to a new
The Cone Penetration Test (CPT) is a valuable device generation of semi-empirical equations where soil
for characterising soil stratigraphy and estimating soil properties can be inferred from tip resistance and
mechanical properties. The penetrometer is inserted sleeve friction by explicitly considering the influence
from the ground surface and initially crosses the of suction. However, the use of these equations in rou
vadose zone, that is the zone above the phreatic sur tine applications remains a challenge due to the diffi
face where pore-water pressure is negative (suction) culty of measuring suction in the field at depths greater
and degree of saturation is usually lower than unity. than 1-2m.
This vadose zone can extend to several meters. The most obvious approach to measure suction
Strength and stiffness of the ground in the vadose during penetration is to provide the penetrometer with
zone are significantly affected by suction variations a High-Capacity Tensiometer to measure pore-water
as a result of rainfall and evapotranspiration and so pressure in the negative range (Tarantino 2002; Mar
do the tip resistance and sleeve friction. Empirical inho et al. 2008). A HCT can also measure pore-water
equations used for characterising soil properties pressure in the positive range and this would enable
from CPT profiles have been developed from tests CPT measurements also in the saturated zone below
conducted in the calibration chamber on soils in dry the phreatic surface. The HCT could offer more accur
or saturated states. These equations clearly do not ate measurements than Piezocone Cone Penetration
hold for unsaturated soils. Data acquired when cross Test (CPTU) because it would not suffer from desat
ing the vadose zone are therefore difficult if not uration when passing through the unsaturated vadose
impossible to interpret in routine practice. zone (Mondelli et al. 2009; Sandven 2010).
In recent years, the research groups at the Univer This paper presents the development a new cone
sity of New South Wales in Australia (Pournaghiazar penetrometer named ‘Tensiocone’, which incorporates
et al. 2012; Yang & Russell 2015, 2016) and the Uni a high-capacity tensiometer for the measurement of
versity of Oklahoma in the US (Miller et al. 2018; pore-water tension during cone penetration. The major
Miller & Collins 2019) have conducted extensive challenge with this design is to ensure proper contact
experimental campaigns where CPTs where performed between the high air-entry porous filter and the
DOI: 10.1201/9781003308829-31
248
ground. In laboratory measurements using standing-
alone high-capacity tensiometers, a soil paste is inter
posed between the high-capacity tensiometer and the
soil to ensure continuity between water in the soil
pores and water in the tensiometer water reservoir.
This paste cannot be used on the HCT installed on the
penetrometer as it would be taken off as soon as the
HCT touches the ground.
Figure 2. High-Capacity Tensiometer (HCT) installed in

2 TENSIOCONE CONCEPT AND DESIGN OF the cone penetrometer shaft.
FIRST PROTOTYPE
filter (800 kPa) was cut and glued into the HCT
The concept of the Tensiocone is illustrated in using epoxy. The same epoxy was also initially used
Figure 1. An adaptor is placed between the friction to fix the tensiometer to the penetrometer adaptor.
sleeve and the cone tip and incorporates a High-
Capacity Tensiometer (HCT) (Tarantino & Mongioví,
2002). The first prototype presented in this paper was 2.2 Reliability of HCT measurement
designed without the cone tip and friction sleeve since The performance of two tensiocone HCTs was initially
the aim was to test the pore-water pressure measure tested via simultaneous measurement on a kaolin
ment. However, the same adaptor could be easily sample compacted to 300 kPa vertical stress at 30%
incorporated into a fully functional cone penetrometer. water content. Measurement was carried out in the suc
The measurement of pore-water pressure is car tion box described in Tarantino & Mongioví (2003).
ried out laterally to minimise the stresses on the The box lid has two holes equipped with O-rings to
porous filter compared to the case of a porous filter install the two HCTs. O-rings are also interposed
located on the penetrometer tip. The major challenge between the annular cylinder hosting the sample and
with this new instrument is to ensure proper contact the base and lid respectively to prevent evaporation
between the porous ceramic filter and the ground. from inside the suction box. A small air gap was pre
sent between the top surface of the sample and the lid.
The two HCTs with an initially dry porous filter
were subjected to ‘one-shot’ saturation directly in
the saturation chamber by pressuring water to 4 MPa
for 5 days. The response of the HCTs was then
tested by generating water tension up to cavitation
(by wiping the porous filter). Cavitation occurred at
the air-entry value of the porous ceramic filter. The
HCTs were then placed again in the saturation cham
ber at 4 MPa for 5 days.
The long-term measurement test is shown in
Figure 3. The two HCTs successfully measured the
same pore-water tension despite the fluctuations due to
water drops condensing on the lid and falling periodic
ally on the sample surface. Cavitation occurred as the
Figure 1. Piezocone and Tensiocone for pore-water pres 800 kPa air-entry value of the porous filter was
sure measurement below and above the phreatic surface approached.
respectively.
2.1 Tensiocone HCT

The HCT is installed horizontally into the adaptor and
its size must be kept small in order to maintain the rear
electrical connections near the axis of the cone and
allow the electrical cables to run through the hollow
push rods. Figure 2 shows the design concept for
the HCT.
It consists of an integral strain-gauge diaphragm
and an extension to support a perforated board with
four pins to connect the thin strain-gauge wires to Figure 3. Long term measurement using two
the thick cable wires transferring the signal through High-Capacity Tensiometers installed simultaneously on
the hollow push rods. A high air-entry value porous the same clay sample.
249
3 TENSIOCONE MESUREMENT IN MOCK-UP pressure measurements of the two HCTs installed on
LABORATORY TEST the tensiocone and top surface respectively.
There was a lowest value of water pressure
A series of penetration tests were carried out in sam detected by the tensiocone HCT during penetration.
ples placed in a mould 100 mm diameter and Following penetration, pore-water pressure increased
150 mm height (Figure 4). A transparent lid was used to -550 kPa and equalised to this value. The top
to seal the upper surface of the sample to avoid evap HCT equalised to a slightly lower pressure (-560
oration during the test. Two holes were made into the kPa) therefore showing that adequate contact could
lid; the larger one was used to insert the tensiocone be established between the porous ceramic filter of
equipped with the HCT whereas the smaller one was the tensiocone HCT and the sample.
used to install a HCT on the top of the sample to
benchmark the measurement by the Tensiocone HCT.
The aim of the test was to verify whether adequate
contact could be established between the HCT porous
and the soil during penetration in the absence of the
clay paste typically used in HCT measurements.
Figure 5. Penetration test on lightly compacted sample

(20% clay/ 80% sand).
3.2 Moderately compacted samples

The second part of the experimental program
involved moderately compacted samples. Three sam
ples with different coarse/fine fraction ratio and com
paction water content were prepared for this purpose.
The soil was first humidified and cured for at least
24h in sealed plastic bags and then compacted in
three layers inside the mould using a loading frame
Figure 4. Mock-up scale test to simulate measurement until reaching the target dry density of 1.4 g/cm3.
during penetration. Figure 6 shows the pore-water pressure measure
ments during penetration in a sample composed of
80% of sand, 20% of kaolin and humidified at 40%
3.1 Lightly compacted sample
water content calculated with respect to the mass of the
The first test was made on a lightly compacted fine fraction (corresponding to an overall water content
sample, which should represent the worst case scen equal to 8%). Similar to the lightly compacted sample,
ario (the lower the density, the lower is the radial the tensiocone pore-water pressure decreased during
pressure pushing the soil against the HCT porous penetration and increased once penetration was stopped
filter). The sample was prepared using 80% of until achieving equilibrium. The top HCT converged to
coarse fraction (sand) and 20 % of fine fraction similar values (the drop at the end of the measurement
(Speswhite kaolin clay) humidified at 30% water was due to the loss of contact with the sample).
content calculated with respect to the dry mass of
kaolinite clay only (corresponding to 6% overall
water content). The two fractions were mixed dry
and water was then sprayed over the mix. After
mixing, the humidified soil was sealed in a plastic
bag for at least 24h for water content equilibration.
In this test, the soil was placed in the mould in
a single layer and compacted manually. After closing
the mould with the transparent lid, the tensiocone was
penetrated for about 100 mm into the sample. Another
HCT was installed on the top surface of the sample.
To improve the contact between the top HCT and the
sample surface, a thin layer of kaolin paste (water con Figure 6. Penetration test on moderately compacted
tent 100%) was interposed between the top HCT and sample (80% sand, 20% kaolin, 40% water content to
the soil. Figure 5 shows the results of the pore-water kaolin dry mass).
250
The second compacted sample was prepared using 4 TENSIOCONE MESUREMENT IN THE FIELD
only Speswhite kaolin humidified at 32% water con
tent. The as-compacted suction was measured using the The tensiocone prototype (only equipped with
top HCT before starting the penetration of the tensio a lateral HCT) was tested on the crest of the Adige
cone. Once the top HCT measurement reached equilib river embankment at about 1 km downriver from the
rium, the tensiocone penetration was started. Figure 7 bridge that connects the local railway station to the
shows the results of the measurement of the pore-water town of Egna (BZ), Italy. The embankment made of
pressure of the two tensiometers during the last part of a silty material was instrumented with two series of
the tensiocone penetration. Again, once penetration conventional tensiometers at depths of 1.8 m, 3.25 m,
stopped, the measurement of the tensiocone HCT and -4.7 m and depths of -2 m, -3.45 m, and
attained the same value recorded by the top HCT. -4.5 m respectively. At the time of the CPT, these ten
siometers showed a pore-water pressure profile nearly
hydrostatic at depths greater than 3 m (associated
with the phreatic surface located at 7m below the
embankment crest). These field measurements were
aimed to provide a reference to benchmark the meas
urements by the tensiocone HCT.
The tensiometer installed in the tensiocone was
preliminarily saturated for more than 24 hours using
a portable saturation chamber (at the constant pres
sure of 4MPa). The pressure in the saturation cham
ber was applied via a piston screw pump (Figure 9a).
Figure 7. Penetration test on moderately compacted
The tensiocone was then removed from the satur
sample (0% sand, 100% kaolin, 32% water content to ation chamber and placed in water for zeroing
kaolin dry mass). (Figure 9b). Finally, the tensiocone was removed
from water and screwed onto the first push rod.
Once removed from water, the porous filter was
The third sample was prepared with 60% coarse immediately covered with kaolin paste to prevent
fraction and 40% fine fraction. The water content cavitation during the short period of time where the
calculated with respect to the dry mass of the fine HCT remained exposed to air before penetrating the
fraction was 40%, corresponding to an overall water ground (Figure 9c). This kaolin paste was chipped
content equal to 16%. The results of this test are off once the tip started penetrating the embankment.
shown in Figure 8. The tensiocone pore-water pres
sure decreased during penetration and then increased
up to ˗145 kPa. This value was markedly different
from the top HCT that equalised at about -40 kPa.
This discrepancy was also observed in other tests not
reported herein.
The reason for such a discrepancy was not due to
poor contact between the soil and the HCT porous
filter. The pore-water pressure recorded by the ten
siocone HCT levelled off suggesting that adequate
contact could be established. In fact, if an air gap
had formed between the soil and the porous ceramic
filter, water from the ceramic filter would have evap
orated into the air gap and pore-water pressure Figure 9. Field test. (a) Tensiocone in the pre-pressuration
would have started declining rapidly over time. chamber. (b) Tensiocone in free water before installation.
(c) Tensiocone at the onset of penetration.
The pore-water pressure measured at different

depths is shown in Figure 10. The thrust system was
stopped four times at four different depths. During
penetration, pore-water pressure always tended to
decrease whereas equilibrium was established in
a relatively short time after penetration was stopped.
The pore-water pressure levelling off once penetration
was stopped was taken as an indication that adequate
contact could be established between the ground and
the HCT porous filter.
Figure 8. Penetration test on moderately compacted However, the equilibrium values where not consist
sample (60% sand, 40% kaolin, 40% water content pect to ent qualitatively and quantitatively with a hydrostatic
kaolin dry mass). pore-water pressure profile controlled by the phreatic
251
surface at 7m depth. It was then speculated that the 6 CONCLUSIONS
HCT readings were affected by spurious mechanical
deformation of the HCT sensing diaphragm. Two This paper has presented the concept of the Tensio
types of stresses can influence the measurement, the cone, a cone penetrometer with the facility to meas
radial stress due to the ground lateral compression ure pore-water pressure in both negative and positive
and the axial stresses imposed by the thrust system. range. A first prototype was tested in the laboratory
and the field to investigate one of the major chal
lenges in tensiocone measurement, i.e. whether
adequate contact can be established between the
ground and the HCT porous filter during penetration.
The clay paste usually applied to the high air-
entry porous filter of standing alone HCTs to ensure
adequate hydraulic continuity between the pore-
water and the water in the HCT reservoir cannot be
used in CPT. The measurement of the negative pore-
water pressure during penetration has to rely on the
contact generated by the radial compression during
penetration of the CPT shaft.
Laboratory tests showed that adequate contact
could be established even under unfavourable condi
Figure 10. Results of field test. tions, i.e. high fraction of course-grained material
(this makes the contact more difficult to achieve)
5 LOADING EFFECTS and lower soil density (low radial stresses develop in
a lightly compacted sample). The field test also
To investigate whether the tensiocone HCT readings showed that equilibrium could be reached after stop
were affected by spurious mechanical deformation ping the penetration, indicating that adequate contact
of the sensing diaphragm, the tensiocone was placed could be established between the penetrometer HCT
in a loading frame. The tensiocone HCT porous filter and the ground in a real case scenario.
was air dried for several days so that the HCT should However, it was observed that the compression of
have measured zero-gauge pressure throughout the the shaft induced deformations in the HCT resulting
mechanical axial compression of the tensiocone. in false pressure readings. This explains the almost
Figure 11 shows the results of the axial loading random values recorded during the field penetration
test. The tensiocone was compressed in steps up to test and the discrepancy between the suction meas
10 kN (0kN, 2 kN, 5 kN and 10 kN) and then ured by the tensiocone HCT and the HCT placed at
unloaded in steps (10kN, 5 kN, 2 kN and 0 kN). Step the top of the sample in the laboratory tests. The
duration was 15 seconds. The HCT recorded signifi next challenge in the design of the penetrometer
cant ‘false’ changes in pressure up to 180 kPa. In the shaft is therefore to isolate mechanically the HCT
field test, circumferential stresses might have also from the shaft and this will be the focus of future
generated additional spurious readings. tensiocone design developments.
The first HCT prototype was designed as an inte
gral strain gauge diaphragm and any deformation of
the HCT body due to the compression of the shaft ACKNOWLEDGEMENTS
directly affected the response of the sensing dia
The financial contribution of the European Commis
phragm. This design is clearly not suitable for the
sion to this research through the Marie Curie Industry-
HCT to be incorporated into the tensiocone.
Academia Partnership and Pathways Network MAGIC
(Monitoring systems to Assess Geotechnical Infrastruc
ture subjected to Climatic hazards) – PIAPP-GA-2012
324426 – is gratefully acknowledged.
REFERENCES
Marinho, F. A. M., Take, W. A. & Tarantino, A. 2008.
Measurement of matric suction using tensiometric and
axis translation techniques. In Laboratory and field test
ing of unsaturated soils. Geotechnical and Geological
Engineering, 26: 615–631. 10.1007/s10706-008-9201-8.
Miller, G. & Collins, R. 2019. A method for predicting the
influence of matric suction changes on CPT tip resist
Figure 11. Mechanical effect on the HCT of the first ten ance. E3S Web of Conferences 92: 18005. 10.1051/
siocone prototype. e3sconf/20199218005.
252
Miller, G., Tan, N., Collins, R., Muraleetharan, K. 2018. proceedings of the third international conference,
Cone Penetration Testing in Unsaturated Soils. Trans UNSAT2002, vol 3, pp. 1005–1017. Lisse, the Nether
portation Geotechnics. 17. 10.1016/j.trgeo.2018.09.008. lands: A. A. Balkema (Taylor & Francis).
Mondelli, G., De Mio, G., Giacheti, H.L. & Howie, J. A. Tarantino, A. & Mongiovì, L. 2002. Design and construc
2009. The use of slot filter in piezocone tests for site tion of a tensiometer for direct measurement of matric
characterization of tropical soils. In M. Hamza et al. suction. In J. F. T. Jucá, T. M. P. de Campos and
(Eds.), Proceedings of the 17th International Confer F. A. M. Marinho (Eds.), Unsaturated soils: proceedings
ence on Soil Mechanics and Geotechnical Engineering. of the third international conference, UNSAT2002,
IOS Press. 10.3233/978-1-60750-031-5-973 vol. 1, pp. 319–324. Lisse, the Netherlands:
Pournaghiazar, M., Russell, A., Khalili, N. 2012. Linking A. A. Balkema (Taylor & Francis).
cone penetration resistances measured in calibration Tarantino, A. & Mongiovì, L. 2003. Calibration of ten
chambers and the field. Géotechnique Letters 2: 29–35. siometer for direct measurement of matric suction.
10.1680/geolett.11.00040. Géotechnique 53(1): 137–141. 10.1680/geot.2003.
Sandven, R. 2010. Influence of test equipment and proced 53.1.137.
ures on obtained accuracy in CPTU. In Mitchell et al. Yang, H-W. & Russell, A. 2015. The cone penetration test
(eds.), 2nd International Symposium on Cone Penetra in unsaturated silty sands. Canadian Geotechnical Jour
tion Testing, Huntington Beach, CA, USA, May 2010. nal 53. 10.1139/cgj-2015-0142.
Tarantino, A. 2002. Panel lecture: direct measurement of Yang, H-W. & Russell, A. 2016. The cone penetration test
soil water tension. In J. F. T. Jucá, T. M. P. de Campos in unsaturated silty sands. E3S Web of Conferences 9:
and F. A. M. Marinho (Eds.), Unsaturated soils: 09008. 10.1051/e3sconf/20160909008.
253
© 2022 the Author(s), ISBN 978-1-032-31259-0
New portable pressiocone system for carrying out CPT+FDP tests

G. Vinco
Geotechnical Engineer, Technical Manager of URETEK SOLUTIONS Group
M. Sacchetto
Geotechnical Engineer, Freelance Consultant
ABSTRACT: Geotechnical investigations in the foundation soil of existing buildings have always been
challenging due to limited space and difficult access. URETEK has developed a portable integrated system
for simultaneously carrying out a CPT and a pressiometric test with Full Displacement Pressuremeter. The 30
kN thrust penetrometer to be used is very small. The reaction is given by two “microanchors”. The cone is
a standard 10 cm2 digital memory cone (no cable), capable of measuring qc, fs, U every centimeter. Above
the cone there is the FDP equipment with a rubber sheath covered by steel plates and connected, by a tube
filled with water, to a device for creating pressure to inflate the sheath and measure pressure-volume curves as
in a standard pressiometric test. The pressure-volume device and the depth transducer are connected to
a microcomputer that is programmed to carry out CPT+FDP tests in an easy-to-use/user-friendly way.
1 INTRODUCTION energy per driving stroke is equal to 60 kJ/ m2 and

the number of strokes required for the penetration of
The recovery of existing buildings is the key to 10 cm represents the resistance of the ground.
building development in Western countries where Dynamic tests 30-20 are mainly used for
the population does not grow, and housing standards a comparison between the cone resistance of the nat
have long since reached a satisfactory level. Tax ural soil and the treated soil. Only in particular cases
incentives and funding introduced by national gov (clean sands), the measured values can be reliably
ernments and the European Community have correlated to the geotechnical parameters of resist
strongly oriented the offer towards systems and solu ance and deformability. The success of the consoli
tions that promote the recovery and adaptation of dation technique with expanding resins, the
structures to current safety and energy saving stand publication of the design guide and the dissemin
ards. Among the various sectors, geotechnics has ation of dedicated software have created the increas
seen the birth and development of new technologies ing need for more precise information on the
capable of restoring the mechanical characteristics of geotechnical parameters of the treated soils (Mana
the soil of existing foundations with a minimum ssero et Al.2014). For this reason, the R&D sector of
impact. Injections of expanding resins represent an Uretek, a multinational company operating in the
example of this. It is a technology that can be done consolidation sector with injections of expanding
in structures in operation by applying a simple and resins, has been involved for some time in the con
non-invasive process (Dei Svaldi et Al. 2005). The struction of a light and easy-to-use instrument cap
subsequent phase of verification and control of the able of detecting directly with the due accuracy the
effectiveness of the injection intervention is carried resistance and deformability of the treated soil.
out by in situ geotechnical tests performed with
small portable lightweight equipment, capable of
operating in narrow and confined spaces. For the 2 GENERAL DESIGN OF THE EQUIPMENT
verification of the improvement of the ground fol
lowing the injection of resins, the use of a small Dynamic tests (especially the ones with low energy,
portable penetrometer, called 30-20 (Cestari, 2013) like the above mentioned 30-20) are considered reli
consisting of a falling weight of 0.3 kN which is able only in certain contexts (clean sands or granular
dropped from a height of 20 cm onto the axis of soils); therefore, it was decided to design equipment
a battery of rods with a conical tip screwed to the for static tests instead than dynamic, with piezocone
bottom. The vertical fall of the mass makes the cone CPTU or electrical cone CPTE. In fact, for the
penetrate the ground for a certain depth. The specific CPTU/CPTE are available many correlations that
DOI: 10.1201/9781003308829-32
254
allow to reliably calculate the geotechnical param a diameter of 3.56 cm, to have more space for digital
eters of the subsoil (Cestari F. 2013). However, the electronics and especially for having no problems in
CPTU/CPTE are not deformative tests, but the resist interpreting the data (the 10 cm2 cones are the most
ance is measured breaking and displacing the soil, widely used and the correlations are very solid).
meaning that geotechnical parameters are obtained Therefore, the cone is a custom model that responds
in an empirical/semi-empirical way and therefore do to the most recent standards and allows the measure
not give (if not indirectly) parameters related to the ment of qc, fs, U; the design of the cone and electron
deformation and/or the stress history of the subsoil ics leaves the possibility of making a CPTE cone
(Robertson P.K. 2009, Robertson P.K., Cabal K.L. with only the qc and fs or a cone equipped with add
2012). itional sensors (for instance inclinometers, seismic,
For this reason, it was chosen from the beginning resistivity, thermal, etc.).
to couple an FDP (Full Displacement Pressiometer) The cone is designed for extremely easy mainten
to the CPTU, so that the penetrometric and pressi ance, with the tip, filter and sleeve that can be
ometer data could complement each other and pro removed on site. The filter for the U can be either in
vide a more complete picture of the geotechnical sintered metal saturated with deaerated silicone oil
characteristics of the subsoil (H.S. Yu et Al., 1996, J. or slotted (slot filter) saturated with silicone grease.
L. Briaud, 1992, J.L. Briaud et Al. 1979,
P. Cosentino 2018, Amar S. et. Al. 1991, Robertson,
P.K et Al.1983).
The pressiometer is an instrument that, in the
standard version with large diameter (60 mm),
Figure 1. Memory piezocone (including battery and
would require the careful execution of a pre-hole,
electronics).
but at this stage it was decided not to design a rotary
drilling machine (albeit of small dimensions) as it In order not to have problems related to the man
would not adhere to the initial targets: portability, agement of the cable, given that the passage through
low or no impact, speed and ease of use even for the rods would interfere with the pressiometer tube,
operators who are not experienced in drilling and creating many problems on site, it was decided to
pressiometer testing. design a custom system with memory.
There are few small diameter (36 or 44 mm) FDP The cone electronics are structured as follows:
systems available on the market, none of which are
integrated with a CPTU/CPTE system. All the equip – high resolution A/D circuit (24 bits)
ment used so far has been customized (P.E. Failmez – programmable microprocessor
ger et 2005, A. Drevininkas 2017). The combination – Large memory capacity and expandable
of CPT (Cone Penetration Test) and PRESSI – high-capacity and small in length
OMETER (hence the name PRESSIOCONE) – rechargeable NiMh battery (60 hours of power
requires to be used with a static penetrometer. The before recharge)
penetrometers that are on the market, although some A great effort was carried out to optimize the power
models are small, are not very suitable for working consumption of electronics, for limiting the dimensions
in contexts where we would like to operate. of the piezocone (in length), the total length of the
Some commercial models are small and can be piezocone being comparable to most cones without
disassembled with an external power unit but require memory. It has been designed to be used (eventually)
a considerable effort for setup before each test, being also with cable, transmitting digital signal.
equipment that are not produced in large series they The test procedures with the memory cone are:
are also relatively expensive.
Furthermore, all the small penetrometers on the a. preparation of the cone and connection with
market are anchored with augers for the reaction to a computer or data logger, which is connected to
the thrust, which is almost never feasible in the con the encoder depth transducer
texts of confined spaces inside existing buildings. b. synchronization of the cone and datalogger and
For these reasons we have opted for the design of definition of the time T = 0; starting from T = 0
a “custom” penetrometer that meets the basic the cone starts storing all the values of qc, fs, U as
requirements of ease of use, portability, electrical a function of time (scanning every second or
operation, reduced height, anchoring without augers, half second) and the data logger the values of
with a thrust of at least 30 kN and standard driving depth as a function of time (data logger is con
speed. of 2 cm/s. nected to the depth transducer)
c. execution of the test
d. recovery of the cone
3 CHARACTERISTICS OF THE CONE e. downloading data from the cone and from the
datalogger, at this point there are two data mat
Although there are cones of reduced diameter (5 cm2 rices, one coming from the cone with all the data
with a diameter of about 2.6 cm) we chose to have as a function of time only, one coming from the
a standard cone with an area of 10 cm2 and datalogger with depth as a function of time.
255
f. synchronization of the matrices: the matrices are
automatically synchronized allowing to have the
penetrometric data vs. depth (not only vs. time,
with exception of the eventual dissipation tests).
4 FEATURES OF THE FDP PRESSIOMETER
The FDP (Full Displacement) Pressiometer is made

of a special rubber sheath coated with six shaped
steel plates and mounted on a cylindrical body which
has an attachment for the piezocone in the lower part
and an attachment for the penetrometer rods in the
upper part. The steel lamellae can slide up and down
at both ends so to follow the expansion of the sheath. Figure 4. Setup of the FDP system.
Figure 2. Sheath and lamellae.
Figure 5. Manual pressurization System and manometer.

Figure 3. Lamellae sliding inside the tube while the sheath
expands. The datalogger is programmed not only to
manage the test manually, but also to automatically
The length of the sheath and the lamellae is 34 cm, perform some pressiometer test “sequences”.
the diameter of the sheath (including the lamellae) is Each sequence contains a certain number of steps
34 mm, with an internal volume of around 200 cm3 which define time, volume, pressure. In this way it is
(not inflated), see Figures 2, 3. The pressure gauge, possible (for instance) to set a preset for a certain
unlike the others on the market, has been designed to automatic sequence for a certain type of soil or to
be inflated with water instead of nitrogen. memorize a manual sequence of a test in progress.
The water, in this prototype, emulsified with oil, This procedure ensures maximum operational
is pressurized through a “Rilsan” plastic tube that is flexibility, ease of use even for Users with little or no
placed through the rods; however, we plan to use experience in pressiometer testing and the possibility
special sealed rods to avoid using the internal tubing to manage tests remotely (simply by exchanging
and pressurize directly through the rods. remotely the files that define a certain sequence, thus
The sheath is pressurized by a system essentially fulfilling the requirements of Industry 4.0).
consisting of a cylinder and a piston which will be
driven (in the final version) by a worm screw con
nected to an electric stepping motor, digitally con 5 FEATURES OF THE PENETROMETER
trolled by the programmed data logger.
Each revolution of the motor corresponds to The most important feature of the new type of pene
a movement of the piston and therefore a certain trometer, which distinguishes it from all models on
volume, measured with great accuracy. the market, will be the anchoring, in relation to the
In the cylinder there is a digital pressure cell con operational context in which the tests will be
nected to the data logger (same device that manages performed.
the cone). However, the preliminary trial tests (car In fact, the pressiocone tests will be carried out
ried out for designing the pressiometer) have been mainly under the existing foundations, mostly inside
carried out with a manometer using a preliminary buildings, in the not treated soils first and after in
version of the system, not yet automatized and with treated soils with resins; therefore, it would be
a manually driven piston and a manometer for meas impossible to anchor with augers or use ballasted
uring the pressure penetrometers.
256
For this reason, the penetrometer will be
anchored with two Microanchors (URETEK,
2017), being small tie rods placed in the founda
tions after execution of a small hole (26 mm
diameter) and fixed with expanding resins; numer
ous tests have confirmed that two anchors are more
than sufficient for a thrust of 30 ÷ 50 kN (which
will be the maximum thrust obtainable with the
penetrometer).
Figure 7. Qc, fs, U graph.
The groundwater table in that area is deeper than

the maximum depth reached with the tests, (around
-4 m) so the U seems not to be significant.
For pushing the pressiocone has been used
a prototype of penetrometer, the microanchors have
been installed in the natural soil instead than in the
foundations.
Figure 6. Setup of the penetrometer.
Basically, the new penetrometer consists in

a light frame with three holes, a central one for the
passage of the pressiocone and two lateral ones for
the installation of the two Microanchors for
reaction.
The thrust devices (pistons) will be chosen
according to the available space and the
requested thrust. One more difference with the
standard penetrometers will be that the thrust
devices will not push the rods from the top of
them, but will push dragging the rods laterally,
with special clamps, to leave the mouth at the
top of the rods free, so better managing the Figure 8. Prototype of the penetrometer, anchored with
pressiometer tube and for reducing the height microanchors.
(“through spindle”). At the same time with this
system any rod of any length can be pushed for Before starting the test, a very careful satur
any depth without constrains. ation of the system (pressumeter, Rilsan tube,
manual pump) has been made with water (emulsi
fied with oil), not to leave any air bubble in the
6 TRIAL TESTS circuit.
The pressiometric tests have been carried out
At present we have carried out some preliminary every meter starting from -1.0 ÷ -1.5 stopping
trial tests, for testing separately the piezocone, the the pushing of the rods and making the sheath
FDP pressiometer with different types of sheaths and expand up to the double of the initial volume,
lamellae and the penetrometer; finally, we tested all sometimes with loops, measuring pressure vs.
equipment together. volume at times T=0 (immediately after expan
The field tests have been carried out during sion) and T=180 s (180 seconds after).
summer 2021 in a site (San Felice Extra, north of an example of a pressumeter test graph carried
Verona) where the stratigraphy is gravel and small out at -2.5 m is shown below, the two curves
cobbles embedded in stiff silty clay, as shown in the show the V-P values at time T = 0 and T =
following graph. 180 s
257
REFERENCES
Manassero, Dominijanni, 2014 Consolidamento dei terreni
con resine espandenti – Guida alla progettazione –
McGraw-Hill Education.
Robertson, P.K., Cabal K.L., 2012 Guide to Cone Penetra
tion Testing for Geotechnical Engineering, Gregg Drill
ing & Testing, Inc., 5th Edition,
Robertson, P.K., 2009 Interpretation of Cone Penetration Tests
a unified approach., Can. Geotech. J. 46(11): 1337–1355
A. Drevininkas, M. Manzari, 2017, Evaluation of Full Dis
placement Pressuremeter for Geotechnical Investigation in
Southern Ontario, Proceedings GEO OTTAWA 2017
H.S.Yu, F.Schnaid, I.F.Collins. 1996 Analysis of Cone Pres
sumeter Tests in sands. Journal Of geotechnical Engin
eering. Vol.122 Issue 8
Figure 9. V-P (volume-pressure) graphs at time T = 0 and Briaud, J.L., Shields, D.H., 1979, “A Special Pressuremeter
T = 180 s. and Pressuremeter Test for Pavement Evaluation and
Design, Geotechnical Testing Journal, ASTM 2: 3
Briaud, J. L. 1992. The Pressuremeter. Taylor & Francis
Group plc, London, UK.
Failmezger, P.E. 2005. Development of a Robust Push-In
Pressuremeter. Symposium International ISP5/
7 CONCLUSIONS PRESSIO 2005, Ponts & Chaussees, France
Dei Svaldi, A. Favaretti, M. Pasquetto, A. & Vinco, G.2005.
The first results were very encouraging, and all Analytical modelling of the soil improvement by injections
parts of the instrumentation worked properly. Thus, of high expansion pressure resin. In 6th International Con
the guidelines for the future development of the ference on Ground Improvement Techniques; Congress
instrumentation and for the definition of the test proceedings, Coimbra, 18-19 July 2005: 577–584
F.Cestari 2013 Prove Geotecniche in Situ Flaccovio Editore
execution procedures were completely outlined: P. Cosentino, A. Shaban, A. Boggs 2018. Predicting Bear
anchoring, preparation, and calibration of the piezo ing Ratios of Granular Soils Using Dynamic Cone
cone and of the FDP, programming of the FDP test Penetrometer and Modified PENCEL Pressuremeter
sequences. Tests Proceedings IFCEE 2018
This first phase will be followed in a very Amar S. et. Al., 1991. Utilisation des résultats des essais
short time by comparative validation tests in pressiométriques pour le dimensionnement des fondations
which tests will be carried out (first in “natural” en Europe, 1ére partie: Pressiomètre Ménard/Pressiomètre
soils and then in soils treated with resins) with auto-foreur. Rapport du Comité Technique Régional Eur
the pressiocone and, at a short distance, com opéen n° 4 Pressiomètres, comitat Français de la Mécani
que des Sols et des Fondations, Rotterdam: Balkema,
parative tests with standard instrumentation Robertson, P.K., Hughes, J.M.O., Campanella, R.G., and
(boreholes with sampling and laboratory tests, Sy, A.,1983. Design of Laterally Loaded Displacement
with execution of standard pressiometer tests, Piles Using a Driven Pressuremeter. ASTM STP 835,
SPT, Vane Test, execution of CPTU and DMT Kansas City
dilatometric tests). URETEK, Microanchors, pers.comm.
258
© 2022 the Author(s), ISBN 978-1-032-31259-0
Characterisation of near-surface sediments using a blend of vertical and

shallow rotational penetrometers
D.J. White
S.A. Stanier
University of Cambridge, UK
H. Mohr
University of Western Australia and MSMT Solutions, Perth, Australia
ABSTRACT: The mechanical properties of near-surface sediments – to a depth of approximately half

a metre – are relevant to the design of cables, pipelines and other shallowly-embedded infrastructure, as well
as benthic habitat characterisation. For this depth of interest, vertically-pushed penetrometers – such as the
cone, T-bar or ball – can be supplemented by shallow rotational devices such as the toroid or hemiball. In this
paper, we report vertical and shallow rotational penetrometer test procedures and show results obtained in pro
ject conditions on natural soil samples. By combining these different penetrometers, a wider and more reliable
set of mechanical properties – spanning strength and consolidation behaviour – can be obtained, compared to
conventional practice. The paper concludes with practical advice on testing protocols and interpretation
methods to best characterise the shallow seafloor, including the use of novel shallow penetrometers.
1 INTRODUCTION Schneider et al. 2020a) and numerical simulations

(Stanier & White 2015, Yan et al. 2017, Schneider
1.1 Shallow seabed properties et al. 2020b,c). Some of this work formed part of the
Remote Intelligent Seabed Surveys (RIGSS) Joint
The properties of the shallowest half metre of the
Industry Project (www.rigssjip.com).
seabed are relevant to the design of cables, pipelines
and other shallowly-embedded infrastructure such as
rock dump and scour protection systems, as well as
2 COMBINED PENETROMETER SYSTEM
the study of benthic ecosystems. The importance of
accurately determining the strength and consolidation
properties of the seafloor is emphasized in good prac 2.1 Actuation system
tice guidance (e.g. ISO 19901-4, White et al. 2017). A shallow penetrometer test aims to subject the
Typical practice to characterize the shallowest seabed to a loading history that replicates the condi
seafloor includes (i) box core or drop core sampling, tions around infrastructure, and provides a simple
(ii) penetrometer testing from a seabed frame or into basis to interpret fundamental soil properties, includ
a box core and (iii) index testing of samples. ing the interface shear strength at the penetrometer
In this paper, we describe how practice can be surface. The control requirements involve vertical
enhanced by modern vertical and shallow rotational and rotational motion in load-controlled contact with
penetrometer testing to determine more accurately the the seabed. These requirements exceed current sys
strength and consolidation properties of the seafloor. tems for CPT and T-bar testing, which involve only
vertical motion between specified position limits.
1.2 Novel shallow penetrometer systems A system for vertical and shallow penetrometry
must therefore meet the following requirements:
Shallow penetrometers, which combine vertical
movement and rotation about the vertical axis, were 1. Accurate identification of the touchdown point,
first trialled a decade ago, using toroid and hemiball when the penetrometer contacts the seabed.
shapes (Yan et al. 2010). Subsequent work includes 2. Control via feedback of the vertical load on the
laboratory studies (e.g. Boscardin & de Groot 2015, penetrometer, e.g. during a dissipation step.
DOI: 10.1201/9781003308829-33
259
3. Allow ‘programmable motion control’ so that
sequences (e.g. penetration-dissipation-rotation)
can be performed without intervention.
4. Allow variable speeds of movement to permit
testing over a range of drainage conditions.
An actuator system that meets these requirements
was developed as part of the RIGSS Joint Industry
Project (Figure 1, Schneider et al. 2020a). The cap
abilities are summarised as follows:
1. Controlled motion at rates of 0-3 mm/s (vertical)
and 0-3 deg/s (rotational), allowing drained and
undrained movements in clay and silt soils (drained
Figure 2. Miniature penetrometers: Ø5 × 20 mm T-bar,
in sand), with positioning resolution of 0.0125 μm Ø10 mm ball, and Ø5 mm cone (MSMT Solutions).
and 9 × 10-6 degrees.
2. Load measurement and control to an accuracy of
± 0.1 N, even on-deck under ship motion – as The shallow penetrometer actuator system is cap
proven during offshore trials. This force reso able of operation offshore on deck, in box cores.
lution corresponds to less than 0.02 kPa of However, it is equally suited to use in the laboratory,
undrained strength, for the sizes of shallow pene with recovered samples. Both modes have been
trometer used, so the system accuracy is much used.
better than typical sample variability.
3. A user interface and motion control system
allows pre-programmed steps of consolidation 2.2 Miniature vertical penetrometers
(under set vertical load) and/or cyclic rotation. To measure the shallow strength profiles of samples,
4. Ruggedised packaging in plastic flight cases miniature scaled versions of the conventional cone,
allows transport offshore or by plane. ball and T-bar penetrometers are used. These attach
via simple fittings to the same actuator, and provide
high resolution profiles of shear strength. For the
ball and T-bar, remoulded strength is also obtained
through a cyclic phase. The penetrometers shown in
Figure 2 have replaceable top load cells and were
manufactured by MSMT Solutions (Perth,
Australia).
2.3 Shallow rotational penetrometers

Three shallow rotational penetrometers are shown in
Figure 3: ring and toroid penetrometers of outer diam
eter, Do = 125 mm, and body width or diameter, Db =
25 mm, and a hemiball with overall diameter, D =
100 mm. All penetrometers are roughened by a sand
coating and feature embedded pore pressure trans
ducers with porous stones flush with the curved
surfaces.
2.4 Sampling and testing location

The penetrometer system mounts onto box cores, or
can rest on a frame over a sample. Tests on intact
samples can be performed in box cores on deck, or
jumbo samples can be extracted from a box core
(e.g. in a 250 mm diameter plastic tube) and trans
ported back to the lab for penetrometer testing
(Figure 4).
A typical combined testing programme for a single
box or jumbo core involves (i) vertical penetrometer
tests to assess the strength profile, followed by (ii)
one or more shallow penetrometer tests to measure
Figure 1. Penetrometer system: vertical and rotational interface strength and near-surface consolidation
actuation.
260
Figure 4. Testing on deck (left) and in the lab (right).
Figure 3. Shallow penetrometers: ring, toroid and
hemiball.
properties. If insufficient surface area is available, across both instruments, and agreement with
extra shallow penetrometer tests can be performed standard (D = 35.7 mm) cone data from a seabed
after scraping away the top layer or on remoulded frame system. However, for the sandy silt
samples. example (Figure 5c), the initial penetration resist
ance of the standard cone is lower, perhaps due
to near-surface effects on the penetration mechan
3 STRENGTH PROFILING ism, which have minimal influence on the mini
ature cone.
Example profiles from the miniature vertical
penetrometers are shown in Figure 5, for condi
tions with penetration resistance from 10 to 4 SHALLOW PENETROMETER TESTING
1000 kPa.
T-bar tests in very soft clay (Figure 5a) resolve 4.1 Typical test procedure
the undrained strength, su, to 1/10th of a kPa, and The usual steps of a shallow penetrometer test are
are consistent with shallow penetrometer strength illustrated in Figure 6. The steps are described in
data. Miniature cone and T-bar data from Table 1, which sets out the control applied to the ver
undrained penetration in carbonate silt tical and rotational axes in each step, the end points
(Figure 5b) show consistent penetration resistance of each step and the typical speed or durations
Figure 5. Vertical penetrometer tests in intact seabed samples (i) Clay (ii) Carbonate silt (iii) Carbonate sandy silt.
261
relevant to testing in soft clay. On sandy soils the
periods are more rapid, and the dissipation pauses
may not be required. Figure 7 shows typical raw
results.
4.2 Typical test results

The usual interpretation procedure for a shallow
penetrometer test is set out in Table 2, and the inter
preted data is shown in Figure 8. The interpretation
methods for soil strength and consolidation param
eters are based on bearing capacity theory and dissi
pation solutions. The interface strength parameters
derived from the rotational stages involve
a correction for ‘wedging’ effects for the hemiball
and toroid devices. Ring penetrometer data does not
need this correction.
Ideally, fully undrained and fully drained
strengths can be measured, limited by the maximum
actuator speed and test duration, respectively, rela
Figure 6. Stages of a shallow rotational penetrometer test. tive to the soil coefficient of consolidation. Regard-
Table 1. Typical shallow penetrometer test procedure.
Step Step Control Control values (toroid and Step procedure Step end point
name parameters ring)
0 Prepare Vertical v = 0.5 mm/s Position penetrometer above test loca Visual check: pene
rate, tion in sample and drive penetrometer trometer is ~20 mm
v Rotation to test start point. Take zero readings above sample surface
rate, ω = 0 from all transducers. Confirm response
of pore pressure transducers to hydro
static head.
1 Penetrate Vertical v = 0.4 mm/s Drive penetrometer vertically to When w = 0.5D or
rate, 0.5D embedment (or until the load cell vertical load, V =
v Rotation limit, Vlimit, is reached). Maximum Vlimit
rate, ω = 0 measured vertical load is captured as
Vmax, and used in control of Step 2.
2 Unload Vertical V = Vmax/2 Reduce vertical load on penetrometer Move immediately to
load, V by a factor of 2, to V = Vmax/2, (to Step 3
Rotation avoid excessive settlement during rota
rate, ω = 0 tional stage of test) and hold this load.
(Or unload to a specified V to match
the applied stress relevant for design).
3 Dissipate Vertical V = Vmax/2 Hold penetrometer under load When Δu/Δuini < 0.1
load, V of V = Vmax/2, allowing
Rotation dissipation of excess pore
rate, ω = 0 pressure, Δu.
3b Set Vertical V = (Vmax/2)/OCR (Optional, OCR test option) Reduce When Δu/Δuini < 0.1
OCR, load, V vertical load by a specified ratio; repeat
dissipate Rotation pore pressure dissipation. (Consider the
rate, ω = 0 OCR range relevant to design when
planning tests. If needed, add Step 3b
in some tests).
(Continued )
262
Table 1. (Cont.)
4 4a Vertical Fast-medium-slow, Δθ = Basic monotonic test option: Rotate When θ = 60°

(4a Twitch load, V 20 deg rotation at each device by a specified angle, at fast,
or rotation Rotation speed, ω = 0.4, 0.04, 0.004 medium and then slow speeds, to meas
4b) rate, ω deg/s ure the undrained (initially) and
drained (later) torsional resistance, T.
4b Vertical ω = ±0.4 deg/s for Δθ = Episodic test option: Perform episodes When the cycles are
Episodic load, V 20° of rotation and consolidation. In each complete
rotation Rotation tconsol = 30 min rotation stage turn through a fixed
rate, ω angle at the fast speed, alternating dir
ection between cycles. During consoli
dation period, tconsol, between cycles,
maintain vertical load. Typically the
response stabilises after ~15 episodes.
5 Drained Vertical dV/dt = –Vmax/(2tunload) Rotate penetrometer at constant slow When V = 0 (i.e.
sweep load, V ω = 0.004 deg/s speed while reducing the vertical load when the penetrom
Rotation to zero over the unloading period, eter lifts away from
rate, ω tunload. Typically set tunload ~ 30 min the sample)
utes on clay. This step gives the change
in T with V, measuring the friction
across a range of normal stresses.
6 Retract Vertical v = –0.5 mm/s Withdraw penetrometer vertically to Penetrometer returns
rate, above the sample surface. Take add- to starting position
v Rotation itional zero readings. above sample surface
rate, ω = 0
Figure 7. Typical unprocessed results from shallow penetrometer test with a monotonic rotation step (hemiball in soft
clay).
263
Figure 8. Typical interpreted results from a shallow penetrometer test with a monotonic rotation step (hemiball in soft
clay).
less, an effective stress interpretation with surface sliding on clay and (iii) more detailed
a weighted average pore pressure provides effective and repeatable measurements of drained and
friction parameters, even if the rotation is undrained undrained interface strength parameters.
or partially drained.
The interpreted results in Figure 8 show the fol The shallow penetrometers offer an alternative
lowing derived soil parameters: (i) profile of to the low stress interface shear box. In our
undrained strength, su (sub-figure (a)); (ii) consolida experience, this test can be unreliable compared
tion coefficient during loading path (c); (iii) to shallow penetrometer tests due to extraneous
undrained interface strength (j); (iv) drained interface forces polluting the shear and normal forces
strength (j); (v) consolidation coefficient for shearing measured in the shear box. Further studies into
path (k); (vi) effective stress failure envelope (e) and shallow penetrometers, particularly focusing on
(vii) effective stress friction (as distribution) (g). the ring, are reported by Mohr et al. 2022 and
Singh et al. 2022.
5 RESEARCH STUDIES AND PROJECT USAGE

6 CONCLUSIONS
The penetrometer system has been used for >300
tests at the time of writing, for research activities Vertical and shallow rotational penetrometer tests
and for commercial projects across three contin together provide a new method to obtain a wider and
ents. Typical project test programmes have more reliable set of near-surface seabed mechanical
involved 10-20 shallow penetrometer tests spread properties compared to conventional practice.
over 3-4 distinct soil zones, with parallel vertical The equipment, procedures and interpretation
penetrometer profiling. Reliable and repeatable approach for the new types of shallow penetrometer
measurements of interface strength have been are now well-established, with the key details sum
obtained for clays, silts and sands. Examples of marized in this paper. Drained and undrained inter
added project value have included (i) the first face strength parameters, as well as consolidation
ever data on consolidation coefficient at the very properties, can be objectively derived through rapid
near-surface, (ii) better characterization of the testing in box cores or samples recovered to the
consolidation-hardening behaviour from repeated laboratory.
264
Table 2. Typical shallow penetrometer test interpretation procedure.
Calculation coefficients
Step name/
Step type Input parameter(s) Calculated parameter(s) Calculation approach Ring Toroid Hemiball
sum, ksu = f (V(w)) – use iteration to fit a linear su
1 Penetration Vertical load, V Undrained shear strength, su N/A See S&W reference
profile (su = sum + ksuw) (Stanier & White 2015)
U ¼ ΔΔuuini ¼ f 1 nm with time Tv ¼ cDv t2v
Tv
1þ T 50;v
Excess pore Consolidation coefficient, cv Plot data and equation of Δu/Δuini vs. tv, adjust cv. T50,v = 0.0675
T50,v and m vary with w/D.
3 Dissipation pressure response, (for stress path after vertical Or find tv,50 when Δu/Δuini = 0.5 → cv = T50,vD2 m = 1.00
See Schneider et al. (2020b)
Δu(tv) penetration) /tv,50. (Singh et al. 2022)
tv: time since start of dissipation (i.e. start of
step 3). f n
T
Undrained interface friction, μud f eff n
μud ¼ στint0 ¼ VF ¼ ζ Vr f n ζ ¼1
n0 ud D2
Drained interface friction, μdr μdr ¼ μ0 ¼ tan δ ¼ στint0 n ¼ VF ¼ ζ VrT eff reff ¼ L þ 12Lb L ¼ 12 ðDo - Db Þ
f nn dr
Tr
T 50;r cv;r tr
Vertical load, V μ ¼ μdr - ðμdr - μud Þ0:5 with Tr ¼ D2
Torque, T Consolidation coefficient, cv,r T50,r = 0.25
pos
ζ , reff ; Ac and β ¼ ΔΔuuave
Total friction, (for stress path during rotation) Use same method of fitting as described in n = 0.9 vary with w/D.
4, 5 Rotation Step 3.
μ = T/V See Schneider et al. (2020b,

Excess pore tr: time since start of rotation (i.e. start of step 4).
c)
pressure, Δu(tr) Ac ¼ 2πDb L
where τint ¼ ATc and
Effective stress envelope, pos
β ¼ ΔΔuuave ≈0:7 for pore pres
sure at mid radius
These methods have been developed through >300 experimental modelling of the penetration and dissipa
tests, supported by theoretical and numerical analysis, tion stages. Canadian Geotechnical Journal 57(4):
and have been applied in several commercial projects. 568–579
Schneider M., Stanier S., White D.J. & Randolph M.F.
2020c. Shallow penetrometer tests: theoretical and
ACKNOWLEDGEMENTS experimental modelling of the rotation stage. Can. Geot.
J. 57(4):568–579
We acknowledge support from the RIGSS JIP (Fugro, Singh V., Mohr H., Stanier S.A. Bienen B. & Randolph M.
Total, Woodside and Shell), the EPSRC Offshore F. 2022. Characterisation of interface friction strain-rate
dependency of soft sediments at low stresses using
Renewable Energy Supergen Hub (grant EP/S000747/ a ring penetrometer. Geotechnique. In review.
1) and Prof. Mark Randolph and Dr Mark Schneider. Stanier, S. & White D.J. 2015. Shallow penetrometer pene
tration resistance. ASCE J. Geo Geoenv. Eng. 141
(3):04014117
REFERENCES White D.J., Stanier S.A., Schneider M., O’Loughlin C.D.,
Chow S.H., Randolph M.F., Draper S.D., Mohr H.,
Boscardin, A.G. & DeGroot, D.J. 2015. Evaluation of Morton J.P., Peuchen, J., Chow F.C., Fearon R. &
a toroid for model pipeline testing of very soft offshore Roux A. 2017. Remote intelligent geotechnical seabed
box core samples. Proc. ISFOG2015, Taylor & Francis, surveys – technology emerging from the RIGSS JIP.
London: 363–368. Proc. Int. Conf. OSIG. 1214–1222
ISO 19901-4:2022. Specific requirements for offshore White, D.J, Clukey EC, Randolph MF, Boylan NP,
structures — Part 4: Geotechnical design considerations. Bransby MF, Zakeri A, Hill AJ, Jaeck C. 2017. The
ISO (draft). state of knowledge of pipe-soil interaction for
Mohr H., Stanier S.A. & White D.J. 2022. The ring pene on-bottom pipeline design. OTC 27623, Proc. Offshore
trometer for interface shear testing on sand. Technology Conference, Houston.
Geotechnique. In review. Yan Y., White D.J. & Randolph M.F. 2010. Investigations
Schneider M., Stanier S., White D.J. & Randolph M.F. into novel shallow penetrometers for fine-grained soils.
2020a. Apparatus for measuring pipe-soil interaction Proc. 2nd ISFOG. Perth. 321–326.
behaviour using shallow ‘pipe-like’ penetrometers. Yan Y., White D.J. & Randolph M.F. 2017. Elastoplastic
ASTM Geotechnical Testing Journal 43(3): 622–640. consolidation solutions for scaling from shallow pene
Schneider M., Stanier S., White D.J. & Randolph M.F. trometers to pipelines. Canadian Geotechnical Journal
2020b. Shallow penetrometer tests: theoretical and 54:881–895.
266
© 2022 the Author(s), ISBN 978-1-032-31259-0
Development of an enhanced CPT system for Dogger Bank

Tor Inge Yetginer-Tjelta
Equinor, Stavanger, Norway
Simon Bøtker-Rasmussen
Geo, Copenhagen, Denmark
Mike Rose
SSE Renewables, Perth, Scotland
Tom Lunne & Vaughan Meyer

NGI, Oslo, Norway
Callum Duffy
Logos Geoservices Ltd, Oxford, UK
ABSTRACT: An enhanced seafloor CPT system has been developed to support completion of the soil investi
gation campaign for Dogger Bank. This enhanced system has a demonstrable and significant performance
increase over standard seafloor CPT systems; capable of pushing through dense sand layers with qc>100 MPa
and through tens of meters of very stiff clays. At Dogger Bank, this enhanced system has enabled CPT penetra
tions of more than 40 m below seafloor, in soils where standard systems could only average in the twenties. The
system enhancement has been achieved through the application and adaption of techniques well known in the
geotechnical industry (water lubrication and water injection), but which have never before been combined in an
offshore seafloor CPT system. The performance of the enhanced CPT system has enabled a reliance on seafloor
CPTs to acquire data to beyond monopile toe depths, therefore removing absolute reliance on boreholes to
acquire data at turbine locations and facilitating the fast and efficient development of a geotechnical design basis.
1 INTRODUCTION Soil conditions at Dogger Bank consist mainly of

dense to very dense sands and very stiff clays. For these
A challenge in offshore wind farm development is soils, experience has shown that for “standard” seafloor
designing a soil investigation campaign that effi CPTs, refusal frequently happens between 20-30m
ciently and cost effectively provides the geotechnical below seafloor. This is due to the combined effects of
data required for the design of pile foundations. With high rod friction in the stiff clays and high tip resistance
many years of gradually applying in situ testing as an encountered within the dense sand layers, frequently
active part of building a design basis, it is observed found at the base of Dogger Bank clays, at 25 to 30m
that seafloor CPTs are both efficient and less expen below seafloor. Consequently 40+ m penetration cannot
sive than boreholes, yet they are often compromised be achieved at most locations at Dogger Bank.
by not being able to penetrate to pile toe depths. The seafloor CPT penetration statistics from the
Monopiles are the primary foundation type for the two preliminary soil investigation campaigns in
Dogger Bank wind farm development. As outlined in 2010 and 2012 illustrate these factors:
the companion paper (Yetginer-Tjelta et al., 2022),
the design of monopiles requires soil data to 40 Table 1. Summary of early seafloor CPT penetration
45 m below seafloor. Furthermore, windfarm layouts depths Dogger Bank.
are commonly only confirmed late in project execu
tion, necessitating a rapid turnaround from turbine Final penetration depth (m bsf)
location definition, to the delivery of a geotechnical Year No. CPTs 1) Minimum Maximum Average
design basis. Due to these factors, seafloor CPTs
would be the preferred solution provided that 2010 97 3.8 40.2 22.5
a penetration to greater than 40m can be achieved. 2012 87 8.9 40.1 27.2
DOI: 10.1201/9781003308829-34
267
Whilst these CPTs provide good data for the between the rod and the soil is reduced. According
development of a ground model, the sub 40m pene to ISO22476-8: 2012 the friction reducer must be at
trations would not enable standard seafloor CPTs to least 400 mm behind a 10 cm2 cone in order not to
be relied on as the primary data acquisition tool for influence the measured CPTU parameters. Offshore
monopile design. Therefore, for the purpose of soil investigations with a seafloor rig often use
designing the turbine location specific completion a 15 cm2 cone and 10 cm2 rods, which gives good
soil investigation (SI) campaign, the following results in many cases.
options were considered: Another way of reducing the rod friction is to
inject water or drilling mud at some location
‒ Revisit turbine locations with seafloor CPT above the friction reducer. Jefferies and Funegard
refusals of less than ~40m penetration with (1983) reported such a system and showed that
a drillship and down hole CPTs, or the pushing force can be reduced by up to 50 %
‒ Accept high uncertainty on soil conditions below (Figure 1). Staveren (1995) reported that in stiff
refusal depth and build trust into a good ground overconsolidated Belgian Boom clay, normal
model, or CPTs met refusal at 5 m penetration: using mud
‒ Attempt to reach deeper with an improved “deep” injection, up to 62 m penetration could be
seafloor CPT system. achieved.
Revisiting seafloor CPT refusal locations with In very dense sand, high cone resistance can be
a drillship was deemed not acceptable for reasons a factor limiting penetration. Bayne and Tjelta
of cost and time. Whilst the reliance on the (1987) and Yagi et al. (1988) reported designs of
ground model below refusal depth was similarly cone penetrometers where water could be injected
rejected due to the complex geology and into the soil through channels in the penetrometer
reworked soils. Geophysical reflectors were fre tip. The cone design reported by Bayne and Tjelta
quently broken and made it difficult and often (1987) was intended to be used to investigate how
impossible to follow any layer more than a few much skirt penetration resistance in dense dilatant
hundred meters. Furthermore, the lower sand sand could be reduced by water injection. High
(base of Dogger Bank) is found between 25-35m; negative pore pressures and increased effective stres
coincident with the seabed multiple of the UHR ses, resulting in high cone resistance, may be neu
seismic data. This resulted in very poor confi tralized by adding water during penetration. This
dence for the ground model below this depth cone penetrometer was unfortunately not used in
since all detailed information on layer thickness practice. Yagi et al. (1988) used their cone
and general stratigraphy disappeared in the strong
multiple reflection and little details were visible
below.
From the above it is clear that development of
equipment and procedures for achieving larger pene
tration with seafloor CPTs was of critical importance
to support the efficient and cost-effective delivery of
the Dogger Bank completion SI campaign. This
became a main objective with the various CPT cam
paigns in the period 2018-2021. This is further
detailed in Sections 3, 4 and 5 below. It became
a stepwise approach where trials, results/experience
and modifications were considered on a continuous
basis.
As outlined by Yetginer-Tjelta et al. (2022), 3D
UHRS geophysics in 2019 (and beyond) improved
the resolution and capability of the ground model
but were not sufficient to replace CPT data as the
basis for foundation design.
2 PREVIOUS WORK TO REACH DEEPER

PENETRATION
The most common way of increasing penetration is

to reduce the friction along the CPT rods. Frequently
a friction reducer such as an expanding coupling is
used at some distance behind the cone. Thus, the Figure 1. Total cone push force with and without mud
diameter of the hole is expanded and the friction injection (adapted from Jefferies and Funegard, 1983).
268
penetrometer to carry out tests at various effective and one test meeting the target with still 5 kN thrust
stresses. They controlled this by adding water or air reserve. Figure 2 shows total thrust curves for 4
through the tip during cone penetration. CPTs within an area of 110 x 110m with 3 standard
In summary, for further work at Dogger Bank in CPTs reaching maximum thrust between 19 and 24m
2018 the ideas of rod lubrication and potentially whilst one CPT with rod lubrication reached 35.2m
water injection at or around the cone tip were including penetration through two dense sand layers.
selected for further experimenting. In addition to thrust, pressure and flow rate of
lubrication water are measured. There is a potential
for using these parameters in the interpretation of the
3 ENHANCED CPTS USING ROD test results.
LUBRICATION
A strategic cooperation contract was agreed between 4 DIRECT PUSH CPT WITH WATER
the Dogger Bank project and the Danish contractor INJECTION AT TIP, MEASURING THRUST
Geo with the aim to improve seafloor CPT penetra ONLY (DCPT)
tion below mudline from being in the low 20m’s to
anything towards 40m (later changed to 45 m). In From the testing in 2018 (above) it was clear that rod
2018 the following modifications were introduced to lubrication worked and enabled significantly deeper
reach this aim: penetration. It also meant when the dense sand layers
were met at about thirty meters depth, only 10-15%
1. Increased thrust capacity, i.e. 250 kN net thrust of the thrust capacity was spent on reaching this depth
available at seafloor. (note the relatively constant thrust with depth in
2. A rod lubrication system which reduced or elim mostly clay down to 30m). But some of the sand
inated friction accumulation in the stiff clays to layers were very dense, with qc values of more than
enable deeper penetrations in the clays and thus 100 MPa, and for some locations significantly more
enable more thrust capacity in the dense sand than 10-15% of the thrust was required to overcome
layers at base of Dogger Bank clay units. layered units. Refusal was therefore still frequently
These first tests improved the penetration depth encountered in dense sand layers. To deal with loca
below seafloor significantly, from an average of tions where several sand layers made the rod lubrica
23 m to approx. 35.8 m, and with several CPTs tion less efficient, and where deeper sand layers
reaching close to 40 m (target depth at that stage) added to frequent refusals in the 30 m range, further
work took part along two parallel paths.
Firstly, further improvements were made to
improve the efficiency of rod lubrication. Many
options were tested, including:
‒ variations in water injection hole diameter and
location (radially and axially)
‒ combination with friction reducers of various
thickness and length
‒ high and low water pressures
‒ variations in water volume.
Secondly, to improve the penetration of very
dense sand layers, water injection at the cone was
attempted in different ways. This built on experience
gained over many years from projects such as Gull
faks C (Tjelta et al.,1986), Dudgeon suction anchor
trials 2013 (unpublished), and the OWA Suction
Anchor Trial Installation Campaign 2015 (unpub
lished) where significant reduction in the penetration
resistance of skirts and suction anchors was observed
by injecting water at the skirt tip. The effect comes
from partly reducing dilation in dense sand, remov
ing dilation completely (and reducing effective stres
ses) or in some cases probably by flushing away
sand particles.
Various geometries of the cone with flushing
through the tip, at the face and at the neck were
tried, see two examples in Figure 3. These cone
Figure 2. Thrust for Enhanced CPT with lubrication and 3 types were not instrumented and were named DCPTs
CPTs with no lubrication. (Direct Cone Penetration Test), with the only
269
recording of penetration resistance being total thrust 100-150 MPa), the S-cone was developed, as shown in
(a bit like the mechanical cones in the early days of Figure 5. This is a cone tip which combines the power
cone testing). Water injection at these positions is of a 5 cm2 cone in spearheading into dense sand, com
expected to have some influence on a measured cone bined with large holes for water injection to reduce
resistance, but nevertheless this system did result in sand resistance on the face of the cone with enlarge
deeper penetration and provided information of what ment following closely behind the cone tip. The early
was below the dense sand at which time the water version of this S-cone was non-instrumented and only
injection can be reduced or halted. thrust was recorded. Later versions of the S-cone are
instrumented, with qc and inclination recorded (named
the iDCPT), but due to robustness being prioritised
and space limitations associated with water injection,
the tool has less accuracy than a standard cone accord
ing to ISO specifications. However, as can be observed
in Figure 6, where an enhanced CPT and an iDCPT
were performed only meters apart; the enhanced CPT
met refusal at 24 m whilst the iDCPT using the S-cone
penetrated to 45 m. The correlation between qc from
the robust iDCPT and the enhanced CPT is good, and
it is seen that the iDCPT provides useful information
for layers below the depth of penetration attained by
the enhanced CPT.
Figure 3. Various cone tips with water flushing for DCPT

cones.
Figure 4a shows an enhanced CPT to 45m depth

(red curve) in parallel with a non-instrumented cone,
DCPT (black curve). At this location it was possible
to penetrate both cone types to 45 m (target depth).
At many other locations the enhanced CPT met
refusal at shallower depth and soil stratigraphy had
to be inferred from thrust curves only (Figure 4b
provides an example).
What this parallel test in Figure 4a shows is
a very good correlation between qc and thrust when
rod friction is kept low and relatively constant. This
test (black curve) uses combined rod lubrication and
cone water injection.
Figure 4b shows an enhanced CPT meeting
refusal due to max thrust and high tip resistance (114
MPa) whilst the DCPT reached 40m without any
problem (40m was target depth at this test location),
and clearly provides relevant information of soil
strength and stratigraphy. For instance, the driveabil
ity of large monopiles can be considered feasible
since the dense sand layers at 26 and 34m are proven
to be relatively thin. When soil stratigraphy is very
variable (less homogeneous and/or rod lubrication is
less efficient) the thrust/qc correlation is less accur
Figure 4. a) Location A (top): Comparisons between
ate. Although the DCPT system worked well in
normal CPT and DCPT - enhanced CPT and DCPT both
2019, some locations had to be covered by drilling penetrate to ca. 45 m; b) Location B (bottom): Enhanced
and down-hole CPTs. This experience stimulated CPT refused at 26 m whilst the DCPT reached 40 m (target
further development. depth at the time, could have gone deeper).
5 DCPT WITH MEASUREMENT OF TIP The most efficient combination of water injection
RESISTANCE (iDCPT) (pressure and volume), friction reducers and cone
shapes may depend on the soil conditions and will
To achieve improved penetration in the very invariably include some trial and error in the begin
dense sand layers at Dogger Bank (qc in the range of ning, but the results speak for themselves (Table 2).
270
From pre-2018, the average CPT depth has increased Table 2. Summary of Seabed CPT penetration depths
from 22m to 44m in an area where dense sand layers Dogger Bank.
dominate below 25-30m. These numbers do not
include the iDCPT. A challenge has also been to dis Final penetration depth (m bsf)
tribute water in the most efficient way between the
rod lubrication and the tip flushing. Year No. CPTs 1) Minimum Maximum Average
2010 97 3.8 40.2 22.5

2012 87 8.9 40.1 27.2
CPT equipment enhancements from 2018 onwards
2019 100 14.3 45.6 38.8
2020 120 7.5 50.3 40.1
2021 98 23.5 58.8 44.5
6 DISCUSSION
The main objective set out at the start of 2018 was to

reach sufficient penetration with seafloor CPT testing
without the need for drilling and down-hole CPTs.
Two main targets were identified in these early
days:
1. Reduce or “eliminate” rod friction to enable
deeper penetration and have more thrust capacity
available when dense sand layers appeared at
base of Dogger Bank formation.
2. Improve the ability to penetrate into and through
dense sand layers with tip resistances in excess of
100 MPa.
Figure 5. Instrumented cone tip used with the iDCPT. This objective has been met; seafloor penetration
depth to 40-45m is now possible at Dogger Bank at
most locations. The importance of this objective
goes beyond the operational aspects of seafloor
CPTs being more efficient than down hole CPTs in
a borehole. It allows for reduced time from con
firmed turbine layout to delivery of a geotechnical
design basis “by the speed with which the data can
be acquired, processed and interpreted to provide
geotechnical engineering parameters. This is advan
tageous for the development of offshore wind farm
projects which may have 100 or more WTGs. For
the Dogger Bank WTG locations, continuous sea-
floor CPT data permitted semi-automatic processing
and analysis of data and thus:
1) Rapid development of geotechnical design pro
files (preliminary profiles available in a matter of
minutes);
2) Early phase monopile driveability predictions
for identifying the potential risk of refusal during
installation.” quote from companion paper (Tjelta
et al 2022)
On a side note, the ability to penetrate into
and through dense sand layers with tip resistance
in excess of 100 MPa at more than 30m below
seafloor is not trivial. It requires a significant
reduction of rod friction. In doing so, the com
bined effect of a friction reducer and water injec
tion as described in this paper is necessary, and it
Figure 6. Enhanced CPT and iDCPT.
has taken time and efforts to arrive at a geometry
271
that works. The impact of altered geometry due objective of this development was to improve the
to extreme wear on all parts in the CPT system depth range to more than 40m below seafloor in an
complicates the feedback. Figure 7 indicates the environment where average penetration used to be
changes taking place. Diameter changes, and fric in the low 20m’s. This has been achieved through
tion reducers are ripped off by the continuous systematic development and adaption of techniques
abrasion from dense quartz sand. well known in the geotechnical industry (water
lubrication and water injection). However, these
techniques have never before been combined in an
offshore CPT system to produce a machine capable
of pushing through dense sand layers with qc >
100MPa and very stiff clays. The result has been
a 100% increase in CPT penetration depth com
pared to a few years ago.
This performance enables fast and efficient devel
opment of a geotechnical design basis when com
bined with early boreholes strategically placed in
geological units to build a ground model, as dis
cussed in the companion paper (Yetginer-Tjelta
et al., 2022).
ACKNOWLEDGEMENTS
The authors are grateful to Dogger Bank Wind Farm

for the opportunity to create new solutions in deliver
ing a timely and efficient design basis and for permis
sion to publish. Colleagues in Equinor, SSE, Geo and
NGI are acknowledged for their contribution to the
development of an enhanced CPT tool and to this
paper.
Figure 7. New rod with friction reducer 44mm left and to
the right the wear has reduced diameter to 41 mm and in
parts of the FR it is ripped off completely.
REFERENCES
Outstanding challenges to be addressed are effect Bayne, J.M. and Tjelta, T.I. (1987) “Advanced cone pene
ive lubrication while the CPT string is being built, trometer development for in situ testing at Gullfaks C”.
and effective lubrication in permeable sand layers. Offshore Technology Conference, Richardson, Texas,
The importance of the following elements has Paper No. 5420.
been recognized during the 3-4 years it has taken to Jefferies, M.G. and Funegard, A. (1983) “Cone penetration
develop a fully operational and efficient enhanced testing in the Beaufort Sea”. Proceedings of the Confer
seafloor CPT system: ence on Geotechnical Practice in Offshore Engineering,
Austin, Texas, 220–43, American Society of Engineers
‒ Friction reduction and ideally removal along the (ASCE).
entire rod by a combination of a friction reducer Staveren, M. van (1995) “Advanced deep cone penetration
above the friction-sleeve and water injection testing and backfilling in overconsolidated clay;. Pro
‒ Increased thrust force when stiff clays and dense ceedings of the International Symposium on Cone Pene
sand layers are combined, like at Dogger Bank tration Testing, CPT ‘95, Linkoping, Sweden, 2,
99–104, Swedish Geotechnical Society.
‒ The importance of inclination measurement in the Tjelta, T.I., Guttormsen, T.R., Hermstad, J. (1986) “Large-
cone, both for operational purposes (abort the Scale Penetration Test at a Deepwater Site”, Offshore
CPT if becoming excessive or to push deeper to Technology Conference, Houston, Texas, May 1986,
compensate for rod inclination) and for final pres Paper number OTC.
entation of results with true vertical depth Yagi, N., Enoki, M. and Yatabe, R. (1988) “Development
‒ Special measures at the cone tip to tackle extremely of a penetrometer capable of applying pore pressure”.
dense sand layers with qc above 100 MPa. Proceedings of the International Symposium on Penetra
tion Testing, ISOPT-1, Orlando, 2, 1051–7, Balkema
Pub., Rotterdam.
Yetginer-Tjelta T.I., De Sordi, J., Caferri, L.,Rose, M.,
7 SUMMARY AND CONCLUSIONS Duffy, C., Lunne, T., Blaker, Ø., Strandvik,S. &
Meyer, V. (2022). “The role of cone penetration testing
A seafloor CPT system with improved (enhanced) cap in the Dogger Bank offshore wind farm”. Proc. 5th Int.
abilities has been developed for the Dogger Bank field Symp. on Cone Penetration Testing, CPT’22, Bologna,
investigations in the period 2018-2021. The main Italy, 8–10 June 2022.
272
Session 2: Interpretation
© 2022 the Author(s), ISBN 978-1-032-31259-0
CSi – a joint industry project into CPTUs in silty soils

A.H. Augustesen
Ørsted, Denmark
P. Carotenuto, C. Bilici & T. Lunne

R.C.J. Lindeboom & L. Krogh

Ørsted, Denmark
J. van den Bosch & R. Barth

Fugro, The Netherlands
C. Erbrich & S. Ingarfield

Fugro, Australia
D. Giretti
University of Bergamo, Italy
V. Fioravante
University of Ferrara, Italy
H. Dias
Equinor
M.-C. Sougle
Vattenfall
A. Barwise
RWE
S. de Wit
Shell
D. Burbury
Scottish Power Renewables
N. Adams
Carbon Trust
ABSTRACT: The CPTU constitutes the main in situ offshore investigation tool for geotechnical site charac
terisation and provision of soil input for the design of wind turbine foundations. CPTUs are typically per
formed at all foundation positions. Thus, all obtained results of supporting geotechnical in situ, model and
laboratory testing need to be confidently correlated to the CPTU parameters. Most of the interpretation meth
odologies available for industry practice consider the soil behaviour around the cone either as fully drained or
undrained, and acknowledged and well-proven correlations between CPTU parameters and classification and
engineering properties exist for sand and clay. However, for transitional soils, e.g. silty soils, which may
exhibit partial drainage during standardized cone penetration, such robust interpretation schemes do not exist.
This paper presents the background, the objectives, setup and early field results of a joint industry project into
CPTUs in silty soils for developing schemes for planning, execution and interpretation.
DOI: 10.1201/9781003308829-35
275
1 INTRODUCTION 2.1 Drainage conditions
In recent years the normalized penetration velocity
The Offshore Wind Industry is rapidly expanding
V represents an often-used framework to indicate the
across the globe. This expansion leads to the explor
drainage conditions around the penetrometer and to
ation of offshore wind sites that are characterized by
demonstrate how CPTU parameters change with
thick layers of silty sand and silt mixtures. In contrast
penetration rate v, soil properties (horizontal coeffi
to sampling boreholes, CPTUs are performed on all
cient of consolidation of the soil ch) and cone pene
wind turbine positions, which in many cases can be
trometer diameter d, see e.g. Randolph & Hope
more than 100 positions covering large areas. The
(2004), Kim et al. (2008) and Schneider et al.
CPTU parameters must be confidently correlated to all
(2008). The normalized penetration velocity is
supporting geotechnical tests to mitigate the following
defined as:
risks: a) site characterization and establishing facility
position-specific soil parameters for foundation design,
b) choice of foundation concept and relevant design
methodologies, c) installation predictions, d) cable
design and e) increased project maturation times and
site investigation scopes. However, as opposed to sand
For contractive soils and a given cone diameter,
and clays, the Industry finds it generally challenging to
the cone shoulder pore pressure u2 increases, and
properly identify and capture the behaviour of transi
the cone resistance decreases with increasing V,
tional soil, e.g. silty sand, silts and silt mixtures, with
whereas the opposite may be observed for u2 for
standardized CPTUs. No simple, robust, and standard
dilative soils (Schneider et al. 2007 and Krage &
ized testing approaches and interpretation methodolo
DeJong 2016). Drained penetration is observed to
gies have been calibrated to these types of soils.
occur for V < 0.01-0.3 and undrained penetration
A Joint Industry Project (CSi – CPTUs in Silty
for V > 20-100 (Randolph 2004, Bihs 2021). v and
soils) is currently being executed within the Carbon
d are parameters, which can be fully controlled
Trust Offshore Wind Accelerator programme with
and specified. Therefore, an accurate determin
the main objective to develop robust guidelines tar
ation of ch is important in the estimation of V. ch
geting the industry for planning, specification, exe
depends on stress history, density and grain
cution, and interpretation of CPTUs in silty soils.
composition.
Focus is especially paid to
Determination of the operational value of the coef
• Soil classification and mechanical behaviour ficient of consolidation of the soil is not straight for
o New and/or revised soil classification and soil ward, especially not for the more coarse-grained
behaviour type charts transitional soils. Most of the work to date, numeric
o Identification of transition between drainage ally and experimentally, is anchored in contractive
conditions including rate effects, i.e. of drained clay-like soils and undrained conditions.
to partially drained to undrained conditions ch can be determined from piezocone dissipation
tests (PPDT) accounting for partial consolidation
• Correlations to engineering properties during cone penetration, and hence penetration rate,
o Strength and stiffness.
in the interpretation of the tests (DeJong & Ran
• Guidelines for specifying and executing CPTUs dolph 2012). However, for transitional soils, the
for identifying silty soils and their behaviour. excess pore pressure measured at the cone shoulder
u2 does not always decay monotonically over time
The project targets mainly silty and transitional during dissipation and a dilatory response can be
soils with low plasticity. This paper presents the observed. Burns & Mayne (1998) and Paniagua
background of the CSi project as well as the overall et al. (2016) discuss different approaches to deter
project set-up and early findings from a test site in mine t50, the time associated for 50% dissipation,
Halden, Norway. dealing with both standard and non-standard (dila
tory) dissipation curves. Carrol and Paniagua
2 BACKGROUND (2018) note that different interpretation methods
can lead to significant differences in the estimated
CPTUs carried out at a standard constant penetration t50 and hence ch.
rate (20 mm/s ± 5 mm/s according to ISO 2012) gen The rigidity index Ir is required to interpret dissipa
erally result in an undrained response in clay and tion tests. However, the concept of Ir, developed for
a drained response in sand (Lunne et al. 1997). How fully undrained conditions, may not be appropriate for
ever, for silty soils partially drained conditions partially drained conditions. Krage et al. (2014) pro
may prevail at the standard penetration rate, see vide guidance on how to determine Ir.
DeJong et al. (2013). Understanding the drainage To overcome some of the challenges with the esti
conditions around the penetrating cone is key to mation of ch, Schnaid et al. (2020) propose
interpret CPTUs in silty soils (correlation with a modified version of the normalized penetration vel
engineering parameters and classification charts). ocity depending on v, t50, d and Ir.
276
CPTUs at variable penetration rates (VRCPTU) 3 CSI PROJECT SETUP
are traditionally performed to investigate the drain
age conditions around the cone. Even with a robust 3.1 CSi project scope and approach
methodology to estimate V, a link between V and
engineering properties and soil classification need to The CSi project includes a comprehensive scope of
be developed – also for engineering practice. work, see Figure 1, on different scales and platforms
to meet the project objectives and to exploit current
and new methodologies and hypotheses for address
2.2 Soil classification ing the project challenges, cf. Sections 1, 2.
Soil behaviour type (SBT) charts, like Robertson Initial studies in terms of a literature review and
(1990) and later updates or Schneider et al. (2008), are a project plan provides the background and detail the
often used for classification purposes. The charts are other activities. The guidelines, addressing the pro
based on standard CPTU geometries and a penetration ject objectives, are based on analyses of high-quality
rate of 20 mm/s. factual data gathered in a database. The data are
Research (e.g. Schneider et al. 2008, Bradshaw extracted from different sources.
et al. 2012) and practice support that silty soils can In situ tests, and parallel laboratory tests, will be
plot on a large range of zones in the SBT charts and performed on aged natural silty soils at two comple
thus the SBT charts alone can, at present, not be used mentary onshore test sites (Halden, Norway and
to robustly identify grain composition or plasticity for Voorne-Putten, the Netherlands). VRCPTUs with
a silt deposit. The reason is that the position in the different cone sizes, dissipation tests, seismic
SBT chart is a function of several parameters, includ CPTUs and sampling (the static hydraulic piston
ing grain size, plasticity, in situ density, stress history sampler and the static gel-push sampler) will be
and local geological history. Furthermore, partial undertaken. The onshore test site data are supple
drainage conditions affect where data plots on the mented by Partners’ data from offshore projects
SBT charts (DeJong & Randolph 2012) and use of across the world as well as data from other silt sites
additional data from dissipation tests and VRCPTUs published in the literature, thereby increasing the
may aid the identification of silty soils. applicability of the database.
Calibration chamber testing, centrifuge testing
and associated standard element laboratory tests
2.3 Strength and stiffness complement the onshore test site data by providing
Current practice in the offshore wind industry for test results under well-defined and controlled condi
assessing silty soils is to apply methods anchored in tions. Furthermore, they are also relevant for bench
either clean sand (drained conditions, effective stress marking to existing interpretation schemes for clean
approach) or pure clay (undrained conditions, total sand. VRCPTUs and dissipation tests are performed
stress approach). at different densities and consolidation stresses with
The relation between cone resistance and the full-scale and miniature cones in unaged non-plastic
undrained shear strength through the cone factor silty sand and reconstituted clean sands (Fioravante
Nkt has been investigated in the literature, e.g. Sen et al. 2022).
neset et al. (1988), Blaker et al. (2019), Naeini & A state-of-the-art numerical model will expand
Moayed (2017) and Huang et al. (2021). For the the general applicability of the guidelines by validat
effective stress friction angle Senneset et al. (1988) ing and extrapolating the design space offered by the
suggested an approach, which considers pore pres experimental data. Furthermore, recommendations
sure build-up and is relevant for partially drained with respect to numerical models for predicting
conditions. Bihs (2021) reports that both the CPTUs in transitional soils will be provided. The
drained and undrained strength depends on the critical state based Norsand constitutive model along
choice of penetration rate. Senneset et al. (1988), with large-deformation finite element and cavity
Lunne et al. (1997), Robertson (2009) and Tonni & expansion analyses will be used to calibrate and val
Simonini (2013) discuss constrained and small idate the model based on field, laboratory and model
strain shear modulus. tests (calibration and centrifuge testing).
Correlations between CPTU parameters and
engineering properties are dependent on the quality
of the samples tested in the laboratory for calibration
purposes. Furthermore, for the strength correlations
a unique failure criterion may not exist for undrained
shear of especially dilative transitional soils (Blaker
& DeGroot 2020). It is complicated to retrieve and
prepare intact samples for testing and currently no
robust and well-proved criterion for evaluating
sample disturbance exists. Sampling and handling of
silty samples are thus of importance to the CSi Figure 1. High-level strategy adopted during the CSi
project. project.
277
3.2 CSi organisation bronze and HPDE filters were used for the 5, 10 and
15 cm2 cones, respectively. All filters and cones
The CSi project is developed and led by Ørsted in
chambers were saturated with silicon oil. The filters
partnership with five other offshore wind developers.
were saturated under vacuum and placed in
The project is run as a discretionary project through
a container of saturation fluid before mobilization
the Carbon Trust’s Offshore Wind Accelerator pro-
and until assembling with the cone. The cone cham
gramme. The technical activities are managed and
ber was first saturated with syringes and hereafter
executed by the Norwegian Geotechnical Institute
left submerged in a chamber with silicon oil along
(NGI), supported by Fugro and ISMGEO. Selected
with the filters. Vacuum was also applied for up to
activities are reviewed by an independent technical
90 minutes for the 5 and 10 cm2 cones. The pene
review panel to substantiate the developed method
trometer was assembled in submerged conditions
ologies. A certification body (DNV) ensures that the
and covered in a rubber membrane. The filters were
outcome is practically, applicable and relevant for
replaced after each test.
industry practice and future standardization. Figure 2
illustrates the CSi organizational setup.
Table 1. In situ testing scope, Halden campaign.
Item Description
Cones 10 cm2 single element (u2)

15 cm2 dual element (u1 and u2)
5 cm2 single element (u2)**
Scope per cone 1 CPTU benchmark*, 20 mm/s rate
1 VRCPTU, 0.2 mm/s rate, 2 PPDT
1 VRCPTU, 2.0 mm/s rate, 2 PPDT
1 VRCPTU,200 mm/s rate, 2 PPDT
* 2 benchmark tests carried out for the 10 cm2 cone

** Additional scope funded by the NGTS project
*** Dissipation tests (PPDT) as per Figure 3
Figure 2. Schematic view of the CSi project setup. Attention was also given to monitor the ground
water table and thus the hydrostatic pressure (u0
profile), since u0 enters the calculation of e.g. Bq and
4 FIELD TESTING AT HALDEN, NORWAY is the baseline for dissipation tests. The ground
water pressure was continuously monitored at four
depths by means of standpipes permanently installed
4.1 Test site and test programme at the site, indicating hydrostatic conditions with the
The Halden site is located in southern Norway and is water table at 1.85 m depth.
one of five sites within the Norwegian Geo-Test To investigate the drainage conditions around the
Sites program (NGTS). Extensive geophysical, cone through the normalized penetration velocity V,
in situ and laboratory testing have been carried out the chosen penetration rates cover four orders of
as part of the site characterization (see Blaker et al. magnitude (from 0.2 mm/s to 200 mm/s). The aim
2019) and for subsequent research. Normally con was to span from drained to undrained response
solidated clayey silt is deposited between depths of during penetration. Furthermore, three cone sizes
five and 16 m and divided into two units, Unit II and were used to investigate the effect of diameter. The
Unit III, which are homogenous across the site. u1 sensor (located at the cone face) was used in add
The in situ testing programme (CPTUs and ition to u2 to study the differences in response during
a borehole) was carried out in September 2021 and is both variable rate penetration and during dissipation.
summarized in Table 1 and Figure 3. The CPTUs The approach of having one long stroke at variable
were performed according to ISO (2012), except that rate per soil unit at each location was preferred to the
non-standard penetration rates were also adopted. All alternative twitch tests (see e.g. DeJong et al. 2013)
tests were closely spaced but ensuring no interference because: (a) from VRCPTU tests performed earlier at
with each other. Halden it was found that after changes in rate or dis
Efforts were made to ensure good saturation of sipation tests, the length required to build up u2 was
the filters and cone chamber. At fast rate, the u2 in several cases up to 1.0 m and (b) shorter strokes
sensor must respond quickly and at slow rates, it can be affected by vertical variability and hence are
shall ensure accurate measurements due to the small thought to be less robust for future recommendations
values of excess pore pressures measured. Plastic, for offshore site investigations.
278
4.2 Test results
Figure 4 shows selected results of standard rate CPTUs
and VRCPTUs carried out with the 10 cm2 and 15 cm2
cones. The penetration rate profiles indicate that the rig
accurately controls the speed in all tests.
At depths of 10-12/13m all CPTUs are performed
at standard rate. There is generally a good match of
qt, fs and u2 between the location with the CPTU at
standard rate and the VRCPTU locations for both
cones (10 cm2 and 15 cm2), respectively, indicating
that (a) adjacent locations have similar soil condi
tions and hence the profiles at different rates at dif
ferent locations can be directly compared and (b) the
cones are performing consistently among different
Figure 3. Specified penetration rate profiles for VRCPTUs.
Figure 4. Selected in situ test results, CSi Halden campaign. 10 cm2 cone (upper figures) and 15 cm2 cone (lower figures).
279
locations. The long strokes (up to 1.5 m before any 5 CONCLUSIONS
stop, cf. Figure 3) allow for a clear identification of
the changes in qt, fs and u2 for penetrations at Even with the landmark research undertaken over the
0.2 mm/s compared to the standard rate. These last decades on CPTUs in silty soils, more research is
aspects provide robustness to the VRCPTU interpret needed to establish robust guidelines for use in engin
ation approach. eering practice. The objective of the Joint Industry
During penetration at 0.2 mm/s both cones Project CSi is to provide such guidelines. The back
show an increase in qt and fs and a reduction in ground, objectives, setup, and approach of the CSi
u2 compared to the tests with standard rate. For project are described in this paper along with some
example, qt increases on average from 1.0 MPa initial results from the Halden test site. Field testing at
to 1.6 MPa and Bq decreases from 0.2 % to 0.05 the Voorne-Putten test site as well as centrifuge and
% for the 10 cm2 cone in Unit II. In Unit III, the calibration chamber testing are currently being per
impact of a slower rate appears more markedly formed in parallel with laboratory testing and numer
with qt increasing from 1.5 MPa to 4.2 MPa and ical modelling. More Partners and Contractors are
Bq decreasing from 0.08% to zero. This may be invited to join the project for the opportunity of inves
due to the less plastic and coarser nature of the tigating more innovative equipment and methods,
lower unit. such as partially drained triaxial testing and selected
In contrast, the VRCPTUs at 200 mm/s do not CPTU add-on sensors. Furthermore, field trials for
appear to affect qt or fs compared to CPTUs at validating the project outcome are essential and will
standard rate (though for the 15 cm2 cone the add to the robustness and broaden the applicability of
entire fs profile is shifted at the location). Gener the project.
ally, after a prolonged stop in penetration, either
for dissipation testing or for change in rate, u2
increases smoothly and linearly with depth often ACKNOWLEDGEMENT
starting with an abrupt negative increment in the
pore pressure. These results will be further scru The CSi Project is managed through the Carbon
tinized during the project. Trust’s joint industry Offshore Wind Accelerator
The u1 profile for the 15 cm2 cone shows similar (OWA) program. The Authors acknowledge the provi
trends to the u2 sensor for both the highest and sion of financial and technical support by the following
lowest rates noting that (a) the drop in pore pressure project partners: Ørsted (lead partner), Equinor, Vatten
during slow penetration is more distinct and (b) at fall, Shell, RWE and Scottish Power Renewables.
high rates u1 appears to exceed u1 for the standard
rate CPTU after typically 50-70 cm of penetration
REFERENCES
after having changed rate or after a dissipation test.
Figure 5 shows the results of a representative dis Bihs, A. 2021. Investigation of a coarse silt deposit by
sipation tests performed with the 15 cm2 cone. u1 varied rate CPTU testing. PhD dissertation, NTNU,
decays monotonically over time whereas u2 exhibits Norway.
a dilatory response. This response is similar to what Blaker, Ø. & DeGroot, D.J. 2020. Intact, disturbed and
has been reported for cases in overconsolidated clays reconstituted undrained shear behavior of low-plasticity
(e.g. Lunne et al. 1997). The effect of this difference natural silt. Journal of Geotechnical and Geoenviron
on the estimated coefficient of consolidation will be mental Engineering 146(8): 04020062.
Blaker, Ø., Carroll, R., Paniagua, P., DeGroot, D. &
investigated.
L’Heureux J.-S. 2019. Halden research site: Geotech
nical characterization of a post glacial silt. AIMS Geo-
sciences 5(2): 184–234.
Bradshaw, A.S., Morales-Velez, A.C. & Baxter, C.D.P.
2012. Evaluation of existing CPT correlations in silt.
Geotechnical Engineering Journal of the SEAGS &
AGSSEA 42(1).
Burns, S.E. & Mayne, P.W. 1998. Penetrometers for soil
permeability and chemical detection. Georgia Institute
of Technology, Report No. GIT-CEEGEO-98-1.
Carroll, R. & Paniagua, P. 2018. Variable rate of penetra
tion and dissipation test results in a natural silty soil. In
Pisano, Hicks & Peuchen (eds) Cone Penetration Test
ing (CPT18): 205–212. CRC Press/Balkema - Taylor &
Francis Group.
DeJong, J.T. & Randolph, M.F. 2012. Influence of partial
consolidation during cone penetration on estimated soil
behavior type and pore pressure dissipation
measurements. Journal of Geotechnical and Geoenvir
Figure 5. Dissipation test at 8.5 m depth, 15 cm2 cone. onmental Eng. 138(7): 777–788.
280
DeJong, J.T., Jager, R.A., Boulanger, R.W., Randolph M.F. silt following CPTU at variable penetration rate. In
& Wahl, D.A.J. 2013. Variable penetration rate cone Acosta-Martinez, Kelly & Lehane (eds), Geotechnical
testing for characterization of intermediate soils. In and Geophysical Site Characterization ISC’5: 509–514.
Countinho & Mayne (eds) Geotechnical and Geophys Randolph, M.F. 2004. Characterisation of soft sediments
ical Site Characterization ISC’4. London: Taylor & for offshore applications. In Geotechnical and Geophys
Francis. ical Site Characterization ISC’2: 209–232 Millpress,
Fioravante, V., Giretti, D., Augustesen, A.H., Netherlands.
Lindeboom, R., Krogh, L., Dias, H.F., Sougle, M.C., Randolph, M.F. & Hope, S. 2004. Effect of cone velocity
Barwise, A., de Wit, S. & Burbury., D. 2022. Calibration on cone resistance and excess pore pressures. In IS
cone penetration testing in silty soils. In Cone Penetra Osaka-Engineering Practice Performance Soft
tion Testing (CPT22). Deposits: 147–152.
Huang, A-B., Augustesen, A.H., Leth, C.T., Molyneaux, E. Robertson, P.K. 1990. Soil classification using the cone
D.G & Krogh, L. 2020. A field study in the effects of penetration test. Canadian Geotechnical Journal 27
fines on the interpretation of CPTU. In Geotechnical (1):151–158.
and Geophysical Site Characterization ISC’6: 509–514. Robertson, P.K. 2009. Interpretation of cone penetration
ISO 22476-1 2012. Geotechnical investigation and testing – tests – a unified approach. Canadian Geotechnical Jour
Field testing – Part 1: Electrical cone and piezocone test. nal 46: 1337–1355.
Kim, K., Prezzi, M., Salgado, R. & Lee, W. 2008. Effect of Schnaid, F.G., Dienstmann, E., Odebrecht & Maghous, S.
penetration rate on cone penetration resistance in satur 2020. A simplified approach to normalisation of piezo
ated clayey soils. Journal of Geotechnical and Geoen cone penetration rate effects. Geotechnique 70(7):
vironmental Engineering 134(8): 1142–1153. 630–635.
Krage, C.P. & DeJong, J.T. 2016. Influence of drainage Schneider, J.A., Lehane, B.M., & Schnaid, F. 2007. Vel
conditions during cone penetration on the estimation of ocity effects on piezocone tests in normally and over-
engineering properties and liquefaction potential of silty consolidated clays. Int. Journal of physical Modelling 7
and sandy soils. Journal of Geotechnical and Geoenvir (2): 23–34.
onmental Engineering 142(11): 4016059. Schneider, J.A., Randolph, M.F., Mayne, P.W., &
Krage, C.P., Broussard, N.S. & DeJong, J.T. 2014. Estimat Ramsey, N.R. 2008. Analysis of factors influencing soil
ing rigidity index (IR) based on CPT measurements. 3rd classification using normalized piezocone tip resistance
International Symposium on Cone Penetration Testing and pore pressure parameters. Journal of Geotechnical
(CPT14): 727–735. and Geoenvironmental Engineering 134(11):
Lunne T., Robertson P. & Powell J. 1997. CPT in geotech 1569–1586.
nical practice. New York: Blackie Academic. Senneset K., Sandven R., Lunne T., By T. & Amundsen T.
Naeini, S. & Moayed, R. 2007.Evaluation of undrained shear 1988. Piezocone tests in silty soils. In De Ruiter (ed.),
strength of loose silty sands using CPT results. Inter ISOPT-1, Orlando, FL: 955–966. Rotterdam: Balkema.
national Journal of Civil Engineering 5(2): 104–117. Tonni, L. & Simonini, P. 2013. Shear wave velocity as
Paniagua, P., Carroll, R., L’Heureux, J.S. & Nordal, S. 2016. function of cone penetration test measurements in sand
Monotonic and dilatory excess pore water dissipations in and silt mixtures. Engineering Geology 163: 55–67.
281
© 2022 the Author(s), ISBN 978-1-032-31259-0
Numerical investigation of piezocone dissipation tests in clay: Sensitivity of

interpreted coefficient of consolidation to rigidity index selection
A. Barati Nia, D.M. Moug & A.P. Huffman
Department of Civil and Environmental Engineering, Portland State University, Portland, USA
J.T. DeJong
Department of Civil and Environmental Engineering, University of California, Davis, USA
ABSTRACT: Standard methods for interpreting the coefficient of consolidation (ch) from CPTu dissipation
tests require an estimation of the soil’s rigidity index (Ir=G/su). For the Teh & Houlsby numerically derived
dissipation solution, the values of shear modulus (G) and undrained shear strength (su) were straight forward to
specify due to the use of the elastic-perfectly plastic Von-Mises soil constitutive model. However, estimation of
appropriate G and su values is not as straight forward for field dissipation tests. The objective of this study is to
examine the sensitivity of the interpreted ch to various approaches for estimating Ir. The study is performed with
an axisymmetric direct cone penetration model to simulate piezocone dissipation. Simulations are performed
with the MIT-S1 constitutive model calibrated for Boston blue clay behavior. Analyses examine how the sensi
tivity changes for normally to lightly overconsolidated clay, and slightly to strongly anisotropic hydraulic con
ductivities. The results indicate that ch interpretation is not highly sensitive to Ir estimation when Ir is within
standard values for clay; however, the sensitivity appears to increase as the overconsolidation ratio increases.
1 INTRODUCTION Ir that was empirically determined to closely agree

with the model data. The time to reach 50% dissipa
The piezocone (CPTu) dissipation test is often used as tion at the u2 position (t50) is often used because it
an in situ method to estimate the coefficient of con provides sufficient dissipation data without requiring
solidation in the horizontal direction (ch) and hydraulic large dissipation times in the field. An example esti
conductivity of fine-grained soils. The CPTu dissipa mation of t50 from a CPTu u2 dissipation curve is
tion test records excess pore water pressure (Δu) at the shown in Figure 1. For t50, T* is equal to 0.245.
cone shoulder (u2 position) as porewater pressure
returns to the hydrostatic condition during a pause in
penetration. The dissipation response is primarily con
trolled by the initial distribution of Δu around the cone
and ch, among other factors.
Various researchers have proposed methods for
estimating ch; however, the Teh and Houlsby (1991)
method remains widely used:
where T* is a modified time factor; r is the radius of

the piezocone; Ir is the soil rigidity index; and t is the
time to a degree of dissipation corresponding to T*.
The Teh & Houlsby (1991) method was developed
with the strain path method and Von-Mises soil model
to estimate the initial Δu distribution, then Δu dissipa Figure 1. Comparison of a normalized field-measured u2
tion was modeled with Terzaghi consolidation theory. dissipation curve in BBC and a simulated curve with
This method for interpreting ch uses the square root of MIT-S1 calibrated for BBC.
DOI: 10.1201/9781003308829-36
282
In addition to Teh & Houlsby (1991), other dissipation is simulated to capture the full response
methods to interpret ch from dissipation tests require around the cone during a CPTu dissipation test.
an Ir value (e.g., Burns & Mayne, 1998, Chai et al.,
2012). Ir is the ratio of shear modulus (G) to the
undrained shear strength (su). Therefore, a single Ir
value represents average G and su values for the com
plex loading, strain, and porewater conditions around
the penetrating cone. For the Teh and Houlsby (1991)
numerically derived dissipation solution, the values of
G and su were straightforward to specify due to the
use of the elastic-perfectly plastic Von Mises soil con
stitutive model. However, natural soil behaviors
include strain softening, non-linear behavior, and
anisotropic strengths. As a result, the selection of
appropriate G and su values is not straightforward.
There are three standard approaches for estimating
Ir: laboratories tests, CPT-based, and empirical rela
tionships. Although laboratory shear testing on intact
soil specimens is a direct method to estimate Ir, this Figure 2. Simulated u2 dissipation for OCR = 1 with
approach typically requires the most effort and expense kh/kb = 2 and 10, and OCR = 2 BBC with kh/kv = 2.
compared to CPT-based and empirical estimates.
Accordingly, empirical methods or CPT-based Simulated u2 dissipation for normally consolidated
approaches are frequently used to estimate Ir values. BBC is presented in Figure 1 and Figure 2. The initial
Given the various approaches, there remains uncer conditions are vertical effective stress (σ’vo) of 100
tainty regarding how sensitive interpreted ch values are kPa, horizontal effective stress of 50 kPa, and hydro
to the estimated Ir. static porewater pressure of 100 kPa. A summary of
For this study, different methods for estimating these initial conditions and simulation results for
Ir will be presented and used to estimate Ir values undrained penetration are presented in Table 1. The
for Boston blue clay (BBC). The various Ir values horizontal hydraulic conductivity (kh) for dissipation
will be used to interpret ch from simulated dissipa is 2x10-7m/s, and is two times larger than the vertical
tion tests in normally and lightly overconsolidated hydraulic conductivity (kv). Although these values of
BBC and for slightly anisotropic and strongly kh and kv are at least an order of magnitude higher
anisotropic hydraulic conductivities. The inter than typical values for clayey soils (Kulhawy &
preted ch values from the simulations will be com Mayne, 1990), the objectives of this study are not
pared to the ch values estimated from assigned compromised since dissipation is simulated from
model properties. undrained conditions. In Figure 1, the u2 dissipation
curve is normalized by the initial Δu and t50 to com
pare it to a field-measured dissipation test in BBC
2 SIMULATED PIEZOCONE DISSIPATION IN (Baligh & Levandoux, 1986). The normalized simu
BOSTON BLUE CLAY lated dissipation results in OCR=1 of BBC agree well
with the field-measured results.
CPTu dissipation was simulated with the finite differ
ence program FLAC (Itasca, 2019) and the MIT-S1
constitutive model (Pestana & Whittle, 1999). MIT
Table 1. Initial conditions and results for simulated
S1 is a bounding surface plasticity model that can undrained cone penetration in BBC with OCR=1.
capture the anisotropic undrained strength behavior of
clays, which is advantageous for direct cone penetra Value
tion simulations (Moug et al. 2019). To capture CPTu
dissipation, first, undrained steady state penetration Initial conditions
was simulated with MIT-S1 calibrated for BBC Pore pressure, u0 100 kPa
behavior. The penetration model is a direct axisym Vertical effective stress, σ’vo 100 kPa
metric model that uses an Arbitrary Lagrangian Euler Horizontal effective stress, σ’ho 50 kPa
ian rezoning and remapping algorithm to Simulated penetration results
accommodate large deformations around the penetrat Cone tip resistance, qt 355 kPa
ing cone. A description of the model and implementa Pore pressure at the cone shoulder, u2 259 kPa
tion with the MIT-S1 model, including validation for
undrained penetration in BBC, is published in Moug
et al. (2019). After reaching steady-state stress and 3 RIGIDITY INDEX ESTIMATION METHODS
porewater pressure conditions, at approximately 25
cone diameters of simulated penetration, penetration Estimation of ch relies on estimated Ir values. Because
is paused and brought to static conditions. Then, Δu Ir values for natural clay deposits may vary between
283
50 to 600, the ch estimation can be different by 50% of the strain level required to reach the yield
a factor of up to 4 for a given clay (Schnaid et al., stress. Past researchers estimated G at 50% of the
1997). This variability of Ir may reflect changes due yield stress (Keaveny & Mitchell, 1986; Schnaid
to OCR and PI; however, there may also be differ et al., 1997); however, this approach was overly stiff
ences in Ir values depending on the method used to for the simulated results. This likely reflects sample
estimate the value. In the following subsections, dif disturbance effects from obtaining and preparing the
ferent approaches for estimating Ir are briefly intro sample that are not captured in the simulated
duced and used to obtain Ir values for OCR = 1 BBC. response. An examination of published CIUC and
CAUC tests for BBC (Landon, 2007) indicated there
was about a 35% difference between Ir values esti
3.1 Laboratory test approaches
mated with 50% of strain and 50% of stress to yield
Laboratory testing of intact soil samples is a direct stress. Consequently, 50% of the strain level to yield
method for estimating Ir. However, uncertainties stress was considered for G50 to calculate Ir values
remain for estimating Ir from laboratory shear tests. su from CIUC and CAUC.
depends on initial stress state, loading conditions,
loading rate, stress history, degree of fissuring, bound
3.2 CPT-based approaches
ary conditions, etc. For example, one clay tested with
isotropically consolidated undrained triaxial compres Several methods have been proposed to estimate Ir
sion (CIUC) and anisotropically consolidated values from CPTu test data. These approaches have
undrained triaxial compression (CAUC) will likely the advantage that the required data to estimate Ir are
yield different su between the two tests, reflecting the collected during the process of performing CPTu dis
different initial consolidation conditions. Ir interpreted sipation test.
from CAUC tests was used for several research studies Krage et al. (2014) analyzed a database of CAUC
that examined dissipation test interpretation (Krage test results with corresponding CPTu data to develop
et al., 2014; Schnaid et al. 1997); however, in practice, two approaches to estimate Ir. The two approaches
CIUC is a more common and simple to perform a test. are called “method A” (IrA) and “method B” (IrB).
Ir values also depend on G estimation. The secant Both methods use a functional reduction of Gmax to
shear modulus at 50% of the yield stress (G50) and approximate G50, where G50/Gmax = 0.26 and Gmax
25% of the yield stress (G25) were investigated by is estimated from Vs measurements obtained from
Schnaid et al. (1997) for dissipation test interpret the seismic CPT profile. Method A estimates su from
ation. G50 was considered by Schnaid et al. (1997) to laboratory testing. Consequently:
reasonably represent the stress and strain levels in
the vicinity of the cone.
For estimation of Ir in this study, CIUC and
CAUC tests were simulated with single element
models in FLAC and the MIT-S1 model calibrated
for BBC. The simulated CIUC and CAUC results for Since this relationship was developed with CAUC
deviatoric stress (q) versus deviatoric strain (εq) are laboratory test results, the simulated CAUC su value
shown in Figure 3. The values of Ir from the stress- was used to estimate IrA for this study.
strain paths in Figure 3 are summarized in Table 2. Method B from Krage et al. (2014) generalizes
The G values for estimating Ir from CIUC and Method A with a CPT-based approximation of su.
CAUC single element simulations were calculated at The estimation of su is based on relationships
between qt and OCR (i.e., Chen & Mayne, 1994)
and SHANSEP principles of su normalization by
σ’vo and OCR. The estimation of IrB is:
Using the simulated qt value from penetration in nor

mally consolidated BBC resulted in OCR < 1. There
fore, the estimation of IrB used OCR = 1 as opposed
to the qt-estimated OCR in the denominator of Equa
tion 3. Gmax was approximated from MIT-S1 stiff
ness parameters, which were selected in Moug et al.
(2019) to be consistent with shear wave velocity
Figure 3. Single element CIUC & CAUC simulations for
OCR=1 BBC. data in BBC (Landon 2007).
284
Mayne (2001) developed a theory-based estima- The average PI of BBC was reported in Landon
tion of Ir with spherical cavity expansion and critical (2007) as 20. Therefore, for OCR = 1 BBC, Ir = 162
state soil mechanics (SCE-CSSM): by the Keaveny & Mitchell (1986) estimation.
Młynarek et al. (2018) performed multivariate
regression analysis to investigate relationships
between Ir and PI, OCR, and liquidity index (LI) for
different soil groups such as overconsolidated &
aged clay, organic & young clay, aged till and young
till. Although the analysis was performed with
in which M = (6sinϕ′)/(3-sinϕ′) and ϕ’ is the effective
regionally-limited soils and one of the authors’ con-
friction angle. M was determined from the critical
clusions was that Ir values and relationships should
state ϕ’ for the MIT-S1 calibration for BBC (Pestana
be regionally developed, the relationship for organic
et al., 2002), which resulted in an M value of 1.35.
and young clays is included herein for comparison:
For OCR = 1 BBC, a LI of 1.14 was used based on

data in Landon (2007), resulting in Ir = 226.
3.4 Summary of estimated Ir

A summary of the estimated Ir values for OCR = 1
BBC is provided in Table 2 and Figure 4. Generally,
the values are consistent with typical Ir for OCR = 1
clay. The values estimated from single element simula
tions of CIUC and CAUC tests do not show a large
difference: 212 and 261 for CIUC and CAUC, respect
ively. The Ir values estimated with Krage et al. (2014)
methods are comparable between themselves but are
larger than the laboratory-estimated values, with IrA =
329 and IrB = 315. As discussed above, the Ir value
estimated with Mayne (2001) results in a notably
Figure 4. Comparison of Ir values estimated with different lower value than the other estimation methods but is
methods for OCR=1 and OCR = 2 BBC.
sensitive to small changes in qt or u2. Ir estimated with
empirical methods yielded values of 162 and 226 for
Keaveny & Mitchell (1986) and Młynarek et al.
The Mayne (2001) approximation of Ir resulted in (2018), respectively. These estimates are consistent
a low value of 36. However, it should be noted that with typical values for clay; the Keaveny & Mitchell
this estimate of Ir is relatively sensitive to changes in (1986) relationship estimates an Ir value lower than the
qt and u2. For example, a 10% increase in qt results laboratory-based estimates, while Młynarek et al.
in an Ir value that is four times larger, or a 10% (2018) is consistent with the laboratory-based
increase in u2 results in an Ir value that is over 10 estimates.
times larger than the original estimate.
3.3 Empirical relationships

Table 2. Ir values for OCR=1 BBC estimated with various
Empirical relationships are used to estimate Ir from methods.
soil index properties and/or OCR. Keaveny & Mitch-
ell (1986) observed that for ordinary clays, Ir Method Ir
increases by decreasing OCR and decreasing plasti- Simulated CIUC 212
city index (PI). Mayne (2001) provided an equation- Simulated CAUC 261
based estimate of the Keaveny & Mitchell (1986) Krage et al. (2014) method A 329
relationships with: Krage et al. (2014) method B 315
Mayne (2001) 36
Keaveny & Mitchell (1986) 162
Młynarek et al (2018) 226
285
4 INTERPRETED ch SENSITIVITY To Ir
ch values from a simulated CPTu u2 dissipation test in

OCR = 1 BBC are estimated using the various Ir
values presented in the previous section and with
Equation 1. The dissipation test was performed for
BBC with a kh/kv = 2, representing slightly anisotropic
hydraulic conductivity conditions. The simulated u2
dissipation curve is shown in Figure 2, which has a t50
of 32 seconds. The ch values interpreted from simu
lated u2 dissipation (ch,interpreted) are compared to ch
values assigned to the cone penetration model
(ch,model). This comparison is shown in Figure 5.
The range of ch,interpreted/ch,model in Figure 5 is
between 0.56 and 1.69 for the various Ir estimation
methods. Generally, the sensitivity of ch,interpreted
/ch,model to the estimated
ffi Ir is low since ch,interpreted is
proportional to in the Teh & Houlsby (1991) rela
tionship. It should be noted that estimating Ir with
CIUC or CAUC does not significantly impact
ch,interpreted indicating that there may not be a strong Figure 5. Comparison of ch estimated from simulated
advantage of one test condition over the other for this results and different Ir with ch assigned to the dissipation
application. Additionally, across the various Ir values, models for OCR = 1, OCR = 2, and kh/kv = 2 and 10 BBC.
the variability of ch,interpreted appears to be within
a typical range due to the variability of soil properties,
estimation of t50, and methods for interpreting ch (e.g.,
Huffman & Moug, 2022).
Comparing ch,interpreted to ch,model for OCR = 1 and
OCR = 2, indicates that as OCR increases, the sensi
5 THE EFFECT OF PERMEABILITY tivity of ch,interpreted to estimated Ir increases. The
ANISOTROPY AND OCR ON ch ESTIMATION ratios of ch,interpreted and ch, model for OCR = 2 are in
some cases twice as large as the ratio for OCR = 1
The sensitivity of ch to Ir is further explored by exam with the same hydraulic conditions, with the excep
ining dissipation in lightly overconsolidated and tion of ch,interpreted using the Ir values from Mayne
highly anisotropic BBC. To this end, u2 dissipation (2001). For OCR = 2 the ratio of ch,interpreted to
tests were also performed for OCR = 1 with kh/kv = ch,model ranged from 0.44 to 2.45.
10; and OCR = 2 with kh/kv = 2. These u2 dissipation
curves are shown in Figure 2. These simulated results
show that although the kh for both OCR = 1 and 6 CONCLUSIONS
OCR =2 models is the same, t50 decreases as OCR
increases from OCR = 1 to OCR = 2. This may be This study examined the sensitivity of CPTu dissipa
attributed to a smaller Δu distribution around the tion test interpretation to estimated Ir values. For
penetrating cone as OCR increases and increasing a given soil, there are various approaches to estimate
soil stiffness as OCR increases and, therefore, less Ir. Simulated CPTu dissipation tests in BBC were
porewater pressure migration to dissipate the Δu field. used to examine how the interpreted ch from CPTu
Estimated Ir values for OCR = 2 BBC are com dissipation tests changes with different methods to
pared to those estimated in the previous section for estimate Ir. ch was interpreted from simulated u2 dis
OCR = 1 in Figure 4. Although the differences in Ir sipation following undrained cone penetration using
among different approaches are still considerable, the Teh & Houlsby (1991) method, which requires
the differences between OCR = 1 and OCR = 2 for an estimation of Ir. Ir was estimated with several
every method are relatively small. approaches that represent laboratory test-based,
The comparison between ch,interpreted with the vari CPT-based, and empirical-based approaches. The
ous Ir values and ch,model is shown in Figure 5. Com sensitivity of ch to various Ir estimations was evalu
paring the ratio of ch, interpreted to ch, model for OCR = ated for BBC with OCR = 1 with slightly anisotropic
1 with different hydraulic conductivity anisotropy, hydraulic conductivity, OCR = 1 with strongly aniso
the differences between kh/kv = 2 and 10 across the tropic hydraulic conductivity, and OCR = 2 with
various Ir estimation methods are similar. The ratio slightly anisotropic hydraulic conductivity. Ultim
of ch,interpreted to ch,model for kh/kv =10 ranged from ately, interpreted ch was compared with the modeled
0.48 to 1.46. This indicates that ch estimation using ch to assess the variability of estimated ch over
various Ir estimation methods is not strongly affected a range of reasonable Ir values. Although this study
by hydraulic conductivity anisotropy. was performed for a single soil, it does provide some
286
insight into the role of Ir estimation on CPTu dissipa Burns, S. E., & Mayne, P. W. (1998). Monotonic and dila
tion test interpretation. The results support the fol tory pore-pressure decay during piezocone tests in clay.
lowing conclusions: Canadian Geotechnical Journal, 35(6), 1063–1073.
https://dx.doi.org/10.1139/cgj-35-6-1063
1. Due to a square root of Ir in the Teh & Houlsby Chai, J., Sheng, D., Carter, J. P., & Zhu, H. (2012). Coeffi
(1991) method (Equation 1), the sensitivity of ch cient of consolidation from non-standard piezocone dis
to Ir is relatively low. For OCR = 1 BBC, the sipation curves. Computers and Geotechnics, 41, 13–22.
range of Ir estimates was large: between 36 and https://dx.doi.org/10.1016/j.compgeo.2011.11.005
329. However, the ch,interpreted values ranged from Chen, B., & Mayne, P. (1994). Profiling the overconsolida
0.56 to 1.69 times the ch,model with slightly aniso tion ratio of clays by piezocone tests. In Rep. No. GIT
tropic hydraulic conductivity. When the hydraulic CEEGEO-94 (p. 294). https://scholar.google.com/scholar?
hl=en&btnG=Search&q=intitle:Profiling+the+Overconsoli
conductivity was strongly anisotropic, the
dation+Ratio+of+Clays+by+Piezocone+Tests#0
ch,interpreted values ranged from 0.48 to 1.46 times Huffman, A. P. E., & Moug, D. M. (2022). Interpretation of
ch,model. However, it should be noted that ch inter Field-Measured and Simulated Non-Monotonic CPTu
pretation methods may also contribute to discrep Dissipation Tests. For Proceedings of the 2022 ASCE
ancies between ch,interpreted and ch,model, in GeoCongress.
addition to the estimated Ir value. Itasca. (2019). Fast Lagrangian Analysis of Continua (8.1).
2. Increasing the OCR from 1 to 2 appeared to Itasca Consulting Group, Inc.
result in ch,interpreted values that were increasingly Keaveny, J. M., & Mitchell, K. (1986). Strength of fine-
sensitive to the estimated Ir values. Estimated Ir grained soils using the piezocone. In S. P. Clemence
(Ed.), Use of In siti tests in geotechnical engineering
values for OCR = 2 ranged from 11 to 343, and
(pp. 668–685). ASCE.
ch,interpreted values were between 0.44 and 2.45 Krage, C. P., Broussard, N. S., & Dejong, J. T. (2014). Esti
times larger than ch,model. The larger discrepancy mating rigidity index (IR) based on CPT measurements.
between ch,interpreted and ch,model for OCR = 2 3rd International Symposium on Cone Penetration Test
compared to OCR = 1 may be due to a smaller ing (CPT14), Teh 1987, 727–735.
Δu distribution around the penetrating cone and Kulhawy, F. H., & Mayne, P. W. (1990). Manual on Esti
increased soil stiffness, resulting in faster dissipa mating Soil Properties for Foundation Design. In Osti
tion and a smaller t50. The smaller t50 leads to gov (p. 299). https://www.osti.gov/energycitations/
more sensitivity to Ir even though the range of product.biblio.jsp?osti_id=6653074
Landon, M. M. (2007). Development of a non-destructive
estimated Ir values was smaller.
sample quality assessment method for soft clays. Uni
3. Minor differences were observed between esti versity of Massachusetts Amherst.
mated ch values using the Ir values from CIUC Mayne, P. W. (2001). Stress-strain-strength-flow parameters
and CAUC testing in OCR = 1 and OCR = 2 from enhanced in-situ tests. International Conference on
BBC. Because CIUC is considered a more simple In-Situ Measurement of Soil Properties & Case Histor
test to perform, with further investigation, it may ies, 27–48.
be considered reasonable to estimate Ir values for Młynarek, Z., Wierzbicki, J., & Stefaniak, K. (2018).
CPTu dissipation test interpretation. Rigidity index (Ir) of soils of various origin from CPTU
and SDMT tests. Cone Penetration Testing 2018 - Pro
ceedings of the 4th International Symposium on Cone
Penetration Testing, CPT 2018, 441–446.
ACKNOWLEDGEMENTS Moug, D. M., Boulanger, R. W., DeJong, J. T., &
Jaeger, R. A. (2019). Axisymmetric Simulations of Cone
Funding for this research was provided by the Penetration in Saturated Clay. Journal of Geotechnical
National Science Foundation (award CMMI and Geoenvironmental Engineering, 145(4), 04019008.
https://dx.doi.org/10.1061/(asce)gt.1943-5606.0002024
1927557). Any opinions, findings, conclusions, or
Pestana, J. M., & Whittle, A. J. (1999). Formulation of
recommendations expressed in this material are a unified constitutive model for clays and sands. Inter
those of the authors and do not necessarily reflect the national Journal for Numerical and Analytical
views of the NSF. Methods in Geomechanics, 23(12), 1215–1243.
https://dx.doi.org/10.1002/(SICI)1096-9853(199910)
23:12<1215::AID-NAG29>3.0.CO;2-F
REFERENCES Pestana, J. M., Whittle, A. J., & Gens, A. (2002). Evalu
ation of a constitutive model for clays and sands: Part II
Agaiby, S. S., & Mayne, P. W. (2018). Evaluating - Clay behaviour. International Journal for Numerical
undrained rigidity index of clays from piezocone data. and Analytical Methods in Geomechanics, 26(11),
Cone Penetration Testing 2018 - Proceedings of the 4th 1123–1146. https://dx.doi.org/10.1002/nag.238
International Symposium on Cone Penetration Testing, Schnaid, F., Sills, G. C., Soares, J. M., & Nyirenda, Z.
CPT 2018, 1985, 65–71. (1997). Predictions of the coefficient of consolidation
Baligh, M. M., & Levandoux, J.-N. (1986). Consolidation from piezocone tests. Canadian Geotechnical Journal,
after undrained piezocone penetration. II: Interpretation. 34(2), 315–327. http://dx.doi.org/10.1139/t96-112
Journal of Geotechnical Engineering, 112(7), 727–745. Teh, C. I., & Houlsby, G. T. (1991). An analytical study of
https://dx.doi.org/10.1061/(ASCE)0733-9410(1986) the cone penetration test in clay. Geotechnique, 41(1),
112:7(727) 17–37. https://dx.doi.org/10.1680/geot.1991.41.1.17
287
© 2022 the Author(s), ISBN 978-1-032-31259-0
Correlation of CPT measurements and VibroCore penetration speed for

medium, calcareous sands: A case study of the cable route survey at the
North Sea
K. Bartczak & G. De Vries
Marine Sampling Holland B.V., The Netherlands
ABSTRACT: The Cone Penetration Test (CPT) is one of the most popular in-situ testing methods used in
qualitative and quantitative research of characteristics of the subsurface sediments, both onshore and offshore.
In offshore cable route surveys, CPT tests are often accompanied by VibroCoring (VC) boreholes. The article
endeavours to verify whether there is a clear correlation between the relative density of cohesion-less soils
and the VC penetration characteristics. The research has been based on an offshore site investigation cam
paign in the North Sea consisting of over 100 CPTs and VCs, supplemented by an extensive laboratory testing
program. After a strict selection of locations, around one-fifth of those tests have been utilised in the correl
ation studies. The ultimate goal of this study was to demonstrate if processing data recorded during VC testing
can increase certainty in the prediction of soil’s strength parameters.
1 INTRODUCTION possible on the grounds of the laboratory testing pro

gram, which comprised both standard index tests and
VibroCoring is a technique of extracting a sediment advanced testing.
core from the seabed, in which a steel tube with an
inner plastic liner is vibrated into the seabed by the
action of two counter-rotating eccentric weights 2 EVALUATION OF RELATIVE DENSITY
driven by an electric- or hydraulic motor. To collect
the fieldwork data used in this paper, a slightly modi 2.1 Introduction
fied VibroCorer Viking VKG-6 was used. The corer
is characterized by 30 kN of vibrating force, vibrating The chosen 33 locations create a triangle-shaped area
frequency of 28 Hz, the inside diameter of the core around 7×10 km big. Elevation of around one-third of
barrel of 96 mm and its net length of 6:0 m. The those was between -10 to -20 metres Lowest Astro
penetration is measured employing sensors and mag nomical Tide (LAT), while for the other two-thirds, the
nets placed each 5:0 centimetres along the barrel. The range was between -25 to -35 metres LAT. The
data is logged with a frequency of 10 Hz. In common usage of tests with highly varying elevations might be
practice, a live view of the current penetration is used problematic when analysing and correlating the VC
to determine if the penetration is properly succeeding and CPT data. As each test starts at a slightly different
and whether it should be already terminated. elevation, using many tests is marked with noise and
This paper poses the question whether records of peaks in data, which affect the latter interpretation.
this data can add value to the data obtained from pro Therefore, the analysis on CPT and VC data sets was
cessing the cores and possibly other tests, such as made only as a function of depth. To simplify the ana
CPTs, performed at the site. A database of over 100 lysis, the 33 picked locations were divided into seven
CPTs and VCs, which were a part of the cable route groups in a manner to obtain the most uniform soil pro
survey in the North Sea, was utilised in this research. file in each group.
From all the tests, a lot of 33 locations characterised
by homogeneous stratigraphy was selected for the 2.2 Laboratory testing data
studies. The soil profile mainly consisted of calcareous,
medium sands in the medium-dense to very dense The determination of a lab-based Relative Density
states. In two groups, there was an underlying clay (RD) profile was made on a selection of lab testing
layer starting at around 4:0 to 4:5 m deep. Field testing results consisting of 22 minimum and maximum
of each location comprised of both VC and CPT tests. density tests and 161 dry density tests. The resulting
Correlation of CPT results and relative density was mean (and standard deviation) of RD value was
DOI: 10.1201/9781003308829-37
288
equal to 69% (32%) for all 150 tests, while for the
selected locations 73% (33%). The evaluated RD
profile is shown in Figure 1 below.
Figure 2. CPT Cone resistance in Depth.
Figure 1. Lab-based relative density in a function of

elevation.
For plotting the depth-trend (solid line), the Python

implementation (SciPy) of Savitzy-Golay (S-G) Filter
(Savitzky & Golay 1964) has been used. For smooth
ing the VC data presented in the latter part of this
paper, a polynomial interpolation algorithm had to be
found. The S-G filter has been chosen as the best solu
tion, and the same algorithm was proposed for the lab-
based RD to avoid extra noise in further analysis. The
window length (M), and the polynomial degree (N),
were chosen via an iterative process of minimization
values of the cost function, i.e. Mean Square Error Figure 3. CPT- and LAB-based RD in-depth for two
(MSE) between the lab data points and smoothed selected groups.
values.
2.3 CPT data

3 VIBROCORE DATA INTERPRETATION
In Figure 2, an average value of CPT-tip resistance is
plotted for each of the seven groups. Several correl
ations for CPT-based RD have been tested to find the 3.1 Data clearing
best fit with laboratory-based values. The one that In these specific soil conditions, the core barrel’s full
provided the best fit was the one proposed for penetration was reached after around 2 minutes of
Ticino, Toyoura, Hokkusund Sand (Jamiolkowski, vibrations. The penetration data recording frequency
Lo Presti, & Manassero 2003) presented in Equa was 10Hz, which gives over 1200 readings per 500
tion 1 below. centimetres of penetration on average. As the reso
lution of depth measurement is 1.67 cm, this means
that each centimetre of penetration has been logged 4
times, which results in high accuracy, but also a very
high level of noise (penetration readings tend to
“jump” up and down by over a dozen of centimetres).
where: C0 ¼ 0:175; C1 ¼ 0:500 and C2 ¼ 3:1. Other aspects that induced the need for rigorous data
The fit between the correlation mentioned above clearing include turning the vibrating motor off and on
and the laboratory-based data (visibly rougher again in case of an obstacle or penetration hindrance
curves) for two selected (and the biggest) groups is and the free-fall penetration of the barrel when it
shown in Figure 3. touches the seabed.
289
Therefore, VC data had to be cleared from the top m (3:70 m for the 5th and 7th group because of the
20 centimetres of data classified as unreliable and underlying clay layer). The statistical relationship
then filtered according to the steps shown below. measured with the Pearson’s coefficient has been sum
The clearing steps explained below correspond to marised in Table 1 below.
the numbers of VC lines in Figure 4.
1. If the consecutive penetration reading is smaller Table 1. Correlation studies between CPT-based Rd and
than the previous one - the previous value is kept VC SP .
and the smaller one ignored. This prevents from
appearing of any negative penetration speed values; Group ρ (0:2 - 5:0m) ρ (0:2 - 1:7m) ρ (1:7 - 5:0m)
2. Exponential Smoothing (Statsmodels.org, Hynd
man2018) of penetration readings with dumping 1 -0.59 -0.44 0.44
factor α ¼ 0:15; 2 -0.91 -0.68 -0.27
3. Applying the Savitzy-Golay Filter (SciPy) with 3 -0.73 -0.20 -0.51
N ¼ 7 and M of 53 for VC penetration speed, SP , 4 -0.75 0.04 -0.20
above 0:075 m/s and 203 for penetration speed 5 -0.68 -0.07 -0.51
below 0:075 m/s along at least 50 consecutive 6 -0.93 -0.76 -0.63
readings. 7 -0.78 -0.63 -0.62
AVG -0.77 -0.39 -0.33
Based on the correlation studies of VC penetration

speed, SP , and CPT-based RD, Rd , an important con
clusion has been drawn that the correlation is not con
stant in-depth. Although the correlation for the whole
penetration dataset bracket together was strong
(ρ ¼ -0:77), it is not the case when the data is separ
ated. The correlation is better in the upper section,
apart from two groups that showed no correlation. In
the upper section (until 1:7m of penetration on aver
age), the strong negative correlation is severely
affected by an initial uneven penetration speed of
a “wavy” nature. Even though the data has been thor
oughly cleaned and filtered, the noise generated by
vibrations does not allow formulating a clear and
undoubtful correlation in each test. The correlation in
the bottom section could be improved by introducing
Figure 4. CPT- and LAB-based RD and VC penetration an additional parameter, which could be based on, for
speed in depth - chosen test from Group 3. example, a net surcharge or penetration depth.
3.3 VC-based RD formula

The reason for choosing the S-G filter was a trial
and-error process of testing several algorithms with To verify what mechanisms govern VC penetration
decision criteria including lack of distortion of readings speed and whether it is possible to correlate it to the
in-depth tendency, and correct depiction of average non-cohesive soils’ RD, physical aspects linking
value at each depth with no severe over or underesti those parameters need to be researched on.
mation. In addition, the S-G filter has two convolution The relationship between the cone tip resistance,
coefficients, which iterative customisation allows redu qc , and the relative density, Dr , was initially pro
cing the MSE between the original data-set and the posed by Lancellota (Lancellotta 1983) as shown in
fitted curve. Equation 2 below:
3.2 Relationship between CPT-based RD and VC

penetration speed
where: X ¼ ln½ðσ0qv0c Þα ] A0 , B0 and α = empirical cor
The relationship between the CPT and VC data sets is relation factors.
depicted in Figure 5, where the CPT-based RD and It is assumed that the VC-based RD should be for
VC penetration speed are plotted together for the mulated in the same way, but with the cone resist
chosen group. The correlation between the mean in- ance, qc , replaced by the VC Penetration Resistance
depth values of those data sets is measured with Pear qVC . The qVC parameter should depend on the depth
son’s coefficient, ρ. The correlation studies are based of penetration and the penetration speed, which is
only on the depth range between 0:20 and 5:00 mainly governed by:
290
Figure 5. CPT-based Relative Density and VC penetration speed and their correlation in Depth for Group 6.
– Liquefaction susceptibility, which mainly depends

on relative density and grain size distribution and
uniformity coefficient (Saikia & Chetia 2014);
– Vibrations frequency and intensity;
– Frictional resistance, which depends on the depth
of the penetration, angle of friction at the inter
face between a cohesion-less soil and a steel core
barrel, friction coefficient and water content
(Mosaddeghi, Hosseini, Hajabbasi, Sabzalian, &
Sepehri 2021), is assumed to be constant.
The intensity and frequency of vibrations generated
by the motor are approximately constant; however, the
vibrations transmitted along the barrel are dumped by
the medium surrounding the barrel, i.e. firstly by the
water and later the soil. Hence, the deeper the barrel is
penetrated, the more vibrations dissipate to the soil
above the cutting shoe at the bottom of the barrel,
Figure 6. Histograms of qc and SP from all 33 tests for
limiting the liquefaction occurrence. depths from 0.20 to 5.00 m.
The dependence of penetration resistance, F, on
velocity for non-deforming projectiles is represented where Li is incremental barrel penetration, d is bar
by the Poncelet equation (Bless, Omidvar, & Iskan rel’s diameter, Ks is coefficient of earth pressure, δ is
der 2018), which is a relatively similar problem in angle of friction at interface and σ0 v is effective over
physical meaning and is shown below. burden pressure. Summing up all the relationships
listed above, following dependency is proposed:
where D2 is a correlation factor and is negative,

where M is a projectile mass, v is penetration velocity, fðLi Þ is a function that links difficulty in liquefying
ρ is the bulk density of sand, C is Poncelet drag, soil beneath the cutting shoe with the depth of pene
which ranges from 0.7 for loose to 1.1 for dense sand tration, and D3 is an empirical factor that is positive.
(Bless, Omidvar, & Iskander 2018), R is bearing For a better understanding of those parameters’ rela
strength (frictional resistance) that for the core barrel tionship, the distributions of the tip resistance and
can be estimated with unit skin resistance equation: VC penetration speed are plotted in Figure 6.
The distribution of qc is almost uniform, with
approximately the same density in between 0 and 35
MPa. The distribution of penetration speed is
291
approximately log-normal. Therefore, the penetration 5 SUMMARY
speed cannot be incorporated linearly in the equation as
the tip resistance is. After substitution of qVC to Equa 5.1 Conclusions
tion 2 and a preliminary verification process, which
resulted in a few changes to the formula, the following This study was based on geotechnical data acquired
relation is proposed: from VC and CPT tests performed at 33 locations,
which were divided into 7 groups with very selected,
homogeneous stratigraphy consisting of medium-
dense to very dense medium sand.
The main conclusion is that the penetration speed,
where DR;VC is relative density in percents, SP2 is VC together with the effective vertical stress and pene
penetration speed in m/s, fðLi Þ is simplified to Li , tration depth, is not enough for the precise estima
i.e. the penetration depth expressed in meters, is tion of the relative density of cohesion-less soil.
effective vertical stress in kPa. Studies presented in this paper have shown that there
The proposed formula has been validated firstly is a strong correlation between VC penetration speed
with correlation factors same as in Equation 2, and and CPT-based relative density in the upper 1:50
employing iterative process, the constants were cali 2:00 meters of penetration. Furthermore, the For
brated to the following values: D0 = 60,D1 = 6.2, D2 mula 6 provided relative density estimations of the
= -0.45, D3 = 0.02. D4 = 0.2, D5 = 0.005, D6 = 0.6. correct order of magnitude and increased correlation
in most of the groups. However, the reliability and
accuracy of relative density estimation with the
4 VC-BASED FORMULA VALIDATION usage of VC data is very low and not enough for
applications in geotechnical design.
The formula proposed in the previous section has The main aspects negatively affecting the estima
been tested along with all seven groups. The main tions’ accuracy are the erratic penetration speed in the
problems that appeared in this process, and are limit upper section (between 0:0 - 1:7 m of depth) and
ing the reliability of the results, are as follows: the effects of exponentiation of penetration speed in the
bottom section, which gives wrong estimations of rela
– The value of D2 constant should be high to keep
tive density. One way of tackling those issues would be
the meaning of SP significant. However,if this
to divide the formula into two parts, i.e. one part for the
value is too high compared to D3 and D4 , the
penetration speed above 0:075 m/s and the other part
value in the logarithm becomes negative. This
for values below this threshold. Another way would be
gives a small threshold of reasonable D2 values;
to incorporate acceleration into the equation instead of
– The constant of D4 had to be introduced to minim
penetration speed or reduce the exponent from 2 to
ise the value second phrase of the nominator - with
a smaller number.
out it, the values of RD;VC in the deeper section of
the test were unrealistic;
– There are limitations connected with the fact that 5.2 Limitations
SP is to the power of 2. For low values of penetra There are several limitations related to the derived
tion speed, it results in practically no influence of correlation factors and formulas:
SP values on the computed value of RD;VC .
– Related to the coring technique and its physical
The exemplary results are shown in Figure 7 for meaning:
the same group as before. The correlation itself has
visibly increased, but the accuracy of RD estimations - Ignoring the friction of the core barrel in the
are still very low. The results have been summarised calculations. The friction raises with penetra
in Table 2 below, analogically to the previous Table 1. tion and should be incorporated in the calcula
tions in the other way than it is in CPT-based
formulas for relative density;
Table 2. Correlation studies between CPT-based Rd and
VC-based RD;VC .
- The influence of the net surcharge has been
calculated in the same way as in CPT-based
Group ρ (0:2 - 5:0m) ρ (0:2 - 1:7m) ρ (1:7 - 5:0m) formula;
1 0.72 0.85 -0.64 – Related to the data processing:

2 0.94 0.90 0.29
3 0.90 0.84 0.55 - The penetration data records have been cleared
4 0.69 0.15 0.14 out of not-significant records. Whether the
5 -0.03 0.52 0.79 record is significant or not is up to the engin
6 0.97 0.92 0.62 eering judgments, which is ambiguous. Due to
7 0.37 0.71 0.61 the high frequency of VC data logging (10
Hz), the records are noisy; hence, each user
AVG 0.65 0.70 0.35 would end up with different cleared data sets;
292
Figure 7. CPT-based Relative Density and VC-based Relative Density and their correlation in depth for Group 6.
- The filtering technique presented in Subsec optimisation algorithm integrating dynamic

tion 3.1 has better performance for tests with Bayesian networks could be of great help in this
penetration time longer than 30 seconds, matter.
while for quick sampling, the results are still
prone to errors;
- All limitations, which apply to Savitzky- REFERENCES
Golay Filter and its SciPy implementation,
apply to the results of this analysis as well. Bless, S. J., M. Omidvar, & M. Iskander (2018). Poncelet
coefficients for dry sand. AIP Conference Proceedings
1979(July), 1–6.
5.3 Recommendations for further research Jamiolkowski, M., D. C. F. Lo Presti, & M. Manassero
(2003, jan). Evaluation of Relative Density and Shear
As concluded above, the topic is promising in terms Strength of Sands from CPT and DMT. In J. Germaine,
of increasing certainty in estimations of soil param T. Sheahan, and R. Whitman (Eds.), Soil Behavior and
eters; however, the literature and research on this Soft Ground Construction, Reston, VA, pp. 201–238.
subject are very limited. Therefore, the following American Society of Civil Engineers.
concerns could be further elaborated on: Lancellotta, R. (1983). Analisi di Affidabilità in Ingegneria
Geotecnica”. In Atti dell’Istituto di Scienza delle Costru
– The phenomenon of the vibration-induced lique zioni, Volume 625, Torino. Politecnico di Torino.
faction underneath the core barrel and its influ Mosaddeghi, M. R., F. Hosseini, M. A. Hajabbasi,
ence on the penetration speed of the core barrel; M. R. Sabzalian, & M. Sepehri (2021). Chapter Two -
– Studies on relating the relative density not only to Epichloë spp. and Serendipita indica endophytic fungi:
the penetration speed but also the acceleration Functions in plant-soil relations. Volume 165, pp.
could bring new conclusions on the usage of VC 59–113. Academic Press.
data in the estimation of relative density; Saikia, R. & M. Chetia (2014). Critical Review on the
Parameters Influencing Liquefaction of Soils. 3(4),
– The relative and absolute error estimation should 110–116.
be incorporated into the correlation studies. Savitzky, A. & M. J. Golay (1964). Smoothing and Differ
– Calibration of coefficients in the formula relating entiation of Data by Simplified Least Squares Pro,
the VC penetration speed and relative density cedures. Analytical Chemistry 36(8), 1639–1643.
should be performed more thoroughly. An SciPy. Savitzky-Golay filter.
293
© 2022 the Author(s), ISBN 978-1-032-31259-0
CPT data interpretation for an improved characterization of the paleosol

stratigraphy in the Po River Valley, Italy
I. Bertolini, M. Marchi, L. Tonni & G. Gottardi
DICAM Department, University of Bologna, Italy
L. Bruno
Department of Chemical and Geological Sciences, University of Modena, Italy
A. Amorosi
Department of Biological, Geological and Environmental Sciences, University of Bologna, Italy
ABSTRACT: Post-depositional phenomena can produce significant changes in the geotechnical properties of
soils. Although the identification of the effect of such phenomena on the soil structure may contribute signifi
cantly to the interpretation of the soil mechanical behaviour, their characterization is not routinely carried out in
practice. In this context, the novel application of paleosol features in the definition of geotechnical stratigraphic
models can result in improved characterization of soil deposits in alluvial sites. Sedimentology uses paleosols
(i.e. fossil soils with peculiar geologic properties) as markers for stratigraphic correlations in alluvial areas at
large scale. This paper outlines the main properties of these geological objects and investigates the strategies for
their identification in cone penetration tests (CPT). The analyses are discussed with reference to boreholes and
CPTs data in 3 well documented sites in the Po River Valley (Italy). First, sedimentological identification of
paleosols from borehole corings is introduced; then, geotechnical interpretation of CPT is carried out on such
a basis, showing how poorly developed paleosols (Inceptisols) found in the investigated sites, are difficult to be
identified through CPT logs and how they plot in the well-established Robertson’s classification charts.
1 INTRODUCTION The stratal architecture of the Holocene succession

of the Po coastal plain in Figure 2 shows the spatial
The Po Plain-Adriatic sea system is an elongated basin distribution of the facies associations along the investi
originated from the Po River catchment. It consists of gated section. The Holocene deposits, down to a depth
a > 7km thick Pliocene - Quaternary sedimentary suc between -10 and -20 m a.s.l, consist predominantly of
cession, with a 30m-thick Holocene succession that a thick swamp clay succession that grades distally into
overlies Late Pleistocene deposits. The Holocene
depositional history can be investigated through inte
grated stratigraphic, sedimentological and paleonto
logical analyses. A total of 22 sedimentary facies,
grouped into 5 main depositional systems, have been
identified in the Po River catchment (Amorosi et al,
2017).
The present work focuses on three locations in
the Po Plain, in Province of Ferrara (Figure 1),
characterized by well-drained floodplain deposits,
which belong to the alluvial plain depositional
system. These deposits consist of different mix
tures of clay and silt and show peculiar paedo
genetic features: paleosols. The investigation
started from the description of three continuous
corings EM-S3, EM-S1 and EM-S5, located in
Gaibana, Dogato and Ostellato, respectively.
Then, three CPTU tests, each one performed in
proximity to one coring, were analysed. Figure 1. Location of the investigated site.
DOI: 10.1201/9781003308829-38
294
Figure 2. Stratal architecture with indication of the different facies association along section A-A in Figure 1.
lagoonal and bay-head delta deposits. Below the Holo close to the ground level and then, there is no vadose
cene succession, Pleistocene deposits are predomin zone. Due to the fact that river continues to supply
antly well-drained floodplain silts and clays with new sediments, soil formation is hindered. When the
subordinate fluvial-channel deposits (Amorosi et al, river enters a new stable phase of bed incision,
2017). renewed pedogenesis occurs on the interfluvial areas
(paleosols 2 and 3 in Figure 3 top).
Holocene paleosols are more discontinuous and
2 PALEOSOL ORIGIN AND DEVELOPMENT immature (Entisols, Amorosi et al., 2014). Paleosol
identification is a valid help in the stratigraphic ana
The floodplain facies association is characterized by lysis of mud-dominated Late Pleistocene and Holo
the presence of paleosols, buried ancient soils cene deposits. For this reason, they are largely used as
formed on a ground surface of the past and subjected regional stratigraphic markers (Amorosi et al, 2017).
to physical, biological and chemical modifications
caused by the interactions between soil, atmosphere
and vegetation. Their development is strictly con 3 PALEOSOL IDENTIFICATION
nected to fluvial activity which in turn depends on
river sediment supply/discharge. A schematic repre In current practice, paleosols in floodplain deposits can
sentation of paleosol horizons (Ab, Bk, Bw) develop be identified by the physical analysis of core samples
ment is shown in Figure 3. When fluvial incision (visual inspection). The Ab horizon usually exhibits
takes place in response to eustatic or climatic con a darker colour with respect to the illuvial Bk - Bw
trol, the adjacent interfluvial areas are prevented horizons, due to the higher quantity of organic matter.
from river flooding and they experience soil devel The Bk horizon typically shows carbonate nodules
opment (paleosol 1 in Figure 3 top). In case of pro and often a microstructure created by calcareous
longed subaerial exposure, vegetation starts to grow cementation between particles. Figure 3 (bottom)
on the top soil, enriching the Ab horizon in organic shows a typical succession of horizons Ab-Bk-Bw. Con
matter. Moreover, weathering in the vadose zone sidering the results of the pocket penetrometer, repre
causes leaching of carbonates from horizon Ab and sented in Figure 4 for the EM-S3 borehole, it can be
their precipitation into the underlying Bk, in form of observed that Ab (whose top is identified with the
millimetric to centimetric nodules. The horizon Bw paleosol symbol shown in the legend, just below the
is characterized by incipient weathering, while hori poorly drained floodplain) shows higher values of
zon C does not experience pedogenetic modifica pocket penetrometer (PP) due to its overconsolidation
tions. In case of high sediment supply/discharge, state created by aging and pedogenetic processes,
fluvial incision is rapidly filled by the transported while Bk is characterized by medium-high values of
sediments, and interfluvial areas experience frequent PP, as a result of the carbonate accumulation and/or of
flooding events and crevasse lobe accumulation. the calcareous cementation between solid particles.
Under these conditions, the water table is usually In addition to the visual core inspection, simple
295
4 ROBERTSON’S CLASSIFICATION CHARTS
The observations of cores, extracted with

a continuous perforating system, and, as additional
tool, the results of classification tests carried out in
the geotechnical laboratory, are a valuable aid for the
identification of paleosol horizons in alluvial
deposits. These types of investigations require
a significant budget, especially when the study
involves large areas. The economic effort, as well as
the duration of the investigation campaign, could be
significantly reduced by means of other geotechnical
in situ tests. To this scope, CPTs, being rapid, cost-
effective and widely used in practice, are taken into
account as a possible additional investigation
method, to be used with the aim of extending punc
tual information from boreholes to larger areas. In
this paper, the application of CPT tests is analysed
with respect to determination of the soil stratigraphy
deduced from the soil behaviour type index (SBT)
(Robertson, 2010). Paleosols are geological horizons
with peculiar mechanical behaviour due to micro-
structure creation as a consequence of pedogenetic
processes and aging. A microstructured soil is char
acterized by a higher yield stress, peak strength and
small-strain stiffness Go with respect to an ideal soil.
At large strain, the soil is subjected to destructura
tion and it may show a contractive behaviour (Ler
Figure 3. Top: schematic representation of development of
paleosol horizons (Ab, Bk, Bw). Bottom: picture of the oueil and Hight, 2003). In order to investigate the
cores extracted at 20 and 25m depth from EM-S3 effectiveness of this approach, the well-drained
borehole. floodplain deposits identified in the three CPTU pro
files, close to EM-S1, EM-S3 and EM-S5, were plot
ted on different soil classification charts available in
the literature. Firstly, the Robertson (1990) chart was
laboratory tests have been performed in order to con considered (Figure 5). Potentially, CPT-based charts
trol experimental data that characterize the paleosols could have the capability of recognizing paleosols
stratigraphy (Bruno et al, 2020). since they rely on in situ mechanical behaviour and
In Figure 4, standard characterization tests have not directly on soil classification criteria based on
been performed on two samples of the Ab horizon of grain size distribution and plasticity, such as the
core EM-S3 (11.55m and 19.55m), on one sample of USCS (Unified Soil Classification System) (Tonni
the Bk horizon (12.55m) and two samples of the et al., 2019a, Tonni et al., 2019b). On the normalized
C horizon (20.55m and 14.35m). The carbonate con Qt-Fr chart the majority of records belong to SBT
tent showed low values (close to zero) for the Ab zone 3 (clay), on the right side of the normal consoli
samples. The organic content determination con dated central area, characterized by increasing OCR.
firmed that the highest values belong to the Ab hori Data which have CPT-based SBT in either zone 4
zon, which is coherent with the presence of or 5, could reflect the same soil (in terms of grain
a vegetation cover during soil development. An size), but stiffer, with higher OCR (horizon Ab), or
increase in particle grain size was observed from a coarser layer as a consequence of a higher percent
paleosol horizon Ab to Bk at the depth of 11.50 age of sand or carbonate nodules (Bk horizons).
12.50 m. Indeed, the deposition in the Bk horizon of Similar results can be deduced from the Qt-Bq charts
Fe and Mn oxides and the formation of secondary (lower part of Figure 5), where the majority of the
calcite nodules causes the enrichment in coarse grain points are located in zone 3 (clay) and, to a lesser
material (recognized as sand in the laboratory pro extent, in zones 4 (silty mixtures) and 5 (sand mix
cedure). On the other hand, the plasticity index is tures). Points showing SBT 4 and 5 plot around the
greater in the Ab horizon. A likely explanation is vertical line Bq = 0. Comparing these groups of data
accumulation and mineralization of the organic (SBT 4 and 5) in the Qt-Fr and Qt-Bq charts, it can be
matter and its association with clay minerals (Bruno noticed that a larger amount of points plots into zones
et al, 2020). Samples from C horizons, not subjected 4 or 5 in the Qt-Bq chart rather than in the Qt-Fr chart.
to pedogenic processes, show an organic content This likely reflects the presence of a significant silty
close to 2.3%, PI around 16-17%, carbonate content component in this part of the deposits (see also grain
~20%, clay content of ~15% and silt content ~75%. size distribution in Figure 4 of samples EM-S3 and its
296
Figure 4. Physical characterization of five soil samples belonging to the continuous coring EM-S3. Two samples (11.55
and 19.55m deep, respectively) belong to horizon Ab; The sample at 12.55m belong to horizon Bk and the remaining two
samples belong to horizon C.
Robertson’s chart in Figure 5). These layers can be

more effectively detected using the Bq parameter,
which captures the dilatant behavior of silts (or paleo
sols) and the possible occurrence of coarser thin
layers, not revealed by Fr. In a recent update of the In addition to Robertson’s approaches, a different
classification charts, Robertson (2016) suggested to classification system, proposed by Schneider et al
use CD = 70 line in the Qtn-Fr chart (Figure 5) to rec (2008) was also taken into account. Its application
ognize microstructured soils. Indeed, the CD = 70 line showed that the investigated soils have
should separate contractive from dilative behaviours a predominantly contractive behaviour, belonging to
and it can be expressed with the following formula: the clay like contractive soils (CC) and the transi
tional contractive soils (TC). Schneider et al (2008)
also suggests that microstructured soils plot above
a line of equation Qtn = -U2+20, where U2 is defined
as the ratio between the excess pore water pressure
With Qtn normalized cone resistance and Fr nor and the effective stress in the point. No points that
malized friction ratio. satisfy this condition were found within the investi
In case of microstructured soils with contractive gated depth. Figure 6 shows the results of the CPTU
behaviour at large strains, but due to the increased test performed in close proximity to EM-S5, in terms
strength and stiffness for cementation or aging, of corrected cone resistance qt, fs and u2. Using the
higher values of Qtn may be recorded. As a result, sedimentological characterization of facies in core
cemented soils wrongly plot on the dilative region of EM-S5, and the SBT classification system, the geo
the chart. In Figure 5 the majority of the points plots logical-geotechnical stratigraphy was identified. It
below the boundary line CD = 70 and only few can be easily observed that the recognition of the dif
points, that may show some microstructure, above it. ferent sedimentary facies starting solely from CPTU
On the Qtn-Fr chart, the modified SBTn boundaries classification, in case no geological information is
(black continuous lines) proposed by Schneider et al available, is not straightforward, as clearly shown by
(2012) have been superimposed on the original the comparison between central lagoon, crevasse
SBTn zones (black dashed lines) by Robertson splay and floodplain deposits. Focusing on the allu
(1990). The modified contours are based on the vial plain deposit (floodplain), it is interesting at this
Modified Soil Behaviour Index IB expressed as stage to compare how points plot with depth in the
follows: classification charts (Figure 6). The comparison
297
Figure 5. Soil Behaviour Type data plotted in the Robertson (1990) Fr-Qt, and Bq-Qt charts and in Robertson (2009) Fr-Qtn
chart for the floodplain deposits identified in EM-S5, EM-S1 and EM-S3 boreholes.
suggests that crevasse splay horizons, in particular sedimentary facies in the alluvial Po Plain, they do
those at a depth of 18 – 19.5 m and 22.5 – 23 m, not provide univocal identification of paleosol hori
show SBT 4 and 5 in almost all types of charts. This zons without the support of their geological prelim
is due to the presence of cm-thick sandy layers that inary recognition. This evidence applies to low
characterize this type of facies. On the contrary, the mature paleosols, peculiar horizons of the flood
floodplain layers, where paleosol horizons have been plain facies subjected in the past to microstructura
identified, plot in SBT zone 4 only on the Qt-Bq tion for cementation (Bk) or aging (Ab): they do not
Robertson (1990)’s chart, while the other classifica appear to be easily identified by using only infor
tion charts recognize them as homogeneous SBT 3. mation from cone penetration tests (qt, fs, u2). Inte
This peculiarity could be justified by the fact that in grating the visual inspection of paleosols from
fine grained soils the Bq parameter is more sensitive cores with basic classification laboratory tests is
than Fr to the presence of structure variations within crucial to a correct geotechnical-geological inter
thin layers, which cannot be detected through the Fr pretation, especially in relatively homogeneous,
parameter. mud-prone alluvial deposits, where sand bodies dis
play strongly lenticular geometries.
The paleosol horizons investigated in this study
5 CONCLUSIONS cannot be identified as microstructured soils in the
classification charts available in the literature.
In this paper, it has been shown how CPT data col This fact is probably a sign of the low maturity of
lected in floodplain deposits of the river Po plot on these paleosols (Inceptisols), which only account
well-known classification charts (Robertson 1990; for relatively short periods of subaerial exposure
Robertson, 2009, Schneider et al, 2008). of the ancient topsoil. Nonetheless, a characteristic
Although CPTU classification charts proved to feature of paleosol horizons seems to be shown
be a potential valuable help in the identification of by the Qt-Bq Robertson’s (1990) chart, which
298
Figure 6. Left side: stratigraphic log of the continuous coring EM-S5. From left to right: corrected tip resistance qt, sleeve
friction fs, pore water pressure u2 and soil behaviour type (Robertson, 2009) vs depth of the CPTU performed in close
proximity to EM-S5. The final five columns represent with a red line the depth where SBT 4 or 5 have been found in the
classification charts.
highlights the presence of SBT 4, where all the Robertson, P.K. 2016. Cone penetration test (CPT)-based
other charts provide SBT 3. Further investigations soil behaviour type (SBT) classification system-an
through dilatometer or other in situ tests will be update. Canadian Geotechnical Journal.
carried out in the next future to expand the results Robertson, P.K. 1990. Soil classification using the cone
of this study. penetration test. Canadian Geotechnical Journal, 27(1):
151–158.
Robertson, P.K. 2009. Interpretation of cone penetration
REFERENCES test-a unified approach. Canadian Geotechnical Journal,
46(11): 1337–1355.
Amorosi, A., Bruno, L., Campo, B., Morelli, A., Rossi, V., Schneider, J.A., Randolph, M.F., Mayne, P.W., Ramsay, N.
Scarponi, D., Hong, W., Bohacs, K.M, Drexler, T.M. R. 2008. Analysis of factors influencing soil classifica
2017. Global sea-level control on local parasequence tion using normalized piezocone tip reistance ans pore
architecture from Holocene record of the Po Plain, Italy. pressure parameters, Journal of Geotechnical and
Marine and Petroleum Geology. Geoenvironmental Engineering, 134(11): 1569–1586.
Amorosi, A., Bruno, L., Rossi, V., Severi, P., Hajdas, I. Schneider, J.A., Hotstream, J.N., Mayne, P.W., Radolph, M.
2014. Paleosol architecture of a late Quaternary basine F. 2012. Comparing CPTU Q-F and Qt-Δu2/б’vo soil
margin sequence and its implications for high reso classification charts. Geotechnique Letters 000:1–7.
lution, non-marine sequence stratigraphy. Glob. Planet. Tonni, L., García Martínez, M.F., Maurini, D.,
Change 112:12–25. Calabrese, L. 2019a. Developing a regional-scale geo
Bruno L., Marchi M., Bertolini I., Gottardi G., Amorosi A. technical model of the north-western Adriatic coastal
2020. Climate control on stacked paleosols in the Pleis area (Italy) for urban planning and robust geotechnical
tocene of the Po Basin (northern Italy). Journal of Qua design. Proc. 17th European Conference on Soil Mech
ternary Science, vol 35(4): 559–571. anics and Geotechnical Engineering, ECSMGE.
Leroueil, S. & Hight, D.W. 2003. Behaviour and properties Tonni, L., García Martínez, M.F., Rocchi, I. 2019b. Recent
of natural soils and soft rocks. In Tan et al (eds), Char developments in equipment and interpretation of cone
acterization and engineering properties of natural soils, penetration test for soil characterization. Rivista Italiana
vol.1. di Geotecnica 53(1): 71–99.
299
© 2022 the Author(s), ISBN 978-1-032-31259-0
Application of CPT to the evaluation of permeability in a Po river

embankment prone to backward erosion piping
Ilaria Bertolini, Guido Gottardi, Michela Marchi & Laura Tonni
University of Bologna, DICAM, Italy
Agnese Bassi & Alessandro Rosso

Interregional Agency for the Po River, AIPO, Italy
ABSTRACT: Backward erosion piping is one of the most critical issues for the stability of many river embank
ments worldwide. The article presents results from an extensive geotechnical campaign carried out on a section of
the Po river embankments (Italy), prone to backward erosion piping. The aim of the investigations was twofold:
firstly, identifying the stratigraphic arrangement of the river system causing favourable conditions for piping occur
rence, and secondly determining the permeability of the aquifer and the finer top stratum (blanket) in their undis
turbed state, as well as that of the preexisting eroded zone in proximity of the sand boil (i.e. the volume around
the exit hole). Indeed, analytical and numerical approaches available in the literature are particularly sensitive to
the value of permeability of the aquifer, which should therefore be determined in undisturbed conditions. CPT
tests turned out to be suitable and economical to this scope. This study presents an integrated analysis of in situ
testing data, collected from boreholes, piezocone test and Lefranc test, aimed at estimating the saturated perme
ability of the different soil units using different experimental methods and interpretation approaches.
A comparison between alternative CPT-based methods and field permeability tests is proposed. A good agreement
between the different methods was found, thus suggesting the suitability of CPT for the hydraulic characterization
of soils.
1 INTRODUCTION Netherlands, among others. A widespread occurrence

of sand boils has been also reported for the Danube. In
Backward erosion piping is an internal erosion pro Italy, along the Po River, more than 130 sand boils
cess which may occur in stratigraphic contexts char have been registered and the majority of them (80)
acterized by a shallow thin clayey layer (“blanket”) reactivate periodically during high-water events (Merli
overlying a highly permeable stratum (aquifer), et al., 2015; Aielli et al., 2019; Marchi et al., 2021).
where significant seepage flow can develop. Such An in-depth understanding of the mechanism and
stratigraphic conditions are typical of alluvial plains a reliable assessment of piping risk require both the
where major lowland rivers flow and of delta areas, preliminary development of a detailed stratigraphic
protected by high-rise embankments. When the model of the whole river embankment system and
water level in the river rises, a high hydraulic head the accurate estimate of soil permeability. The differ
difference may be attained, resulting in an increase ent approaches, either analytical or numerical (e.g.
of the seepage flow in the aquifer. In such condi USACE 2000; Sellmeijer et al. 2011; García Marti
tions, the water flow may concentrate towards local nez et al., 2020), for the modelling of this phenom
discontinuities where the hydraulic gradients enon, are indeed based on a careful characterization
increase, possibly creating flow exit pipes and the of the hydraulic properties of soils. As a result, due
consequent initiation of the backward erosion pro to the significant extent of the areas to be investi
cess. The phenomenon could then progress eroding gated and to the fact that characterization involves
sand particles at the interface between the aquifer coarse grain soils, field investigations turn out to be
and the blanket, creating at the same time sand boils the only possible practical approach which can be
on the ground surface. adopted for the definition of the soil parameters
Backward erosion piping is recognized to represents required in the analyses. This paper discusses the
a serious threat to dams and levees. Indeed, this phe determination of soil permeability from CPTU tests
nomenon typically affects the surroundings of embank carried out in a piping-prone section of the Po river
ments of major lowland rivers in China, USA, and The embankments.
DOI: 10.1201/9781003308829-39
300
Figure 1. Geographical localization of the investigated site in Guarda Ferrarese (Ro, Italy). In photo 1b, the sand boil
reactivation in November 2014 in the investigated river embankment section.
2 THE STUDY AREA recurrent and intense high-water events and conse
quently further aggravates the problem.
The Po River is the main Italian watercourse. It flows On that account, over the last years local authorities
across northern Italy from the Western Alps to the have launched a number of investigation campaigns
Adriatic Sea near Venice, for an overall length of on the most critical areas, with the aim of mitigating
650 km. In the last two centuries, the height of Po the flood risk connected to backward erosion piping.
river embankments has been progressively increased, This paper is based on the experimental dataset
with the obvious benefit of preventing overtopping recently collected in one of the most investigated sec
while increasing at the same time the susceptibility to tions of the Po river, located in Guarda Ferrarese
piping of the whole area. The higher the water level (Ferrara province) (Figure 1a). Here, an historical
in the river, the greater the water head difference at sand boil of remarkable size (the sand volcano
the toe of the embankment landside, which, in turn, reached an external diameter higher than 2.5 m),
results in an increase of the probability of sand boil located landside of the riverbank section, at the
occurrence. In addition, the variation of flood dynam bottom of a ditch, reactivates recurrently during high
ics, induced by climate changes, generates more water events.
Figure 2. Stratigraphic model of Section A-A in Figure 1, with CPTU log profiles.
301
3 GEOTECHNICAL INVESTIGATIONS AND blanket thickness approaching the ditch may give
STRATIGRAPHIC MODEL a reason for the position of the sand boil. Moreover, the
thickness of the lesser permeable unit C, roof of the
Figure 1 shows the location of the main in situ tests aquifer, decreases from 1.76m in CPT1bis to 1.2m in
performed in the investigated river embankment sec CPT5.
tion, in three different stages of geotechnical cam Laboratory tests have been carried out on 14 dis
paigns: in 2010, 2016 and 2021. During the 2016 turbed soil samples collected during borings. Sam
investigation, dissipation tests have been performed in ples have been tested to measure grain size
units B and C, in order to gain a better insight into the distributions (Figure 3A) and plastic properties of
permeability characteristics of the subsoil, in terms of the finer samples (Figure 3B). Unit A (no. 5 sam
the horizontal permeability kh. In 2021, the geotech ples) turns out to have an average percentage of fine
nical campaign included continuous coring boreholes and coarse materials equal to 14% (standard devi
with execution of Lefranc permeability tests and SPT. ation, SD, equal to 16%) and 86% (SD = 16%),
In addition, Casagrande piezometers have been respectively. The organic content is extremely low
installed in the bored holes to monitor pore water pres (0.28%) and the carbonate content equal to 10%.
sures changes in both the confined and the phreatic Unit C (no.3 samples) and E (no. 2 samples) are
aquifer (blanket), in relation to the variation of the dominated by fine-grained soils, with a percentage
water level in the river (PZ1, PZ2, PZ3, PZ4 and PZ5 of clay and silt equal to 52% (SD = 20%) and 60.8%
in Figure 1). (SD = 6.5%), respectively. Unit B, consisting of an
The stratigraphic model in Figure 2 shows profiles alternation of silty mixtures, is composed of 39% silt
of the corrected cone resistance (qt) and the pore and 8% clay, with plasticity index equal to 14% (SD
pressure (u2) for the 2016 piezocone tests, which = 7.8%) and organic content 0.72%. The samples of
have been interpreted according to the classification the investigated units B and C are located in the left
framework proposed by Robertson (2009). Samples side of the Casagrande chart (Figure 3) and therefore
extracted during borings have been also taken into can be classified as low plasticity silts and clays.
account to support stratigraphic modelling.
As evident from Figure 2, five different soil units
have been identified. Unit A (aquifer) is the sandy 4 DETERMINATION OF THE SATURATED
layer where the erosion process takes place in case of PERMEABILITY
piping. On the top of this unit, there are the continuous
clayey and silty horizons C and B, creating favorable Estimates of the saturated permeability of the differ
conditions for backward erosion piping initiation and ent soil units have been first obtained using the well-
progression. It is worth noting that the thickness of the known CPTU-based empirical correlations proposed
‘’blanket’’ (comprising units B and C) tends to reduce by Robertson (2010), given by:
in close proximity to the sand boil, as deduced from
CPT1bis (2021) and CPT5 (2021). In particular, the
thickness changes from 3.72m to 2.96m moving
toward the ditch, being the CPT5 (2021) test closer to
the longitudinal axis of the ditch. The reduction of the
Figure 3. (Left) Particle size distribution of soil samples collected from boreholes; (Right) Soil classification according to
the Plasticity Chart.
302
Figure 4. CPTU-based logk profiles in four different vertical alignments of the investigated river embankment section, in
conjunction with permeability data deduced from Lefranc and dissipation tests.
Robertson’s correlations are function of Ic, which plot shows that the coarser layers A and D exhibit
generally relates well to soil grain size distribution, a bell-shaped distribution, typical of a normal distribu
although this index is mainly intended to identify an tion, while the heterogeneous unit B, characterized by
in-situ soil behaviour. alternations of silty-sands and sandy-silts, presents
The permeability profiles deduced using eqs. (1) a bimodal distribution. In this case, the higher peak is
and (2) are shown in Figure 4. found at log(k) = -4.75 (i.e. k=10-4.75 m/s), typical of
In addition, the frequency distributions of log(k), as silty-sands, whereas the smaller peak corresponds to
deduced from CPT data, are provided in Figure 5.This log(k) = -6.8, consistent with sandy-silts. In unit C, the
frequency distribution of log(k) is asymmetric, with
a peak at -9.75 (thus low permeability clay) and mean
value equal to -8.3, the latter value corresponding to
a permeability typical of silty-clayey soils. This out
come reflects the fact that unit C is an alternation of
clay and silty-clay materials.
The analysis of soil permeability was also devel
oped using no. 7 dissipation tests, carried out in the
2016 CPTU tests. These data have been analysed
using two approaches: the well-established Parez and
Fauriel (1988) simplified formulation, which provides
estimates of the permeability kh from the time for
50% excess pore pressure dissipation (t50), and the
correlation based on the 1D constrained modulus M
and the coefficient of consolidation in horizontal dir
ection ch, according to the following formula:
where γw is the unit weight of water. In eq. (3), ch

Figure 5. Frequency distribution of permeability data (log has been approximated according to Robertson
k) from CPTU for each soil unit under investigation. (2010):
303
Table 1. Mean values of soil permeability, as deduced from CPTU interpretation in terms of Ic and Parez and Fauriel
(1988) method, and results from Lefranc tests (standard deviation given in brackets).
log k (k in m/s) k (m/s)

From From
Lefranc Test dissipation From CPTU (Ic) Lefranc Test dissipation From CPTU (Ic)
A -6.05 (0.74) -5.71 (0.38) -5.04 (0.42) 8.91E-07 1.95E-06 9.12E-06

B -5.79 (0.22) -5.99 (0.79) 1.62E-06 1.02E-06
C -8.33 (0.44) -8.34 (0.79) 4.68E-09 4.57E-09
D -6.54 (0.52) 1.62E-07
E -6.08 (1.20) 1.17E-05
a chamber filled with gravel located at the bottom of

a borehole and then in the measurement of the water
level lowering with time. The permeability can be
whereas the constrained modulus M has been esti calculated using information on the geometry of the
mated using the Robertson (2009) formulation, sug internal chamber (area, length and diameter) and the
gested for soils with Ic > 2.2 and thus including a wide measurements of the variable head during the test.
variety of soils such as sandy silts, silts and clays: The determination of the permeability by means of
Lefranc test is particularly suitable for soils having
permeability higher than 10-7 m/s, which corresponds
to units B and A of the investigated subsoil.
The estimates of k, as obtained from the different
with methods described above, are all shown in Figure 4.
It can be observed that local determination of the
permeability, provided by interpretation of dissipa
tion tests or by Lefranc tests, are generally in good
agreement with CPTU-based profiles of k predicted
using eqs. (1) and (2). As regards the vertical align
ment n° 4, located on the right side of the ditch,
close to the sand boil, the permeability obtained
being Qtn the stress-normalized corrected cone from the Lefranc test, at 14.5 m in depth, could be
resistance, as defined in Robertson (2009). only compared with the permeability profile com
Results from the analysis of dissipation tests are puted from CPTU3_2016, which is the closest piezo
shown in Figure 4. It is worth observing here that cone test though located on the opposite side of the
the indirect determination of k described above road embankment. Table 1 summarizes the average
requires a reliable interpretation of dissipation tests, values of permeability obtained from the different
which could be affected by partial drainage effects methods. The table confirms what previously
when performed in predominantly silty sediments observed in Figure 4, i.e. that minor differences exist
(such as unit B). In this case, the estimate of t50 and between dissipation-based-values of k and CPTU-
ch from dissipation tests, using standard approaches, based predictions provided by eqs. (1) and (2).
might be somewhat inaccurate (Tonni et al., 2019). An additional data analysis is provided in Table 2,
The latest site investigation campaign, performed where the average and the standard deviation of log(k),
in 2021, also included n° 7 Lefranc tests (LF). These as computed from piezocone tests, are reported for
tests consist in the injection of water by gravity into
Table 2. Mean values (standard deviation in brackets) of the permeability of each soil unit, as computed from each cone
penetration test.
log k (k in m/s) D C A B C2 B2 C3 E
CPTU1_2016_Berm -6.86 -8.48 -5.11

(0.52) (0.51) (0.39)
U277_2010_Crest -8.57 -5.18 -6.54 -8.57 -5.25 -9.25 -6.08
(0.83) (0.54) (1.07) (1.06) (0.87) (0.29) (1.20)
CPTU02_2016_Outerslope -6.59 -7.71 -5.06 -8.23 -6.80
(0.45) (1.22) (0.45) (0.52) (1.07)
CPTU3_2016_BankToe -8.48 -4.95 -6.11 -8.94 -8.91
(0.62) (0.41) (1.03) (0.76) (0.24)
304
every soil unit, taking into account the increasing dis In this study, the use of a few CPTU, performed in
tance of the test from the river course. In this analysis, different locations of the river embankment section
the fine-grained unit C has been divided into 3 distinct (namely the berm, the crest, the toe of the embankment
sub-units, namely the upper unit C and the deeper soil and in proximity of the sand boil) has provided infor
units C1 and C2 (see Figure 6). Similarly, the predom mation on the spatial variability of permeability.
inantly silty unit B has been divided into the shallowest A negligible variation of this parameter has been
unit B and the deep unit B2. observed, especially for the aquifer. Such an evidence
The differences in the permeability values within potentially simplifies the analysis of the phenomenon
each soil unit appear to be small, thus indicating since a unique mean value can be assumed as represen
a substantial homogeneity in the horizontal direction. tative of the permeability of each soil unit.
Such condition facilitates the numerical modelling of Such outcomes, though preliminary, appear to be
the phenomenon (García Martínez et al., 2018). very promising in view of an intensive application of
the method for the assessment of piping susceptibil
ity of the river embankments along the mid-lower
stretch of the Po.
REFERENCES
Aielli S., Pavan S., Parodi S., Rosso A., Tanda M.G.,
Marchi M., Vezzoli G., Pantano A., Losa D., Sirtori M.
2019. Collection and analysis of the reactivation data of
the historical sand boils in the po river levees. Lecture
Notes in Civil Engineering, 17, 327–335.
Figure 6. Stratigraphic section of the investigated river García Martínez M.F., Gottardi G., Marchi M., Tonni L.
embankment with indication of the different soil units 2018. Modelling a sand boil reactivation in the
specified in Table 2. middle-lower portion of the po river banks. Numerical
methods in geotechnical engineering IX, 2, 1219–1225.
García Martínez M.F., Tonni L., Marchi M., Tozzi, S.,
In particular, a minor variability is observed Gottardi G. 2020. Numerical tool for prediction of sand
between the average values of permeability in the boil reactivations near river embankments. Journal of
aquifer A. Moving from the centre to the toe of the Geotechnical and Geoenvironmental Engineering 146
embankment, a slight increase of k is observed, (12): 06020023.
likely due to a reduction of porosity beneath the Marchi M., García Martínez M.F., Gottardi G., Tonni L.
2021. Field measurements on a large natural sand boil
embankment due to its self- weight. along the river Po (Italy). Quarterly Journal of Engineer
ing Geology and Hydrogeology 54(4): qjegh2020–097
5 CONCLUSIONS Merli C., Colombo A., Riani C., Rosso A., Martelli L.,
Rosselli S., Severi P., Biavati G., De Andrea S.,
Fossati D., Gottardi G., Tonni L., Marchi M., García
The assessment of backward erosion piping suscepti Martínez M.F., Fioravante V., Giretti D., Madiai C.,
bility of river embankment systems requires the Vannucchi G., Gargini E., Pergalani F., Compagnoni M.
accurate estimate of the permeability of the soil units 2015. Seismic stability analyses of the Po River Banks.
affected by underseepage, irrespective of the approach Engineering Geology for Society and Territory 2: Land
(numerical or analytical) adopted for the analysis. slide Processes, 877–880.
In this study, the well-known Robertson’s (2010) Parez, L. & Fauriel, R. 1988. Le piézocône améliorations
empirical formulation, providing profiles of the perme apportées à la reconnaissance des sols. Revue française
ability k as a function of the Soil Behaviour Type de Géotechnique, 44: 13–27.
Index, Ic, has been used to estimate the permeability of Robertson, P.K. 2009. Interpretation of cone penetration
tests - a unified approach. Canadian Geotechnical Jour
an aquifer and the blanket forming the subsoil of a river nal 46(11): 1337–1355.
embankment prone to piping, along the major Italian Robertson, P. K. 2010. Estimating in-situ soil permeability
watercourse. The comparison with alternative determin from CPT & CPTu. In: Proc. 2nd International Sympo
ations of k, based on dissipation tests or Lefranc perme sium on Cone Penetration Testing. Huntington Beach,
ability tests, showed a good agreement between the California.
estimates provided by the different methods. Sellmeijer, H., Lopez de la Cruz, J., van Beek, V.,
According to results presented in this paper, cone Knoeff, H. 2011. Fine-tuning of the backward erosion
penetration tests have proven to be a valuable strat piping model through small-scale, medium-scale and
egy for the definition of the saturated permeability of IJkdijk experiments. European Journal of Environmen
tal and Civil Engineering, 15(8):1139–1154.
the soil layers underlying the embankment. The esti Tonni L., García Martínez M.F., Rocchi I. 2019. Recent
mation of permeability from Ic results in developments in equipment and interpretation of cone
a continuous profile of k with depth, which appears penetration test for soil characterization. Rivista Italiana
to be an important added value in comparison with di Geotecnica 53(1): 71–99.
a limited number of local measurements from dissi USACE. 2000. Design and Construction of Levees, U.S.
pation or Lefranc tests. Army Corps of Engineers. Washington, DC.
305
© 2022 the Author(s), ISBN 978-1-032-31259-0
Dissipation tests to evaluate the equilibrium pore pressure

F.A.B. Danziger, G.M.F. Jannuzzi & A.V.S. Pinheiro
Coppe and Polytechnic School, Federal University of Rio de Janeiro, Brazil
ABSTRACT: Dissipation tests are carried out with two main purposes: the estimation of the horizontal coeffi
cient of consolidation, ch, and the evaluation of the equilibrium pore pressure. Regarding the first case, even in
soft clays the duration of the test doesn´t need to be more than 40 minutes, in most situations. However, in the
latter case a long time is needed to reach equilibrium, even when probes with small diameters are used. The
present paper presents results of long-term dissipation tests performed in a soft clay layer, aiming at the decision
of executing or not an underpinning of shallow foundations of a damaged warehouse which suffered significant
settlements.
1 INTRODUCTION used. The present paper presents results of long-term

dissipation tests performed aiming at the decision of
According to Campanella & Robertson (1988), the executing or not an underpinning of a damaged
pore pressure measurement during the penetration of warehouse.
a probe was carried out for the first time at the begin
ning of the 1970s, as described by Wissa et al. (1975) 2 DISSIPATION TESTS
and Torstensson (1975). The main purpose of the
Wissa et al.´s (1975) probe was to determine flow pat 2.1 General
terns through dams and pore pressure fluctuations due
to waves and tides. However, this could only be Pore pressure dissipation occurs as soon as the probe
achieved after the dissipation of the pore pressures penetration is halted, even for a simple rod addition
generated during probe installation. Wissa et al. in the rod stem. In general, tests are carried out on
(1975) and Torstensson (1975) realized the enormous specific depths of particular interest for the soil
potential of the pore pressure measured during pene investigation. The time for dissipation depends on
tration, and Torstensson (1975) mentioned the possi the coefficient of consolidation and the diameter of
bility of assessment of the coefficient of consolidation the filter, the higher the coefficient of consolidation
from the time necessary to pore pressure dissipation. and the smaller the filter diameter the faster is the
It must be pointed out that Janbu & Senneset dissipation (e.g., Torstensson, 1977). In most soil
(1974) and Schmertmann (1974) also used a pore types and regular piezocone 10 cm2 penetrometers,
pressure probe and a Geonor electric piezometer, full dissipation can be achieved in less than
respectively, to investigate the pore pressure gener one hour. However, when tests are performed in soft
ation during the probe installation. clays full dissipation may need a few days. Never
It is generally accepted that the piezocone was theless, since most methods use t50 to evaluate the
formally born (i.e., the joint measurement of cone coefficient of consolidation, in most cases 30-40
resistance, qc, sleeve friction, fs, and pore pressure, minutes of test duration is enough for the estimation
u) at the beginning of 1980s, from the publication of the coefficient of consolidation.
of several papers in the 1981´s ASCE Symposium There are some methods to estimate the coeffi
on Cone Penetration Testing and Experience held cient of consolidation from dissipation tests (e.g.,
in St. Louis, and the 1982´s II European Sympo Torstenson, 1977, Levadoux & Baligh, 1986,
sium on Penetration Testing held in Amsterdam. Houlsby & Teh, 1988, Danziger et al., 1997). The
Since then, dissipation tests have been used both methods based on the Strain Path Method seem to be
to estimate the horizontal coefficient of consolida more used in practice.
tion, ch, and the evaluation of the equilibrium (or
hydrostatic) pore pressure. 2.2 The evaluation of the equilibrium pore
However, in the latter case, a long time is needed to pressure
reach equilibrium when tests are carried out in soft Some engineering applications require the measure
clays, even when probes with small diameters are ment of the equilibrium pore pressure, both onshore
DOI: 10.1201/9781003308829-40
306
and offshore. In the offshore case, the verification of materials, therefore regular 10 cm2 penetrometers may
the presence of gas on sediments to estimate the be used.
safety of pile foundations of platforms was carried out In all cases, an issue with the measurement of the
by Lunne et al. (1996). The determination of the so- equilibrium pore pressure is the accuracy of the zero
called “drilling window” for the installation of jetted (reference) reading. Any possible variation must be
conductors in clays (e.g., Jeanjean et al., 2015) also avoided, since significantly influences the reliability
requires the equilibrium pore pressure measurement. of the measurements. A way to solve this issue is to
In the onshore case, the previously mentioned use dual pore pressure penetrometers, which have
determination of flow patterns through dams and been adopted by the Federal University of Rio de
pore pressure fluctuations due to waves and tides Janeiro since 1996 on regular basis (Bezerra,1996,
(Torstensson, 1975) are possible applications. Danziger et al., 1997). The use of two independent
In all those cases, since tests have been performed pore pressure sensors eliminates any doubts regarding
in clays, the time required for full dissipation is sig the reliability of the measurements. An example is
nificant, thus probes with small filter diameters have presented in Figure 2, in which a pore pressure smal
been used. Figure 1 presents pore pressure probes ler than the hydrostatic (u0), assumed as the water
used offshore (Peuchen & Klein, 2011). unit weight multiplied by the distance between the
water table position and the tested depth, is verified.
Figure 2. Dissipation test performed on the main Brazilian

motorway with a dual pore pressure piezocone (Danziger
et al., 1998, Danziger, 2007).
Since the material tested is dilative, the pore pres

sure u2 at the onset of dissipation (time t=0), meas
ured at the cone shoulder, recorded a negative excess
pore pressure, because this position is dominated by
the shear stresses. The pore pressure u1, measured at
cone face, however, recorded a positive excess pore
pressure at t=0, because this position is dominated by
octahedral stresses. After a certain time both u1 and
u2 record the same pore pressure, below the hydro
static pressure. If only one sensor was available,
doubts regarding the reliability of the measurements
would certainly exist, because at that time there was
no apparent reason for such an event. However, the
Figure 1. Small diameter pore pressure probes (Peuchen & measurements carried out triggered an investigation
Klein, 2011). which discovered that a beer factory (not close to the
tested site) was taking water from the ground, gener
ating a reduction in pore pressure in depth.
Another increasing application is the determination
of pore pressure equilibrium for tailings dams, how
ever in such cases probes with small filter diameters 3 THE CASE
are not required, because in most cases sandy silts or
silty sands constitute the materials to be tested. The A warehouse of an industry in the state of Rio de
time for full dissipation is not significant in those Janeiro, Brazil, presented serious damages due to
307
excessive settlements. The warehouse is founded on when penetration is carried out on unsaturated mater
shallow foundations constructed on an embankment ial, the Coppe/UFRJ procedure is always to saturate
overlying a soft clay layer. The company in charge of the filter with water, following a procedure developed
the repair wanted to know whether the settlements had by NGI (Lacasse, 1980) in which vacuum is used;
already ceased or were still under development. If they a parallel CPT is then carried out in the unsaturated
had ceased, just repairing the damaged construction soil; ii) the pore pressure is recorded during probe
would take place, whereas underpinning with deep lowering inside the casing filled with water, which is
foundations in addition to repairing the structure then used to check the sensor calibration.
would be necessary to significantly reduce further dis
placements of the structure.
The soil profile from a CPT is presented in
Figure 3, where it can be observed an upper embank
ment until 10.2 m depth, followed by a soft clay
layer until 15.0 m depth. The water table is approxi
mately 5.3 m below ground level.
Figure 4. Pore pressure probe used for the determination of

equilibrium pore pressure.
The pore pressure probe was pushed to 11.11 m,

approximately 1 m below the top of the clay layer,
then the dissipation was allowed to occur. The data
acquisition system was kept inside a room adjacent
to the test site, and the chart pore pressure versus
time (in a log scale) was observed, and pictures were
sent to the Coppe/Ufrj team, which had returned to
the laboratory in Rio de Janeiro, to follow the
Table 1. Dissipation tests performed.

Figure 3. Soil profile.
Test Depth Duration (day)
It was not known whether the consolidation of the 1 11.11 5
clay layer due to the embankment and surcharges on 2 12.11 7
the ground had already ceased or not. Therefore, it 3 13.11 2
was decided to verify whether the pore pressures gen 4 14.10 1
erated by the loading had dissipated or not, by the 5 14.85 4
measurement of the equilibrium pore pressures at dif
ferent depths inside the clay material. Since the
expected time for full dissipation would be significant,
a pore pressure probe with a filter with a smaller diam
eter than a regular penetrometer (10 cm2) would be
used. The available probe has only one filter, 11.6 mm
in diameter, made of sintered bronze (Figure 4), rather
than two filters in the piezocone penetrometer rou
tinely used by Coppe/UFRJ.
Since the interest of the measurements was only
the soft clay layer, a casing borehole was executed
until 8 m depth, which was filled with water. The
probe was then inserted into the casing and pushed at
a constant rate of 20 mm/s, the standard rate used for
piezocone testing. It must be highlighted that: i)
although several penetrometers are saturated with Figure 5. Picture showing a time close to the end of dissi
fluids like glycerin and silicon oil to keep saturation pation, 11.11 m test depth.
308
Figure 6. Dissipation tests performed.
Figure 8. Pore pressure measured during installation and

Figure 7. Equilibrium pore pressure versus depth of all tests. dissipation.
309
dissipation remotely. Figure 5 shows one of such Danziger, F. A. B. (2007). In Situ Testing of Soft Brazil
pictures, in which dissipation was considered almost ian Soils. Studia Geotechnica et Mechanica XXIX,
completed. When this occurred, a member of the 5–22.
team traveled to the test site and pushed the probe to Danziger, F.A.B., Almeida, M.S.S. & Sills, G.C. (1997).
the next depth. Table 1 presents the tested depths The significance of the strain path analysis in the inter
and the duration of each test until complete dissipa pretation of piezocone dissipation data. Géotechnique
tion. Figure 5 presents some oscillations in the final 47, No. 5, 901–914.
Danziger, F.A.B., Almeida, M.S.S. & Bezerra, R.L. (1997).
part of the test, which was attributed to loading – Piezocone research at COPPE/Federal University of Rio
unloading activities near the test. Those activities de Janeiro, 1997. Proc., Int. Symp. on Recent Develop
were requested to be interrupted to not disturb the ments in Soil and Pavement Mechanics, Rio de Janeiro,
tests, especially in its final part. pp. 229–236.
Figure 6 presents all dissipation tests, and Danziger, F.A.B., Almeida, M.S.S., Paiva, E.N., de
Figure 7 shows the equilibrium pore pressure versus Mello, L.G.F.S. & Danziger, B.R. (1998). The piezo
depth of all tests. The assumed hydrostatic pore pres cone to determine stratigraphy and soil classification (in
sure, based on the water depth on top of the soft Portuguese). Proc., XI COBRAMSEG, Brasília, Vol. II,
pp. 917–926.
clay, is also shown in the figure. It can be observed
Houlsby, G.T. & Teh, C.I. (1988). Analysis of the piezo
that there is still an excess pore pressure in the soft cone in clay. Proc., ISOPT-1, Orlando, Vol. 2, pp.
clay, especially in shorter depths. 777–783.
Figure 8 shows the pore pressure measured Janbu, N. & Senneset, K. (1974). Effective stress interpret
during installation (including during lowering the ation of in situ static penetration test. Proc., ESOPT,
probe inside the 8 m long casing) and the dissipa Stockholm, Vol. 2-2, pp. 181–193.
tion tests. Jeanjean, P., Zakeri, A., Al-Khafaji, Z., Hampson, K.
Based on the obtained results, it was decided Clukey, E. & Liedtke, E. (2015). Geotechnics for wells
to underpin the shallow foundations of the top-hole section and conductor. Frontiers in Offshore
warehouse. Geotechnics III – Meyer (Ed.), Taylor & Francis Group,
London, Vol. 1, pp. 95–128.
Lacasse, S. (1980). Procedure for deairing the pore pres
4 SUMMARY AND CONCLUSIONS sure probe in the laboratory. NGI report 40015-6,
Oslo.
Long-term dissipation tests with a pore pressure probe Levadoux, J.N. & Baligh, M.M. (1986). Consolidation
after undrained piezocone penetration. I: prediction.
have been carried out in a soft clay layer. The tests Journal of Geotechnical Engineering, ASCE 112, No. 7,
aimed at the evaluation of the equilibrium pore pres 707–726.
sure to know whether the soft clay was still under con Lunne, T., Isa, O.M. & Tan, M. (1996). Shallow gas prob
solidation due to an embankment. A few days were lem at Duyong B offshore Malaysia. Proc., 11th Off
necessary to reach full dissipation, even with the small shore South East Asia Conf., Singapore.
diameter of the probe (11.6 mm) used. An excess pore Peuchen, J. & Klein, M. (2011). Prediction of formation
pressure was measured in three out of the five tests pore pressures for top hole well integrity, Proc., Off
performed, indicating that the clay was still under shore Technology Conference, Houston, TX, paper
consolidation. 21301.
Schmertmann, J.H. (1974). Penetration pore pressure
Based on these measurements, it was decided that effects on quasi-static cone bearing, qc. European Sym
an underpinning of the shallow foundations with posium on Penetration Testing, Stockholm, Vol. 2-2, pp,
deep foundations was necessary, in addition to 345–351.
repairing the damaged warehouse. Torstensson, B.A. (1975). Pore pressure sounding
instrument. Proc., Specialty Conf. on In Situ Measure
ment of Soil Properties, ASCE, Raleigh, Vol. II, pp.
REFERENCES 48–54.
Torstensson, B.A. (1977). The pore pressure probe. Fjell
Bezerra, R.L. (1996). Third generation of Coppe/UFRJ sprengningsteknikk, Bergmekanikk/ Geotecknikk, 34.1–
piezocone penetrometer (in Portuguese). D.Sc. Thesis, 34.15.
COPPE/UFRJ, Rio de Janeiro. Wissa, A.Z.E., Martin, R.T. & Garlanger, J.E. (1975). The
Campanella, R.G. & Robertson, P.K. (1988). Current status piezometer probe. Proc., Specialty Conf. on In Situ
of the piezocone test - state-of-the-art report. Proc., Measurement of Soil Properties, ASCE, Raleigh, Vol. I,
ISOPT-1, Orlando, Vol. 1, pp. 93–116. pp. 536–545.
310
© 2022 the Author(s), ISBN 978-1-032-31259-0
Dynamic characteristics of the soils by Cone Penetration Tests (CPT)

A. Cavallaro
CNR-ISPC, Catania, Italy
ABSTRACT: A geotechnical project requires a site-specific investigation to collect data regarding the sub
surface conditions. Soil explorations must be made to determine the presence and identification of underlying
strata, groundwater conditions, types of geomaterials, their depths and thicknesses, and the associated engin
eering parameters required for geotechnical design. The Cone Penetration Test (CPT) represents one of the
most widely used on-site investigation methods in geotechnical engineering. This paper presents some empir
ical correlations available by Cone Penetration Tests (CPT) for determining the main geotechnical characteris
tics of soils. Moreover it aims to evaluate the small strain shear modulus by means of empirical correlations
based on penetration tests results, CPT, Standard Penetration Tests (SPT) and Flat Dilatometer Marchetti
Tests (DMT) or laboratory geotechnical investigations. It is aimed to achieve a better understanding of the
obtained geotechnical parameters and the empirical correlations, justifying the investigation effort and enab
ling reliable input data for advanced dynamic analysis.
1 INTRODUCTION lightly cemented silty-clayey-sand (Cavallaro et al.,

2003a; Cavallaro & Maugeri, 2003).
Currently, CPT static sounding is one of the most The pre-consolidation pressure σ’p and the over-
popular field study as its specificity provides rich consolidation ratio OCR ¼σ’p =σ’vo were evaluated
quasi-continuous data at depth. The CPT results from the 24h compression curves of incremental
enable to infer the soil profile as well as strength and loading (IL) oedometer tests.
stiffness parameters even at great depths in a cost Moreover, Marchetti’s flat dilatometer tests
effective way (Robertson et al., 1986, Kulhawy & (DMT) were used to assess OCR and the coefficient
Mayne, 1990; Fellenius & Eslami, 2000; Powell & of earth pressure at rest Ko following the procedure
Lunne, 2005; Robertson, 2009; Mayne, 2020). suggested by (Marchetti, 1980).
The results presented in the article add new data to For depths of about 15 m, DMT results show an
the documented knowledge. The results of field studies OCR from 1 to 4.5 (Ko = 0.5 ÷ 1.0).
and guidelines on the soil of Noto area have been The OCR values inferred from oedometer tests
described in previous publications (Cavallaro et al., (OCR from 1 to 3) are lower than those obtained
2003a, 2003b; Cavallaro & Maugeri, 2003), and the from in situ tests.
data from soundings are the basis for determining soil One possible explanation of these differences
parameters (Mayne, 2016) and engineering could be that lower values of the pre-consolidation
calculations. pressure σ’p are obtained in the laboratory because of
In the light of the above, linking the numerical sample disturbance.
results of CPT static sounding to the genesis of soil The value of the natural moisture content wn
will bring significant benefits to geotechnicians and prevalently range from between 12 - 37 %. Character
engineering geologists in terms of data interpretation. istic values for the Atterberg limits are: wl = 37 - 69
This paper intends to propose a critical evaluation % and wp = 17 - 22 %, with a plasticity index of PI
of geotechnical parameters (Cavallaro, 2020) with = 15 - 47 %. The obtained data indicate a low degree
special attention for small strain shear modulus Go. of homogeneity with depth of the deposits (Cavallaro
et al., 2003b; Cavallaro & Maugeri, 2003).
Shear modulus G and damping ratio D of Noto
2 INVESTIGATION PROGRAM AND BASIC soil were obtained also in the laboratory by resonant
SOIL PROPERTIES column tests (RCT). These tests were performed on
Shelby tube specimens retrieved from Noto site. The
The Pliocene Noto deposits of Trubi Formation Resonant Column/Torsional shear apparatus were
mainly consist of a medium stiff, over-consolidated used for this purpose.
DOI: 10.1201/9781003308829-41
311
Figure 1. Borehole and index properties of Noto soil; where R: Landfill; S(L): Silty Sand; L(A): Clayey silt; S: Sand; S,G:
Sand and Gravel.
3 EVALUATION GEOTECHNICAL
PARAMETERS FROM CPT
CPT tests are widely used to investigate the subsoil

in order to obtain information on the different phys for clays and sands.
ical-mechanical properties of the soils. Empirical
laws are therefore available for determining the main
geotechnical characteristics of soils.
The following geotechnical parameters have been
determined:
- Total unit weight g;
- Angle of shear resistance f’;
- Undrained resistance cu;
- Shear modulus Go.
It is possible to evaluate the total weight unit g
by empirical correlations of CPT in situ
measurement:
- Mayne et al. (2010): Figure 2. Total unit weight g by empirical correlations
based on CPT.
Figure 2 shows a comparison between the results

obtained in the laboratory and those obtained by the
empirical relationships proposed.
It is possible to observe that the results
where z = depth [m], fs = sleeve friction resistance obtained are comparable even though the
[kN/m2] and qt = corrected cone resistance [kN/m2]. relationship proposed by Robertson & Cabal
- Robertson & Cabal (2010): (2010) is better aligned with the experimental
results.
Numerous empirical correlations have been pro
posed for the determination of f’:
- Meyerhof (1951):
where gw = water weight unit [kN/m3], Rf = (fs/qt) =

friction ratio [-], qt = corrected cone resistance [kN/
m2] and Pa = atmospheric pressure, expressed in the
same unit of measurement of qt. where qc = cone resistance [kg/cm2].
- Mayne & Peuchen (2012): - De Beer (1965):
312
Regarding the results obtained for the angle of
shear resistance ϕ', reported in Figure 3, it is
observed that the values obtained in the laboratory
where σ’VO = effective vertical stress [kg/cm2]. by direct shear tests intersect all the empirical rela
tionships proposed. Probably because these rela
- Dourgunouglu & Mitchell (1975): tions fail to, evaluate correctly the stratigraphic
variations of the soil. High values of f’ were
obtained in correspondence of the most superficial
layers where no laboratory data is available. Over
all, the Robertson & Campanella (1983) equation
the terms qc and σ’vo are expressed in the same unit seems to approximate better the results of the direct
of measurement [kg/cm2]. shear tests.
As for the evaluation of the undrained resistance
- Robertson et al. (1983): cu the following empirical expressions were used:
- Lunne et al. (1976):
the terms qc and σ’vo are expressed in the same unit

of measurement [kPa].
where PI = plasticity index.
- Kulhawy & Mayne (1990):
- Lunne & Kleven (1981) and Lunne at al. (1997):
where σatm = pa = atmospheric pressure, expressed in

the same unit of measurement of qc [kPa]. where Nk = empirical factor for bearing capacity
dependent on depth and opening angle of the pene
- Marchetti (1997): trometer cone (11 - 19).
Figure 4 shows the cu values obtained by labora
tory tests and empirical correlations. The laboratory
data aligns well with the cu values derived from the
correlations. High values of undrained resistance
were obtained by the expression of Lunne & Kleven
(1981) on the surface layers and at a depth of
where KD = horizontal stress index by about 16 m.
~33 · · · KD Þ.
DMT ðqc =σ’vo ¼
- Jefferies & Been (2006):
where fcv = constant volume friction angle and

Qtn ¼ ðqt -σvo Þ=σ’vo normalized cone resistance.
Figure 4. Undrained resistance cu by empirical correlations

based on CPT.
4 EVALUATION GO FROM CPT
The small strain shear modulus Go can be evaluated by

laboratory Resonant Column Tests (RCT) (Capilleri
et al., 2014; Castelli et al., 2016a) or in situ by Down
Figure 3. Angle of shear resistance ϕ’ by empirical correl Hole tests and SDMT (Castelli et al., 2016b; Cavallaro
ations based on CPT. et al., 2012a, 2012b; Marchetti et al., 2008; Cavallaro
313
& Grasso, 2021) using the relationships: Go = ρVs2 where: Go, σ’v and pa are expressed in the same unit;
(where: ρ = mass density) based on theory of elasticity. pa = 1 bar is a reference pressure; gD and Ko are
An attempt was made to evaluate the small strain respectively the unit weight and the coefficient of
shear modulus by means of the following empirical earth pressure at rest, as inferred from DMT results
correlations based on penetration tests results, CPT, according to Marchetti (1980).
SPT and DMT or laboratory results available in lit - Jamiolkowski et al. (1995):
erature (Figure 5).
- Imai & Tomaichi (1990):
The Jamiolkowski et al. (1995) method was

for any soil. applied considering a given profile of void ratio.
- Mayne & Rix (1993): The values for parameters, which appear in equa
tion (20), are equal to the average values resulting
from laboratory tests performed on quaternary Italian
clays and reconstituted sands.
Equation (20) incorporates a term, which
for clayey strata;
expresses the void ratio; the coefficient of earth pres
where: Go and qc are both expressed in [kPa] and
sure at rest only appears in equation (19). However
e is the void ratio. Equation (15) is applicable to clay
only equation (19) tries to obtain all the input data
deposits only.
from the DMT results.
As regard Noto soil the Go values obtained with
- Simonini & Cola (2000): the methods above indicated for CPT and SPT are
plotted against depth in Figure 5.
It is also possible to evaluate the small strain

shear modulus using the relation Go = ρ · Vs by the
following equations proposed by Ohta & Goto
(1978) and Yoshida and Motonori (1988) for the
shear waves velocity Vs:
- Ohta & Goto (1978):
where: Vs = shear wave velocity (m/s), N60 = Figure 5. Small strain shear modulus Go by empirical cor
number of blow/feet from SPT with an Energy Ratio relations based on CPT and SPT.
of 60 %, Z = depth (m), FG = geological factor
(clays = 1.000, sands = 1.086), FA = age factor The results obtained show a greater similarity
(Holocene = 1.000, Pleistocene = 1.303) between the Go results obtained by the empirical cor
- Yoshida and Motonori (1988): relations proposed for SPT, which, moreover, are
quite close together. The Go values, obtained through
the correlation equations proposed for CPT are more
dispersed and higher.
The lowest values of the shear modulus are obtained
by the equation proposed by Mayne & Rix (1993). Only
where: Vs = shear wave velocity (m/s), NSPT = by the Down Hole test it is possible to identify the rapid
number of blows from SPT,σ’VO =vertical pressure, β increase of Go at a depth of 20 m in correspondence
= geological factor (any soil=55, fine sand=49). with some layer characterized by higher mechanical
- Hryciw (1990): characteristics that both CPT and SPT cannot identify.
The N60 values, experimentally determined
during SPT, did not show any important variation in
the transition zone at depth of 20 m, where the char
acteristics of the soil change from silty sand to sand
and then to sand with gravel.
314
Standard Penetration Tests were performed at
intervals from 1.5 to 3.0 m. The quite large interval
used could explain why the thin sand layers were not
detected. Consequently, the obtained Go values, in
the transition zone, resulted to be quite low.
Unfortunately, the depth investigated by DMT is
not able to intercept the most consistent layers of
sand and sand with gravel. However, the method by
Hryciw (1990) is the best one to follow the trend of
the results obtained from the Down Hole tests, as
can be seen in Figure 6.
Figure 7. Go from Down Hole vs Go by empirical

correlation.
paper. On the basis of the data shown it is possible to

draw the following conclusions:
- empirical correlations between the small strain
shear modulus and penetration test results were
used to infer Go from SPT, CPT, DMT and Down
Hole. This comparison clearly indicates that
a certain relationship exists between Go and the
penetration test results, which would encourage to
establish empirical correlations for a specific site.
Figure 6. Go from Down Hole, empirical correlation and
RCT. This approach makes it possible to consider the
spatial variability of soil properties in a very cost
effective way.
From a comparison between Figure 5 and - The values of Go were compared to those meas
Figure 6 all the considered methods show very dif ured with DMT and DH tests. This comparison
ferent Go values of the Pliocene Noto soil. On the indicates that some agreement exists between
whole, Down Hole seem to provide the most accur empirical correlations by DMT and DH test.
ate trend of Go with depth even if the available data - relationships like those proposed by Jamiolkowski
are unable to investigate the behavior of the soil for et al. (1995) seem to be capable of predicting Go
depths greater than 20 m. The method by Jamiolk profile with depth only in the initial strata. The
owski et al. (1995) was applied considering a given accuracy of these relationships could obviously be
profile of void ratio but while guaranteeing continu improved if the parameters, which appear in the
ity of results, it fails to intercept the most consistent equations, were experimentally determined in the
layers of sand and sand with gravel. laboratory for a specific site.
In Figure 6 the RCT results are also reported. The - Down Hole test only is probably able to investigate
data obtained also show as the dynamic laboratory correctly the various layers of soil, identifying even
tests are able to interpret the Go trend obtained from the smallest variations in the mechanical
the Down Hole test. characteristics.
Figure 7 reports a comparison between DH test
measurements and the corresponding empirical correl
ations. Since the DH test was performed with an inter
val of about 5 m, it was necessary to interpolate the
REFERENCES
intermediate values. From the comparison obtained it Capilleri, P., Cavallaro, A. & Maugeri, M., 2014. Static and
is possible to observe a more disordered initial phase Dynamic Characterization of Soils at Roio Piano (AQ).
in correspondence with the more superficial layers Italian Geotechnical Journal, vol. XLVIII, no. 2, Aprile -
while subsequently a horizontal linear relationship is Giugno 2014, Patron Editore, 38–52.
highlighted with the exception of the depth of Castelli, F., Cavallaro, A., Grasso, S. & Ferraro, A., 2016a.
16.20 m. In Situ and Laboratory Tests for Site Response Analysis
in the Ancient City of Noto (Italy). Proc. of the 1st
IMEKO TC4 Int. Workshop on Metrology for Geotech
nics, Benevento, 17-18 March 2016, 85–90.
5 CONCLUDING REMARKS Castelli, F., Cavallaro, A. & Grasso, S., 2016b. SDMT Soil
Testing for the Local Site Response Analysis. Proc. of
A site characterization for a possible advanced the 1st IMEKO TC4 Int. Workshop on Metrology for
dynamic response analysis has been presented in this Geotechnics, Benevento, 17-18 March 2016, 143–148.
315
Cavallaro A., Massimino M. R., Maugeri M., 2003a. Noto Scandinavian Soft to Medium Stiff Clays. Canadian
Cathedral: Soil and Foundation Investigation. Construc Geotechnical Journal, 13, 430–441.
tion and Building Materials, no. 17, 2003, 533–541. Lunne, T. & Kleven, A., 1981. Role of CPT in North Sea
Cavallaro, A., Maugeri, M. & Ragusa, A., 2003b. Small Foundation Engineering. Symposium on Cone Penetra
Strain Stiffness from in Situ and Laboratory Tests for tion Engineering Division, ASCE, 49–75.
the City of Noto Soil. Proc. of the 3nd International Lunne, T, Robertson, P. & Powell, J., 1997. Cone Penetra
Symposium on Deformation Characteristics of Geoma tion Testing in Geotechnical Practice. Blackie Academic
terials, Lyon, 22-24 September 2003, 267–274. and Professional, Chapman and Hall.
Cavallaro, A. & Maugeri, M., 2003. Site Characterization Marchetti, S., 1980. In Situ Tests by Flat Dilatometer. Jour
by In-Situ and Laboratory Tests for the Microzonation nal of Geotechnical Engineering, ASCE, 1980, no. GT3.
of Noto. Proc. of Symposium on Geotechnical Analysis Marchetti, S., 1997. The Flat Dilatometer Design
of Seismic Vulnerability of Monuments and Historical Applications. Proc. of the 3rd Geotechnical Engineering
Sites, Catania, 15 November 2001, Geotechnical Ana Conference, Cairo University, 5-8 January 1997.
lysis of Seismic Vulnerability of Monuments and Histor Marchetti, S., Monaco, P., Totani, G. & Marchetti, D.,
ical Sites, Patron Editor, 2003, Edited by Maugeri 2008. In Situ Tests by Seismic Dilatometer (SDMT).
M. and Nova R., 237–256. From Research to Practice in Geotechnical Engineer
Cavallaro, A., Grasso, S., Maugeri, M. & Motta, E., 2012a. ing, ASCE Geotech. Spec. Publ. no. 170, New Orleans,
An Innovative Low-Cost SDMT Marine Investigation 9-12 March 2008.
for the Evaluation of the Liquefaction Potential in the Mayne, P.W. & Rix, G.J., 1993. Gmax-qc Relationships for
Genova Harbour (Italy). Proc. of the 4th Int. Conf. on Clays. Geotechnical Testing Journal, vol. 16, no. 1, 54–60.
Geotechnical and Geophysical Site Characterization, Mayne, P.W, Peuchen, J. & Bouwmeester, D., 2010. Soil
ISC’4, Porto de Galinhas, 18-21 September 2012, vol. Unit Weight Estimated from CPTu in Offshore Soils.
1, 2013, 415–422. Frontiers in Offshore Geotechnics II (Proc. ISFOG
Cavallaro, A., Grasso, S., Maugeri, M. & Motta, E., 2012b. 2010, Perth), Taylor & Francis Group, London:
Site Characterisation by in Situ and Laboratory Tests of 371–376.
the Sea Bed in the Genova Harbour (Italy). Proc. of the Mayne, P.W. and Peuchen, J., 2012. Unit Weight Trends
4th Int. Conf. on Geotechnical and Geophysical Site with Cone Resistance in Soft to Firm Clays. Geotech
Characterization, ISC’4, Porto de Galinhas, 18-21 Sep nical and Geophysical Site Characterization 4, vol. 1,
tember 2012, vol. 1, 637–644. CRC Press, London: 903–910.
Cavallaro, A., 2020. Use of CPT for the Study of the Mayne, P.W., 2016. Evaluating Effective Stress Parameters
Dynamic Properties of the Soils. Proc. of the IMEKO and Undrained Shear Strengths of Soft-Firm Clays from
TC4 International Conference on Metrology for Archae CPTu and DMT. Proc. of the 5th International Confer
ology a Cultural Heritage, Trento, 22-24 October 2020, ence on Geotechnical and Geophysical Site Character
242–247. ization, ISC 2016, vol. 1, 400, July 2016, 19–39.
Cavallaro, A. & Grasso, S., 2021. Small Shear Strain Mayne, P.W., 2020. Use of In-Situ Geotechnical Tests for
Modulus Degradation by the Seismic Dilatometer Foundation Systems. Proc. Széchy Károly Emlékkonfer
Marchetti Tests (SDMTs). Proc. of the 6th International encia, no. 402, 12–73 September 2020.
Conference on Geotechnical and Geophysical Site Meyerhof, G.G., 1951. The Ultimate Bearing Capacity of
Characterisation, Budapest, 26-29 September 2021. Foundations. Geotechnique, 2, 301–332.
De Beer, E.E., 1965. Bearing Capacity and Settlement of Ohta, Y. & Goto, N., 1978. Empirical Shear Wave Velocity
Shallow Foundations on Sand. Proc. of the Symposium Equations in Terms of Characteristic Soil Indexes.
on Bearing Capacity and Settlement of Foundations, Earthquake Engineering and Structural Dynamics, vol.
Duke University, 15–33. 6, 1978.
Dourgunouglu, H.T. & Mitchell, J.K., 1975. Static Penetra Powell, J.J.M. & Lunne T., 2005. Use of CPTU Data in
tion Resistance of Soils: I - Analysis. Proc. of the ASCE Clays/Fine Grained Soils. Studia Geotechnica et
Specialty Conference on In-situ Measurement of Soil Mechanica, vol. 27, no. 3-4, 431, 29–66.
Parameters. Raleigh, N.C., 151–171. Robertson, P. K., Campanella, R.G. & Wightman A., 1983.
Fellenius, B.H. & Eslami, A., 2000. Soil Profile Interpreted SPT-CPT Correlations. Journal of the Geotechnical
from CPTu Data. Proc. of the Geotechnical Year 2000, Engineering Division, vol. 108, no. GT 11, 1449–1459.
Asian Inst. of Technology, Thailand, 27-30 November Robertson, P.K., Campanella. R.G., Gillespie, D. &
2000, p. 18. Greig J., 1986. Use of Piezometer Cone Data. In-Situ’86
Hryciw, R.D., 1990. Small Strain Shear Modulus of Soil by Use of In-situ testing in Geotechnical Engineering, GSP
Dilatometer. JGED, ASCE, vol. 116, no. 11, 1700–1715. 6, ASCE, Reston, VA, Specialty Publication, SM 92,
Jamiolkowski, M., Lo Presti, D. C. F. & Pallara, O., 1995. 1263–1280.
Role of In-Situ Testing in Geotechnical Earthquake Robertson P. K., 2009. Interpretation of Cone Penetration
Engineering. Proc. of the 3rd Int. Conf. on Recent Tests a Unified Approach. Canadian Geotechnical Jour
Advances in Geotechnical Earthquake Engineering and nal, vol. 46, 435, no. 11, 1337–1355.
Soil Dynamics, St. Louis, Missouri, 2-7 April 1995, vol. Robertson, P. K. & Cabal, K. L., 2010. Guide to Cone
II, 1523–1546. Penetration Testing for Geotechnical Engineering,
Jefferies, M.G. & Been, K., 2006. Soil Liquefaction Gregg Drilling & Testing.
A Critical State Approach. Taylor & Francis, 478 Simonini, P. & Cola S., 2000. On the Use of the Piezocone
pages. to Predict the Maximum Stiffness of Venetian Soils.
Kulhawy, F.H. & Mayne, P.W., 1990. Manual on Estimat Journal of Geotechnical and Geoenvironmental Engin
ing Soil Properties for Foundation Design. Report EL eering, vol. 126, no. 4, 378–382.
6800, Electric Power Research Institute, Palo Alto, Ca. Yoshida, Y., and Motonori, I. 1988. Empirical Formulas of
Lunne, T.J., de Ruiter, J. & Eide, O., 1976. Correlation SPT Blow-Counts for Gravelly Soils. Proc. of ISOPT-1,
between Resistance and Vane Shear in Some Orlando (USA).
316
© 2022 the Author(s), ISBN 978-1-032-31259-0
Validating cone penetration test in partially drained conditions using

a simplified numerical modelling method
R.W.L. Chia, Z.Z. Wang & S.H. Goh
Department of Civil and Environmental Engineering, National University of Singapore, Singapore
ABSTRACT: A simplified numerical modelling method, the Press-Replace Method (PRM), is applied to simu
late the cone penetration process under conditions that range from fully undrained to fully drained. In this study,
rigorous validations of the PRM technique for simulating the cone penetration test were performed. The results
for a fully undrained and drained penetration are compared against published information using cone factor solu
tions. Coupled-consolidation analyses at different cone penetration rates were then carried out for a smooth cone,
and the results processed to obtain the dimensionless backbone curves for comparison with published data. The
consolidation analyses were further extended to a rough cone subjected to different penetration rates, for which
there is little, if any, numerical simulation results reported in the literature. The results and computed backbone
curves obtained using the smooth and rough soil-cone interfaces are presented and compared in this paper.
1 INTRODUCTION associated with the large distortions of the discretized

finite element mesh, many of which require add
The cone penetration test (CPT) is a widely used site itional subroutines. In this paper, a simplified
investigation tool for the characterization of soil and method, the Press-Replace Method (PRM), is used to
the estimation of engineering properties through empir simulate the process of cone penetration. PRM can
ical and theoretical correlations. This method of testing be easily implemented in standard small-strain finite
provides a continuous profiling of the ground. The element programs without the need for additional
cone penetration in a clayey soil would typically be subroutines to execute remeshing or a mesh update.
considered as an undrained process, whereas in PRM was first introduced to model suction
a sandy soil, it is assumed as a drained process. Prob anchors (Andersen et al 2004), and was refined and
able issues may arise when penetration occurs in inter employed to model the installation of piles and spud-
mediate soil layers whereby the penetration process is can (Engin et al., 2015; Sivasithamparam et al.,
neither undrained nor drained. The majority of existing 2015; Wang & Goh, 2018). A preliminary study of
theoretical and empirical solutions to determine soil cone penetration was conducted by Engin et al.
properties are based on the assumption of either a fully (2015) by comparing cone factors with other pub
undrained or a fully drained condition. As such, par lished literatures. Paniagua et al. (2014) compared
tially drained conditions could complicate the inter numerical results obtained through PRM with labora
pretation of results. tory scale cone penetration experiments. PRM was
In this study, finite element analysis is employed to also used to simulate undrained cone penetration tests
model the penetration process in partially drained con by Lim et al. (2018), who validated its performances
ditions, by simulating the penetration at different rates. in stress magnitude, principal stress direction and
Experimental studies in the published literature indi excess pore pressure against other existing numerical
cate that, when the rate of pore pressure dissipation is approaches such as LDFE, SSFE and MPM. More
faster than the penetration rate, the soil near the cone recent advances in PRM for cone penetration prob
consolidates during penetration. As a result, a higher lems were presented by Paniagua et al. (2021), in
tip resistance is expected as compared to undrained which soil stresses and pore pressure measurements
conditions (Lehane et al., 2009; Oliveira et al. 2011). of a model scale penetration test with varying pene
To model the large deformation scenarios like tration rates were compared against the results from
cone penetration, different types of modeling tech PRM back-analyses. However, the study only con
niques (e.g. Updated Lagrangian, Arbitrary Lagran sidered three different orders of penetration rates.
gian Eulerian (ALE) and Remeshing Interpolation As the application of PRM to simulate cone pene
Technique combined with Small-Strain (RITSS)) tration test in partially drained conditions has not been
have been used to overcome the numerical issues widely reported in the literature, an attempt to
DOI: 10.1201/9781003308829-42
317
rigorously validate its performance in partially drained
conditions is presented in this paper. This paper first
provides a concise summary of PRM and its imple
mentation in the finite-element program PLAXIS 2D.
Secondly, results obtained in partially drained condi
tions will be processed and compared against informa
tion from published literatures. Lastly, the influence of
different soil-cone roughness on the penetration pro
cess under partially drained conditions will be briefly
examined.
Figure 1. An example of PRM technique for cone penetra

2 METHODOLOGY tion; a) Initial Setup, b) First “press” phase c) “replace”
phase d) Subsequent “press” phase.
PRM models penetration problems via a displacement
control scheme (Engin et al., 2015), in which 3 NUMERICAL MODELLING
a prescribed displacement is first applied to the cone
shaft to simulate the incremental pressing of the cone
3.1 Model dimensions and soil models
into the soil. This is followed by the replacement of
that slice of the soil (through which the cone has The FEM program used in this study is PLAXIS 2D
incrementally penetrated) by the cone material. CONNECT Edition V21. Investigations were carried
This systematic ‘press-and-replace’ process is pre using a standard cone diameter (D) of 36mm and
sented in Figure 1. Figure 1 illustrates a standard a tip angle of 60° with an axisymmetric setup. As
36mm diameter cone with some initial embedment. this paper involves a comparison study using pub
The intended downward incremental movements of lished data from multiple sources (Yi et al., 2012;
the cone are represented by slices below the cone Sheng et al., 2014; Ceccato et al., 2016; Orazalin
that are set up during the mesh generation process. et al, 2018) with different model set-ups and
The thickness of the slices is defined as ts. Interface assumptions, the following standardized procedures
elements were added to the sides of these slices to were adopted in all the PRM models of this study for
model the interaction between the soil and cone. simulating the different cases. These are: 1) As
These interfaces were extended slightly beyond the recommended by Engin et al. (2015), Δuy = ts = iint-ext
corners of the slices; these interface extensions help of 0.1D will be adopted; 2) The initial setup begins
reduce stress fluctuations around the angled corners with a soil body, which is followed by the first
of the cone. The length of these extensions is “press” that represents the embedment of the cone
defined as iint-ext. (Figures 2, 3) When required, the overburden stress in
The process starts with a prescribed downward the soil will be modeled using a surcharge loading
displacement uy (= 0.1D, where D is the cone diam imposed by a 0.05m thick linear-elastic soil layer
eter) of the cone over the thickness of one slice ts with the appropriate unit weight. The model dimen
(also equal to 0.1D). This represents a “press” sions adopted will follow those reported in the
phase. Within the same phase, the interfaces and respective source literatures. Key details of these lit
the interface extensions bounding the cone are acti eratures and their soil models are summarized in
vated. Upon completion of this phase, the stress Table 1. Figure 2 illustrates a typical finite element
states of the soil elements in the deformed mesh setup used in this study.
will correspond to those caused by the pressing of
the cone over the prescribed incremental depth. In
3.2 Soil-cone interface
the next phase, the soil material that was originally
occupying the slice below the cone (in the preced Interfaces elements were used to model the inter
ing phase) will be changed to the cone material, action between the soil and cone surfaces. With each
thus making this the “replace” phase. To increase incremental advancement (Δuy) of the cone, the
the penetration depth to 0.2D, the “press” phase is interface elements bounding the entire penetrometer
repeated and uy is increased from 0.1D to 0.2D. As are activated. Interface elements allow for slippage
the cone advances downward, the cycle of “press” (relative movement between the cone and the adja
and “replace” is repeated. Accordingly, more slices cent contacting soil) to occur as the cone displaces
must be created for greater penetration depths. To downwards.
ease calculation time, after the initial “press” phase, Separate interface materials were created and
successive “press” and “replace” phases can be assigned to the shaft and cone of the penetrometer. The
combined into one phase. Negligible or little differ general properties of the interface materials follow the
ence in the computed cone resistance was observed neighboring soil; however, shear strength properties
(Lim, 2017). were varied depending on the desired roughness.
318
Table 1. Key set-up details of respective literatures used in this study.
Soil Properties Input
Author/ Soil model Soil-Cone φ Overburden
Approach Input Contact E’(kPa) λ κ (°) M k(ms-1) e0 K0 OCR stress (kPa)
Yi et al Mohr’s Cou Smooth 0.026 + - - 23 - 5E-10 - 0.6 1 0

(2012)/ lomb Model* 380z
Updated
Lagrangian
Sheng et al. Modified Smooth - 0.3 0.05 - 1 1E-7 2 0.5 1.394 50
(2014)/ Cam-clay
Updated model
Lagrangian
Ceccato Modified Smooth - 0.205 0.04 - 0.92 1.02E-3 to 1.41 0.68 1 50
et al. Cam-clay 1.02E-8 ^
(2016)/ model
Material
Point
Method
Orazalin Modified Smooth - 0.205 0.04 - 0.92 1.02E-3 to 1.41 0.68 1 50
(2018)/ Cam-clay 1.02E-8 ^
Lagrangian model
FEM with
an auto
mated pro
cedure for
remeshing
Note:
E’: effective Young’s modulus; φ: effective friction angle; λ: virgin compression index; κ: swelling index; z: depth; k:
hydraulic conductivity; e0: initial void ratio; K0: Coefficient of earth pressure at rest; OCR: overconsolidation ratio.
^: Hydraulic conductivity is varied to obtain partial drainage conditions
*: A Drucker-Prager Soil model was adopted in Yi et al. (2012)
For undrained penetrations, the results are presented

in the form of the cone tip factor (Nc), which is calcu
lated as Nc = (qc-σv0)/su, where qc is the calculated cone
tip resistance from the analysis, su is the equivalent
undrained shear strength corresponding to the friction
angle φ or friction coefficient M prescribed in the
model, and σv0 denotes the total overburden stress (Lu
et al., 2004).
The calculated Nc values from the present PRM
analysis are plotted against the normalized cone
depth (H/D, where H is the penetration depth and
D is the cone diameter), as shown in Figure 3. The
Figure 2. Typical finite element mesh of a first “press” results agree well with the data reported by Yi et al.
phase with 18,359 elements and 161.903 nodes. (2012), with a stabilized Nc value of approximately
9. This stabilized Nc value also agrees well with
4 RESULTS other published solutions (Teh & Houlsby, 1991; Yu
et al., 2000; Abu-Farakh et al., 2003; Lu et al., 2004;
4.1 Cone factor comparison for undrained and Walker & Yu, 2006).
drained penetration For a drained penetration, the cone factor Nq is
used for comparison. It is calculated as Nq=qc’/σ’v0
Using a case study from Yi et al. (2012), preliminary (Yu, 2004; Ahmadi et al., 2005), in which qc’ denotes
investigations for both undrained and drained condi the effective cone resistance obtained from the ana
tions were carried out and compared. To achieve an lysis and σ’v0 the effective vertical stress. Figure 4
undrained penetration, a penetration rate of shows the computed Nq values plotted against the
0.0278 m/s is selected (Yi et al., 2012). Similarly, to normalized cone depth (H/D). The results obtained
achieve drained penetration, a slow penetration rate from the present PRM analysis agree well with the
of 1E-10 m/s was used (Yi et al., 2012). reported solution from Yi et al. (2012). The stabilized
319
Nq value of approximately 7.5 also falls within the The penetration rate is represented using the non-
range of 5 ~ 8 reported in other publications. dimensional velocity V (= vD/cv) proposed by Ran
dolph & Hope (2004), in which v, D and cv denote the
cone penetration velocity, cone diameter and the soil’s
coefficient of consolidation respectively. The cv value
in the MCC model is given by Equation 1, in which
e0 is the initial void ratio, k and σ’v0 represents the
hydraulic conductivity and effective vertical stress, λ
denotes the virgin compression index and γw is the
unit weight of water.
4.2.1 Comparison with Sheng et al. (2014)

Results obtained using PRM is presented in
Figure 5. Sheng et al. (2014) associated an undrained
and drained penetration in a soil body (with perme
Figure 3. Comparison of calculated Nc with literature (Yi ability 10-7m/s) to a penetration rate of v = 200cm/s
et al., 2012) and published solutions. and v = 0.0002cm/s respectively. Figure 5 plots the
variation of the net cone resistance (qc-σv0) with nor
malized depth (H/D), in which good agreement was
obtained between the results from the PRM analyses
and those reported by Sheng et al. (2014) for the
drained and undrained penetration conditions.
In the same figure, the results from consolidation
analyses at various penetration rates proposed by
Sheng et al. (2014) are also shown. At a penetration
depth 10H/D, the net cone resistances of 213kPa and
137 kPa are achieved for the drained and undrained
conditions respectively. Penetration velocities
between v = 200cm/s and v = 0.0002cm/s resulted in
the net cone resistances falling between the range of
137 to 213 kPa. The results from the various penetra
tion rates were processed, following which the
dimensionless backbone curve as illustrated in
Figure 6 was obtained. In Figure 6, a normalized
cone resistance (qn) (Randolph& Hope, 2004) is
Figure 4. Comparison of calculated Nq with literature (Yi plotted against a normalized penetration rate repre
et al., 2012) and other published solutions. sented by v/k, where v and k denote the penetration
velocity and hydraulic conductivity respectively.
Instead of adopting the normalized velocity
4.2 Consolidation analysis of the penetration V proposed by Randolph & Hope (2004), Sheng et al.
process at different cone penetration rates suggested the use of the normalized term v/k, which
provides a more direct relationship that could facili
In this section, the effect of different cone penetra
tate the processing of data. The normalized cone
tion rates is explored and validated. The effect of
resistance qn is defined as:
varying the penetration rates will result in partial
drainage conditions (Suzuki and Lehane, 2014). Par
tial drainage occurs when there is concurrent gener
ation and dissipation of excess pore pressures during
the penetration process. In this section, the results
from two published numerical case studies are
adopted for comparison: (i) Sheng et al. (2014) and in which qref is the reference cone resistance
(ii) Ceccato et al. (2016) and Orazalin et al. (2018). obtained from a fully undrained penetration. From
For both studies, the modified Cam-clay (MCC) the fitted dimensionless backbone curve of Figure 6,
model (Roscoe & Burland, 1968) is adopted to simu a stabilized qn value at v/k = 0.1 can be obtained for
late the soil behavior in a more realistic manner. the drained condition. The qn values predicted by
320
PRM and Sheng et al (2014) at v/k = 0.1 are 1.56 4.2.2 Comparison with Ceccato et al. (2016) and
and 1.68 respectively with a 7% difference in value. Orazalin et al. (2018)
PRM predicts an undrained penetration response In this section, the PRM results will be studied com
when v/k > 200,000 (approximately), whereas pared against those reported by Ceccato et al. (2016)
a drained behavior is observed when v/k < 20 and Orazalin et al. (2018). These two published
(approximately). For intermediate values of works adopted different numerical techniques to
v/k between 20 and 200,000, the effect of partial simulate the same cone penetration set-up proposed
drainage is observed. A similar behavior was by Ceccato et al. (2016). Both Ceccato et al. and
observed in the results of Sheng et al. (2014), for Orazalin et al. adopted a MCC soil model to capture
which the best fit curve showed partial drainage the soil behavior. However, for simulating the pene
occurring between v/k values of 30 and 400,000 tration process, Ceccato et al. (2016) used the Mater
(approximately). Overall, there is good agreement ial Point Method (MPM), whereas Orazalin et al.
between the PRM results and those reported by Sheng (2018) adopted the Lagrangian finite element
et al. (2014) in terms of the qn values as well as the method with an automated remeshing procedure.
intermediate values of v/k (indicating partial drain In this section, the normalized net tip resistance
age). For this example, these comparisons affirm the (qn.net) is plotted against the normalized velocity
validity of the PRM to predict reasonable results V (Randolph & Hope, 2004) to obtain the dimen
under partial drainage conditions. sionless backbone curve. The normalized net tip
resistance is defined as follows:
Normalized Nettip Resistance ðqn:netÞ ¼ ¼

qc - σv0
qref - σv0
Using PRM, the cone penetration at different rates

were modeled using consolidation analysis.
A comparison of the computed cone resistances for
the fast and slow penetration are presented in
Figure 7 and Figure 8 respectively. These results plot
the total cone resistance against the normalized
penetration displacement in an undrained (fast) and
drained (slow) penetration. The normalized penetra
tion displacement is defined as the ratio of the pene
tration displacement to the cone diameter.
Figure 7 shows good agreement between the results
Figure 5. Comparison of undrained and drained net cone reported by Ceccato et al. (2016) and Orazalin et al.
resistance between PRM and literature (Sheng et al., 2014). (2018), while the PRM an undrained cone resistance
which is about 7% higher. The results obtained by
Lim (2017) are also plotted. Overall, the PRM results
for the undrained response of this particular case study
Figure 7. Comparison of calculated total cone resistance

Figure 6. Comparison of dimensionless backbone curve with literature (Ceccato et al., 2016; Lim, 2017; Orazalin
between PRM and literature (Sheng et al., 2014). et al., 2018) in an undrained penetration.
321
compare favorably with those reported by Orazalin 4.3 Effects of cone roughness
et al. (2018), Ceccato et al. (2016) and Lim (2017).
This section explores the influence of cone roughness
For a drained penetration, Figure 8 shows that
with a case example taken from Yi et al. (2012). Cone
PRM predicts a greater cone tip resistance of 263kPa
roughness affects the stress-transfer interaction between
compared to the MPM value of 210 kPa reported by
the cone and soil surfaces, and are characterized by the
Ceccato et al. (2016), which is a difference of about
friction angle φint and cohesion cint’ assigned to the
25%. Despite the good agreement observed for the
interface elements.
undrained response, the cone tip resistance obtained
For this study, the interface cohesion cint’ is
by Orazalin et al. (2018) for the drained response is
assumed to be negligible, although a small value of
about 250kPa, or about 20% higher than the MPM
0.001 kPa is used to avoid numerical instability.
value. For the drained penetration, Orazalin et al.’s
Hence, the interface friction φint is the only param
results are closer to those obtained using PRM.
eter controlling the interaction between the 2 sur
The cone resistance results at different penetration
faces. The analyses are performed for φint values of
rates obtained by Ceccato et al. (2016), Orazalin
0, 8°, 16° and 23° to capture the transition from
et al. (2018) and the PRM approach are processed
a perfectly smooth to a fully rough cone-soil contact.
and plotted, in the form of qn.net against the normal
Figure 10 plots the PRM computed undrained and
ized velocity V, to produce the backbone curves
drained responses for the perfectly smooth (φint = 0°)
shown on Figure 9. This figure shows that the PRM
and fully rough (φint = 23°) interface conditions, in the
predictions of the normalized tip resistance qn.net are
form of the normalized depth (H/D) against qc. For
approximately the average of those predicted by
both the undrained and drained conditions, a higher qc
Ceccato et al. (2016) and Orazalin et al. (2018).
value is obtained when φint = 23°, the increase being
much more significant for the drained response.
Figure 8. Comparison of calculated total cone resistance

with literature (Ceccato et al., 2016; Orazalin et al., 2018)
in a drained penetration.
Figure 10. Illustration of cone tip resitance for an undrained
and drained penetration, when φint/° = 0 and φint/° = 23.
Subsequently, PRM analyses are performed for

different penetration rates with different values of
φint. The results are summarized and presented in the
form of backbone curves obtained by plotting qn.net
against V, as shown in Figure 11 for 3 different
depths (15H/D, 20H/D, 25H/D). Four distinctive
back bones curves may be fitted to the data. These
correspond to φint values of 0, 8°, 16° and 23°, with
drained qn.net values of 2.38, 2.90, 3.4, 3.61 respect
ively. Figure 11 shows that increasing the cone
roughness results in higher qn.net values across the
whole spectrum of penetration rates. Further studies
should be performed to examine the rate of increase
of qn.net, as well as the absolute cone resistance
Figure 9. Comparison of dimensionless backbone curve values (qc), for different φint values at different pene
between PRM and literature (Ceccato et al., 2016; Orazalin tration rates. This would provide a better understand
et al., 2018). ing of the relationship between φint with qn.net and qc.
322
Ahmadi, M. M., Byrne, P. M., & Campanella, R. G. (2005).
Cone tip resistance in sand: modeling, verification, and
applications. Canadian geotechnical journal, 42(4),
977–993.
Ceccato, F., Beuth, L., & Simonini, P. (2016). Analysis of
piezocone penetration under different drainage condi
tions with the two-phase material point method. Journal
of Geotechnical and Geoenvironmental Engineering,
142(12), 04016066.
Durgunoglo, H. T., & Mitchell, J. K. (1975). Static penetra
tion resistance of soils I-analysis. In Proc. ASCE Conf.
In-Situ Measure. Soil Properties (p.1).
Engin, H. K., Brinkgreve, R. B. J., & Van Tol, A. F. (2015).
Simplified numerical modelling of pile penetration–the
press-replace technique. International Journal for
Numerical and Analytical Methods in Geomechanics, 39
(15), 1713–1734.
Lehane, B. M., O’loughlin, C. D., Gaudin, C., &
Figure 11. Qn.net against non-dimensional velocity, V, for Randolph, M. F. (2009). Rate effects on penetrometer
φint = 0°, 8°, 16° and 23°. resistance in kaolin. Géotechnique, 59(1), 41–52.
Lim, Y. X. (2017). Numerical study of cone penetration test
5 CONCLUSION in clays using press-replace method (Doctoral disserta
tion, National University of Singapore (Singapore)).
This paper presents a simplified approach, the Press- Lim, Y. X., Tan, S. A., & Phoon, K. K. (2018). Application
of press-replace method to simulate undrained cone
Replace Method (PRM), for modeling large deform
penetration. International Journal of Geomechanics, 18
ation penetration problems using a small strain finite (7), 04018066.
element program. The PRM approach was used in Lu, Q., Randolph, M. F., Hu, Y., & Bugarski, I. C. (2004).
this paper to model cone penetration in soils under A numerical study of cone penetration in clay. Géotech
different drainage conditions. This paper compares nique, 54(4), 257–267.
the PRM predictions against the results of 4 Oliveira, J. R., Almeida, M. S., Motta, H. P., &
examples taken from the published literature. Almeida, M. C. (2011). Influence of penetration rate on
The PRM results from drained and undrained pene penetrometer resistance. Journal of Geotechnical and
trations were presented and compared against the pub Geoenvironmental Engineering, 137(7), 695–703.
Orazalin, Z. Y., & Whittle, A. J. (2018). Realistic numerical
lished results for the 4 examples (Yi et al., 2012;
simulations of cone penetration with advanced soil
Sheng et al., 2014; Ceccato et al., 2016; Orazalin et al., models. In Cone Penetration Testing 2018 (pp.
2018). In general, the PRM results (undrained and 483–489). CRC Press.
drained) are close to and within the range obtained by Paniagua, P., Nordal, S., & Engin, H. K. (2014). Back calcula
other researchers using different modeling techniques. tion of CPT tests in silt by the Press-Replace technique.
The effect of partial drainage was studied and val Paniagua, P., Nordal, S., Emdal, A., Engin, H. K., &
idated in this paper by performing consolidation ana Kim, Y. (2021, May). Back-Calculation of Stresses and
lysis with different penetration rates. The results, in Pore Pressures Around a Penetrating Cone in Silt. In
the form of dimensionless backbone curves obtained International Conference of the International Associ
ation for Computer Methods and Advances in Geome
by plotting normalized resistances against normal
chanics (pp. 260–267). Springer, Cham.
ized velocity, compare favorably with those reported Randolph, M., & Hope, S. (2004). Effect of cone velocity
by Sheng et al. (2014) and Orazalin et al. (2018). on cone resistance and excess pore pressures. Proc. 1st
Some discrepancies were observed when compared Int. Symp. on Engineering Practice and Performance of
to the MPM results of Ceccato et al. (2016), which Soft Deposits, Osaka, 147–152.
could be due to the use of a different tip geometry. Roscoe, K., & Burland, J. B. (1968). On the generalized
The effects of cone roughness were also explored stress-strain behaviour of wet clay. In Engineering Plas
in this paper. The results show an increase in qc and ticity, Cambridge University Press, 535–609
qn.net values with the interface friction angle φint. Sheng, D., Kelly, R., Pineda, J., & Bates, L. (2014).
Numerical study of rate effects in cone penetration test.
In 3rd international symposium on cone penetration
testing (pp. 419–428).
REFERENCES Sivasithamparam, N., Engin, H. K., & Castro, J. (2015).
Numerical modelling of pile jacking in a soft clay. In
Abu-Farsakh, M., Tumay, M., & Voyiadjis, G. (2003). Numer
Computer Methods and Recent Advances in Geomecha
ical parametric study of piezocone penetration test in clays.
nics: Proceedings of the 14th International Conference
International Journal of Geomechanics, 3(2), 170–181.
of International Association for Computer Methods and
Andersen, K. H., Andresen, L., Jostad, H. P., &
Recent Advances in Geomechanics, 2014 (IACMAG
Clukey, E. C. (2004, January). Effect of skirt-tip geom
2014) (pp. 985–990). Taylor & Francis Books Ltd.
etry on set-up outside suction anchors in soft clay. In
Suzuki, Y., & Lehane, B. (2014). Field observations of
International Conference on Offshore Mechanics and
CPT penetration rate effects in Burswood clay. In 3rd
Arctic Engineering (Vol.37432, pp. 1035–1044).
International Symposium on Cone Penetration Testing
323
(pp. 403–410). International Symposium on Cone Pene Yi, J. T., Goh, S. H., Lee, F. H., & Randolph, M. F. (2012).
tration Testing. A numerical study of cone penetration in fine-grained
Teh, C. I., & Houlsby, G. T. (1991). An analytical study of the soils allowing for consolidation effects. Géotechnique,
cone penetration test in clay. Geotechnique, 41(1), 17–34. 62(8), 707–719.
Walker, J., & Yu, H. S. (2006). Adaptive finite element analysis Yu, H. S., Herrmann, L. R., & Boulanger, R. W. (2000).
of cone penetration in clay. Acta Geotechnica, 1(1), 43–57. Analysis of steady cone penetration in clay. Journal of
Wang, Z. Z., & Goh, S. H. (2018). Spudcan installation and Geotechnical and Geoenvironmental Engineering, 126
post installation behaviour in soft clay: The press- (7), 594–605
replace method. In Numerical Methods in Geotechnical Yu, H. S. (2004). The First James K. Mitchell Lecture In
Engineering IX (pp. 1503–1510). CRC Press. situ soil testing: from mechanics to interpretation. Geo-
Yasufuku, N., & Hyde, A. F. L. (1995). Pile end-bearing cap mechanics and Geoengineering: An International Jour
acity in crushable sands. Géotechnique, 45(4), 663–676. nal, 1(3), 165–195.
324
© 2022 the Author(s), ISBN 978-1-032-31259-0
Cone penetration testing to constrain the calibration process of a sand

plasticity model for nonlinear deformation analysis
A. Chiaradonna
T.J. Carey
The University of British Columbia, Vancouver, Canada
K. Ziotopoulou & J.T. DeJong

University of California at Davis, Davis, USA
ABSTRACT: A reliable prediction of liquefaction-induced damage typically requires nonlinear deformation ana
lyses with an advanced constitutive soil model calibrated to the site conditions. The calibration of constitutive
models can be performed by relying primarily on a combination of commonly available properties and empirical
or semi-empirical relationships, on laboratory tests on site-specific soils, on in-situ penetration tests, or
a combination thereof. Chiaradonna et al. (2022) described a laboratory-based calibration approach of the
PM4Sand constitutive model and evaluated the prediction accuracy against the response of a centrifuge experiment
of a submerged slope. This paper addresses an alternate calibration approach in which the PM4Sand model is cali
brated using centrifuge in-situ CPT data. The model performance for the resulting calibration is evaluated against
the centrifuge experimental data and prior simulations from Chiaradonna et al. (2022). In this case, the CPT-based
calibration resulted in more accurate estimations of the dynamic response and permanent displacements.
1 INTRODUCTION Chiaradonna et al. (2022) modelled the dynamic

response of a centrifuge experiment consisting of
A reliable prediction of liquefaction-induced damage a submerged 10-degree slope of a poorly graded
usually requires performing nonlinear deformation ana clean sand at a DR of 63% (Carey et al. 2022a).
lyses by adopting advanced constitutive soil models. Shaking was imposed by applying a 1 Hz at the base
Constitutive model calibration protocols have been of the model container. CPTs were pushed using
developed to guide the selection of parameters, firstly a 10 mm cone in the experiment prior to and follow
driven by the goal to reproduce the soil element behav ing shaking, but were not considered in the calibra
ior as observed in laboratory element tests and, if tests tion performed by Chiaradonna et al. (2022).
are not available, against the broader body of data and The critical state compatible, stress ratio-based,
engineering relationships in the literature. For larger bounding surface plasticity constitutive model
scale experiments, the use of in-flight miniature Cone PM4Sand (Boulanger & Ziotopoulou 2017), imple
Penetration Tests (CPT) in centrifuge testing has pro mented in the commercial finite-difference platform
vided system level soil properties and better definition FLAC (Itasca, 2016) was adopted in the simulations.
of soil conditions before and after any applied shaking The primary input parameters of the model are the
(Kim et al. 2016; Khosravi et al. 2018; Moug et al. apparent relative density (DR), the shear modulus
2019; Darby et al. 2019; Carey et al. 2020). Darby coefficient (Go), and the contraction rate parameter
et al. (2019) used CPT soundings collected prior to (hpo). A laboratory-based calibration was defined on
multiple shaking events to define liquefaction trigger the results of undrained cyclic direct simple shear
ing correlations in a centrifuge experiment. The cyclic tests performed on reconstituted samples (Humire
resistance for the investigated sand in the experiment et al. 2022). The simulation of the centrifuge test by
was lower than that inferred from case history-based adopting the laboratory-based calibration reasonably
liquefaction triggering correlations. Data from in-flight simulated the pore pressure and acceleration time
miniaturized CPTs strongly constrain several soil histories, while the permanent horizontal displace
parameters, e.g., relative density DR, and were there ments were overpredicted by a factor of three (Chiar
fore useful to define model calibration parameters. adonna et al. 2022).
DOI: 10.1201/9781003308829-43
325
In this paper, the aforementioned simulation was Modeling located at the University of California,
revisited by calibrating PM4Sand parameters using the Davis (Figure 1). The soil was a clean poorly graded
cone tip resistance as measured in the centrifuge by sand, hereafter called 100A sand (Sturm 2019). The
Carey et al. 2022a. The DR is the target value of 63%, physical properties of the 100A sand are emin = 0.579,
which was verified through a pre-shaking cone pene emax = 0.881, D50 = 0.18mm, Cu = 1.68, and Gs =
tration test (Carey et al. 2022a). The soil behavior at 2.62. The embankment was dry-pluviated to a target
small strains (i.e., Go) was estimated by the measured DR = 63%, overlying a dense sand layer (DR > 90%)
cone tip resistance through the application of several of the same soil. The model was saturated with
literature relationships, whose efficacy was verified methylcellulose pore fluid that had a viscosity that was
against the shear moduli based on the measured shear 40x that of water (40 CSt).
wave velocity. The cyclic resistance ratio (CRR), Instrumentation within the model included pore
which is primarily controlled by the contraction rate pressure transducers and accelerometers, which
parameter (hpo) in PM4Sand, was estimated through enabled monitoring the coupled excess porewater pres
the normalized cone tip resistance measured in the sures and acceleration responses. Piezoceramic bender
experiment and the CPT-based empirical triggering element pairs were placed at two depths in the model
liquefaction chart by Boulanger & Idriss (2014). to measure the shear wave velocity. Measurements
The comparison between simulated and experimen were performed at 1 g and at 40 g, before and after
tal soil response is made to verify that in-flight CPT shaking, for a total of 6 measurements. The processing
testing in the centrifuge experiments properly meas of bender element time histories is reported by Carey
ures the cyclic strength of soils. In addition, the com et al. (2021). Horizontal displacement time-histories
parison between the CPT-based calibration and the were measured using highspeed videos of the deform
calibration based on cyclic laboratory tests performed ing embankment’s cross-section recorded through the
by Chiaradonna et al. (2022) is also discussed. transparent side walls of the model container and
GEOPIV image analysis software (Carey et al. 2022a).
The ground motion sequence included four shak
2 EXPERIMENTAL PROGRAM ing motions, all of which included a linear ramp to
the maximum acceleration, a hold at the maximum
2.1 Overview of centrifuge test acceleration for a certain number of cycles, and
a non-linear decay. All motions had a prototype fre
A 14 m-high submerged embankment with a 10 quency of 1 Hz but varied in their number of cycles
degree slope constructed with a uniform profile of and amplitude of the hold cycles. Further details
sand was tested in a rigid container at 40g using the about the motion are given by Carey et al. (2022a, b).
9-m radius centrifuge at the Center for Geotechnical Herein the system response to the motion shown in
Figure 1. (a) Cross-section of the centrifuge model with accelerometers and porewater pressure sensors (length in prototype
scale); (b) FLAC numerical grid used in the simulations overlaid on a photo of the centrifuge model test cross-section; and
(c) recorded input motion from the centrifuge experiment used in this study.
326
Figure 1c with a maximum acceleration of 0.14g is 3 ESTIMATION OF MODULUS AND CRR
analyzed. For brevity, only the shallowest sensors of
the mid-slope array were analyzed (Figure 1a,b). 3.1 Estimation of small-strain shear modulus
The shear wave velocity measurements made prior
2.2 Experiment characterization using a CPT to and following the shaking event were used to cal
A 10 mm-diameter cone penetrometer was culate the small strain shear modulus, with the
pushed before and after the completion of the values shown in Figure 2b. The small-strain shear
ground motion sequence. Cones were pushed into modulus at the depth of the atmospheric pressure,
the soil 457 mm at the model scale at Gmax,1, was calculated according to Gmax,1= ρVS2,
a penetration rate of 1 cm/s using a hydraulic where ρ is the soil saturated density of 1,958 kg/m3
actuator (Carey et al. 2022a). Figure 2a presents at the DR = 63%.
the CPT profile prior to shaking, measured in the Several literature relationships expressing the nor
upper bench of the slope. malized shear wave velocity, VS1, as a function of
the cone tip resistance measured in sands were
selected and applied using qc1N and are given in
Table 1. The laws are expressed by an exponential
function:
where m and n are the coefficients listed in Table 1.

The vertical profile of Gmax as a function of the
mean effective stress, p', is expressed as:
where Pa is the atmospheric pressure (101.3 kPa)

and Go is the shear modulus coefficient calculated by
imposing Gmax= Gmax,1 in Eq. (2) for an effective
vertical stress, σ’v , equal to the atmospheric pressure
(Table 1). The mean effective stress, p' is related to
the depth, z, as follows:
Figure 2. (a) CPT sounding measured prior to the start of

the shaking sequence; and (b) comparison of Gmax values
based on bender elements in the model test with predicted
Gmax profiles from literature. where K0 is the coefficient of earth pressure at rest,
assumed equal to 0.5, and γ’ is the unit weight of the
submerged soil. Eq. (2) can be expressed as
a function of the depth. As evident in Figure 2b, the
Shown in Figure 2a is the pre-shake CPT pene
bender element-based Gmax values were generally
trated to depths that exceeded an overburden stress
consistent with the predicted Gmax profiles.
of 1 atm. Hence, the depth of 10.45 m corresponds
to an overburden effective stress of 1 atm and the
cone tip resistance, referred to as qc1, was 12.8 MPa. Table 1. Considered VS1 – qc1N relationships.
The qc1 measurement was sufficiently deep to avoid
shallow penetration effects in the model (Kim et al. m n VS1 Go
2016; Sawyer 2020) and was used to calculate the
normalized corrected cone tip resistance, qc1N, as Relationship m/s m/s
defined by Boulanger & Idriss (2014), which was
rounded to 126.0 for the considered case. Since the Baldi et al. 1989 110 0.13 206 964
considered soil has a fines content equal to zero, the Robertson et al. 1992 60.3 0.23 183 762
‘equivalent clean sand’ normalized and corrected Hezagy & Mayne 1995 72.8 0.192 184 769
cone tip resistance, qc1Ncs, is also 126.0. Andrus et al. 2004 62.6 0.231 191 830
327
3.2 Estimation of the cyclic resistance ratio estimated by the in-situ CPT is 1.4 times higher than
that measured by direct simple shear tests.
The estimation of the CRR for the 100A sand was
obtained through the normalized cone tip resistance
qc1Ncs and the CPT-based triggering liquefaction
relationship developed by Boulanger & Idriss
(2014), as shown in Figure 3. For the qc1Ncs of
126.0, the CRR was equal to 0.186.
Figure 4. Cyclic resistance curves used in the simulations

(CPT calibration) vs calibration based on direct simple
shear data (Lab calibration) and experimental data.
5 NUMERICAL SIMULATION RESULTS

Figure 3. Estimation of CRR from the CPT-based liquefac
tion triggering curve by Boulanger & Idriss (2014). The centrifuge test was numerically simulated with
the finite difference program FLAC (Itasca, 2016).
The geometry of the analysis domain was based on
4 PM4Sand CALIBRATION the centrifuge prototype dimensions (Figure 1a). The
discretized domain is shown in Figure 1b, with fur
The PM4Sand model calibration based on the ther details available in Chiaradonna et al. (2022).
experimental CPT data is referred to as “CPT cali The predicted and observed time histories of the
bration”. The calibration focused on defining the horizontal displacement at the surface of the mid-
three primary input parameters, DR, Go, and hpo. The slope (Figure 1b) are shown in Figure 5. The experi
secondary parameters were left to their default mental trend exhibits a progressive accumulation of
values (Boulanger & Ziotopoulou 2017). Specific displacements, with a final permanent value of 7 cm.
ally, DR was set to 63%, controlled by the centrifuge
test design, and Go was set to 863, the average of the
upper (964) and lower (762) bounds of the Go rela
tionships in Table 1. The hpo parameter of 0.21 was
iteratively calibrated via single element undrained
cyclic stress-controlled direct simple shear (DSS)
simulations until a satisfactory match between the
Cyclic Stress Ratio (CSR) and CRR of 0.186 was
reached for a triggering criterion of 3% shear strain
in 15 cycles.
Figure 4 Illustrates CSR versus number of cycles
to liquefaction for a 3% single amplitude shear strain
triggering criterion for the “CPT calibration”. This
curve is generated from a series of single element
DSS simulations using the calibrated PM4Sand pri Figure 5. Comparison of horizontal displacement measured
at the surface of the embankment slope and numerical
mary variables, subjected to a range of CSRs.
simulations from the CPT and Lab calibrations.
Experimental points as measured through cyclic
DSS tests (green points) for two different overbur
den effective vertical stresses are also plotted for ref The magnitude of the displacement oscillations per
erence. The data for a σ'vo = 50 kPa were used as the cycle is relatively minor, resulting in a clear ratcheting
dataset for the laboratory-based calibration (“Lab of downslope displacement. The CPT calibration pre
calibration”) in Chiaradonna et al. (2022). dicted displacements that were practically identical to
The cyclic strength at 15 cycles is 0.186 for the the experiment with a permanent horizontal displace
“CPT calibration” and 0.133 for “Lab calibration”, ment was 5 cm. The oscillations in displacements for
implying that the soil resistance to liquefaction as the CPT calibration are larger compared to the
328
experiment and predict a dynamic upslope movement. history the predicted accelerations in the simulation are
For context, the lab calibration-based simulation pre lower. These trends are also evident in the clear agree
dicted significantly higher displacements, with the ment in the spectral acceleration plots for AH9, with
accumulation primarily occurring during the last two show consistency across all periods. For AH10 the
cycles of shaking at full acceleration amplitude and higher spectral acceleration at the predominant period
during the decay. of 1 Hz is evident. For periods less than 0.7 s, spectral
The agreement between the experimentally meas accelerations are better captured better by the “Lab
ured and CPT Calibration simulation of the mid- calibration” due to the dilation spikes in the time his
slope displacement also extended to the global tories of acceleration; however, these high frequencies
deformation patterns across the centrifuge experi have a minor contribution to the overall movement of
ment. This is evident in Figure 6 where contours of the embankment.
horizontal displacement of the experiment are pre The agreement in deformation, acceleration, accel
sented. The magnitude and spatial distribution of the eration response spectra, and excess porewater pres
displacement field is nearly identical. sure trends stand in clear contrast to the trends for the
While the agreement of the displacement fields is “Lab Calibration” simulation. As evident in Figures 6
central for performance-based design, examining the and 7 the displacements that accumulate near the end
pore pressure and acceleration time histories as well and after shaking are significantly larger at the mid-
as the response spectra at the mid-slope during shak slope surface and throughout much of the model. The
ing, is also insightful. Figure 7 presents the pore pres primary reason for these differences is attributed both
sure and acceleration time histories of the upper two (1) to the lower cyclic resistance ratio (CRR) of the
locations on the mid-slope. The pore pressure gener laboratory data which contributes to an early onset of
ation time histories for P7 and P8 show different pat liquefaction and its associated deformations, and (2)
terns of accumulation, as well as different residual the continuing deformation after the end of shaking
values at the end of shaking. Excess pore pressure due a high excess pore pressure.
ratio, ru, peaks in P7 are better captured by the “Lab
calibration” compared to the “CPT calibration”. The
experiment reached an ru of 1 after several cycles, 6 DISCUSSION
while the simulation did not. However, at the end of
shaking both the experimental and the CPT calibration The numerical simulation presented in this paper
simulation have a ru value less than about 0.7, imply demonstrates the utility and value of CPT measure
ing that resedimentation and re-establishment of the ments in centrifuge experiments. Simulations based
effective stress profile has been partially taking place. on a CPT-based calibration of the PM4Sand constitu
tive model provided a very satisfactory match to the
observed system level responses. Past parametric
investigations by Chiaradonna et al. (2022) had
shown that a CRR higher than the one obtained from
DSS tests would likely justify the observed
responses.
However, that study was inconclusive as to
whether the higher in-situ CRR was an artifact of
arching or sloping ground conditions or that in gen
eral the DSS data in this case had misrepresented the
in-situ centrifuge conditions. The present study dem
onstrated that the CPT measurements provided
a significantly improved characterization of the
cyclic resistance of the centrifuge model and thus
a more successful validation of the response, particu
larly with respect to displacements.
Future work will investigate the (i) influence of
other contributing factors such as 3D effects and
Figure 6. Contour fields of displacement from the shake arching, (ii) effect of sloping ground conditions on
measured in the centrifuge experiment, and the numerically cyclic strength, and (iii) reasons between the dis-cre
predicted contours using the CPT and Lab PM4Sand pancy of the DSS-based CRR and the CPT-
calibrations. based CRR.
The acceleration time histories between the meas 7 CONCLUDING REMARKS

ured experimental response and the CPT calibration
simulation are nearly identical for the lower AH9 This paper addressed the calibration of a critical
accelerometer, time history, and in the first part of the state compatible, stress ratio-based, bounding sur
AH10 time history. However, later in the AH10 time face plasticity constitutive model aiming at
329
Figure 7. Comparison of the measured experimental and numerically simulated responses for ‘CPT Calibration’ vs. ‘Lab
Calibration’. Results are for the select sensors (Figure 1) in terms of porewater pressure ratio, acceleration time histories,
and response spectra (5% damping).
numerically simulating the dynamic behavior of REFERENCES

a poorly-graded sand slope. While deficiencies in
the horizontal displacement prediction were high Andrus, R.D., Piratheepan, P., Ellis, B.S., Zhang, J. &
lighted by a calibration obtained from cyclic Juang, C.H. 2004. Comparing liquefaction evaluation
laboratory tests (Chiaradonna et al. 2022), the use methods us-ing penetration-VS relationships. Soil Dyn.
of CPT data as a calibration basis produced Earthq. Eng. 24: 713–721.
Baldi, G., Bellotti, R., Ghionna, V., Jamiolkowski, M. &
a better prediction of both displacements and
Lo Presti, D.C.F. 1989. Modulus of sands from CPTs
accelerations. The reliability of the simulation can and DMTs. Proc. 12th Int. Conf. on Soil Mech. and
be ascribed mainly to two principal reasons: (i) Found. Eng., Rio de Janeiro, 13–18 August 1989.
the cyclic strength of the soils is directly esti Vol. 1: 165–170. Rotterdam: Balkema.
mated by combining the measured cone tip resist Boulanger, R.W. & Idriss, I.M. 2014. CPT and SPT lique
ance of the soil tested in centrifuge with the CPT- faction triggering procedures. Report No UCD/GCM
based triggering liquefaction chart, and (ii) the 14/01, University of California at Davis, California,
small-strain shear modulus is well reproduced by USA.
the applied literature relationships for clean sands. Boulanger, R.W. & Ziotopoulou, K. 2017. PM4Sand (Ver
sion 3.1): A sand plasticity model for earthquake engin
This study strengthens the importance of in-flight
eering applications. Technical Report No. UCD/CGM
CPT measurements in centrifuge tests and repre 17/01, Center for Geotechnical Modeling. University of
sents a step forward in the calibration of California, Davis.
advanced constitutive models using measurements Carey, T.J., Chiaradonna, A., Love, N., DeJong, J.T.,
from in-situ CPT tests. Further applications to Ziotopoulou, K., Martinez, A. 2021. Effect of soil grad
available centrifuge tests with different relative ation on the response of a submerged slope when sub
densities and gradation of soils will be performed jected to shaking –centrifuge data report for TJC01.
to generalize the obtained results. Report No. UCD/CGM-21/03, University of California
at Davis, California, USA.
Carey, T.J., Chiaradonna, A., Love, N., Wilson, D.W., Ziot
opoulou, K., Martinez, A. & DeJong, J.T. 2022a. Effect
ACKNOWLEDGEMENTS of soil gradation on embankment response during lique
faction: a centrifuge testing program. Soil Dyn. Earthq.
The National Science Foundation (NSF) provided Eng. (Under review).
funding for this work (Grant No. CMMI-1916152) Carey, T.J., DeJong, J.T., Ziotopoulou, K., Martinez, A. &
and for the Natural Hazards Engineering Research Chiaradonna, A. 2022b. The effects of gradation on the
Infrastructure (NHERI) centrifuge facility at UC dynamic response of sloping ground. Proc. 20th Intern.
Davis (Grant No. CMMI-1520581). Any opinions, Conf. on Soil Mech. & Geotech. Eng., Sydney, 1–5 May
findings, and conclusions, or recommendations 2022 (Accepted).
expressed in this material are those of the author(s) Carey, T.J., Gavras, A. & Kutter, B.L. 2020. Comparison of
and do not necessarily reflect those of the NSF. The LEAP-UCD-2017 CPT Results. In Model Tests and
Numerical Simulations of Liquefaction and Lateral
lead author was supported by the U.S. Fulbright
Spreading: 117–129. Cham: Springer.
Scholar Program.
330
Chiaradonna, A., Ziotopoulou, K., Carey, T.J., DeJong, J.T. Khosravi, M., Boulanger, R.W., DeJong, J.T., Khosravi, A.,
& Boulanger, R.W. 2022. Dynamic Behavior of Uni Hajialilue-Bonab, M. & Wilson, D.W. 2018. Centrifuge
form Clean Sands: Evaluation of Predictive Capabilities modeling of cone penetration testing in layered soil. In
in the Element- and the System-Level Scale. GeoCon Geotechnical Special Publication 290, S. J. Brandenberg
gress 2022, Charlotte, 20–23 March 2022 (Accepted and M. T. Manzari (eds.); Proc. Geotechnical Earth
paper). quake Engineering and Soil Dynamics V, Austin, 10–13
Darby, K.M., Boulanger, R.W., DeJong, J.T. & Bronner, J.D. June 2018.
2019. Progressive Changes in Liquefaction and Cone Kim, J.H., Choo, Y.W., Kim, D.J. & Kim, D.S. 2016.
Pene-tration Resistance across Multiple Shaking Events in Miniature cone tip resistance on sand in a
Cen-trifuge Tests. J. Geotech. Geoenviron. Eng., ASCE centrifuge. J. Geotech. Geoenviron. Eng., ASCE 142
145(3). (3).
Hegazy, Y.A. & Mayne, P.W. 1995. Statistical correlations Moug, D.M., Price A.B., Bastidas, A.M.P., Darby, K.M.,
between VS and cone penetration data for different soil Bou-langer, R.W. & DeJong, J.T. 2019. Mechanistic
types. Proc. 1st International Symposium on Cone Pene Develop-ment of CPT-Based Cyclic Strength Correl
tration Testing, CPT ’95, Linkoping, 4-5 October 1995, ations for Clean Sand. J. Geotech. Geoenviron. Eng.,
2: 173–178. ASCE 145(10).
Humire, F., Ziotopoulou, K. & DeJong, J.T. 2022. Evaluating Robertson, P.K., Woeller, D.J. & Finn, W.D.L. 1992.
shear strain accumulation of sands exhibiting cyclic mobil Seismic CPT for evaluating liquefaction potential.
ity behavior. Proc. 20th Intern. Conf. on Soil Mech. and Canadian Ge-otechnical Journal 29: 686–695.
Geotech. Eng., Sydney, 1–5 May 2022 (Accepted paper). Sawyer, B.D. 2020. Cone penetration testing of coarse-
Itasca 2020. FLAC – Fast Lagrangian Analysis of Continua, grained soils in the centrifuge to examine the effects of
Version 8.1. Minneapolis, MN: Itasca Consulting Group. soil gradation and centrifuge scaling. Master’s Thesis.
Kamai, R., & Boulanger, R.W. (2011). Characterizing local UC Davis.
ization processes during liquefaction using inverse analyses Sturm, A.P. 2019. On the Liquefaction Potential of Grav
of instrumentation arrays. Meso-Scale Shear Physics in elly Soils: Characterization, Triggering and Perform
Earthquake and Landslide Mechanics, 219–238. ance. PhD Dissertation. UC Davis.
331
© 2022 the Author(s), ISBN 978-1-032-31259-0
Probabilistic delineation of soil layers using Soil Behavior Type Index
S. Collico & M. Arroyo

Universidad Politécnica de Cataluña, Barcelona, Cataluña, Spain
M. DeVincenzi, A. Rodriguez & A. Deu

Igeotest s.l. Figueres, Cataluña, Spain
ABSTRACT: CPTu-based soil profiling has become a key component in the geotechnical design process.
However, this is an interpretative process, affected by the inherent variability of soil properties, measurement
noise and subjective heuristics. These are difficult to communicate to other interpreters or, even for the same
interpreter, to transfer across profiles. A semi-automated tool for CPTu data interpretation is presented as an aid
in this interpretation process. A probabilistic-based algorithm is employed to elicit the implicit heuristics in
CPTu-based soil profiling and facilitate transference. Univariate normal distributions fit Soil Behavior Type
Index data. Soil class boundaries, taken from a conventionally accepted chart, are sequentially activated with
user-specified refinement. Thin layers under cone resolution are merged using well-established criteria. An appli
cation to CPTu records on finely interlayered deltaic deposits is illustrated, in which output delineations resulting
from different analyst choices are compared among themselves and with one based on core description.
1 INTRODUCTION vertical and horizontal inherent variability and

measurements errors, that might differ during
Differentiation of ground units by means of strati each test, intervene. Therefore, such lack of
graphic profiling is a key step of site characterization transparency and not methodical process of
for geotechnical design. Among all the in-situ tests, layers’ identification, should be addressed.
Cone Penetration Tests CPTu are widely employed Different probabilistic methodologies have
for layer’s identification based on the soil response to addressed the problem of stratigraphic profiling
the probe. An aid for CPTu postprocessing results is based on CPTu. Some works (Jung et al., 2008;
given by Soil Behavior Type (SBT) charts (Douglas Wang et al., 2013) put the focus on the basic classifi
& Olsen, 1981; Robertson, 1990, 2009) that allow cation scheme applied. They point that generally
every reading to be classified. Analysists can then established soil class boundaries are inherently
establish boundary layers by inspecting the profile of inaccurate to derive site-specific soil profiling and
soil class assignment derived from these charts. address this issue by updating the classification
This is an interpretation task, and as such is scheme according to site-specific observations.
strongly affected by analyst knowledge, experience Another line of thinking is the one that enable the
and heuristics. The interpretation process followed is integration geological unit’s information to CPTu
sometimes difficult to explain and might appear arbi data (Depina et al., 2016; Krogstad et al., 2018).
trary and be a subject of disputes if other parties are These works employ SBT charts only as a reference
involved. Even if that is not the case, the same ana template, and CPTu soil class assignment refers to
lyst may find it difficult to ensure coherence of inter geological units identified from retrieved core sam
pretation across different soundings of a site. ples and/or previous studies.
As an example, when a thin silt layer (e.g., This work follows a different line of thinking and
20-30 cm) is embedded in two thick sand units, it is based on a more modest approach, manly
the analyst, based on his experience, may or may oriented for practical use. SBT charts- here Robert
not consider relevant the presence of such layer. son, (2009) - are considered as conventionally
An additional and more sensitive issue arises accepted template for soil classification with given
when dealing with multiple CPTu sounding (not uncertain) class boundaries. The probabilistic
records and the so called cross-site variability approach is applied only to assess inherent variabil
(Zhang et al., 2004). Layer’s identification, by ity and measurements error of CPTu readings. The
accounting for information of different locations, probabilistic parameters introduced are then oriented
is an even more complex process since both to quantify the degree of internal coherence of each
DOI: 10.1201/9781003308829-44
332
identified layer, and, at the same time, eliciting ana
lyst heuristics and interpretation criteria.
The tool fits a univariate normal distribution
to interpret data originated from one or various
CPTu probes, which are gradually plotted on
Robertson (2009) chart, identifying soil units in
a probabilistic manner. Class boundaries are
sequentially activated with user-specified refine
ment. Thin layers (layers below a specified min
imum thickness resolution) are assessed through Figure 1. Soil Behavior Type chart based on Soil behavior
a practical merging scheme. This work applies Type Index Ic .
the methodology previously proposed by Collico
et al., (2020) to Robertson, (2009) SBT chart. In
Collico et al., (2020), a bivariate normal distri 2.2 Statistical model for CPTu data
bution approach was applied to fit CPTu obser
vations on Robertson, (1990) SBT chart. The When plotting CPTu observations classified by SBT,
flexibility of such work is here exploited to fit inherent variability and measurements errors, gener
Soil Behavior Type Index observations according ate some data scatter (Figure 2). In this study
to Robertson, (2009) class boundaries. a univariate normal distribution is fitted to SBT data
As illustrated section 2, the proposed methodology of a given stretch of CPTu:
consists of three main parts: a probabilistic model to
CPTu data; a staged classification procedure and an
option for automated thin layer consolidation.
2 METHODLOGY with μIC ; σIC mean and standard deviation of IC data

sample. An example of 3D representation of fitted
2.1 Soil behavior type index Ic univariate normal distribution for a 5m IC profile
(Figure 2b) is reported in Figure2b. The area of
Among all SBT charts, one of the most popular is
color, under the density function, indicate the extent
the Robertson (2009) chart (Figure 1). Based on the
of an interval pre-established representative data
work of Douglas & Olsen, (1981) and Jefferies &
interval. As explained below, the interplay between
Davies (1991), Robertson (2009) highlighted how
this interval and SBT class boundaries is exploited
Robertson (1990) class boundaries could approxi
to obtain a systematic procedure for layer
mate by a concentric circle whose radius can be
identification.
defined by a unified parameters denoted as soil
behavior class index Ic which combine both normal
ized CPTu parameters as:
with:
Figure 2. a) Synthetic IC record. b) Corresponding univari

ate distribution and pre-established representative interval
plotted on Robertson (2009) chart.
where n � 1 is a soil-type dependent exponent
which normalizes for the effect of stress level on tip
2.3 Class assignment criteria
resistance.
Robertson (2009, 2016) highlighted that despite Once the CPTu SBT values corresponding to
being a single-valued index, by including the stress a particular stretch are fitted to a univariate normal
exponent n, IC based soil delineation might provide distribution, the fitted data will be assigned to the
more useful information on soil behavior than one soil class in which the distribution mean is found.
based on Robertson (1990) classes. Two user-specified acceptance thresholds (i.e.,
333
model parameters) are used to divide the CPTu
record into different stretches:
• Noise-threshold parameter,
• Class-mixture tolerance parameter, m
The meaning of these two parameters is rather
intuitive. The P parameter (Figure 2b), expressed in
term of standard deviations, establishes the width of
the representative data sample interval (P ¼ μIC ± σIC ).
It is introduced to account for the possibility of
extreme values of IC within the dataset (e.g., shelly
inclusions in clays). As an example, by considering
a value of P = 2, the representative interval would
contain 95% of the underlying data and 5% will be
considered as noise.
The class-mixture tolerance parameter m is intro
duced so that the user can modulate the relevance of
the underlying classification system in the soil profil
ing exercise. The value of m is defined as the area Figure 3. Class boundary at first and second staged classifi
cation on Robertson (2009) chart.
proportion of specified representative interval that is
allowed to cross soil class boundaries. In other word,
the m value explicit how strictly adhered to are pre 2.5 Thin layer treatment
established class boundaries when the CPTu is sub Any semi-automated analysis of CPTu records is
divided into layers. Parameter m expresses the degree liable to end identifying thin layers (e.g., thinner than
of belief on the underlying classification adopted. 20 cm). It is well recognized that the cone diameter
employed limits the ability of CPTu to resolve thin
layers, as the measured cone tip resistance is affected
2.4 Classification levels
by both development and sensing distances (Boulan
The original Robertson (2009) chart introduces ger & Dejong, 2018). Due to those two limits, the
six different soil classes. In some circumstances measured soil resistance does not coincide with that
simpler classifications might appear more suit of the material at the tip location, i.e., measured soil
able to the analyst. This possibility has been resistance in absence of both development and sens
enabled in this work introducing a staged classi ing distances. Therefore, the cone is only able to
fication procedure, in which SBT-based classifi unambiguously identify layers above a minimum
cations of progressive refinement are introduced layer resolution of 150- 200 mm. A simplified prac
sequentially. tical approach to deal with this difficulty is to merge
The coarser level uses a dual classification, distin layers below the minimum thickness with adjacent
guishing only between: ones of similar soil behavior type (Ganju et al. 2017).
This strategy is also adopted here, with details
• Clay-like behavior (C-L)
slightly dependent on the classification level selected.
• Sand-like behavior (S-L)
At the first and second refinement levels thin layers
These two soil classes, representative of are merged based on the closeness on mean value of
undrained and drained response for C-L and IC . At the third level, when using the full (Robertson,
S-L respectively, are identified by the IC bound 2009) classification chart, the situation is analogous
ary value 2.6 (Figure 3), (Robertson 2009). That to that of Ganju et al. (2017) and the same auxiliary
boundary is here onwards designed as a primary criteria based on class groupings are applied.
boundary. A second, more refined, level of clas
sification introduces a soil-mixture class to 2.6 Single CPTu analysis workflow
account for partially drained soil response to the
cone probe, along with a sand-like and clay-like The main workflow steps are reported in Figure 4.
classes. The boundaries separating these three The analysist has to feed in a CPTu record to analyze
classes (Figure 3) are designed as secondary and select the level of classification, minimum layer
boundaries. Finally, the most refined classifica resolution and the P and m values. The code can then
tion level introduces all six soil classes of the start analyzing the CPTu input record by selecting
original Robertson, (2009) proposal. The limits a segment of CPTu data pairs long enough (e.g., 10
between them and here designated as tertiary points) to fit an initial PDF (IC ). The PDFðIc ) is then
boundaries. updated by adding the next CPTu data point, moving
The analyst might choose directly a particular clas downwards through the record. Such updating allows
sification level or might run sequentially through all the IC density to move on the SBT line until the rep
levels. resentative interval (given by P) surpasses any class
334
boundaries by a larger proportion than that allowed CPTu locations is less than typical value of horizon
by m. At that time, data that fed the univariate dens tal scale of fluctuation (e.g., 40 - 80 m, Phoon &
ity are assigned to the identified layer, whose soil Kulhawy 1999) lateral continuity could be assumed.
class is assigned depending on location of the mean
value of the univariate distribution. The procedure
resumes by analyzing the next segment in the CPTu
record until all the record is analyzed (Figure 4). An
example of layer identification is depicted in Figure 5
for a synthetic Ic profile and P ¼ 1:5 and m ¼ 0
using a first level (binary) classification.
Figure 5. General workflow of single CPTu analysis.
Figure 4. General workflow of single CPTu analysis.
2.7 Multiple CPTu analysis

The same methodology can be applied to multiple
CPTu records. The aim is to identify common layers
at different locations according to a pre-established
depth correspondence criterion between the records. Figure 6. Common layer identification by employing two
This may just be a hypothesis of layers parallel to CPTu sounding record.
the ground level, or following a linear trend through
out the site or any other criteria.
Once selected the depth criterion, the univariate
density is now initialized by nCPTðuÞ length (initial data
segment) data points, with nCPTðuÞ number of CPTu
records analyzed (Figure 6a). The univariate density is
then updated by nCPTðuÞ data (one for each sounding
record analyzed) until the boundary condition is
attained according to P and m threshold (Figure 6b).
After stoppage, to ensure that identified layers belong
to the same soil type, data assigned to the layer for
each CPTu sounding is analyzed separately, and nCPTu
univariate density distributions are computed.
A common soil layer is identified only if the nCPTu
mean values are within the same class boundaries,
(Figure 6b). The procedure then resumes until all the
CPTu records are analyzed.
Qualitative considerations on layer lateral con (a) (b) (c)
tinuity may be then inferred depending on thickness
of identified common layers and relative distance of Figure 7. Layer delineation a) from core sample inspec
CPTu locations. As an example, if a thick common tion. From CPTu using all SBT classes with b)
soil layers is identified and the relative distance of P = 1.8; m = 0.05. c) P = 1; m = 0.2.
335
3 ILLUSTRATIVE EXAMPLE parameters as in (Figure 7b). Most boundaries
established by core inspection are recovered by
3.1 Case study the binary classifier, whereas that is not the case
for the second level profile. It is interesting to
The proposed methodology is tested on CPTu observe how the coarser binary classification
records obtained at Barcelona harbor (Spain). results in very fast layer alternance in the first
Two adjacent CPTu sounding profiles (CPTu74, 14 m, Figure 8a whereas the introduction of the
CPTu73), each up to 21m depth from seabed “mixtures” class results in a rather homogenous
surface (Figure 7) are analyzed. An independ profile (Figure 8b).
ently established soil layer profile was available This result is also clear when another CPTu is
for the site, which was derived from cores analyzed using the same control parameters
retrieved at the same location where CPTu74 (Figure 9). Comparing the different profiles
was performed. Laboratory samples were obtained, the analyst can decide which set of
retrieved each 5 m depth and core description choices (classification level, P and m values)
followed UNE-EN ISO 14688-1 from which results in the level of detail that is more mean
a detailed soil profile was derived. The boundar ingful for engineering design process. Once the
ies identified (Figure 7a) carried significant analyst heuristics are thus elicited, they can be
uncertainty due to the complex structure and systematically applied at all site locations.
strong heterogeneity of the site. The layering of
Figure 7a was directly checked against cone
results. For each layer, CPTu74 data were 3.3 Multiple CPTu records
assigned to one of the six classes of Robertson As an example of simultaneous analysis, the two
(2009). Percentages of IC data are reported in CPTu records discussed above were considered.
Table 1. Up to 15.5 it is evident how core-based Based on previous study on this area a depth
layers are mainly composed by soil-mixture correspondence criterion parallel to the seabed
class (class 3 and 4 of Robertson, 2009). How surface was selected. By assuming the same P
ever, a significant heterogeneity is evident and m value, 2D cross sections for three differ
within each layer. ent level of classifications are reported in
Figure 10. For identified common layers, lateral
3.2 Case study-single CPTu record continuity was assumed due to the relatively
close distance between CPTu (e.g., about 50 m).
The procedure previously described was applied At first level 15% of data was classified as S-L,
to the CPTu74 record. The P and m model 59% as C-L, and 26% corresponded to “not
parameters were varied to try to match the core- common” layers. Increasing the classification
based delineation. A third staged delineation was refinement level at every profile the continuity
initially selected due to the detailed core-based between different site investigation points is
description. By considering P = 1.8 (e.g., about reduced. Using second and third level classifica
92% representative interval) and 5% threshold tions, the percent of data in “not common”
mixture, all core-based boundaries (except one layers increase to 40% and to 58%, respectively.
at 17.4 m depth) were identified (Figure 7).
However, several additional layers were detected
as expected from the results in (Table 1). To
illustrate the effect of varying the fitting control
parameters, the same record is reanalyzed using Table 1. Percent of soil class data within each geological
unit.
the m = 0.2 value and P = 1 (i.e., representative
interval of about 68%) (Figure 7c). The param Depth SBT1 SBT2 SBT3 SBT4 SBT5 SBT6
eters of the reanalysis result in a simpler profile,
but it is noticeable that the layers disappearing [m] % % % % % %
are not simply the thinner ones (e.g., at 1-5.1 - 2 34 43 21
6 m depth) but others (e.g., at 19 m depth) that 5.1-7.05 - 1.5 68.9 13.8 15.8
were less statistically contrasted with their 7.05-7.5 - - 56.6 43.4 -
neighbors. 7.5-12.05 - 15 63 15 6
An alternative route to profile simplification is 12.05-15.5 - 30 31 29 10
to use a coarser classification level. This is 15.5-17.4 - 68 22 3 7
explored in Figure 8a, b using the same CPTu74 17.4-21 - 83.3 13.6 2.7 1.4
record and maintaining the same control
336
(a) (b)
Figure 10. a) Soil delineation at second staged classifica
Figure 8. CPTu profile layering resulting from tion for P=1.86; m=0.05 b) Soil delineation at third staged
P=1.8; m=0.05 a) using a first level classification b) using classification for P=1.8; m=0.05.
a second level classification.
delineation process. The proposed methodology is

application to multiple CPTu records to help discern
continuous layering. Although the method has been
illustrated using SBT it is easily adaptable to any
other chart-based classification procedure (Collico
et al. 2020).
REFERENCES
Boulanger, R. W., & Dejong, J. T. (2018). Inverse filtering
procedure to correct cone penetration data for thin-layer
and transition effects. Cone Penetration Testing 2018
Proceedings of the 4th International Symposium on
Cone Penetration Testing, CPT 2018, 25–44.
(a) (b) (c) Collico, S., Devincenzi, M., & Rodriguez, A. (2020). Semi
automated probabilistic soil profiling using CPTu. 6th
INTERNATIONAL CONFERENCE ON GEOTECH
Figure 9. Soil delineation for P=1.8; m=0.05 for CPTu73;
NICAL AND GEOPHYSICAL SITE
a) first level classification. b) second level classification. c)
CHARCTERISATION.
third level classification..
Depina, I., Le, T. M. H., Eiksund, G., & Strøm, P. (2016).
Cone penetration data classification with Bayesian Mix
ture Analysis. Georisk, 10(1), 27–41. http://dx.doi.org/
4 CONCLUSION 10.1080/17499518.2015.1072637
Douglas, B. J., & Olsen, R. S. (1981). SOIL CLASSIFICA
This work describes a novel probabilistic procedure TION USING ELECTRIC CONE PENETROMETER.
for soil delineation based on CPTu sounding records October 1981, 209–227.
aiming to facilitate communication in the process of Jefferies, M., & Davies, M. (1991). Soil Classification by
the cone penetration test: Discussion. Can. Geotech. J.,
layer delineation. Subjective heuristics, applied 28, 173–176.
within CPTu records interpretation, are elicited and Jung, B. C., Gardoni, P., & Biscontin, G. (2008). Probabil
quantified facilitating consensus in the soil istic soil identification based on cone penetration tests.
337
Geotechnique, 58(7), 591–603. http://dx.doi.org/ Robertson, P. K. (2016). Cone penetration test (CPT)
10.1680/geot.2008.58.7.591 based soil behaviour type (SBT) classification system —
Krogstad, A., Depina, I., & Omre, H. (2018). Cone penetra an update. 1927(July), 1910–1927.
tion data classification by Bayesian inversion with Robertson, P. K., & Wride, C. E. F. (1998). Evaluating
a Hidden Markov model. Journal of Physics: Conference cyclic liquefaction potential using the cone penetration
Series, 1104(1). http://dx.doi.org/10.1088/1742-6596/1104/ test.
1/012015 Uzielli, M., Vannucchi, G., & Phoon, K. K. (2015).
Phoon, K. K., & Kulhawy, F. H. (1999). Characterization of Random field characterisation of stress-normalised cone
geotechnical variability. Canadian Geotechnical Journal, penetration testing parameters. Risk and Variability in
36(4), 612–624. http://dx.doi.org/10.1139/t99-038 Geotechnical Engineering, 1, 3–20. http://dx.doi.org/
Robertson, P. K. (1990). Soil classification using the cone 10.1680/ravige.34860.0001
penetration test. Canadian Geotechnical Journal, 27(1), Wang, Y., Huang, K., & Cao, Z. (2013). Probabilistic iden
151–158. http://dx.doi.org/10.1139/t90-014 tification of underground soil stratification using cone
Robertson, P. K. (2009). Interpretation of cone penetration penetration tests. 776(May), 766–776.
tests — a unified approach. Canadian Geotechnical Zhang, L., Asce, M., Tang, W. H., Asce, H. M., Zhang, L.,
Journal, 46(11), 1337–1355. http://dx.doi.org/10.1139/ & Zheng, J. (2004). Reducing Uncertainty of Prediction
t09-065 from Empirical Correlations. May, 526–534.
338
© 2022 the Author(s), ISBN 978-1-032-31259-0
Clustering analysis to improve total unit weight prediction from CPTu

S. Collico & M. Arroyo
Department of Civil and Environmental Engineering (DECA), Universidad Politecnica de Cataluña, Barcelona,
Spain
M. DeVincenzi, A. Rodriguez & A. Deu

Igeotest s.l. Figueres, Spain
ABSTRACT: Accurate estimates of soil unit weight are fundamental for correctly post process CPTu data
and making use of Soil Behavior Type-based classification systems. Soil-specific and global regressions have
been proposed for this purpose. However, soil-specific correlation might pose a problem of pertinence when
applied at new sites. On the other hand, global correlations are easy to apply, but generally carry large system
atic uncertainties. In this context, this work proposes a data clustering technique applied to geotechnical data
base aiming to identify hidden linear trends among dimensionless soil unit weight and normalized CPTu
parameter according to some unobservable soil classes. Global correlations are then revisited according to
such data subdivision aiming to improve accuracy of soil unit weight prediction while reducing transform
ation uncertainty. A new probabilistic criterion for soil unit weight prediction is also obtained. The potential
benefits of the proposed procedure are illustrated with data from a Llobregat delta site (Spain).
1 INTRODUCTION apply at a particular site because they have less

coverage and require a previous soil classification
CPTu interpretation is almost always based on stress step, which may introduce additional uncertainty.
normalized cone readings. For instance, soil classifi One possible way out of this problem is to develop
cation charts (Robertson, 1990; Been & Jefferies, soil-specific regressions for unit weight using
2006; Schneider et al., 2008) use stress normalized a global database that is segmented into soil classes
tip resistance, friction and excess pore pressure for regression purposes. How to define those classes?
parameters. Total soil unit weight γt is thus In this work we employ a Gaussian Mixture Model
a necessary input in any CPTu interpretation exercise. (a.k.a., GMM) technique to identify hidden classes in
Soil unit weight is measured on undisturbed sam a global database described by Mayne, (2014) with the
ples from boreholes. Such samples are not always purpose of establish more accurate regressions. GMM
available, particularly for granular soils and/or in early have been previously applied to CPTu data analyses
stages of site investigation. As a simpler alternative, (Depina et al., 2016; Krogstad et al., 2018) to identify
total soil unit weight can be estimated indirectly from soil classes for stratigraphic delineation. In those stud
piezocone and seismic piezocone readings. In the last ies, a Bayesian perspective was introduced, as having
decades several proposals have been presented to the possibility of updating the stratigraphic groups as
achieve this purpose (Mayne, 2009; Mayne et al., more information was gathered was deemed essential.
2010; Robertson & Cabal, 2010; Mayne, 2014; Leng In the present work a fixed database in considered and
keek, 2018). therefore the Bayesian updating aspect has been
These proposals are all based on empirical regres omitted.
sion using databases in which cone readings are A key step of applying such data clustering tech
paired with soil unit weight measurements. A trade- nique is to express the reference database, including
off typically arises between coverage (the scope of soil unit weight and CPTu data, as a multivariate
the original soil database) and precision (the predict normal distribution before GMM is applied. To this
ive power of the regression). Global regressions, usu end we use a methodology laid out by Ching et al.,
ally described as applicable to soils with a "normal" (2014), to rationally account for predictor co-
or common mineralogy, are widely applicable but dependence and to establish or revise correlations
less precise than regressions developed only for between different variables.
a certain class of soil (i.e., soil specific). On the other In what follow, key steps to construct valid multi
hand, soil specific regressions are more difficult to variate distribution are reported followed by a brief
DOI: 10.1201/9781003308829-45
339
description of finite Gaussian Mixture Models. In
the results section, an example of data subdivision
and assignment to identified hidden classes is
reported using dimensionless soil unit weight and
normalized CPTu parameters. Existing correlations
are re-examined using the new soil classes and
a new one is proposed.
2 METHODOLOGY
2.1 Setting up a multivariate distribution

Ching et al., (2014) and Phoon & Ching, (2018) pro
posed a systematic cumulative transform procedure
to build standard multivariate normal distributions
for multivariate databases. Such approach is adopted
in this study to treat the Mayne, (2014) database.
Essentially the steps involved are:
1. For a given set of selected observations Ω (e.g.,
Ω ¼ ½Ω1 ; Ω2 ] with Ω1 = [nx1] and n number of
observations) define marginal distributions Ωi for
each of the component variates in the database;
2. Transform the different marginals into standard
ized normal distributions Xi
3. Assume that transformed observations
X ¼ ½X1 ; X2 ] follow a multivariate normal distri
bution f ðXÞ;
4. Obtain a random sample from f ðXÞ, and trans
form it back to the non-normalized space, obtain
ing the simulated set Ωsim
5. Check by inspection the overlap between Ωsim Figure 1. Fitting Johnson probability density function to a)
and Ω. ln (Qtn Þ; b) γt =γw data of Mayne, (2014) database.
A key step in the procedure is the definition of
marginal distributions for the different variates in the
database (in this case dimensionless soil unit weight
and normalized CPTu parameters). Lognormal distri underlying components (the hidden classes that we
butions have been frequently used in geotechnical are trying to identify). This is formalized using
applications (Phoon & Kulhawy, 1999). The lognor a Gaussian Mixture Model, (Depina et al., 2016;
mal is included in the Johnson system of distribu McLachlan & Peel, 2000), that expresses the trans
tions (Ching et al., 2014), which has three main formed -standardized- multivariate distribution as
families: the lognormal system SL , the bounded a linear combination of K multivariate gaussians:
system SB and unbounded system SU . Choosing
between the whole system of Johnson distributions
offers versatility while maintaining the interesting
property of having analytical expressions to trans
form into standard normal variables and back (Ching
et al. 2014). Each j represents a component of the mixture and
The selection of an appropriate Johnson distribu is a multivariate normal probability density function
tion for a particular dataset is based on the first four with the same structure as f ðXÞ. Therefore, each
moments of a sampling distribution (George & component has its own statistical parameters, denoted
Ramachandran, 2011). Examples of Johnson distri by Yj In our case they includeP a vector of means and
butions fitted to different variates of the Mayne, a covariance matrix (i.e., μj , j). The collection of
(2014) global database are given in (Figure 1). all the Yj is denoted by Y. Each component has
a weight πj (proportion of data assigned to j );
weights are chosen so that they add up to one and are
2.2 Multivariate as a gaussian mixture collected in a vector I ¼ (π1 ,.., πK ). ζ ¼ collection
The multivariate distribution that represents the data of all the unknown parameters of the mixture
base might be conceived as a combination of several model ði:e:; ½Y; I]Þ.
340
Depending on the values that ζ finally takes each 3 RESULTS
observation xi would have a certain probability pij
of belonging to a particular component j . 3.1 Cluster definition and analysis
Gaussian Mixture analysis consists of estimating
the most probable ζ, b ζ. This is generally done There are six variates in the Mayne, (2014) database.
through the Expectation-Maximization (EM) algo To illustrate the methodology proposed in a simple
rithm (Samé et al., 2011; Huang et al., 2017; Liu setting, we consider only two of them, namely
et al., 2019) which alternates an expectation step in dimensionless unit weight γt =γw and normalized tip
which observations are assigned exclusively to resistance, . After fitting a bivariate distribu
a particular gaussian component and a maximization tion to all the data pairs we run the
step, in which the log-likelihood function for the expectation-maximization algorithm to identify
incomplete dataset is maximized. a GMM with two components.
The concept of GMM could be also formulated An example illustrating the probabilities of data
in a Bayesian framework, by integrating prior belonging to a particular component of the Gaussian
knowledge p (ζ) and observations Ω to obtain mixture is given in Figure 2a. Clustering is based on
posterior estimates of ζ. Due to lack of previous such probabilities. A simple choice is to assign data
studies, this work only applies GMM. For more to the component in which they have the largest
detailed information about the Bayesian formula probability of belonging. For two components this is
tion, analysis about optimal number of hidden equivalent to enforcing a probability threshold of 0.5
classes to consider and most plausible normalized as clustering criteria. The result of doing so in this
CPTu parameters to be considered, the reader is case is reported in Figure 2b, while statistics for the
referred to Collico, (2021). two clusters in the Ω space are reported in Table 1.
To understand the meaning of this newly identi
fied soil classes, the clustered data is plotted in Soil
Behavior Type charts (Robertson, (2016); Schneider
et al. (2008) (Figure 3). Results show that the first
hidden class identified is dominated by Clay-Like-
Contractive soils (C - C, Figure 3a), while
the second hidden correlation identified is repre
sented by a wider range of conventional soil types.
Table 1. Covariance matrix and mean vector of

γt =γw -lnðQtn Þ hidden classes identified.
Class Mean Covariance
1 μγt =γw γt =γw

0.04 -0.036
1.59 1.71
-0.036 0.125
2 μγt =γw γt =γw
0.025 0.04
1.95 3.32 0.04 0.94
3.2 Generating cluster-based correlations

The clustered data can be used to generate new correl
ations. Such correlations need not be based on the
same restricted subset of the global database that was
used to generate the clustering. For instance, we use
here correlations that follow a template proposed by
Mayne, et al., (2010) (i.e., γt =a+ b·log(z) + c·logðf s Þ+
d·logðqt Þ). Results in terms of coefficient of determin
ation, R2 and regression standard error, σεT are
reported in Table 2, while regression coefficients are
reported in Table 3. The standard error, after GMM
subdivision, is at least 13% smaller than the one of
global database. The coefficient of determination R2 is
Figure 2. A) Assigned probability of belonging to compo also smaller for the clusters than for the global data
nent 1 of the GMM b) Scatter and marginal distributions of base. These effects are particularly strong for Hidden
observations for the two hidden classes identified. class 2.
341
computed at each trial, for each one of the correl
ations (global, cluster 1, cluster 2). 500 such simu
lations were performed. Results, reported in
Figure 4 for both Hidden classes, highlight the
benefit of the GMM as the distribution of error
norms is clearly shifted towards lower values.
Table 3. Coefficient of the regression for linear form of

Mayne (2010) after BMA subdivision.
a b c d
Global 8.78 -0.67 2.24 1.457

Class1 -0.78 -4.2 7.77 -0.57
Class2 15.34 0.052 0.02 1.94
Figure 3. Scatter plot of data belonging to each hidden

cluster identified on SBT charts: a) Robertson, (2016); b)
Schneider et al., (2008).
Table 2. Coefficient of determination and regression

standard error before and after database subdivision.
R2 σ εT
Global 0.66 1.44

Hidden class1 0.6 1.25
Hidden class 2 0.39 1.22
However, those comparisons are somewhat mis

leading, as the statistics are computed using differ
ent observations. The improvement of predictive
strength after clustering is more clearly identified
through a cross-validation procedure. To this end
we run a simulation exercise in which we ran
domly selected 85% of the data in each cluster and
used them to fit cluster specific correlations, as
well as a global one in which no cluster distinction
was made. Then we applied those correlations, to Figure 4. Cross validation of revisited regression and
comparison with global literature correlation for a)
the remaining 15% of the dataset -the validation
Hidden class1. B) Hidden class 2.
data. The sum of squared residuals jjSSres jj was
342
3.3 Application to new sites: Decision boundary The same is done using the global correlation, obtained
using the whole database, without clustering. Summary
To assign CPTu observations at a new site to a cluster,
results are also presented in Table 4. They include the
we use the (non- normalized) Robertson, (1990) SBT
chart. Plotting the clustered data in that chart we apply
discriminant analysis (Ghojogh & Crowley, 2019) to Table 4. Statistics of mean value of total unit weight
establish a user-friendly separation criteria for new prediction.
CPTu observations. The decision boundary (Figure 5)
μγt σμγt
obtained takes the shape of a quadratic in Robertson
(1990) chart. This boundary line between the two clus Global correlation 17.47 0.48
ters has the following expression: Cluster-based correlation 18.1 0.42
Laboratory 19.45 0.66
with σboundary systematic uncertainty associated

with decision boundary (σboundary ¼ 0:13Þ.
Figure 5. Mean estimate and 95% Confidence Interval of

decision boundary.
3.4 Illustrative example

The cluster-based correlations just developed are now
applied to results of a particular CPTu campaign per
formed at a Llobregat site (SW Barcelona-Spain).
Several infrastructures have been developed around
this site during the last decades, requiring extensive
onshore-nearshore geotechnical investigation. Along
with in-situ investigation, laboratory tests were per
formed on selected sub-samples from undisturbed
Shelby tubes recovered every 5m in each borehole.
For this study we consider data from 20 CPTu and 44
total soil unit weight measurements. More detail on
this campaign is given by Deu et al., (2021).
All the CPTu data were plotted on Robertson
graph, showing that the site is dominated by silty
soils. Then each CPTu observation was assigned to
class 1 or 2 according to the mean value of the deci
sion boundary (Figure 6a). Most of the data were
assigned to Hidden class 2. Figure 6. A) CPTu data assigned to class 1 and class 2
The cluster-based correlation was applied to the 20 according to the proposed criterion. b) Profile of total unit
CPTu data and the results are plotted in (Figure 6b), weight prediction for cluster-based and global correlations
indicating the mean and spread of the predicted values. and laboratory observations.
343
overall predicted mean estimate as well as the standard of Modern Applied Statistical Methods, 10(2),494–504.
deviation of the means evaluated at different depths. https://doi.org/10.22237/jmasm/1320120480
The cluster-based correlation clearly exhibits an Ghojogh, B., & Crowley, M. (2019). Linear and Quadratic
improvement of total unit weight prediction at the Discriminant Analysis: Tutorial. 4, 1–16. http://arxiv.
site (Figure 5b), when compared global correlation. org/abs/1906.02590
It has a lower dispersion and its mean is closer to Huang, T., Peng, H., & Zhang, K. (2017). Model selection
that of the laboratory measurements. It is clear, how for Gaussian mixture models. Statistica Sinica, 27
(1),147–169. https://doi.org/10.5705/ss.2014.105
ever, that even the cluster-based correlation under Krogstad, A., Depina, I., & Omre, H. (2018). Cone penetra
estimates the laboratory mean. There are several tion data classification by Bayesian inversion with
possible explanations for this discrepancy. One is a Hidden Markov model. Journal of Physics: Confer
that samples of silty soils generally tend to be densi ence Series, 1104(1). https://doi.org/10.1088/1742-6596/
fied upon sampling (Lim et al. 2018). The other is 1104/1/012015
that the reference database is somewhat scarce in the Lengkeek, H. J. (2019). CPT based unit weight estimation
silt area were most datapoints in the example lie extended to soft organic soils and peat. January 2018.
(compare Figures 6a and 3a). Lim GT, Pineda JA, Boukpeti NA, Fourie AN, Carraro JA.
Experimental assessment of sampling disturbance in cal
careous silt. Géotechnique Letters. 2018 Sep;8(3):240–7.
4 CONCLUSION Liu, Y., Ye, L., Qin, H., Ouyang, S., Zhang, Z., & Zhou, J.
(2019). Middle and Long-Term Runoff Probabilistic
Forecasting Based on Gaussian Mixture Regression.
This study describes a first attempt to apply GMM Water Resources Management, 33(5),1785–1799.
clustering to geotechnical database. https://doi.org/10.1007/s11269-019-02221-y
The clustering technique has been applied in Mayne, P.W., Peuchen, J., & Bouwmeester, D. (2010). Soil
a space parametrized by dimensionless soil unit unit weight estimation from CPTs. 2nd International
weight and normalized cone tip resistance, assuming Symposium on Cone Penetration Testing, 2 (May),8.
the existence of two unobserved classes, selected at https://doi.org/10.1201/b10132-41 10.1201/b10132-41
50% probability. One of emerging hidden classes can Mayne, P. W. (2014). Interpretation of geotechnical param
eters from seismic piezocone tests. 3rd International
be associated to Clay-Like-Contractive class, while
Symposium on Cone Penetration Testing (CPT`14),
the second is representative of a wide range of soil 47–73.
types. Assignation of new data to the hidden classes Mayne, Paul W, Coop, M. R., Springman, S. M., &
is based on a discriminant line generated on non- Zornberg, J. G. (2009). Geomaterial behavior and test
normalized SBT charts (Robertson, 1990). CPTu data ing. Proceedings of the 17th International Conference
subdivision according to such classes has been shown on Soil Mechanics and Geotechnical Engineering: The
to increase the predictive strength of correlations. Academia and Practice of Geotechnical Engineering, 5,
This has been illustrated using a correlation template 2777–2872.
from Mayne et al., (2010) in which unit weight is pre McLachlan, G. J., & Peel, D. (2000). Finite mixture
dicted from shaft friction, depth and tip resistance. models. John Wiley & Sons.
Phoon, K. K., & Ching, J. (2018). Risk and Reliability in
Beyond the particular application explored here, Geotechnical Engineering. In Risk and Reliability in
the technique presented here may prove useful to Geotechnical Engineering. https://doi.org/10.1201/
improve accuracy and precision of empirical CPTu b17970
based correlations established using global databases. Phoon, Kok Kwang, & Kulhawy, F. H. (1999). Character
ization of geotechnical variability. Canadian Geotech
nical Journal, 36(4),612–624. https://doi.org/10.1139/
REFERENCES t99-038
Robertson, P. K. (1990). Soil classification using the cone
Ching, J., Phoon, K., & Chen, C. (2014). Modeling piezo penetration test. Canadian Geotechnical Journal, 27
cone cone penetration (CPTU) parameters of clays as (1),151–158. https://doi.org/10.1139/t90-014
a multivariate normal distribution. 91(July 2012), 77–91. Robertson, P. K. (2016). Cone penetration test (CPT )
Collico, S., (2021). Geotechnical characterization of marine based soil behaviour type ( SBT ) classification
sediments via statistical analysis (Unpublished doctoral system — an update. 1927(July), 1910–1927.
dissertation). Universidad Politècnica de Cataluña Robertson, P. K., & Cabal, K. L. (2010). Estimating soil
(UPC), Barcelona, Spain. unit weight from CPT. May.
Depina, I., Le, T. M. H., Eiksund, G., & Strøm, P. (2016).
Samé, A., Chamroukhi, F., Govaert, G., & Aknin, P.
Cone penetration data classification with Bayesian Mix
(2011). Model-based clustering and segmentation of
ture Analysis. Georisk, 10(1),27–41. https://doi.org/
time series with changes in regime. Advances in Data
10.1080/17499518.2015.1072637
Analysis and Classification, 5(4),301–321. https://doi.
Deu, A., Marti X., Peña S., Tarragò D., Gens A., &
org/10.1007/s11634-011-0096-5
Devincenzi, M. (2021) DMT, CPTU and laboratory tests Schneider, J. A., Randolph, M. F., Mayne, P. W., &
comparison for soil classification and strength param Ramsey, N. R. (2008). Analysis of factors influencing
eters of deltaic soft soils in Barcelona Port, Proceedings soil classification using normalized piezocone tip resist
of the 6th International Conference on Site Characteriza ance and pore pressure parameters. Journal of Geotech
tion ISC-6 Budapest nical and Geoenvironmental Engineering, 134
George, F., & Ramachandran, K. M. (2011). Estimation of (11),1569–1586. https://doi.org/10.1061/(ASCE)1090
parameters of johnson’s system of distributions. Journal 0241(2008)134:11(1569)
344
© 2022 the Author(s), ISBN 978-1-032-31259-0
Combining CPTU and UMASW to characterise Irish offshore deposits

M. Coughlan
School of Civil Engineering, University College Dublin, Belfield, Dublin, Ireland
SFI Research Centre for Applied Geosciences (iCRAG), Ireland
Gavin and Doherty Geoslutions Ltd., Dublin, Ireland
A. Trafford, M. Long & S. Donohue

School of Civil Engineering, University College Dublin, Belfield, Dublin, Ireland
SFI Research Centre for Applied Geosciences (iCRAG), Ireland
S. Corrales
APEX Geophysics, Gorey, Wexford, Ireland
ABSTRACT: The north Irish Sea is earmarked for the development of several offshore wind farms. The
complex Quaternary history of this area has resulted in a legacy of geotechnically challenging deposits,
including heterogeneous glacial deposits, soft sediments and gas-charged sediments. Multichannel Analysis
of Surface Waves (MASW) is commonly used onshore to produce shear wave velocity (Vs) profiles and is
increasingly being used offshore (UMASW). When used in tandem with in-situ Cone Penetration Testing
(CPTU) profiles it is a powerful tool for characterising offshore deposits. UMASW profiles were obtained
from sites across the north Irish Sea coincident with CPTU profiles. Vs profiles from UMASW show good
agreement with CPTU parameters in a range of geological settings. The purpose of this paper is to present
a characterisation of geological deposits using a combined approach of CPTU and UMASW. By doing so the
aim is to further develop classification charts that can give reliable characterisations of offshore deposits by
combing non-invasive geophysical techniques and traditional CPTU.
1 INTRODUCTION The north Irish Sea has experienced a complicated

geological history comprising of ice-sheet advance
1.1 Background and retreat coupled with marine transgression and
complex sea-level rise (e.g. Scourse et al., 2021;
The use of the Multichannel Analysis of Surface
Ward et al., 2016). As a result, glacial deposits have
Waves (MASW) technique is used extensively in
been deposited on bedrock, which in turn are over
onshore geotechnical investigations, providing
laid by marine sediments that are known to host shal
a profile of shear wave velocity (Vs) against depth.
low gas (Coughlan et al., 2019). These geological
Additional geotechnical parameters, such as the small
conditions create a number of potential geohazards
strain stiffness (Gmax), can be derived using this Vs
and geotechnical constraints that may inhibit the
profile and subsequently used as input for foundation
extensive development of offshore wind planned for
design. Typically, measurement of Gmax requires high-
the area (Coughlan et al., 2020; Guinan et al., 2020).
quality undisturbed samples for laboratory-based test
Shear wave velocity (Vs) profiles were acquired
ing, which are challenging to acquire and only measure
using the UMSAW technique from a number of loca
discrete sampling locations. In this way, in-situ geo
tions where CPTU data existed. This paper presents Vs
physical testing like UMASW offers a rapid method of
and CPTU data in order to characterise challenging
measuring the bulk response of undisturbed sediment
ground conditions, such as heterogeneous glacial
(Long & Donohue, 2007). When combined with other
deposits and soft marine sediments hosting shallow gas.
in-situ testing, such as Cone Penetration Testing
(CPTU), Vs profiles derived from the Underwater
MASW (UMASW) have been shown to be powerful 1.2 Site location and description
complimentary tools in characterising marine sedi
The study area is outlined in Figure 1. In this area
ments (e.g. Long & Donohue, 2010) and in offshore
UMASW data were collected at 19 locations, corres
geotechnical site investigations (Long et al., 2020).
ponding with locations where previous geotechnical
DOI: 10.1201/9781003308829-46
345
Figure 1. Study area map with data site locations. Areas of shallow gas and glacial deposits are from Coughlan et al.
(2020).
data had been acquired. The data presented in this margins mapped onshore and is likely correlative to
paper were obtained at 2 locations representing differ the Clogher Head and/or Killard Point Stadial, both
ing geological conditions: soft sediments with shallow of which are exposed in onshore sections, in
gas (Site 1) and glacial deposits (Site 2). Site 1, located a period spanning 18.2 to 16.5 ka BP (McCabe
in an area known as the Western Irish Sea Mud Belt et al., 2005).
(WIMSB: Belderson, 1964). Coughlan et al. (2019),
identified four stratigraphic units. These consist of
a basal subglacial (lodgement) till emplaced by the 2 MATERIALS AND METHODS
Irish Sea Ice Stream (ISIS) as it advanced across the
area, underlain by irregular bedrock. As the ISIS 2.1 Cone Penetration Testing (CPTU)
retreated during deglaciation,, ice-proximal outwash
gravels, sands and silts were deposited in CPTU data was acquired in the study area in 2014
a glaciomarine to glaciolacustrine. As deglaciation con onboard the RV Celtic Explorer, using the University
tinued, the area became increasingly ice-proximal and of Bremen developed Geotechnical Offshore Seabed
dark muddy sands were deposited in a glaciomarine to Tool (GOST) system. The hardened stainless steel
marine environment. Finally, Holocene marine muds cone tip had a cross-sectional area of 5 cm2 and
varying in thickness lie above an erosive horizon mark sleeve area of 75 cm2 recording tip resistance, sleeve
ing the base of the Holocene (Woods et al., 2019), and friction and differential pore pressure with a sampling
form the contemporary seabed. These Holocene sedi frequency of ~30 Hz, resulting in a vertical resolution
ments are known to host shallow gas (Coughlan et al., of ~2 mm with a maximum tip resolution of 0.06
2021; Yuan et al., 1992). MPa and a range up to 120 MPa. The GOST system
At Site 2, to the north and north-west, the Holo has an operational weight of between 2 and 8 tonnes
cene muds and silts thin out and coarsen to sands depending on the addition of weighted plates for
with the underlying glacial units out- and sub- extra stability. It has 8 tonnes of hydraulic push
cropping at the seafloor. In the vicinity of Dundalk power with variable hydraulic pressure of 0-20 MPa.
Bay, a large arcuate moraine ridge is evidence of Exact control on push velocity during penetration
a readvance of the Irish Ice Sheet eastward (Callard allows for data to comply to the highest international
et al., 2019). This feature is correlated with ice including DIN 4904 and ISO 22476-1:2012
346
requirements. The interface was a digital one of from the dispersion curves defines both the upper and
industrial RS485 BUS using direct A/D converting of lower depth limits of the inversion process (Figure 2).
a measured variable with overvoltage and reverse pro The shallowest resolvable depth is approximately 1/3
tection. Raw CPTU data were calibrated using stand to 1/2 of the maximum phase velocity wavelength
ard methods according to (Lunne et al., 1997). (c/f). The maximum resolvable depth is defined in the
same way by the minimum phase velocity wave
length. In this instance the range of reliable Vs values
2.2 Underwater Analysis of Surface Waves is typically between 3 and 25 m.
(UMASW)
UMASW data were acquired as part of the Developing 3 RESULTS
Site Investigation Methodologies and Constraint Map
ping Products for Offshore Renewable Energy
(DeSIRE) Survey in 2020. The set-up comprised two 3.1 Site 1
Geometrics Model DHA-7 bay cables, each consisting CPTU and UMASW for Site 1 is shown in
of 24 hydrophones moulded onto the receiver cable, Figure 3. The data show 2.5 m to 5 m of a Very
that was deployed onto the seabed from a main vessel Soft clayey silt (Unit 1) overlying a Soft to Firm
(Ocean Navigator) using a smaller auxiliary vessel clay layer to circa. 18 m (Unit 2). At 18 m there is
(Fionn MacCumhaill). Hydrophone spacing on the a transition to a Firm to Stiff clayey Silt/sandy Silt
cable was 3.125 m and it was attached to a mooring (Unit 3), with a corresponding drop in u2. The tran
weight at the Ocean Navigator end, whilst a clump sition from the base of Unit 2 into Unit 3 at Site 1
weight deployed by the Fionn MacCumhaill secured it correlates with the presence of shallow gas as deter
at the other end. The Fionn MacCumhaill also acted as mined by seismic sparker profiles in Coughlan et al.
a gunboat for the seismic source, which comprised (2019) and Figure 4.
a Sercel 12 cu.in. Mini G. Gun, allowing for shots to During the acquisition of seismic sparker profiles
be taking at different positions along the cable the presence of shallow gas, predominantly contained
(McGrath et al., 2016). within the Holocene fine-grained sediments, results in
Processing of UMASW data involved the Fourier high acoustic reflectivity at the top of the gas unit. This
transform of time series shot data to produce phase reduces the signal to noise ratio manifesting itself as
velocity (c) – frequency (f) intensity plots, often acoustic turbidity within the gas and acoustic blanking
referred to as dispersion curves images. The method below the gas. These effects result in a lower reso
relies on the dispersive nature of surface waves lution of the underlying units, reducing the ability to
where different frequencies sample different depths identify and pick horizons from the seismic profiles.
within the ground travelling at different velocities. (e.g. Judd and Hovland, 1992; Tóth et al., 2014). In
Surfseis software developed by the Kansas Geo the DeSIRE study area, shallow gas had previously
logical Survey was used for the processing and ana been mapped by Coughlan et al. (2019, 2021;
lysis of the surface wave data (Park et al., 1999, Figure 1).
2007). Once the surface wave response has been iso The division between Unit 1 and 2 is most
lated and the high semblance curve picked the soft clear from the u2 data, where the pore pressure
ware carries out a least squares inversion of the data picks up in the clay but can also be seen from
to produce a 1D Vs profile. The UMASW data were the change in slope of the qt and fs profiles. The
processed using a default 10-layer variable thickness division between units 2 and 3 is clear from all 3
model based on the input picked dispersion curves. sets of CPTU data. As the cone penetrates from
These parameters are routinely used to give a good the clay to the more coarse material, qt increases,
approximation of the seismic layer thickness and fs (and Rf) reduces and u2 drops back towards
resulting Vs profile. The frequency content resolved the hydrostatic value.
Figure 2. Dispersion curves produced in 45 m water depth at Site 1 and Site 2 in 25 m water depth.
347
Figure 3. CPTU data for Site 1 is presented in the form of (a) corrected tip resistance (qt), (b) sleeve friction (fs) and fric
tion ratio (Rf) and (c) pore pressure generated in the u2 filter position and normalised pore pressure parameter (Bq) versus
depth (Lunne et al., 1997). In addition the parallel UMASW profiles for Site 1 are shown in (d).
Figure 4. Seismic profile for line highlighted in Figure 1 (see Coughlan et al. (2019) for further details).
The UMASW profiles (Figure 3d) are consistent charts but suggests that the material is an organic-
with the CPTU data, correlating well with the qt type material. The 5 m to 18 m clay sediment
values. The profile identifies Very Soft sediments (Unit 2) is classified correctly as a sensitive fine-
(< 100m/s) to the base of Unit 1 at c. 5 m increasing grained material or a clay. Finally, the deeper mater
from Soft to Firm to Stiff (100 – 180 m/s) at the base ial (Unit 3) is classified as either a clay or a silty
of Unit 2 at c. 18 m. Unit 3 corresponds to Stiff clay.
material (>180 m/s) below this depth. Stiffness des
ignations according to the classification system pro
3.2 Site 2
posed by Poulos (2021).
The Robertson et al. (1986) soil behaviour type Data for Site 2 is shown on Figure 6. Only limited
chart (SBT) for Site 1 is shown on Figure 5 and is CPTU data is available here as no fs and u2 data
consistent with the findings presented above. Both were recorded. The qt values confirm that the mater
the Bq/qt and Rf/qt components of the chart work ial is highly competent with qt values increasing rap
well for this case. The data for the upper 5 m of the idly with depth up to a value of about 18 MPa with
sediment (Unit 1) is close to the boundaries of the refusal at 3.25 m. However, the UMASW data was
348
checking across profiles and with available lab
data.
In this study the UMASW technique was success
fully applied across this variety of geological condi
tions and a range of water depths from 17 to 45 m,
obtaining reliable Vs profiles to c. 25 mbsb. The two
sites presented in this paper highlight the presence of
soft Holocene sediments with shallow gas (Site 1),
as well as glacial deposits (Site 2).
UMASW proved effective at Site 2 in characteris
ing heterogenous glacial deposits at depths below
CPTU refusal. At Site 1 MASW showed good cor
Figure 5. Soil behaviour type classification for Site 1 relation with CPTU to similar depths, however both
CPTU data (from Robertson et al., 1986). methods were ineffective in determining the pres
ence of shallow gas, which was identified using seis
mic sparker profiles. Vs and CPTU profiles show
good agreement and can be used to develop classifi
cation charts for offshore deposits in complex geo
logical areas.
The UMASW application is shown to be a proven
method of providing geotechnical parameters for
sub-seabed characterisation. When used in combin
ation with other complimentary methods such as
CPTU and seismic reflection a more robust ground
model can be delivered.
ACKNOWLEDGEMENTS
This project was carried out at the SFI Centre of

Research in Applied Geoscience (iCRAG) with fund
ing from the Geological Survey of Ireland (GSI) as
well as the Sustainable Energy Authority of Ireland
Figure 6. CPTU data for Site 2 is presented in the form of (SEAI) under their Research, Development and Dem
(a) tip resistance (qt) (Note the different x-axis scales from onstration Funding Funding Programme 2019 (Award:
Figure 3). In addition, the corresponding UMASW profile 19/RDD/411). Dr. Mark Coughlan is funded under an
is shown in (b).
Irish Research Council Enterprise Partnership Scheme
Postdoctoral Fellowship (EPSPD/2020/109) and in
able to resolve data to c. 15 m. The classification part by a research grant from Science Foundation Ire
system proposed by Poulos (2021) suggests the land (SFI) under Grant Number 13/RC/2092, with
material designation as increasing Firm, through support from Gavin and Doherty Geosolutions Ltd
Stiff and Very Stiff to Hard with depth. and the Geological Survey of Ireland (GSI). The
authors would like to thank Tony Lombard, Ian Shar
key (Apex Geophysics) and Eoin Grimes (Irish Com
4 CONCLUSIONS mercial Charter Boats) for their skill and support in
acquiring UMASW data. CPTU data was acquired
The area of the north Irish Sea studied here onboard the RV Celtic Explorer (CE14001: Develop
exhibits a lateral variability of complex deposits, ing Geotechno-stratigraphies), which was funded
which are potentially challenging to the deploy through the Marine Institute Ship-time Programme
ment of offshore renewable energy infrastructure and the Sustainable Energy Authority of Ireland
such as foundations. These deposits include soft, (SEAI) Ocean Energy Research and Development
fine-grained Holocene sediments, gas-charged Programme 2014 (Award: OCN 00016).
sediments, and heterogeneous glacial deposits.
Previous efforts at CPTU data acquisition in these
sediments had mixed results, with limited penetra REFERENCES
tion in particular in glacial deposits. The use of
Belderson, R. H. (1964). Holocene sedimentation in the
SBT based on Robertson et al. (1986) is con western half of the Irish Sea. Marine Geology, 2,
sidered suitable here due to the relatively shallow 147–163.
depths and nature of the sediment known in the Callard, S. L., Plets, R., Cooper, A., Long, A., Kelley, J.,
area. Future work will include comparison with Belknap, D., Edwards, R., Quinn, R., & Jackson, D.
updated charts (e.g. Robertson, 2016) and cross (2019). The evolution of a submerged glacial and
349
deglacial landscape from east Ireland (Dundalk Bay and Conf. on Geotechnical and Geophysical Site Character
Carlingford Lough). International Union for Quaternary isation (ISC’5) (pp. 911–916). Australian Geomechanics
Research. Society.
Coughlan, M., Long, M., & Doherty, P. (2020). Geological Park, C. B., Miller, R. D., & Xia, J. (1999). Multichannel
and geotechnical constraints in the Irish Sea for offshore analysis of surface waves. GEOPHYSICS, 64(3),
renewable energy. Journal of Maps, 16(2), 420–431. 800–808. https://doi.org/10.1190/1.1444590
https://doi.org/10.1080/17445647.2020.1758811 Park, C. B., Miller, R. D., Xia, J., & Ivanov, J. (2007).
Coughlan, M., Roy, S., O’Sullivan, C., Clements, A., Multichannel analysis of surface waves (MASW)
O’Toole, R., & Plets, R. (2021). Geological settings and Active and passive methods. Leading Edge (Tulsa, OK),
controls of fluid migration and associated seafloor seep 26(1), 60–64. https://doi.org/10.1190/1.2431832
age features in the north Irish Sea. Marine and Petrol Poulos, H. G. (2021). Use of shear wave velocity for foun
eum Geology, 123. https://doi.org/10.1016/j. dation design. Geotechnical and Geological Engineer
marpetgeo.2020.104762 ing, in press.
Coughlan, M., Wheeler, A. J., Dorschel, B., Long, M., Robertson, P. K. (2016). Cone penetration test (CPT)-based
Doherty, P., & Mörz, T. (2019). Stratigraphic model of soil behaviour type (SBT) classification system — an
the Quaternary sediments of the Western Irish Sea Mud update. Casnadian Geotechnical Journal, 53(12),
Belt from core, geotechnical and acoustic data. Geo- 1910–1927. https://doi.org/10.1139/cgj-2016-0044
Marine Letters, 39(3), 223–237. Robertson, P. K., Campanella, R. G., Gillespie, D., &
Guinan, J., McKeon, C., O’Keeffe, E., Monteys, X., Greig, J. (1986). Use of Piezometer Cone data. ASCE
Sacchetti, F., Coughlan, M., & Nic Aonghusa, C. Speciality Conference In-Situ ’86: Use of In-Situ Testing
(2020). INFOMAR data supports offshore energy devel in Geotechnical Engineering. American Society of
opment and marine spatial planning in the Irish offshore Engineers (ASCE), 1263–1280.
via the EMODnet Geology portal. Quarterly Journal of Scourse, J. D., Chiverrell, R., Smedley, R., Small, D.,
Engineering Geology and Hydrogeology, 54, Burke, M., Saher, M., Van Landeghem, K.,
qjegh2020-033. https://doi.org/10.1144/qjegh2020-033 Duller, G. A. T., Ó Cofaigh, C., Bateman, M.,
Judd, A. G., & Hovland, M. (1992). The evidence of shal Benetti, S., Bradley, S. L., Callard, S. L., Evans, D.,
low gas in marine sediments. Continental Shelf Fabel, D., Mc Carron, S., Medialdea, A., Moreton, S.,
Research, 12(10), 1081–1095. Ou, X., … Clark, C. (2021). Maximum extent and read
Long, M., & Donohue, S. (2007). In situ shear wave vel vance dynamics of the Irish Sea Ice Stream and Irish
ocity from multichannel analysis of surface waves Sea Glacier since the Last Glacial. Journal of Quater
(MASW) tests at eight Norwegian research sites. Can nary Science, 1–25. https://doi.org/10.1002/jqs.3313
adian Geotechnical Journal, 44(5), 533–544. https://doi. Tóth, Z., Spieß, V., & Jensen, J. (2014). Seismo-acoustic
org/10.1139/t07-013 signatures of shallow free gas in the Bornholm Basin,
Long, M., & Donohue, S. (2010). Characterization of Nor Baltic Sea. Continental Shelf Research, 88, 228–239.
wegian marine clays with combined shear wave velocity https://doi.org/10.1016/j.csr.2014.08.007
and piezocone cone penetration test (CPTU) data. Can Ward, S. L., Neill, S. P., Landeghem, K. J. J. Van,
adian Geotechnical Journal, 47(7), 709–718. https://doi. Scourse, J. D., Bradley, S. L., & Uehara, K. (2016). Sen
org/10.1139/T09-133 sitivity of palaeotidal models of the northwest European
Long, M., Trafford, A., McGrath, T., & O’Connor, P. shelf seas to glacial isostatic adjustment since the Last
(2020). Multichannel analysis of surface waves Glacial Maximum. Quaternary Science Reviews, 151,
(MASW) for offshore geotechnical investigations. 198–211. https://doi.org/10.1016/j.
Engineering Geology, 272, 105649. https://doi.org/ quascirev.2016.08.034
10.1016/j.enggeo.2020.105649 Woods, M. A., Wilkinson, I. P., Leng, M. J., Riding, J. B.,
Lunne, T., Robertson, P. K., & Powell, J. J. M. (1997). Vane, C. H., Lopes dos Santos, R. A., Kender, S., De
Cone Penetration Testing In Geotechnical Practice. Schepper, S., Hennissen, J. A. I., Ward, S. L.,
Blackie Academic & Professional. Gowing, C. J. B., Wilby, P. R., Nichols, M. D., &
McCabe, A. M., Clark, P. U., & Clark, J. (2005). AMS Rochelle, C. A. (2019). Tracking Holocene palaeostrati
C dating of deglacial events in the Irish Sea Basin and fication and productivity changes in the Western Irish
other sectors of the British – Irish ice sheet. Quaternary Sea: A multi-proxy record. Palaeogeography, Palaeo
Science Reviews, 24, 1673–1690. https://doi.org/ climatology, Palaeoecology, 532, 109231. https://doi.
10.1016/j.quascirev.2004.06.019 org/https://doi.org/10.1016/j.palaeo.2019.06.004
McGrath, T., Long, M., O’Connor, P., Trafford, A., & Yuan, F., Bennell, J. D., & Davis, A. M. (1992). Acoustic
Ward, D. (2016). Multichannel Analysis of Surface and physical characteristics of gassy sediments in the
Waves (MASW) Offshore Geotechnical Investigations. western Irish Sea. Continental Shelf Research, 12(10),
In A.-M. & K. Lehane (Ed.), Proceedings of the 5th Int. 1121–1134.
350
© 2022 the Author(s), ISBN 978-1-032-31259-0
Cone factor from CPTU tests in very soft clays at the east of Mexico’s valley
J.M. De La Rosa R. & F.A. Flores López
Ingenieros Geotecnistas Mexicanos. S.C., México.
ABSTRACT: From the results obtained during an exploration campaign, the correlation between CPTU resist
ance and the undrained shear strength of soft clays at the valley of Mexico, is evaluated in this paper. This cam
paign consisted of 50 piezocone penetration tests, 50 soil sampling borings and laboratory tests carried out on more
than 300 undisturbed soil samples recovered along 10 square kilometers. The samples were tested by means of
unconfined compression and unconsolidated undrained triaxial compression tests. The correlation of the determined
undrained shear strength and the experienced CPTU results was evaluated and the empirical cone factor (Nkt) was
determined for all tests. The main innovation lies in the fact that for the eastern zone of the Valley of Mexico, the
geotechnical characterization is not abundant and there are information gaps for this and other design parameters.
Keywords: CPTU, Nkt, Undrained shear strength, Valley of Mexico
1 INTRODUCTION 2 BACKGROUND
1.1 Origins of Mexico’s valley 2.1 Empirical cone factor

Mexico City is located on a lacustrine plain in the The many existing formulations to correlate Cone
Basin of Mexico that occupies an approximate area of penetration tip resistance (qc) from CPT/CPTU, results
9,600 km2 and is located at 2,250 m above sea level. with the undrained shear strength of clays (Su) are
This basin remained open (exoreic) until 700,000 years summarized in many publications. Most of those inves
ago, when extensive volcanic activity formed the Sierra tigations found Su linearly proportional with the cor
de Chichinautzin, which closed the basin (Mooser, rected tip resistance of the cone as shown in
1963) and obstructed drainage to the Balsas River. For equation (1):
this reason, the water was stored and gave rise to sev
eral lakes. At the foot of the mountains and due to the
sudden change in the slopes of the rivers, large alluvial
deposits of highly variable composition and cross-
stratification are located, evidence of an erosive
dynamic due to periods of intense rain. This area of the Where,
city was enabled as a flood control zone through the Su: Undrained shear strength.
creation of artificial lagoons causing surface depres qc: Tip resistance of the cone/piezocone.
sions by pumping. Texcoco Project (1969). The soils σv0: Total overburden pressure.
found in this zone correspond to very soft and saturated Nkt: Empirical cone factor.
lacustrine clays, subject to a regional subsidence pro
cess due to the extraction of water by deep pumping.
2.2 Previous research and typical values
The use of CPTU is a common option to know the
stratigraphic sequence for the determination of settle- Most of the research aimed at estimating the value of
ments magnitude. (Nk) that correlates CPT/CPTU with Su. Variation
Currently, the urban development of the city has ranges has been estimated between 11 and 19 by
been pushing towards the east, driven by a new airport Lunne & Kleven (1981) for normally consolidated
under construction which was canceled three years Scandinavian marine clays, also Jorss (1998) estimated
ago. However, during its execution this project trig- values ranging from 15 to 20 for marine y boulder
gered the infrastructure needs of the area, so it is neces- clays, respectively. Mexico City’s clays present Nk
sary to expand the knowledge about its geotechnical values between 5 and 11, according to Santoyo et al.
characterization. This situation makes this publication (1989). Some values range between 5 and 28 for
necessary and useful. different types of soils, according to Chen (2001) and
DOI: 10.1201/9781003308829-47
351
Gebre-selassie (2003). Also, Ovando (2011) related The exploratory drilling was carried out accord
Su = qc/Nk with Nk values of 13 for Mexico City. Typ ing to the corresponding ASTM standard, in the
ically, Nk is between 10 and 18, with an average of 14 case of mixed SM drilling, the SPT technique was
(Robertson & Cabal, 2015). The marine clays of the combined with the selective recovery of undis
Gulf of Mexico present Nk values between 12 and 20, turbed samples using a thin-walled tube (Shelby)
according to Cruz et al. (2019). and the Electronic Friction Piezocone Test was
performed using a digital equipment that complies
with the ASTM D5778-20 standard and that
2.3 Geotechnical input information
allows the measurement of resistance to tip pene
tration (qc), lateral friction (fs) and pore pressure
2.3.1 Field exploratory survey behind de cone (u2).
Based on the dimensions of the project structures,
a geotechnical exploration campaign was executed to
define the stratigraphic profile, from which the labora
tory program was planned. The exploration campaign
consisted of boreholes between 25 and 50 m deep
(SPT/SHB mixed, SM and piezocone tests, CPTU), in
addition to a surface exploration campaign, consisting
of 150 open-pit wells (PCA), 100 plate load Tests (PP)
and strategically located piezometric stations.
Figure 2. Location of CPTU tests.
2.3.2 Laboratory tests

All the samples obtained in the exploration work
were macroscopically identified and classified
both in the field and in the laboratory, according
Figure 1. Location of SM boreholes. to the Unified Soil Classification System. Tests
352
were carried out to determine the following index
properties: Natural water content (150 to 400%).
Limits of consistency (liquid limit, plastic limit)
and plasticity index (100 to 300%). Grain size
analysis. Percentage of fines (90 to 100%). Rela
tive specific weight and density of solids (2.30 in
average, maximum of 2.65 and minimum
of 2.15).
The mechanical properties were determined based
on the following tests: Unconfined compression
shear resistance. Triaxial compression test UU with
determination of Young’s Modulus (E). Triaxial
compression test CU. Compressibility in one-
dimensional consolidation tests.
2.4 Analysis of parameters

The fulfillment of the objective for geotechnical
characterization is not easy due to the variability
of the results obtained in the field and laboratory
tests. Another difficulty that causes uncertainty
in representative soil results is the considerable
size of the study area and the relatively large
distances between boreholes and measurement
points. To achieve a satisfactory description of
the properties and stratigraphic characteristics of
the subsoil, the use of descriptive statistics tools
was used. In the next section, the analysis cri
teria and selection of parameters used are
presented.
2.5 Criteria for quantitative analysis and Figure 3. Histogram of simple and triaxial compressive
variability of the parameters strength UU and its normal probability function.
By having a soil property measured repeatedly,

we get a data set called a “data sample” that
can be analyzed quantitatively using statistical mostly obtained the number of blows N equal to
parameters. For the values obtained in the pro zero, which does not provide relevant data for
ject, although there are many parameters that characterization purposes. The histogram of all
can be estimated, the following were evaluated: the correlated data was generated and the
Mean, median, mode, variance, standard normal logarithmic probability distribution
deviation. function that best fit the measured values was
The analysis of variability of the soil proper obtained (Figure 3). Figure 4 shows the results
ties was graphically complemented by means of obtained by correlating the tip resistance with
a histogram calculating its corresponding prob the laboratory results for the UU resistance of
ability distribution in order to select the data the soils, where a linear increase can be seen
to be considered to describe the geotechnical as a function of depth, it is also observed that
parameters. there is more variation near the soil surface
associated with the surface crust.
Figure 5 shows the average values of Undrained
3 RESULTS OBTAINED Resistance obtained in the laboratory and Tip
Penetration Resistance, obtained on site for all
3.1 Mechanical properties samples.
Based on the analysis of variability and the
selection of representative values, the results of
the CPTU tests on site and the laboratory results 3.2 Cone factor values
were correlated. For this project, the results Once the corresponding analyzes had been
obtained from the electrical cone (tip resistance) carried out and the results obtained in the field
are more representative, since the SPT tests and laboratory correlated, the average values of
353
Figure 4. Undrained resistance obtained by correlating
laboratory tests and electrical cone results.
Figure 6. Average Cone Factor Nkt, depending on the

depth range analyzed.
Figure 7. Average Cone Factor Nkt, for each CPTU

analyzed.
Figure 5. Average values of Soil Undrained Resistance

(Su) and Tip Penetration Resistance along the explored
depth for all samples. 4 DISCUSSION AND CONCLUSIONS
4.1 Variability of results

Nkt, shown in Figure 7, were obtained for the The results obtained during the analysis for this
38 most representative CPTU. The results research, indicate that in 87% of the cases, the cone
obtained range between 8 and 21 with an aver factor Nkt is between 10 and 20, with an average of
age of 14; while Figure 6 shows the average 14 of the 12,500 data analyzed, the maximum value
Nkt values obtained by depth range every five found was 25 and the minimum 7. Figure 6 shows
meters, also for the 38 most representative a progressive variability of the Nkt factor as a function
CPTU. of depth during the first 25 meters of depth, with
354
values between 11 and 16, while for greater depths similar standards, they do not usually include
a stabilization around values higher than the average is temperature and verticality sensors, which have
observed (16 to 19). Figure 8 shows the variability ana proven necessary to guarantee the quality of the
lysis between the proposed cone factor Nkt average results obtained.
and the results from the direct correlation of the Tip These trends, point to a less widespread use of
Resistance (qc) and the laboratory UU resistance (Su). the electric cone due to the costs involved in
Excluding the values obtained near the surface, due to this modernization, as well as more accurate
what is mentioned in section 3.1, an R = 86% is results in cases where this technique is applied.
obtained in the adjusted regression line. However, these improvements are unlikely to
replace or lessen the need for laboratory testing,
while the main objective of using the electric cone
will remain unchanged: to determine the strati
graphic sequence in previously explored areas
where the stratigraphy is partially known in
advance.
ACKNOWLEDGMENTS
In this section we express our gratitude to the
geotechnical team of the company CONSULTORIA
INTEGRAL EN INGENIERÍA, who carried out
much of the supervision of the field and laboratory
Figure 8. Nkt average/correlated vs depth. work, as well as the collection and primary processing
of the information obtained. Also, in a special way to
the COMISION NACIONAL DEL AGUA, as the
The variability of Nkt shown in Figure 6 with government dependency responsible for the project
respect to depth is due in part to clay samples from which the analyzed data was obtained.
obtained with water contents between 50% and
450%. For extreme cases, sample recovery becomes
challenging and results in non-representative sam
ples. This variability as function of depth is a typical REFERENCES
behavior for marine clays, and the lacustrine nature Carrillo, N. et al. 1969. Proyecto Texcoco. Secretariat of
of the virgin, high-salinity clays to the east of the Finance and Public Credit of Mexico.
valley is consistent with this trend. Chen, C. 2001. Evaluating undrained shear strength of
Klang clay from Cone penetration test, In: International
Conference on In Situ Measurement of Soil Properties
4.2 Future lines of research and Case Histories, Parahyangan Catholic University of
In addition to the results obtained so far, there is suffi Indonesia.
cient data to establish correlations between Su and Cruz, D. et al. 2019. Los suelos marinos de la Sonda de
Campeche y Litoral Tabasco, Golfo de México. XVI
Index properties. Data consist of Atterberg limits from
Panamerican Conference on Soil Mechanics Geotech
500 samples and elastic Young’s modulus from 200 nical Engineering in the XXI Century: Lessons learned
samples. and future challenges.
Gebreselassie, B. 2003. Experimental, analytical and
numerical investigations of excavations in normally
4.3 Current trends in the valley of Mexico consolidated soft soils, PhD thesis, University of Kassel.
The current trends among the predominant players Jörß, O. 1998. Erfahrungen bei der Ermittlung von cu-
Werten mit der Hilfe von Drucksondierungen in bindi
of Mexican geotechnics can be observed with
gen Böden, Geotechnik, 21.
a tendency to promote the use of automated equip Lunne, T. & Kleven, A. 1981. Role of CPT in North Sea
ment to control the driving speed of the tip with the Foundation Engineering, In: Symposium on Cone Pene
valid argument that there is some relevant variation tration Engineering Division, ASCE.
between the parameters obtained in an automated Mooser, F. 1963. Historia tectónica de la Cuenca de
way by comparing them with those derived from México: Boletín de la Asociación Mexicana de Geólo
rudimentary field estimates, such as the most gos Petroleros.
common practice of drawing division marks of pre Ovando, E. 2011. Some geotechnical properties to charac
determined length on the drilling rods and maneu terize Mexico City Clay, 14th Pan-American Conference
on Soil Mechanics and Geotechnical Engineering.
vering the drill so that the guide adapts to the
Robertson, P. & Cabal, K. 2015. Guide to Cone Penetration
required speed of penetration. Testing for Geotechnical Engineering. Gregg drilling.
In addition, the application spaces for nation 6th edition.
ally produced equipment have decreased Santoyo, E. et al. (1989). El cono en la exploración geotéc
because, although they comply with ASTM or nica. México: TGC Geotecnia.
355
© 2022 the Author(s), ISBN 978-1-032-31259-0
Piezocone testing in Nordic soft clays: Comparison of high-quality databases

M. D’Ignazio & B. Di Buò
Ramboll Finland Oy, Tampere, Finland
T. Länsivaara
J.-S. L’Heureux & P. Paniagua

Norwegian Geotechnical Institute, Trondheim, Norway
J. Selänpää
Destia Oy, Tampere, Finland
ABSTRACT: Soft and sensitive clays are widespread in Scandinavia. Piezocone correlations for Norwe
gian clays have been previously proposed based on high-quality block samples from several sites.
Recently, a large database of Finnish soft clays was compiled by Tampere University from piezocone
measurements as well as high-quality laboratory tests on specimens from large tube samples. Finnish and
Norwegian clays exhibit some differences in terms of basic properties. Norwegian clays show lower water
content, lower organic content, higher silt content and lower plasticity than the clays from Finland. This
may be linked to the source of the materials, their depositional and post-depositional processes that in turn
impact on the mechanical behaviour.
This paper aims to compare piezocone Norwegian and Finnish data with focus on strength and stress history.
The database trends are compared for relevant engineering parameters. The data and its variability are critically
discussed considering differences in geological history, basic properties, sampling techniques and disturbance.
1 INTRODUCTION carried out studies on CPTU testing and correlations

in Finnish soft clays (Di Buò et al. 2016, Di Buò et al.
Piezocone (CPTU) testing is one of the most reliable 2018, Selänpää et al. 2018, Di Buò 2020, Di Buò
site investigation tools in soft clays, as it has proven et al. 2020, Selänpää 2021). The calibration database
repeatability of measurements and can be used to includes testing on large tube samples retrieved with
detect soil layering and derive a large number of a Laval-type tube sampler developed by Tampere
geotechnical parameters (e.g., Lunne et al. 1997a, University (Di Buò et al. 2019). Both the Norwegian
Robertson 2009, Di Buò 2019). During cone pene and the Finnish databases include laboratory index
tration, the standard CPTU equipment measures tests, constant-rate-of-strain (CRS) oedometer tests,
cone tip resistance qc, sleeve friction fs and excess undrained triaxial compression (CAUC) and exten
pore pressure above the cone tip u2. sion tests (CAUE) and direct simple shear
The use of CPTU is widespread in Norway, regard tests (DSS).
less of the project size. CPTU correlations for Norwe The scope of this paper is to compare the Nor
gian clays derived from high-quality laboratory data wegian and Finnish clay data to better understand
on block samples (Karlsrud et al. 2005, Paniagua how the basic clay properties impact on the CPTU
et al. 2019) have been in use for nearly two decades. response in these clays. The database trends are
On contrary, CPTU testing is still on the rise in Fin compared for two key engineering parameters, i.e.
land, where weight soundings, dynamic penetration undrained shear strength and overconsolidation
testing and field vane testing are among the most pre ratio. The observations in terms of cone factors
ferred site investigation tools. This is mainly and calibration coefficients are critically discussed,
a consequence of the lack of local correlations as well considering differences in basic properties, geo
as limited experience by the operators. Nevertheless, logical history, sampling techniques and features of
in recent years the Tampere University (TUNI) has the databases.
DOI: 10.1201/9781003308829-48
356
2 GEOLOGICAL HISTORY OF NORDIC CLAYS 3 EVALUATION OF ENGINEERING
PROPERTIES OF CLAY FROM CPTU
A detailed study on the geological formation of
fine-grained sediments in Finland was conducted by 3.1 CPTU parameters
Gardemeister (1975). Fine-grained soil sediments in
Finland originated in the late Pleistocene, during the Engineering properties of clays are derived from
retreat of the continental ice sheet in the Weichselian both measured and normalized CPTU parameters.
ice age (11,700 years ago). The entire Scandinavian Among these:
region was covered by a large ice sheet named Fenno- • Normalized cone resistance Qt=(qt-σv)/σ’v
Scandian that spread out from the Scandinavian Moun • Normalized excess pore pressure Qu=(u2-u0)/σ’v
tains to Northwest Russia, UK, and The Netherlands. • Normalized effective cone resistance Qe=(qt-u2)/σ’v
The stratigraphy of Finnish soil deposits is the result of • Pore pressure ratio Bq=(u2-u0)/(qt-σv)
a series of processes that occurred at the end of the last
glacial period when the Fenno-Scandian ice sheet where qt is the corrected cone tip resistance, σv is the
retreated and during the Holocene (i.e. c. last 10,000 total overburden vertical stress, σ’v the vertical
years ago). The glacier meltwater accumulated between effective stress, u2 the pore pressure measured above
the front of the ice sheet and the southern shores, the cone tip, u0 the in-situ pore pressure.
giving rise to what currently is the Baltic Sea. This area In addition, (qt-σv), (u2-u0) and (qt-u2) are com
underwent four environmental stages in the postglacial monly referred to as qnet, Δu and qe, respectively.
progression of the Baltic basin, known as Baltic Ice
Lake, Yoldia Sea, Ancylus Lake, and Littorina Sea. 3.2 Overconsolidation ratio
(Gardemeister 1975)
Despite the extensive studies conducted on this The over-consolidation ratio (OCR) is defined as the
topic, the process of the formation of the Baltic Sea is ratio of effective preconsolidation stress σ’p and
not completely clear. Its connection with the Atlantic the vertical effective stress σ’v. Several authors verified
Ocean during the different phases made the salinity the dependence of σ’p and OCR on the cone tip resist
vary with location, depth, and time. The complex ance and excess pore pressure parameters, with σ’p
origin and development of this area may explain the and OCR increasing with increasing qnet, σu, qe
different geotechnical properties characterizing the clay and Qt, Qu, Qe respectively. (e.g. Chen and Mayne
deposits located in Finland and Sweden compared with 1996, Lunne et al. 1997, D’Ignazio et al. 2019).
the Norwegian ones. The Baltic Ice Lake originated In practice, the relationship between σ’p and qnet
during the retreat of the Weichselian glacier, when is the most used (Equation 1). The relationship
meltwater accumulated and formed a freshwater lake. between OCR and Qt is used in the same way as
At this stage, the connections with the North Sea and (Equation 2):
the Atlantic Ocean were closed because the ground on
the entire depression rose faster than the sea level.
However, a short connection with the sea across central
Sweden occurred during the Yoldia Sea stage. At the
early stage, the depositional environment was still char
acterized by low salinity owing to the heavy water flow
from the continental ice sheet. The salinity increased
after 200 years of the ingression of salt water, creating where k is an empirical parameter. Similar equations
the condition for a brackish depositional environment. are found in the literature for Δu, qe and Qu, Qe. An
Afterward, the isostatic uplift of the Baltic basin closed average value of k ≈ 0.32 is suggested by Chen and
the connection with the Atlantic Ocean and the Yoldia Mayne (1996) based on statistical analysis of piezo
Sea turned into Ancylus Lake. This stage lasted until cone-oedometer data involving a variety of different
a new connection with the North Sea was established clays. D’Ignazio et al. (2019) found k in the range
owing to the continuous rising of the water level of 0.15-0.5 for clays with OCR ≈ 1-5. Paniagua et al.
Ancylus Lake, forming the Littorina Sea. Finally, the (2019) found k = 0.20-0.75 for Norwegian clays,
continuous land rise made the connection with while Di Buò (2019) suggested k = 0.28 with coeffi
the Ocean shallower, thus creating the conditions for cient of variation (COV) ≈ 0.1 for Finnish soft clays.
the formation of the current Baltic Sea, which is char
acterized by brackish water. (Gardemeister 1975)
It is evident that the combination of the sea water 3.3 Undrained shear strength
intrusion and freshwater flow from the melting glacier
created a heterogeneous depositional environment The net cone resistance qnet is related to the undrained
characterized by variable salinity content. Although the shear strength su by means of the cone factor Nkt as:
salt leaching process is considered as the main factor
explaining the high sensitivity of Scandinavian marine
clays, further studies are needed for Finnish clays.
357
Similarly, Δu and qe are related to su by means of (St = 2–240). The OCR ranges from 1 to 6, while the
the cone factors NΔu and Nke respectively as: clay content varies between 21 and 65%.
4.2 Finland
The Finland TUNI’s CPTU database in summarized
in the works carried out by Di Buò (2019), with
focus on preconsolidation stress, and Selänpää
(2021), with focus on undrained shear strength. Data
For low OCR offshore and onshore clays, Low was collected from 5 test sites located in Southern
et al. (2010) reported Nkt = 8.6-15.3 and NΔu = Finland. Both CPTU and laboratory measurements
3.3-8.8 for triaxial compression (CAUC) and Nkt = are available from the test sites.
11-20 and NΔu = 4.8-11.9 for field vane test (FVT). The laboratory tests were carried out on speci
Paniagua et al. (2019) found Nkt = 5-16 and NΔu = mens obtained from large diameter (Ø132 mm)
5-10 for CAUC in onshore Norwegian clays with Laval-type sampled designed by Tampere University
OCR less than 6. Paniagua et al. (2019) further and presented in detail by Di Buò et al. (2019).
observed Nke ≈1,5-10 for CAUC decreasing with Some of the tests were performed on specimens
increasing Bq. Selänpää (2021) suggested Nkt = 9.1, obtained from a mini-block (Ø150 mm) Sherbrook
NΔu = 7.7, and Nke = 4.5 with COV ≈ 0.1 for CAUC sampler (Emdal et al. 2016).
in Finnish soft lightly overconsolidated clays. The database includes laboratory index tests, n.99
D’Ignazio & Lehtonen (2021) observed NΔu ≈ 11 for constant-rate-of-strain (CRS) oedometer tests, n. 17
FVT in a soft organic sulphate rich lightly overcon anisotropically consolidated undrained triaxial com
solidated clay from Finland. D’Ignazio et al. (2020) pression (CAUC) and n. 14 extension tests (CAUE),
found Nkt = 20-32 (CAUC) for an overconsolidated n. 14 direct simple shear tests (DSS) and n. 14 field
North Sea clay with OCR ≈ 4-20, with Nkt increasing vane tests (FVT).
with increasing OCR. Sample quality was assessed according to the
Lunne et al. (1997b) criterion based on the relative
void ratio change at reconsolidation Δe/e0 from
4 HIGH-QUALITY CLAY DATABASES CAUC as well as CRS tests. CAUC data points
from Ø132 mm samples fall within sample
4.1 Norway quality categories “Very good to excellent”, while
the CRS data points and fall within “Very good to
The Norway CPTU database is presented in detail excellent” and “Good to fair” sample quality
by Paniagua et al. (2019). The database consists of categories.
61 high-quality block samples data points collected Soil properties were measured from specimens
from 17 Norwegian clay sites located all over the collected down to a maximum depth of 9 m. The
country. For these points, both CPTU and laboratory clay properties cover a plasticity index Ip varying
measurements are available. The laboratory tests between 16 (low-medium plastic) and 59 (very
were carried out on specimens obtained from large high plastic), a wide range of water content (w = 66–
diameter (Ø250 mm) block samples and mini-block 127%), a wide range of sensitivity (St) values
(Ø150 mm) Sherbrook samples (Emdal et al. 2016). (St = 16–98). The OCR ranges from 1 to 2, while the
For Norwegian clays, block samples seem to ensure clay content varies between 40 and 100%.
higher quality than the more traditional Ø54 mm or
Ø72 mm piston samples (e.g., Lunne et al. 2006,
L’Heureux et al. 2018). 5 COMPARISON OF CPTU DATABASES OF
The database includes laboratory index tests, con- NORWEGIAN AND FINNISH CLAYS
stant-rate-of-strain (CRS) oedometer tests, anisotrop
ically consolidated undrained triaxial compression
5.1 Overconsolidation ratio
(CAUC) and extension tests (CAUE) and direct
simple shear tests (DSS). Figures 1, 2 and 3 show plots of OCR versus the nor
Sample quality was assessed according to the malized values of cone resistance Qt, pore pressure Qu
Lunne et al. (1997b) criterion based on the normal and effective cone resistance Qe. The plots suggest
ized change in void ratio Δe/e0 from CAUC tests. a well-defined trend line for the Finland data, while the
Data points fall within sample quality categories Norway data appears to be characterized by higher
“Very good to excellent” and “Good to fair”. scatter.
Soil properties were measured from specimens Even though the uncertainty associated with OCR
collected down to a maximum depth of 22 m. The for Finnish clays appears to be lower than that for the
clay properties cover a wide range of plasticity Norwegian clays, it must be noted that the Finland
index, with Ip varying between 4 (low plastic) and data is obtained from 5 sites all located in the South
49 (very high plastic), a wide range of water content ern part of the country; while the Norwegian data is
(w = 28–72%), a wide range of sensitivity (St) values collected from 17 sites spread all over the country,
358
where areas might have undergone different geo
logical histories. However, the OCR range of the
Finland data (OCR = 1-2) is lower than that of
the Norway data (OCR = 1-7). For OCR < 2, the
mean trends of the two databases appear to be
consistent, despite the larger scatter in the Norway
data. Consistency is also found between the OCR
based on mini-block samples and Ø132 mm sam
ples of Finnish soft clays.
The coefficient k of Equation 2 is ≈0.3 in
Figure 1 for Finnish clays. Such a value is often
assumed in practice in absence of site-specific
oedometer tests.
Figure 3. OCR vs Qe.
Figure 1. OCR vs Qt.
Figure 4. K = OCR/Qt vs plasticity index.
5.2 Undrained shear strength

The undrained shear strength from triaxial CAUC
tests (suC) is taken as a reference and used for com
parison of the Finland and Norway databases.
Figures 5, 6 and 7 illustrate the variation of suC
with qnet, Δu and qe respectively. The suC from Finn
ish clay deposits is generally lower than that of Nor
wegian clays. It must be noted that the sampling
depths are limited to 9 m below ground level for the
Finland data, while depths reach up to 22 m for
the Norway data. Therefore, the deeper samples in
Figure 2. OCR vs Qu. the Norway database experienced larger stress relief
than the samples from Finland. Nevertheless, for the
depth range 0-9 m, suC of Norwegian clays is in the
Data in Figures 1, 2 and 3 show a remarkable vari range 26-62 kPa, while suC of Finnish clays is 12-24
ability of the calibration coefficients, particularly for kPa. One possible explanation for this is the variation
the Norwegian database. Figure 4 shows the variation of OCR with depth. As shown in Figure 8, the range
of k (=OCR/Qt) with plasticity index (Ip). The coeffi of OCR is larger for the Norwegian data. Given that
cient k appears to vary as a function of plasticity, su increases with increasing OCR (e.g. D’Ignazio
decreasing with increasing Ip. Such a trend could not et al. 2016, Selänpää 2021), this explains the higher
be observed when treating the two databases separ strength of Norwegian clays at shallow depths.
ately. Nevertheless, for Ip < 20%, the data from Fin The cone factor values for Norwegian and Finnish
land is consistent with measurements at larger Ip. clays are discussed in detail by Paniagua et al. (2019)
359
and Selänpää (2021). Cone factors Nkt, NΔu and Nke
present some variability as illustrated in Figures 5, 6
and 7. As for the OCR, the variability appears to be
the lowest for the Finland data. The variability of cone
factors for Norwegian clays was discussed by Karlsrud
et al. (2005) and Paniagua et al. (2019). These studies
suggested a dependency of cone factors on OCR
and lightly on index properties, although the proposed
correlations are characterized by a non-negligible
scatter.
Figure 9 presents the variation of the normalized
CAUC strength suC/σ’v with OCR. In agreement with
previous analyses, the Finland data shows overall
a well-defined trend and a lower scatter. For OCR
< 2, the normalized suC/σ’v of Finnish clays appears
to be higher than that of the Norwegian clays, with Figure 7. Qe vs suC.
the Miniblock samples data from Finland showing
slightly higher values than the Laval Ø132 mm
Karlsrud & Hernandez-Martinez (2013) and
samples. This may result from the fact that the lower
Paniagua et al. (2019) discussed how the variabil
OCR points in the Norway database are observed
ity of Norwegian clays in Figure 9 may be cap
at greater depths than in the Finnish database
tured by the variability of water content.
(Figure 8). This may be an indicator of the stress
Considering that the Finland data is characterized
dependency of the normalized shear strength of clays.
by higher water content, this may further support
the results of these studies.
Figure 5. Qqnet vs suC.

Figure 8. OCR vs depth.
Figure 6. Δu vs suC. Figure 9. suC vs OCR.
360
6 DISCUSSION AND CONCLUSIONS Di Buò, B., Selänpää, J., Länsivaara, T. & D’Ignazio, M.
2019. Evaluation of sample quality from different sampling
This study has compared two high-quality clay data methods in Finnish soft sensitive clays. Canadian Geotech
bases from Finland and Norway. Both databases nical Journal, 56(8), 1154–1168.
include piezocone (CPTU) measurements and labora Di Buò, B., D’Ignazio, M., Selänpää, J., Länsivaara, T., &
tory testing from different sites. Laboratory specimens Mayne, P. W. 2020. Yield stress evaluation of Finnish
were mainly retrieved from large diameter block sam clays based on analytical piezocone penetration test
(CPTu) models. Canadian Geotechnical Journal, 57(11),
pler (Norway) and large diameter Laval-type tube sam 1623–1638.
pler (Finland). Constant-rate-of-strain (CRS) D’Ignazio, M., Phoon, K.K., Tan, S.A., & Länsivaara, T.T.
oedometer tests results and undrained triaxial compres 2016. Correlations for undrained shear strength of Finnish
sion (CAUC) tests were used as reference tests to com soft clays. Canadian Geotechnical Journal, 53(10),
pare the overconsolidation ratio (OCR) and undrained 1628–1645.
shear strength from triaxial compression tests (suC) and D’Ignazio, M., Lunne, T., Andersen, K. H., Yang, S., Di
their trends with respect to CPTU parameters. Buò, B., & Länsivaara, T. 2019. Estimation of preconsoli
The study showed that the shear strength of Finn dation stress of clays from piezocone by means of
high-quality calibration data. AIMS Geosciences, 5(2);
ish clays is lower than that of Norwegian clays. This
104–116.
is mainly due to the stress histories of the soils, i.e., D’Ignazio, M., Andersen, K. H., Engin, H. K.,
the OCR of Finnish clays is generally lower when Sivasithamparam, N., Jostad, H. P & Yetginer, G. 2020.
comparing samples from the same depth interval. Interpretation of piezocone tests in overconsolidated clays
The larger scatter in both OCR and suC observed for using SHANSEP. International Symposium on Frontiers in
the Norwegian data might be related to the fact that the Offshore Geotechnics (ISFOG) 2020, Austin, Texas
Finnish data is collected from 5 sites all located in D’Ignazio, M., & Lehtonen, V. 2021. Using SHANSEP for
Southern Finland, while the Norwegian database com verification of unreliable piezocone data in clays. In IOP
prises 17 sites spread all over Norway, where areas Conference Series: Earth and Environmental Science (Vol.
might have experienced different depositional histories 710, No. 1, p. 012011). IOP Publishing.
Emdal, A., Gylland, A., Amundsen, H.A., Kåsin, K., &
and relative fall of sea level. Furthermore, the Finnish Long, M. 2016. Mini-block sampler. Canadian Geotech
data has been collected using the same piezocone and nical Journal, 53(8), 1235–1245.
sampling equipment for all sites, while the Norwegian Gardemeister, R. 1975. On engineering-geological properties
data has been acquired over a period of 30+ years of fine-grained sediments in Finland. Dissertation. Tech
using different equipment. Even though data has been nical Research Center of Finland, Building Technology
collected according to best practice, differences in and Community Development, Publication 9. 91 p.
measured CPTU parameters may be observed when Karlsrud, K., Lunne, T., Kort, D.A., & Strandvik, S. 2005.
using different cones (Lunne et al. 2018). CPTU correlations for clays. In Proceedings of the inter
national conference on soil mechanics and geotechnical
Future research shall focus on the mineralogy of
engineering, Vol. 16 (2), AA Balkema.
clays. Mineral composition can have a large effect on Karlsrud, K., & Hernandez-Martinez, F.G. 2013. Strength and
the flocculation potential and building of strength of deformation properties of Norwegian clays from laboratory
the soil (e.g. Torrance 2014). Depositional and post tests on high-quality block samples. Canadian Geotech
depositional processes for both countries are also nical Journal, 50(12), 1273–1293.
important aspects to investigate in depth. For a more L’Heureux, J. S., Gundersen, A. S., D’Ignazio, M.,
thorough comparison, it is recommended to add data Smaavik, T., Kleven, A., Rømoen, M., … & Hermann, S.
from Central and Northern Finland to the Finnish 2018. Impact of sample quality on CPTU correlations in
database. clay—example from the Rakkestad clay. In Cone Penetra
tion Testing 2018 (pp. 395–400). CRC Press.
Low, H. E., Lunne. T, Andersen, K. H., Sjursen, M. A., Li, X.
& Randolph, M. F. 2010. Estimation of intact and
REFERENCES remoulded undrained shear strengths from penetration tests
in soft clays. Géotechnique, 60(11), 843–859.
Chen, B.S., & Mayne, P.W. 1996. Statistical relationships Lunne, T., Robertson, P.K. & Powell, J.J.M. 1997a. Cone
between piezocone measurements and stress history of penetration testing in geotechnical practice. London, Spon
clays. Canadian Geotechnical Journal, 33(3), 488–498. Press. 312p
Di Buò, B. 2020. Evaluation of the Preconsolidation Lunne, T., Berre, T., & Strandvik, S. 1997b. Sample disturb
Stress and Deformation Characteristics of Finnish ance effects in soft low plastic Norwegian clay. In Pro
Clays based on Piezocone Testing. PhD Thesis, Tam ceedings of the Conference on Recent Developments in
pere University, Tampere, Finland. ISBN 978-952-03 Soil and Pavement Mechanics, Rio de Janeiro, Brazil, Rot
1468-2 terdam: Balkema, 81–102.
Di Buò B., D’Ignazio M., Selänpää J. & Länsivaara T. 2016. Lunne, T., Berre, T., Andersen, K. H., Strandvik, S., &
Preliminary results from a study aiming to improve ground Sjursen, M. 2006. Effects of sample disturbance and con
investigation data. Proceedings of the 17th Nordic Geo- solidation procedures on measured shear strength of soft
technical Meeting: 187–197. marine Norwegian clays. Canadian Geotechnical Journal,
Di Buò, B., Selänpää, J., Länsivaara, T., & D’Ignazio, M. 43(7), 726–750.
2018. Evaluation of existing CPTu-based correlations for Lunne, T., Strandvik, S., Kåsin, K., L’Heureux, J. S.,
the deformation properties of Finnish soft clays. In Cone Haugen, E., Uruci, E., … & Kassner, M. 2018. Effect of
Penetration Testing 2018 (pp. 185–191). CRC Press. cone penetrometer type on CPTU results at a soft clay test
361
site in Norway. In Cone Penetration Testing 2018, Thesis, Tampere University, Tampere, Finland. ISBN 978
417–422, CRC Press. 952-03-2156-7.
Paniagua, P., D’Ignazio, M., L’Heureux, J. S., Lunne, T. & Selänpää, J., Di Buò, B., Haikola, M., Länsivaara, T., &
Karlsrud, K. 2019. CPTU correlations for Norwegian D’Ignazio, M. 2018. Evaluation of existing CPTu
clays: an update. AIMS Geosciences, 5, 82–103. based correlations for the undrained shear strength of
Robertson, P.K. 2009. Interpretation of cone penetration tests soft Finnish clays. In Cone Penetration Testing 2018.
—a unified approach. Canadian Geotechnical Journal, 46 CRC Press.
(11), 1337–1355. Torrance, J. K. 2014. Chemistry, sensitivity and quick-clay
Selänpää, J. 2021. Derivation of CPTu Cone Factors for landslide amelioration. In Landslides in Sensitive Clays,
Undrained Shear Strength and OCR in Finnish Clays. PhD 15–24. Springer, Dordrecht.
362
© 2022 the Author(s), ISBN 978-1-032-31259-0
Prediction of resilient modulus of cohesive subgrade soils from CPTU data

using polynomial neural networks
Wei Duan
College of Civil Engineering, Taiyuan University of Technology, Taiyuan, China
School of Transportation, Southeast University, Nanjing, China
Zening Zhao
Guojun Cai*
School of Civil Engineering, Anhui Jianzhu University, Hefei, China
Anhui Wang
China Construction Industrial & Energy Engineering Group Co., Ltd., Nanjing, China
Ruifeng Chen
Anand J. Puppala
Zachry Department of Civil and Environmental Engineering, Texas A&M University, USA
Songyu Liu
Surya Sarat Chandra Congress

ABSTRACT: Although various empirical models have been proposed to predict the resilient modulus (Mr)
from laboratory test parameters, less effort has been made in developing reliable empirical models using
piezocone penetration test (CPTU) data. Moreover, the prediction accuracy of the existing empirical models is
not high enough. In the present study, a novel empirical model was proposed to predict the Mr from CPTU
data based on the polynomial neural network. To this end, a comprehensive database comprising 16 different
sites in Jiangsu province, China, was firstly compiled, which contains 124 sets of Mr, cone tip resistance (qc),
sleeve friction (fs), pore water pressure (u2), moisture (w), and dry density (γd) values at the in-situ stress con
dition. Taking the in-situ Mr values as reference values, three empirical models was developed using the
group method of data handling (GMDH) neural network. The results show that proposed GMDH model 3
(GMDH method) with the input parameters of qc, fs, w, and γd can accurately predict the Mr. The obtained
specific expressions for prediction of Mr further prove the reliability of the GMDH model. Overall, the new
GMDH method can more accurately predict the Mr of subgrade soil and guide engineering practice.
1 INTRODUCTION an important index that describes the nonlinear stress-

strain behavior of soil materials with respect to
The properties of subgrade soils are the key contents repeated traffic loading. The definition of Mr can be
in the design and performance of a pavement structure expressed as the ratio of the repeated deviator stress
(Liu et al., 2016; Ghorbani et al., 2020). Among these to the recoverable axial strain. The most usually used
properties of dry density, moisture content, stress method to determine such property of Mr is using the
state, the resilient modulus (Mr), and so on, the Mr is repeated load triaxial test with a closed loop servo
*Corresponding author
DOI: 10.1201/9781003308829-49
363
pneumatic loading system (AASHTO, 1993; NCHRP, Mr evaluation have been presented. A comparison of
2004). However, most of the results from laboratory the predictions from the existing empirical correl
tests generally run into difficulties: size limitations, ations are presented in detail. Salient findings and
disturbance, and actual maintenance effects (Duan conclusions are shown based the above results and
et al, 2019). Instead, the in situ tests can provide more analysis.
direct and accurate results, thus, several attempts were
made to determine and understand the behavior of
resilient modulus property based on different soil 2 GMDH ALGORITHM
properties (Mohammad et al. 2002; Tarawneh et al.
2014; Liu et al., 2016; Heidaripanah et al., 2017; The GMDH algorithm an inductive self-organizing
Ghorbani et al. 2018). approach to estimate black-box models with an
In fact, some empirical correlations between Mr unknown correlation between variables. The kernel
and other test indices such as undrained shear of GMDH is to build an analytical function in
strength, unconfined compressive strength, and Cali a feed-forward network based on a quadratic node
fornia Bearing Ratio (CBR) has been established. transfer function and use regression technique to
Additionally, the in situ testing indices are also very determine the coefficients of partial quadratic poly
well-received when related the indices to Mr. Among nomials (Duan et al. 2021a). Assume yi and ^ yi repre
these in situ tests, the cone penetration test (CPT), sent the actual output and predicted outputs of
especially for the piezocone penetration test (CPTU) GMDH neural network, respectively.
is treated as the most accurate method that have the For a given input vector xi = [xi1, xi2, …, xin]T,
advantage of near-continuous, repeatable measure i =1, 2, …, M, the mathematical relationship between
ments that provide a detailed profile of the subsur the input and the output variables of GMDH neural
face soil layers (Ghorbani et al. 2018). The past network can be simplify expressed by a system of
studies have confirmed that the Mr is a function of partial quadratic polynomials consisting of only two
CPT measured parameters (cone tip resistance (qc), variables (neurons) as
sleeve friction (fs)), moisture content (w), applied
stress and confining pressure (Mohammad et al.,
2007). several empirical correlations have been pro
posed that related to Mr to soil properties, stress
states, and CPT testing indices (Mohammad et al.
2002; Liu et al. 2016; Ghorbani et al., 2020). How
ever, the it appears that most prior studies suggested The coefficients of each quadratic function (G) can
the empirical correlations may be limited to site- be obtained for an optimal fit for the output in the
specific conditions due to rather selective and limited input-output data pairs. The optimal fit is based on the
in-situ data obtained from only a few sites. More least root mean square error (RMSE) described as:
over, the studies related to CPTU indices are very
limited (Liu et al. 2016; Ghorbani et al., 2020).
With recent advances in computational software
and hardware (Zhao et al., 2021a), the use of the arti
ficial neural network (ANN) method as a statistical
regression technique to approximate input-output
relationships in geotechnical engineering has gained The purpose of GMDH neural network is to deter
impetus for its ability to effectively deal with the mine the coefficient ai in Eq. (1) to minimize the
complex relationships in recent years (Juang et al., RMSE by regression techniques. A number of partial
1999; Zhao et al., 2021b). However, the major draw descriptions are formed based on all possibilities of
back of ANN is that the detected dependencies are obtaining the output variable from two independent
hidden in the ANN structure, and the correlations are variables to construct the regression polynomial that
often not expressed intuitively (Duan et al., 2021a). best fits the dependent observations (yi, i =1, 2, …,
The group method of data handling (GMDH) type M). The normal equations can be solved using the
neutral network is a polynomial neural network that least squares method as follows:
using a powerful identification technique to model
complex relationships between multiple variables
(Moayed et al., 2017). Hence, in the present study,
the GMDH model is proposed for predicting Mr
based on these parameters of cone tip resistance (qt),
sleeve frictional resistance (fs), dry density (γd) and where a = [a0, a1, a2, a3, a4, a5]T is the vector of
moisture (w). unknown coefficients of the quadratic polynomial in
In this present study, the feasibility of the GMDH Eq. (1); Y = [y1, y2, y3, …, yM]T is the vector of
algorithm to predict the Mr of cohesive subgrade output values from observation. The best estimation
soils has been conducted. Details of the development of the coefficients in Eq. (1) from the whole set of
of various GMDH-type network models for in-situ M data triples can be determined through the Eq. (3).
364
3 DATABASE including qc, fs, w, γd and Mr is compiled and used
for the following analysis.
3.1 Site description
Geotechnical investigations were conducted in
Jiangsu Province, China and sixteen (16) different
test sites are investigated and considered in this
study as shown in the Figure 1. The geological for
mations of Quaternary clays mainly contain Marine,
Yangtze River Delta, Floodplain of Long River,
Floodplain of Abandoned Yellow River, Lagoon of
Lixia River and Lagoon of Taihu Lake. Based on the
previous analysis, the subsoil profile in these test
sites mainly clay deposits.
Figure 2. Typical profiles of piezocone penetration test.
4 CONSTRUCTION OF GMDH MODELS
4.1 GMDH models development

The data points can be separated into three subsets in
the ratios of 7:2:1 for training, testing, and valid
ation. Actually, the subsets of testing and validation
are always merged into one. Thus, the data sets sep
Figure 1. Location of the testing sites.
arated with a ratio of 7:3 for the training and testing
(including validating), respectively. Of the 124 data
sets, 87 data points were randomly selected for the
3.2 Laboratory tests and field CPTU tests training while 37 data points were used only for val
Based on the soil samples from the testing sites, idating the developed models. In the present study,
laboratory tests were performed to obtained the w, γd four variables (qc, fs, w, γd were considered to dem
and Mr properties. The resilient modulus tests were onstrate the prediction ability of the GMDH net
conducted on the soil samples by repeated load triaxial work. Three categories are considered as follows: i)
tests. The CPTU soundings have been conducted at the w, γd; ii) qc, fs; iii) qc, fs, w, γd. The relative root
adjacent locations near the boreholes where the soil mean square error (RRMSE), absolute fraction of
samples were collected to ensure the predicted Mr variance (R2) are utilized for the performance evalu
value with the in-situ stress level. In the next section, ation of GMDH models.
the CPTU technique along with representative profile
is presented. 4.2 GMDH models results
CPTU field tests were conducted using a light
weight truck with a 20-ton-capacity hydraulic The values of the RRMSE, and R2 for each model
system, which is in accordance with international are listed in Table 1. It can be observed that all
standards (Lunne et al., 1997; ASTM D5778, 2012; developed GMDH models have shown relative
Duan et al., 2021b). The sizes of the CPTU probe good performance. Comparing the performance of
are as follows: The section area of the cylindrical model 1 using soil indices from laboratory tests, the
cone penetrometer is 10 cm2 with a tip angle of predicted performance has been improved using the
60°, the surface area of the friction cylinder is CPT parameters based on the results of model 2.
150 cm2. The penetration rate was set as 20 mm/s, The CPT parameters seems to be more useful in
and nearly continuous data were produced at inter predicting Mr values. From the perspective of per
vals of about 50 mm. Compared with CPT, the pore formance evaluation, model 3 was found to be the
pressure can be measured simultaneously and the best among other GMDH models. This is not sur
ground water table (GWT) can be identified based prise because the input parameters are comprehen
on CPTU test. Based on the test results, the GWT sive and GMDH has strong predictive ability. Thus,
value varied with the range of 0.4 m to the accuracy of predictions can be consistently
4.5 m. Figure 2 shows a typical profile of the improved if more indices are involved in the
SCPTU sounding. Thus, a database of 124 data sets GMDH model.
365
Table 1. Performance of GMDH models in predicting Mr
in testing dataset.
Model Inputs RRMSE R2
1 w, γd 0.5571 0.6896
2 qc, fs 0.4730 0.7762
3 qc, fs, w, γd 0.1314 0.9827
Based on the analysis, the specific equations are

presented subsequently.
Figure 3. Correlation of Mr with qc and fs using GMDH
1) Model 1 model 2.
2) Model 2
3) Model 3
Figure 4. Comparison between targeted Mr and predicted Mr.
Taking model 2 as an example, the targeted

values (scatter points in red) and predicted curved
surface values (colored) using Eq. (5) is shown in
Figure 3. It can be observed that the data points of
targeted values are closed to the curved surface and
they are uniformly distributed on both sides of the
predicted curved surface, indicating the ease and
practicability of the GMDH model using CPT
parameters (qc, fs). Thus, under the condition that the Figure 5. Variation in the relative errors obtained from
accuracy requirement is satisfied, model 2 can be GMDH method (Model 3).
chosen as an option for its simplicity, whereas model
3 is a better choice when high precision is required. GMDH model 3 (Figure 4). Compared with model 1
The points scatter of predicted (calculated by and 2, the predicted points by GMDH model 3 are
GMDH model 3 along with model 1 and model 2) closer (near 1:1 line) to those targeted values. Thus,
versus targeted ψ values has further demonstrated the GMDH model 3 was considered as the GMDH
the accuracy of resilient modulus evaluation using method for predict the resilient modulus of subgrade
366
soils. Overall, the predicted values by the GMDH Clay in Jiangsu, China. Transportation Geotechnics
models are smaller than the targeted values, which 20:100253.
will lead to conservative engineering design. Duan, W., Congress, S.S.C., Cai, G., Liu, S., Dong, X.,
Figure. 5 Plots the relative errors (defined as the Chen, R. & Liu, X. 2021a. A hybrid GMDH neural net
percentage difference between the GMDH predicted work and logistic regression framework for state
and the targeted values) for the training and testing parameter-based liquefaction evaluation. Canadian Geo-
sets. Observely, most of the GMDH estimations of the technical Journal online.
Duan, W., Congress, S. S. C., Cai, G., Puppala, A. J.,
data patterns for model 3 fell within ±12% of the tar Dong, X., & Du, Y. 2021b. Empirical Correlations of
geted values. Overall, it is feasible to apply GMDH Soil Parameters based on Piezocone Penetration Tests
method to evaluate the resilient modulus of subgrade (CPTU) for Hong Kong-Zhuhai-Macau Bridge (HZMB)
soils. Project. Transportation Geotechnics, 100605.
Ghorbani, B., Sadrossadat, E., Bazaz, J. B., &
Oskooei, P. R. 2018. Numerical ANFIS-based formula
5 CONCLUSIONS tion for prediction of the ultimate axial load bearing cap
acity of piles through CPT data. Geotechnical and
Geological Engineering, 36(4), 2057–2076.
In this study, the feasibility of GMDH polynomial
Ghorbani, B., Arulrajah, A., Narsilio, G., Horpibulsuk, S.,
neural networks in modeling a regional database & Bo, M. W. 2020. Hybrid Formulation of Resilient
containing Quaternary clay from different geologic Modulus for Cohesive Subgrade Soils Utilizing CPT
formations in Jiangsu province, China. Three Test Parameters. Journal of Materials in Civil Engineer
GMDH models using all or part of the parameters ing, 32(9), 06020011.
including Mr at the in-situ stress condition, CPTU Heidaripanah, A., Nazemi, M., & Soltani, F. 2017. Predic
indices (qc, fs), and laboratory indices (w, γd). tion of resilient modulus of lime-treated subgrade soil
The relative good prediction results are obtained using different kernels of support vector machine. Inter
from all GMDH models and it is confirmed that the national Journal of Geomechanics, 17(2), 06016020.
GMDH is capable of effectively capturing the non Liu, S., Zou, H., Cai, G., Bheemasetti, T. V., Puppala, A. J.,
& Lin, J. 2016. Multivariate correlation among resilient
linear relationships. The most significant advantage for modulus and cone penetration test parameters of cohe
GMDH is that specific expressions can be given in sive subgradesoils. Engineering Geology, 209, 128–142.
predicting geotechnical parameters. Among three types Lunne, T., Robertson, P.K. & Powell, J.J.M. 1997. Cone
of GMDH models, model 3 including qc, fs, w and γd Penetration Testing in Geotechnical Practice. CRC
is selected as the GMDH method for its high accuracy. Press, London.
Mohammad, L. N., Herath, A., Abu-Farsakh, M. Y.,
Gaspard, K., & Gudishala, R. 2007. Prediction of resili
ACKNOWLEDGEMENTS ent modulus of cohesive subgrade soils from dynamic
cone penetrometer test parameters. Journal of Materials
The majority of the work presented in this paper was in Civil Engineering, 19(11), 986–992.
funded by the National Key R&D Program of China National Cooperative Highway Research Program
(NCHRP), 2004. Guide for Mechanistic Empirical
(Grant No. 2020YFC1807200), the National Natural Design of New and Rehabilitated Pavement Structures.
Science Foundation of China (Grant No. 52108332, Part 2, Design Inputs. Final Rep. No. NCHRP 1-37A,
No. 41877231, and No. 42072299), and the Project Washington, D.C.
of Nanjing Construction System Science and Tech Tarawneh, B., & Nazzal, M. D. 2014. Optimization of
nology(Ks2153). The financial supports are greatly resilient modulus prediction from FWD results using
acknowledged. artificial neural network. Periodica Polytechnica Civil
Engineering, 58(2), 143–154.
Zhao, Z., Duan, W. & Cai, G. 2021a. A novel PSO-KELM
REFERENCES based soil liquefaction potential evaluation system using
CPT and Vs measurements. Soil Dynamics and Earth
American Association of State Highway and Transportation quake Engineering 150: 106930.
Officials (AASHTO). Guide for design of pavement Zhao, Z., Congress, S.S.C., Cai, G. & Duan, W. 2021b.
structures. Washington, D. C., 1993. Bayesian probabilistic characterization of consolidation
Duan, W., Cai, G., Liu, S., Puppala, A.J. & Chen, R. 2019. behavior of clays using CPTU data. Acta Geotechnica
In-Situ Evaluation of Undrained Shear Strength from 1–18.
Seismic Piezocone Penetration Tests for Soft Marine
367
© 2022 the Author(s), ISBN 978-1-032-31259-0
Effect of sand bio-cementation on cone tip resistance: A numerical study

M.El Kortbawi, K. Ziotopoulou & J.T. DeJong
University of California, Davis, CA, USA
D.M. Moug
Portland State University, Portland, OR, USA
ABSTRACT: Understanding the effect of soil cementation on cone measurements is important for the iden
tification of naturally cemented soil deposits and for the verification of soil improvement achieved by various
forms of artificial cementation, including bio-cementation. This paper presents the results of an effort to con
nect cone tip resistances with fundamental, constitutive level, bio-cementation-induced changes in soil behav
ior. To this end, a direct axisymmetric cone penetration model using the Mohr-Coulomb constitutive model
and grid rezoning and remapping algorithms is used to model cone penetration in bio-cemented sands. By
connecting the results of cone penetration simulations in a Mohr-Coulomb material to real data and estab
lished relationships, this work will guide selection of equivalent strength properties of these challenging
materials. More specifically, the apparent cohesion, peak friction angle, dilation angle, and small-strain shear
modulus within the Mohr-Coulomb constitutive model are varied parametrically across a reasonable range of
parameter values informed by past laboratory, bench-scale, and centrifuge tests. Results show that cone pene
tration resistance in bio-cemented sands is mostly influenced by the interconnected apparent cohesion and the
small-strain shear modulus, while other parameters play a secondary role.
1 INTRODUCTION primarily derived through empirical analyses of field

data obtained in cemented soils, although some full
The cone penetration test (CPT) is a widely used field scale 1-g and reduced scale centrifuge modeling data
test for soils in subsurface explorations due to its abil are also available. Therefore, the CPT has been pro
ity to provide continuous information on the stratig posed as a field test able to overcome the limitations of
raphy, and to be correlated to physical and engineering sampling and to provide direct indications of the pres
properties of the subsurface strata (e.g., Robertson & ence of the cementation in the field. In general, the
Campanella 1983, Lunne et al. 1997, Robertson 2016). cone tip resistance and the shear wave velocity both
While the CPT has been mostly used for the character increase with cementation, which suggests the devel
ization of sands and clays, it may also be used for the opment of a framework relating these two parameters
identification of naturally cemented soils deposits (e.g., as an indicator of possible cementation. Currently,
Puppala et al. 1995, Roy 2008) and artificially such a framework is not available for the interpretation
cemented and improved soils, including bio-cemented of the CPT data and their correlation to soil strength
ones (e.g. Gomez et al. 2018, Darby et al. 2019). parameters for bio-cemented sands.
Bio-cementation is a relatively new ground This paper presents the results of an effort to
improvement technique in which microorganisms pre connect cone tip resistances with fundamental,
sent in the soil are stimulated under specific conditions constitutive level, bio-cementation-induced
to precipitate calcite which coats and bridges soil par changes in soil behavior. To this end, a direct axi
ticle contacts. The artificially precipitated calcite is symmetric cone penetration model using the Mohr-
analogous to natural cementation and can be con Coulomb constitutive model and grid rezoning and
sidered as a proxy for naturally cemented sands (e.g., remapping algorithms (Moug 2017) is used to
DeJong et al. 2006, DeJong et al. 2010). To date, the model cone penetration in bio-cemented sands. By
characterization of bio-cemented sands and, similarly, connecting the results of cone penetration simula
naturally cemented sands in the field remains tions in a Mohr-Coulomb material to real data and
a challenging task due to difficulties in sampling and established relationships, this work will guide
its high cost. selection of equivalent strength properties of these
Classification and characterization approaches for challenging materials. More specifically, the appar
bio-cemented sands developed to date have been ent cohesion, peak friction angle, dilation angle,
DOI: 10.1201/9781003308829-50
368
and small-strain shear modulus within the Mohr- model with few parameters and nicely describes
Coulomb constitutive model are varied parametric a broad range of responses without unnecessary com
ally across a reasonable range of parameter values plexities. Based on numerous bench-scale and large-
informed by past laboratory, bench-scale, and cen scale experiments on bio-cemented sands, a linear
trifuge tests. relationship between the “apparent” cohesion ðcÞ and
The axisymmetric penetration model is briefly the change in shear wave velocity ðΔVs Þ was devel
introduced, followed by a summary of the equivalent oped to estimate the value of cohesion from the meas
Mohr Coulomb input parameters. Simulation results ured post-cementation in-situ Vs (El Kortbawi et al.,
are presented, interpreted, and compared against under review).
a published soil behavior type chart as well as
experimental data from a reduced scale centrifuge
model test. Conclusions pertaining to the validity of
the proposed framework and the results are drawn.
where ΔV s = the change in shear wave velocity
in m/s and c = the “apparent” cohesion in kPa.
2 NUMERICAL INVESTIGATION Following an extensive literature review on bio
cemented sands (e.g., Nafisi et al. 2019, Nafisi et al.
2.1 Axisymmetric penetration model 2020, Wu et al. 2020) and using the above empirical
The axisymmetric model presented in Figure 1 simu relationship, the values for the input parameters in
lates the steady-state penetration of a standard 10 cm2 Table 1 was defined.
(3.57 cm-diameter) cone into the soil column. Bound
ary conditions are imposed for a soil flowing into the
bottom of the model upwards relative to a fixed cone.
Mohr-Coulomb interface elements are applied at the
interfaces between the cone and soil to represent an
interface roughness factor of 0.60, which is the ratio
of the interface friction angle to the soil friction angle.
Stresses and Mohr-Coulomb material properties are
initialized for a “wished-in-place” condition at the
depth of interest. Initial stress conditions correspond
to an at-rest Ko condition and a fully drained penetra
tion is simulated. The cone penetration is velocity-
controlled at the gridpoints across the top boundary.
The right radial boundary is far enough from the cone
to avoid any boundary effects and is thus defined as
an infinite elastic boundary. The bottom boundary is
sufficiently far from the cone’s zone of influence to
maintain the prescribed in-situ stress conditions. Large
deformations are handled with a user-implemented
Arbitrary Lagrangian-Eulerian (ALE) algorithm which
performs grid rezoning and remapping during the Figure 1. Geometry and boundary conditions of the numer
cone penetration. Penetration is simulated until ical cone penetration model (Moug, 2017).
a steady state penetration resistance is achieved
(Moug et al. 2019), which is approximately 25 cone
diameters of penetration.
Table 1. Variables for parametric study.

2.2 Input parameters and calibration
The application of bio-cementation treatment to sands, Variable Tested values
as well as the geologic-time scale process where sand
Cohesion, c (kPa) 0, 10, 20
is gradually cemented, induce changes in the stiffness
Friction angle,
and strength. While difficult to characterize in the field, 30, 35
(degrees)
coupling the simulated cone tip resistance in a Mohr-
Dilation angle,
Coulomb type of material and established relationships 0, 10
ψ (degrees)
can guide the selection of equivalent strength param
“Functional” shear
eters for these challenging geomaterials. The Mohr- modulus,
Based on the relationship between
Coulomb constitutive model was selected because it ΔVs and cohesion
Gsec (kPa)
defines a failure envelope with a cohesion intercept Confining stress,
and a peak friction angle which can be representative 35, 100
σ 0vo (kPa)
of bio-cemented sands. More broadly, it is a simple
369
Due to the large strains around the penetrating of 0º. Figure 2b illustrates the change in qc with
cone, the small-strain shear modulus ðGmax Þ as an the change in Vs due to the presence of a cohesion
input model parameter does not reflect the complex (Eq.1) for the given combinations of stress states
deformations and strain softening in the plastic and friction angles. Figure 2c illustrates the effect
region near the cone tip which would result in an of the confining stress on qc by normalizing it
overestimation of the cone tip resistance (Teh & according to . Collectively,
Houlsby 1991, Lu et al. 2004). Therefore, a reduction these figures confirm that: (1) the cone tip resist
factor is applied to Gmax (where Gmax ¼ ρVs 2 and ρ ance increases considerably with the increase in
is the density) to account for the shear modulus soft the cementation level, (2) the change in qc is due
ening and a secant shear modulus ðGsec Þ is used, to the changes in cementation only, and (3) the
where Gsec ¼ Gmax =F and F ¼ 0:15 c þ 3 based on “cemented” friction angle may have an effect on
a calibration process and comparison of simulated the increase in qc but to a lesser extent than the
cone penetration with experimental results (El shear stiffness.
Kortbawi et al., under review). The chosen initial Vs Figure 3 Presents the variation of the tip resist
values normalize with respect to the varying ance as a function of the confining stress with vary
confining stresses of 35 and 100 kPa, hence the ing strength parameters, cohesion from 0 to 20 kPa,
results of these simulations are examined in absolute and friction angle from 30 to 35 degrees. Circle and
terms such as qc instead of incremental terms (Δqc Þ.
The input to each set of simulations is summar
ized in Tables 2 and 3 for the confining stresses of
35 and 100 kPa, respectively.
Table 2. Simulation input parameters for = 35 kPa.
Cohesion (kPa)
0 10 20
Vs,ini (m/s) 150 150 150

ΔVs (m/s) 0 189 378
Reduction factor F 3 4.5 6
Gsec (kPa) 12,750 43,149 78,262
Table 3. Simulation input parameters for = 100 kPa.
Cohesion (kPa)
0 10 20
Vs,ini (m/s) 200 200 200

ΔVs (m/s) 0 189 378
Reduction factor F 3 4.5 6
Gsec (kPa) 22,667 56,814 93,783
3 RESULTS
The focus of the subsequent analyses and illustra

tions is the cone tip resistance qc . Open and closed
symbols correspond to confining stresses of 35 and
100 kPa, respectively. Circle and square symbols
correspond to friction angle of 30 and 35 degrees,
respectively. Figure 2a presents the variation of Figure 2. (a) Variation of qc as a function of Vs , (b)
the absolute qc with an increasing shear stiffness variation of Δqc with varying ΔVs , and (c) variation of nor
(proportional to Vs according to Gmax ¼ ρVs 2 ), fric malized Qt as a function of Vs , for varying of 30º and 35º
tion angle, and confining stress at a dilation angle and different confining stresses at ψ of 0º.
370
square symbols correspond to friction angles of 30 and classification of soils which include cemented
and 35 degrees, respectively, whereas the shading sands. As the change in shear stiffness plays a major
corresponds to increasing the cohesion from 0 to role in the behavior of bio-cemented sands and
20 kPa. cemented sands in general, the small-strain shear
Similar to Figure 2, a positive trend exists between modulus Gmax and the stress-normalized cone tip
qc and where the following observations are made: resistance Qt can be correlated to guide the soil
(1) for σ0vo ¼35 kPa, qc for c = 20 kPa is around 4 behavior type and extend it to these cemented soils.
times the qc for c = 0 kPa, whereas for a higher = Figure 5a presents the simulation results for the dif
100 kPa, this ratio decreases to 2.5 times regardless of ferent scenarios. In addition to the simulation results
the friction angle, (2) for c = 0 kPa, qc for = 100 from this work, experimental data obtained from
kPa is around twice the qc for σ0vo = 35 kPa, whereas cones pushed in a bio-cemented specimen in
for c = 20 kPa, this ratio decreases to 1.3 times regard a centrifuge model (Darby et al. 2019) are plotted.
less of the friction angle, and (3) for σvo = 35 kPa, qc Several levels of cementation were established by
for = 35º is around 1.5 times qc for = 30º whereas the experimentalists targeting light, moderate, and
for σ0vo ¼100 kPa this ratio slightly decreases to 1.4 heavy cementation levels. The datapoints from the
times, regardless of the cohesion. These observations cones pushed in treated sands plot reasonably within
suggest that: (1) cohesion is the major contributor to the region corresponding to soils with “ageing
the enhanced tip resistance, followed by the confining cementation”, with few datapoints plotting close to
stress, and then followed by the friction angle, and (2) the “uncemented” region due to their light level of
the effect of the cementation is more significant at low cementation (Darby et al. 2019). Specimens with
confining stress. higher levels of cementation follow the trend dis
cussed earlier. Moreover, Figure 5a shows that the
simulation results for = 35 kPa fall within the
range of the experimental data from Darby et al.
(2019) with a similar confining stress of 35 kPa,
hence validating the proposed approach.
Figure 3. Variation of qc as a function of for varying

and c at ψ of 0º.
Figure 4 Incorporates, in addition to the cohesion,

friction angle and confining stress, the effect of the
dilation angle ψ. Figures 4a and 4b correspond to
= 30º and 35º, respectively, and the diamond and tri
angle symbols correspond to a ψ of 0º and 10º,
respectively. The results show that the increase in
the dilation angle leads to a larger cone tip resist
ance, due to stronger dilation of the soil which in
turn results in a higher resistance of the soil to shear
ing. The qc with a ψ = 10º (a limit for bio-cemented
sands suggested by the literature e.g., Wu et al. Figure 4. Variation of qc as a function of c for varying
2020) is 3 to 4 times higher relative to qc with a ψ = ψ and at, (a) of 30º, and (b) of 35º.
0º, depending on the strength parameters with this
ratio increasing for a higher confining stress ( of
100 kPa). The CPT-based classification chart according to
The current state of practice uses previously Qt and Go =qt is reproduced in Figure 5b and the
established charts for the interpretation of CPT data simulation and experimental results are overlaid.
371
Figure 5b shows a reasonable agreement between towards a higher normalized tip resistance and lat
the expected soil behavior from the input parameters erally to a higher normalized stiffness. When other
(i.e., uncemented sands with zero cohesion versus parameters remain constant, increasing the friction
cemented sands with nonzero cohesion) and the sug angle results in a larger qc hence the soil moves ver
gested soil behavior from the chart. For example, the tically towards a higher normalized tip resistance but
simulations with c = 0 kPa plot in the “sand” behav horizontally towards a smaller normalized stiffness
ior type on the chart as expected for a cohesionless due to the normalization with qc .
uncemented sand. Similarly, the simulations with
c > 0 kPa plot in the region corresponding to soils
with “ageing cementation” on the chart. Moreover, 4 CONCLUDING REMARKS
the aforementioned relative magnitude of Qt and Go
/qt between the cemented sands and the uncemented The cone penetration test has been commonly used
sand is also evident on the soil classification chart. in soils to indicate the soil behavior type and to
assess the soil’s resistance in bearing capacity and
liquefaction mitigation applications. Its use can be
extended to bio-cemented or cemented sands. Some
limitations are associated with pushing cones in
cemented sands, such as the inability of the cone to
detect light cementation (equivalent to calcite con
tent less than 3%) and the destructive nature of the
test to any pre-existing cementation in the soil. How
ever, it is still a common and valuable test able to
provide proxies of the expected soil behavior and its
degree of cementation in the absence of more viable
exploration methods.
In this paper, a parametric study on the input
parameters to an axisymmetric cone penetration
model for sands was presented in the context of its
application to bio-cemented sands. The numerical
model implemented the Mohr-Coulomb constitutive
model which was used to estimate equivalent
“cemented” strength parameters. Although the
Mohr-Coulomb constitutive model prescribed con
stant strength parameters to strain-dependent soil
conditions, the approximation with a “functional”
shear modulus allows the model to reasonably pre
dict the elastic-plastic behavior of bio-cemented
soils in the failure zone near the cone tip. The input
parameters varied in this paper are the cohesion, fric
tion angle, dilation angle, and confining stress. The
results were analyzed in terms of cone tip resistance,
and the study confirmed that it increased with
increasing the level of cementation, and suggested
that the cohesion value is the main contributor to this
improvement in the soil resistance. The dilation and
friction angles also affect the soil’s resistance but to
a lesser extent. Simulation results were also com
pared to experimental data from CPTs pushed in bio
cemented sands and to previously established soil
behavior type charts for “unusual” soils. The simula
tion results fell within the range of the experimental
Figure 5. Soil classification chart based on normalized data and they both plotted in the “ageing cementa
cone resistance and small-strain shear modulus, reproduced tion” region of the chart. Hence the proposed frame
from Lunne et al. (1997), and overlaid with simulation work appears reasonable to estimate the soil’s
results from this work and experimental data from the cen strength and overcome the challenges in sampling
trifuge test from Darby et al. (2019). cemented sands.
Oftentimes, considerable effort is put into the
selection of input parameters for numerical models.
The results suggest that as cohesion increases, The above observations related to the effect of input
i.e., the level of cementation precipitated at the par parameters on the cone tip resistance are insightful
ticle contacts increases, the soil moves vertically for prioritizing the choice of input parameters and
372
can be potentially used to guide parameter selection DeJong, J. T., Mortensen, B. M., Martinez, B. C., &
for numerical models, based on CPT data, to repre Nelson, D. C. (2010). Bio-mediated soil improvement.
sent cemented materials in numerical simulations. Ecological Engineering, 36(2), 197–210.
The ranges of input parameter values used in this DeJong, J. T., Fritzges, M. B., & Nüsslein, K. (2006).
study reflect the typical ranges found in the literature Microbially induced cementation to control sand
for lightly and moderately bio-cemented sands. The response to undrained shear. Journal of Geotechnical
trends presented here may be cautiously extrapolated and Geoenvironmental Engineering, 132(11),
1381–1392.
to reasonably higher strength parameters and confin El Kortbawi, M., Moug, D. M., Ziotopoulou, K.,
ing stresses; however, the relative magnitudes may DeJong, J. T., Boulanger, R. W. (under review). Appli
warrant further exploration. cation of an axisymmetric cone penetration model to
In the developing field of bio-cemented sands, bio-cemented sands. Journal of Geotechnical and
numerical models that are validated and calibrated Geoenvironmental Engineering.
against existing data can be leveraged to populate Gomez, M. G., DeJong, J. T., & Anderson, C. M. (2018).
and synthesize data that are not yet available. While Effect of bio-cementation on geophysical and cone
some approximations such as the estimation of the penetration measurements in sands. Canadian Geotech
nical Journal, 55, 1632–1646.
cohesion value and the reduced secant shear modu
Lu, Q., Randolph, M. F., Hu, Y., & Bugarski, I. C. (2004).
lus are made herein, the numerical model is still A numerical study of cone penetration in clay. Géotech
a valuable tool to, in the first stage, study the sensi nique, 54(4), 257–267.
tivity of the results to input parameters and, in Lunne, T., Robertson, P. K., & Powell, J. J. M. (1997).
the second stage, synthesize data needed to develop Cone penetration testing in geotechnical practice.
correlations between these various parameters. El Moug, D. M. (2017). Axisymmetric cone penetration model
Kortbawi et al. (under review) extend this work to for sands and clays. Doctoral Dissertation, University of
develop a correlation between the cone tip resistance California, Davis.
and the cohesion in order to implement it in Moug, D. M., Price, A. B., Parra Bastidas, A. M.,
a plasticity constitutive model for bio-cemented Darby, K. M., Boulanger, R. W., & DeJong, J. T. (2019).
Mechanistic development of CPT-based cyclic strength
sands. correlations for clean sand. Journal of Geotechnical and
Adding to the importance of the CPT in indicating Geoenvironmental Engineering, 145(10).
the behavior of the present soil, it can and should be Nafisi, A., Mocelin, D., Montoya, B. M., & Underwood, S.
a preferred method for post-cementation verification (2019). Tensile strength of sands treated with micro
in the field. The CPT can thus serve as a tool in the bially induced carbonate precipitation. Canadian Geo-
design of ground improvement methods and in technical Journal, 57, 1611–1616.
the QA/QC of the final improvement, especially in Nafisi, A., Montoya, B. M., & Evans, T. M. (2020).
the field of bio-cementation ground improvement Shear Strength Envelopes of Biocemented Sands with
Varying Particle Size and Cementation Level. Journal
techniques.
of Geotechnical and Geoenvironmental Engineering,
146(3).
ACKNOWLEDGMENTS Puppala, A. J., Acar, Y. B., & Tumay, M. T. (1995). Cone
penetration in very weakly cemented sands. Journal of
This material is based upon work primarily sup Geotechnical Engineering, 121(8), 589–600.
Robertson, P. K. (2016). Cone penetration test (CPT)-based
ported by the National Science Foundation (NSF)
soil behaviour type (SBT) classification system — an
under NSF Award Number EEC-1449501. Any opin update. Canadian Geotechnical Journal, 53,
ions, findings and conclusions, or recommendations 1910–1927.
expressed in this material are those of the author(s) Robertson, P. K., & Campanella, R. G. (1983). Interpret
and do not necessarily reflect those of the NSF. The ation of cone penetration tests. Canadian Geotechnical
authors are grateful to Professor R. W. Boulanger for Journal, 20(4), 1–81.
discussions and insights that set the framework of Roy, D. (2008). Coupled use of cone tip resistance and
this work in its original stages. small strain shear modulus to assess liquefaction
potential. Journal of Geotechnical and Geoenvironmen
tal Engineering, 134(4), 519–530.
REFERENCES Teh, C. I., & Houlsby, G. T. (1991). An analytical study of
the cone penetration test in clay. Géotechnique, 41(1),
Darby, K., Hernandez, G., DeJong, J. T., Boulanger, R. W., 17–34.
Gomez, M. G., & Wilson, D. W. (2019). Centrifuge Wu, S., Li, B., & Chu, J. (2020). Stress-dilatancy behavior
model testing of liquefaction mitigation via microbially of MICP-treated sand. International Journal of Geome
induced calcite precipitation. Journal of Geotechnical chanics, 21(3), 04020264. https://doi.org/10.1061/(asce)
and Geoenvironmental Engineering, 145(10). gm.1943-5622.0001923.
373
© 2022 the Author(s), ISBN 978-1-032-31259-0
A data-driven approach to predict shear wave velocity from CPTu

measurements
I. Entezari & J. Sharp
ConeTec Group, Burnaby, British Columbia, Canada
P.W. Mayne
Georgia Institute of Technology, Atlanta, Georgia, USA
ABSTRACT: The use of machine learning modelling to predict shear wave velocity (VS) from piezocone
penetration tests (CPTu) is presented. A large dataset of paired VS-CPTu data (n = 104,054) compiled from
seismic piezocone (SCPTu) soundings completed in a wide variety of soil types with various stress histories
and geological environments was used to develop machine learning models to directly estimate VS from
CPTu data. The impact of soil microstructure on the results was investigated and separate models were devel
oped to predict VS in cemented and uncemented soils. The results of machine learning models outperformed
the existing widely used CPT-based relationships to predict VS.
1 INTRODUCTION projects. When actual measurements of VS are not

practical, estimates can still provide useful additional
Shear wave velocity (VS) is an important property of information. Existing empirical CPT relationships for
geomaterials and is widely used to evaluate the VS have been developed using statistical approaches.
dynamic and elastic properties of soils in geotechnical This paper explores the use of a data-driven approach
design. VS measurements provide the fundamental via machine learning modelling to predict VS from
stiffness of the ground in terms of the small-strain CPTu. Machine learning requires little or no priori
shear modulus (Go), specifically: Go = ρVS2, where ρ assumptions to be considered and thus are more flex
= γt/ga = soil total mass density, γt = soil total unit ible than statistical models.
weight, and ga = gravitational acceleration constant. The development dataset used in this study is com
VS measurements can be obtained using a variety of prised of VS-CPTu data pairs from ConeTec SCPTu
test methods. The value of VS can be measured in the soundings collected from 2017 to early 2021. The
laboratory using high quality undisturbed samples and soundings have been completed in a wide variety of
special equipment (resonant column, bender elements), soil types with various stress histories and are from
which is costly and restricted to a limited number of geological environments around the world. The dataset
samples. In-situ measurements of VS are preferable to is tested with a random forest algorithm to develop
preserve site-specific conditions and minimize errors a model for the prediction of VS from CPTu data. The
due to sampling disturbance and stress release. results of the machine learning models are compared to
In-situ measurements of VS can be obtained through empirical equations proposed by Mayne (2006) and
downhole and crosshole tests, seismic piezocone tests Robertson (2009). Furthermore, the impacts of soil
(SCPTu), spectral analysis of surface waves (SASW), microstructure and cementation on estimated VS results
and multichannel analysis of surface waves (MASW). are discussed and separate models are developed for
SCPTu method is often preferred as it is a rapid and the categories of uncemented and cemented soils.
cost-effective technique to measure in-situ wave vel
ocities in conjunction with CPTu parameters, including 2 BACKGROUND
cone tip resistance (qt), sleeve friction (fs), and
dynamic porewater pressure (u2) in a single direct push
2.1 Seismic Piezocone Tests (SCPTu)
sounding.
Although performing site-specific testing is the pre The SCPTu is similar to the CPTu probe with the
ferred method to determine shear wave velocity, sev addition of one or more geophones or accelerometers
eral empirical relationships have been developed to located behind the cone tip. As shown in Figure 1, the
estimate VS from the basic CPTu for lower risk equipment required to perform SCPTu includes the
DOI: 10.1201/9781003308829-51
374
age. Robertson (2009) also developed a generalized
soil relationship where VS is a function of net tip
resistance (qnet = qt - σvo), total vertical stress (σvo),
atmospheric pressure (σatm), and soil behaviour type
index (Ic). An overview on some of the CPT relation
ships to predict VS has been provided in Wair et al.
(2012).
In this study, the estimated VS results from the
machine learning models are compared to the results
obtained by the empirical expressions proposed by
Mayne (2006) and Robertson (2009), shown in
Equations 1 and 2, respectively:
Figure 1. Schematic of seismic CPTu equipment.

where VS is in m/s in both equations, fs is in kPa in
seismic source on the ground surface, seismic sensors Eq. 1, and σatm is in same units as qt and σvo in Eq. 2.
behind the cone probe, a data-acquisition system, and
a data recording trigger circuit (Styler et al. 2016). 2.3 Impact of soil microstructure
The seismic testing is conducted at selected
depth intervals (typically every 1m), while the The existing empirical correlations developed for
penetrometer is pushed into the ground. The shear interpretation of CPT results have been generally
waves are generated by striking a horizontal beam developed using silica-based uncemented soils with
pressed firmly against the ground. Paired leftand little or no microstructure (Robertson 2016). There
right-strikes are used to define either the first fore, caution should be exercised when CPT based
arrival time of shear waves, or first crossover, or relationships are used in soils with microstructure.
both. VS is calculated using the difference in According to Robertson (2016), the empirical par
arrival times of the shear wave traces between the ameter, K*G, can be used to determine whether soils
source and geophone at two successive depths. are cemented or not. K*G is calculated as
The SCPTu has several advantages including cor (Go/qnet)(Qtn)0.75 (Robertson 2016), where Qtn is the
related VS and CPTu results, the capability to do normalized tip resistance. Soils with K*G of less than
deep tests, the ability to measure compression 330 are likely young and uncemented with little or
wave velocity (VP), and availability of soil prop no microstructure, while soils with K*G of greater
erty interpretations based on small-strain rigidity than 330 can be classified as cemented and micro-
index (IG = Go/qnet, where qnet is cone net tip structured soils.
resistance). The cemented versus uncemented soils are con
sidered in this study for the evaluation of the per
formance of the machine learning model.
2.2 Existing CPT relationships Furthermore, individual models are developed spe
Relationships between CPT data and VS have been cifically for uncemented and cemented soils.
studied previously by various researchers. Hegazy &
Mayne (1995) developed various expressions to esti
mate VS using parameters including tip resistance (qc), 3 DESCRIPTION OF DATASET
sleeve friction (fs), vertical effective stress (σ’vo), and
in-situ void ratio (e) for Quaternary clays, sands, and To investigate the potential of a data-driven
mixed soils. Various relationships have also been pro approach to estimate VS from CPTu data, a dataset
posed by Piratheepan (2002) for the estimation of VS of paired VS-CPTu data was compiled using Cone
based on tip resistance (qc), sleeve friction (fs), vertical Tec’s geospatial database. The database was queried
effective stress (σ’vo), depth (z), and soil behaviour to find the relevant information that resulted in
type index (Ic) for Holocene clays, sands, and other 14,855 SCPTu tests worldwide with more than
soils. Mayne (2006) showed a relationship where the 248,500 VS-CPTu data pairs. For this study, sound
VS is a function of the sleeve friction (fs) for Quater ings collected after 2016 were selected in order to
nary soils. Another correlation developed by Andrus only utilize modern tests with increased quality con
et al. (2007) for Holocene and Pleistocene soils is trol. Procedural changes in ConeTec’s SCPTu meth
based on tip resistance (qt), depth (z), soil behaviour odology yielded slightly higher accuracy in VS
type index (Ic), and a time factor depending on the soil measurements due to signal enhancement and signal
375
stacking after this date (Styler & Weemees 2016). fall in the young and uncemented soils category
Consequently, the dataset was reduced to 104,809 (soils with little or no microstructure), where K*G is
VS-CPTu data pairs from 7171 independent SCPTu less than 330 (green points in Figure 2). Another
soundings worldwide. Most of the soundings are 40,314 data points fall in the cemented soils category
from North America, with additional contributions (soils with microstructure), where K*G is greater
from various sites in South America, Australia, than 330 (blue points in Figure 2).
Europe, and Asia. To pair the CPTu parameters with
VS measurements at a given depth, the median of
3.2 Training and test datasets
CPTu parameters over a window size equal to the VS
depth interval was calculated. Only depth intervals The dataset was split into training and test sets. The
equal to or less than 1 m were considered to minim training set was used to calibrate the model whereas
ize variations due to potential soil heterogeneity. The the test set was used to evaluate the model perform
CPTu parameters paired with VS included corrected ance. The data collected from 2017 to 2019 was used
tip resistance (qt), sleeve friction (fs), porewater as the training set and data collected in 2020 and
pressure (u2) and depth (z) at each VS measurement. early 2021 provided the test set. This allows for an
Additional parameters including normalized tip unbiased performance evaluation of the model (a
resistance (Qtn), normalized friction ratio (Fr), nor blind test) where the potential errors due to variation
malized porewater pressure (Bq), net tip resistance in stress histories and geological environments are
(qnet), total stress (σvo), and effective stress (σ’vo) as taken into account. The number of paired VS-CPTu
well as small-strain shear modulus (Go), small-strain data points for the training and test sets are listed in
rigidity index (IG), and K*G were also calculated. Table 2.
Calculation of a number of these parameters
required the soil unit weight (γ). The machine learn
ing model based on corrected tip resistance (qt),
sleeve friction (fs), porewater pressure (u2) and depth
(z) developed by Entezari et al. (2021) was used to
estimate unit weight at each depth. The measured
equilibrium pore pressure profile of each SCPTu
sounding, combined with the estimated unit weight
profile, was used to determine in-situ vertical
stresses.
Data points with net tip resistance (qnet) of less
than 100 kPa were screened out to remove fluid-like
tailings from the dataset. Also, data points with
sleeve friction (fs) of less than 1 kPa were screened
out in order to remove data where the soil-sleeve
friction was less than internal o-ring friction. The
final dataset used included 104,054 VS-CPTu data
pairs. Table 1 lists the summary statistics of the
paired dataset.
Table 1. Summary statistics of the VS-CPTu dataset.

Figure 2. The dataset plotted in the Qtn-IG chart.
Min Max Mean
VS (m/s) 9 1000 251

qt (MPa) 0.1 94.1 8.4 Table 2. Number of data pairs in the training and test sets.
fs (kPa) 1.0 1577 117.6
All Uncemented Cemented
u2 (kPa) -87.2 5489 245.0
z (m) 0.3 129.6 17.3 Training set 73,010 45,386 27,624
σ’vo (kPa) 0.1 2185 215.4 Test set 31,044 18,354 12,690
Total number of data pairs = 104,054.
4 MACHINE LEARNING MODELLING

3.1 Soils with microstructure
Machine learning models acquire information from
The plot of normalized tip resistance (Qtn) versus prior data, allowing the computers to discover predict
small-strain rigidity index (IG) for the dataset is ive rules applicable for future data. Machine learning
shown in Figure 2. Accordingly, 63,740 data points models are generally data-hungry and need large
376
datasets for training. In general, as more data become The performance of the model was also assessed
available, the more accurate and robust the predictions on uncemented and cemented soils. When only unce
become. Machine learning is widely used in numerous mented soils were considered in the test set, the bias
disciplines and has gained interest in geotechnical and error of the estimated VS results are –23.4 and
engineering. Example applications of machine learning ±34.8 m/s, respectively. For cemented soils, the bias
for CPT interpretations can be found in Erzin & and error of the estimated VS were observed to be
Ecemis (2016), Reale et al. (2018), Wang et al. (2019), 27.2 and ±62.8 m/s, respectively.
Erharter et al. (2021), Rauter & Tschuchnigg (2021),
and Entezari et al. (2020, 2021).
5.2 Uncemented and cemented soils models
In this study, the random forest algorithm (Brei
man 2001) was employed to calibrate CPTu data to Using the uncemented and cemented soil categories
VS measurements. It is one of the most widely used in the training set, two separate models were devel
machine learning algorithms for classification and oped for the estimation of VS in these types of soils.
regression tasks. Random forest is an ensemble of Figure 4 shows the relationship between the SCPTu
several decision trees and thus overcomes the short measured and random forest predicted VS for the
comings of traditional decision trees, predominantly fraction of the test set in uncemented soils. As can
overfitting. The models here were trained using four be seen, the correlation between estimated and meas
input parameters including corrected tip resistance ured VS significantly improved compared to the all-
(qt), dynamic porewater pressure (u2), sleeve friction soils model shown in Figure 3 (R2 of 0.79 compared
(fs), and depth (z). to 0.58). The bias and error of the estimated results
were observed to be 0.6 and 28.2 m/s, respectively.
4.1 Performance evaluation
The performance of the random forest models is
evaluated using the properties of the cumulative dis
tribution function (CDF) of errors on the test set.
The error is calculated as the discrepancy between
the measured VS from SCPTu and predicted VS from
the random forest models. The 50th percentile in the
CDF is taken as the bias of the prediction. Assuming
the errors follow a normal distribution, the CDF
values at 15.9% and 84.1% correspond to ±1 stand
ard deviation. The average of the two CDF values at
15.9% and 84.1% is considered as the overall error
of the model.
To compare the performance of the machine
learning models to existing CPTu expressions, simi
lar performance evaluation is performed on the test
set using the predicted VS obtained from the equa
tions proposed by Mayne (2006) and Robertson
(2009).
5 RESULTS
Figure 3. Relationship between measured and random
forest estimated VS on the test set using all soils.
5.1 All-soils model
An all-soil model was developed using the random
forest model trained with all data points in the train Evidently, the random forest model is far better
ing set. The relationship between measured VS from able to model the relationship between VS and CPTu
SCPTu and the estimated VS from the random forest parameters in uncemented soils, compared to the all-
model is shown in Figure 3. This relationship is soil scenario.
shown for the test set. The R2 of the model on the The results of the random forest model developed
test set was observed to be 0.58. The error analysis for cemented soils are also shown in Figure 4.
using CDF of errors showed that the bias and error The bias and error were observed to be -12.3 and
of the estimated results are –8.5 and 49.5 m/s, 54.1 m/s, respectively. Compared to the all-soil
respectively. The bias –8.5 m/s means that random model, this model performs better on cemented soils,
forest model overestimates the measured VS by but the bias and error are still high. This is presum
8.5 m/s overall. The error of 49.47 m/s means that ably because microstructure can have a variety of
68.2% of the estimated VS values fall within impacts on CPTu parameters. Thus, the learnt rela
±49.5 m/s of the measured VS from SCPTu testing. tionship between VS and CPTu parameters in
377
cemented soils of the training set may not be applic 52.5 m/s, respectively. The bias and error were calcu
able on the cemented soils of the test set. lated to be 51.1 and 82.8 m/s, respectively, on
cemented soils.
In case of Robertson (2009) model, the bias
5.3 Existing relationships
and error were 21.5 and 64.3 m/s, respectively,
Figures 5 shows the relationships between the esti on all soils in the test set. The bias and
mated VS calculated using the methods of Mayne error were observed to be -6.1 and 50.3 m/s,
(2006) and Robertson (2009) with the measured VS respectively, for uncemented soils, compared to
using SCPTu on the test set. For the Mayne (2006) 69.2 and 57.8 m/s for cemented soils. Overall, it
model, the bias and error were observed to be 12 and can be seen that these expressions perform
68.6 m/s, respectively, when error assessment was better on uncemented soils, as expected.
done on all soils. When only uncemented soils were A summary of model performances is presented
considered, the bias and error were dropped to -7.2 and in Table 3. It should be noted that no limits
were applied to the two existing methods
because the intent was to compare the results to
those obtained from the random forest models
developed using a wide range of soil types.
Limiting the range of applicable data to be used
in the existing methods would be prudent and
may result in a better average correlation and
error.
Table 3. Performance of different models.
Bias±Error (m/s)
Model All Soils Uncemented Cemented

RF-All Soils -8.5±49.5 -23.4±34.8 27.2±62.8
RF-Uncemented NA 0.6±28.2 NA
RF-Cemented NA NA -12.3±54.1
Mayne (2006) 12.0 -7.2±52.5 51.1±82.8
±68.6
Robertson (2009) 21.5 -6.1±50.3 69.2±57.8
±64.3
5.4 Example SCPTu Vs profile

An example SCPTu profile of VS estimated
using the random forest models developed in
this study is shown in Figure 6. The estimated
VS values from the expressions of Mayne (2006)
and Robertson (2009), as well as the measured
VS profile, are displayed along with the
results of this study. The SCPTu sounding is
from 2020 and is thus part of the test set. As
evident, both all-soils and uncemented models
are in agreement with the measured VS. The
uncemented model, however, shows less fluctu
ations and a better performance compared to
the all-soils model. Both of these models appear
to outperform the Mayne (2006) and Robertson
Figure 4. Relationships between measured and random forest
(2009) models. The analysis of K*G revealed
estimated VS of the uncemented (top) and cemented soils that the soils are uncemented for the whole
(bottom) in the test set when separate models were trained profile except for depth ranges between
using uncemented and cemented soils in the training set. 3.5-9 m and 15-17.5 m.
378
Figure 6. Example profile of Vs estimated from models
developed in this study and existing relationships.
When a priori information on soil categories is avail

able, the soil-specific models developed in this study
could lead to better results than the all-soils model
and the equations proposed by Mayne (2006) and
Robertson (2009).
In future work, the dataset compiled in this study
will be used to investigate the potential of machine
learning algorithms to classify and identify cemented
and uncemented soils from CPTu parameters. How
ever, K*G of 330 as a threshold to distinguish
cemented from uncemented soils has been determined
empirically and may not be an absolute metric.
In addition to the models developed and presented
in this paper, random forest models were trained by
adding normalized tip resistance (Qtn), normalized
friction ratio (Fr), normalized porewater pressure
Figure 5. Relationships between measured vs. estimated (Bq), and effective stress (σ’vo) to the input variables,
VS using Mayne (2006) (top) and Robertson (2009) but no significant improvements were observed.
(bottom) methods.
7 CONCLUSIONS
6 DISCUSSION
Machine learning models using a random forest
In practice, a priori information on the soil micro- algorithm were developed to directly predict VS
structure is required in order to be able to employ from CPTu data. A dataset of paired VS-CPTu data
soil-specific models developed in this study to compiled from 7171 SCPTu soundings completed at
estimate VS (uncemented and cemented soils various sites with a wide variety of soil types, stress
models). Determining the soils categories based on histories, and geological environments was used to
K*G is not practical without knowing VS. Therefore, develop machine learning models. Results showed
information on soil categories should be available that the all-soils model developed using random
from other sources such as previous SCPTu testing forest algorithm can estimate VS with ±49.5 m/s
in the region under investigation or information on error. The model developed for uncemented soils
the geology of soils. showed a significant improvement and could predict
When such information is not available, the VS with ±28.2 m/s error. The model developed for
results of this study showed that the developed all- cemented soils achieved an accuracy of ±54.1 m/s.
soils model performs better than the Mayne (2006) All the developed machine learning models outper
and Robertson (2009) models when CPTu is pushed formed the studied existing relationships from litera
in regions with both cemented and uncemented soils. ture. Although actual measurement of VS is always
379
preferrable, it appears to be more crucial when deal Mayne, P.W. 2006. In-situ test calibrations for evaluating
ing with soils that have microstructure. The models soil parameters. Proc. Characterization and Engineer
developed are from a very large dataset compiled ing Properties of Natural Soils II, Singapore, Vol. 3:
from SCPTu soundings from various geological 1601–1652.
regions and are therefore considered to be robust, Piratheepan, P. 2002. Estimating shear wave velocity
however engineering judgement should always be from SPT and CPT data. MSc Thesis, Clemson
exercised when using any empirical statistics or University.
Rauter, S. & Tschuchnigg, F. 2021. CPT data interpretation
models. employing different machine learning techniques.
Geosci. J. 11(7), 265.
REFERENCES Reale, C., Gavin, K., Librić, L., Jurić-Kaćunić, D. 2018.
Automatic classification of fine-grained soils using CPT
Andrus, R.D., Mohanan, N.P., Piratheepan, P., Ellis, B.S., measurements and Artificial Neural Networks. Adv.
and Holzer, T.L. 2007. Predicting shear wave velocity Eng. Inform. 36: 207–215.
from cone penetration resistance, Proc. 4th Intl. Conf. Robertson, P.K. 2009. Interpretation of cone penetration
on Earthquake Geotech. Engrg., Thessaloniki, Greece. tests – a unified approach, Can. Geotech. J. 46
Breiman, L. 2001. Random forests. Machine Learning, (11):1337–1355.
45(1): 5–32. Robertson, P.K. 2016. Cone penetration test (CPT)-based
Entezari, I., McGowan, D., and Sharp, J. 2020. Tailings soil behaviour type (SBT) classification system-an
characterization using cone penetration testing and update. Can. Geotech. J. 53: 1910–1927.
machine learning, Proc. Tailings & Mine Wastes 2020, Styler, M.A. & Weemees, I. 2016. Quantifying and redu
University of British Columbia, Vancouver, 695–704. cing uncertainty in down-hole shear wave velocities
Entezari, I., Sharp, J., and Mayne, P.W. 2021. Soil unit weight using signal stacking. Proc. ISC’5, Gold Coast,
estimation using the cone penetration test and machine Australia.
learning, Proc. GeoNiagara 2021, Niagara Falls, Canada. Styler M.A., Weemees, I., Mayne, P.W. 2016. Experience
Erharter, G.H., Oberhollenzer, S., Fankhauser, A., and observations from 35 years of seismic cone penetra
Marte, R., and Marcher, T. 2021. Learning decision tion testing (SCPTu), Proc. GeoVancouver 2016: www.
boundaries for cone penetration test classification, Com- cgs.ca
puter-Aided. Civil & Infrastructure Eng. 1: 1–15. Wair, B.R., DeJong, J.T., and Shantz, T. 2012. Guidelines
Erzin, Y. & Ecemis, N. 2016. The use of neural networks for Estimation of Shear Wave Velocity. PEER Rept.
for the prediction of cone penetration resistance of silty 2012/08, Pacific Earthquake Engineering Research
sands. Neural Comput. Appl. 28: 727–736. Center, Berkeley, CA: 95 p.
Hegazy, Y.A. & Mayne P.W. 1995. Statistical correlations Wang, H., Wang, X., Wellmann, J.F., Liang, R.Y. 2019.
between VS and cone penetration data for different soil A Bayesian unsupervised learning approach for iden
types. Proc. CPT ’95, Linkoping, Sweden, Vol. 2: tifying soil stratification using cone penetration data.
173–178. Can. Geotech. J. 56: 1184–1205.
380
© 2022 the Author(s), ISBN 978-1-032-31259-0
Interpretation and comparison of CPT derived soil properties to static and

cyclic laboratory tests on unique fine-grained soils in Western Washington
and Oregon
B. Exley, A. Pyrch, J. Jacoby & B. Thunder
Haley & Aldrich, Inc., USA
ABSTRACT: CPT interpretation methods are evaluated against laboratory and in-situ methods to
identify and evaluate the soil properties of multiple non-textbook soil deposits located in Western
Washington and Oregon. These deposits include unique glacially overridden soils which are wide
spread in the Puget Sound region, as well as younger soils which vary widely in depositional history
and stress history.
The glacially overridden deposits can be very thick and are heavily overconsolidated, often show
ing signs of fissures and slickensides when they are fine-grained, with SPT blowcounts frequently
much larger than N=100 in coarse grained deposits. Based on the CPTs included in this paper, the
fine-grained overconsolidated materials are typically identified by the CPT as sand-like and dilative,
and generally have Ic values between 2.0 and 2.6 despite frequently having index properties which
classify the soils as stiff to hard high plasticity clay. This paper documents the characteristics of these
soils to facilitate identification of them by the CPT.
Additionally, the correlated strength, preconsolidation pressures, and cyclic resistance of the
younger fine-grained soils are evaluated against laboratory testing methods. This includes documenting
a site in Beaverton, Oregon with several CPTs paired with constant rate of strain consolidation labora
tory tests and both monotonic and cyclic direct simple shear tests in the fine-grained Missoula Flood
Deposits as part of a study on the susceptibility of the silt to liquefaction or cyclic softening.
1 INTRODUCTION 2 GLACIALLY OVERRIDDEN SOILS
Western Washington and Oregon each consist of soil 2.1 Fine-grained material
conditions which generally do not fall under the clas
sical soils. The Puget Sound Lowlands in Western The glacially overconsolidated fine-grained soils in the
Washington generally are a deep basin that has been Puget Sound Lowlands are known to widely vary in
glacially overridden several times, with the most strength due to the presence of blocky texture and
recent occurring approximately 15,000 years ago slickensides in zones. However, when intact, they
during the Vashon Stade of the Fraser glaciation exhibit significant strength and stiffness. Unfortunately,
(Booth & Goldstein, 1994). The glacially overridden the soils are also brittle and easily sheared due to the
soils were subjected to approximately 5,000 feet of ice rotary action of a pitcher barrel when undisturbed sam
(Easterbrook 1969), resulting in very dense and very pling is attempted. Accordingly, being able to reliably
hard soil deposits that engineers generally do not estimate the soil properties using the CPT presents sig
attempt to push CPTs in. nificant value where the soils are intact. However, ini
Holocene age deposits tend to be interlayered tial Soil Behavior Type screening tools generally
due to depositional environment and frequently describe the soils as having a more coarse-grained
consist of relatively silty sands or low plasticity behavior than would be measured by traditional index
silts. These characteristics generally make them dif testing.
ficult to collect and test using high quality undis This trend has been observed on multiple sites,
turbed samples. When silt-rich, these soils do with observed tip resistances in the fine-grained
not behave like classical sand-like or clay-like material often about 100 tsf or more with SBT index
materials. values of approximately 2 to 2.2. Documented on
DOI: 10.1201/9781003308829-52
381
Figure 1 are three CPT sounding results from a site design properties. The use of the modified Norwegian
which was located along Interstate 405 south of Bel Institute of Technology (NTH) method (Ouyang &
levue, Washington (Site 1). Three CPTs were pushed Mayne 2019) results in an interpreted friction angle
into a likely glacially overconsolidated fine-grained of approximately 35 degrees, which is consistent with
material. However, the materials had lower qt values the results of consolidated undrained triaxial tests on
than previously mentioned, ranging from about 50 to undisturbed samples of similar clays in the region.
75 tsf. The Ic index was generally about 2.2. This friction angle can then be used to refine the esti
mated preconsolidation pressure, resulting in a refined
estimate as documented in Mayne and Miller (2021).
3 SUSCEPTIBILITY TO LIQUEFACTION OR
CYCLIC SOFTENING OF RECENTLY
DEPOSITED FINE-GRAINED SOILS
As previously noted, the Holocene aged fine-grained

soils in the Pacific Northwest are commonly silt-rich
with relatively low plasticity, and are characterized as
“transitional” soils when using the SBT Index. This
causes significant difficulties for consultants in identify
ing whether the material has a “sand-like” or “clay-like”
behavior when subjected to cyclic loading. These soils
do not consistently identify with an Ic>2.6, further con
Figure 1. Site 1 basic CPT results. founding identification of them through traditional filter
ing methodology when assessing sand-like behavior. An
in-depth study of these soils in the Pacific Northwest has
The CPTs were advanced near existing borings been completed and is documented by the Oregon
which had index testing completed on the samples Department of Transportation (Dickenson et. al 2021).
that identified the fine-grained soil as a high-plasticity The study includes a publicly accessible database that is
clay (CH). A qualitative screening of the sounding being continually updated with new data. Currently, the
shows large positive excess pore pressure values with database includes over 300 cyclic tests targeted at evalu
Bq values generally between 0.3 and 0.5. Evaluating ating the response of these silty materials to cyclic load
the soil type using a technique which considers pore ing. In general, their findings show that the materials do
pressure response (Schneider 2008), the CPT correctly not exhibit the significant degradation of shear stiffness
identifies the material as a clay. An example of this is and strengths that sands do as the porewater pressure
interpretation is shown on Figure 2 for HC-CPT-2-20. increases. Gingery (2014) showed that, for sands with
relative densities of less than 40%, sands tend to exhibit
post-liquefaction accumulation of shear strains per cycle
of approximately 2% per cycle. Conversely, while silts
do accumulate shear strains during cyclic loading, the
accumulated shear strain per cycle is typically much
smaller. For example, a sample of the Columbia River
Silt with a PI = 1 and an OCR = 1 accumulates approxi
mately 0.7% shear strain per cycle for single amplitude
shear strains larger than 3% (Dickenson et. al 2021).
Further, unlike sands, these soils do not tend to show the
large loss of strength following cyclic loading which
imposes large pore pressure ratios (Ru) or single ampli
tude shear strains larger than 3%. Accordingly, the term
“liquefaction” is misleading when used to describe
response to cyclic loading of these soils. Documented
herein is a case study of Willamette Silt encountered in
Beaverton, Oregon (Site 2) with evaluation of the cyclic
resistance of those soils relative to traditional liquefac
tion triggering methods.
Figure 2. Example schneider plot interpretation of Site 2 is a well-documented site that has had 19
HC-CPT-2-20. CPTs advanced through the soil profile, as well as 21
traditional SPT borings and 24 test pits. In support of
a recent seismic assessment of the site, several CPTs
Correctly identifying the material as a clay facili and borings were advanced (included in the totals
tates the use of appropriate methods to estimate above), the latter of which included undisturbed
382
samples used for advanced laboratory testing. The For a preliminary evaluation of the susceptibility
samples were collected and extruded using methods to liquefaction of the fine-grained soils encountered
generally consistent with those recommended by at the project site, we used Bray & Sancio (2006)
Ladd & DeGroot (2003), including cutting and criteria based on soil index tests. According to the
delaminating the soil from the tube to minimize findings in Bray & Sancio (2006), the amount and
sample disturbance during preparation of the sample type of clay minerals in the soil best indicate lique
for testing. In addition to index testing, constant rate faction susceptibility, which can be quantified by PI
of strain (CRS) consolidation testing, static direct and the ratio of water content and liquid limit (wc/
simple shear testing (DSS), and cyclic direct simple LL). Based on the available soil index tests for the
shear testing (CDSS) was completed. The DSS and project site, 9 out of 22 samples (41%) tested in the
CDSS tests were completed using a SHANSEP type depth range of 5 to 55 feet bgs are considered to be
testing methodology to explicitly evaluate the effects susceptible or moderately susceptible to liquefaction
of overconsolidation on the samples and attempt to by the Bray & Sancio (2006) criteria.
overcome some of the impacts of sample disturbance The use of a typical Ic cut-off for the transition
that may have occurred during testing. between sand-like and clay-like behavior of 2.6
Fill materials, if encountered, ranged between would result in analyzing most of the profile as
approximately 2 and 15 feet thick and generally con a sand, calculating a magnitude 7.5 normalized cyclic
sisted of either reworked native silty/clayey soils or resistance ratio (CRR7.5) of approximately 0.15.
imported gravels. More significant thicknesses of fill However, as noted, Dickenson et. al and others have
were encountered on the south end of Site 2 in proximity observed that these materials behave in a manner
to Cedar Mill Creek. Materials interpreted as native allu more similar to clays. An assessment of the profile
vium generally consisted of very soft clays underlying using a methodology similar to that presented by
thicker fills in proximity to Cedar Mill Creek. The gen Idriss & Boulanger (2006) results in an improved
erally silty Missoula Flood deposits were encountered in assessment of the soil’s response to cyclic loading.
all of the explorations near the surface or at depths of up
to approximately 35 feet below ground surface (bgs)
3.1 Overconsolidation ratio profiling
underlying fill and alluvium. All of Haley & Aldrich’s
explorations encountered high- to low-plasticity clay The use of the generalized expression for stress history
interpreted to be the Hillsboro Formation underlying the (Agaiby & Mayne 2019) with dependency on Ic did
Missoula Flood deposits at approximately 22 to 75 feet not adequately reflect the laboratory test results. This is
bgs. All the explorations terminated in materials inter due to the insensitivity of the Ic index to changes in
preted as the Hillsboro Formation down to approxi soil type in these soils. Agaiby & Mayne presented
mately 101.5 feet, the maximum depth explored. While alternative relationships to estimate the exponent m’
the layering is generally consistent across the site, there on D50 and fines content. However, these generally did
is significant intra-deposit variation across the site. not provide consistent differentiation between the silts
Therefore, the assessment of an individual CPT is pre and clays, and a relationship associated with Bq was
sented to simplify the case study. However, the results observed. As such, a modified m’ exponent was devel
were broadly applicable across the site. oped as a function of the pore pressure ratio. The pre
The basic results of CPT-6 are presented on consolidation pressure was estimated according to
Figure 3. The Ic index in the fine-grained soils is on Equations 1 and 2, which are modifications of the
average about 2.5. Interestingly, the index appears to be methods presented by Agaiby & Mayne (2018):
insensitive to index properties of the soils, however the
pore pressure ratio was more responsive to transitions
between silt and clay.
The resulting overconsolidation ratios calculated

for CPT-6 are presented on Figure 4, with nearby
CRS results from nearby borings. There was a general
trend on the site that the clays had significantly larger
OCR’s than the silts and sands, which is reflected in
the reduction of OCR with the drop in Bq that reason
ably fits the trends observed in the lab tests. Some dif
ferences in elevation of the transitions between silts
and clays is expected between the borings and the
CPT. This likely explains the depth offset between the
CPT predicted OCR for B7-S6 and B9-S12. This
Figure 3. Site 2 basic CPT results for WHQ CPT-6. trend was observed in other CPTs as well.
383
Figure 5. DSS undrained shear strength ratios.
Figure 4. OCR profile for WHQ CPT-6.
3.2 Shear strength profiling

The undrained shear strength of the materials can be
calculated using critical state soil mechanics for the
appropriate stress path (Mayne 2005). The effective
stress friction angle can be calculated by the modified
NTH method (Ouyang & Mayne 2018). Equation 3
can be used to calculate the undrained shear strength
for the direct simple shear (DSS) stress path. When
calculating the undrained shear strength, it was found
that using a Λ=0.7 works well for these silty soils.
The laboratory test results from DSS testing for

depths between 10 and 30 feet deep are presented on Figure 6. CPT-6 calculated normally consolidated shear
Figure 5. Figure 6 provides a comparison of the CPT strength ratios compared to lab test model fit.
derived properties with the laboratory-based model fit.
The “a” coefficient can be described as a function
3.3 Cyclic resistance profiling of the normally consolidated undrained shear strength
ratio. Using model parameters of a = 0.8*Su/σv’ =
Dickenson et. al (2021) presents the cyclic resistance
0.213 and b = 0.15 results in a reasonable fit for the
ratio (CRR) of silts in the Pacific Northwest accord
soils as tested for a single amplitude shear strain of
ing to Equation 4, where N is the number of cycles:
3%, as shown on Figure 7. The model fit may be
increased by approximately 0.05 for larger shear
strain amplitudes up to approximately 7%. The model
fit results in a CRR at 30 cycles equivalent to 50% of
the undrained shear strength, which can be used in
384
4 CONCLUSIONS
When evaluating the behavior of glacially overcon

solidated soils in Western Washington, it has been
shown that penetration of the CPT through the over-
consolidated fine-grained soils of the Puget Sound
Lowlands can be readily accomplished. When evalu
ating those soils, practitioners should consider the
use of Schneider plots to facilitate identification of
fine-grained soils that may otherwise not be identi
fied by SBT.
Additionally, the assessment of cyclic resistance
ratios of transitional soils in the Pacific Northwest
is documented. The soils generally have an Ic less
than 2.6, resulting in most practitioners using sand-
like liquefaction assessment methods. This signifi
Figure 7. Cyclic shear strain accumulation as a function of cantly underestimates the soils cyclic resistance
undrained shear strength and number of cycles. ratio. Classic liquefaction behavior was not
observed in any of the cyclic laboratory tests in
soils associated with an Ic index as low as 2.2. Ic
values generally ranged from 2.3 to 2.6 for depths
in the soils tested with plasticity index values as
low as 5.
A methodology was presented to assess the
cyclic resistance to cyclic softening of the tran
sitional soils that is consistent with other soils
in the Pacific Northwest that can be used in
combination with existing stress history and
shear strength profiling techniques. Further
refinement to OCR profiling methods in soils
other than clay would be beneficial. An example
of such refinement would be the development of
a unified approach which provides a theoretical
Figure 8. Peak Ru generation normalized to shear strength framework for both drained and undrained
and cyclic loading. materials, which also addresses the impact of
site-specific shear strength parameters in
a consistent manner.
combination with the values presented in Figure 6 to
develop a CRR7.5 profile. This fit is less than the CRR
= 0.8*Su recommended by Idriss and Boulanger REFERENCES
(2006) for clays, but results in a significantly larger
CRR than the sand-like behavior presented previously Agaiby, S.S. & Mayne, P.W. 2019. CPT evaluation of yield
for soils that are overconsolidated. This reduction stress profiles in soils. Journal of Geotechnical and
may be due to the use of SHANSEP-like testing Geoenvironmental Engineering 145(12).
methodology to directly control the sample OCR, Booth, D.B. & Goldstein, B 1994. Patterns and processes
resulting in a loss of the in-situ soil stiffness. of landscape development by the Puget lobe ice sheet.
Washington Division of Geology and Earth Resources
The “a” coefficient is generally less than the 80: 207–218.
coefficients presented by Dickenson et. al for other Bray, J.D. & R.B. Sancio 2006. Assessment of the liquefac
soils in the Pacific Northwest. This is consistent tion susceptibility of fine-grained soils. Journal of Geo-
with observations that under cyclic loading these technical and Geoenvironmental Engineering, 132(9),
soils tend to generate more excess pore pressures pp.1165–1177.
than most of the silts in the Pacific Northwest data Dickenson, S.E., Khosravifar, A., Beaty, M., Bock, J.,
base, with max Ru values of approximately 0.9, as Moug, D., Schlechter, S., & Six, J. 2021. Cyclic and
shown on Figure 8, suggesting the silts on Site 2 post-cyclic behavior of silt-rich, transitional soils of the
are more sand-like than many other silts in the Pacific Northwest; a database for geo-professionals in
practice and research. Salem, OR: Oregon Department
region. of Transportation.
When the CRR is exceeded, resulting in an Ru Easterbrook, D.J. 1969. Pleistocene chronology of the
greater than 0.8, the post cyclic shear strength was Puget Lowland and San Juan Islands, Washington. Geo
approximately 50% of the undrained shear strength. logical Society of American Bulletin 80: 2273–2286.
385
Gingery, J.R. 2014. Effects of liquefaction on earthquake Mayne, P.W. & Miller, B. 2021. Analytical solutions for
ground motions. Doctoral Dissertation, UC San Diego. evaluating CPTu soundings in overconsolidated Hart
Idriss, I. M. & Boulanger, R. W. (2008). Soil liquefaction ford clay. ISC-6 Paper 100.
during earthquakes. Monograph MNO-12, Earthquake Ouyang, Z. and Mayne, P.W. 2018. Effective friction angle
Engineering Research Institute: Oakland, CA of clays and silts from piezocone. Canadian Geotech
Ladd, C.C & DeGroot D.J. 2003. Recommended practice nical Journal 55(9): 1230–1247.
for soft ground site characterization: Arthur Casagrande Robertson, P.K. & Cabal, K.L. 2015. Guide to Cone Pene
Lecture. Proc. of the 12th Panamerican Conference on tration Testing for Geotechnical Engineering 6th Edi
Soil Mechanics and Geotechnical Engineering 1:3–57. tion. Singal Hill, CA: Gregg Drilling & Testing, Inc.
Cambridge, MA. Schneider, J.A., Randolph, M.F., Mayne, P.W. &
Mayne, P.W. 2005. Integrated ground behavior: in-situ and Ramsey, N.R. 2008. Analysis of factors influencing soil
lab tests. H.D Benedetto, T. Doanh, H. Geoffroy, & classification using normalized piezocone tip resistance
C. Sauzeat (eds), Deformation Characteristics of Geo- and pore pressure parameters. Journal Geotechnical and
materials 155–180; London, UK; Taylor and Francis Geoenvironmental Engineering. 134 (11): 1569–1586.
Group plc. Gingery (2014)
386
© 2022 the Author(s), ISBN 978-1-032-31259-0
Evaluation of cyclic softening potential using CPTu and assessment with

cyclic triaxial test results: A case study
K. Fakharian, M. Bahrami & M. Kashkooli
Department of Civil & Environmental Engineering, Amirkabir University of Technology, Tehran, Iran
H. Vaezian & T. Bahrami

Pars GeoEnviro Inc., Tehran, Iran
ABSTRACT: An extensive geotechnical investigation was performed on an oil terminal site located in
a seismically active field. Combination of in situ and laboratory tests were performed including Piezocone Pene
tration Tests (CPTu), some of which were adjacent to the borehole locations, providing a detailed stratigraphic soil
profile and different physical and mechanical parameters. In this paper, the application of the CPTu test for analyz
ing cyclic liquefaction is evaluated and discussed. Both “liquefaction” and “cyclic softening” potentials are evalu
ated depending on the magnitude of Ic factor. A clayey silt layer was identified having an Ic factor of about 2.5
between 12 and 19 m of depth. The average qc of this layer was 6 MPa and friction ratio (RF) was 2.5. The
induced pore pressure, u, was measured between 300 to 1800 kPa from 12 to 19 m of depth. The equivalent SPT
counts from correlations with CPT were 15 while the real SPT counts were about 50 or higher. According to CPT
data, the factor of safety of liquefaction was calculated having an average of 0.56 for a 0.49g earthquake with
M=7. Undisturbed samples were taken from the layer onto which cyclic undrained liquefaction tests were carried
out. The results, however, revealed that specimens were resisting the cyclic liquefaction with sufficiently high
factor of safety. The soil classification in the laboratory showed that the soil was clay and not silty clay. Detailed
analyses and interpretations are carried out to find out the reasons of the differences between CPTu correlations
and real response of the specimens to cyclic loading. The results show that modifications might be required at
times in interpretations of soil classification as well as liquefaction potential evaluation for soils at the borderline
of low plasticity clay and silty clay.
1 INTRODUCTION cementation, affect “sand-like” soil response to the

cyclic liquefaction (Robertson 2009). Generally, loose,
Liquefaction can be defined as a loss of strength and young, uncemented “sand-like” soil is known to be
stiffness in soils, divided into “cyclic liquefaction” highly susceptible to cyclic liquefaction. In contrast,
and “cyclic softening” categories. The terms “sand- “clay-like” soils do not reach zero effective stress due
like” and “clay-like” were introduced to refer to soils to the developed pore pressure under the same
whose stress-strain behavior under undrained mono- undrained cyclic loading condition. In “clay-like” soil,
tonic and cyclic loading is similar to that of sands the buildup pore pressure is controlled by key fea
and clays, respectively (Boulanger & Idriss 2004). tures, including in situ state, sensitivity, age, fabric,
Soils like gravels, sands, and very low plasticity silts and cementation. Soft normally to lightly overconsoli
that are recognized as “sand-like” can experience dated and sensitive clays are considered to tolerate
some level of cyclic liquefaction. However, the term significant shear strains. They are susceptible to cyclic
“clay-like” is reserved for clay and plastic silts which softening that cause lateral and vertical deformations.
can experience cyclic softening (Robertson 2009). Monotonic and cyclic undrained loading tests data
Developing high positive pore pressure under show that soil transition from sand-like to clay-like
undrained cyclic loading can reach the “sand-like” behavior appears over fine-grained soils on a fairly
soil at the state of zero effective stress causing narrow range of plasticity indices (Boulanger &
a condition under which the initial soil structure is Idriss 2006). To distinguish between fine-grained
lost, and large deformations occur. Alongside the size soils that may exhibit sand-like behavior versus
and duration of the imposed cyclic loading, in situ clay-like behavior, researchers have developed sev
state parameters and soil structure, including relative eral criteria. The Chinese criteria have been widely
density, effective confining stress, age, fabric, and for evaluating the liquefaction susceptibility of silts
DOI: 10.1201/9781003308829-53
387
and clays (Boulanger & Idriss 2006). Based on An industrial site on northern shorelines of Gulf
Atterberg limits for those reported data from sites in of Oman was lately developed where boreholes and
China, Seed & Idriss (1971) categorized clayey soils CPTu tests were carried out. At a small zone of this
having less than 15% finer than 5 μm, LL<35, and site near sea, the CPTu tests identified a clayey silt
wc>0.9 LL as vulnerable to severe strength loss as to silty clay with Ic factor of slightly below 2.5
a result of earthquakes (Seed 1982). Andrews & (average of 2.497). The Robertson classification
Martin (2000) reviewed a few case histories, con chart used correlations of the sand-like soil with Ic
cluding that soils are susceptible to liquefaction if <2.5 for this layer and concluded that this layer is
they have 10% finer than 2 μm and LL<32. Soils are highly liquefiable with a factor of safety, FS=0.54
not susceptible to liquefaction if they have 10% finer for the 0.49 g peak ground acceleration, PGA, of the
than 2 μm and LL>32. Further study is required for field. This is while SPT correlations indicate a much
soils that meet one, but not both, of these criteria greater FS=0.9 to 1, also assuming that the soil is
(Andrews & Martin 2000). Bray et al. (2004) pro sand-like. Cyclic triaxial tests were carried out on
vided new liquefaction criteria for soils containing undisturbed specimens revealing that the soil is not
significant fine contents. The criteria are considered liquefiable.
fine-grained soils with plastic index, PI<12 and wc In this paper, the geotechnical condition of the
>0.85 LL as susceptible to liquefaction. Boulanger site is characterized presenting the CPTu and SPT
& Idriss (2006) suggested that for practical purposes, data along with blow count profile of a precast con
fine-grained soils exhibit clay-like behavior if they crete pile driven in this zone. The soil classification
have PI≥7. If a soil plots as CL-ML, the PI criterion results from CPTu data and soil sampling and lab
may be reduced to PI≥5. Fine-grained soils that do data are compared and attempts are made to account
not meet these criteria should be considered sand- for the contradictions between CPTu data and other
like liquefiable, unless appropriate in situ and labora methods.
tory testing have shown otherwise (Bray et al. 2004).
The most common CPT-based classification sys
tems are based on behavior characteristics and are 2 STUDY SITE
often referred to as a soil behavior type (SBT) classi
fication (e.g., Robertson, 1990). The SBT charts are 2.1 Location and morphology
behavior-based descriptions, while the general soil
classification is textural-based (Robertson 2016). The study site is located near Jask in Hormozgan
The charts proposed by Robertson (1990) and the Province, southern Iran, on the northern shorelines
present update (Robertson 2009) have become very of Gulf of Oman. The site is approximately
popular, in which cone resistance, qc, and sleeve 300 m away from sea. This area is located in the
resistance, fs, are normalized to account for in situ Zagros orogenic zone and generally consists of marl,
stress level with depth. Since then, many CPT soil sandstone, and silty/clayey sediments. The site area
behavior type charts have been developed (e.g., Jef is situated in a seismically active zone. The PGA
feries & Davies 1993, Ramsey 2002, Schneider et al. with a return period of 475 years is estimated equal
2012). Robertson & Wride (1998) had suggested that to 0.49g on the basis of a site-specific seismology
Ic = 2.6 was an approximate boundary between soils study carried out by Pars GeoEnviro Consulting
that were either more sand-like or more clay-like. Engineers.
The Robertson SBTn chart works well in ideal soils
but can be less effective in structured soils. The term 2.2 Geotechnical conditions
“ideal soil” describe soils with little or no micro-
structure that are predominately young and unce An extensive geotechnical investigation was per
mented. The term “structure” is used to describe formed at the study site. A combination of in situ
features either at the deposit scale, macrostructure, and laboratory tests were carried out including
e.g., layering and fissures, or at the particle scale, Piezocone Penetration Tests, CPTu, some of which
microstructure, e.g., bonding and cementation were adjacent to the borehole locations, providing
(Robertson 2016). Older natural soils tend to have a detailed stratigraphic soil profile and different
some microstructure due to the post-deposition. physical and mechanical parameters. Total of 12
Many researchers have contributed to developing the CPTu points as well as 6 boreholes were considered
CPT-based models to evaluate liquefaction based on for the geotechnical investigation objectives.
cyclic liquefaction case histories that were limited to Subsurface layers across the site mostly consist of
predominately silica-based ideal soils (e.g., Robert cohesive soil containing clay and silty clay, and non-
son & Wride 1998). The updated method proposed cohesive soil containing silty sand to sandy silt. After
by Robertson (2009) is the most widely used method processing all data from field exploration, in situ tests
to assess the potential for liquefaction. The method including CPTu and SPT, as well as lab tests, the site
was initially developed for clean sands and it pro area has been divided into three main zones in terms
posed a unified approach, in which two sets of equa of geotechnical characteristics and subsoil conditions
tions are created for sand-like and clay-like behavior as shown for BH5. Zone b, which is the focus of
in liquefaction. this paper, mainly consists of a 6 m thick layer from
388
non-plastic silt and sand categorized as medium
dense to dense starting from ground surface; under
neath this layer, a layer consisting of plastic lean clay
is located down to depth of around 12 m and categor
ized as firm stiff to stiff. Third layer starts from depth
12 m and extends down to the end of investigated
BHs and CPTs consisting of clay and silty clay with
PI=10-11; this layer is categorized as very stiff.
In most locations across the site, there has where qt is CPT corrected cone resistance, qt = qc
been a good compatibility between geotechnical + u2 (1-a). The “a” is net area ratio of the cone
data obtained from borehole sampling and the determined from laboratory calibration for the
ones obtained on the basis of CPTu processed cone, with a typical value between 0.7 and 0.85, fs
data and soil classification charts. However, in is CPT sleeve friction, σv0 and are in situ total
Zone b in which two CPTu tests as well as one and effective vertical stresses, respectively. The
borehole were executed, some incompatibilities primary ground water table is evaluated from site
are observed, especially between the results from observations and rechecked from drilled adjacent
CPT6 and BH5. BH5 which was continuously borehole as 5 m from the surface at the CPT loca
drilled down to 20 m is located adjacent to CPT6 tion. Figure 1 also shows the recorded data of
penetrating to 19.1 m. Below 11 m, incompatibil CPT6 in Zone “b”. Profile of soil behavior classifi
ities were more severely observed in classifica cation based on Roberson (2009) method is plotted
tion, equivalent SPT numbers, and most next to the CPT data alongside of the profile of
significantly, liquefaction/cyclic softening poten soil classification from laboratory test.
tial analyses. The data points scattering over the Qt−Fr chart
are plotted in Figure 2. The soil layering according
3 CPT DATA INTERPRETATIONS to CPT classification are three layers: Layer I: the
top 5 m can be categorized as “clean sand to silty
sand”, Layer II: between 5 to 12 m, dominantly
3.1 Soil classification
“clay to silty clay”, and Layer III: between 11 to
Among proposed SBT charts on the basis of CPT 19.5 m, classified as “silt mixtures, clayey silt to
data, the updated unified approach by Robertson silty clay”. Layer III is categorized as a transition
(2009) is employed. The effective overburden stress layer in the SBT chart. To further evaluate and
is used to normalize tip and sleeve friction resist verify the classification, another chart proposed by
ances as Qt and Fr, as shown in Eqs. 1 and 2, Eslami & Fellinius (1997) was attempted as shown
respectively. in Figure 3.
Figure 1. CPT6 cone resistance, sleeve friction, friction ratio, and pore pressure profile accompanied by soil classification
profile based on Robertson (2009) and soil classification based on laboratory tests on adjacent borehole BH5.
389
Considering plotted data on Figures 2 and 3.2 Liquefaction & cyclic softening
Figure 3, the two proposed classification SBT
A procedure suggested by Robertson & Wride
charts show an accumulation of points over 12
(1998) is used to evaluate the potential for
to 19 m representing by red squares. The accu
cyclic softening and cyclic liquefaction. The
mulation of points is mainly categorized as
procedure used the basic methodology, devel
clayey silt to silty clay over the SBT chart pro
oped by Seed & Idriss (1971), calculating
posed by Robertson (2009) on Figure 2. These
cyclic stress ratio, CSR, induced by the earth
points have Ic value close to the suggested
quake, and cyclic resistance ratio, CRR of the
boundary, Ic=2.6, distinguishing between sand-
soil. If CSR>CRR, then liquefaction can occur.
like or clay-like behavior by Robertson & Wride
The proposed procedure estimates CRR based
(1998). The proposed boundary Ic=2.6 is the
on the “sand-like” or “clay-like” classification,
approximate boundary between soils that can be
separated by the defining boundary of Ic=2.6.
either claylike or sandlike, however, it is not
The updated procedure is well established by
always a good fit, except for predominately
Robertson (2009).
young uncemented, and normally consolidation
To get an outlook of the potential for
soils referred to as ideal soil (Robertson, 2016).
cyclic liquefaction, Robertson (2009) presented
Also, the other CPTu tests performed at the site
a SBT Qt−Fr chart by defining “dilative” and
in different locations show the existence of this
“contractive” zones as a guide. Figure 4 shows
behavior type layer which is recognized as
the plotted data points over the chart for CPT6.
a transition layer between sand-like and clay-
The Ic boundaries of 2.5 and 2.7 are plotted
like behavior.
with dashed lines. The red dots of the layer
12-20 m are dominantly in the transition zone
and some points slightly below 2.5. The plots
indicate that the points are at the “dilative” zone
of the chart.
In all assessments, the earthquake magnitude,
M, is estimated 7 with PGA equal to 0.49
g. Figure 5 shows the profile of safety factors
for cyclic liquefaction for sand-like soil classifi
cation and cyclic softening for claylike soil
classification with depth. The calculations
demonstrate that the soil beneath the performed
location has potential for either of the cyclic
liquefaction and cyclic softening. In layers
with clay-like behavior, Ic>2.7, green points,
FS is approximately around 0.73 in cyclic soft
ening. While in sand-like layers, Ic<2.5, red
points; Ic between 2.5 and 2.7, yellow points,
Figure 2. Scattered data points on Qt-Fr soil behavior type the average safety factor against cyclic liquefac
chart proposed by Robertson (2009) for CPT6. tion is 0.56.
Figure 3. Scattered data points on qe-fs classification chart Figure 4. Data points on soil behavior type chart proposed
proposed by Eslami & Fellenius (1997) for CPT6. by Robertson (2009) for CPT6.
390
Figure 5. Liquefaction analysis of layers at location of Figure 6. Comparison between direct results and correlated
CPT6; “liquefaction: is for sand-like (Ic<2.5) and transition results: (a) N-SPT from actual SPT test and correlated from
zone (2.5<Ic<2.7); “cyclic softening” is for clay-like CPT, (b) blow count values from actual pile tests program.
(Ic>2.7).
4 COMPARISON WITH BOREHOLE SAMPLES
4.1 Atterberg limits and classification

The results of Atterberg Limits tests performed
according to ASTM D-4318 shows that a thick plas
tic layer is situated with PI=10-11 extended between
12 to 20 m (Figure 1). CPTu results show that in this
layer, silt mixtures containing clayey silt to silty clay
are dominant. According to Ic value, this layer is
located in sand-like zone as far as classification is
concerned (Ic≤2.6). It seems that lab testing shows
more dominant clayey soil than CPT classification.
SPT test was carried out with intervals of 2 m in
soil layers as well. N-SPT obtained from SPT test in
BH5 is shown in Figure 6. In addition, N-SPT is esti Figure 7. Assessments of liquefaction: (a) CSR values
mated from available correlations proposed by Robert from SPT numbers based on Youd et al. (2001) and CRR
son (2012) as well as Jefferies and Davies (1993) values based on Seed & Idriss (1971), (b) factor of safety
based on CPTu test results, as shown (Figure 6). Com against liquefaction.
paring actual SPT number with correlated values from
CPT shows that in upper 6 m sand, there is fairly An evaluation of liquefaction is also performed
good compatibility between SPT and CPTu data; in based on SPT results according to the method pro
deeper depths consisting of clay, N-SPT values from posed by Youd et al. (2001). The SPT-based assess
SPT test are somewhat greater than values correlated ments for cyclic liquefaction is presented in Figure 7.
from CPTu down to 10 m within the firm clayey The results show that both CPT and SPT confirm the
layer. Below this, the N-SPT from the tests is signifi probability of liquefaction occurrence in the layer at
cantly higher than CPT correlation, highlighting a very the depth of 12 to 20 m. However, SPT-based assess
stiff layer. The blow count data of a 400 mm precast ment generally shows higher FS values close to unity.
concrete square pile is also plotted in Figure 6 versus
depth to further validate the soil stratification. Good
correlations exist in layers I and II, while the blow 4.2 Cyclic triaxial tests
counts rapidly reached 20 blows per 100 mm penetra The liquefaction resistance could be evaluated using
tion and beyond and pile driving was stopped. The undrained cyclic triaxial tests with two-way, symmet
blow count record further confirm the trend of N-SPT rical loading in compression and extension (Yang &
counts rather than N-SPT correlated from CPT. Sze, 2011). For the mode of cyclic mobility, the usual
391
approach is to define failure criterion as the point that The lab testing on the specimens classified the soil
is accompanied by the occurrence of 5% double- as CL with PI=10-11 and relatively high shear
amplitude (DA) axial strain (Toki et al., 1986; Ishihara, strength under cyclic loading. The SPT N-value is
1996). greater than the N-value correlated with qc. CPTu
A series of cyclic triaxial tests were carried out results show very high induced pore-water pressure
for the assessment of liquefaction and cyclic soften for the subject layer even up to 1800 kPa, therefore,
ing occurrence of the 3rd layer. An undisturbed the layer is highly compressible. This is while the
sample extracted from depth 12 m was studied by SBC has recognized the soil as dilative. Cyclic tri
cyclic triaxial test. Assuming peak ground acceler axial test confirm that soil is not liquefiable even
ation of amax=0.49g, the average stress for liquefac after 25 cycles.
tion triggering was calculated equal to CSR=0.37 It looks like that an Ic value of slightly below and
based on the method proposed by Seed & Idriss above 2.5 mandated using “cyclic liquefaction” cor
(1971). After saturating the soil sample and being relations of CPTu. This is while the soil is in fact
consolidated under σ3c= 120 kPa, cyclic deviatoric clay-like and “cyclic softening” correlations should
stress considering CSR=0.37 varying in the range of have been used instead. When the soil is classified at
±100 kPa has been applied on the sample. The the transition zone, special cares are required in
results show that no sign of failure exist (Figure 8). adopting correlations with CPTu data to evaluate
For an earthquake with M=7, according to Boulanger liquefaction.
& Idriss (2004), the number of cycles equivalent to
an earthquake for triggering liquefaction is 25 for
clay-like soil samples. Figure 8 shows that at cycle REFERENCES
No. 1000, DA axial strain is equal to 3.4%, which is
much less than the proposed criterion. Thus, Andrews, D. C. A., & Martin, G. R. 2000. Criteria for
liquefaction of silty soils. Proc., 12th World Conf. on
although CPTu correlated results for assessing cyclic Earthquake Engineering, 1–8.
liquefaction shows high risk of cyclic softening at Begemann, H. K. 1965. The friction jacket cone as an aid
the layer between depth of 12 and 20 m, the per in determining the soil profile. Proc. 6th Int. Conf. on
formed triaxial test shows no evidence of liquefac SMFE, 1, 17–20.
tion under assumed seismic condition. Boulanger, R., & Idriss, I. 2004. Evaluating the potential
for liquefaction or cyclic failure of silts and clays.
Center for Geotechnical Modeling, Dept. of Civil and
5 CONCLUSIONS Environmental Engineering, Univ, of California. Davis,
California, USA.
Boulanger, R. W., & Idriss, I. M. 2006. Liquefaction Sus
The layer between 12 to 19 m is classified using two
ceptibility Criteria for Silts and Clays. Journal of Geo-
classifications systems as clayey silt to silty clay. technical and Geoenvironmental Engineering, 132
(11),1413–1426.
Bray, J. D., Sancio, R. B., Riemer, M. F., & Durgunoglu, T.
2004. Liquefaction susceptibility of fine-grained soils.
Proc., 11th Int. Conf. on Soil Dynamics and Earthquake
Engineering and 3rd Int. Conf. on Earthquake Geotech
nical Engineering, 1, 655–662.
Douglas, B. J., & Olsen, R. 1981. Soil classificaion using
electric cone penetrometer. Symp. on Cone Penetration
Testing and Experience, Geotech. Eng. Div., 209–227.
Eslami, A., & Fellenius, B.H. 1997. Pile Capacity by direct
CPT and CPTu methods applied to 102 case histories.
Canadian Geotechnical Journal, 886–898.
Ishihara, K. 1996. Soil behaviour in earthquake geotech
nics. Oxford: Clarendon Press.
Jefferies, M. G., & Davies, M. P. 1993. Use of CPTU to
estimate equivalent SPT N 60. Geotechnical Testing
Journal, 16(4),458–468.
Ramsey, N. 2002. A calibrated model for the interpretation
of cone penetration tests (CPTs) in North Sea quaternary
soils. Offshore Site Investigation and Geotechnics’
Diversity and Sustainability. Proceedings of an Inter
national Conference of offshore Site Investigation and
Geotechnics, PP. 341–356. London, UK.
Robertson, P.K. 2009. Performance based earthquake
design using the CPT. Conference on PBD in Earth
quake Geotechnical Engineering, Tokyo.
Figure 8. Cyclic triaxial test results at the end of 1000th Robertson, P.K. 2009. Interpretation of cone penetration
cycle: (a) loops of deviator stressaxial strain, (b) Effective tests - A unified approach. Canadian Geotechnical Jour
stress paths on Cambridge p’-q diagram. nal, 46(11), 1337–1355.
392
Robertson, P.K. 2016. Cone penetration test (CPT)-based Toki, S., Tatsuoka, F., Miura, S., Yoshimi, Y., Yasuda, S.,
soil behavior type (SBT) classification system - An & Makihara, Y. 1986. Cyclic undrained triaxial strength
update. Canadian Geotech. Journal, 53(12),1910–1927. of sand by a cooperative test program. Soils Found. 26,
Robertson, P.K. 1990. Soil classification using the cone No. 3, 117–128.
penetration test. Canadian Geotechnical Journal, 27 Yang, J., & Sze, H.Y. 2011. Cyclic behaviour and resistance
(1),151–158. of saturated sand under non-symmetrical loading
Robertson, P.K., Wride, C.E. 1998. Evaluating cyclic lique conditions. Géotechnique, 61:1,59–73.
faction potential using the cone penetration test. Discus Youd, T.L., Idriss, I.M., Andrus, R.D., Arango, I.,
sion 1. Canadian Geotec. Journal, 37(1), 442–458. Castro, G., Christian, J.T., Dobry, R., Finn, W.D.L.,
Schneider, J.A., Hotstream, J.N., Mayne, P.W., & Harder, Jr. L.F., Hynes, M.E., Ishihara, K., Koester, J.
Randolph, M.F. 2012. Comparing CPTU Q–F soil clas P. Liao, S.S.C., Marcuson III, W.F., Martin, G.R.,
sification charts. Géotechnique Letters, 2(4),209–215. Mitchell, J.K., Moriwaki, Y., Power, M.S.,
Seed, H.B., & Idriss, I.M. 1971. Simplified procedure for Robertson, P.K., Seed, R.B., & Stokoe, K.H. II. 2001.
evaluating soil liquefaction potential. Journal of the Soil Liquefaction Resistance of Soils: Summary Report
Mechanics and Foundations Division, 97(9), 1249–1273. from the 1996 NCEER/NSF Workshops on Evaluation
Seed, H.B. 1982. Ground motions and soil liquefaction of Liquefaction Resistance of Soils. Journal of Geo
during earthquakes. Earthquake Engineering Research technical and Geoenvironmental Engineering, 127
Institute, Berkeley, California, USA. (10), pp. 817–833.
393
© 2022 the Author(s), ISBN 978-1-032-31259-0
Application of integrated Game Theory-optimization subground

stratification (-IGTOSS) model to Venetian Lagoon deposits
M.S. Farhadi & T. Länsivaara
Faculty of Built Environment, Department of Civil Engineering, TERRA Research Center
L. Tonni
Department of Civil, Chemical, Environmental and Materials Engineering (DICAM)
University of Bologna, Bologna, Italy
ABSTRACT: The geographical and geological spatial variabilities raise challenges for geotechnicians to
devise globally applicable subground stratification models working based on cone penetration testing (CPT).
Recently, a novel CPT-based stratification and classification model was proposed in Tampere University, Fin
land. It combines the soil behavior type (SBT) classification chart proposed by Robertson (1990) with a novel
integrated Game Theory-optimization subground stratification model (denoted herein as RIGTOSS). The model
has already been verified based on few test sites results from Taiwan and the U.S. Therefore, in this paper, the
RIGTOSS model is developed further, and it is evaluated based on the stratification profiles provided by CPT
experts for the Venetian Lagoon deposits. The test site has been selected because of high variability of CPT
measurements and the thin transient soil layers in the area. The results indicated comparable stratification pro
files from the developed model(s) and the profiles by experts, derived based on field and laboratory tests.
1 INTRODUCTION been recently appealed to researchers, and several

models have been proposed (Wang et al. 2013,
Stratifying and classifying the soil behavior type Ching et al. 2015, Wang et al. 2019). They often
(SBT) based on loads of sampling and laboratory recognize strata based on computing a consistency
testing can be currently in conflict with the sustain factor among the succeeding CPT measurements in
able design of a project. Furthermore, it may not depth. A probable problem of this approach has
finally lead to a desirable stratification- been with the recognition of thin strata. There have
classification profile due to the probable discon been recent improvements, though. On the other
tinuities and soil disturbances in sampling. This hand, several methods are proposed which seek the
problem is more challenging in highly variable and change points in the succeeding CPT measure
stratified soils. Then the thin transient layers may ments, as the boundaries of strata. They may dis
not be determined accurately. A solution can be cover the thin strata better, compared to the former
sought in the advantageous continuous measure approach. Following the latter approach, a novel
ments of cone penetration test (CPT). Although, model is proposed in Tampere University, Finland,
there are already challenges in the interpretation of named herein integrated Game Theory-optimization
the CPT measurements, and consequently in strati subground stratification (-IGTOSS) model (Farhadi
fying soils based on them. & Länsivaara 2021).
In the past, much investigation has triggered the The -IGTOSS model has been previously combined
SBT classification and stratifying soils based on the with the classification chart proposed by Robertson
CPT measurements. Several SBT classification (1990), so called RIGTOSS. In this study, it is devel
charts have been successfully proposed based on oped further, and applied to the CPTu (‘-u’ indicates
data sets from around the world (Douglas 1981, the pore water pressure measured in CPT) measure
Robertson 1990, Schneider et al. 2012, Eslami et al. ments at the highly nonhomogeneous deposits of the
2017); although, in several studies, their applicabil Venetian lagoon, Italy. Herein, the modified model is
ity has been criticized for several soil types and explained briefly, and the resulting stratification pro
geographical regions (Ricceri et al. 2002, Gylland files of the site are compared with the ones suggested
et al. 2017). Besides, stratification models have by experts.
DOI: 10.1201/9781003308829-54
394
2 STRATIFICATION MODEL - Normalized excess pore pressure, Δu2/σ'v0,
which equals BqQt;
The basics of the utilized stratification model are
described in (Farhadi & Länsivaara, 2021). In general, where, qn is the net corrected cone tip resistance,
the proposed model consists of the following steps: σ'v0 is effective vertical stress, qt is total corrected cone
tip resistance, σv0 is total vertical stress, fs is sleeve fric
1. Importing and interpreting CPTu measurements tion, u2 is pore pressure measured at the cone shoulder,
2. Denoising interpreted measurements u0 is the in-situ pore pressure prior to cone penetration,
3. Stratifying and classifying soils based on inte and Δu2 is the excess pore pressure measured at the
grated Game Theory-soil classification charts cone shoulder in penetration. As generally utilized, qt
4. Illustrating stratification profile is the corrected measured cone tip resistance, qc, based
on water content and unequal end effect of the piez
In the Game Theory model, the previously opti ometer: qt= qc+u2(1-a), where, a is the cone area ratio.
mized parameters are utilized herein as well. The For computation of the in-situ vertical effective
steps 1 and 3 are briefly explained below. stress, σ'v0, unit weight of soil, γ, is computed based
on the equation by Robertson & Cabal (2010), which
2.1 Classification charts and data interpretation provides a continuous profile in depth:
The previously published model contained only the
Fr-Qt Robertson SBT classification chart (1990). It is
developed in this study, and currently, three other
CPTu-based classification charts are implemented in
the model. Hence, the modified model consists of:
where, friction ratio, Rf, equals (fs/qt)×100, γw is the
a. RIGTOSSFr-Qt (sub)model: stratification based on unit weight of water in same unit as γ, and pa is the
the Fr-Qt classification chart, proposed by atmospheric pressure in the same unit as for qt.
Robertson (1990).
b. RIGTOSSBq-Qt (sub)model: stratification based
on the Bq-Qt classification chart, proposed by 2.2 Digitized classification charts
Robertson (1990). In order to implement the charts in computations,
c. SIGTOSSFr-Qt (sub)model: stratification based on
different equations have been fitted to the boundary
the Fr-Qt classification chart, proposed by lines of each classification chart. The fitted equations
Schneider et al. (2012). for the Fr-Qt chart of Robertson (1990) is previously
d. SIGTOSS(Δu2/σ'v0)-Qt (sub)model: stratification
published in Farhadi & Länsivaara (2021). The fitted
based on the (Δu2/σ'v0)-Qt classification chart, equations for the other used charts are presented in
proposed by Schneider et al. (2012). Figures 1-3.
The initial R/S letters of the mentioned (sub)models In Figure 1, the SBT zones in the Bq-Qt chart of
names represent either of the employed charts pro Robertson (1990) are defined as:
posed by Robertson (1990) or Schneider et al. (2012). 1. Sensitive, fine-grained soils
In the classification charts, several normalized 2. Organic soils and peat
parameters are used, which are interpreted from the 3. Clays (clay to silty clay)
CPTu measurements as: 4. Silt mixtures (silty clay to clayey silt)
5. Sand mixtures (sandy silt to silty sand) or
- Normalized cone tip resistance, Qt: cemented soil
6. Sand (silty sand to clean sand)

7. Sand to gravelly sand
8. Sand (clayey sand to ‘very stiff’ sand)
9. Very stiff, fine-grained, overconsolidated
Note that the SBTs of 8 and 9 only exist in the Fr-
- Friction ratio, Fr:
Qt chart, not in the Bq-Qt chart.
Figures 2-3 illustrate the classification charts pro
posed by Schneider et al. (2012) and the fitted equa
tions to the boundary lines. In these two charts, the
SBTs are defined as:
1a. Low-IR clays (IR=G/Su; where, IR, G and Su
- Pore pressure ratio, Bq: represent rigidity index, shear modulus, and
undrained strength, respectively)
1b. Clays
1c. Sensitive clays
3. Silts and transitional soils
395
2. Essentially drained sands and sand mixtures
3 TEST SITE
The data of the CPTu measurements are derived

from a long-lasting project in the Venetian lagoon
basin, Treporti test site, Italy. The importance of
studying the Treporti site originated from the regional
land subsidence in 1970s, and designing submersible
gates to protect Venice from recurrent flooding in
1990s. There exist predominantly Pleistocene silty
sediments of the Venetian lagoon basin with a high
variability of strata (Tonni & Gottardi 2019), which
can be a challenging soil condition to stratify.
Figure 2. The equations fitted to the boundary lines of the

Fr-Qt classification chart proposed by Schneider et al.
(2012). The boundary lines numbers are shown in grey.
Figure 1. The equations fitted to the boundary lines of the

Bq-Qt classification chart proposed by Robertson (1990).
The boundary lines numbers are shown in grey.
A sandy loading bank was constructed for study

ing the compressibility behavior of silty deposits at Figure 3. The equations fitted to the boundary lines of the
the site. For details of the Treporti site refer to Tonni (Δu2/σ’v0)-Qt classification chart proposed by Schneider
& Gottardi (2011). Several CPTu tests were per et al. (2012). The boundary lines numbers are shown in
formed before and after construction of the loading grey.
396
The solid lines in Figures 5f-5i show the identified
SBT versus depth after using directly the classifica
tion charts of Fr-Qt and Bq-Qt by Robertson (1990)
and Fr-Q and (Δu2/σ’v0)-Q by Schneider et al. (2012),
respectively. It can be observed that the succeeding
data points are highly variable on the charts and too
many SBTs/strata are identified. Thus, it may be
really challenging for a geotechnician to decide on the
number of strata and their boundary depths in highly
heterogeneous soils, such as in the Venetian lagoon.
In this regard, (R/S)IGTOSS models facilitate the
stratification procedure. Their resulting stratification
profiles (illustrated with colored contours) for the test
CPTu14 can be compared with the profiles resulting
from direct use of classification charts in Figures 5f
5i. In the colored stratification profiles, the tone of
colors varies from blue to yellow; where yellow
means the highest probable SBT, and vice versa.
Then, it is observed that after applying the (R/S)
IGTOSS models, less strata can be detected generally.
Figure 5j illustrates the distribution of all CPTu
measurements points on the used classification charts.
As observed, especially in the Fr-Q classification chart
of Schneider et al. (2012), numerous points are located
out of the boundary lines of the chart, indicated by
SBT=‘f’ in Figure 5h. They may result from different
factors, such as uncertainties in measurements, inter
pretation methods and parameters, or incompatibility
of the chart with the soils at the Treporti site. How
ever, Figures 5f-5i unveil that after application of
(R/S)IGTOSS models, a large number of the data
points close to the boundary lines have been regarded
Figure 4. Location of CPTu tests at Treporti test site within the zones. It evidences that the model considers
(TTS), from Tonni & Gottardi (2011). The diameter of the the proximity of the succeeding points in detecting
loading bank is 40 m. strata.
bank. In this study, the CPTu measurements before 4.2 Spatial variability of sediments
the bank construction are utilized, which are indi
cated as ‘Phase 2’ in Figure 4. Figure 6 illustrates the highly horizontal variabil
ity of deposits in the Venetian lagoon for tests
CPTu11, CPTu12 and CPTu13, located around
4 RESULTS AND DISCUSSION the perimeter of the loading bank. Although the
(R/S)IGTOSS models have identified the strata
based on each CPTu (Figures 6b-6m), it is still
4.1 Stratification profiles
challenging to find similar layers with almost the
The (R/S)IGTOSS models were applied to nine same boundary depths at the test site. This is due
CPTu measurements: CPTu 11-CPTu 19. to the highly spatial variability of soils at the
In the first step, data were imported into the Treporti site.
model and were interpreted.
In the second step, the interpreted data were
denoised using the locally estimated scattered 4.3 Comparison of classification charts
smoothing (LOESS) method. The results of the Despite the spatial variability of deposits at the Tre
smoothing for the four interpreted parameters of test porti site, and highly alternation of different grain-
CPTu14 (performed at the center of the loading sized sediments, Gottardi & Tonni (2005) reported
bank) are presented in Figures 5a-5d. As can be the following strata (as illustrated in Figures 5e
observed, the benefit of the smoothing has been and 6a):
denoising the outliers (for instance, the abrupt large
fluctuations at depths of 10-17 m), such that the gen - very soft silty clay, from ground level to 2 m in
eral trends of variations would be preserved. depth,
397
Figure 5. Illustration of CPTu measurements, experts stratification profile, RIGTOSS and SIGTOSS stratification profiles,
and distribution of CPTu measurements on classification charts, for test CPTu14: a-d) smoothed versus unsmoothed inter
preted CPTu parameters, e) expert-based stratification reported for Treporti site (Tonni & Gottardi 2011), f-i) directly chart-
based stratification profile, presented by solid line, versus the profiles by the RIGTOSSFr-Qt, RIGTOSSBq-Qt, SIGTOSSFr-Q
and SIGTOSS(Δu2/σ'v0)-Q models, respectively, and, j) distribution of measurements points on the classification charts (the
color of data points gets darker with depth). In (R/S)IGTOSS stratification profiles, SBTs of ‘10ʹ and ‘f’ mean that the data
points located out of the boundaries of the classification charts.
Figure 6. The expert-based stratification profile (a) by Tonni & Gottardi (2011) compared with the profiles from RIGTOSS
and SIGTOSS models for tests: b-e) CPTu11, f-i) CPTu12, and, j-m) CPTu13. Every four profiles, i.e. a-e, f-i, and j-m, are
derived from RIGTOSSFr-Qt, RIGTOSSBq-Qt, SIGTOSSFr-Q and SIGTOSS(Δu2/σ'v0)-Q, respectively.
- medium-fine sand (to approximately 8 m in - silty sand,

depth), - alternate layers of silty sand, sandy silt, and
- silt with thin layers of sandy to clayey silt, from clayey silt, with occasional presence of peat, at
8 to 20 m in depth, depths greater than 24 m.
- dense clean sand (interbedded within the silty
unit, though not everywhere), Figures 5-6 indicate highly variable behaviour of
soil at depths less than 2 m. However, they are
398
mostly silt, or silty clays, which is almost similar to REFERENCES
the observation of Gottardi & Tonni (2005).
At the depth range of approximately 2-8 m, an Ching, J., Wang, J.-S., Juang, C.H. & KU, C.-S. 2015. Cone
almost homogeneous layer is identified with SBT of 6, penetration test (CPT)-based stratigraphic profiling using
that means sand (silty sand to clean sand) according to the wavelet transform modulus maxima method. Canadian
the Robertson chart, and based on the Schneider et al. Geotechnical Journal 52: 1993–2007.
Douglas, B. 1981. Soil classificaion using electric cone
chart, SBT is 2, that means essentially drained sands
penetrometer. Symp. on Cone Penetration Testing and
and sand mixtures. Although the SBT from two charts Experience, Geotech. Engrg. Div., 1981. ASCE, 209–227.
are almost similar, the thickness of the layer based on Eslami, A., Alimirzaei, M., Aflaki, E. & Molaabasi, H. 2017.
each chart is a little different. For example, Figure 6h Deltaic soil behavior classification using CPTu records—
shows SBT of 2 from surface to approximately Proposed approach and applied to fifty-four case histories.
7.5 m deep, while Figure 6f indicates SBT of 6 for Marine Georesources & Geotechnology 35: 62–79.
1-8.5 m deep. Farhadi, M.S. & Länsivaara, T. 2021. Development of an
From 8 to 40 m in depth, highly variable deposits integrated game theory-optimization subground stratifi
are identified. Their SBTs based on the Robertson cation model using cone penetration test (CPT)
measurements. Engineering with Computers: 116.
charts are 3 or 4 and at some depths it is 5. Generally,
https://doi.org/10.1007/s00366-020-01243-0
they range from clay/silty clays to sandy silts/silty Gottardi, G. & Tonni, L. 2005. The Treporti test site:
sands. On the other hand, the chart by Schneider et al. Exploring the behaviour of the silty soils of the Venetian
(2012) indicates almost all SBT classes from 1b to 3, lagoon. Proceedings of the 16th International Confer
which means from clays and sensitive clays to drained ence on Soil Mechanics and Geotechnical Engineering,
sands and sand mixtures. Therefore, using each chart 2005. IOS Press: 1037–1040.
leads to different profiles of stratigraphy. Such obser Gylland, A.S., Sandven, R., Montafia, A., Pfaffhuber, A.A.,
vations may lead to the necessity of a site-specific clas Kåsin, K. & Long, M. 2017. CPTU classification dia
sification chart, such as the Bq-Fr chart by Ricceri grams for identification of sensitive clays. Landslides in
Sensitive Clays: 57–66. Springer.
et al. (2002).
Ricceri, G., Simonini, P. & Cola, S. 2002. Applicability of
piezocone and dilatometer to characterize the soils of
the Venice Lagoon. Geotechnical & Geological Engin
5 CONCLUSIONS eering 20: 89–121.
In this study, a stratification-classification model is penetration test. Canadian Geotechnical Journal 27:
modified and employed for the highly variable 151–158.
deposits of Venetian lagoon, Treporti site, Italy. Robertson, P.K. & Cabal, K. 2010. Estimating soil unit
The model classified soil behavior based on four weight from CPT. 2nd International Symposium on
Cone Penetration Testing, 2010: 2–40.
classification charts. It was observed that the
Schneider, J.A., Hotstream, J.N., Mayne, P.W. &
model was capable of stratifying highly variable Randolph, M.F. 2012. Comparing CPTU Q–F and
deposits. Generally, the obtained stratification pro Q–Δu2/σ′v0 soil classification charts. Géotechnique Let
files have been comparable with the profile pro ters 2: 209–215.
vided by experts, which have been not only based Tonni, L. & Gottardi, G. 2011. Analysis and interpretation
on CPTu tests, but also based on the field and of piezocone data on the silty soils of the Venetian
experimental tests. lagoon (Treporti test site). Canadian Geotechnical Jour
The differences between the model and experts nal 48: 616–633.
stratification profiles originate from numerous fac Tonni, L. & Gottardi, G. 2019. Assessing compressibility
characteristics of silty soils from CPTU: lessons learnt
tors, such as the incompatibility of the classification
from the Treporti Test Site, Venetian Lagoon (Italy).
charts with the testing site condition, uncertainties AIMS Geosci 5: 117–144.
in measurements, or different criteria in classifying Wang, H., Wang, X., Wellmann, J.F. & Liang, R.Y.
soils in contrast to those of the available classifica 2019. A Bayesian unsupervised learning approach
tion charts. for identifying soil stratification using cone penetra
tion data. Canadian Geotechnical Journal 56:
1184–1205.
ACKNOWLEDGEMENTS Wang, Y., Huang, K. & Cao, Z. 2013. Probabilistic identifi
cation of underground soil stratification using cone
The study was supported partly by a grant from penetration tests. Canadian Geotechnical Journal 50:
Tekniikan Edistämissäätiö for the first author. 766–776.
399
© 2022 the Author(s), ISBN 978-1-032-31259-0
Application of two novel CPTu-based stratification models
M.S. Farhadi & T. Länsivaara

Faculty of Built Environment, Department of Civil Engineering, TERRA Research Center
J.-S. L’Heureux & T. Lunne

Norwegian Geotechnical Institute, Geotechnics and Natural Hazards, Oslo, Norway
ABSTRACT: Two novel CPTu-based stratification models are applied to the measurements at four Norwe
gian sites, including a broad range of soil behavior types (SBTs). The two models, called RIGTOSS and SIG
TOSS, are developed based on the recently proposed integrated Game Theory-optimization subground
stratification (-IGTOSS) model. The RIGTOSS model includes two submodels classifying the SBTs based on
either the Fr-Qt or Bq-Qt charts by Robertson (1990); and similarly, the SIGTOSS model includes two submo
dels classifying the SBTs based on either the Fr-Qt or (Δu2/σ'v0) Qt charts by Schneider et al. (2012). The result
ing stratification profiles of the four submodels were compared with the ones provided by experts, derived from
extensive field and experimental testing. Similarity was observed and the results are promising. Differences at
this stage can be attributed to incompatibility of the classification charts with the Norwegian soils and more spe
cifically soil heterogeneities at the Øysand test site.
1 INTRODUCTION classification charts with computational methods for

stratifying soils based on CPTu measurements
A significant advantage of cone penetration test is the (Ching et al. 2015, Wang et al. 2019, Shuku et al.
continuity of measurements in depth. This valuable 2020).
feature can contribute to a better understanding of sub- In Tampere university, Tampere, Finland, a novel
ground, which may not be obtained only from sam CPT-based stratification-classification model, named
pling. As an example, it can lead to one of the main -IGTOSS here, is proposed integrating several com
applications of CPTu testing, which is the stratification putational models: a Game Theory model, the evolu
and classification of the soil behaviour type (SBT). tionary Grey Wolf Optimizer (GWO) and the
Currently, the stratification-classification needs com classification chart proposed by Robertson in 1990
bining CPTu measurements with experimental and (Farhadi & Länsivaara 2021). It was previously
probably other field tests. However, it is desirable to evaluated based on several CPTu tests in Taiwan and
stratify and classify soils only based on the CPTu the US. In this study, the model is developed further
measurements, which is targeted in this study. including three other SBT classification charts. Then,
The CPTu-based stratification-classification has it is applied to several CPTu tests from Norway con
been extensively investigated. Initially, a generally taining a broad range of SBTs.
applicable method was sought to classify soils based
on the CPTu measurements, which resulted into sev
eral classification charts (Robertson 1990, Schneider 2 -IGTOSS MODEL
et al. 2012, Eslami et al. 2017). Despite the break
throughs, the observed inaccuracies of the charts for 2.1 Normalized CPTu parameters
different sites have left the question open, and
researchers are still proposing soil-type- or site- The utilized normalized parameters interpreted from
specific classification charts (Ricceri et al. 2002, the CPTu measurements are (Schneider et al. 2012):
Gylland et al. 2017). However, using only the classi i) Normalized cone tip resistance, Qt:
fication charts could not provide accurate stratifica
tion profiles while they represented largely high
number of strata especially in highly variable soils.
Therefore, several researchers focused on combining
DOI: 10.1201/9781003308829-55
400
ii) Friction ratio, Fr: As illustrated in Figure 1, in the first step, the
CPTu data are imported and interpreted. A constant
soil unit weight, γ, for the whole depth of cone pene
tration is used to interpret Qt, Fr, Bq and Δu2 =σ0 v0 .
In the second step, the interpreted data is
smoothed using a spatial regression method, called
iii) Pore pressure ratio, Bq: locally estimated scatterplot smoothing (LOESS).
1% of data is used as the spatial regression range.
In the third step, the integrated stratification-
classification model specifies the strata boundaries
depths, using a Game Theory model, as described in
Farhadi & Länsivaara (2021). Then, the SBT is deter
mined from a classification chart, in each of the fol
iv) Normalized excess pore pressure, Δu2/σ’v0, lowing submodels:
which equals BqQt;
a) RIGTOSSFr-Qt based on the Fr-Qt chart by Robert
where, qn is the net corrected cone tip resistance, son (1990),
is the effective vertical stress, σv0 is the total verti b) RIGTOSSBq-Qt based on the Bq-Qt chart by
cal stress, qt is the total corrected cone tip resistance, Robertson (1990),
fs is sleeve friction, u2 is the pore pressure measured c) SIGTOSSFr-Qt based on the Fr-Qt chart by Schnei
at the cone shoulder, u0 is the in-situ pore pressure der et al. (2012),
prior to cone penetration, and Δu2 is the excess pore d) SIGTOSSðΔu2 =σ0 v0 Þ - Qt based on the
pressure measured at the cone shoulder in penetration. Qt chart by Schneider et al. (2012).
As generally utilized, qt is the measured cone tip The introductory R or S letters in the names of the
resistance, qc, corrected based on water content and RIGTOSS and SIGTOSS models indicate the util
unequal end effect of the piezometer: qt= qc+u2(1-a), ized classification charts.
where, a is the cone area ratio. In the charts by Robertson (1990), the SBTs are
defined as:
2.2 Developed stratification model
1. Sensitive, fine-grained
The basics of the stratification model are described 2. Organic soils - peats
in Farhadi & Länsivaara (2021). In this study, it has 3. Clays - clay to silty clay
been developed further containing three other classi 4. Silt mixtures - clayey silt to silty clay
fication charts. A concise flowchart of the modified 5. Sand mixtures-silty sand to sandy silt
model is presented in Figure 1. 6. Sands, clean sand to silty sand
7. Gravelly sand to sand
8. Very stiff sand to clayey (heavily overconsoli
dated or cemented) sand
9. Very stiff fine-grained
In the charts by Schneider et al. (2012), the SBTs
are defined as:
1a. Low-IR clays (IR=G/Su; where, IR, G and Su
represent rigidity index, shear modulus, and
undrained strength, respectively)
1b. Clays
1c. Sensitive clays
3. Silts and transitional soils

2. Essentially drained sands and sand mixtures
In the final step, the identified SBTs of the strata
are plotted three-dimensionally with colored con
tours, as explained in Farhadi & Länsivaara (2021).
The interested readers may contact the first author
for the MATLAB code; which will be probably
described further in a separate paper in future.
3 TEST SITES
Figure 1. Concise flowchart of the proposed (R/S)ITGOSS The developed (R/S)IGTOSS models are applied to
model(s). the CPTu data from four sites in Norway (Figure 2):
401
The Øysand site consists of a 20 m thick glacio
fluvial mostly sandy deposit. The site includes sev
eral strata of gravelly sand (fluvial deposit), fine silty
sand (deltaic soils), and clay-and-silt. The sand layer
includes fine to medium uniform sand, predominantly
of quartz minerals, some plagioclase and micas.
The Tiller-Flotten site consists of marine and gla
ciomarine sediments with a thick layer of sensitive
clay from 8 m below terrain.
4 RESULTS AND DISCUSSION
4.1 Denoising data

Figure 3 shows the interpreted smoothed Fr, Qt, Bq
and Δu2 =σ0 v0 parameters for tests HALC19 and
HALC20. It can be observed that the major variations
are preserved after smoothing, and the minor sharp
Figure 2. Location of the test sites in Norway (provided by variations are smoothly approximated. Thus, the
MATLAB, hosted by Esri, and Google Earth Pro). smoothing rarely impacts the resulting stratification
profile while the model finds the strata boundaries
Table 1. Soil properties of the test sites, in brief (Blaker
et al. 2019, Gundersen et al. 2019, Quinteros et al. 2019,
and L’Heureux et al. 2019).
Halden Onsøy Øysand T-F*
Property min max min max min max min max
wn** (%) 20 35 45 65 12 33 30 53
γðkN=m3 Þ 18.3 20.8 15 19 13 23 16.8 19.1
FC (%) 15 99
CC (%) 2 17 45 68 0 18 40 70
LL (%) 27.5 37.5 46 77 27 53
Ip (%) 6 13 25 50 7 29
St 5 19 0 360
OC (%) 30 50
SC (g/L) 8 36 2 3
* T-F stands for Tiller-Flotten.

** wn: natural water content; γ: unit weight; FC: fines con
tent; CC: clay content; LL: liquid limit; Ip: plasticity index;
St: sensitivity; OC: organic content; and SC: salt content.
Figure 3. Illustration of the smoothing impact on the param
eters interpreted from the CPTu tests at the Halden site.
a silt site at Halden, a soft clay site at Onsøy,
a medium dense sand site at Øysand, and a quick mostly based on the large change points. However,
clay site at Tiller-Flotten (L’Heureux & Lunne 2019). most importantly it removes the outliers in data, such
The Halden site consists of a natural fjord marine as those caused by stopping the cone penetration.
deposit including mostly a low plasticity silt.
A majority of bulky angular grains was observed at the
scanning electron microscope (SEM) test: 41% quartz 4.2 Stratification-classification profiles
and 42% feldspar (Carroll & Paniagua 2018). Water
table was 2.5 m below the surface. Further soil proper 4.2.1 Halden
ties for all sites are presented in summary in Table 1. Figure 4 illustrates the stratification-classification
The Onsøy site consists of a marine clay deposit profiles for test HALC19.
with similar behaviour as that observed in Canada, Figures 4a-4d show that the interpreted param
Japan, Southeast Asia, Sweden, and Finland. The site eters look similar before and after smoothing;
has a thick layer of uniform very soft to soft clay. although, several abrupt changes are removed. For
402
Figure 4. Illustration of CPTu measurements, experts stratification profile, RIGTOSS and SIGTOSS stratification profiles,
and distribution of CPTu measurements on classification charts, for test HALC19, from Halden investigation site: a-d)
smoothed versus unsmoothed interpreted CPTu parameters, e) expert-based stratification reported for Halden site (Blaker
et al. 2019), f-i) directly chart-based stratification profile, presented by a solid line, versus the profiles provided by the
RIGTOSSFr-Qt, RIGTOSSBq-Qt, SIGTOSSFr-Qt and SIGTOSSðΔu2=σ0 v0Þ-Qt models, respectively, and, j) distribution of meas
urements points on the classification charts (the color of data points gets darker with depth). In (R/S)IGTOSS stratification
profiles, SBTs of ‘10ʹ and ‘f’ mean that the data points located out of the boundaries of the classification charts. [In profiles,
the yellow color shows higher probability of the SBT.].
example, in Figure 4a, two sudden changes at (Figures 4f-4g). Thinner layers can be
approximate depths of 0.5 and 1.4 m are removed observed with different probabilities at these thick
after smoothing. strata.
Figure 4e illustrates the experts’ judgement based Figure 4j illustrates the location of the CPTu
on the in-situ and laboratory tests performed by Nor measurements on the classification charts. The colour
wegian Geotechnical Institute (NGI). Four main soil of the points gets darker with depth. They show the
deposits are identified: a loose to medium dense silty variability of the soil at the Halden site. In addition,
clayey sand from 0 to 5 m (Unit I), two clayey silt comparing them with Figures 4f-4i showed that the
layers with the same geologic origin and almost the (R/S)IGTOSS profiles are correctly determined.
same material (Units II and III), and a deeper unit con
sisting of medium stiff clay (Unit IV) (Blaker et al. 4.2.2 Onsøy
2019). Figure 5 shows the (R/S)IGTOSS results for tests
In Figures 4f-4i, the solid lines show the stratifica ONSC19, ONSC20, and ONSC21, performed at the
tion profiles after using directly the classification south-east corner of the Onsøy site. The distances
charts. A large number of thin layers can be identi between their locations were less than 2.1 m.
fied with dissimilar SBTs, compared with the (R/S) Figures 5a-5d indicate the impact of the
IGTOSS contoured profiles (Figures 4f-4i) and the smoothing method, and an appropriate repeatabil
experts’ profile (Figure 4e). The identification of ity of CPTu tests; although, the ONSC20 measure
numerous strata originates from the location of the ments deviate from the other two tests at some
CPTu measurements points on a classification chart; depths.
which may be located close to the SBT boundary Comparing Figures 5e with two sets of
lines, but on different sides. Figures 5f-5i and 5j-5m indicates a generally appro
The colour-contoured profiles in Figures 4h-4i priate similarity in recognizing the soil type as clay;
show that the SIGTOSSFr-Qt and SIGTOSS- although, the strata identified by experts are more
models indicate three thick strata. than the strata identified by the (R/S)IGTOSS
However, the RIGTOSSFr-Qt and RIGTOSSBq-Qt models. On the other hand, several thin layers are
models indicate approximately 5 thick strata identified by the (R/S)IGTOSS models.
403
Figure 5. Comparison of the CPTu measurements at the Onsøy test site: a-e) smoothed and unsmoothed measurements of tests
ONSC19, ONSC20, and ONSC21, f-i) respectively, RIGTOSSFr-Qt, RIGTOSSBq-Qt, SIGTOSSFr-Qt and SIGTOSS(Δu2/σ’v0)-Qt pro
files for ONSC19, and, j-m) respectively, RIGTOSSFr-Qt, RIGTOSSBq-Qt, SIGTOSSFr-Qt and SIGTOSS(Δu2/σ’v0)-Qt profiles for test
ONSC20.
Figure 6. Comparison of (a) experts’ stratification profile Figure 7. Comparison of (a) experts’ stratification profile
with the (b) RIGTOSSFr-Qt, (c) RIGTOSSBq-Qt, (d) with the (b) RIGTOSSFr-Qt, (c) RIGTOSSBq-Qt, (d)
SIGTOSSFr-Qt, and (e) SIGTOSSðΔu2=σ0 v0Þ-Qt models pro SIGTOSSFr-Qt, and (e) SIGTOSSðΔu2=σ0 v0Þ-Qt models pro
files, for test OYSC40, Øysand site. In the experts’ profile, files, for test TILC17, Tiller-Flotten site (L’Heureux et al.
Unit I is sand (silty, fine; layers and seams of medium to 2019).
coarse sand), Unit II is sand (fine to coarse, gravelly), Unit
III is sand (fine, traces of organic material), Unit IV is sand
(fine to medium, silty), Unit V is sand (medium to coarse, judgement at Øysand; although, the SBT characteriza
gravelly), Unit VI is sand (fine to medium, silty), Unit VII tion is quite similar.
is silt (sandy, clayey), and Unit VIII is sand (fine to
medium, silty) (Quinteros et al. 2019). 4.2.4 Tiller-Flotten
Figure 7 compares the SBT profiles by the experts
(L’Heureux et al. 2019) and (R/S)IGTOSS models
4.2.3 Øysand for test TILC17. It is interesting that only
Figure 6 shows the experts’ stratification profile (Quin the SIGTOSSðΔu2=σ0 v0Þ-Qt model has clearly recog
teros et al. 2019) compared with the (R/S)IGTOSS pro nized the sensitive clays. Albeit, in the
files for test OYSC40. The experts recognized several SIGTOSSFr-Qt model the SBT has been mainly
strata of sandy soils for the whole depth, except for either 1c, sensitive clay, or 3, i.e. silts and transi
Unit VII, which is mainly silty. Similarly, the (R/S) tional soils. Besides, the RIGTOSS model could not
IGTOSS models identified mostly sand mixtures; recognize the sensitive clays, i.e. SBT = 1, well;
except SIGTOSSFr-Qt that identified silts and transi however, it found them as clays or silt mixtures. Fur
tional soils, i.e. SBT=3, but at depths different from ther, it can be observed that the strata change depths
Unit VII. In general, the strata boundaries recognized are recognized differently by the models and the
by the models differ significantly from the experts’ experts.
404
4.2.5 Comparison of classification charts caused by the incompatibility of the chart with the
Comparing the RIGTOSSFr-Qt and RIGTOSSBq-Qt Norwegian soils.
profiles in Figures 4-7 for each test, indicates almost
similar stratification profiles. However, differences
can be due to the different numbers of SBT zones on ACKNOWLEDGEMENT
the classification charts, and different interpreted
parameters used in each of them. The authors acknowledge the support from Norwe
Similarly, comparing the SIGTOSSFr-Qt and gian GeoTest Sites infrastructure (NGTS) for access
SIGTOSSðΔu2=σ0 v0Þ-Qt profiles in Figures 4-7 indi to data. The study was also supported partly by
cates almost similar profiles with respect to the Tekniikan Edistämissäätiö (Finnish Foundation for
strata boundaries depths; but the SBT differences Technology Promotion), Grant No. JNR. 8375, for
are visible. Besides, data points may be the first author.
located out of the boundaries of the
SIGTOSSðΔu2=σ0 v0Þ-Qt chart, i.e. SBT=‘f’ in the REFERENCES
profiles. This may originate from several factors,
such as the incompatibility of the chart with the Blaker, Ø., Carroll, R., Paniagua Lopez, A.P., Degroot, D.
Norwegian soils, using different interpretation J. & L’Heureux, J.-S. 2019. Halden research site: geo
methods in this study in contrast to the ones technical characterization of a post glacial silt. AIMS
used in Schneider et al. (2012), and the small Geosciences 5(2): 184–234. doi: 10.3934/
zone of ‘1c’ on the chart. geosci.2019.2.184.
Carroll, R. & Paniagua, P. 2018. Variable rate of penetra
Figures 4-7 indicated that judging based on
tion and dissipation test results in a natural silty soil.
the Schneider’s chart might have advantage over Cone Penetration Testing 2018. CRC Press.
using the Robertson chart for the Norwegian Ching, J., Wang, J.-S., Juang, C.H. & Ku, C.-S. 2015. Cone
fine-grained soils; although, it can be confusing penetration test (CPT)-based stratigraphic profiling
to differentiate clays from silts based on the using the wavelet transform modulus maxima method.
Schneider’s chart. Canadian Geotechnical Journal 52(12): 1993–2007.
On the other hand, comparing the RIGTOSS and https://doi.org/10.1139/cgj-2015-0027.
SIGTOSS profiles, less strata are determined based Eslami, A., Alimirzaei, M., Aflaki, E. & Molaabasi, H.
on the charts proposed by Schneider et al. (2012), 2017. Deltaic soil behavior classification using
CPTu records—Proposed approach and applied to
while they include less SBT zones. Then, it may
fifty-four case histories. Marine Georesources &
affect the precision of stratification and the recog Geotechnology 35: 62–79. doi: 10.1080/
nized number of strata; which can be observed in 1064119X.2015.1102185.
Figures 5 and 6, as examples. Farhadi, M.S. & Länsivaara, T. 2021. Development of
an integrated game theory-optimization subground
stratification model using cone penetration test
5 CONCLUSIONS (CPT) measurements. Engineering with Computers.
https://doi.org/10.1007/s00366-020-01243-0.
A stratification-classification model by Farhadi & Gundersen, A., Hansen, R., Lunne, T., L’Heureux, J.-S.
& Strandvik, S.O. 2019. Characterization and engin
Länsivaara (2021) is developed further in this study
eering properties of the NGTS Onsøy soft clay site.
to include different CPTu-based soil classification AIMS Geosciences 5: 665–703. doi: 10.3934/
charts. The developed models are applied to four geosci.2019.3.665.
Norwegian soils. As observed, the stratification pro Gylland, A.S., Sandven, R., Montafia, A.,
files of the models are comparable with the profiles Pfaffhuber, A.A., Kåsin, K. & Long, M. 2017.
presented by experts –which resulted from extensive CPTU classification diagrams for identification of
field and laboratory experiments. However, there are sensitive clays. Landslides in Sensitive Clays.
still differences between them, principally originat Springer.
ing from: L’Heureux, J.-S., Lindgård, A. & Emdal, A. 2019. The
Tiller–Flotten research site: Geotechnical characteriza
- using different soil classification viewpoints or tion of a very sensitive clay deposit. AIMS Geosciences
parameters implemented in the available classifi 5(4): 831–867. doi: 10.3934/geosci.2019.4.831.
cation charts, L’heureux, J.-S. & Lunne, T. 2020. Characterization and
- incompatibility of the used classification charts engineering properties of natural soils used for
geotesting. AIMS Geosciences 6(1): 35–53. doi:
with the Norwegian soils, especially for fine- 10.3934/geosci.2020004.
grained soils, and, L’Heureux, J.-S., Carroll, R., Lacasse, S., Lunne, T.,
- spatial variability of soils, resulting into almost dif Strandvik, S.O., Degago, S., Instanes, A., Nordal, S. &
ferent stratification profiles for the CPTu tests per Sinitsyn, A. 2017. New Research Benchmark Test Sites
formed close to each other. in Norway. Geotechnical Frontiers 2017.
Quinteros, S., Gundersen, A., L’Heureux, J.-S., Carraro, A.H.
In addition, it was observed that the CPTu meas & Jardine, R. 2019. Øysand research site: Geotechnical
urements points may be located out of the bounds of characterisation of deltaic sandy-silty soils. AIMS Geo-
the Schneider’s Δu2 =σ0v0 - Q chart, which could be sciences 5(4): 750–783. doi: 10.3934/geosci.2019.4.750
405
Ricceri, G., Simonini, P. & Cola, S. 2002. Applicability Letters 2(4): 209–215. https://doi.org/10.1680/
of piezocone and dilatometer to characterize the geolett.12.00044.
soils of the Venice Lagoon. Geotechnical & Geo Shuku, T., Phoon, K.-K. & Yoshida, I. 2020. Trend estima
logical Engineering 20: 89–121. doi: 10.1023/A: tion and layer boundary detection in depth-dependent
1015043911091. soil data using sparse Bayesianlasso. Computers and
Robertson, P.K. 1990. Soil classification using the Geotechnics 128, 103845. https://doi.org/10.1016/j.
cone penetration test. Canadian Geotechnical compgeo.2020.103845.
Journal 27(1): 151–158. https://doi.org/10.1139/ Wang, H., Wang, X., Wellmann, J.F. & Liang, R.Y. 2019.
t90-014. A Bayesian unsupervised learning approach for identify
Schneider, J.A., Hotstream, J.N., Mayne, P.W. & ing soil stratification using cone penetration data. Can
Randolph, M. F. 2012. Comparing CPTU Q–F and adian Geotechnical Journal 56(8): 1184–1205. doi:
Q–Δu2/σv0′ soil classification charts. Géotechnique 10.1139/cgj-2017-0709.
406
© 2022 the Author(s), ISBN 978-1-032-31259-0
Numerical simulation of CPT in sands using DeltaSand and Hardening

Soil models
M. Fetrati
MARUM–Center for Marine Environmental Sciences, University of Bremen, Bremen, Germany
V. Galavi
Deltares, Delft, The Netherlands
M. Goodarzi
COWI A/S, Hamburg Branch Office, Germany
S. Kreiter
T. Mörz
Geo-Engineering.org GmbH, Bremen, Germany
ABSTRACT: In this paper, Hardening Soil model with small-strain stiffness (HSsmall) and Delta-
Sand model are used as constitutive soil models to simulate the Cone Penetration Test (CPT) using
the Material Point Method (MPM). Both models are formulated within the double hardening frame
work, in which independent yield surfaces represent mechanical behavior of soil under deviatoric and
volumetric loadings. DeltaSand is a new advanced state-dependent constitutive model in which the
relative density is incorporated in the formulation to represent the mechanical behavior of soil under
deviatoric and volumetric loading in different stresses and relative densities. The numerical simulations
are compared with CPTs in Cuxhaven Sand in a calibration chamber. Both constitutive soil models are
compared with each other and DeltaSand is found to be capable to capture the soil behavior during
quasi-static CPT.
1 INTRODUCTION Martinelli & Galavi 2021). Each of these approaches

has its pros and cons. For instance, investigation of
The Cone Penetration Test (CPT) is a widespread the interactions of the soil particles with each other
tool in both onshore and offshore for soil character and with the cone is possible through DEM simula
ization, (Lunne 2010) due to repeatability, accuracy, tion in a precise manner. However, a high number of
and simplicity. Simulating CPT, as a mean to further particles and, followed by that, a significant amount
interpret and understand the CPT results, is challen of time is required for a realistic boundary value
ging because of the large deformation near the cone problem. FEM has the ability to model realistic
and complex behavior of soil under various type of boundary value problems and lower computational
loading. Different numerical approaches have been time but suffers from mesh distortion in the simula
used to simulate CPT, considering the large deform tion of large deformation problems. On the other
ation calculations, such as the Particle Finite Elem hand, MPM discretizes the continuum medium in
ent Method in Geomechanics (G-PFEM) (Monforte material mass points and has the ability to simulate
et al. 2017, 2018, Schweiger & Hauser 2021, Carbo large deformations’ problems without mesh distor
nell et al. 2022), Finite Difference Method (FDM) tion. MPM utilization in different geotechnical prob
(Moug et al. 2019), Finite Element Method (FEM) lems is comparatively new (Coetzee et al. 2005,
(Goodarzi et al. 2018), Discrete Element Method Zabala & Alonso 2011, Sołowski & Sloan 2015,
(DEM) (Khosravi et al. 2020), and the Material Goodarzi & Rouainia 2017, Ghasemi et al. 2018,
Point Method (MPM) (Ghasemi et al. 2018, Galavi et al. 2019).
DOI: 10.1201/9781003308829-56
407
In addition to the numerical approach to simulate In standard MPM, MPs inside elements are used
CPT, employing an appropriate constitutive soil as integration points. Therefore, the accuracy of the
model to predict the soil behavior in different soil integration depends on the location and numbers of
states is indispensable. Various soil models have MPs. The Gauss integration method, in contrast, util
been used to predict the soil behavior in CPT simula izes a fixed number of integration points in
tions, such as Mohr-Coulomb (Huang et al. 2004), a perfectly suited location to obtain accurate integra
Drucker–Prager (Susila & Hryciw 2003), Hardening tion similar to FEM (Beuth & Vermeer 2013). Beuth
Soil (HS) (Tolooiyan & Gavin 2010), and Hypoplas et al. (2007) suggested a combination of both inte
tic with intergranular strain (Fan et al. 2018). How gration methods to mitigate stress oscillations in
ever, either they need recalibration for different elements due to MPs crossing between elements.
relative densities or face several issues pertaining to This is based on an assumption that stresses of MPs
the numerical stability. in a fully-filled linear element are identical, and
In this paper, the Material Point Method, along with therefore, stress calculation and integration can be
Hardening Soil model with small-strain stiffness (HS performed on one Gauss Point in the elements. This
small) and DeltaSand model (Galavi 2021), is used. numerical scheme leads to less computational effort
The latter model is a new advanced state-dependent than the standard material point integration (Marti
constitutive soil model, which is an elasto-plastic con nelli & Galavi 2021). This mixed integration method
stitutive model formulated based on the double harden is used in this study.
ing framework. The simulation outcomes are In problems in which the inertia effect is insignifi
compared with CPT tests in the MARUM Calibration cant, mass scaling can reduce the computational cost
Chamber (MARCC) in order to evaluate the capability by scaling the density. This scheme increases the
pffiffiffiffiffiffiffi
of the new constitutive model in capturing the soil time step size by a factor of MS:
response throughout the simulation of CPT.
2 NUMERICAL MODEL
2.1 MPM formulation

where Δtcrit is the critical time step; Lmin is the smal
The benefits of combined point-based and mesh-
lest element length; E is young modulus; ρ is density,
based strategies in MPM is the avoidance of mesh
and MS is the mass scaling factor. Sensitivity ana
distortion and collecting the history of simulation in
lysis is required to determine the amount of mass
material points. (Sulsky et al. 1994).
scaling factor such that it does not lead to inaccurate
The entire domain of the problem is covered by
results. More information about the MPM formula
computational mesh similar to FEM. Additionally
tion can be found in (Martinelli & Galavi 2021).
the continuum body is divided into a set of Material
Points (MPs), which can relocate in the whole com
putational mesh contrary to the fixed position of 2.2 Constitutive soil models
Gaussian integration points within elements in FEM.
In the material point method, the momentum bal This paper uses two constitutive soil models, HS-
ance equations are solved in the computational mesh small and DeltaSand, to model the cone penetration
nodes, and constitutive equations and mass conser into the soil. The list of parameters for the HS-small
vations are solved in MPs. The required information model and their values for the Cuxhaven sand for dif
for solving balance equations in the mesh is trans ferent relative densities are given in Table 1 and 2.
ferred from MPs to the degrees of freedoms at nodes The full description of HS-small model can be
in the beginning of each calculation step. After solv found in (Benz 2007). The parameters of the HS-small
ing the balance equations at nodes, the position of model are calibrated for the Cuxhaven Sand based on
MPs, acceleration, and velocity at the MPs will be laboratory experiments (Kluger 2014). Since the state
updated. The computational grid does not store any dependency is not implemented in the HS-small
permanent information; therefore, all nodal data are model, unrealistically high dilation during shearing
discarded at the end of each calculation step. might occur. The maximum dilation angle is, therefore,
A contact algorithm is employed to model the decreased to stop unrealistic volume increase during
frictional contact between the cone and the soil. In extensive deformations (Martinelli & Galavi 2021).
order to enhance the precision of the contact algo DeltaSand captures the soil behavior under volu
rithm and simulate the cone penetration into the soil, metric and deviatoric loading in different soil states
the moving mesh concept (Al-Kafaji 2013) is util (Galavi 2021). Deviatoric, volumetric, and tension
ized here. In this concept, the deformable mesh is cut-off yield surfaces are incorporated in the model.
located between the fixed boundary and the moving The deviatoric part in DeltaSand includes three
mesh part, adjusting its dimensions with time. The deviatoric yield surfaces such that two of which are
refined mesh at the contact will always stay around implemented to reproduce the cyclic behavior of
the cone in this concept. sands. An extended Marsouka-Nakai criterion
408
Table 1. Description of HS-small model parameters.
parameters Units Description
Eref ½kPa] The secant stiffness from triaxial test where q and p´ are Von Mises deviatoric stress
50
when the ratio of deviatoric stress to and mean effective stress, respectively; pc is
the deviatoric stress at failure is 0.5 effective pre-consolidation stress; Mv is a shape
Eref
oed ½kPa] The tangent stiffness obtained from factor that is acquired from a 1D consolidation
oedometer tests loading. A full explanation of the constitutive
E50
ur ½kPa] Unloading/reloading stiffness relations of the DeltaSand model can be found in
ur Unloading/reloading Poisson’s ratio (Galavi 2021).
Effective friction angle
½kPa] Effective cohesion
ψ Dilation angle 2.3 The schematization of CPT in the calibration
m Rate of stress dependency in stiffness chamber
behavior
The calibration chamber was developed in the
γ0:7 Shear strain at which the shear modulus
decreases to almost 70% of its initial Marine Engineering Geology working group of
value the Center for Marine Environmental Sciences
Gref ½kPa] Shear stiffness at small strains (MARUM), University of Bremen (Fleischer et al.
0
Rf Failure ratio qf =qq (ratio of deviatoric 2016, Stähler et al. 2018). MARCC samples have
stress at the failure to an asymptotic a diameter of 30 cm and a height of 55 cm.
value of deviatoricstress) A piezocone with a diameter of 12 mm is used in the
chamber. This chamber has the capability of con
ducting the tests under “BC5” boundary conditions
(Fleischer et al. 2016).
Table 2. HS-small model parameters.
The numerical model geometry used as equivalent
Relative Densities (%) to MARCC BC5 (simulated field boundary condition)
is shown in Figure 1. BC5 presumes that the specimen
parameters Units 83.48 85.63 86.85 92.53 is in an infinite soil volume with known horizontal
stiffness (Kluger et al. 2021). The ratio of chamber
Eref
50 83000 85500 86500 92500 diameter to cone diameter in the numerical model is
Eref
oed 83000 85500 86500 92500 125 for field conditions to avoid boundary effects.
E50 249000 256500 259500 277500 A thin layer of 5 cm is modeled with a linear elastic
ur
ur 0.2 0.2 0.2 0.2 material to simulate the vertical effective stresses in
41.4 41.7 41.8 42.4 experiments. Vertical effective stress equals 200 kPa,
1 1 1 1 and the K0 value is 0.45 as it was in the laboratory
ψ 0.01 0.01 0.01 0.01 experiments. Due to the symmetry of the geometry,
m 0.439 0.432 0.429 0.411 a 2D axisymmetric material point formulation (Galavi
γ0:7 0.00012 0.00011 0.00011 0.00011 et al. 2019) is employed. The meshes are fine near the
cone, and their size increase as they get close to the
Gref
0 116700 118200 119000 122900 far-field boundary. Quadrilateral elements are used in
Rf 0.896 0.893 0.891 0.884 all simulations.
The mini-cone is modeled as a rigid body that
penetrates into the soil under drained condition
(Matsuoka & Nakai 1974) is used for the deviatoric with a prescribed velocity of 2 cm/s like in the
yield surfaces. The hardening-softening rule that is laboratory experiments. The initial position of
used in this model is given in Eq. (2): the cone is below the soil surface, and it is
pushed into the soil until it reaches a steady-
state for measuring the cone resistance qc. The
contact friction angle of 15 degrees is chosen as
the cone surface is smooth (Al-Mhaidib 2005).
At the beginning of the simulations, the MPs’
where fd ðeÞ is a function to connect the0 hardening positions are defined uniformly in rectangular
softening rule to the current void ratio; ’i and are clusters in the whole domain. This uniform dis
initial effective and critical friction angles, respect tribution is independent of the elements’ size
ively; A is a parameter to control plastic stiffness. and type. In order to avoid the formation of
The volumetric section of DeltaSand is an ellip unrealistic empty elements, the number of mater
tical yield surface with an associated flow rule. This ial points was chosen so that they fill all elem
yield surface is expressed as follows: ents throughout the penetration.
409
Table 3. DeltaSand parameters.
parameters Unit Description
fGref ½-] Reference elastic stiffness

n ½-] Power stiffness parameter
hs ½stress] Hardness parameter
Coed ½-] Oedometer parameter
α ½-] Parameter to control peak friction
angle
β ½-] Parameter to control volumetric
behavior
Amat ½-] Plastic stiffness parameter
emin ½-] Minimum void ratio
ecrit ½-] Critical void ratio
emax ½-] Maximum void ratio
½� ] Friction angle in constant volume
½-] Ratio of stiffness in small strains to
medium strains
Figure 1. Schematic illustration of the model geometry. γr ½-] Reference shear strain at
Yellow: cone, green: soil, blue: elastic layer as a surcharge. Gratio ¼ 0:722
*
Gratio ¼ G0 =G
3 RESULTS
3.1 Calibration of DeltaSand Table 4. Set of parameters used in all numerical

simulations.
Prior to the simulation of CPT in the calibration
chamber, the parameters of the DeltaSand model for
Cuxhaven sand were obtained from experiments.
The oedometer tests with four relative densities and
920 0.3 30 1.1 1.65 0.95 0.016 0.47
triaxial tests with three relative densities are used to
cover medium-dense to dense sand samples (Kluger
2014). In this study, the kinematic hardening parts of
the model, implemented in the model to capture
the mechanical behavior of soil under cyclic 0.8 0.85 32 5 0.0001
loadings, were ignored. Therefore, only param
eters related to the isotropic hardening parts of
the model are calibrated and presented here.
A brief description of DeltaSand parameters is
presented in Table 3. The calibration procedure
was conducted using the SoilLab feature of
PLAXIS (Plaxis 2021). The results of the calibra
tion are depicted in Figure 2 and Figure 3, and
the final set of parameters is given in Table 4.
Detailed explanations and steps for calibrating the
DeltaSand model are represented in (Galavi
2021).
3.2 Simulation results vs. laboratory tests

In this section, the capabilities of the constitutive
soil models are shown by comparing the numerical
outcomes with laboratory test results. The same
geometry, boundary conditions, number of elements,
number of material points are used for comparing
the constitutive soil models. The steady-state cone
resistance was reported here obtained from aver Figure 2. Simulation and experimental results of oed
aging the values of cone resistances for depths ometer tests on loose to dense sand (Solid lines: experi
between 18-32 cm, similar to the experiments. ment; dashed lines: simulations).
410
Figure 5. Mesh size effect on qc-Dr results (Black:
Dr = 83.48%, red: Dr = 86.85%, and blue: Dr = 92.53%).
Figure 3. Simulation (DeltaSand and HSsmall) and experi

mental results of drained monotonic triaxial tests on
medium dense and dense sand.
3.2.1 DeltaSand constitutive model

A sensitivity analysis to identify the optimum mesh
size was conducted for DeltaSand. The results of
simulations for fine, medium-fine, and coarse mesh
Figure 6. Comparison of experimental and numerical

results using DeltaSand and HSsmall model (Upper bound
obtained using φpeak and lower bound using φcritical in
HSsmall model).
were compared to each other. The obtained averaged

qc regarding depths between 18-32 cm for fine and
coarse mesh are within 2% and 5% of the ones
obtained with fine mesh. The number of elements for
coarse mesh is 929 and for medium-fine and fine
mesh are 2463 and 3680, respectively. The average
element size exactly beneath the tip for coarse mesh
is 4 mm and for medium-fine and fine mesh are
1.7 mm and 1.1 mm, respectively. Therefore, the
medium-fine is an optimum choice for analysis, con
sidering that the results do not differ tremendously
Figure 4. Different mesh size adjacent to cone.
411
from the ones obtained from finer mesh. Moreover, ACKNOWLEDGMENT
using medium-fine mesh is computationally more
efficient. Figure 4 shows different mesh sizes adja The authors would like to acknowledge the support
cent to the cone. Figure 5 illustrate the results of qc of the project “VCPTu2PDA”, FKZ: 03EE3025A by
in different depth for different mesh sizes. Besides, the Federal Ministry for Economic Affairs and
the experiment result for relative density 83.48% is Energy (BMWi) and the MARUM—Center for
included in this figure. Marine Environmental Sciences, University of
Figure 6 illustrates the qc – Dr curve for both Bremen. The help and assistance of Geo
numerical and experimental data under BC5 condi Engineering.org GmbH, RWE Renewables GmbH,
tion. The obtained results with DeltaSand match Deltares and COWI A/S as project partners are also
well with the experimental data. A set of parameters greatly appreciated.
is used for all relative densities, which demonstrates
the capability of DeltaSand.
REFERENCES
3.2.2 HS-small constitutive model
Al-Kafaji, I. K. J. (2013) Formulation of a Dynamic Mater
The input parameters for HS-small model are cali ial Point Method (MPM) for Geomechanical Problems,
brated based on relative density for the Cuxhaven University of Stuttgart.
sand and were presented in Table 2 and Figure 3. Al-Mhaidib, A. I. (2005) ‘Shearing rate effect on interfacial
Figure 3 represents the calibration results based on friction between sand and steel’, in The Fifteenth Inter
two different friction angles- peak and critical fric national Offshore and Polar Engineering Conference.
tion angles are shown by solid and dashed lines, Benz, T. (2007) Small strain stiffness of soils and its conse
respectively. Figure 6 compares the obtained qc – quences. Universität Stuttgart.
Dr curve from numerical modeling with laboratory Beuth, L., Benz, T., Vermeer, P. A., Coetzee, C. J.,
tests under the BC5 condition. Contrary to the Delta- Bonnier, P. and Van Den Berg, P. (2007) ‘Formulation
and validation of a quasi-static material point method’,
Sand model, the Hardening Soil has no state-
in Proceedings of the 10th international symposium on
dependent parameter. It should be noted that the numerical methods in Geomechanics (NUMOG),
simulations are run with two different friction Rhodes, Greece, pp. 189–195.
angles, the critical friction angle as lower bound and Beuth, L. and Vermeer, P. A. (2013) ‘Large deformation
the peak friction angle upper bound, while other analysis of cone penetration testing in undrained clay’,
parameters are constant. The experimental data are Installation effects in geotechnical engineering. Taylor
located between the lower and upper bound numer \& Francis Group London.
ical simulations. Another shortcoming of the HS- Carbonell, J. M., Monforte, L., Ciantia, M. O., Arroyo, M.
small soil model is that different sets of parameters and Gens, A. (2022) ‘Geotechnical particle finite elem
ent method for modeling of soil-structure interaction
have to be used for each relative density.
under large deformation conditions’, Journal of Rock
Mechanics and Geotechnical Engineering. Elsevier.
Coetzee, C. J., Vermeer, P. A. and Basson, A. H. (2005)
4 CONCLUSION ‘The modelling of anchors using the material point
method’, International journal for numerical and ana
In this study, the cone penetration in a small volume lytical methods in geomechanics. Wiley Online Library,
calibration chamber MARCC was simulated using 29(9), pp. 879–895.
two different constitutive soil models, namely Hard Fan, S., Bienen, B. and Randolph, M. F. (2018) ‘Stability
ening Soil with small-strain stiffness and DeltaSand. and efficiency studies in the numerical simulation of
cone penetration in sand’, Geotechnique Letters, 8(1),
Prior to simulations, material parameters were
pp. 13–18. doi: 10.1680/jgele.17.00105.
calibrated for both constitutive soil models based on Fleischer, M., Kreiter, S., Mörz, T. and Huhndorf, M.
the laboratory tests. The results showed that Delta- (2016) ‘A small volume calibration chamber for cone
Sand can capture the sand behavior better than the penetration testing (CPT) on submarine soils’, in Sub
HSsmall. marine Mass Movements and their Consequences.
The result of simulations showed that the Delta- Springer, pp. 181–189.
Sand model can capture the soil behavior around the Galavi, V. (2021) ‘DeltaSand: A state dependent double
cone for different relative densities with one set of hardening elasto-plastic model for sand: Formulation
parameters. Simulating CPT with HS-small soil and validation’, Computers and Geotechnics, 129(April
2020). doi: 10.1016/j.compgeo.2020.103844.
models, however needs recalibration of material
Galavi, V., Martinelli, M., Elkadi, A., Ghasemi, P. and
parameters for different relative densities and neither Thijssen, R. (2019) ‘Numerical simulation of impact
calibration of the model at peak strength nor at crit driven offshore monopiles using the material point
ical state strength could reproduce the experimental method’, Proceedings of the XVII ECSMGE - Geotech
results for a complex problem like CPT. Moreover, nical Engineering foundation of the future, (September).
the effect of element size on simulation outcomes doi: 10.32075/17ECSMGE-2019-0758.
was analyzed in this study. The simulations showed Ghasemi, P., Calvello, M., Martinelli, M., Galavi, V. and
that the mesh size has a negligible effect on the Cuomo, S. (2018) ‘MPM simulation of CPT and model
results for fine and medium-fine meshes, while calibration by inverse analysis’, Cone Penetration Test
ing 2018 - Proceedings of the 4th International
a coarse mesh can affect the results.
412
Symposium on Cone Penetration Testing, CPT 2018, pp. Finite Element Method (PFEM)’, Computers and Geo
295–301. technics, 82, pp. 144–156. doi: 10.1016/j.
Goodarzi, M. and Rouainia, M. (2017) ‘Modelling Slope compgeo.2016.08.013.
Failure Using a Quasi-static MPM with a Non-local Monforte, L., Arroyo, M., Carbonell, J. M. and Gens, A.
Strain Softening Approach’, Procedia Engineering. (2018) ‘Coupled effective stress analysis of insertion
Elsevier B.V., 175, pp. 220–225. doi: 10.1016/j. problems in geotechnics with the particle finite element
proeng.2017.01.015. method’, Computers and Geotechnics. Elsevier, 101, pp.
Goodarzi, M., Stähler, F. T., Kreiter, S., Rouainia, M., 114–129.
Kluger, M. O. and Mörz, T. (2018) ‘Numerical simulation Moug, D. M., Boulanger, R. W., DeJong, J. T. and
of cone penetration test in a small-volume calibration Jaeger, R. A. (2019) ‘Axisymmetric Simulations of
chamber: The effect of boundary conditions’, in Cone Cone Penetration in Saturated Clay’, Journal of Geo-
Penetration Testing 2018. CRC Press, pp. 309–315. technical and Geoenvironmental Engineering, 145(4), p.
Huang, W., Sheng, D., Sloan, S. W. and Yu, H. S. (2004) 04019008. doi: 10.1061/(asce)gt.1943-5606.0002024.
‘Finite element analysis of cone penetration in cohesion- Plaxis (2021) ‘PLAXIS 2D Manual’. Bentley Systems.
less soil’, Computers and Geotechnics, 31(7), pp. Schweiger, H. F. and Hauser, L. (2021) Numerical Simula
517–528. doi: 10.1016/j.compgeo.2004.09.001. tion of CPT with the Clay and Sand Model (CASM)
Khosravi, A., Martinez, A. and Dejong, J. T. (2020) ‘Dis Including Effects of Bonding, Lecture Notes in Civil
crete element model (DEM) simulations of cone pene Engineering. Springer International Publishing. doi:
tration test (CPT) measurements and soil classification’, 10.1007/978-3-030-64514-4_11.
Canadian Geotechnical Journal, 57(9), pp. 1369–1387. Sołowski, W. T. and Sloan, S. W. (2015) ‘Evaluation of
doi: 10.1139/cgj-2019-0512. material point method for use in geotechnics’, Inter
Kluger, M. O. (2014) Hypoplastizität: Von der Parameter national Journal for Numerical and Analytical Methods
bestimmung bis zur numerischen Modellierung eines in Geomechanics. Wiley Online Library, 39(7), pp.
Cuxhavener Sandes. Universität Bremen. 685–701.
Kluger, M. O., Kreiter, S., Stähler, F. T., Goodarzi, M., Stähler, F. T., Kreiter, S., Goodarzi, M., Al-Sammarraie,
Stanski, T. and Mörz, T. (2021) ‘Cone penetration tests D., Stanski, T. and Mörz, T. (2018) ‘Influences on
in dry and saturated Ticino sand’, Bulletin of Engineer CPT-results in a small volume calibration chamber’, in
ing Geology and the Environment. Bulletin of Engineer Proceedings of China-Europe Conference on Geotech
ing Geology and the Environment, (2006). doi: 10.1007/ nical Engineering, pp. 730–733.
s10064-021-02156-y. Sulsky, D., Chen, Z. and Schreyer, H. L. (1994) ‘A particle
Lunne, T. (2010) ‘The CPT in offshore soil method for history-dependent materials’, Computer
investigations-a historic perspective’, Proc. CPT, 10, pp. methods in applied mechanics and engineering. Else-
71–113. vier, 118(1–2), pp. 179–196.
Martinelli, M. and Galavi, V. (2021) ‘Investigation of the Susila, E. and Hryciw, R. D. (2003) ‘Large displacement
Material Point Method in the simulation of Cone Pene FEM modelling of the cone penetration test (CPT) in
tration Tests in dry sand’, Computers and Geotechnics. normally consolidated sand’, International Journal for
Elsevier Ltd, 130(November), p. 103923. doi: 10.1016/j. Numerical and Analytical Methods in Geomechanics, 27
compgeo.2020.103923. (7), pp. 585–602. doi: 10.1002/nag.287.
Matsuoka, H. and Nakai, T. (1974) ‘Stress-deformation and Tolooiyan, A. and Gavin, K. (2010) ‘Finite element ana
strength characteristics of soil under three different prin lysis of the CPT for design of bored piles’, in 2nd Inter
cipal stresses’, in Proceedings of the Japan Society of national Symposium on Cone Penetration Testing,
Civil Engineers, pp. 59–70. Huntington Beach, CA, USA.
Monforte, L., Arroyo, M., Carbonell, J. M. and Gens, A. Zabala, F. and Alonso, E. E. (2011) ‘Progressive failure of
(2017) ‘Numerical simulation of undrained insertion Aznalcóllar dam using the material point method’, Géo
problems in geotechnical engineering with the Particle technique. Thomas Telford Ltd, 61(9), pp. 795–808.
413
© 2022 the Author(s), ISBN 978-1-032-31259-0
CPT calibration in centrifuge: Effect of partial saturation on cone resistance

V. Fioravante
Department of Engineering, University of Ferrara, Italy
D. Giretti
Department of Engineering and Applied Science, University of Bergamo, Italy
E. Dodaro, C.G. Gragnano & G. Gottardi

Department of Civil, Chemical, Environmental and Materials Engineering, University of Bologna, Italy
ABSTRACT: When dealing with unsaturated soil conditions, the influence of matric suction on cone tip
resistance of the soil above the ground water table is typically neglected in engineering practice, with conse
quent possible misinterpretation of soil features. In the last decades, various researchers have investigated the
influence of suction on cone resistance for sands, whilst still little is known for silty materials, whose contri
bution can be significant and extended for many meters above the groundwater table. Such issue is especially
relevant for compacted earth structures, like river embankments, typically made of a heterogeneous mixture
of intermediate soils. With the aim of providing a contribution and stimulating its correct implementation into
geotechnical practice, a set of miniature piezocone tests have been carried out in a centrifuge on both saturated
and partially saturated silty sand models. The interpretation of CPT results is discussed, highlighting the
effect of partial saturation on cone tip resistance.
1 INTRODUCTION contribution of suction to the effective stress is fre

quently neglected, due to difficulties in assessing the
Due to its reliability and time and cost effectiveness, in-situ moisture content and in obtaining a reliable pore
the cone penetration test (CPT) represents a valuable water pressure distribution, often resulting in exces
tool for continuous stratigraphy profiling and geotech sively conservative design approaches and in the incor
nical soil properties estimation. Most of the existing rect evaluation of soil features (Russell and Khalili
approaches for analyzing CPT results are based on 2006). Various studies have been recently carried out
fully saturated or dry conditions, for which interpret to gain insights on the influence of unsaturated condi
ation methods are well established and have a solid tions on CPT data, mostly limited to sandy soils
theoretical background (Robertson and Campanella (Hryciw and Dowding 1987; Bolton et al. 1999; Rus
1983a, b; Lunne et al. 1997; Mayne 2007; Robertson sell et al. 2010; Pournaghiazar et al 2013; Jarast and
2009). However, in several cases CPT soundings may Ghayoomi 2018), all showing evidence that the cone
cross a vadose zone, conventionally extended from the penetration resistance, qc, can be significantly increased
ground level to the water table, where partially satur by suction. Conversely, limited research has been car
ated soil states are very likely to occur. Thus, a reliable ried out on CPTs in unsaturated silty materials (Silva
interpretation of CPT data in unsaturated soil layers is and Bolton 2005; Tan 2005; Yang and Russell 2016),
of pivotal importance for the design, optimization and due to the intrinsic complexity related to partial drain
management of the engineering works interacting with age, occurring during penetration at the standard rate
soils at shallow depths (e.g. foundations, road pave of 20 mm/s (Paniagua et al. 2014), and to the microfab
ments) or influenced by infiltration, evapo-transpiration ric of intermediate soils. Most of these studies have
and transient groundwater flow, such as river embank been performed on reconstituted samples, under the
ments, earth dams or backfill of retaining walls. controlled laboratory environment of calibration cham
As observed by Yang and Russell (2016), for an bers or centrifuges, in order to eliminate the typical
accurate analysis of CPT results in partially saturated uncertainties related to soil heterogeneity and param
soils, the in-situ stress state needs to be accurately eters estimation, while only few field tests have been
evaluated, taking into account the variations with depth performed so far to evaluate the effect of soil moisture
of the matric suction and the effective degree of satur content on the cone resistance, qc (Lehane et al. 2004;
ation of soils. In the current practice, instead, the Collins and Miller 2014; Giacheti et al. 2019). The
DOI: 10.1201/9781003308829-57
414
present paper aims at contributing to a better under section containing a second 9.8 kN load cell, used to
standing of the effect of matric suction on CPT results measure tip resistance plus sleeve friction. In addition,
interpretation. For this purpose, a set of piezocone tests the cone has a 35-bar capacity Druck PDCR pressure
have been carried out on a compacted mixture of sand transducer for interstitial pressure measurements.
and finer material in the 240 g-ton geotechnical centri Physical models were reconstituted in layers of pre
fuge facility at the Experimental Institute for Geotech scribed height to obtain a 1g dry density of 90%
nical Modelling (Italian acronym: ISMGEO) of Seriate of γd,max and using an initial water content of
(Bergamo, Italy), in both saturated and partially satur about 17%. The container was a cylindrical box,
ated conditions. The use of monitoring sensors allows 400 mm in diameter. With a ratio D/d (D is the
to clearly define the pore pressure distribution of the container diameter) equal to about 36, boundary
models under different water table depths. Further side effects were minimized. During the reconsti
more, implications of using various assumptions on the tution, pore pressure transducers (ppts M, P, Q,
calculation of the effective stress states during penetra N, R) and tensiometers (tens 1, 2, 3) were
tion, starting from matric suction measures, are prelim embedded in the model at prescribed heights
inarily discussed and CPT calibration in the centrifuge (Figure 1) and at a distance of 50 mm from the
is attempted. box axis. Once the total height was achieved, the
soil saturation was completed applying to the
model a continuous vacuum pressure of about -70
2 EXPERIMENTAL CAMPAIGN kPa for 12 hours. Then a rigid frame which
holds a linear displacement transducer to monitor
2.1 Equipment, tested material and test procedure
A scheme of models tested is presented in Figure 1.
The testing soils are Ticino sand (TS, Baldi et al.,
1982, 1986, Fioravante, 2000, Jamiolkowski et al.,
2003, Fioravante & Giretti, 2016) and Pontida clay
(PON, Ventini et al. 2021). TS is a coarse to medium,
uniform silica sand, of alluvial origin, mainly com
posed by angular grains; PON is a low plasticity kaoli
nitic clayey silt, deposited in a post-glacial lake
environment. The overall experimental campaign has
been performed considering different mixtures of TS
and PON. However, in this contribution, only the
results of a test carried out on a mixture of 85% by
weight of TS and 15% by weight of PON are dis
cussed. The main physical properties of the mixture,
obtained from an accurate laboratory characterization,
are listed in Table 1. In particular, the minimum and
maximum dry density have been obtained following
ASTM 4254 - Method A (2016) and ASTM D1557
12e1 (2012) - Modified Proctor method, respectively.
Figure 2 compares the grain size distribution of TS,
PON and the mix.
Table 1. Main physical properties of the mixed soil 85%

TS+15%PON.
SOIL Gs d50 CU CC γd,min γd,max

kN/ kN/
- mm - -
m3 m3
85%TS+15%
2,695 0,499 13,2 10,9 13,92 18,15
PON
The ISMGEO miniaturized piezocone used in the

tests has a diameter d = 11.3 mm and a total cone
area of 100.3 mm2. It incorporates a 60° cone tip with
a load cell to measure tip forces up to 9.8 kN and Figure 1. Sketch of the cylindrical strongbox containing
a 36.9 mm long shaft, which connects to an upper details of geometry, transducers and in-flight miniature probe.
415
10 d of penetration from the ground level were
removed (Gui and Bolton 1998). The plot shows the
variation with the dimensionless depth (i.e. the ratio
between penetration depth, z, and cone diameter, d)
of the sleeve friction resistance, fs, and of the cor
rected cone tip resistance, qt, this latter expressed as:
where qc is the measured cone tip resistance, a =

0.785 is the net area ratio, determined from labora
tory calibration, and u2 is the pore pressure gener
ated during cone penetration and measured just
behind the cone. In addition, the pore pressure
values measured by ppts (in the positive range) and
tensiometers (both in positive and negative ranges)
are also plotted, with circles and triangles, respect
ively, together with u2 data. The dimensionless
depth of the water level has been determined from
Figure 2. Grain size distribution of the tested soils. the measures of the ppt placed at the bottom of the
the soils surface settlement, the miniaturized

piezocone and the actuator was fixed to the top
of the container.
The container was loaded onto the centrifuge and
accelerated to the target (50g at the soil surface). After
the in-flight consolidation, a first CPTU was carried out
in the central axis of the model. At the end of penetra
tion, the penetrometer was lifted, the centrifuge was
stopped, the equipment was moved 70 mm from the
original position and the model was re-accelerated.
When the pore pressure equilibrium was achieved
again, as identified by real-time pore pressure monitor
ing data, an outflow was imposed to the soil model by
opening a hydraulic valve placed at the bottom of the
cylindrical box. The outflow was interrupted when the
water table reached almost mid depth. Following pore
pressure stabilization, a second CPTU was carried out
in a model partially saturated in the upper part and sat
urated below, with matric suction and pore pressure
data continuously recorded. It has to be noticed that,
due to settlement induced by saturation, centrifuge
accelerations and desaturation of the model, the soil
sample underwent a progressive increase in density,
with an average void index value equal to 0.480 (at the
beginning of the test in saturated conditions) and to
0.475 (at the beginning of the test in partially saturated
conditions).
3 CENTRIFUGE TEST RESULTS
The data recorded during the piezocone advance

ment in both fully saturated soil conditions (continu
ous lines) and with the phreatic surface below the Figure 3. Variation with the dimensionless depth, z/d, of
ground level (dotted lines) are presented in Figure 3. the sleeve friction resistance, fs, of the pore pressure gener
To take into account the progressive mobilization of ated during cone penetration, u2, together with the pore
the cone resistance from the free model surface pressure measurements (circles and triangles) and of cor
(Schmertmann, 1978), the data registered in the first rected cone tip resistance, qt.
416
model; for the test in unsaturated conditions, its σ’v, the pore pressure, u0, and the degree of saturation
value, z/dw,UNSAT, is equal to 18.7 and is drawn in are plotted with dimensionless depth in Figure 4. Con
bold hatched line (Figure 3). sidering a constant 0.5 value of the degree of saturation
The ppts and tensiometers monitoring data fit quite (UNSAT,1), a noticeable increase in the effective stress
well the u2 measured during penetration, showing distribution is produced. Instead, due to the significant
a hydrostatic distribution of pore pressures in the posi percentage of sand in the tested material, the degree of
tive range of values in both experiments (SAT and saturation tends to rapidly reduce with the increase of
UNSAT) and a less than hydrostatic distribution in the matric suction (absolute) values, evidencing in case
unsaturated area (UNSAT model), highlighting that the UNSAT,2, a limited impact on the effective stress as
process of lowering the water table led to a hydraulic calculated with equation (2). It is now possible to try to
equilibrium of pore pressure in the saturated soil area determine general trends in the soil response to CPT
also for the unsaturated test. On the other hand, distri advancement.
butions of qt and fs show significant differences
between the two tests, i.e. higher cone tip resistance
and sleeve friction in the area where desaturation
occurred, where the suction effect is clearly tangible. It
should be noticed that, since the cone tip and sleeve
friction resistances vary with the relevant overburden
stress for the same material, the CPT data plotted in
Figure 3 require a stress normalization for a proper
interpretation and comparison of resistance profiles.
4 STRESS NORMALIZATION
Stress conditions are significantly different in the

two tests. The change in the water table depth, in
fact, leads to a variation of the saturation level and
produces suction states in the partially saturated soil
in the UNSAT experiment and subsequent effects on
the test results. At least two independent stress state
variables should be considered for an accurate
description of the relevant soil behaviour (i.e. Mor
genstern, 1979; Fredlund et al., 2012); however, it is
also well known that the most commonly used CPT
charts and correlations are based on a single-valued
effective stress approach. To show the implications
of adopting a single-valued effective stress approach,
the Bishop’s effective stress equation (1959) for
unsaturated soils has been here used, such as:
with (σ – ua) being the net stress, Sr the degree of satur

ation and (ua – uw) the matric suction. For such test
interpretation, it is therefore required the knowledge of
the distribution of Sr and of (ua – uw), typically not
readily available in traditional engineering applications.
Two different assumptions have been thus made here,
with a first simplified case (UNSAT,1) considering
a hydrostatic distribution of pore pressure above the
groundwater line and a constant degree of saturation
equal to 0.5, while a second case (UNSAT,2) aims at
representing the experimental conditions closely, with
a matric suction distribution determined on the base of
tensiometer measurements and a variable degree of sat Figure 4. Calculated values of total, σv, and effective, σ’v,
uration, provided by the water retention curve obtained stresses acting on vertical direction, pore pressure, u0, and
from physical soil properties through the procedure degree of saturation, Sr, are plotted with dimensionless depth,
suggested in Aubertin et al. (2003). For both cases, the z/d, under a simplified assumption (UNSAT,1 top graph) and
calculated values of total stress, σv, and effective stress, based on tensiometer data (UNSAT,2 bottom graph).
417
According to Robertson’s (2009) very popular However, for the test performed after the water
unified approach, the normalized cone resistance, table lowering, significant variations can be found in
Qtn, and the Soil Behavior Type index, Icn, are calcu the values of Qtn when comparing the advancement
lated using a stress exponent, n, that varies with in the saturated versus the unsaturated zones. For
soil type and stress level. Specifically: dimensionless depths lower than z/dw,UNSAT (18.7),
none of the two assumptions on the effective stress
and saturation degree provide uniform Qtn profiles,
despite the material is essentially the same in the two
experiments, showing higher normalized cone resist
ances in the unsaturated with respect to the saturated
zone. Indeed, below the water table, the assumption
based on monitoring data (UNSAT,2) tends to pro
duce similar values to the test performed in fully
saturated conditions, rather than for the case of
a a simplified assumption (UNSAT,1). Analogous
observations can be done for the Icn data, always
within the range of sandy materials (1.31 – 2.05),
but with substantial differences between the satur
where pa and pa2 are reference pressures in the same ated and the unsaturated zones.
units of qc, σv and σ’v, while F is the normalized friction Hence, adopting a single-valued effective stress
ratio. Results obtained from the present CPT tests per approach for stress normalization to interpret the CPT
formed before (SAT) and after (UNSAT) the water table data in the unsaturated zone appears to be not fully
lowering are plotted in Figure 5, in terms of Qtn and reliable.
Icn, considering both the simplified (UNSAT,1) and the
more accurate (UNSAT,2) assumptions for the calcula
tion of stress conditions (see Figure 4). Considering the 5 CONCLUSIONS
test conducted in fully saturated conditions (SAT),
values of Qtn and Icn tend to be relatively constant with Results from small-scale laboratory tests, which
depth and typical of sandy materials, in good agree included CPT execution with pore pressure and suction
ment with Robertson’s approach. measurements in a centrifuge environment, have been
presented herein. The experiments described are
referred to a soil mixture, made of mainly coarse-
grained particles with a limited fine fraction, tested
under different saturation conditions. In fact, the only
variation in the two presented cases was related to their
water table depth, either at almost the ground surface
or at a lower level, obtained through a dewatering pro
cess. Monitoring sensors, pore pressure transducers and
tensiometers, located along the model depth, played an
essential role in determining the soil suction distribu
tion above the water table and in identifying the
hydraulic equilibrium reached at the end of the outflow
phase. Data measured during cone penetration tests (fs,
qt and u2), before and after the dewatering process,
show only limited differences in the saturated zones,
further reduced by adopting the stress normalization of
Robertson’s unified approach.
On the other side, when comparing cone penetra
tion data measured in the unsaturated versus the sat
urated zones, substantial differences in the selected
stress-normalized results (Qtn and Icn), regardless of
the assumption on suction and saturation degree pro
files above the water table, can be detected. There
fore, from the data presented herein, it would appear
that the effect of partial saturation on cone tip resist
ance provided by the matric suction is not duly taken
into account by simply applying a stress normaliza
tion by the Bishop’s equation for unsaturated soils.
Figure 5. Variation with the dimensionless depth, z/d, of In other terms, it seems that the combined use of
the normalized cone resistance, Qtn, and the Soil Behavior a single effective stress variable (as typically
Type index, Icn, calculated using the stress exponent, n. assumed when more specific information on
418
unsaturated soil behavior is lacking) with the stand on Geotechnical Site Characterization. Atlanta, Georgia:
ard CPT charts and correlations cannot produce 1063–1068.
a similarly reliable data interpretation. Additional Hryciw, R.D. & Dowding, C.H. 1987. Cone penetration of
investigations, including the use of other materials, partially saturated sands. Geotechnical Testing Journal. 10
are clearly required to better define such critical (3): 135–141.
issue. Jamiolkowski, M.B., Lo Presti, D.C.F., & Manassero, M.
2001. Evaluation of Relative Density and Shear Strength
from CPT and DMT. ASCE Geotechnical Special Publica
ACKNOWLEDGMENT tion. 119: 201–238.
Jarast, P. & Ghayoomi, M. 2018. Numerical modeling of cone
This research was funded under the scheme for penetration test in unsaturated sand inside a calibration
“Research Projects of National Relevance” (in Italian: chamber. International Journal of Geomechanics. 18
Progetti di Ricerca di Rilevante Interesse Nazionale- (2).10.1061/(ASCE)GM.1943-5622.0001052.
Lehane, B.M., Ismail, M.A. & Fahey, M. 2004. Seasonal
PRIN), Bando 2017, grant number 2017YPMBWJt, dependence of in situ test parameters in sand above the
promoted by the Italian Ministry of Education, Univer water table. Géotechnique. 54(3): 215–218
sity and Research (in Italian: Ministero dell’Istruzione, Lunne, T., Robertson, P.K. & Powell, J. 1997. Cone penetra
dell’Università e della Ricerca-MIUR). tion testing in geotechnical practice. London: Blackie Aca
demic and Professional.
Mayne, P.W. 2007. Cone penetration testing. A synthesis of
REFERENCES highway practice. Washington, D.C.: Transportation
Research Board.
ASTM D1557-12e1. 2012. Standard Test Methods for Morgenstern, N. R. 1979. Properties of compacted soils, Proc.
Laboratory Compaction Characteristics of Soil Using 6th Pan American Conference on Soil Mechanics and
Modified Effort (56,000 658 ft-lbf/ft3 (2,700 kN-m/m3)). Foundation Engineering. Lima, Peru. 3: 349–354.
ASTM International, West Conshohocken, PA. Paniagua, P. & Nordal, S. 2015. Influence of variable rates of
ASTM D4254-16. 2016. Standard test methods for minimum penetration on silt behavior and its SBT classification.
index density and unit weight of soils and calculation of Proc. 15th Pan-American Conference on Soil Mechanics
relative density. ASTM International, West Conshohocken, and Geotechnical Engineering. Buenos Aires, Argentina:
PA 509-516.
Aubertin, M., Mbonimpa, M., Bussière, B. & Chapuis, R.P. Pournaghiazar, M., Russell, A.R. & Khalili, N. 2013. The
2003. A model to predict the water retention curve from cone penetration test in unsaturated sands. Géotechnique.
basic geotechnical properties. Canadian Geotechnical 63(14): 1209–1220.
Journal. 40: 1104–1122. Robertson, P.K. 2009. Interpretation of cone penetration tests -
Baldi, G., Bellotti, R., Ghionna, V., Jamiolkowski, M. & A unified approach. Canadian Geotechnical Journal. 46
Pasqualini, E. 1982. Design parameters for sand from (11): 1337–1355.
CPT. Proc. 2nd European Symposium on Penetration Test Robertson, P.K., and Campanella, R.G. 1983a. Interpretation
ing. Amsterdam: 425-432. of cone penetration tests. Part I: Sand. Canadian Geotech
Baldi, G., Bellotti, R., Ghionna, N., Jamiolkowski, M. & nical Journal. 20(4): 718–733.
Pasqualini, E. 1986. Interpretation of CPTs and CPTU’s. Robertson, P.K., and Campanella, R.G. 1983b. Interpretation
2nd Part. Proc. 4th International Geotechnical Seminar. of cone penetration tests. Part II: Clay. Canadian Geotech
Nanyang Technological Institute, Singapore: 143–156. nical Journal. 20(4), 734–745.
Bishop, A. W. 1959. The principle of effective stress. Teknisk Russell, A.R., and Khalili, N. 2006. On the problem of cavity
Ukeblad. 106(39): 859–863. expansion in unsaturated soils. Computational Mechanics.
Bolton, M.D., Gui, M.W., Garnier, J., Corte, J.F., Bagge, G., 37(4): 311–330.
Laue, J. & Renzi, R. 1999. Centrifuge cone penetration Russell, A.R., Pournaghiazar, M. & Khalili, N. 2010. Inter
tests in sand. Géotechnique. 49(4): 543–552. preting CPT results in unsaturated sands. Proc. 2nd Inter
Collins, R. & Miller, G.A. 2014. Cone penetration testing in national Symposium on Cone Penetration Testing.
unsaturated soils at two instrumented test sites. Proc. 6th Huntington Beach, USA.
International Conference on Unsaturated Soils. Sidney, Schmertmann, J.H. 1978. Guidelines for Cone Penetration
Australia: 1489-1494. Test (Performance and Design). U.S. Department of Trans
Fioravante, V. 2000. Anisotropy of small strain stiffness of portation, Federal Highway Administration. Report
Ticino and Kenya Sands from seismic wave propagation FHWA-TS-, Washington: 787–209.
measured in triaxial testing. Soils and Foundations. 40(4): Silva, M. & Bolton, M. 2005. Interpretation of centrifuge
129–142. piezocone tests in dilatant, low plasticity silts. Proc. of
Fioravante, V. & Giretti, D. 2016. Unidirectional cyclic resist International Conference on Problematic Soils. Cyprus,
ance of Ticino and Toyoura sands from centrifuge cone Greece.
penetration tests. Acta Geotechnica. 11(4): 953–968. Tan, N.K. 2005. Pressuremeter and cone penetrometer testing
Fredlund, D.G., Rahardjo, H., & Fredlund, M.D. 2012. Unsat in a calibration chamber with unsaturated Minco Silt. PhD
urated soil mechanics in engineering practice. New York: dissertation. University of Oklahoma.
John Wiley & Sons. Ventini, R., Dodaro, E., Gragnano, C.G., Giretti, D., &
Giacheti, H., Rodrigues, R., Bezerra, R. & Rocha, B. 2019. Pirone, M. 2021. Experimental and numerical investiga
Seasonal influence on cone penetration test: An unsatur tions of a river embankment model under transient seepage
ated soil site interpretation. Journal of Rock Mechanics conditions. Geosciences. 11(5), 192.
and Geotechnical Engineering. 11(2): 361–368. Yang, H. & Russell, A.R. 2016. Cone penetration tests in
Gui M. W. & Bolton M. D. 1998. Geometry and scale effects unsaturated silty sands. Canadian Geotechnical Journal,
in CPT and pile design, Proc. 1st International Conference 53(3): 431–444.
419
© 2022 the Author(s), ISBN 978-1-032-31259-0
Calibration cone penetration testing in silty soils

V. Fioravante
University of Ferrara, Italy
D. Giretti
University of Bergamo, Italy
T. Lunne & P. Carotenuto

A.H. Augustesen, R.C.J. Lindeboom & L. Krogh

Ørsted, Denmark
H. Dias
Equinor
M.-C. Sougle
Vattenfall
A. Barwise
RWE
S. de Wit
Shell
D. Burbury
Scottish Power Renewables
ABSTRACT: The Offshore Wind Industry’s rapid expansion across the globe requires geotechnical model-
ling of sites that are often characterized by layers of silty sand and silt mixtures. The CPTU is the main in situ
offshore investigation tool for defining the ground conditions and for establishing facility position and soil
parameters for foundation design, but no simple and robust methodologies exist for characterizing transitional
soils. This paper presents some results of CPTUs carried out in a large calibration chamber and in a centrifuge
aimed at contributing to the development of guidelines for planning, specification, execution, and interpret
ation of CPTUs in transitional soils.
1 INTRODUCTION are unexpectedly encountered and not predicted by

the CPTU interpretation. This is mainly due to the
After the installation of the first Offshore Wind Farm lack of robust methodologies for characterizing silty
in Denmark in 1991, the Offshore Wind Industry sand and silt mixtures based on CPTU, as existing
(OWI) has been increasing exponentially in Europe correlations between CPTU parameters and classifi
and recently also in new markets such as the US cation and engineering properties have been devel
East Coast and the Asia-Pacific Countries. oped for sand and clay. Consequently, the cost and
Soil modelling and spatial mapping for the design risk of developing offshore wind farms in these
and installation of foundations for offshore wind tur regions are high.
bines is generally based on the combined and staged The goal of the CSi – CPTU in silty soils – Joint
use of seismic surveys, piezocone penetration testing Industry Project is ultimately to develop guidelines for
(CPTU) and boreholes with soil sampling. However, set-up, execution and interpretation of CPTUs in silty
in many regions where the OWI is expanding, layers soils, see Augustesen et al. (2022). This goal is pur
of transitional soils, i.e. neither clean sand nor clay, sued by a combination of research activities including
DOI: 10.1201/9781003308829-58
420
in situ testing, numerical modelling and using The tests on clean TS11 were meant to assess if
a significant number of CPTUs, carried out in a large CC and CCC cone penetration tests were comparable
calibration chamber and in a geotechnical centrifuge at with each other and with previous studies carried out
the ISMGEO laboratory (Italy, Baldi et al. 1982, Baldi using TS and the same facilities. In addition, they
et al. 1986). were aimed at validating the use of the centrifuge as
With the main aim of highlighting the impact of a calibration tool of CPTUs in sandy soils. Indeed,
fines content on soil strength and stiffness, the tests are centrifuge CPTUs have the advantage, with respect
carried out on a clean sand and on the sands mixed to calibration chamber tests, of giving a qc-profile
with non-plastic fines to obtain 15% and 30% fines over a wide range of vertical stress, rather than
content (grain size < 0.063 mm). Some of the prelimin a single qc value associated to the specific level of
ary results of the calibration chamber and centrifuge the applied stress of a single sample. This is under
tests are presented in this paper, together with a brief the condition that the effects of rigid boundaries and
description of the testing apparatuses and procedures. scale effects are minimised. In addition, CCC models
are smaller and a test requires few days compared to
about two weeks for a CC test. If validated, CCC
2 TESTING SOIL AND PROGRAM CPTUs can be extensively used to explore the effect
of variable density, fine content, stress level, over-
2.1 Ticino Sand and Ticino Filler consolidation ratio on the penetration resistance.
The sand and silty sand used for the experimentation 2.2.1 The ISMGEO calibration chamber
are Ticino Sand (TS) and Ticino Filler (TF). TS is The calibration chamber specimens are 1.4 m high
a clean silica sand used extensively in the past for cali and 1.2 m in diameter. The CC is a flexible-wall
bration chamber, centrifuge and laboratory tests (Baldi chamber and it can impose four different boundary
et al. 1982, 1986, Fioravante 2000, Jamiolkowki et al. conditions (BC):
2003, Fioravante & Giretti 2016). The batch used for
the CSi project is named TS11, which is a natural, • BC1: constant vertical and horizontal stresses, σv
coarse to medium clean sand, with principal compo = const and σh = const;
nents of quartz (36% by weight), feldspar (40%), mica • BC2: zero vertical and horizontal strains, Δεv =
(11%). TF is the natural flour obtained by sieving the Δεh = 0;
coarser fraction of Ticino sand and has similar mineral • BC3: constant vertical stress and zero horizontal
ogical composition (21% quartz, 47% feldspar, 16% strain, σv = const and Δεh = 0;
mica). • BC4: constant horizontal stress and zero vertical
Optical microscope analysis evidenced that in strain, σh = const and Δεv = 0.
both materials quartz is mainly present in sub- Two cells enclose the specimen. This allows
angular, equidimensional grains, feldspars are in obtaining a zero average lateral strain boundary con
both round and prismatic form, while micas are in dition by keeping the pressure in the outer cell equal
lamellae. A diffractometric study shows that the to the pressure, developed by the specimen, in the
mineralogical composition of the two materials is inner cell. Vertical and horizontal stresses can be
compatible with dominant origin from metamorphic applied independently in a controlled manner to the
rocks. Stereo microscope observations denote a type boundaries of the sample. Vertical stresses are
of transport that is relatively low in energy and of applied to the specimen through a piston (positioned
short duration, compatible with poorly worked sand. at the bottom of the chamber) raised by pressured
Grain size distribution of TS11 and TF is shown water and the horizontal stresses are applied by the
in Figure 1. Table 1 lists the main index properties. pressure of water surrounding the specimen.
In this paper the preliminary results of tests carried
out on clean TS11 and on a mix of TS11 and TF char
acterized by 15% FC (MIX15%) are discussed, see
Table 1. Grain size and index properties of testing soils.
Figure 1 and Table 1. The minimum and maximum
void ratios reported in Table 1 were measured accord TS11 TF MIX15%
ing to the method proposed by Knudsen et al. (2020),
validated for silty sand with FC as high as 14%. D60 [mm] 0.49 0.098 0.43
D50 [mm] 0.46 0.075 0.38
2.2 Test program and procedure D10 [mm] 0.32 0.009 0.028
Uc [-] 1.53 11 15.4
Tables 2 and 3 provide the main characteristics of the GS [-] 2.695 2.772 2.721
calibration chamber (CC) and centrifuge (CCC) *γd.max [kN/m3] 16.18 - 18.03
models discussed in this paper. The values of void *γd.min [kN/m3] 13.05 - 13.93
ratio e, relative density DR and dry unit weight γd refer emin [-] 0.634 - 0.48
to the end of consolidation. The test layout is sketched emax [-] 1.026 - 0.916
in Figure 2. All the models were normally
consolidated. * Knudsen et al. (2020)
421
Figure 1. Grain size distribution of TS11, TF and
MIX15%.
Table 2. CC Test Program.
DR
test Soil (%) σ’v (kPa) e (-) γd (kN/m3) V*
1 TS11 48 50 0.84 14.39 V1&V2

2 49 200 0.83 14.41 V1&V2
* V1 = 20 mm/s, V2 = 100 mm/s
Table 3. CCC Test Program (N=63).
test Soil DR (%) σ’v (kPa) e (-) γd (kN/m3) V*
3 TS11 43 50-180 0.86 14.22 V1

4 44 0.85 14.26 V2
9 MIX15% 53 60-220 0.7 15.85 V1
11 53 0.7 15.85 V2
* V1 = 20 mm/s, V2 = 100 mm/s
The CC specimens are enclosed at the sides and

base by a membrane, sealed at the top around an alu
minium plate, which confines the specimen and
transfers the thrust of the chamber piston from the
specimen to a top lid. A 120 mm diameter hole is
present in the centre of the lid; by changing the seal
ing hollow bush it is possible to press devices of dif
ferent sizes into the specimen.
The loading frame, which counteracts the vertical
load transferred to the lid during the compression of
the specimen, also holds the hydro-mechanical press
which pushes the test devices into the chamber during
penetration tests. A hollow jack mounted inside the
loading frame is used to counteract the vertical load; it
is automatically controlled by a closed loop system,
which equalizes the compression force in real time and
allows to keep the lid position fixed and independent Figure 2. CC (a) and CCC (b) models.
422
from the frame deformations. Thus the specimen to minimise any scale effects on the results (Bolton
deformation can be monitored by measuring the cham et al. 1999). The ratio of the cone diameter to the
ber piston displacement. mean particle size is dc/D50 ≈ 25 for TS11 and dc/D50
The penetration probes used during the tests is ≈ 30 for MIX15%. CCC models were instrumented
a standard piezocone 35.7 mm in diameter, with with pore pressure transducers (PPT in Figure 2),
a total area of 10 cm2 and an apex angle of 60°. Two located at the base to monitor the water table and at
load cells measure the tip resistance and the lateral three relevant depths of penetration, at a distance of
friction, independently; a pressure transducer meas one cone diameter from the penetration axis.
ures the pore water pressure behind the tip (u2). For each test condition two penetration rates were
The CC specimens were reconstituted in 15 strata adopted; a standard rate V1 = 20 mm/s and a higher
using the undercompaction method (Ladd, 1978), rate V2 = 100 mm/s (both velocities properly scaled).
Hereafter, they were saturated through an upwards
flow of deaerated water and then by application of
a back pressure. Reaching a Skempton B-value equal 3 TEST RESULTS
or larger than 0.95, the specimens were consolidated
by applying the target vertical and horizontal stresses. 3.1 CPTUs in clean TS11
To allow for comparison with the centrifuge tests,
during which a rigid strong box houses the models and The results of the tests discussed in this section are
prevents the development of horizontal strains, the shown in Figure 3 and were obtained from soil models
BC3 condition was adopted in the CC. In consequence, of clean TS11 reconstituted at a relative density DR
horizontal effective stresses imposed in CC were cali slightly lower than 50%. In Figure 3 the corrected cone
brated during the consolidation step to avoid radial resistance qt is plotted as a function of the vertical
deformations. A standard rate (V1 = 20 mm/s) was effective stress σ’v. The tests are numbered according
adopted in the upper part of the specimen and the max to Tables 2 and 3. The black and white squares repre
imum velocity possible for the loading system (V2 = sent the representative qt measured in the CC speci
100 mm/s) in the lower part, see Figure 2a. mens (Tests 1 and 2). For each CC test, two qt values
are plotted: one is the average value measured in the
2.2.2 The ISMGEO geotechnical centrifuge upper half of the model, with the probe penetrating at
The ISMGEO geotechnical centrifuge is a beam the standard rate V1; the second value refers to the
centrifuge made up of a symmetrical rotating arm faster rate V2 adopted in the lower half of the model.
with a diameter of 6 m, a height of 2 m, a width of The black and grey lines in Figure 3 are the qt profiles
1 m, and a nominal radius of about 2.2 m to the measured in the CCC (Tests 3 and 4). The vertical
model base (Baldi et al.1988, Fioravante et al. effective stresses in the centrifuge models are com
2021). The miniaturised piezocone used for the puted referring to: i) the average soil unit weight at the
tests has a diameter dc = 11.3 mm, an apex angle of end of the in-flight consolidation, ii) the depth of the
60° and a sleeve friction of 11 mm in diameter and water table (estimated from PPT measurements) and
37 mm in length. One load cell measures the cone iii) the acceleration field distortion. Figure 3 also shows
resistance and another one measures the cone resist the qt profile (dashed line) estimated using the equation
ance plus the shaft friction, up to forces of 9.8 kN. of Jamiolkowsi et al. (2003). This allows to express the
A pressure transducer is installed behind the tip for cone resistance as function of the vertical effective
interstitial pressure measurements (u2). stress and relative density, using correlation coefficients
The centrifuge specimens are 470 mm high and calibrated by the Authors for Ticino sand and account
400 mm in diameter and were reconstituted at 1g ing for the saturation effects. It is worth noting that the
using the undercompaction method within a rigid correlation was calibrated on the base of CC tests car
strong box. They were saturated under vacuum using ried out using the same apparatus employed for the pre
deaerated water and subjected in flight to an acceler sent experimentation. For the centrifuge tests, the u2
ation field of 63 g imposed at mid depth (geometrical profiles, compared with the hydrostatic lines derived
scaling factor N = 63). The scaling factor and the from the PPT measures, are given in Figure 4.
angular velocity adopted allowed to obtain a vertical The centrifuge test results show that the soil
stress of about 50 kPa at a depth of 120 mm from models were rather homogeneous and the tests are
ground surface (which is the depth at which the cone repeatable, as the two qt profiles are almost superim
resistance qc is no more affected by top boundary posed. The penetration was, as expected, drained for
effects) and of 200 kPa at a distance of 150 mm from both penetration rates, see Figure 4. A very good
the container bottom (depth beyond which qc can be agreement between CC and CCC results can also be
affected by the rigid bottom boundary); 50 kPa and observed, see Figure 3. In addition, the measured
200 kPa are the vertical effective stresses imposed in cone resistance is very well described by the correl
the calibration chamber (tests N. 1 and 2 in Table 2). ation proposed by Jamiolkowski et al. (2003). This
The CCC boundary conditions are: D/dc = 35, is considered an important result, as it demonstrates
where D is the internal diameter of the container and that CC and CCC give comparable CPT results,
sc/dc = 17, where sc is the distance between the CPT which are also consistent with previous studies car
and the side wall. These values are sufficiently large ried out using TS and the same facilities; these data
423
run at V2. Figures 5 and 6 show the results. Inspecting
the u2 profiles (Figure 6), the penetration appears to be
practically drained irrespective of the penetration rate,
similar to the results on TS11.
However, as to the effect of FC on the penetration
resistance (Figure 5), MIX15% had a penetration resist
ance about 40% lower than TS11, even though the
void ratio is higher and the relative density is lower for
TS11 compared to MIX15% (see Table 3). It’s worth
noting that drained and undrainded triaxial tests on
reconstituted samples indicate a shearing resistance
angle at critical state of 36° and 35° for TS11 and
MIX15%, respectively. On the other hand, the two
materials proved to have different volumetric behavior
during shearing. The results of 4 drained triaxial tests
carried out on medium dense TS11 and MIX15% sam
ples are shown in Figure 7, in the void ratio e – mean
effective stress p’ and in the stress deviator q – p’
plane. All the samples were reconstituted in strata at
medium density (similar to the CC and CCC speci
mens, see Table 2 and 3) and were K0-consolidated
Figure 3. Centrifuge (solid lines) and Calibration Chamber under a vertical stress of 50 and 200 kPa. While
(squares) CPTUs on TS11 - cone resistance qt. MIX15% manifested a contractive behavior, TS11
dilated.
In general, the cone penetration resistance qt of an
uncemented and unaged soil depends on the material
properties and the state of the soil (stress level and
density). The state of the soil governs the direction of
volumetric strains, (dilation or contraction) during
shearing, which, in turns, controls the stress increment
around the tip. Consequently, a dilative soil will
develop a larger stress increment around the tip and
will oppose larger resistance to penetration than
a contractive material, as observed for TS11 and
MIX15%.
Figure 4. Centrifuge CPTUs on TS11 – PPT measures,

hydrostatic line, pore pressure u2.
are used as benchmark for the following testing

stages and, in the first instance, used to evaluate the
effect of 15% FC on the cone resistance, for the
same relative density and stress state.
3.2 CPTUs in MIX15%

Two centrifuge tests (9 and 11 in Table 3) on MIX15%
were carried out using the same test conditions as
those of tests 3 and 4, i.e. similar relative density
(about 50%), stress range and penetration rate. Test 9
was carried out using the standard rate V1; test 11 was Figure 5. Centrifuge CPTUs on MIX15% - cone resistance qt.
424
Figure 6. Centrifuge CPTUs on MIX15% – PPT measures,
hydrostatic line, pore pressure u2.
4 CONCLUSIONS
A very good agreement between results of CC and

CCC CPTUs in clean TS11 was observed and the
measured cone resistance is very well described by the
correlation proposed by Jamiolkowski et al. (2003).
The sand mixed with none plastic silt experiences
a drop of cone resistance in the centrifuge, which
Figure 7. Drained triaxial tests on TS11 and MIX15%: a)
cannot be attributed to effect of partial drainage e-p’ plane; b) q- p’ plane.
during penetration, as the measured u2 profile is
straight and coincident with the hydrostatic line and
no excess pore pressures developed neither at the the development of acknowledged, simple and
standard penetration rate nor during the faster robust guidelines for specification, execution
penetration. and interpretation of CPTUs in silty soils/silt
The drop of cone resistance is attributed to con mixtures.
tractive behaviour of the silty sand at the test density
and stress in contrast with the dilative behaviour of
the clean sand. ACKNOWLEDGMENTS
The centrifuge has proved to be a reliable CPT
calibration tool in clean sand, alternative to the The CSi Project is managed through the Carbon
calibration chamber, with the advantage of provid Trusts’s joint industry Offshore Wind Accelerator
ing a continuous cone penetration resistance pro (OWA) program. The Authors acknowledge the pro
file over a wide range of stress level in vision of financial and technical support by the fol
significantly less time. If a good agreement lowing project partners: Ørsted (lead partner),
between centrifuge and calibration chamber results Equinor, Vattenfall, Shell, RWE and Scottish Power
will be gained also for the MIX15%, the centrifuge Renewables.
alone will be employed in a further stage of
experimentation, during which 10 additional cali
bration CPTUs will be carried out on a mix of REFERENCES
TS11 and TF characterized by 30% FC
Augustesen, A.H., Carotenuto, P., Bilici, C., Lunne, T.,
(MIX30%). Samples will be reconstituted varying Lindeboom, R.C.J., Krogh, L., van den Bosch, J.,
soil density and overconsolidation ratio and Barth, R., Erbrich, C., Giretti, D., Fioravante, V.,
will be tested varying the penetration rate. The H. Dias, H., Sougle, M-C., Barwise, A., de Wit,
calibration CPTUs and the complementary labora S., Burbury, D., Adams, N. 2022. CSi – a joint industry
tory tests will be interpreted in the frame of project into CPTUs in silty soils. Submitted for publica
CSi project with the final goal of contributing to tion in the Proceedings of CPT’22.
425
Baldi, G., Bellotti, R., Ghionna, V., Jamiolkowski, M. and cone penetration tests. Acta Geotechnica, 11:953,
Pasqualini, E. 1982. Design parameters for sand from doi:10.1007/s11440-015-0419-3.
CPT. Proc. ESOPT 2, Amsterdam. Fioravante, V. Giretti, D., Airoldi, S., Moglie J. 2021.
Baldi, G., Bellotti, R., Ghionna, N., Jamiolkowski, M. Effects of seismic input, fine crust and existing structure
and Pasqualini, E. 1986. Interpretation of CPTs and on liquefaction from centrifuge model tests. Bulletin Of
CPTU’s, 2nd Part. Proc 4th International Geotech Earthquake Engineering, vol. 19, p. 3807–3833, ISSN:
nical Seminar, Nanyang Technological Institute, 1570-761X, doi: 10.1007/s10518-021-01139-4.
Singapore, 143–156. Jamiolkowski, M.B., Lo Presti, D.C.F., Manassero, M.
Baldi, G., Belloni, G., Maggioni, W. (1988). The ISMES 2003. Evaluation of Relative Density and Shear Strength
Geotechnical Centrifuge. In Centrifuge 88, Paris, Corté from CPT and DMT. Soil Behavior and Soft Ground
J. F. Ed., Balkema, Rotterdam, 45–48. Construction, Ladd Symposium, MIT, Cambridge Mass.
Bolton, M.D., Gui, M. W., Garnier, J., Corte, J. F., Geotechnical Special Publication No. 119, ASCE,
Bagge, G., Laue, J. & Renzi, R. (1999). Centrifuge Reston, Virginia, pp. 201–238.
Cone Penetration Tests in Sand. Geotechnique, 49(4), Knudsen, S., Powell, J.J.M, Lunne, T., Thomsen, N.V.,
543–552. Krogh, L., Barwise, A. 2020. Development of new
Fioravante, V. 2000. Anisotropy of small strain stiffness of robust procedures for the determination of maximum
Ticino and Kenya Sands from seismic wave propagation and minimum dry densities of sand. Submitted to
measured in triaxial testing. Soils and Foundations, Vol. ISFOG2020.
40, No.4, 129–142. Ladd, R.S. 1978. Preparing tests specimens using under-
Fioravante, V. & Giretti, D. 2016. Unidirectional cyclic compaction. Geotechnical Testing Journal, GTJODJ,
resistance of Ticino and Toyoura sands from centrifuge 1(1), 16–23.
426
© 2022 the Author(s), ISBN 978-1-032-31259-0
Comparison between coefficients of consolidation from CPTu and

laboratory tests for Guaratiba’s soft soil, Rio de Janeiro, Brazil
M.M. Freire, M.E.S. Marques & M.C. Tassi
Military Institute of Engineering, Rio de Janeiro, Brazil
L.A. Berbert
Prodec Consultoria, Rio de Janeiro, Brazil
ABSTRACT: This paper presents a discussion about vertical (cv) and horizontal (ch) coefficients of consoli
dation of Guaratiba’s very soft clay, located at the lowlands of the West Zone of Rio de Janeiro city, Brazil.
Due to the thickness of the deposit, high compressibility of the clay, and low undrained strength, the settle
ment stabilization, even with vertical drains, are known to take of about 30 months to occur at these lowlands.
Thus it is essential to know cv, ch and ch/cv relationship in order to predict when settlement stabilization will
occur. The variability of cv and ch values is very high, probably due to the wide range of organic matter, sand
lenses and the sand content throughout the soft soil deposit. The aim of the paper is to compare these values
of coefficients of consolidation obtained from the dissipation tests and from the consolidation tests and to
discuss an approach to determine them for design considering their variability.
1 INTRODUCTION Permeability ratio (kh/kv) is used for the estima

tive of ch or cv, but regardless of the type of test, for
Vertical (cv) and horizontal (ch) coefficients of con very soft, very organic soils, there is a wide range of
solidation are used to predict settlement evolution consolidation coefficient values obtained for these
with time, and are essential for the design of deposits.
embankments over soft soils, mainly when vertical This paper presents a discussion about coefficients
drains are used. of consolidation values from dissipation tests and
Coefficients of consolidation are estimated from consolidation tests carried out at a very soft organic
dissipation tests, carried out at a prestabilished depth clay deposit of Rio de Janeiro city.
of CPTu, from CRS and incremental consolidation
tests, consolidation on triaxial tests and from field
monitoring. They are related to permeability (k), 2 SITE DESCRIPTION
which is a parameter that, for soils, has a range of
values on the order of about 106 m/s. This parameter The site is located at Guaratiba, a neighborhood
depends on several factors as: diameter, distribution, area of increasing population density in the city of
shape, texture and mineral composition of particles; Rio de Janeiro, Brazil (Figure 1), limited on one
flow and fluid type; soil structure; water content; side by the coastal massifs and on the other side by
void ratio and temperature. the ocean.
Thus, it does not come as a surprise that the The area of the site is 1.5 km² and the thickness
values from field monitoring, field tests and labora of the soft soil in the deposit reaches 13.5 m. The
tory are different, due to: water table is coincident with ground level, and the
site is under water at the rainy season.
- The boundary conditions are very different in the The geotechnical investigation consisted of 98
field, where lateral deformations exist, and in the boreholes with Standard Penetration Test carried
laboratory consolidation test, where the deform out in order to map clay thickness. After that,
ation is unidirectional; 15 geotechnical investigation clusters (see
- The occurrence of sand lenses in the field, redu Figure 2) were implemented: 15 CPTu with 41
cing the stabilization time, a fact that is not repro pore-pressure dissipation tests, 15 undisturbed
duced in the laboratory; Shelby samples, 40 oedometer consolidation
- Secondary consolidation takes place in the field tests, field vane tests, characterization and min
and is disregarded in the analysis of field data. eralogy tests.
DOI: 10.1201/9781003308829-59
427
and dissipation field tests carried out with
a CPTu.
From dissipation tests, ch values can be obtained
from Equation 1, proposed by Houlsby and Teh
(1988):
where:
Th – Time Factor;
R – Piezocone Radius;
Ir – Rigidity Index (G/Su);
G – Shear modulus
Su – Undrained strength
t – Dissipation Time/Period.
In order to obtain cv values from ch values of

Figure 1. Site localization. CPTu tests, it is necessary to obtain the permeability
ratio (kh/kv) as follow.
Some geotechnical characteristics of the clay are:
natural water content values vary from 50% to
150%; the plasticity index is between 62% and
103%; initial void index vary from 1.51 to 3.65;
organic matter content from 1.74% to 13.3% and the Conventional and special oedometer tests were car
undrained strength is in the range of 9 to 36 kPa. ried out on Sarapuí organic soft clay, located at Gua
The compression ratio, Cc/(1+e0), vary from 0.27 to nabara Bay, 60 km from Guaratiba (Coutinho, 1976;
0.51, preconsolidation pressure from 23 to 50 kPa Lacerda et al., 1977, 1995). These tests were carried
and OCR values from 1.07 to 3.34. All those charac with radial drainage in order to obtain ch. The ch/cv
teristics indicate a very compressible clay. relationship obtained was between 1.0 and 2.0 at
From macro-drainage projects in this region, con normally consolidated range, which is typical for
ducted by the city of Rio de Janeiro, the embank very soft clays, as shown in Table 1, for worldwide
ments must level up to 3m in order to avoid floods. deposits. Thus, the permeability ratio used for design
The work schedule of embankment construction in engineering practice for the soft clay deposits of
over soft soils city is directly related to vertical (cv) Rio de Janeiro City has been 1.5.
and horizontal consolidation (ch) values, which are
in the order of 2-3 x 10-8 m²/s at the West Zone of
Rio de Janeiro, under normally consolidated stresses. Table 1. Ranges of possible field values of permeability
ratio of soft clay deposits (Jamiolkowski et al. 1985).
Nature of clay kh/kv
No macrofabric, or only slightly developed 1.0 to

macrofabric, essentially homogeneous deposits 1.5
From fairly well to well developed macrofabric, 2.0a4.0
e.g. sedimentary clays with discontinuous lenses
and layers of more permeable material
Varved clays and other deposits comtaining 3.0 to
embedded and more or less continuous perme- 15.0
able layers
In order to obtain laboratory ch of Guaratiba, from

Figure 2. Geotechnical investigation clusters. consolidation tests, the samples were rotated 90º
from its original in situ position. Using this proced
3 LABORATORY AND CPTU TESTS RESULTS ure, cv values were obtained, as shown in the schema
of Figure 3, from 10 HS samples and ch from 10 VS
The main tests used to determine the coefficient samples, using square root of time method proposed
of consolidation are laboratory consolidation test by Taylor, as shown in Table 2.
428
where ch(oc) at overconsolidated range was obtained
from equation (1) and Cr/Cc was 0.15. Table 3 shows
results of ch values from both tests.
Table 3. The ch values from consolidation tests and CPTu

tests at normally consolidated range.
cv (HS) x 1.2 ch (VS) ch (CPTu)*

Depth
Vertical (m) (x10-8 m2/s)
Figure 3. Sampling details.
Z1-04 3 3.5 -
5 83.3** -
Since the effective stress at the top of the clay deposit Z1 14 2 3.4 - 17.3
was very low, consolidation tests loading began at 3.0 6 1.4 - 6.3
kPa, with load increment ratio equal to one (Δσ/σ= 1), 8 1.5
and at vertical effective stress of 100 and 200 kPa the Z1-18 1,5 - 3.8
samples were at the normally consolidated range. From 3,5 - 3.1
average values, ch/cv was in the order of 1.2 for Guara Z1-20 3 3.0 - 2.9
tiba clay (Berbert, 2016). Thus, it was possible to esti 4,7 - - 2.3
mate from ch = 1.2 cv, the values of ch for another set of 3 - 1.8 2.5
20 consolidation tests carried out on HS samples. Z1-27 5 - 1.6 3.1
7 - 2.2 6.1
2 - 2.2 16.7
Table 2. Ch and cv from consolidation tests at normally
4 - - 7.0
consolidated range.
Z1-29 6 - 2.1 8.7
ch cv 8 - - 10.1
(VS) (HS) ch AV cv AV 10 2.1 - 13.1
Z1-33 2 3.7 -
Depth Stress -8
(x10 m /s)2
kh 5 1.6 -
Vertical (m) (kPa) /kv Z1-36 3 - 2.1 5.0
7 - 2.2
Z1-27 3 100 2.3 1.9 1.8 1.6 1.2 Z1-50 2 4.9 -
200 1.4 1.3 Z1-60 4 - 4.9
5 100 2.1 1.3 1.6 1.3 1.3 Z1-66 3 2.7 -
200 1.2 1.3 Z1-76 3 1.6 -
7 100 2.3 3.0 2.2 2.9 0.8* 5 2.1 -
200 2.0 2.7 4 1.6 - 3.7
Z1-18 1,5 100 4.4 8.4 3.8 5.5 0.7* 6 - - 3.1
200 3.3 2.5 Z2-06 7 1.68 -
3,5 100 3.6 2.5 3.1 2.3 1.3 8 - - 6.2
200 2.5 2.0 10 2.28 - 4.7
Z1-29 2 100 2.6 2.9 2.2 2.7 0.8* 3 - 3.7
200 1.8 2.6 Z2-09 5 1.68 - 5.7
6 100 2.3 1.6 2.1 1.8 1.1 7 2.28 - 6.8
200 1.9 2.1 Z2-14 2 1.56 - 11.8
Z1-36 3 100 2.3 1.4 2.1 1.3 1.6 4 1.68 - 6.2
200 1.9 1.2
7 100 2.1 1.9 2.2 1.7 1.3 * CPTu values at normally consolidated range
200 2.3 1.4 ** not considered
Z1-60 4 100 7.2 4.7 4.9 3.7 1.3
200 2.6 2.6 Sample Z1-04 at 5m had a high cv value.
A possible cause of this high value is the fact that
* kh/kv assumed=1 this sample had the lowest initial void index, natural
moisture and organic matter value.
From CPTu dissipation tests, the ch at normally
consolidated range was obtained from Equation 3
(Jamiolkowski et al. 1985): 4 DISCUSSION
Guaratiba clay is a little less compressible than most

clays from the western region of Rio de Janeiro City
429
Figure 4. Cv and ch values from consolidation tests at normally consolidated range.
Figure 5. Ch values from consolidation tests and dissipation tests at normally consolidated range.
and also less plastic. However, the values of cv could be attributed to sample disturbance, since sam
and ch of Guaratiba clay are inside the range of pling those very compressible clays is very difficult.
the cv values obtained for western region Rio de Almeida and Marques (2013) showed that the great
Janeiro clays for the normally consolidated range: variability of ch values is not uncommon for Rio de
10-9m2/s to 2 x 10-7 m2/s for Recreio and Barra Janeiro clays, and the difference between cv values
da Tijuca clays (Almeida and Marques, 2013). from laboratory data and field instrumentation can be
From consolidation tests, it is not possible to as high as 10.
obtain a specific range for cv or ch, but the
majority of data shows higher ch, at a certain
depth as shown in Figure 4. 5 CONCLUSIONS
Figure 5 shows ch results from CPTu compared
with laboratory consolidation tests. It seems that This paper showed the comparison between cv and
CPTu values are higher than laboratory tests, which ch values at normally consolidated range from special
430
consolidation tests and CPTu tests. These parameters laboratory testing of soils. In: Proc. XI ICSMFE. San
are important to obtain settlement prediction with Francisco, 1985. v.1, p. 57–153.
time, thus essential for deadlines of earthworks. Rosa, A. C. & Marques, M.E.S. 2019. Estimative of coeffi
The ch values from CPTu tests and from consoli cient of consolidation from piezocone dissipation tests
dation tests (HS and VS samples) are inside the wide (in Portuguese). GEOSUL 2019. Joinville, Santa Catar
range of magnitude of values of Rio de Janeiro clays. ina, Brazil.
However, the variation of values makes it difficult to Santos, H.M.C. 2004. Physical, chemical, mineralogical
and geotechinal characterization of gleissolos of baixa
adopt a ch value for design. In this case, which ch or cv das de Jacarepaguá, Guaratiba and Santa Cruz – from
is recommended for the practice? Average values does the city of Rio de Janeiro (in Portuguese). D.Sc. Thesis,
not seem adequate, since it could underestimate the Department of Geology, Federal University of Rio de
deadlines of the work. It is suggested to choose a value Janeiro, CCMN, Rio de Janeiro.
that meets most cases, considering that the work sched Houlsby, G. T. & TEH, C. I. 1988. Analysis of the piezo
ule could be adjusted depending on field monitoring. cone in clay. Proceedings of the International Sympo
Therefore, together with field and laboratory tests, sium on Penetration Testing, ISOPT-1, Orlando, 2,
field instrumentation is necessary to monitor the work. Balkema Pub., Rotterdam: 777–783.
Coutinho, R. Q. 1976 Consolidation characteristics from
radial drainage tests on a soft Clay of Fluminense Low
REFERENCES lands (in Portuguese). Master’s Thesis – COPPE/UFRJ,
Rio de Janeiro.
Almeida, M. S. S. & Marques, M. E. S. 2013. Design and Lacerda, W. A.& Almeida, M. S. S. & SANTA
performance of Embankments on very soft soils. MARIA, P. E. L. & Coutinho, R. Q. 1995. Interpret
Londres. 228p. ation of radial consolidation tests. Proceedings of
Berbert, L.A. 2016. Evaluation of vertical and horizontal the International Symposium on Compression and
consolidation coefficients of soft soil deposit aiming Consolidation of Clayey Soils, Yoshikuni, Kusabe
settlements prevision of infrastructure works (in (eds), Hiroshima, Japan, Vol. 2, 1091–1096. Rotter
Portuguese). M.Sc. Dissertation, Transportation Engin dam: Balkema.
eering Program, Military Institute of Engineering, Rio Lacerda, W. A. & Costa Filho, L. M. & Coutinho, R. Q. &
de Janeiro, 182 p. duarte, A. R., 1977. Consolidation characteristics of Rio
Jamiolkowski, M. & Ladd, C. C. & Germaine, J. T. & de Janeiro soft clay. Proceedings of Conference on Geo-
Lancellotta, R. 1985. New developments in field and technical Aspects of Soft Clays, Bangkok, 231–244.
431
© 2022 the Author(s), ISBN 978-1-032-31259-0
Quantitative modelling of spatial variability of piezocone data from Venice

lagoon silty soils
G. Gottardi, M. Ranalli & L. Tonni
Department of Civil, Chemical, Environmental and Materials Engineering, DICAM, Alma Mater Studiorum,
University of Bologna, Italy
M. Uzielli
Department of Civil and Environmental Engineering, DICEA, University of Florence, Italy
ABSTRACT: As geotechnical research and design codes rely increasingly on probabilistic approaches, site
characterization should also be conducted in the light of the explicit quantification of the uncertainty and spa
tial variability in soil properties. This paper provides a practical case-study application of spatial variability
analysis of piezocone data obtained at a test site in the Lagoon surrounding the historic city of Venice, in
North-Eastern Italy, where unusually dense and regularly spaced CPT test data were available. Empirical
semivariograms are calculated for cone resistance, sleeve friction, and porewater pressure, along with the soil
behavior classification index obtained from these measurements at a set of reference measurement depths.
A number of theoretical semivariogram models are fitted comparatively and best-fit models are selected based
on objective criteria and subjective judgment. Characteristic semivariogram parameters, providing informa
tion on correlation distance and small-scale variability, are retrieved. Modeling options are explained and
results are analyzed and assessed critically.
1 INTRODUCTION character of evolutionary design codes based on limit-

state design and reliability analysis, attest to the rele
The definition of subsoil properties for engineering vance of indetermination in geotechnical data and
purposes is a difficult task due to the indetermination models, and of the importance of considering its effects
in geotechnical data which is employed in the charac on characterization and design.
terization process. Such indetermination can be In the technical literature, the terms variability and
ascribed to both the soils and the investigators. Soils uncertainty are often employed interchangeably. Here,
are natural materials, which are formed and continu variability is defined as an observable manifestation of
ously modified by complex geological processes, as heterogeneity of one or more physical parameters and/
discussed in detail by Hight & Leroueil (2003). The or processes, while uncertainty pertains to the model
variety and complexity of such processes result in ler’s state of knowledge and strategy, and reflects the
physical heterogeneity and, consequently, in the spatial decision to recognize and address the observed vari
and temporal variability of quantitative parameters. In ability in a qualitative or quantitative manner. Inherent
any type of real-world geotechnical problem, more soil variability describes the variation of properties
over, it is impossible to obtain exhaustive values of from one spatial location to another within a soil mass.
data at every desired point because of practical and Such variability is parameterized quantitatively by
economical constraints. The testing methods by which aleatory uncertainty. Epistemic uncertainty, on the con
such data are obtained are always imperfect and non- trary, stems from the investigator’s invariably limited
repeatable to some degree. Moreover, if measured information, imperfect measurement and modelling
parameters require transformation for use in geotech capabilities.
nical analyses and design, they are to serve as inputs to In investigating the spatial variability of
transformation models which, whether theoretical, a geotechnical property, both aleatory and epistemic
empirical or experimental, are invariably incapable of uncertainties are addressed, as both contribute to the
perfectly replicating the complexity of the physical uncertainty which is modelled, processed, and
world. A steadily increasing bulk of research and reported. The aim of this paper is to provide a practical
experience, not to mention the uncertainty-based case-study of characterization of spatial variability of
DOI: 10.1201/9781003308829-60
432
the mechanical properties of soils as parameterized by
piezocone data from a site located near the historical
city of Venice, in North-Eastern Italy. Spatial variabil
ity analysis is performed through the calculation of
empirical semivariograms and the subsequent fitting of
theoretical semivariogram models.
2 TEST SITE DESCRIPTION
The Venetian lagoon is underlain by about 800 m of

Quaternary deposits which originated from alternate
phases of continental and marine sedimentation associ
ated to marine regressions and transgressions occurred
over the last 2 million years. Over the last forty years,
the upper 100 m sediments of the Venetian lagoon Figure 1. Treporti CPTU test site.
have been thoroughly investigated, in relation to the
regional land subsidence (e.g. Ricceri & Butterfield,
1974) as well as the foundation design of the submers reference to the detection of the most compressible
ible gates intended to protect the historical city of soil layers.
Venice against recurrent flooding (e.g. Simonini et al., A very detailed ‘second phase’ testing program
2007). These sediments have proven to be highly het was then performed just on the subsoil beneath the
erogeneous and stratified in a continuous alternation of test bank to be built. All the second phase tests,
slightly overconsolidated and predominantly silty sedi numbered from 11 to 20, were located within a
ments, ranging from medium-fine sands (SP-SM) with 45 m diameter circular area. They were lined up
sub-angular grains to silts (ML) and very silty clays along three diameters, splitting the circular area in
(CL). Coarse sediments are predominantly composed three 120° identical slices, and placed on two concen
of silicates and carbonates, whilst silts and silty clays, tric circumferences, inner (D=30 m) and outer
which originated from mechanical degradation of (D=45 m) with respect to the loading bank (D=40 m).
sands, display a content of clay minerals (illite with All tests were pushed as deep as possible, depending
minor quantities of chlorite, kaolinite and smectite) on the maximum available thrust, typically to
never exceeding 20% in weight. Generally, cohesive about 42 m.
soils are slightly overconsolidated due to aging or In addition to the main investigation program,
oxidation. two further testing phases were carried out. Imme
A representative test site of such heterogeneous and diately after the loading bank construction, the
stratified subsoil was located outside Treporti, ‘third phase’ was performed from top of the
a historical fishermen’s village on the Cavallino coast bank, whereas a final ‘fourth-phase’ testing cam
line facing the North Eastern lagoon. At this site an paign was launched after the loading bank removal.
extensive research program was recently carried out by The aim was to examine the stress-history effect on
the Italian Universities of Padova, Bologna and the soil mechanical properties along a number of
l’Aquila, with the aim of better understanding the verticals already investigated in both the pre-bank
behavior of such heterogeneous sediments. The and post-bank construction phases (Tonni & Got
research program, consisting of a detailed geotechnical tardi, 2009).
characterization through in-situ tests, continuous- Due to the importance of ensuring the quality of the
coring boreholes and high-quality laboratory tests, also dataset in terms of homogeneity of loading conditions,
included the construction and the subsequent gradual only testing data from the first and second phases were
removal of a full-scale, 6.5 m high and 40 m in diam used in this analysis. Typical piezocone log profiles of
eter, vertical-walled cylindrical test bank, in conjunc Venetian subsoil are reported in Figure 2, showing the
tion with a very detailed monitoring system of subsoil corrected cone resistance qt and the pore pressure u of
strains and pore water pressures (Tonni & Gottardi, CPTU tests located along the W-E cross section. Pro
2011, 2019). files of CPTU 6 and CPTU 14, located approximately
The extensive experimental program carried out 7-8 m north from the selected alignment, are superim
in Treporti included no. 24 piezocone tests. The posed on the same plot.
campaign was conducted in four phases, in relation The figure clearly shows the high stratification
to the different loading conditions associated to the and marked horizontal spatial variability of stratig
presence of the bank. Figure 1 shows a plan of the raphy, which are typical features of Venetian lagoon
area together with the location of CPTU tests. subsoil. Indeed, while piezocone log profiles provide
The ‘first phase’ of the in-situ testing campaign, immediate evidence of a well-defined, 6-7 m thick top
including no. 7 piezocone tests located throughout layer of silty sand, the underlying, dense alternation of
the selected area, was intended to give an initial thin layers of sandy and clayey silts is significantly
description of the test site subsoil, with specific variable from one location to another. The continuous
433
horizontal variation of the soil stratigraphy is particu The quantitative parameterization of spatial cor
larly evident in the pore pressure logs, where u rarely relation given by the empirical semivariogram
follows up the hydrostatic level, often falls below it, includes both aleatory uncertainty (given by the
but never develops high Δu values, typical of pure nor measurable scatter in the parameters of interest) and
mally consolidated or slightly overconsolidated clays. epistemic uncertainty (comprising measurement
A potentially more compressible fine-grained unit, error and inherent variability at sub-measurement
which is mainly composed of silts, can be detected in interval scale). The analysis was performed in
the CPTU profiles from 7-8 m to 20 m depth. How a pseudo-3D approach, which means estimating the
ever, the thickness of such silty unit is clearly not con spatial variability only in the horizontal planes at
stant throughout the whole test site area, as it is often nominal analysis depth, ranging from 1m to 40m
interbedded with a clean sand layer of variable thick below ground surface at 1-meter intervals. This
ness. Soil heterogeneity and high degree of interbed approach requires the availability of an exceptional
ding make the analysis of field data rather difficult. database in terms of the number of CPTu verticals at
different distances from each other, within the area
to be analysed.
In practice, the calculation of empirical semivar
iograms relies on the establishment of a lag toler
ance, both on distance (by accepting data pairs
which separation distance is acceptably close to the
nominal lag distance) and direction (by allowing
a preset degree of deviation from the nominal direc
tion between two sampled locations). The reason for
the acceptance of such tolerances (which should not
be too large in order to avoid the introduction of
excessive distortion and bias) lies in the opportunity
to have a greater number of pairs, which results in
Figure 2. Typical soil profile of Venetian subsoil at Tre a more statistically reliable semivariogram structure.
porti site. In order to identify an optimum lag interval (in terms
of regularity and accuracy of empirical semivariance
values), a parametric analysis on Δh was carried out.
3 SPATIAL VARIABILITY ANALYSIS In particular, empirical semivariograms with lag
intervals of 5 m, 10 m and 15 m were calculated at
3.1 Spatial correlation analysis of CPTU data each analysis depth and for all three CPTU test
Spatial variability can be investigated quantitatively parameters; namely: tip resistance qt, sleeve friction
using several approaches. Semivariogram-based fs, and induced pore pressure u2. In all cases, the dis
tance tolerance h ~ was set equal to one half the lag
modeling ranks among the most widely used due
to its close relationship with geostatistical kriging interval Δh (Isaaks & Srivastava 1990), i.e. =2.5 m,
interpolation. Semivariogram-based modeling accom 5 m and 7.5 m, respectively. Thus, all data points are
modates weakly (but not necessarily strongly) station considered at least once, with no redundancies at any
ary data sets, i.e., data presenting spatial trends, lag distance value. An intermediate lag interval of
provided that spatial correlation depends on measure Δh=10 m (with an associated lag tolerance ~ h=5 m)
ment spacing but is independent of specific measure was deemed to be the most suitable, since it is small
ment location. The empirical absolute semivariogram enough to preserve the inherent variability of semi-
of a dataset is computed as variance while providing a clearer structure of spa
tial correlation than smaller lag intervals.
Example empirical semivariograms, pertaining to
qt at analysis depth of 27 m and calculated for the
three different values of lag spacing, are shown in
Figure 3. In the plots, for each estimated semivariance
value (black points), the corresponding number of
where N(h) is the number of data pairs sampled contributing data pairs is indicated. is the calculation
at locations separated by a spatial distance h sample numerosity makes it possible to assess the
(hereinafter referred to as lag distance), ξi and ξj reliability of each semivariance value (with larger
are the data values at the ith and jth locations, N(h) ensuring lower statistical uncertainty in calcu
respectively. lated semivariance).
Calculating empirical semivariance by Eq. (1) for
different lag distances allows quantification of the
magnitude and variation of spatial correlation with 3.2 Modelling of empirical spatial variability
lag distance. Small γ(h) values are representative of The fitting of suitable semivariogram models to
high spatial correlation, while high γ(h) values attest empirical semivariograms is a fundamental step in
for low correlation. the quantitative modelling of soil spatial variability
434
the structure (model), magnitude (sill), spatial extent
(range), and level of epistemic uncertainty (measure
ment uncertainty and small-scale variability),
respectively, of spatial correlation. Such an assess
ment is based on a combination of objective
approaches (provided by the possible use of regres
sion techniques) and subjective belief (e.g., manual
fitting, critical assessment of results of regression).
The quantification of the semivariogram nugget is
not a univocal procedure. Operationally, the assign
ment of a semivariogram nugget should rely on the
user’s “backward extrapolation” of a best-fit semivario
gram model to zero-lag distance. In practice, this pro
cedure provides reliable results only if reliable
empirical semivariance values are available for small
lag distances. This is arguably not the case with the
Figure 3. Empirical semivariogram of qt at 27 m depth, for Treporti data sets. Consequently, in order to provide
different values of lag interval and lag tolerance. a consistent and conservative criterion for estimating
semivariogram nuggets, it is necessary to resort to
existing knowledge regarding the phenomena contrib
using geostatistical approaches.,.Numerous semivar uting to the nugget effect. The piezocone test is known
iogram models are available in the literature (e.g., to have one of the lowest total measurement errors
Isaaks & Srivastava 1990). Among the most widely among in-situ test in current practice (Uzielli 2008).
used models are the spherical, exponential, and CPTU measurement error represents a combination of
Gaussian models, characterized by the following several sources of variability and cannot generally be
equations, respectively: evaluated in routine geotechnical analysis. It is thus
necessary, in most, to refer to literature approaches and
values of measurement uncertainty. The latter is most
often parameterized by the coefficient of variation of
measurement uncertainty (COVm). Based on the above,
nuggets were tentatively calculated as:
in which
where C0 is the semivariogram nugget, C + C0 is a “characteristic” conservative estimate of the

is the sill (i.e., the maximum value of semivariance), CPTU parameter ξ of interest. Comprehensive tables
and a is the range (i.e., the lag distance at of measurement uncertainty parameters for laboratory
which semivariance attains the sill value). and in-situ testing are available in Uzielli (2008). Kul
The spherical model displays a linear behavior at hawy & Trautmann (1996) reported COVm of meas
short lag distances; subsequently, it flattens and urement uncertainty for qc and fs in CPTU testing as
reaches the sill at a. The exponential model and the COVm(qc)=0.07 and COVm(fs)=0.12. Gauer et al.
Gaussian model instead approach their sill value (2002) estimated the COVm of measurement uncer
asymptotically. For these models, the range a is tainty for u2 at COVm(u2)=0.07. The corrected cone
defined as the lag distance at which the semivario tip resistance qt is not measured directly but is
gram value attains 95% of the sill value. Like the obtained from qc and u2 by the model given in cor
spherical model, the exponential model is linear at rected cone tip resistance model. Assuming that qc
very small lag distances but reaches the sill value and u2 are independent and statistically uncorrelated
more gradually. The Gaussian model, instead, shows parameters, the COVm of measurement uncertainty of
a parabolic behaviour near the origin, and is charac qt can be approximated using the following additive
terized by an inflection point. model:
The three models served as candidates in the ana
lysis of spatial correlation of CPTU parameters
within the Treporti site. The choice of the semivario
gram model reflects the user’s assessment regarding
435
The standard deviation of measurement uncer C0; sill C0+C) and relative nugget C0/(C0+C), for qt,
tainty can be expressed by inverting the definition of fs and u2.
the coefficient of variation:
Assigning a constant value to a coefficient of vari

ation implies the acceptance of the fact that the stand
ard deviation varies proportionally with the expected
value of the parameter. As qc, fs and u2 generally
increase with overburden stress (i.e., with depth) in
homogeneous soil layers, measurements at greater
depths are assumed to be less precise and accurate than
measurements taken at shallow depths. This hypothesis
is consistent with piezocone practice, as deeper meas
urements potentially suffer from the undesirable effects
of inclination, elastic compression of rods, filter
obstruction, etc. Such effects hinder the accuracy and
precision of testing results, ultimately increasing meas
urement uncertainty. The characteristic value in ξN was
taken as the arithmetic mean m(ξ) of an extended data
sample (comprising measurements pertaining to the
depth interval defined by the nominal analysis depth ±
a depth tolerance, established at 0.10 m to account for
possible measurement offsets) increased by 3 sample
standard deviations s(ξ) of the same data sample.
Fitting of semivariogram models was performed on
empirical semivariograms for each parameter and for
each analysis depth, for a total of 120 fitting oper
ations. With the exception of the assignment of
semivariogram nuggets as previously described, the Figure 4. Semivariogram models fitting to the empirical
semivariograms of qt at 32 m (a), of u2 at 29 m and of fs at
fitting procedure was carried out manually, with no 27 m.
implementation of numerical optimization tech
niques. It is well known that manual semivariogram A number of observations can be drawn regarding
fitting is preferable over automatized fitting (e. the best-fit semivariogram models and their respective
g. Isaaks & Srivastava 1990; Journel & Huijbregts parameters. The most frequent best-fit model was
1978). With the aim of obtaining the best possible found to be the spherical model. Concerning the range
adjustment to the empirical data, the three a, neither a relationship between the values relative to
models were tested comparatively: semivariogram qt and fs at each analysis depth, nor a particular depth-
model parameters a, C0 and C were varied subject wise trend is identified for either parameter. CPTU
ively, and the goodness-of-fit of the resulting semi- parameters qt and fs show instead a quite strong correl
variance values was assessed qualitatively. In the ation regarding the nugget term C0. More specifically,
latter process, a greater weight was awarded to the a positive statistical correlation (quantitatively param
adherence of theoretical models to empirical values eterized by a linear correlation coefficient 0.7) between
at small lag distances, which result in a greater the C0 values of qt and those of fs was noted. For these
number of data pairs and are thus more reliable two parameters, C0 shows a proportional increase with
than those pertaining to large lag distances. depth. This is at least partly a consequence of the cal
Despite the typical lack of complete regularity culation method adopted for C0, which partially
and noted in most of the calculated empirical semi relies on the definition of a characteristic value which
variograms, the overall qualitative assessment of the is related to second-moment sample statistics. As both
goodness-of-fit fit of semivariogram models was at qt and fs tend to increase with depth due to in-situ
least satisfactory. For illustrative purposes, best-fit stress effects, it may be expected that increasing
cases of spherical, exponential and Gaussian models sample statistics would induce the above noted correl
are shown in Figure 4. These cases refer to qt at ation. Comparative examination of the relative nugget
depth 32 m, u2 at depth 29 m and fs at depth 27 m. values showed that the contribution of measurement
Figure 5 illustrates instead numerical tables with, error and small-scale variation to spatial variability is
for each analysis depth, the best-fit semivariogram more relevant (approximately by one order of magni
model (S: spherical; G: Gaussian; E: exponential), tude) for fs than for qt and u2. This observation is con
best-fit semivariogram parameters (range a; nugget sistent with the well-known fact that sleeve friction
436
well-established knowledge regarding the degree of
spatial correlation and the magnitude of measurement
uncertainty in cone resistance, sleeve friction, and
pore pressure measurements.
Aside from their inherent utility in describing
spatial correlation structures, the analyses per
formed herein could serve as inputs to two categor
ies of analyses aimed at non-deterministic
geotechnical site characterization; namely: geosta
tistical kriging and random field modelling. In the
first case, semivariogram model parameters serve
as direct inputs to geostatistical kriging methods,
specifically in the calibration of coefficients in the
linear kriging interpolation methods. In the second
case, model parameters can be used to estimate
scales of fluctuation, which are necessary for the
definition of random fields. Both approaches are
cost-efficient means in providing estimates of soil
properties at unsampled locations and to param
eterize the associated uncertainty for seamless
inclusion in rapidly developing statistical and
probabilistic geotechnical methods.
REFERENCES
Hight, D.W. & Leroueil, S. 2003. Characterization of soils
for engineering purposes. Characterization and engin
eering properties of natural soils, Volume I, Tan et al.
Eds, Balkema Publishers.
Isaaks, E.H. & Srivastava, R.M. 1990. An introduction to
applied geostatistics. Oxford University Press,
New York.
Journel, A.G. & Huijbregts, C.J. 1978. Mining geostatis
tics. Academic Press, Inc.
Kulhawy, F.H. & Trautmann, C.H. 1996. Estimation of
in-situ test uncertainty. Uncertainty in the Geologic
Environment: From Theory to Practice, Geotechnical
Special Publication No. 58: 269–286. New York: ASCE.
Ricceri, G. & Butterfield, R. 1974. An analysis of com
pressibility data from a deep borehole in Venice. Geo-
technique, 24(2), pp. 175–192.
Simonini, P., Ricceri, G., and Cola, S. 2007. Geotech
nical characterization and properties of the Venice
lagoon heterogeneous silts. In Proc. 2nd International
Workshop on Characterization and Engineering Prop
erties of Natural Soils, Singapore, 29 November –
Figure 5. Best fit semivariogram models and parameters 1 December 2006, pp. 2289–2327. London: Taylor &
for qt, fs and u2. Francis.
Tonni, L. & Gottardi, G. 2009. Partial drainage effects in the
measurements are affected by a larger measurement interpretation of piezocone tests in Venetian silty soils. In
uncertainty. Proc. 17th Int. Conf. on Soil Mech. & Geot. Eng., Alex
andria, Egypt, 5-9 October 2009, Vol. 2, pp. 1004–1007.
Tonni, L. & Gottardi, G. 2011. Analysis and interpretation
4 CONCLUSIONS of piezocone data on the silty soils of the Venetian
lagoon (Treporti test site). Canadian Geotechnical Jour
This paper has illustrated the background concepts nal 48(4): 616–633.
and operational details of the analysis of spatial cor Tonni, L. & Gottardi, G. 2019. Assessing compressibility
characteristics of silty soils from CPTU: lessons learnt
relation of CPTU parameters measured at the Treporti from the Treporti Test Site, Venetian Lagoon (Italy).
site through semivariogram-based modelling. Despite Aims Geosciences 5 (2): 117–144.
the particularly high stratigraphic complexity of the Uzielli, M. 2008. Statistical analysis of geotechnical data.
site, the inspection of output characteristic parameters Keynote paper, Proceedings of the 3rd International
of best-fit semivariogram models provided interesting Conference on Site Characterization, Taiwan, April 1-4,
qualitative and quantitative correspondences with 2008. The Netherlands: Taylor & Francis.
437
© 2022 the Author(s), ISBN 978-1-032-31259-0
Thin-layer detection from the cone resistance rate of change

H.B. Hammer
Dr.techn. Olav Olsen AS, Trondheim, Norway
S. Nordal
Norwegian University of Science and Technology (NTNU), Norway
J.-S. L’Heureux & H. Skrede

Norwegian Geotechnical Institute (NGI), Norway
ABSTRACT: True determination of soil parameters for thin clay layers in sand or silts from CPTU meas
urements is challenging. Cone resistance in thin layers is influenced by the surrounding layers resulting in
unrepresentative measurements. Measurements should be corrected for this effect prior to interpretation of
parameters. Such correction requires accurate information about the thickness of the thin layer. Previous stud
ies have shown that the pore pressure parameter of the CPTU may not properly identify the layer interfaces
for thin clay layers in sand. In this study a “cone resistance rate of change” parameter is suggested for thin-
layer interface detection. Results from recently performed physical experiments at NTNU as well as from
prior studies are used to evaluate the ability of this parameter to detect thin layers of clay in sand. The param
eter appears to detect layer interfaces with good accuracy, even for layers with thickness as thin as the cone
diameter. The results suggest that the approach may estimate layer thickness quite well and aid towards effi
cient correction of cone resistance to achieve more realistic soil parameters for thin clay layers in sand.
1 INTRODUCTION evaluations in the current study relies exclusively on

the cone resistance measurements and soil type (i.e.,
The near-continuous measurements of CPTU sand or clay), rather than the geotechnical parameters
parameters cone resistance (qc, qt), pore pressure of the soils.
(u2) and skin friction (fs) provide great details of
the subsurface. The combined response of these
parameters can be used for characterization of 2 TRANSITION AND THIN-LAYER EFFECTS
materials for thick homogenous sediments. Close
to layer interfaces and in thin layers the measure Values of the cone resistance that only reflect
ments may be significantly influenced by multiple a single, homogenous material is labeled the
materials simultaneously. Accurate characteriza characteristic cone resistance of the material.
tion and interpretation of geotechnical parameters Close to a layer interface between different
of thin layers requires correction of these effects, materials, the measured cone resistance may devi
which in turn depends on the layer thicknesses. It ate from the characteristic cone resistance. This
is therefore important to obtain detailed informa is due to factors such as the cone geometry and
tion of the layer interfaces. The u2 parameter difference in stiffness and strength between the
may not provide accurate measurement of the materials. The distances of transition effects in
thickness of such thin clay layers (Hird et al. the materials are labeled sensing- and developing
2003, van der Linden et al. 2018, Hammer et al. distance, as illustrated in Figure 1 (a).
in press). In a thin layer, where the layer thickness is less
This study assesses the possibility of detecting layer than the sum of the sensing- and developing distance
interfaces from cone resistance measurements based of the thin layer, the extreme value will not equal the
on its rate of change. Layering between different com characteristic cone resistance, known as thin-layer
binations of sand and clay layers are assessed from effects. An example of this is illustrated in Figure 1
physical experiments. There are complex relations (b). The characteristic cone resistance in thin layers
between the mechanical properties of different mater can be estimated through correction factor, KH
ials and the measured cone resistance. However, the (Youd & Idriss 2001):
DOI: 10.1201/9781003308829-61
438
i, i.e., qi. These values have a corresponding measure
ment depth zi, where values of depth are increasing
with increasing indices.
Assuming a constant characteristic cone resist
ance for each layer, such as for the examples in
Figure 1, the derivate of the measured cone resist
ance is expected to reflect the transition effects. The
derivative can be expressed as the change of meas
ured cone resistance over the distance between
measurements. The derivative of q becomes:
Figure 1. Illustration of transition- and thin-layer effects The unit of this parameter is given in units of
for (a) a two-layered composition and (b) a three-layered stress, e.g., MPa.
composition.
Hammer et al. (in press) proposed a normalized
parameter of cone resistance rate of change, q’,
defined as Δq/ΔzN divided by the average cone
resistance between the two depths:
Values of KH is less than or equal 1 in thin weak

layers, e.g., Figure 1(b). For the opposite case, in
thin strong layers, values are greater or equal 1. Fac
tors of KH are proposed for thin sand layers in clay
(i.e., thin strong layers) based on field data and
experiments. It is however a lack of proposed correc
tion factors for thin weak layers.
An advantage of normalizing the derivative of the
The proposed correction factors are typically
average measurement is an increased emphasize on
dependent on the thin layer thickness compared to
transition effects in weak materials. This parameter
the cone diameter (H/dc) and the contrast between
showed promise for detecting interfaces for thin clay
the characteristic cone resistance of the thin layer
layers.
and the surrounding layers (i.e., qthin/qsurrounding).
The procedure of Boulanger & DeJong (2018) con
More advanced methods of thin layer corrections
sisted of three main components, where the first two
have been proposed, most noticeably the inverse fil
corrects measurements of cone resistance for thin layer
tering procedure (Boulanger & DeJong 2018).
effects. The last component attempts to correct for tran
sition effects. Profiles corrected for thin-layer effects
are evaluated based on the rate of change to identify
3 CONE RESISTANCE RATE OF CHANGE
and approximate sharp transition (i.e., interfaces). The
resistance rate of change was defined as:
The cone resistance is in this study presented against
the depth of the cone tip (z) normalized on the cone
diameter (dc):
Interfaces was in the study determined primarily

The depth is furthermore referenced to the depth based on whether the values of m (calculated from
of a layer interface (zint): cone resistance profiles corrected for thin-layer
effects) were greater than 0.1.
Both q’ and m are parameters describing the relative
change in cone resistance over normalized distances
(ΔzN). The two parameters are compared for various
relations of qi+1/qi over one cone diameter distance
The subscript of the cone resistance measurement (ΔzN =1) in Figure 2. The parameter m is for this situ
type is omitted in this study. I.e., the symbol q is used ation represented with f1 while f2 represents q’. Two
rather than qc or qt. The latter is the corrected cone additional relationships are added for comparison, f3
resistance for pore pressure due to unequal area effects. and f4, these represent Δq/ΔzN normalized on qi and
Each measurement of q can be labeled with and index qi+1 respectively. From this it is evident that there is
439
a negligible difference between q’ and m for small
relative changes, qi+1/qi. Due to the numerical advan
tages of computing the function f1 rather than f2, the
parameter m is used in this study for the cone resist
ance rate of change. Each measurement of mi is
assigned to the average depth between zi and zi+1.
Note that if the distances between measurements are
approximately constant, this only results in a change
the reference depth.
Figure 3. Example of values of m compared to Δq/ΔzN.

Measurements from de Lange (2018), containing multiple
thin clay and sand layers of thicknesses of H=3.2dc.
between sand layers. The names of the series indicate the
applied surcharge.
The physical experiments are performed on con

structed samples of layered sands and clays in cham
Figure 2. Different measurements of cone resistance rate bers. Preparation methods and material properties of
of change. sands and clays vary between the different experi
ments. In general, the sands in the created samples are
described as clean, uniform, and homogenous. Clay
layers are also homogenous and primarily described as
The advantages of the cone resistance rate of soft. The relations between the material properties and
change m compared to the derivative Δq/ΔzN is the measured cone resistance are not evaluated in this
highlighted through an example of one of the study.
physical experiments from de Lange (2018). The CPT probes used in all experiments have 60°
Figure 3 presents values of q together with apex while the cone diameters are varying from 1 to
calculated values of m and Δq/ΔzN. The soil 3.6 cm. The diameter size (rounded in millimeters)
sample has thin alternating clay and sand layers of of the different experiments are shown in Table 1.
H=3.2dc between two sand layers. Three measure The results of Hammer (2020) and Skrede (2021)
ments of the same sample were performed with the are corrected for unequal area effects, however, the
sample exposed to a surcharge load of 25, 50 and differences in thin clay layers were found to be
200 kPa. neglectable. The measurements from literature are
The two primary advantages of m compared to primarily not corrected for unequal area effect.
Δq/ΔzN are firstly the apparent independency on the
stress level, as observed from the figure. Secondly,
values of m appear to reach extreme values at con
stant distances to the layer interfaces. The second Table 1. Physical experiments of cone penetration in cali
bration chambers of layered sands and clays evaluated in
advantage will be explored in chapter 5.
this study.
Reference Materials dc [cm]

4 DATA FROM PHYSICAL EXPERIMENTS
van der Berg (1994) Sand, clay 3.6
4.1 Experiments Teh et al. (2010) Sand, clay 1
Młynarek et al. (2012) Sand, clay 3.6*
The normalized cone resistance rate of change was Tehrani et al. (2017) Sand 3.2
applied to measurements from physical experiments van der Linden et al. (2018) Sand, clay 2.5
on layered sand and clay. The aim to determine its de Lange (2018) Sand, clay 2.5
ability to detect layer interfaces as well as character Wang (2019) Clay 1
ize thin-layer- and transition effects. Results Hammer (2020) Sand, clay 3.6
of recent experiments performed at NTNU are evalu Skrede (2021) Sand, clay 3.6
ated together with multiple studies from literature.
A list of all references of the experiments evaluated * Probe properties were not found, a standard size (10cm2)
in this study is presented in Table 1. is assumed
440
4.2 Results from chamber tests at NTNU
The problem of thin clay layers in sand has been studied
through large scale physical experiments in a CPTU
chamber at NTNU (Norwegian University of Science
and Technology) in Trondheim during the last two
years. Samples were constructed in a chamber of
1.2 m diameter and 1 to 2.2 m height. Multiple cases of
varying thin-layer thickness were tested at varied stress
states. Details on the experiments and results are pre
sented in the MSc theses Hammer (2020) and Skrede
(2021). A combined six sample cases were constructed
and tested in the two studies, named E1 - E6.
4.3 Results from literature

Multiple studies have been performed on physical
experiments of cone penetration in layered sand and/or
clays. The experiments that are found to be relevant for
this study are summarized in Table 1. The cone resist
ance measurements of these studies were digitized
from figures in a detailed manner. Depth measurements Figure 4. q and m profiles for layered sand, experiments of
were converted to the normalized depth of the cone tip Tehrani et al. (2017). Top: dense over loose sand. Bottom:
below a layer interface (zN;int). The measurements Loose over dense sand.
from literature were interpolated at 0.3dc intervals
since the actual intervals between measurements are
unknown. For a standard cone with area 10 cm2 this
corresponds to measurements each 1 cm.
5 RESULTS
5.1 Sand
The profiles of two experiments with dense over loose
sand from Tehrani et al. (2017) are presented in
Figure 4 (top). Minimum values of m coincide with the
layer interface with values of -0.2 and -0.3. The
study defined sensing lengths of 5.1dc for both experi
ments, reflecting approximately the distance from
where m changes sign to the layer interface. Developing
distances were defined as 2.2dc and 2.4dc, respectively.
Figure 4 (bottom) show the opposite layering, i.e.,
loose sand over dense sand. The maximum values
of m for the two experiments were 0.5, these values occur
at a distance 1-2dc prior to the layer boundary. Sensing
distances were described as 2.8 dc and 3 dc, respectively, Figure 5. q and m profiles for layered clay, experiments
while developing distances were 3.8 dc and 3.9 dc. from Wang (2019). Top: stiff over soft clay. Bottom: soft
over stiff clay.
5.2 Clay
Transition effects between clay layers was evaluated 5.3 Sand and clay
in the study of Wang (2019). Results of experiments
with stiff clay over soft clay are presented in the top Various studies have included experiments on thick
of Figure 5. Values of m decrease only from layers of sand and clay. These include van der Berg
a distance of about 1dc prior to the layer interface. (1994), Teh et al. (2010), Młynarek et al. (2012), van der
The transition towards the characteristic cone resist Linden et al. (2018) and Skrede (2021). Four of these
ance of the soft clay layer (i.e., the developing are presented in Figure 6 for a thick sand layer over
length) appear to be over about 4-6dc. clay. Three of the q-profiles are increasing until 2-3dc
Results of experiments with soft over stiff clay distance to the interface due to the proximity to the top
are shown in Figure 5 (bottom). Sensing distance is surface of the sample. This causes greater uncertainty in
for this layering case about 1dc as well, while the the interpretation of sensing distance. An approximation
developing distance appear to be about 2dc. of the sensing distance may be 2-3dc for these three
441
measurements and 5-6dc for Młynarek et al. (2012). decreasing layer thickness. Sensing and developing
Extreme values of the m-profiles were between -0.8 and distances in the sand can be approximated to 2-3dc
-0.6, reached within 1dc of the actual layer boundary. and 2dc, respectively.
5.5 Multiple thin sand and clay layers

Numerous experiments on samples with multiple
thin sand and clay layers of equal thickness is pre
sented in the study of de Lange (2018). Here, three
of these are presented. The first is the “exploratory
test 4” presented in Figure 3 with thicknesses 3.2dc.
Figure 9 and Figure 10 present respectively “soil
model 02” with layer thicknesses 1.6dc and “soil
model 08” with layer thicknesses 0.8dc. Both the
experiment with thicknesses 3.2dc and 1.6dc reaches
extreme values of m very close to or at the inter
faces. Extreme values of m prior to clay layers and
Figure 6. q and m profiles for two-layered samples of sand sand layers are respectively about -0.6 and 1-1.8. For
over clay from multiple studies. the sample with layer thicknesses 0.8dc the values
of m are significantly lower than the other two
experiments. However, the shape of the m-profile is
The opposite layering is presented in Figure 7. largely the same and extreme values of m correspond
Similar to the two-layered clay in section 5.2, the to interfaces fairly well.
clay layers exhibit a very short sensing distance of
about 1dc. The contrast (i.e., ratio) between the char
acteristic cone resistance of the bottom sand layer and
the clay layer varies from 25 to 100. The developing
distance displayed in the sand layer appear to be
approximately 3dc. Extreme values of the m-profiles
vary from 2.6 to 4.4 and occur at the layer interface.
Figure 7. q and m profiles for two-layered sand and clay

from multiple studies. Top: sand over clay. Bottom: clay
over sand.
5.4 Thin clay layer in sand

Measurements from NTNU experiments E2
(Hammer 2020) and E5 (Skrede 2021) of thin clay
layers with thickness 0.56-2.2dc are presented in
Figure 8. The characteristic cone resistance of the
sand layers were about 10 times that of the clay
layer. The m-profiles for the three different layer
thicknesses primarily have extreme values for the
top transition of between -0.6 and -0.5. These occur Figure 8. q and m profiles for thin clay layers in sand.
very close to the clay layer interface. Extreme values Experiments from Hammer (2020) and Skrede (2021). Top:
of m for the bottom transition occur 0-1 dc below the H= H=2.2dc. Middle: H=1.1dc. Bottom: H=0.56 dc (each
bottom interface. The extreme values decrease with measurement point is marked with dots).
442
key in understanding the thin-layer effects acting
in sand and clays. Figure 11 illustrates the
expected thin-layer effects in (a) a thin clay layer
in sand and (b) a thin sand layer in clay due to the
difference in sensing and developing distances.
6.2 Layer interface detection

Based on the studied experiments, layer interfaces
between sand and clay may be interpreted based
on representative extreme values of m. The top
interface of clay layers can be approximated at
Figure 9. q and m profiles for multiple thin layers of sand a depth with an extreme value of m less than -0.4.
and clay, H=1.6dc. Experiments from de Lange (2018). For layers with small contrast in characteristic
cone resistance, such as the two-layered sand/
or clay, a value of about -0.2 can indicate a layer
interface. A bottom interface of a clay layer
may be interpreted at the depth where the
parameter m reaches an extreme positive value of
at least 0.5.
7 CONCLUSIONS
A cone resistance rate of change parameter shows

promising possibilities to detect layer interfaces
Figure 10. q and m profiles for multiple thin layers of sand between sand and clays. The parameter additionally
and clay, H=0.8dc. Experiments from de Lange (2018). yields useful information on transition effects. It
allows for efficient interpretation of layer boundaries
and may even detect layers as thin as the cone diam
eter. The use of the parameter for interface detection
depends on a significant contrast in characteristic
cone resistance between layers. Testing on field data
is needed to look deeper into the possibilities and the
limitations for this method of layer interface detec
tion in naturally formed deposits.
REFERENCES
Boulanger, R., & DeJong, J. 2018. Inverse filtering proced
Figure 11. Illustration of the expected thin-layer effects in ure to correct cone penetration data for thin-layer and
(a) a thin clay layer embedded in sand and (b) a thin sand transition effects. Proceedings of the 4th International
layer embedded in clay. Black lines are the characteristic Symposium on Cone Penetration Testing (CPT’18),
cone resistance. Delft, 21-22 June.
de Lange, D.A. 2018. CPT in Thinly Layered Soils (J. van
Elk & D. Doornhof, Eds.; No. 1209862–006-GEO-0007).
Hammer, H.B. 2020. Physical experiments on CPTU thin-layer
6 DISCUSSION effects of thin clay layers embedded in sand (Master’s
thesis). Norwegian University of Science and Technology
6.1 Sensing and developing lengths (NTNU). Trondheim, Norway. 11250/2689484
Hammer, H.B., Nordal, S. and L’Heureux, J.-S. In press.
It is evident from the experiments that the sensing Detection of thin clay layers in sand using a standard
and developing distances of sand and clay are sig CPTU probe. 20th International Conference on Soil
nificantly different. While penetration in sands Mechanics and Geotechnical Engineering (ICSMGE),
exhibit an almost equal sensing- and developing Sydney, 1-5 May.
distances, there are large differences in clays. In Hird, C., Johnson, P., & Sills, G. 2003. Performance of
clays, the sensing distance appear to typically be miniature piezocones in thinly layered soils. Geotechni
que, 53(10),885–900. 10.1680/geot.2003.53.10.885
about one cone diameter in front of the cone tip, Młynarek, Z., Gogolik, S., & Półtorak, J. 2012. The effect
while the developing distance are up to six times of varied stiffness of soil layers on interpretation of
the amount. This difference is considered to be CPTU penetration characteristics. Archives of civil and
443
mechanical engineering, 12(2),253–264. 10.1016/j. van der Linden, T.I., De Lange, D.A., & Korff, M.,
acme.2012.03.013 2018. Cone Penetration Testing in Thinly
Teh, K. L., Leung, C. F., Chow, Y.K. & Cassidy, J. 2010. Inter-Layered Soils. Geotechnical Engineering.
Centrifuge model study of spudcan penetration in sand 10.1680/jgeen.17.00061
overlying clay. Géotechnique, 60(11),825–842. 10.1680/ Wang, Y. 2019. Centrifuge Modelling and Numerical
geot.8.P.077 Analysis of Penetrometers in Uniform and Layered
Tehrani, F.S., Arshad, M. I., Prezzi, M., & Salgado, R. 2017. Clays. (Doctoral dissertation, The University of
Physical modeling of cone penetration in layered sand. Western Australia). 10.26182/5d14603a91815
Journal of Geotechnical and Geoenvironmental Engin Youd, T.L. and Idriss, I.M. 2001. Liquefaction resist
eering, 144 (1).10.1061/(ASCE)GT.1943-5606.0001809 ance of soils: summary report from the 1996
Skrede, H. 2021 CPTU-detection of thin clay layers in NCEER and 1998 NCEER/NSF workshops on
sand (Master’s thesis). Norwegian University of Science evaluation of liquefaction resistance of soils. Jour
and Technology (NTNU). Trondheim, Norway. nal of Geotechnical and Geoenvironmental Engin
van den Berg, P. 1994. Analysis of soil penetration. (Doc eering, ASCE, 127(4),297–313. 10.1061/(ASCE)
toral dissertation, Delft University of Technology). 1090-0241(2001)127:4(297)
444
© 2022 the Author(s), ISBN 978-1-032-31259-0
Recalculation of in-situ CPTu in intermediate soils using G-PFEM

L. Hauser, S. Oberhollenzer, A. Gharehaghajlou, H.F. Schweiger & R. Marte
Institute of Soil Mechanics, Foundation Engineering and Computational Geotechnics, Graz University of
Technology, Graz, Austria
C. Fabris
Geoconsult ZT GmbH, Puch bei Hallein, Austria
ABSTRACT: Piezocone penetration in intermediate soils is often associated with partially drained
behaviour which has to be taken into account when interpretating the results. As part of the research
project PITS, the effect of partial drainage on in-situ testing in postglacial silty deposits in Austria is
investigated by means of in-situ testing campaigns, laboratory testing and numerical simulations. The
present work presents a numerical recalculation of in-situ CPTu, carried out with different penetration
velocities in a silty soil layer at the test site Rhesi near lake Constance, using the application
G-PFEM. The results agree reasonably well with the in-situ measurements considering the given het
erogeneity of the soil layer and suggest that partial drainage occurs during the penetration process.
1 INTRODUCTION 2 IN-SITU TESTING CAMPAIGN
The interpretation of piezocone testing (CPTu) in inter As part of the ongoing research project PITS
mediate soils is a challenging task considering the fact a comprehensive in-situ testing campaign was
that standard testing with a penetration velocity, v, of carried out at test site Rhesi, located in the fore-
2 cm/s often takes place under partially drained condi lands of the river Rhine near lake Constance in
tions. This has been shown, e.g. by Paniagua et al. Austria. CPTu tests with different penetration
(2013) on the basis of small-scale experiments where velocities of 0.1, 0.5, 2 and 10 cm/s were car
local drainage around the cone tip was observed. ried out along with dissipation tests at different
Therefore, specific approaches have been developed to depths using u1 and u2 probes. Additionally, soil
quantify partial drainage at normalized penetration vel samples were recovered for laboratory testing
ocities (Randolph and Hope, 2004; Schnaid et al., and seismic flat dilatometer tests (SDMT)
2020) and to account for this effects when deriving were performed. Figure 1 shows the obtained
parameters, such as the coefficient of consolidation profiles of qt, fs, u1 and u2 over depth for the
(DeJong and Randolph, 2012), from CPTu data. penetration velocities of 0.5 and 2 cm/s. For the
The research project PITS, initiated at TU Graz in following numerical study, layer L2, character
cooperation with the Federal Chamber of Architects ized by sand-silt alterations, is of main interest
and Chartered Engineering Consultants, aims to and a 0.5 m thick sublayer located at a depth of
improve the characterization of postglacial silty around 11.3 to 11.8 m is selected for recalcu
deposits in Austria by means of in-situ testing lation. According to EN ISO 14688-1, the
addressing, inter alia, the effect of partial drainage. sublayer is classified as sa’ cl’ Si. Overall,
The present work focusses on the numerical recal the in-situ results show a clear scatter due to the
culation of CPTu tests carried out in a silty layer, sand-silt alterations. However, the calculated
where partially drained behaviour is observed, with mean and median values of qt, fs, u1 and u2
penetration velocities of 0.5 and 2 cm/s. First, the in- are in good agreement for the considered sub
situ testing campaign and the numerical model are layer. A shear wave velocity of around 170 m/s
outlined. Then, the results of the recalculations are was measured at the considered depth and
presented and discussed highlighting the partially a friction angle of 30° results from direct shear
drained behaviour during penetration. testing.
DOI: 10.1201/9781003308829-62
445
Figure 1. In-situ measurements (qt, fs, u1, u2) over depth at test site Rhesi; depth used for the numerical recalculation
marked at 11.5 m below the ground surface.
3 NUMERICAL MODEL equations in an Updated Lagrangian manner is used

in connection with low equal order approximations
The numerical simulation of cone penetration testing (linear triangular elements) of the displacement and
is a challenging task involving large displacements pore pressure fields. Additionally, the determinant of
and deformations of the soil around the penetrometer, the deformation gradient is introduced as a nodal
nonlinear material behaviour and frictional contact variable (Monforte et al., 2017b). Constraints due to
between the ideally rigid cone and the deformable frictional contact between cone and soil, according
soil. Advanced numerical methods have already been to a Coulomb friction law, are enforced using
successfully applied to model cone penetration (Cian a penalty method. Also, no water flow normal to the
tia et al., 2016; Ceccato et al., 2016; Sheng et al., contact interface is allowed as the rigid body is
2009; Yuan et al., 2019). For the present work, the impervious. The interested reader is referred to Mon-
application G-PFEM (Carbonell et al., 2022), an forte et al. (2017a) and Carbonell et al. (2022) for
implementation of the Particle Finite Element Method a more detailed outline of the application G-PFEM.
(PFEM) developed for modelling large deformation For the present study, the Clay and Sand Model
problems in geotechnics within the framework Kratos (CASM), a non-associated elastoplastic model pro
(Dadvand et al., 2010), is used. posed by Yu (1998), is used allowing for a flexible
definition of the yield surface geometry by means of
the model parameters r and n. For particular combin
3.1 G-PFEM
ations of r and n the surfaces of the Original Cam
In the PFEM, a given boundary value problem is Clay model (OCCM) and the Modified Cam Clay
solved performing frequent remeshing of critical model (MCCM) may be obtained as shown in
regions of the integration domain which is treated as Figure 2. Adapted to finite strains according to Mon-
a cloud of particles carrying all information. At the forte et al. (2015), the model features the hyperelas
beginning of a time step, particles/nodes may be tic model by Houlsby (1985) with the modified
added or removed in certain regions. Subsequently, swelling index κ* and the shear modulus
the boundaries of the domain are defined and the G. Coupling of volumetric and deviatoric response is
mesh is created which is used for solving the compu not considered, thus, the parameter α equals zero.
tation step by means of the Finite Element Method The classical critical state hardening rule is adopted
(FEM). Finally, an updated cloud of nodes/particles according to which the preconsolidation pressure
is obtained at the end of the time step (Oñate et al., evolves with plastic volumetric strain depending on
2011). Large deformations of the domain can be κ* and the modified compression index λ*. The
treated efficiently as excessive mesh distortion is implemented version of the CASM and its extension
avoided, however, at the cost of increased computa for structured material as well as the FD_MILAN
tional effort. A stabilized, mixed formulation of the model have already been used in connection with
quasi-static linear momentum and mass balance G-PFEM for modelling cone penetration (Hauser
446
and Schweiger, 2021; Monforte et al., 2021; Oliynyk 4 RESULTS OF THE RECALCULATION
et al., 2021).
The recalculation was performed for a 0.5 m thick sub
layer at an average depth of 11.55 m leading to the
3.2 CPTu model
constant initial stress field defined through σ’v = 128
The penetration process is modelled in 2D as an axi kPa, σ’h = 64 kPa and u0 = 110 kPa with K0 being 0.5.
symmetric problem consisting of a rectangular, For this study, the yield surface of the CASM is
deformable soil domain (measuring 2.2 m in height assumed to be a MCC ellipse (r = 2, n = 1.5) since no
and 1 m in width) and a rigid cone with 15 cm² base experimental evidence for an alternative, more suitable
area and a tip angle of 60°. Overburden pressure is choice of r and n is available. This results in an associ
applied on top of the domain while the lateral and ated flow rule which influences the pore pressure
lower boundaries are fixed in normal direction (see development. However, the effect of a non-associated
Figure 3). Initially, the probe is already inserted in the flow rule is not addressed in this study. The overconso
soil and the downwards movement of the rigid body lidation ratio (OCR) is equal to 1. G0 is calculated
starts from there. A constant initial stress field is from Vs for a density of 1.83 g/cm³ and a ratio G0/G =
applied assuming weightless soil and the stationary 5 is used to estimate the shear modulus at working
values of qt, fs, u1 and u2 are reached after an initial strains, G, of 10600 kPa. Marchetti and Monaco
transient phase of roughly 10 radii of penetration. The (2018) suggest a similar value of around 4.2 for G0/G
initial mesh, the contact stiffness and the criteria for for a clayey silt at a site in Italy. Assuming uncoupled
mesh refinement were chosen based on previous stud volumetric and deviatoric behaviour in the hyperelastic
ies (Hauser and Schweiger, 2021). model (α = 0), κ* is determined in such a way that
a Poission’s ratio of 0.2 is obtained for the initial mean
stress and the given G. Eventually, λ* is defined
through the ratio λ*/κ* = 4. The parameter set, as sum
marized in Table 1, yields an undrained shear strength
of 39 kPa which is obtained from a numerically simu
lated triaxial compression test for the given initial
stress state. A contact friction angle of 9° at the inter
face between penetrometer and soil is used which
proved to be a reasonable assumption in previous stud
ies (Hauser et al, in press).
Table 1. Material parameters for the numerical recalcu

lation of a silty layer at a depth of 11.55 m at test site Rhesi.
γ [kN/m³] φ [°] r [-] n [-] OCR [-]
0 30 2 1.5 1
κ* [-] G [kPa] α [-] λ* [-] K0 [-]
0.0061 10600 0 0.0244 0.5
Figure 2. CASM yield surfaces for different combinations
of r and n in the q-p’ space normalized with respect to
respective preconsolidation pressure p’0.
Figure 4. Results of the recalculation of qt, fs, u2 over nor

Figure 3. Axisymmetric model and refined mesh during malized depth compared to the in-situ tests executed for
penetration. v = 2 cm/s with a u1 und u2 piezocone, respectively.
447
Figure 5. Comparison of in-situ measurements (square markers and dashed lines) and numerical recalculations for penetra
tion velocities of 0.5 and 2 cm/s as well as permeabilities of 1*10-7 and 2*10-7 m/s in terms of qt, fs, u1, u2 and pore pres
sure dissipation curves. Note that the in-situ u2 pressure for v = 0.5 cm/s is not available for comparison.
448
Recalculations were carried out for the penetration
velocities of 0.5 and 2 cm/s as well as for isotropic
permeabilities equal to 1*10-7 and 2*10-7 m/s. The
results are presented in Figure 5 comparing the in-situ
and the recalculated qt, fs, u1, u2 along with the pore
pressure dissipation curves. Thereby, the numerical
results for qt, fs, u1, u2 are obtained from the penetra
tion curves as the average between 16 and 17 radii of
penetration (see Figure 4) while the in-situ values rep
resent the average over the considered 0.5 m thick
sublayer. As expected, the lower permeability k =
1*10-7 m/s leads to more “undrained like” behaviour,
resulting in lower qt, fs and higher u1, u2 as well as
longer dissipation times, compared with the higher
permeability k = 2*10-7 m/s. However, better agree
ment between the in-situ results and the recalcu
lations, in terms of qt, fs, u1, u2, is reached for k =
2*10-7 m/s as both penetration velocities match the
in-situ results reasonably well. Numerically, the dissi
pation curves at the u1 und u2 position tend to coin
cide after around 10 to 12 s which is not the case for
the in-situ measurements. For k = 1*10-7 m/s, the
recalculated dissipation curves lie between the in-situ
ones, however, the initial pore pressure is significantly
higher.
For k = 2*10-7 m/s, the dissipation time and the
initial pore pressures are reduced suggesting better
agreement with the in-situ measurements in terms of
initial pore pressure. Additional calculations with
permeabilities of 1*10-1° m/s and 1*10° m/s were
performed in order to obtain fully undrained and
fully drained references cases.
5 DISCUSSION
The recalculations employing the higher permeabil

ity of 2*10-7 m/s appear to match the in-situ qt, fs
u1 and u2 best for the two considered penetration
velocities. In case of faster penetration (2 cm/s) the
numerical response is slightly more “undrained
like” compared to the in-situ measurements while
the response is rather on the drained side for slower
penetration (0.5 cm/s). The difference between the
in-situ and the recalculated dissipation curves is
more pronounced as the pore pressure at the u1 pos
ition decays significantly faster compared to the in-
situ measurement. However, considering the hetero
geneity of the layer and the fact that the in-situ
curves for u1 and u2 originate from two different
tests, carried out in close proximity and at roughly
the same depth with the u1 and u2 probe, respect
ively, the overall behaviour, especially the non-
Figure 6. Tip resistance Qt and excess pore pressures monotonic decrease of the u2 pressure, is properly
Δu1 and Δu2 normalized with respect to the undrained reproduced by the numerical model. Also, aniso
tip resistance Qt,ref and the undrained excess pore pres tropic permeability was not considered in the pre
sures Δu1,ref and Δu2, ref, respectively over the penetra sent numerical study.
tion velocity v normalized by the permeability k (Sheng The recalculated CPTu quantities appear to be
et al., 2014). sensitive to the considered range of penetration
449
velocities and permeabilities suggesting partially Ciantia, M.O., Arroyo, M., Butlanska, J. and Gens, A.
drained behaviour. Figure 6 shows the tip resistance (2016), “DEM modelling of cone penetration tests in a
Qt and the excess pore pressures Δu1 and Δu2 nor double-porosity crushable granular material”, Com
malized with respect to the undrained reference puters and Geotechnics, Vol. 73, pp. 109–127.
Qt,ref, Δu1,ref and Δu2,ref, respectively, plotted against Dadvand, P., Rossi, R. and Oñate, E. (2010), “An
the penetration velocity normalized by the perme Object-oriented Environment for Developing Finite
ability (Sheng et al., 2014). Therefore, the undrained Element Codes for Multi-disciplinary Applications”,
Archives of Computational Methods in Engineering,
calculation for v = 2 cm/s and k = 1*10-1° m/s is Vol. 17 No. 3, pp. 253–297.
used as reference case. Compared to the undrained DeJong, J.T. and Randolph, M. (2012), “Influence of
reference, the tip resistance increases by a factor of 3 Partial Consolidation during Cone Penetration on
for v = 0.5 cm/s and k = 2*10-7 m/s and by a factor Estimated Soil Behavior Type and Pore Pressure Dis
of more than 5 for the drained case. Similarly, Δu1 sipation Measurements”, Journal of Geotechnical
decreases by a factor of up to 2.8 and Δu2 by a factor and Geoenvironmental Engineering, Vol. 138 No. 7,
of up to 5 with respect to the maximum undrained pp. 777–788.
pressure suggesting that the shape of the pore pres Hauser, L. and Schweiger, H.F. (2021), “Numerical study on
undrained cone penetration in structured soil using
sure bulb changes with drainage condition. Also, the
G-PFEM”, Computers and Geotechnics, Vol. 133, p.
recalculated tests lie within the proposed range for 104061.
partial drainage, i.e., 2*101 < v/k < 2*106 (Sheng Houlsby, G.T. (1985), “The use of a variable shear modulus
et al., 2014). in elastic-plastic models for clays”, Computers and Geo
technics, Vol. 1 No. 1, pp. 3–13.
Marchetti, S. and Monaco, P. (2018), “Recent Improve
6 CONCLUSION ments in the Use, Interpretation, and Applications of
DMT and SDMT in Practice”, Geotechnical Testing
The numerical recalculation of in-situ CPTu, per Journal, Vol. 41 No. 5, p. 20170386.
formed in a silty postglacial deposit in Austria, con Monforte, L., Arroyo, M., Carbonell, J.M. and Gens, A.
(2017a), “Numerical simulation of undrained insertion
firmed that partial drainage governs the penetration problems in geotechnical engineering with the Particle
process as the obtained qt, fs, u1 and u2 are highly Finite Element Method (PFEM)”, Computers and Geo
sensitive to the considered range of penetration vel technics, Vol. 82, pp. 144–156.
ocities and soil permeabilities. Overall, the numer Monforte, L., Arroyo, M., Gens, A. and Carbonell, J.M.
ical results match the in-situ measurements (2015), “Integration of elasto-plastic constitutive
reasonably well for the considered 0.5 m thick sub models in finite deformation: An explicit approach”,
layer, which is, to a certain degree, heterogeneous in Oñate, E., Owen, D.R.J., Peric, D. and
due to the presence of alternating thin layers of Chiumenti, M. (Eds.), Computational Plasticity XIII
Fundamentals and Application: Proceedings of the
sand and silt. Hence, the dissipation behaviour
XIII International Conference on Computational
could not be reproduced as accurately as qt, fs, u1 Plasticity, International Center for Numerical
and u2 resulting in faster dissipation of pore pres Methods in Engineering (CIMNE), Barcelona, Spain,
sure at the u1 position compared to the in-situ pp. 398–406.
results. Nevertheless, the study demonstrates that Monforte, L., Carbonell, J.M., Arroyo, M. and Gens, A.
the application G-PFEM is capable of modelling (2017b), “Performance of mixed formulations for the par
cone penetration under partially drained conditions ticle finite element method in soil mechanics problems”,
in a realistic way. Computational Particle Mechanics, Vol. 4 No. 3, pp.
269–284.
Monforte, L., Gens, A., Arroyo, M., Mánica, M. and
Carbonell, J.M. (2021), “Analysis of cone penetration in
ACKNOWLEDGEMENT
brittle liquefiable soils”, Computers and Geotechnics,
Vol. 134, p. 104123.
The authors acknowledge the financial and logistical Oliynyk, K., Ciantia, M.O. and Tamagnini, C. (2021),
support provided by the Austrian Research Promo “A finite deformation multiplicative plasticity model
tion Agency and the Federal Chamber of Architects with non–local hardening for bonded geomaterials”,
and Chartered Engineering Consultants. Computers and Geotechnics, Vol. 137, p. 104209.
Oñate, E., Idelsohn, S.R., Celigueta, M.A., Rossi, R.,
Marti, J., Carbonell, J.M., Ryzhakov, P. and
REFERENCES Suárez, B. (2011), “Advances in the Particle Finite
Element Method (PFEM) for Solving Coupled
Carbonell, J.M., Monforte, L., Ciantia, M.O., Arroyo, M. Problems in Engineering”, in Oñate, E. and
and Gens, A. (2022), “Geotechnical particle finite elem Owen, D.R.J. (Eds.), Particle-based methods: Funda
ent method for modeling of soil-structure interaction mentals and applications, Computational Methods in
under large deformation conditions”, Journal of Rock Applied Sciences, Vol. 25, Springer, Dordrecht,
Mechanics and Geotechnical Engineering. London, pp. 1–49.
Ceccato, F., Beuth, L., Vermeer, P.A. and Simonini, P. Paniagua, P., Andò, E., Silva, M., Emdal, A., Nordal, S.
(2016), “Two-phase Material Point Method applied to and Viggiani, G. (2013), “Soil deformation around
the study of cone penetration”, Computers and Geotech a penetrating cone in silt”, Géotechnique Letters, Vol. 3
nics, Vol. 80, pp. 440–452. No. 4, pp. 185–191.
450
Randolph, M.F. and Hope, S. (2004), “Effect of cone vel Sheng, D., Nazem, M. and Carter, J.P. (2009), “Some com
ocity on cone resistance and excess pore pressures”, putational aspects for solving deep penetration problems
Proceedings of the IS Osaka - Engineering Practice and in geomechanics”, Computational Mechanics, Vol. 44
Performance of Soft Deposits, pp. 147–152. No. 4, pp. 549–561.
Schnaid, F., Dienstmann, G., Odebrecht, E. and Yu, H.S. (1998), “CASM: a unified state parameter model
Maghous, S. (2020), “A simplified approach to normal for clay and sand”, Int. J. Numer. Anal. Meth. Geomech.,
isation of piezocone penetration rate effects”, Géotech Vol. 22, pp. 621–653.
nique, Vol. 70 No. 7, pp. 630–635. Yuan, W.-H., Zhang, W., Dai, B. and Wang, Y. (2019),
Sheng, D., Kelly, R., Pineda, J. and Lachlan, B. (2014), “Application of the particle finite element method for
“Numerical study of rate effects in cone penetration large deformation consolidation analysis”, Engineering
test”, Las Vegas, Nevada, USA. Computations, Vol. 36 No. 9, p. 370.
451
© 2022 the Author(s), ISBN 978-1-032-31259-0
Full-flow CPT tests in a nearshore organic clay

S. Hov
GeoMind, Stockholm, Sweden
K. Borgström
GeoMind, Stockholm, Sweden
P. Paniagua
Norwegian Geotechnical Institute and Norwegian University of Science and Technology, Trondheim, Norway
ABSTRACT: The T-bar is a full-flow CPT with a larger probe surface area compared with the conventional
CPT probe. The tip has the shape either of a ball or as an upside-down T, a so-called T-bar. The term ‘full
flow’ comes from the assumption that the earth ‘flows’ around the tip, which is a realistic assumption for soils
with extremely low undrained shear strength. This paper presents a case study where a large number of T-bar
tests have been performed in a nearshore organic clay with high water content and very low undrained shear
strength. The test site is located in Stockholm, Sweden, where planned land reclamation and capping of con
taminated top soils are challenging from a stability perspective due to the low strength of the soil. T-bar tests
were thus performed to characterise the shear strength profile of the soil in detail, especially at shallow depths
where sampling was difficult and the shear strength values were under 5 kPa. A N-factor relating the net cone
resistance (qnet) and the undrained shear strength of the soil (cu) was evaluated based on T-bar measured param
eters and direct simple shear tests and undrained triaxial tests on samples taken at greater depths. This allowed
to estimate shear strength profiles with depth. For all tests, both the penetration and extraction cone resistance
were measured, and a good correlation was obtained between this ratio and the soil sensitivity measured in the
laboratory. In addition, the sensitivity was correlated to the organic content of the clay. These correlations were
found to be OCR-dependent. Further, cyclic tests were performed, and their results were correlated with the
remoulded shear strength values. This facilitated mapping of the soil conditions across the site.
1 INTRODUCTION – The correction of overburden stress and pore pres

sure measurements are one tenth, i.e. they are
Conventional CPT tests have long been used in almost negligable
all types of soil to interpret strength characteris – The resistance can be measured during penetration
tics. In extremely soft clays the penetration resist and extraction, giving additional data on soil type
ance is however very small, and hence full-flow and behaviour
penetration tests was developed during the late
This paper presents a case study where full-flow
1990s, mainly for offshore applications (Stewart
CPT tests have been performed nearshore as part of
& Randolph, 1994; Randolph, 2004). The full-
a land reclamation project. An old industrial area,
flow probe has the shape of a ball or a upside- located in the northeastern part of central Stockholm,
down ‘T’ in order to increase its probe area and is undergoing residential development. Highly con
hence the penetration resistance. taminated soils are also planned to be capped, i.e.
The probe area is normally enlarged 10 times a fill layer is placed on the seabed to minimize dis
compared to conventional CPT tests. As the full- persion of contamination. The soil consists of
flow probe is mounted on conventional CPT probe, extremely soft organic clay and gyttja, and hence the
only replacing the tip, the probe area is 10 times filling is challenging from a stability persective. The
larger than the drilling rod. This gives several advan full-flow CPT tests were performed to obtain high-
tages in extremely soft soils: quality detailed strength properties, especially in the
‒ The penetration resistance is 10 times higher, i.e. upper part of the soil. In addition, the area had local
the meaurement uncertianties originating from the variation due to earlier unknown works such as
load cell decrease dregding, and hence mapping of the nearshore area
DOI: 10.1201/9781003308829-63
452
was important. The full-flow CPT tests was shown
to be an excellent method to quickly map the area,
as a complement to sampling and laboratory testing.
2 BACKGROUND
2.1 Previous work on full-flow CPT tests

The full-flow CPT tests have been used during the
last 20 years for field applications (e.g., Yafrate
et al., 2009; Peuchen & Terwindt, 2016; Nakamura
et al., 2009; Boylan et al., 2011; Schaeffers & Wee
mees, 2012), as well as in laboratory tests including
centrifuge tests (Almeida et al., 2011; Levacher
et al., 2016; Sahdi et al., 2014). In addition, the full-
flow test has been analysed theoretically by e.g. Ran
dolph & Andersen (2006) and Zhu et al. (2020).
In Sweden, a research project on sensitive clays
on land was performed around 10 years ago (Larsson
Figure 1. The Iskymeter (top left), column penetration test
et al., 2014; Åhnberg & Larsson, 2012). The purpose (top right) and T-bar used in this study (bottom). The T-bar
was to investigate cyclic strength degradation of was Ø40 mm and width 250 mm (tip area 10,000 mm2).
highly sensitive and quick clays. The drilling rod is Ø36 mm (area ~1,000 mm2) Sources:
Randolph (2004), DeJong et al. (2010) and Lunne Kallstenius (1961), Massarsch (2014), Geotech (2019).
et al. (2011) provide excellent all-inclusive descrip
tion of the method. et al., 2010). It is found to be dependent on soil type,
sensitivity and rate of strain softening.
2.2 Correction of tip resistance According to theoretical studies by Randolph &
Andersen (2006) the NTbar is in the range 11-13, not
Similar as for the conventional CPT test, the meas noticeably dependent on strength anisotropy if the
ured tip resistance is corrected by overburden stress average strength is used. For cyclic tests, i.e. several
and measured pore pressure. This corrected tip cycles of penetration and extraction over an interval,
resistance (qnet) is calculated by Eq. 1 (DeJong et al., a remoulded strength can be interpreted with a NTbar
2010): varying between 10.5-15.
Larsson et al. (2014) studied 13 different Swedish
soils and found that NTbar was highly dependent on
the soil’s plasticity, in particular its liquid limit.
A range between ~7 for low-plastic soils to ~16 for
high-plastic soils were found.
where qc = measured tip resistance; p0 = total vertical
overburden stress; u2 = measured pore pressure, a = 2.4 Comparison with similar tests
area ratio, As = area of rod; and Ap = tip area. This
equation assumes that the soil ‘flows’ around the Although the shape of the T-bar probe is unique,
probe and applies a vertical stress on the upper sur there are two other penetration tests which are
face. As the ratio As/Ap ≈ 0.1, the effect of overbur remarkably similar. The ‘Iskymeter’ was developed
den stress and pore pressure is one tenth of that for by the Swedish Geotechnical Institute during the
conventional CPT tests. It is thus often negligible. 1930s for soft clays. The Iskymeter consists of two
foldable wings which are folded during penetration
and unfolded when extracted where the extraction
2.3 Evaluation of shear strength resistance is measured. Calibration of shear strength
was done by comparing extraction resistance, fall
The undrained shear strength (cu ) is calculated as: cone test and in situ vane tests, and a N-factor of
around 10-15 was found, dependent on both sensitiv
ity and organic content (Kallstenius, 1961).
The Iskymeter is no longer in use but was sup
posedly the origin for the column penetration tests
which is used as a quality control of dry deep mixing
Usually, site specific calibrations are recom columns. For this application, a N-factor of 10 is
mended, however, the NTbar factor normally is found used (Axelsson, 2001), but this has undergone sur
to be around 10-13 (Lunne et al., 2011; DeJong prisingly little research.
453
Photographs of the Iskymeter, the column pene
tration and T-bar probe are shown in Figure 1. Not
ably, the Iskymeter, the standard dimension column
penetration test and the full-flow CPT have equal
probe areas, i.e. 10,000 cm2. The N-factor for inter
pretation of shear strength is also similar.
2.5 Extraction ratio

The extraction ratio for full-flow CPT tests, i.e. ratio
between penetration (qin ) and extraction (qout ) resist
ance is often used to interpret the soil’s sensitivity.
For low sensitive soils, the ratio is typically 0.6-0.8
decreasing to around 0.3 for highly sensitive soils
(DeJong et al., 2010).
3 SITE AND METHODS
3.1 Soil profile

Figure 2 shows a typical soil profile in one of the
boreholes with a water depth of ~18 m. The sedi
ments consist of clayey gyttja from seabed down to
appr. 10 m depth. From around 10 m the soil is cat
egorised as clay. As the organic content and water
content decreases with depth, the density increases.
The plastic limit is normally not determined in
Swedish engineering practice. Instead, only the
liquid limit is used for empirical correlations,
including silt and organic content. Figure 3 shows
values of liquid limits vs. organic content from tests
on the soil profile shown in Figure 1. This correl
ation allows a simple mapping of soil type only by
liquid limit values. Figure 2. Typical soil profile (water depth ~18 m). cl =
clay, gy = gyttja, sh = shales, v = varved, si = silt. Soil clas
sification according to EN ISO 14688-1 and -2.
3.2 Execution of the field work
The nearshore area had limited water depths, and all
tests and sampling were done using a pontoon with
supporting legs as a stable working platform. A total
of around 30 T-bar tests were performed in an area
of around 400x500 m. However, only a few is pre
sented herein due to limited space. Around half was
done with data acquisition also during extraction,
and a few was performed with cyclic tests.
All T-bar tests were performed with the drilling
rod within a casing to prevent excess deflection of the
rod. Divers attached the T-bar probe under the casing,
and also noted the time of penetration into seabed.
Sampling was done using the standard Swedish
piston sampler (50 mm and 60 mm diameter).
3.3 Laboratory tests Figure 3. Liquid limit vs. organic content on samples from
borehole shown in Figure 1.
A total of around 40 samples were retrieved. Routine
analyses, i.e. bulk density, natural water content,
liquid limit and intact and remoulded shear strength On selected samples, CRS oedometer, direct
with fall cone (FC) tests were performed on all simple shear (DSS) and triaxial compression and
samples. extension tests were performed.
454
For normal ranges of liquid limits for inorganic
clays, i.e. liquid limits around 40-80%, the NTbar varies
between ~10 and ~13. These values are similar to
those found by Larsson et al. (2014), Nakamura et al.
(2009), Randolph & Andersen (2006) and others.
For higher values of liquid limits, i.e. organic
clays and gyttja, the NTbar is up to ~20 according to
Equation 3. This is in the same range as reported for
tests in peat (e.g., Long & Boylan, 2012; Boylan
et al., 2011).
4.2 Evaluation of sensitivty

Figure 4. Liquid limit vs. sensitivity on all samples. The extraction resistance was measured for several of
the T-bar tests. An example is given in Figure 6 where
both the extraction (qout ) and penetration (qin ) resist
4 RESULTS ance is shown. The qout is consistenly lower than qin ,
as expected, due to the remoulding occuring around
4.1 Evaluation of undrained shear strength the probe.
Figure 7 plots all extraction ratios (qout =qin ) vs. sen
The strength anisotropy is larger for silty clay than sitivity values from laboratory FC tests where T-bar
organic clay and gyttja. It was thus decided to cali tests and sampling were done in the same location.
brate the NTbar against the undrained direct simple The ratios varies between 0.45 and 0.9, decreasing
shear strength (cDSS
u ). This eliminates any anisotropy with increasing sensitivity. This is in the same range as
effects in the penetration tests as it is done in reported by e.g. DeJong et al. (2010) for low sensitive
a stratigraphy of both organic and silty clay soils.
(example in Figure 2). Interpreting the cDSS
u is also There is a clear correlation between the two vari
normal practice for conventional CPT tests in ables, despite the relative small variation in sensitiv
Sweden. ity. This relationship is also seen in e.g. Yafrate et al.
Trials of correlating qnet with strengths from FC (2009). The correlation is however OCR-dependent
and DSS tests were done using several soil param (the OCR in the OC area is around 1.2–3).
eters, however, the NTbar was found to be best
expressed as a function of the liquid limit. Notably,
the liquid limit reflects the type of soil, its organic 4.3 Evaluation of remoulded shear strength
content and sensitivity (Figure 3 and 4). The follow Cycles were performed in an attempt to correlate the
ing expression was evaluated: penetration resistance with the remoulded strength
measured in the laboratory (using FC tests). Cycles
over 1 m intervals are shown in Figure 6, and
a detailed resistance plot of the upper cycle is shown
wL is here given in decimal form. A typical example in Figure 8. This cycle was perfomed in clayey
of a strength profile with depth is shown in Figure 5. gyttja with a water content of 150–200%. A clear
Here, Equations 1–3 gives strength values similar to decrease in both penetration and extraction resist
FC and DSS tests. The figure also shows cDSS calcu ance is seen for each cycle.
u
lated by empirical correlations with preconsolidation Figure 9 plots the average penetration resistance
stress from CRS oedometer tests and cDSS (Hov (qin) for each cycle vs. number of cycles. It seems
u
et al., 2021). Figure 5 also shows results from tri that after 5-7 cycles, the decrease in resistance seems
axial tests at shallow depths, although these were not to level off. This is similar to the findings from e.g.
used for intepretation of the T-bar tests. Yafrate et al. (2009).
As seen in the figure, the strengths are The measured remoulded undrained shear
extremely low just below the seabed, in practice strength in the laboratory was 1.1 kPa. The N-factor
zero at seabed, but increasing with around 1.2 kPa/ in the clayey gyttja was thus evaluated to:
m. By correlating the T-bar tests with laboratory
data on greater depths, it is thought that a strength
extrapolation towards the seabed is realistic. The
T-bar tests seemed to confirm this. Obviously,
strengths evaluations from full-flow tests are more It should be noted that cycles were performed in
certain than conventional CPT test due to the a very limited number of locations, so the dependency
larger tip resistance, thus reducing the measure of plasticity, sensitivity or OCR has not been
ment uncertainties. evaluated.
455
Figure 5. Typical example of evaluated strength by Eqs.
1-3 plotted together with results from laboratory tests. Figure 6. Typical example of penetration and extraction
Strength “Emp.corr” is based on empirical correlations resistance in one T-bar test. Two sets of cycles were per
with the preconsolidation stress σ0c from CRS oedometer formed during penetration.
tests; cDSS
u ≈ ð0:125 þ 0:205 x wL Þ x σ0c for OCR=1,0-1,3
(Hov et al., 2021).
5 DISCUSSION
The T-bar tests were shown to be very useful for

this type of project. The two main benefits were: a)
detailed analyses of undrained shear strength in the
upper part of the deposit, i.e. for strengths which in
practice increases from zero, and b) a rough map
ping of soil conditions by analysing the extraction
ratio.
The very low strengths are difficult to measure
with conventional CPT probes due to the large
measurement uncertainty. In addition, sampling of
such low strength sediments is very difficult, and in Figure 7. Values of sensitivity from laboratory tests (fall
practice impossible with the Swedish piston sam cone) and evaluation extraction ratio (example given in
pler as it is dependent on soil resistance when Figure 6).
coring (the coring is done by rotation of the drilling
rods, but the outer part of the sampler is kept still the planned capping of the contaminated areas and
using the soil resistance). filling for land reclamation.
The use of T-bar tests showed that the strength The rough mapping of soil conditions was pos
increases almost linearly from zero at seabed. This sible using the correlations between extraction ratio,
was valuable information for stability evaluation of sensitivity, liquid limit and organic content.
456
The sediments consisted of organic clay and gyttja.
The following conclusions are drawn:
‒ The T-bar test is a useful testing equipment for
soils with extremely low shear strengths,
‒ The N-factor was evaluated to vary between ~10
and ~20,
‒ The N-factor increases with increasing plasticity
(i.e. liquid limit),
‒ A good correlation was obtained between the
extraction ratio and sensitivity,
‒ A N-factor of around 35 was found for cyclic tests
to evaluate the remoulded shear strength.
REFERENCES
Almeida, M.S., Oliviera, J.R., Motta, H.P., Almeida, M.C.
& Borges, R.G. (2011). CPT and T-bar penetrometers
for site investigation in centrifuge tests. Soils and Rocks,
34(1), pp. 79–88.
Axelsson, M. (2001). Deep stabilisation with lime-cement
columns – methods for quality control [In Swedish].
Swedish Deep Stabilization Research Centre, Report 8.
Boylan, N., Long., M. & Mathijssen, F.A.J.M. (2011). In
situ strength characteristics of peat and organic soil
using full-flow penetrometers. Can. Geotech. J. 48,
1085–1099.
DeJong, J., Yafrate, N., DeGroot, D, Low, H.E. &
Randolph, M. (2010). Recommended practice for
Figure 8. Example of cyclic test (upper cycle shown in full-flow penetrometer testing and analysis. Geotech
Figure 6). Shear strength evaluated using N = 35. Negative nical Testing Journal, Vol. 33, No. 2.
strength is shown for extraction to make the figure more EN ISO 14688-1 and -2. Geotechnical investigation and
readable. testing Part I and II. International Standard.
Geotech (2019). CPT Geotech Nova – T-bar full-flow
sounding equipment.
Hov, S., Prästings, A., Persson, E., Larsson, S. (2021). On
empirical correlations for normalised shear strengths
from fall cone and direct simple shear tests in soft
Swedish clays. Geotech Geol Eng 39, 4843–4854.
Kallstenius, T. (1961). Development of two modern con
tinuous sounding methods. Proc. 5th Int. Conf. on Soil
Mech. and Found. Eng., Vol. 1, pp. 475–480.
Larsson, R., Åhnberg, H. & Schälin, D. (2014). T-bar tests
in soft and sensitive Swedish clays. Proc. 3rd Int. Symp.
on CPT, Las Vegas, pp. 449–458.
Levacher, D., Razakamanatsoa, A., Gupta, R. &
Katsumi, T. (2016). Comparative su measurements with
vane shear and T-bar testing on soft soils in laboratory.
15th Asian Reg. Conf. on Soil Mech. and Geoth. Eng.,
JGS, pp. 465–468.
Lunne, T., Andersen, K., Low, H.E., Randolph, M. &
Sjursen, M. (2011). Guidelines for offshore in situ test
ing and interpretation in deepwater soft clays. Can. Geo-
Figure 9. Measured strength using N=35 and remoulded tech. J. 48, 543–556.
strength from FC tests for cycle shown in Figures 7 and 8. Long, M. & Boylan, N., (2012). In-situ testing of peat –
a review and update on recent developments. Geotech.
Eng. J. of SEAGS & AGSSEA, Vol. 43, No. 4, pp. 41–55.
Massarsch, R. (2014). Cone penetration testing – A historic
perspective. Proc. 3rd Int. Symp. on CPT, Las Vegas, pp.
97–134.
6 CONCLUSIONS
Nakamura, A., Tanaka, H. & Fukasawa, T. (2009). Applic
ability of T-bar and ball penetration tests to soft clayey
This paper presents a nearshore case study where grounds. Soils and Foundation, Vol. 49, No. 5,
a large number of T-bar tests have been performed. 729–738.
457
Peuchen, J. & Terwindt, J. (2016). Critical appraisal of Vancouver Geotechnical Society Symposium, Soft
T-bar penetration tests. Geot. and Geoph. Site Charac Ground Engineering, Vancouver (BC).
terisation, pp. 351–356. Stewart, D.P. & Randolph, M.F 1994. T-bar penetration
Randolph, M. & Andersen, K. (2006). Numerical analysis of testing in soft clay. J. Geot. Eng. Div., ASCE, 120(12),
T-bar penetration in soft clay. ASCE Int. J. of Geom., 6(6), 2230–2235.
411–420. Yafrate, N., DeJong, J., DeGroot, D. & Randolph, M.
Randolph, M. (2004). Characterisation of soft sediments (2009). Evaluation of remoulded shear strength and sen
for offshore applications. Keynote lecture. Proc. 2nd sitivity of soft clay using full-flow penetrometers. ASCE
Int. Conf. Site Characterisation, Porto, Vol. 1, pp. J. Geot. Geoenv. Eng., Vol. 135, No. 10.
209–231. Zhu, B., Dai, J. & Kong, D. (2020). Assess the effects of
Sahdi, F., Gaudin, C., White, D. & Boylan, N. (2014). loading rate and interface roughness on T-bar penetra
Interpreting T-bar tests in ultra-soft clay. Int. J. of Phys. tion resistance. Soils and Foundations, 60, pp. 266–273.
Mod. In Geot., 14(1), 13–19. Åhnberg, H. & Larsson, R. (2012). Degradation of undrained
Schaeffers, J. & Weemees, I. (2012). Comparison of in-situ shear strength due to dynamic actions and large strains.
shear strength measurement techniques of soft clay. 20th Swedish Geotechnical Institute, Report 75, Linköping.
458
© 2022 the Author(s), ISBN 978-1-032-31259-0
Bayesian supervised learning of 2D subsurface soil stratigraphy using

limited cone penetration tests with consideration of uncertainty
Yue Hu
Yu Wang
Department of Architecture and Civil Engineering, City University of Hong Kong, Hong Kong, China
ABSTRACT: Cone penetration tests (CPT) have been widely used for soil stratification in geotechnical site
investigation for decades. However, due to time and budget limits, the layout of CPT sounding at a specific
project site is often sparse, leading to significant interpolation uncertainty in the development of subsurface
soil 2D cross-section, particularly at locations without CPT measurements. Such development is often com
bined with empirical classification criteria, which further introduce model uncertainty to soil stratification.
These uncertainties may pose great risks to the geotechnical engineering practice. A Bayesian supervised learn
ing method is presented in this paper for probabilistic soil stratification in a 2D cross-section using limited
CPT. The proposed method can not only automatically stratify soils in a 2D cross-section from limited CPT
soundings, but also can properly quantify the associated uncertainties. Complete 2D CPT data cross-section is
firstly learned from limited number of 1D CPT profiles using Bayesian supervised learning. The associated
interpolation uncertainty is modelled numerically using non-parametric random field simulation based on the
results of Bayesian supervised learning. Parametric autocorrelation function of CPT data along either vertical
or horizontal direction is not needed. A probabilistic model is also developed to account for the model uncer
tainty of an empirical soil behavior type classification chart. The interpolation uncertainty and soil classifica
tion model uncertainty are then evaluated simultaneously in a Monte Carlo simulation framework. A simulated
data example is used for illustration. The results suggest that the proposed method performs well.
1 INTRODUCTION Wang et al. 2013). When interpreting 2D vertical

soil cross-section using CPT, interpolation between
Subsurface soil stratification is an indispensable adjacent CPT profiles or stratigraphy should be per
element in geotechnical engineering as required for formed. However, direct 2D interpolation using con
geotechnical designs and analyses. However, due to ventional geostatistical methods is difficult because
time and budget limits, the layout of site investiga of the limitation of CPT soundings number and the
tion points is often sparse, leading to significant chal spatially varying soil layer boundaries which render
lenge and uncertainty in the development of 2D the soil properties in the 2D cross-section highly
subsurface soil stratigraphy which is frequently non-stationary (e.g., Wang et al. 2019). In this case,
adopted in geotechnical analysis. In engineering prac it is even more challenging to reasonably tackle the
tice, such development is often combined with an interpolation uncertainty quantification associated
empirical soil classification system, which inevitably with the 2D subsurface stratigraphy. On the other
introduce additional model uncertainty to soil stratifi hand, CPT-based soil classification relies on empir
cation. These uncertainties may pose great risks to ical soil classification charts to transform the con
the geotechnical designs, analyses, and construction tinuous CPT measurements to discrete soil behavior
process (e.g., Clayton 2001; Mayne 2007). It is there type (SBT). Note that these charts rely on a series of
fore necessary to properly evaluate the uncertainties deterministic SBT classification boundaries which
associated with 2D subsurface soil stratification. are developed from site investigation data globally.
Cone penetration tests (CPT) have been widely These classification charts might not provide consist
used for soil stratification in geotechnical site inves ent results at a specific site and introduce model
tigation. It can be used to classify subsurface soils, uncertainty when used locally (e.g., Boulanger &
identify stratification, and quantify associated uncer Idriss 2014; Maurer et al. 2019). How to incorporate
tainty in a single sounding (e.g., Robertson 1990; the model uncertainty of SBT classification
DOI: 10.1201/9781003308829-64
459
boundaries in the CPT-based 2D soil stratification except for a limited number of non-trivial ones.
remains unsolved. Therefore, the F may be reconstructed approxi
To address the abovementioned challenges, mately if those non-trivial weight coefficients can be
a Bayesian supervised learning method is proposed to identified and estimated using sparse measurements
interpret 2D subsurface soil stratigraphy from limited Y (e.g., CPT profiles data at limited locations),
CPT with explicit evaluation of both interpolation which is a sub-matrix of F with a dimension of
uncertainty and SBT chart model uncertainty. The pro M x1 x M x2 (M x1 55N x1 , M x2 55N x2 ). The relation
posed framework is introduced in the following sec between Y and !2D t is expressed as:
tion and then illustrated using a simulated example.
2 THE PROPOSED METHOD
The proposed framework is based on a Monte Carlo

simulation (MCS) and comprises of three compo in which A2D t is the sub-matrix of B2D t with
nents. First, SBT index Ic of CPT data is firstly inter a dimension of M x1 x M x2 . A2D t just reflects the
polated in the concerned 2D cross-section from measured elements in B2D t . Equation 2 enables those
limited CPT using a Bayesian supervised learning non-trivial coefficients to be learned through max
algorithm in a non-parametric manner. The interpol imum likelihood estimation under a Bayesian frame
ation uncertainty is quantified automatically during work (e.g., Tipping 2001). The learned weight
learning process and then modelled by Ic random coefficient vector is denoted as ω ^ 2D . After deriv
field simulation based on the learning outcome. ation, it is found that the posterior distribution of
Second, a probabilistic SBT chart is developed by ^ 2D given measurement data follows a multivariate
ω
adapting an empirical chart to consider the model Student’s t distribution, with the mean vector and
uncertainty using which random samples of SBT covariance matrix expressed as (e.g., Zhao et al.
chart with random classification boundaries are 2018; Hu et al. 2020):
drawn. Third, each random field sample of Ic is
matched with a random sample of SBT chart to pro
duce a Monte Carlo sample of SBT cross-section
with soil stratigraphy. Statistical analysis is then per
formed on the generated Monte Carlo samples. Both
uncertainties are evaluated simultaneously under
MCS. The three components are introduced in the
following three subsections, respectively. in which J is a matrix with element
2D T
Jt;s ¼ tr½A2Dt ðAs Þ ], (t, s = 1, 2, …, N x1 x N x2 ). “tr”
2.1 Bayesian supervised learning with random represents trace operation in linear algebra. D is
field simulation a diagonal matrix with diagonal elements
Dt,t = αt (t = 1, 2, …, N x1 x N x2 ) in which αt are non
The full Ic data variability in the concerned 2D cross- negative parameters to be determined by a maximum
section are learned from the profiles of limited CPT. likelihood algorithm (e.g., Tipping, 2001). Vtr is
In the context of the Bayesian supervised learning, a N x1 x N x2 vector with element
2D data matrix F (e.g., Ic data matrix of a 2D cross- V tr ðtÞ ¼ tr½YðA2D Þ T
]. c = M x M /2+c; dn = d
t n x1 x2
section), which is spatially varying along coordinates +ðkYk22 -μω^ 2D T H-1 μω^ 2D Þ=2. c and d are non-negative
x1 and x2 (e.g., depth direction and horizontal direc small constants to achieve an uninformative prior in
tion), has a dimension of Nx1 x Nx2 . Mathematically, the Bayesian formulation. Due to the compressibility
F is expressed as a weighted summation of a series of of F, only those Na (Na55N x1 x N x2 ) non-trivial
orthonormal 2D basis functions (e.g., Zhao et al. coefficients need to be estimated, and the rest are
2018; Hu et al. 2020; Wang et al. 2020&2021): zeroed out. Therefore, in Equation 3, the μ!^2D is
reduced to an Na×1 vector and COV!^2D is reduced to
an Na × Na matrix for simplicity.
Given the learning results in Equation 3, random
samples of ! ^2D can be generated through eigen
decomposition (e.g., Hu et al. 2019; Hu and Wang
2020):
in which B2D
t is the t-th 2D basis function, while
2D
!2D
t is the weight coefficients of Bt . Discrete wave
let transform may be selected to construct B2Dt (e.g.,
Donoho et al. 2006). Note that for compressible
images (e.g., spatially correlated CPT data cross
section), most !2D t have negligibly small values
460
in which Ui is the i-th eigen-vector of the covari be mapped, leading to a 2D SBT cross-section.
ance matrixCOV!^2D ; λ!^2D is the i-th eigenvalue of After repeating the process NB times with different
i
COV!^2D ; Zi is a set of independently and identically combination of 2D Ic data RFS and SBT chart
distributed standard Gaussian random variables. random samples, NB 2D SBT cross-sections are
Using Equation 4, random vectors !^2D can be gener generated. Each SBT cross-sections serves as
ated readily through realizations of Zi . After that, a possible state of subsurface soil stratigraphy
extensive RFS of the approximated 2D data F ^ (e.g., given limited CPT. Statistical analysis is then per
Ic data 2D cross-section) are subsequently recon formed for those NB SBT cross-sections. The
structed by substituting the random vectors ! ^2D into probability of soil at a given point (x1, x2) being
Equation 1 as below: mapped to a specific SBT, e.g., SBT=t (t=2, 3,
···, 7), can be calculated as (e.g., Hu and Wang
2020):
Each generated 2D RFS indicates a possible out

come of Bayesian supervised learning. The ensemble
directly reflects the interpolation uncertainty. in which Ntx1 ;x2 is the number of SBT values
at point (x1, x2) that equal to t. Equation 6 quan
titatively measures how likely the soil at a point
2.2 Probabilistic SBT chart (x1, x2) is classified as any one of the six SBT.
The empirical Ic – based SBT chart summarized in The SBT with the highest probability is taken as
Table 1 was developed from a global soil database the most likely SBT at the point (x1, x2). Simi
compiling CPT data obtained predominately within larly, the most likely SBT cross-section is
limited depths (e.g., Robertson 1990; Robertson & obtained. In addition, the classification uncer
Wride 1998). It is expected that the SBT chart tainty is quantified by the SD of SBT samples at
might not provide accurate classification and intro a given point. The higher the SD, the higher the
duce model uncertainty when used locally at uncertainty. Similarly, an SD cross-section can
a specific project site. In other words, those empir be calculated which reflects the pattern of uncer
ical Ic classification boundaries (e.g., see Table 1) tain region in the 2D cross-section. The SD
can vary from site to site (e.g., Boulanger & Idriss cross-section directly assesses the reliability of
2014; Maurer et al. 2019). To consider the model soil stratigraphy.
uncertainty, a probabilistic Ic – based SBT classifi
cation chart is developed. The five SBT classifica
tion boundaries (i.e., 1.31, 2.05, 2.6, 2.95, 3.6) 3 ILLUSTRATIVE EXAMPLE
listed in Table 1 are denoted as B1 to B5 and mod
elled as five Gaussian random variables, respect To illustrate the proposed framework, a simulated
ively. The mean values of B1 to B5 are taken as geological cross-section example is provided in this
their original values, as summarized in the second section. As shown in the Figure 1a, a 2D vertical
column of Table 2. According to literature (e.g., cross-section is simulated. Four soil types exist in
Boulanger & Idriss 2014; Maurer et al. 2019) on Ic this cross-section, i.e., clay, silt mixtures, sand mix
data variability, a set of standard deviation (SD) tures and sand, which correspond to SBT values of 3
values for B1 to B5 is suggested in this study, as to 6, respectively. Spatial variability of Ic data in each
summarized in the third column of Table 2. SD of soil layer is generated using assumed random field
B1, B2 and B5 are taken as 0.1, while the SD of B3 model. The random field parameters used (e.g., mean
and B4 are taken as 0.05. The probabilistic SBT µIc, standard deviation σIc, correlation lengths along
chart allows for the varying nature of Ic – based horizontal direction λh and vertical direction λv) are
classification criteria at different local sites. Exten summarized in Table 3. An exponential correlation
sive random samples of classification boundaries structure is adopted in this simulation. The simulated
can be generated by repetitively sampling B1 to B5, 2D Ic data cross-section is shown by colormap in the
leading to random samples of SBT chart. Each Figure 1b. The Ic data cross-section is a 128×256
random sample is a possible state of the uncertain matrix with resolution of 0.1m for both directions. It
SBT classification chart and will randomly match is regarded as a geological setting at a specific site.
with one RFS of Ic data cross-section for classifica Note that the complete Ic data cross-section is usually
tion and stratification purposes. not available in engineering practice. This example is
just for illustration and validation purposes. Suppose
that six (i.e., M=6) CPT soundings are conducted
2.3 Statistical analysis of Monte Carlo simulation within the cross-section, as denoted by black dash
By pairing 2D Ic RFS with a random sample of lines (e.g., M1-M6) in Figure 1. Ic data profiles of
SBT chart, each point in that 2D cross-section can M1-M6 are shown in Figure 2. These six 1D Ic data
461
profiles are used as input to interpret the soil examples of SBT cross-sections are shown in
stratigraphy under the proposed framework. Figures 4a-4d. These four 2D SBT cross-sections
After constructing the Y matrix in Equation 2 correspond respectively to the four RFS of 2D Ic
from data profiles of M1-M6, the Bayesian super data shown in Figure 3 mapped with random sample
vised learning is implemented. Then NB=500 2D of SBT charts shown above each subplot of Figure 4.
RFSs of Ic data are generated from the learning Note that each of these SBT cross-section indicates
results. Four examples of Ic cross-section are shown a possible soil stratigraphy.
by colormap in Figure 3. Each plot in Figure 3 is Using NB=500 Monte Carlo samples of 2D SBT
a possible interpolation of Ic data in this 2D cross- cross-section, statistical analysis is performed using
section from six CPT soundings. No parametric cor Equation 6. The most likely SBT cross-section is
relation structure is needed for the learning process. shown in the Figure 5a. In the most likely SBT cross-
Next, NB=500 random SBT classification charts are section, four SBT (i.e., SBT3-6) are presented, which
generated from the probabilistic SBT classification is consistent with the underlying true stratigraphy.
boundary model. Note that a Gaussian random vari The original soil zone boundaries are shown by black
able ranges from negative infinity to positive infinity. solid lines for comparison. Note that the most likely
To mitigate the overlapping problem, the Gaussian SBT cross-section is generally comparable to the
probability density functions (PDF) for B1-B5 are underlying true one, although the soil zone boundaries
truncated respectively to a range of mean ± three are not perfectly learned due to interpolation uncer
standard deviation, as summarized in the fourth tainty and model uncertainty in SBT chart. The uncer
column of Table 2. tainties can be evaluated simultaneously through SD
By randomly pairing a RFS of Ic data cross- of NB=500 SBT cross-sections, as shown in
section with one random sample of SBT chart, Figure 5b. It is found that the bright areas with high
uncertainty are generally consistent with the under
lying true boundaries (i.e., the black solid lines). The
results in Figure 5b suggest that the underlying true
soil zone boundaries can be approximated based on
the SD of NB SBT cross-sections. The proposed
framework performs reasonably well in the interpret
ation of 2D soil stratigraphy and uncertainty quantifi
cation, given only six CPT soundings.
Table 1. Ic – based SBT classification chart (after Robert

son 1998).
Range of SBT SBT

index Ic ID SBT description
Ic <1.31 7 Gravelly sand to dense sand

1.31 < Ic < 2.05 6 Sands: clean sand to silty sand
Sand mixtures: silty sand to
2.05 < Ic < 2.60 5
Figure 1. Simulated geological cross-section (a) and Ic data sandy silt
(b) in a 2D cross-section. Silt mixtures: clayey silt to silty
2.60 < Ic < 2.95 4
clay
2.95 < Ic < 3.60 3 Clays: silty clay to clay
Ic >3.60 2 Organic soil: peats
Table 2. Probabilistic model of the Ic – based SBT classi

fication boundaries.
Boundaries Mean SD Range
B1 1.31 0.1 [1.01, 1.61]

Figure 2. Ic profiles of six CPT soundings. B2 2.05 0.1 [1.75, 2.35]
B3 2.6 0.05 [2.45, 2.75]
a Monte Carlo sample of 2D SBT cross-section is B4 2.95 0.05 [2.8, 3.1]
obtained. After repeating this procedure 500 times, B5 3.6 0.1 [3.3, 3.9]
NB=500 SBT cross-sections are obtained. Four
462
4 EFFECT OF SOUNDING NUMBER locations. The proposed method is data-driven and
features fine scale spatial variability to the interpol
To investigates the effect of CPT soundings number ation result when M increases.
M on the proposed framework, an additional scen
ario with M=25 CPT soundings is discussed. Those
25 CPT are performed with equal space in the 2D 5 CONCLUSIONS
cross-section of illustrative example (i.e., see
Figure 1). Following the same procedures as A novel Monte Carlo simulation (MCS) – based
described above, NB=500 Monte Carlo samples of framework was proposed in this paper for interpret
2D SBT cross-section are obtained. The most likely ing soil stratigraphy in a 2D cross-section from
SBT cross-section in this scenario is shown in
Figure 6a. The approximated soil stratigraphy in this
scenario becomes more accurate. The quantified
uncertainty also shrinks significantly, as shown in
the Figure 6b. Those thin bright area with high
uncertainty is in good agreement with boundary
Figure 4. SBT cross-sections for the four Ic samples in

Figure 3. Four examples of 2D Ic data RFS. Figure 3.
463
Bayesian supervised learning. The associated inter
polation uncertainty was modelled by non
parametric random field simulation based on the
learned results. A probabilistic soil behavior type
(SBT) chart was developed for incorporating the
model uncertainty in the empirical chart. The inter
polation uncertainty of Ic data and model uncertainty
in SBT chart were considered simultaneously under
MCS. Key equations and detailed implementation
procedures were provided. Numerical example was
illustrated and showed that the proposed method per
formed reasonably well. Sensitivity study suggested
that the proposed method was data-driven. As the
number of CPT soundings increased, the most likely
SBT cross-section became accurate and associated
interpolation uncertainty reduced significantly.
Figure 5. Statistics of generated SBT cross-sections when

ACKNOWLEDGEMENTS
M=6: (a) the most likely SBT cross-section; (b) standard
deviation of 500 SBT cross-sections. The work described in this paper was supported by
grants from the Research Grants Council of the
Hong Kong Special Administrative Region, China
(Project Nos. CityU 11213119 and CityU 11202121).
The financial supports are gratefully acknowledged.
REFERENCES
Boulanger, R., & Idriss, I. 2014. CPT and SPT based liquefac
tion triggering procedures. Report No. UCD/CGM.-14, 1.
Clayton, C. R., Matthews, M. C., & Simons, N. E. 1995.
Site investigation: a handbook for engineers. Blackwell
Science.
Donoho, D., Maleki, A. & Shahram, M. 2006. Wavelab
850. Software Toolkit for Time-Frequency Analysis.
Stanford University.
Hu, Y., Zhao, T., Wang, Y., Choi, C., & Ng, C. W. W. 2019.
Direct Simulation of Two Dimensional Isotropic or
Anisotropic Random Field from Sparse Measurement
using Bayesian Compressive Sampling. Stochastic
Environmental Research and Risk Assessment, 33(8-9),
1477–1496.
Hu, Y., Wang, Y., Zhao, T., & Phoon, K. K. 2020. Bayes
Figure 6. Statistics of generated SBT cross-sections when
ian supervised learning of site-specific geotechnical
M=25: (a) the most likely SBT cross-section; (b) standard
spatial variability from sparse measurements. ASCE
deviation of 500 SBT cross-sections.
ASME Journal of Risk and Uncertainty in Engineer
ing Systems, Part A: Civil Engineering, 6(2),
04020019.
Table 3. Random field parameters for the simulated 2D Ic Hu, Y., & Wang. Y. 2020. Probabilistic soil classification
cross-section. and stratification in a vertical cross-section from limited
cone penetration tests using random field and Monte
Layer μIc σIc λh (m) λv (m) Carlo simulation. Computers and Geotechnics, 124,
103634.
Layer1 3.2 0.1 15 1 Maurer, B., Green, R., van Ballegooy, S., &
Layer2 2.8 0.07 30 2 Wotherspoon, L. 2019. Development of region-specific
Layer3 2.3 0.1 20 1.5 soil behavior type index correlations for evaluating
Layer4 1.5 0.1 25 1.8 liquefaction hazard in Christchurch, New Zealand. Soil
Dynamics and Earthquake Engineering, 117, 96–105.
Mayne, P. W. 2007. Cone penetration testing. Transporta
tion Research Board.
limited cone penetration tests (CPT) with consider Robertson, P. 1990. Soil classification using the cone pene
ation of uncertainty. Complete CPT data cross- tration test. Canadian Geotechnical Journal, 27
section was learned from limited CPT soundings via (1),151–158.
464
Robertson, P., & Wride, C. 1998. Evaluating cyclic lique Wang, Y., Hu, Y., & Zhao, T. 2020. CPT-based subsurface
faction potential using the cone penetration test. Canad soil classification and zonation in a 2D vertical
ian Geotechnical Journal, 35(3),442–459. cross-section using Bayesian compressive sampling.
Tipping, M. E. 2001. Sparse Bayesian learning and the rele Canadian Geotechnical Journal, 57(7),947–958.
vance vector machine. Journal of machine learning Wang, Y., Hu, Y. & Phoon, K.K., 2021. Non-parametric
research, 1(Jun), 211–244. modelling and simulation of spatiotemporally varying
Wang, Y., Huang, K., & Cao, Z. 2013. Probabilistic identi geo-data. Georisk: Assessment and Management of Risk
fication of underground soil stratification using cone for Engineered Systems and Geohazards (published
penetration tests. Canadian Geotechnical Journal, 50 online), https://doi.org/10.1080/17499518.2021.19
(7),766–776. 71258.
Wang, Y., Zhao, T., Hu, Y., & Phoon, K. K. 2019. Simula Zhao, T., Hu, Y., & Wang, Y. 2018. Statistical interpretation
tion of random fields with trend from sparse measure of spatially varying 2D geo-data from sparse measure
ments without detrending. Journal of Engineering ments using Bayesian compressive sampling. Engineer
Mechanics, ASCE, 145(2), 04018130. ing Geology, 246, 162–175.
465
© 2022 the Author(s), ISBN 978-1-032-31259-0
Estimation of constrained modulus from CPT measurements in case of

Holocene sands
Zsombor Illés, István Kádár, Gábor Nagy, András Mahler & László Nagy
Budapest University of Technology and Economics, Budapest, Hungary
ABSTRACT: In case of major projects there are numerous measurements which provide a sufficient set of
data to determine characteristic values based on statistical methods. In the current paper Cone Penetration Test
(CPT), oedometer and Flat Dilatometer Test (DMT) results are analyzed. Constrained or oedometric modulus
is one of the most important deformation parameters, which has a key role in settlement estimation. There are
various recommendations to derive Eoed from CPT tip resistance values. Three of these recommendations are
compared with the results of the oedometer tests and the correlation used in case of DMT measurements.
Keywords: Constrained modulus, CPT, Tip-resistance, DMT, Normally-consolidated layers
1 INTRODUCTION construction waste as well. Its thickness varies

between 3.5-5.5 m.
1.1 Measurements and methods Alluvial deposits are generated during periods of
high-water level when a predominantly suspended
Several Cone Penetration Tests were conducted at
fine matter settles to the surface of the floodplain. It is
a site in South-Central Hungary near the Danube.
composed of lean clay mixed with firm-stiff organic
The probes penetrated until the depth of 20 - 40
matter, and its thickness is approximately 0.7 m.
meters. Altogether 32 CPTs were analyzed, penetrat
As described by Kádár and Nagy (2018), below
ing at least to the depth of 25 – 26 m, until the
the alluvial deposit, there is eolian sand (transported
bottom of Holocene layers; clay, sand, and gravel.
by wind). The rounded particles are in the size range
Below that, over-consolidated late Miocene
of silt and fine sands; the layer has a thickness of
(Pannon) layers were found.
5.0 m. It is followed by fluvial sand containing fine,
Different methods are evaluated for estimating
medium, coarse and gravelly sands over a thickness
constrained modulus from cone penetration resist
of 5.3 m. It abruptly turns into gravelly sands and
ance of the CPT measurements in the article. These
gravel; it forms a 10.0 m thick gravel terrace. Both
three methods are the following: Sanglerat (1972),
the in-situ measurements, CPT and DMT, and the
Lunne and Christoffersen (1983) and Eslaamizaad
oedometric compressions, are evaluated up to the
and Robertson (1996).
base of the gravelly sand layer. In this study, the
Furthermore, laboratory soil identification and
eolian and fluvial sands are treated as one layer.
oedometric tests were also carried out. In addition,
there is a possibility to compare Cone Penetration
and Flat Dilatometer Test (DMT) results.
2 SOIL CLASSIFICATION AND
LABORATORY MEASUREMENTS
1.2 Geology and stratigraphy of the site
The project site is located in South-Central Hun 2.1 Evaluation of CPT measurements
gary. It used to be the flood plain of the river The CPT measurements were conducted according
Danube until the construction of a dike system. to MSZ EN ISO 22476-1:2013.
During the previous phases of the industrial com The CPT-based soil behavior of the layers; land
pound’s construction, the extracted soil was fill, lean clay, silty sand and gravelly sand are pre
placed on the current site as a backfill. The fill sented in Figure 1. A CPT-based normalized soil
material consists of fine sands, less often silty behavior chart was suggested by Robertson (1990)
sands, mostly grey-yellow silty clay and lean and updated by Robertson (2009). The parameters
clay, and medium sands. The landfill includes used are summarized in Robertson (2016).
DOI: 10.1201/9781003308829-65
466
The stratification of the area was defined accord Table 2. CPT parameters of the lean clay.
ing to traditional geotechnical soil identifications and
CPT measurements. The layers appearing at the CPT Parameter qc [MPa] fs [MPa] Rf [%]
logs are presented in Figure 1. One dot is one layer
in one CPT. According to normalized cone resistance No. of samples 26 26 26
(Qtn) and normalized friction ratio (Fr), the top layer, Average 2.69 53.52 2.65
landfill, is classified as silty sand and sand. The lean Min 0.29 1.60 0.90
clay is regarded as clay, silty clay and sandy silt. Max 7.73 89.53 6.50
This layer is not present in all the CPTs. The layers Std. 1.37 16.22 1.30
of eolian and fluvial sands (Kádár and Nagy, 2018) Co.V. 0.51 0.30 0.49
are classified as silty sand and sand, while the last
normally consolidated layer is sand. The four layers
are not overconsolidated as they more or less fall in
the diagonal of the diagram. Table 3. CPT parameters of the silty sand.
Parameter qc [MPa] fs [MPa] Rf [%]
No. of samples 32 32 32
Average 17.09 170.28 1.02
Min 8.77 74.22 0.52
Max 24.79 309.02 1.83
Std. 3.47 71.21 0.38
Co.V. 0.20 0.42 0.37
Table 4. CPT parameters of the gravelly sand.
No. of samples 33 33 33
Average 23.64 126.19 0.67
Min 16.76 68.75 0.27
Max 30.86 234.51 2.99
Std. 3.23 34.66 0.49
Co.V. 0.14 0.27 0.73
Figure 1. Classification of the layers according to Robert

son (2009). The landfill and the lean clay CPT parameters have
a higher coefficient of variation than the sandy layers.
Each layers’ parameters obtained by CPT are 2.2 Evaluation of DMT measurements
summarized in Tables 1-4, tip resistance (qc), skin
On the test site 11 DMT soundings penetrated to dif
friction (fs) and friction ratio (Rf). To each of the
ferent depth. The flat dilatometer was developed in
three parameters, minimum, maximum, average,
the 1980s by Silvano Marchetti. Shear strength
standard deviation and coefficient of variation are
parameters of the soils such as undrained shear
defined.
strength (cu). and friction angle (’). can be derived.
One of the most helpful information that DMT meas
urements can derive is related to the soil layers’
Table 1. CPT parameters of the landfill. stress history. Overconsolidation ratio (OCR), and
coefficient of lateral earth pressure (K0) can be deter
mined in the case of sands, in which sampling would
No. of samples 33 33 33 be difficult (Marchetti et al. 2001).
Average 10.08 102.30 1.38 In this paper. constrained modulus is estimated by
Min 2.11 36.58 0.28 DMT measurements, which is the key parameter for
Max 22.95 252.76 3.15 settlement calculation.
Std. 5.55 54.03 0.66 Two corrected readings (p0 and p1) are obtained
Co.V. 0.55 0.53 0.48 by the flat dilatometer, from these, material index
(ID), horizontal stress index (KD), and dilatometer
467
modulus (ED) are calculated. Vertical drained con The results of the oedometric tests and the depth
strained modulus MDMT can be determined accord from where the samples were extracted are presented
ing to Eq. (1) (Marchetti, 1980) as: in Figure 2. the modulus is related to the stress state
as well as the depth.
In the case of the silty sand layer, the following
correlation can be derived between the depth and
oedometric modulus:
where RM is a correction primarily depending on
the stress history (KD).
2.3 Oedometer tests

While in case of the gravelly sand layer, a different
To test the stress-strain relationship and the consoli correlation can be made:
dation parameters of each layer oedometer tests were
carried out according to MSZE CEN ISO/TS 17892
5:2010 standard. Undisturbed samples were tested
with 0-100-200-400-600 kPa, it was a usual practice The linear correlations (Eq. 2.-3.) describe the con
to load, unload and reload the samples; a common nection between depth and oedometric modulus well.
load path was 0-2-23-45-100-2-100-200-400 kPa.
A linear assumption was used between each load
step according to Eq. (2). 3 METHODS USED TO DERIVE THE
CONSTRAINED MODULUS
3.1 Sanglerat method

The Cone Penetration Testing (CPT) sounding was
developed in the Netherlands in the 1930s. It was
used to investigate the layers’ penetration resistance
for pile foundation design. A key parameter for
settlement calculation is the restrained modulus (M).
Buisman (1940) proposed the following correlation
between the tip resistance and oedometric modulus
in case of cohesionless soils (sands):
1.5 multiplier in Eq. (5) was replaced to αm by San

glerat (1972):
The values for the coefficient in the case of cohesive

soils were also defined and presented by Sanglerat
(1972), later synthesized by Kumala Sari et al. (2017).
According to Figure 1. the landfill is classified as “silts
of low plasticity”, although it is very heterogenic, and
a high percentage of construction waste can be found
in it. According to Figure 1. and laboratory identifica
tion tests, the lean clay layer is considered a “clay of
low plasticity”, while the last two layers of silty sand
gravelly sand are regarded as “sands”.
3.2 Lunne and Christophersen method

Most correlations between CPT results and the
drained constrained modulus (M) refer to the tangent
modulus, as found from oedometer tests. The refer
Figure 2. Different layers oedometric modulus and the ence value of M is typically based on the effective
depth of the samples. vertical stress σ0v0 (Lunne et al. 1997).
468
Lunne and Christoffersen (1983) reviewed the avail- The kM functions of different OCRs’ are esti
able (at that time) calibration chamber test results and mated by digitalizing the figure in Lunne et al.
made the following recommendations to estimate M in (1997). The estimated functions are the following:
case of normally consolidated uncemented silica sands:
They also included overconsolidated sands in their

studies and made rough estimates for them as well.
3.3 Eslaamizaad and Robertson method

Based on the assessment of extensive calibration tests
on quartz sand (Baldi et al. 1986; Fioravante et al.
1991). Eslaamizaad and Robertson (1996) proposed
an alternative method to estimate M from CPT data.
The method presents a correlation incorporating
normalized cone resistance and normalized vertical
effective stress in the form of:
The method has the advantage that prior knowledge

of relative density is not required (Lunne et al.
where:
1997). When Eslaamizaad and Robertson (1996)
n: stress exponent equal to 0.200 for normally con
compared the estimated values against the ones
solidated sands. and 0.128 for overconsolidated sands.
measured in the calibration chamber, they were
pa : atmospheric pressure. in the same units as
between 75% and 125%. Later the M values are pro
M and qc .
vided for each layer in Figure 5; the ones estimated
kM : is a dimensionless modulus number which
according to Eslaamizaad and Robertson (1996) are
can be determined using Figure 3. based on normal
divided by 1.25.
ized cone penetration resistance and esti
mated overconsolidation ratio (OCR).
4 EVALUATION OF THE RESULTS
4.1 Vertical soil profile

The 32 CPTs average. minimum and maximum tip
resistance (qc) are presented in a vertical profile
(Figure 4). The average cone resistance (qc) was
used for the calculation of restrained modulus (M)
by Sanglerat (1972), Lunne and Christoffersen
(1983) and Eslaamizaad and Robertson (1996)
methods.
There are lower and upper limits for cohesive
soils (landfill and lean clay). The eodometric moduli
of each layer measured on undisturbed samples in
the laboratory are also shown in Figure 4. Eodo
metric (Eoed) and constrained (M) modulus are the
same mechanical parameters. If it was measured in
the laboratory, the Eoed abbreviation is used, while if
it is estimated from field measurements, CPT and
DMT, M is allocated.
The layers’ thickness in Figure 4. is just the aver
Figure 3. Consolidation modulus number of sand as age of the 32 CPTs. It happens that sometimes an
a function of cone resistance and OCR (Eslaamizaad and oedometric or DMT test falls below or above the
Robertson, 1996). assigned stratum.
469
method (Figure 4. Table 5.). However, the results of
the oedometric test vary with depth, Figure 2. Eq. 2-3.
Table 5. Constrained modulus of silty sand and gravelly

sand.
Silty sand Gravelly sand
Parameter Eoed MSang. Eoed MSang.
No. of
33 102 6 85
samples
Average 15.59 25.56 33.03 33.48
Min 5.30 15.78 26.31 25.99
Max 38.46 33.64 40.63 46.08
Std. 7.44 3.44 6.26 4.71
Co.V. 0.48 0.13 0.19 0.14
In lower two strata the DMT measurements result

in a higher constrained modulus (Figure 4.) as the
methods proposed by Lunne and Christoffersen
(1983) and Eslaamizaad and Robertson (1996).
4.2 Different soil layers

The estimated constrained moduli by different methods
for each layer are summarized in Figure 5. From the
applied methods Lunne and Christoffersen (1983) and
Eslaamizaad and Robertson (1996) were calibrated
against tests of silica sand, their capability to estimate
soil parameters in cohesive soils is questionable. It is
assumed that DMT measurements give a better predic
tion of constrained modulus as CPT measurements
(Marchetti, 2015) for two main reasons: (i) blades
cause penetration distortions lower than axy-cylindrical
probes, (ii) modulus by a mini load test relates better to
a modulus than a penetration resistance.
This is supported by many case studies such as
the Sunshine Skyway Bridge in Florida (Schmert
mann, 1998), where the constrained modulus from
back calculation was closes by the DMT measure
Figure 4. Soil strata with CPT values, estimated con ments. It is also common to use the DMTs as
strained (M), measured oedometric (Eoed) moduli. a calibration for CPT sounding, Jacksonville Power
Plant reported by Schmertmann (1998) and a project
In the upper two strata (landfill and lean clay) the in Bucharest (Poenaru et al., 2021).
oedometric moduli measured in the laboratory seem The correlation between CPT and DMT data was
to agree with the minimum value estimated by San analyzed in the top layer of the silty-sand, regarded
glerat (1972). In the top layer. constrained modulus as eolian sand, Figure 6.
derived from DMT measurements are way higher. Only 4 DMT and CPT test results for a single sub
than the results of other methods. The local inhomo layer, 47 measurement pairs are analyzed. The CPT
geneity is probably also playing a part in the extreme data is averaged to be compared with a DMT meas
DMT results. The scatter of the measurements is not urement. R2 value greater than 0.7 indicates that the
really concerning as no foundation will be built in linear correlation describes the investigated param
landfill. The moduli estimated by DMT measurements eters with sufficient accuracy. The following equation
in lean clay spread between the other methods based is on the border of the criteria, where: qc and the con
on CPT. In the lower two strata of silty sand and grav stant is in MPa, so as the obtained MDMT result.
elly sand, the results of oedometer tests distribute
around the values estimated by the Sanglerat (1972)
470
Figure 5. Constrained modulus of the layers (landfill. lean clay. silty sand and gravelly sand).
5 CONCLUSIONS
According to Table 5. the constrained moduli esti

mated by the Sanglerat (1972) method correlates
well with the measured ones in the laboratory by
oedometer.
The calculated oedometric modulus values have
a linear correlation with depth (Figure 2. and
Eq. 3–4.).
The dilatometer disturbs the soil less than the
CPT, so more realistic constrained modulus can be
estimated, which is way higher than the results of
investigated correlations (Eslaamizaad and Robert
son, 1996; Lunne and Christoffersen, 1983)
Figure 4-5.
A simple linear connection (Eq. 17) is derived
between the cone resistance of CPT and the con
strained modulus estimated by DMT in case of the
upper eolian layer of silty sand (Figure 6). It corres
Figure 6. Correlation between qc and MDMT for Eolian
ponds to the mean value, in case of design the char
sand layer. acteristic value must be used.
471
REFERENCES Marchetti, S., 1980. In Situ Tests by Flat Dilatometer.
ASCE Jnl GED 106, 299–321.
Baldi, G., Belotti, R., Ghionna, N., Jamiolkowski, M., Marchetti, S., Monaco, P., Totani, G., Calabrese, M., 2001.
Pasqualini, E., 1986. Interpretation of CPT and CPTU; The Flat Dilatometer Test (DMT) in soil investigation,
2nd part: drained penetration of sands., in: Fourth Inter in: A Report by the ISSMGE Committee TC16,
national Geotechnical Seminar., Singapour., pp. 143–156. Encyclopedia of Earth Sciences Series. T, Washington
Buisman, K., 1940. Grondmechanica. Uitgeverij Waltman, D.C., p. 41. https://doi.org/10.1007/978-3-319-73568
Delft, Neterlands. 9_174
Eslaamizaad, S., Robertson, P.K., 1996. Cone penetration MSZ EN ISO 22476-1:2013, 2013. Geotechnical investiga
test to evaluate bearing capacity of foundation in sands, tion and testing. Field testing. Part 1: Electrical cone
in: 49th Canadian Geotechnical Conference. Presented and piezocone penetration test (ISO 22476-1:2012).
at the 49th Canadian Geotechnical Conference, MSZE CEN ISO/TS 17892-5:2010, 2010. Incremental
St. John’s, Newfoundland, pp. 429–438. loading oedometer test.
Fioravante, V., Jamiolkowski, M., Tanizawa, F., Tatsuoka, F., Poenaru, A., Bilcu, A., Meirosu, A., Saidel, T.,
1991. Results of CPTs in Toyoura quartz sand, in: Sympo Batali, L., 2021. In situ and laboratory soil investiga
sium on Calibration Chamber Testing. Presented at the tions. Correlations between different parameters spe
Symposium on Calibration Chamber Testing, Elsevier, cific to Bucharest area. Presented at the ISC’6,
Potsdam, Germany, pp. 135–146. Budapest, p. 10.
Kádár, I., Nagy, L., 2018. Comparison of determination of Robertson, P.K., 2016. Cone penetration test (CPT)-based
oedometric modulus based on CPT and laboratory test soil behaviour type (SBT) classification system — an
ing in case of pleistocene sand layers. ce/papers 2, update. Can. Geotech. J. 53, 1910–1927. https://doi.org/
683–688. https://doi.org/10.1002/cepa.749 10.1139/cgj-2016-0044
Kumala Sari, P., Sari, T., Mochtar, N., Yogyanta, D., 2017. Robertson, P.K., 2009. Interpretation of cone penetration
Consolidation parameters evaluation of cohesive soils in tests — a unified approach. Can. Geotech. J. 46,
Lianganggam, Kalimantan Indonesia using CPT data. 1337–1355. https://doi.org/10.1139/T09-065
International Journal of Current Research 9, 58509–58513. Robertson, P.K., 1990. Soil classification using the cone
Lunne, T., Christoffersen, H.P., 1983. Interpretation of cone penetration test. Can. Geotech. J. 27, 151–158. https://
penetrometer data for offshore sands. Offshore Technol. doi.org/10.1139/t90-014
Conf.; (United States) 1. Sanglerat, G., 1972. The penetrometer and soil exploration:
Lunne, T., Powell, J.J.M., Robertson, P.K., 1997. Cone interpretation of penetration diagrams theory and
Penetration Testing in Geotechnical Practice. CRC Press. practice, 2nd enlarged ed. ed, Developments in geotech
Marchetti, S., 2015. Some 2015 Updates to the TC 16 nical engineering. Elsevier, Amsterdam Oxford
DMT Report 2001, in: DMT ’15 3rd Int. Conf. on the New York.
Flat Dilatometer. Presented at the DMT ’15 3rd Int. Schmertmann, J.H., 1998. Dilatometers Settle In. Civil
Conf. on the Flat Dilatometer, Rome, Italy, p. 23. Engineering ASCE 4.
472
© 2022 the Author(s), ISBN 978-1-032-31259-0
Evaluation of complex CPTu dissipation tests of B.E.S.T.

E. Imre
BGK, HBM EKIK, Óbuda University, Budapest, Hungary
M. Hegedűs
BME, Budapest, Hungary
Apostol Brc, Budapest, Hungary
L. Bates & S. Fityus

University of Newcastle, Newcastle, Australia
ABSTRACT: The u2 complex pore water pressure dissipation tests made in NC-LOC sandy soils at the
Bolivian test site were evaluated with the variants of a mathematically precise, automatic method in the func
tion of the testing time. The identified c became near constant and its error became small after a so called
minimum testing time tmin < t50 depending on the variant. The traditional t50 method failed in some cases since
the tests were with types I, II, III, V and t50 varied from 4s to a value being larger than 15 min (not measured).
1 INTRODUCTION 2.2 Geology and soil profile

he city of Santa Cruz de la Sierra lies in the southern
The pore water pressure dissipation tests of B.E.S.T
part of the Amazonas. The geology of the area is
were started to be evaluated by using the well-
characterized by a Paleozoic (250 million years old)
known t50 method and some suggested methods.
sedimentary basin.
The complex dissipation tests were made in NC
The soils of interest are quaternary. The main
LOC sandy soils, the t50 varied from 4s to an unknown
agent is the Piray River and its tributaries, which
value being longer that 15 min (in tests with tmax <
past meandering over the area has resulted in
t50), therefore, the t50 method failed in some cases.
a sedimentation-erosion-sedimentation process and
By using the suggested methods, the evaluation was
a geological profile dominated by fine to medium
made in the function of the testing time for all tests.
sands with intermittent layers of clay or clayey sand.
The c identified from truncated tests became near con
Due to the compressible soils, even light buildings
stant with t after a minimum testing time tmin < t50.
need to be supported on piles.
The groundwater table at the site is situated sea
2 MATERIALS AND METHODS sonally between the ground surface and about
0.5 m depth.
2.1 The tests Sandy/plastic soils had negative/positive excess
pore water pressure during penetration and dissipa
The B.E.S.T. site is about 40 m wide and 100 m long. tion, which indicates the soil types. The upper
The geotechnical conditions of the site have been about 10 to 20 m part of the profile consists of LOC
investigated using conventional in-situ and laboratory clays, silts, sands, in various combination and
methods at each single pile location (Figure 1). The fol thickness.
lowing in-situ tests have been performed: SPT, SCPTU In the pore water pressure profile (Figure 2) four
(with dissipation tests), SDMT, PMT, SASW and plastic layers 1 to 4 with positive excess pore water
REMI geophysical tests. Each borehole and field test is pressure were isolated between the following depths:
identified with the letter of its designated test pile. 2.5 to 5.0 m; 11 to 15 m; 15 to 17 m; below 22 m,
resp. Sand-silt is found among the plastic layers.
The measured u2 dissipation curves (Figures 2 to 3)
were with types I, II, III, V, (Sully et al, 1999). The t50
varied between 4 s and t50 > 15 min. A crust was
likely found in layer 1, indicated by the larger t50
Figure 1. Dissipation test locations. values (Figure 2.).
DOI: 10.1201/9781003308829-66
473
Table 1. Dissipation test list, depth, notation, t50.
Layer 2 Layer 4
depth [m] sign t50 [s] depth [m] sign t50 [s]
A2 13,6 21 190 22,0 22 120

A3 13,6 23 220
B1 13,3 24 250
B2 22,5 9 > 850
D2 22,0 10 > 720s
DC 12,8 35 450 20,3 36 4
Table 2. Dissipation test list, depth, notation, t50.
Layer 1 Layer 3
depth [m] sign t50 [s] depth [m] sign t50 [s]
C1 15,17 25 1020
Figure 2. Dissipation tests in plastic soil layers 1 to 4. D2 2,7 34 500 15,63 33 550
E1 2,3 32 500
Table 3. Comparing c [cm2/s] from various methods, t =

tmax.
Depth c [cm2/s] with method

z [m] t50 method k = 1, fast slow
21 13.6 0.02 to 0.13 0.6 0.42
22 22.0 0.03 to 0.42 1.8 1.2
24 13.3 0.01 0.6 0.48
36 20,3 negative (sand) 54 4.8
9 22.5 shorter than t50 0.15 0.06
10 22,0 shorter than t50 0.18 0.12
2.3 The evaluation methods

The one-point model fitting (Teh-Houlsby, 1988;
Lunne at al. 1992) requires time of t50 and Ir which
is difficult to assess (Mayne, 2007).
The one-point model fitting (at t50 cannot
handle more than one initial condition (if t50< 50
s, and if the dissipation is starting from less than
u0 values). The reliability of the identified c is
not tested.
The suggested evaluation methods imply (i) an
automatic and mathematically precise non-linear
inverse problem solution, (ii) some initial condi
tion identification methods which can be used in
partly drained penetration, (iii) some reliability
testing.
The methods are depending on r1 but its value can
Figure 3. (a) to (d): Dissipation curves in layer 1 to 4. In the be taken into account using a model law, resulted
same layer both monotonic, non- monotonic curve may occur. from the derived time factor T (see App,).
474
The newer methods (“fast” and “slow”) are
based on the mathematically precise Least
Squares fitting of a new consolidation model,
with different initial condition identification.
In the “slow” method, the first 1..200 terms of
the analytical solution (large numerical work)
are used, with some pre-elaborated, parametric
shape functions. In the “fast” method, the first
few (1..k) terms are used (small numerical work)
for both the monotonic (k = 1) and the non-
monotonic (1..k >1) time variation of the pore
water pressure.
The c values were identified at testing times t
= 0,3 min; t = 0,5 min; t = 5 min besides at t =
tmax. Since only one case (test 24) was a “real
time” evaluation in the sense, that the evaluation
was made after each measured data sampled gen
erally in every second, the results of this work
are approximate. It can also be noted that the LS
merit function related to the slow method has
several minima, the fast method has a unique
distinct minimum for k = 1, can be quasi-
degenerated, with large error for k =3 (see
Figure 4).
Figure 4. Deepest section of the LS merit function, Test 21. (a)
Slow method, several minima, disctinct minimum for k = 1. (b)
3 RESULTS The minimum is quasi-degenerated, with large error for k = 3.
3.1 Evaluation using maximum testing time Table 4. Test 24, c - t, slow and fast method, t50 = 6,1 min.
According to the results (Table 3), the c values c [cm2/s] with method
identified with the various methods as earlier testing
(Imre et al, 2018). The fast methods gave larger time [min] fast k = 1 slow fast k = 3
c by about a factor of 1.5 than the slow methods.
The one-point t50 method gave smaller c by gen 0,3 1,8 0,48 6
erally about factor of 10 to 30 depending on Ir. 0,5 1,2 0,12 6
The fast k = 1 method usually gave distinct solu 5 0,6 0,54 0,6
tion, a worse fit, a faster decay, and thus a higher 52 0,6 0,48 0,24
c value than the slow method.
Table 5. Test 21, c - t, slow and fast method, t50 = 3,2 min.
3.2 Evaluation of truncated tests, monotonic data
According to the results (see Figures 5 to 7, c [cm2/s] with method
testing
Tables 4 to 6 ), the c - t functions generally time [min] fast k = 1 slow fast k = 3
decreased for the fast method, and was generally
constant for the slow method. The identified 0,33 3,00 0,42 1,80
c became generally about constant at a minimum 0,50 1,80 0,30 6,00
testing time tmin except in one case (test 21). 3,33 1,20 0,60 0,60
The tmin < t50 was met for the slow method in 23,50 0,60 0,42 0,36
every case, the tmin was shorter or longer than
the t50 time for the fast methods. The fast k = 3
method gave better fit but too large parameter Table 6. Test 22, c - t, slow and fast method, t50 = 2 min.
error.
c [cm2/s] with method
testing
3.3 Evaluation of truncated tests with complex data time [min] fast k = 1 slow fast k = 3
According to the results shown in Figures 8 to 10,
0,33 5,40 1,80 6,00
Tables 7 - 8, at the initial – non– monotonic – part of
0,50 4,20 1,20 6,00
the tests, the parameter error was high and c changed.
3,33 1,80 1,20 1,80
The identified c became about constant and the error
18,00 1,80 1,20 1,80
dropped at a minimum testing time tmin.
475
Figure 6. The c-t function. (a) Test 24, layer 2 at B1, t50 =
250 [s]. (b) Test 21, layer 2 at A2, t50 = 190 [s]. (c) Test 22,
layer 4 at A2, t50 =120 [s].
Figure 5. Monotonic curves, measured and fitted data. (a)

Test 24, the dissipation test in layer 2 at B1. (b) Test 21 in
layer 2 at A2. (c) Test 22 in layer 4 at D2.
Table 7. Complex test 34, c with t, k =1 to 3, t50 = 8,3 min.
c [cm2/s] with method

testing
time [min] fast k = 1 slow fast k = 3
0,33 1.80 4.80 3.60

0,50 1.20 4.20 3.00
4,00 0.12 1.20 1.20
60,00 0.18 0.60 0.60
Figure 7. “Real-time” evaluation, k = 1. Test 24. (see

Figure 6(a).).
476
The tmin < t50 was met for the slow method and
the fast k = 1 method in general.
The tests 9 and 10 – made on a more plastic
soil layer, being shorter than t50 –, and test 36 –
made in sand with fully negative excess pore
water pressure – were successfully evaluated with
similar results.
4 DISCUSSION
4.1 The effect of filter position

The c - t function was determined for the mono
tonic dissipation tests of the Bothkennar soft clay
measured after undrained penetration, in filter
positions E to A with the fast k =1 method (Imre
et al, 2010). The c - t function was about con
stant in fps E, in filter position D it became
about constant at a minimum testing time tmin
~30 min which was less than t50 ~90 min
(Figure 11, Table A-1).
In the present study, after partly drained pene
tration, in fps u2 (~fps D and C), the tmin was
the smallest and less than t50 in every case for
the slow method only. For the fast methods, tmin
was less than t50 in about half of the cases. The
minimum testing time tmin can be assessed pre
cisely from real-time evaluation. Further research
Figure 8. Complex dissipation curves, measured and fitted is suggested on this.
data. (a) Test 23, layer 2 at A3. (b) Test 9, layer 4 at B2.
4.2 Complex dissipation of LOC soils

According to Figure 3, in the same layers both
monotonic, non-monotonic curves occurred. Type
I and II curves are commonly understood results
from normally consolidated (NC) and lightly
over-consolidated (LOC) clays, type III to V are
attributed to heavily over-consolidated (HOC)
soils.
Some recent studies (Lim, 2019, Figure 11)
revealed that „Extremely fine mesh around the
cone penetrometer made possible to study dissi
pation curves from very closely spaced stress
points. It is found that type I, II and III curves
are all possible outcomes from NC and LOC
soils.”
4.3 Initial condition and space domain

The suggested methods are based on r1, valid after
undrained penetration. The one-point t50 method
gave smaller c by about factor of 10 to 30 than
the suggested methods. The difference can prob
ably be attributed to the due to undrained value
r1 =37r0.
The c identified can be modified for other r1 using
the model law (see the App.) but no values are avail
able to r1 valid in (partly) drained penetration. Further
Figure 9. Complex dissipation tests. Test 34, layer 1, D2,
research on is needed (Osman, 2021).
t50 = 500 [s]. (a) Function c-t. (b) Coefficient of variation
Cv(c)-t.
477
Table 8. Test 23, Function c-t, (slow:0,6), t50 = 4 min.
fast method
testing
time [min] k=1 k=3
2
c [cm /s] Cv(c) c [cm2/s] Cv(c)
0,33 0,06 18,66 4,20 0,11
0,50 0,06 11,27 3,00 0,12
3,33 1,20 0,30 1,20 0,11
33,50 1,20 0,31 0,60 0,16
5 SUMMARY, CONCLUSION
5.1 The evaluation results

Figure 10. Dissipation test in sand (DC, test 36). In the present study, the complex dissipation curves
of B.E.S.T. test site, measured in filter position u2,
were evaluated in the function of the testing time t.
(1) The c identified with the various methods (t50
and the newer) using the long test data (t = tmax)
showed similar pattern as in previous research.
Some complex or short tests were impossible to be
evaluated with the t50 method.
(2) Truncated tests (t < tmax), were evaluated with
the newer methods with various elapsed time values.
Generally (with one exception), at a minimum test
ing time tmin, the identified c became about constant,
the error of c became as small as in the long tests.
The tmin was always smaller than t50 for the slow
method, and sometimes for the fast methods even in
the case of the complex shaped, short tests.
5.2 Summary, conclusions

The c was identified with various methods only in
a few elapsed times. It became generally constant
Figure 11. Variation of the uncorrected c with test duration τ. after a minimum testing time tmin, which was always
smaller than t50 for the slow method but only in half
of the cases for the fast methods with various k.
It follows that - by evaluating in real time with the
fast methods -, the c can be determined theoretically
earlier than t50. More than one method is suggested
to be used for the evaluation since the solution of the
fast .k=1 method is unique only. The slow method is
suggested to be applied for the evaluation only in
a few elapsed times, due to the large numerical work.
Previous research indicates that tmin can be the
shortest in filter position u3 (Imre et al 2010) due to
the one-dimensional state. Further research is sug
gested on the real time evaluation, on the tmin in vari
ous filter positions and on partly drained penetration
(to assess initial and boundary condition).
ACKNOWLEDGEMENT
The help of Professor Peter Roberson in data sugges

tion is greatly acknowledged.
Figure 12. Shape issues.
478
REFERENCES monotonic data of Bothkennar clay (fast model, first
term solution, k =1).
Baligh, M. M. 1986. Undrained deep penetration, II. pore If the evaluation is made assuming r1 =37r0 – valid
pressures. Geotechnique, 36(4): 487–503. in filter position E but using data of various filter posi
Imre, E.; Rózsa, P.; Bates, L.; Fityus, S 2010. Evaluation of tions, – then from the identified c, the “correct” value
monotonic and non-monotonic dissipation test results. of r1 can be back-calculated with the “c formula”:
Computers and Geotechnics, 37, Issues 7–8, 885–904.
Imre, E; Schanz, T; Bates, L; Fityus, S. 2018: Evaluation of
complex and/or short CPTu dissipation tests. Cone Pene
tration Testing 2018: Proc. of the 4th International Sympo
sium on Cone Penetration Testing. CRC Press, 351–357.
Lunne, T., Robertson P.K. and. Powell, J.J.M 1992. Cone
penetration testing, Blackie Academic & Professional. According to the results (Table A-1), the back-
Lim, Y X; Tan, S A; Kok-Kwang Phoon 2019 Interpretation computed r1 values varies from 37r0 to 17 r0 as the
of horizontal permeability from piezocone dissipation tests sensor positions changes from E to A. As
in soft clays. Computers and Geotechnics, 107. 189–200. a comparison, the value of r1 with the strain path pre
Sully, J.P., Robertson, P.K., Campanella, R.G., Woeller, D.J.
diction of Baligh 1986 varies from 22-23r0 to 37r0.
1999. An approach to evaluate field CPTU dissipation data
in overconsolidated fine-grained soils. CGJ. 36: 369–381. The r1 can be computed on the basis of the meas
Teh, C.I. and Houlsby, G.T. 1988. Analysis of the cone pene ured dissipation times with the “t formula”:
tration test by the strain path method. Proc. 6th Int. Conf.
on Num. Meth. in Geomechanics, Innsbruck. 1:397–400.
Osman, A. 2021. Modelling of penetrometers. Dissipation test
workshop ISC6. Sept. 26th 2021, personal commu-nication.
APPENDIX - SUGGESTED MODEL/MODEL using the r1 value valid for filter position E. According
LAW to the results (Table A-1), the back-computed r1 values
vary from 37r0 to 18r0 as the sensor positions changes
Consolidation model from E to A.
The system of differential equations, new boundary
conditions, for pore water pressure u for displacement
v and the suggested time factor T (Imre et al 2010): Table A-1. The r1 from identified c and measured t50 values.
computed Measured computed

Identified c [cm2/s] r1/r0 t50 (min) r1 [cm]
A 0.09 17,0 8 18.74

B 0.06 20,8 22 29.92
C 0.04 25,5 60 48.28
D 0.02 36,0 90 58.74
E 0.02* 36,0* 110 64.75*
* considered as reference value
where v is displacement, r is space coordinate, t is

time. k is permeability, γv is unit weight of water,
Eoed is oedometric modulus, Jn and Yn are Bessel
functions, c is coefficient of consolidation, the λi, μi
and Ci (i=1..k) depend on the space domain (r0 and r1)
and the initial condition, T time factor.
The value of r1 and initial condition can be deter
mined by the strain path theory using the rigidity
index Ir, the OCR, friction angle and rate of penetra
tion (Figure A-1, Baligh 1986).
Model law
In this section the use of the model law is presented Figure A-1. The zero pore pressure line r1 (after Baligh, 1986).
on some examples, regarding the evaluation of the Left and right: bilinear and hyperbolic modelling. Ir =150.
479
© 2022 the Author(s), ISBN 978-1-032-31259-0
Simulation of CPT penetration in sensitive clay

J. Isaksson
Chalmers University of Technology, Sweden
NCC AB, Sweden
J. Yannie
NCC AB, Sweden
M. Karlsson & J. Dijkstra

Department of Architecture and Civil Engineering, Chalmers University of Technology, Sweden
ABSTRACT: This paper presents the results from numerical simulations of CPTu penetration in a natural clay
combining the SCLAY1S constitutive model with a large deformation Finite Element framework including
a coupled deformation and porewater pressure formulation. The hierarchical model formulation of SCLAY1S cap
tures many features of a natural sensitive clay, such as the evolving anisotropic strength-stiffness response, as well
as the degradation of the initial bonding. A sensitivity analysis is performed varying the overconsolidation ratio
(QCR), bonding and anisotropy, also the hydraulic conductivity (hence, cv) of the clay. The findings indicate that
some soil properties (the cv and OCR) impact both the normalised cone resistance Qt and the generation of excess
porewater pressures. In contrast the sensitivity St of soft soils primarily affects Qt. In the current work it seems
that the effects of the inherent and stress induced (from CPT penetration) anisotropy is not detected using these
normalised plots.
1 INTRODUCTION of CPTu penetration are (i) the capability of the Finite

Element (FE) code to deal with large deformations
The cone penetration test is a widely used method to (ii) the adequate coupling of deformations and the
perform geotechnical site investigation, by continu generation/dissipation of excess porewater pressures
ous measuring of the cone resistance, the sleeve fric (iii) a constitutive model that incorporates the com
tion, and in case of the piezocone (CPTu) the plex features of natural soils.
generated excess porewater pressures, during the A number of numerical methods able to simulate
penetration into a soil. This allows the mapping of the kinematics of CPTu penetration in FE have been
a deposit to be performed in a time-effective manner reported, among others the Arbitrary Lagrangian Euler
with a high resolution (Lunne et al. 1997). Further ian method (Berg et al. 1996, Walker & Yu 2006),
soil characterisation can be performed using classifi Material Point Method (Ceccato et al. 2016), Geotech
cation systems based on statistical correlations of nical Particle Finite Element Method (Hauser &
normalised CPTu results against borehole data, see Schweiger 2021, Monforte et al. 2021), and remeshing
e.g. Robertson (2016) and Schneider et al. (2008). procedures (Hu & Randolph 1998, Orazalin & Whittle
Due to the continuous measurement of the soil 2018, Mahmoodzadeh et al. 2014). In some cases the
response, the CPTu is a great tool to detect differ effects of CPT penetration are captured in an Updated
ences in the response between and within soil layers Lagrangian framework (Yi et al. 2012, Konkol & Bała
by relying on a contrast in hydro-mechanical proper chowski 2018, Mahmoodzadeh et al. 2014).
ties, e.g a change in hydraulic conductivity, overcon Some of the studies (Ceccato et al. 2016, Mon-
solidation ratio or sensitivity (brittleness). forte et al. 2021, Yi et al. 2012, Konkol & Bała
Another approach to establish the relation between chowski 2018, Mahmoodzadeh et al. 2014, Orazalin
soil properties and CPTu response is to use numerical & Whittle 2018) also incorporates a coupled stress
modelling where a prescribed change of a model par formulation enabling the study of partial drainage
ameter of a given constitutive model leads to during penetration. Constitutive models able to
a change in CPTu response. This approach is becom describe advanced soil features such as brittleness
ing increasingly more attainable with the ongoing (Monforte et al. 2021) and anisotropy (Hauser &
developments for numerical analyses. Three model- Schweiger 2021, Orazalin & Whittle 2018) has also
ling aspects that are necessary for accurate simulation been incorporated to simulate CPTu penetration.
DOI: 10.1201/9781003308829-67
480
This paper builds upon those previous studies by allowing for groundwater flow across the boundary.
implementing SCLAY1S in a fully coupled Eulerian At the bottom boundary, vertical groundwater flow
Finite Element (FE) framework. Subsequently, the and soil movement is prevented. The top boundary
relation between the CPTu response and different of the domain is modelled with a prescribed vertical
soil properties is investigated. The model parameters load that is in equilibrium with the total vertical
varied, include the hydraulic conductivity (k), the stress (σv ) and is equal to the sum of σv 0 and the ini
sensitivity of the soil (St ), the fabric anisotropy and tial porewater pressure u0 . The increase in stress due
the overconsolidation ratio (OCR). to the weight of the soil in the domain is set to be
zero to create a uniform soil domain.
2 NUMERICAL MODEL
Natural features of soft clay, such as breakage of ini

tial bonding and fabric anisotropy, are captured by
the SCLAY1S constitutive model (Koskinen et al.
2002) and (Karstunen et al. 2005). The elasto-plastic
model originates from the Modified Cam Clay
(MCC) constitutive model (Roscoe & Burland
1968), in addition to the volumetric hardening of
MCC, SCLAY1S also incorporates rotational hard
ening and gradual degradation of bonding due to
plastic strains in the soil. In short, the evolution of
the initial anisotropy and degradation of strength is
controlled by volumetric plastic strains and deviato
ric plastic strains in the hardening law. The model is
hiearchical, i.e. an appropriate choice of model
parameters leads to the (de-) activation of the model
features that capture (evolution of) anisotropy and
destructuration. Hence, in its simplest form the
model formulation becomes identical to MCC.
For the current work, the SCLAY1S model was
implemented in the Tochnog Professional (Rodde
man 2021) finite element framework that is able to Figure 1. Boundary conditions and mesh in the region
handle large deformations by using an Eulerian close to the penetrating CPTu.
description with a fixed mesh, where the solution
fields for the stress, material velocity and other state
variables of the calculation are advected through the
domain. Penetration of the CPTu into the soil is per All simulations presented in the study were per
formed with the moving boundary method proposed formed using a 60 cone with a diameter (d) of
by Dijkstra et al. (2011). Initially, the cone is con 0.036 m corresponding to a radius (r) of
sidered to be outside of the calculation domain, i.e. 0.018 m. The height of the domain h was set to two
above the soil surface, and the desired stress state times the penetration depth and the width w was set
and other state variables required for the model are to 40r, to prevent numerical disturbance related to
prescribed to establish the initial state in the model. boundary effects. A structured quadrilateral mesh (see
The numerical penetration is then performed by Figure 1) was required in the location of the penetrat
defining a geometric entity representing the CPTu ing cone to ensure geometrical compatibility between
and prescribing the penetration velocity v to all the mesh and the penetrating cone that is prescribed
nodes in this geometry while simultaneously expand with a geometry entity. Quadrilateral elements were
ing the geometry downwards with the same penetra used in a region extending 5 cone radii (r) from the
tion velocity. axis of symmetry. The rest of the domain is filled
The axisymmetric nature of the problem is with unstructured triangular elements. In total, the
exploited using a 2D simplification where the hori model contains 1789 quadrilateral elements and 3579
zontal soil movement and groundwater flow is pre triangular, both with first order shape functions. All
vented perpendicular to the axis of symmetry. The simulations in this paper were performed with
geometry and boundary conditions of the numerical a penetration rate v of 0.02 m/s down to a final pene
model are presented in Figure 1. The initial stress tration depth of 20d. The porewater pressure pre
state is prescribed by the vertical effective stress sented in this study was extracted from a position
(σv 0 ), initial porewater pressure (u0 ) and the initial right above the cone shoulder corresponding to the u2
earth pressure coefficient (K0 ). Horizontal move position. The cone resistance qc was calculated from
ment is prevented at the far right boundary while the total force needed to push the inclined cone tip
keeping the porewater pressure constant to u0 , hence downwards divided by the area of the cone. The net
481
cone resistance qnet was calculated by subtracting the volume (drainage lengths) to the properties of the soil
initial vertical stress σv0 from the cone resistance qc . such as the vertical effective stress initial void
ratio e0 , stiffness λ and the hydraulic conductivity
(via the vertical consolidation coefficient cv ).
Table 1. Model parameters used to investigate the effect DeJong & Randolph (2012) proposed a backbone
of drainage conditions on the CPTu response.
curve of both the net cone resistance and excess pore-
Symbol Parameter Value water pressure normalised with the corresponding
undrained value based on the result from seven differ
Vertical effective stress [kPa] 109 ent studies investigating the change in response for the
u0 Initial porewater pressure [kPa] 70 CPTu under different drainage conditions and confin
K0 Initial earth pressure coefficient [-] 0.61 ing stress p. Mahmoodzadeh & Randolph (2014) also
OCR Overconsolidation ratio [-] 1.02 proposed a backbone curve based on a series of centri
e0 Initial void ratio [-] 1.41 fuge test of CPTu penetration in kaolin clay. The net
λ Virgin compression index [-] 0.205 cone resistances are normalised with the results from
κ Swelling/recompression index [-] 0.044 the undrained penetration simulation and are presented
Poisson’s ratio 0.3 in Figure 2. Whereas, the results for the normalised
M Slope of CSL line [-] 0.9 excess porewater pressure are presented in Figure 3.
χ0 Initial amount of bonding [-] 0 Both figures also show the two proposed backbone
a Rate of destructuration [-] 0 curves.
b Rate of destructuration due to 0 The transition of the simulated net cone resistance
to deviator strain [-] from the undrained to the intermediate and drained
α0 Initial anisotropy [-] 0 state are in good agreement with both backbone
ω Rate of rotation [-] 0 curves. The relative magnitude of the net cone resist
ωd Rate of rotation due to 0 ance in the drained state, however, is considerably
deviator strain [-] larger when compared to the proposed backbone
curves. As this study is with equal strength in the soil
as in the element near the interface, the contact
between the CPTu and the soil can be considered
3 VARIATION OF HYDRAULIC rough. Monforte et al. (2021) performed an additional
CONDUCTIVITY sensitivity study on the impact of the interface rough
ness on the CPTu simulations. The normalised net
Initially, the effect of the drainage conditions on the cone resistance for the rough interface ( = 19)
CPTu response was studied using a MCC model for increased with about 40 % from the smooth interface
mulation, by varying the hydraulic conductivity k in (included in Figure 2). In contrast, the normalised
the range 5:510-3 m/s and 1:110-8 m/s. An isotropic excess porewater pressure response is not greatly
hydraulic conductivity was used in all performed affected by the interface formulation. Looking at
simulations. All the model parameters used in the Figure 2 the results from this study fit in between the
numerical study are presented in Table 1 and are smooth and the rough interface response reported by
based on those derived for kaolin clay, as used for Monforte et al. (2021).
the numerical studies of the CPTu in Mahmoodzadeh The normalised excess porewater pressure from
et al. (2014). The normalised penetration velocity V this study is slightly shifted compared to the other
is used to define the current drainage conditions for studies (Figure 3). This is due to the presence of some
quasi-static penetration problems, as it enables the numerically locked-in porewater pressures in a single
comparison between various test conditions. V is element near the cone shoulder, i.e. at the u2 position
defined as: and is most prominent for very low hydraulic con
ductivities corresponding to a practically undrained
state with normalised penetration velocities above 50.
The porewater pressure presented herein, are
unsmoothed and taken from the u2 position and is
where v is the penetration rate, d is the diameter of considered to be accurate when looking at the relative
the CPT cone and cv is the vertical consolidation change in response between the analyses in the sensi
coefficient of the soil. tivity study.
4 CPTU IN SOFT CLAYS
The numerical investigation into the impact of soil

The normalised penetration velocity helps to correct properties on the CPTu penetration in soft (sensi
for experimental scaling conditions by linking the tive) clays was performed starting from a normally
penetration velocity and size of the object and soil consolidated and isotropic reference state without
482
Figure 2. Normalised cone resistance over normalised penetration rate. Comparison between results from this study and
Mahmoodzadeh & Randolph (2014) and DeJong & Randolph (2012). Results from Monforte et al. (2021) is included to
indicate the effect of interface properties on the CPTu response.
Figure 3. Normalised excess porewater pressure over normalised penetration rate. Comparison between results from this
study and Mahmoodzadeh & Randolph (2014) and DeJong & Randolph (2012). Results from Monforte et al. (2021) is
included to indicate the effect of interface properties on the CPTu response.
Table 2. Parameters used for investigation of the CPTu CPTu response was studied by introducing an ini
response in soft clays. tially inclined yield surface, that evolves with
deviatoric and volumetric strains, in the model for
Symbol Parameter Value mulation. Table 2 presents the range of the
SCLAY1S parameters used. The rate parameters
OCR Overconsolidation ratio [-] 1.2, 1,5, 1,8
and anisotropy α0 are assumed based on Gras et al.
χ0 Initial amount of bonding [-] 2, 5, 10, 20, 50
(2017) for natural clays, whilst keeping the ori
a Rate of destructuration [-] 6
ginal parameters from the kaolin clay. This ensures
b Rate of destructuration due to 0.4
consistency of model parameters between simula
to deviator strain [-]
tions. Although this approach captures the soft soil
α0 Initial anisotropy [-] 0.352
features found in natural clays, the dataset does not
ω Rate of rotation [-] 10
represent a natural clay deposit.
ωd Rate of rotation due to 0.374
Robertson (1990) proposed a classification system
deviator strain [-]
based on the normalised cone resistance Qt and pore
pressure ratio Bq , where
initial bonding, using a normalised penetration vel
ocity of V = 200 for the CPTu. First, the impact of
OCR on the soil response was investigated by
increasing the OCR in three increments from 1.02
to 1.8. The brittleness of the soil was also investi
gated by varying the SCLAY1S state parameter for
destructuration χ0 between 0 (no initial structure)
and 50 (clay with a high sensitivity). Although this
parameter is closely related to the sensitivity of the The Qt is the relation between the net cone resist
soil it should not be considered to be similar. ance from the CPTu measurements and the initial
Finally, the impact of fabric anisotropy on the effective vertical stress. Bq is the excess porewater
483
pressure divided by the net cone resistance. This
classification system is shown in Figure 4 with the
results from the present numerical study. The
arrows that annotate the data points correspond to
each model parameter and are showing the direc
tion of the normalised CPT response when the par
ameter is increased in the numerical analysis.
Distinct trends for each parameter are clearly
identified and are in good agreement with trends
proposed by Robertson (1990), for both St
and OCR.
The numerical results are also presented in the
classification chart (Figure 5) originally proposed
by Schneider et al. (2008), which is based on Qt
and the excess porewater pressure (Δu) normalised
with the initial vertical effective stress The
impact of changing St , OCR and cv indicates clear
trends that are in good agreement with the
response suggested by Schneider et al. (2008). The
effect of fabric anisotropy α only shows limited
impact on the results. The results only slightly
changed, due to the lower Qt and excess porewater
pressures when compared to the isotropic model
results.
Figure 5. The effect on CPTu response from changing con

solidation coefficient cv ; overconsolidation ratio OCR; sen
sitivity St and considering fabric anisotropy α in the
charaterisation chart for CPTu proposed by Schneider et al.
(2008).
5 CONCLUSIONS
This paper presents the results from a series of CPTu

simulations using a large deformation Finite Element
framework in which partial consolidation during
penetration is considered by linking the material
deformations to the coupled response of porewater
flow. An Eulerian framework, in which the mesh is
fixed and the soil is able to move independently of
the mesh, has been used, to avoid mesh distortions
from large deformations associated with the CPTu
penetration. The SCLAY1S model is implemented
for these analyses, as it captures the evolving aniso
tropic strength-stiffness response, as well as the deg
radation of the initial bonding present in natural
sensitive clays.
In the first part of the paper the effect of different
drainage conditions is quantified and the overall trend
compares well with prior work. Further studies need,
however, to be conducted to improve the accuracy of
the calculated porewater pressures at the shoulder of
Figure 4. The effect on CPTu response from changing the the CPTu. Extending the study to also include the
consolidation coefficient cv ; overconsolidation ratio OCR; response on the friction sleeve of the CPTu could fur
sensitivity St and considering fabric anisotropy α in the ther expand the conclusions of this study.
characterisation chart for CPTu proposed by Robertson The impact of features that are fundamental to
(1990). soft soils, i.e. hydraulic conductivity, OCR, sensitiv
484
ity and anisotropy, on the CPTu response have been calibration for soft soils using multi-objective
investigated in a hierarchical manner. The following optimisation. Computers and Geotechnics 90, 164–175.
can be concluded after integrating the results in the Hauser, L. & H. Schweiger (2021). Numerical study on
CPTu classification charts: undrained cone penetration in structured soil using
g-pfem. Computers and Geotechnics 133, 104061.
• Increasing the hydraulic conductivity leads to an Hu, Y. & M. Randolph (1998). A practical numerical
increase in normalised penetration resistance approach for large deformation problems in soil. Inter
while the normalised excess porewater pressure is national Journal for Numerical and Analytical Methods
decreasing. in Geomechanics 22, 327–350.
• Increasing the OCR is associated with an Karstunen, M., H. Krenn, S. Wheeler, M. Koskinen, &
increase in both the normalised cone resistance R. Zentar (2005). Effect of anisotropy and destructura
tion on the behavior of murro test embankment. Inter
and the normalised excess porewater pressure.
national Journal of Geomechanics 5(2), 87–97.
• Increasing St leads to a considerable decrease in Konkol, J. & L. Baachowski (2018). Large deformation
the normalised cone resistance while leaving the modelling of cpt probing in soft soil—pore water pres
normalised excess porewater pressure nearly sure analysis. In M. A. Hicks, F. Pisanò, and J. Peuchen
unaffected (Eds.), Proceedings of the 4th International Symposium
• The simulated CPTu response is practically on Cone Penetration Testing (CPT’18), London, pp.
unaffected by soil anisotropy. 371–376. CRC Press.
Koskinen, M., M. Karstunen, & S. Wheeler (2002). Model-
The conclusions of this study are in good agreement ling destructuration and anisotropy of a natural soft clay.
with suggestions from Robertson (1990) and Schnei In Mestat (Ed.), Proc., 5th European Conf. Numerical
der et al. (2008) for the anticipated response from Methods in Geotechnical Engineering, Paris, pp. 11–20.
a change in cv , OCR and St . Hence, the results of Presses de l’ENPC/LCPC.
this study contribute to the interpretation of the Lunne, T., P. Robertson, & J. Powell (1997). Cone penetra
widely used classification charts, by linking it to the tion testing in geotechnical practice. New York: E &
FN Spon/ Routledge.
fundamental features of natural soils. Mahmoodzadeh, H. & M. Randolph (2014). Penetrometer
The extensive empirical evidence used to establish testing: Effect of partial consolidation on subsequent
the relation between CPT and soil characteristics is dissipation response. Journal of Geotechnical and
in good agreement with the numerical results, Geoenvironmental Engineering 140 (6), 04014022.
increasing the confidence in the ability to accurately Mahmoodzadeh, H., M. Randolph, & D. Wang (2014).
simulate penetration into soft soils with the proposed Numerical simulation of piezocone dissipation test in
numerical method. Finally, the numerical simulations clays. Geotéchniqe 64 (8), 657–666.
should be validated further against in-situ CPTu data. Monforte, L., A. Gens, M. Arroyo, M. Mánica, &
J. Carbonell (2021). Analysis of cone penetration in brit
tle liquefiable soils. Computers and Geotechnics 134,
104123.
ACKNOWLEDGEMENTS Orazalin, Z. & A. Whittle (2018). Realistic numerical simula
tions of cone penetration with advanced soil models. In
The authors acknowledge the financial support pro M. A. Hicks, F. Pisanò, and J. Peuchen (Eds.), Proceed
vided by SBUF (Development fund of the Swedish ings of the 4th International Symposium on Cone Penetra
construction industry, grant 13614) and BIG (Better tion Testing (CPT’18), London, pp. 483–489. CRC Press.
Interaction in Geotechnics, grant A2019-19, from Robertson, P. (1990). Soil classification using the cone
the Swedish Transport Administration) penetration test. Canadian Geotechnical Journal 27,
151–158.
Robertson, P. (2016). Cone penetration test (cpt)-based soil
REFERENCES behaviour type (sbt) classification system — an update.
Canadian Geotechnical Journal 53, 1910–1927.
Berg, P., R. Borst, & H. Huétnik (1996). An eulerean finite Roddeman, D. (2021). Tochnog professional user’s manual,
element model for penetration in layered soil. Inter october 21. https://www.tochnogprofessional.nl/man
national Journal for Numerical and Anaylytical uals/user/user.pdf. Accessed: 2021- 10-21.
Methods in Geomechanics 20, 865–886. Roscoe, K. & J. Burland (1968). On the generalized stress-
Ceccato, F., L. Beuth, P. Vermeer, & P. Simonini (2016). strain behaviour of wet clay. In Engineering plasticity,
Two-phase material point method applied to the study of Cambridge Univ. Press, Cambridge U.K., pp. 553–609.
cone penetration. Computers and Geotechnics 80, Schneider, J., M. Randolph, P. Mayne, & N. Ramsey
440–452. (2008). Analysis of factors influencing soil classification
DeJong, J. & M. Randolph (2012). Influence of partial con using normalized piezocone tip resistance and pore pres
solidation during cone penetration on estimated soil sure parameters. Journal of Geotechnical and Geoenvir
behavior type and pore pressure dissipation onmental Engineering 134(11), 1569–1586.
measurements. Journal of Geotechnical and Geoenvir Walker, J. & H. S. Yu (2006). Adaptive finite element ana
onmental Engineering 138 (7), 777–788. lysis of cone penetration in clay. Acta Geotechnica 1,
Dijkstra, J., W. Broere, & O. Heeres (2011). Numerical 43–57.
simulation of pile installation. Computers and Geotech Yi, J. T., S. H. Goh, F. Lee, & M. Randolph (2012).
nics 38, 612–622. A numerical study of cone penetration in fine-grained
Gras, J., N. Sivasithamparam, M. Karstunen, & J. Dijkstra soils allowing for consolidation effects. Geotéchniqe 62
(2017). Strategy for consistent model parameter (8), 707–719.
485
© 2022 the Author(s), ISBN 978-1-032-31259-0
A CPT-based method for estimation of undrained shear strength of sands

and transitional soils
K. Kaltekis & J. Peuchen
ABSTRACT: This paper presents a practical approach for developing a site-specific CPT-based method for
monotonic undrained shear strength (su) in sands and transitional soils, using results of laboratory undrained tri
axial compression (CU) tests on reconstituted and undisturbed specimens as reference. The methodology includes
use of net cone resistance values normalised to vertical effective stress, a procedure for pairing of CPT data with
CU test results, and definition of a practical failure criterion for deriving su from CU test data. The presented
approach is particularly useful for application in offshore wind, where the economics of wind farm development
favour performing only a single cone penetration test (CPT) per wind turbine location. This setting drives develop
ment of CPT-based methods for key geotechnical parameters for foundation design.
1 INTRODUCTION This paper includes an example of the site-

specific approach, using input data taken from two
The characterisation of undrained behaviour of sands wind farm sites offshore Netherlands, namely the
and transitional soils (e.g. silty sands, low plasticity Hollandse Kust (west) site and the Hollandse Kust
silts) is important for large foundations subject to sig (noord) site (Figure 1, HKW and HKN respect
nificant short-duration loading and cyclic loading. ively). The input data are in the public domain, as
Therefore, undrained shear strength (su ) of sands and per the European INSPIRE (2018) directive for spa
transitional soils is an important geotechnical param tial information.
eter that can be used (i) as direct input in calculation
models for fully undrained modelling, and (ii) for
defining a reference for normalisation of cyclic soil 2 DATABASE
parameter values.
This study presents a practical framework for 2.1 Geological setting
developing a site-specific CPT-based method for su in
sands and transitional soils, using results of laboratory The HKW site and the HKN site are located in the
undrained triaxial compression (CU) tests on reconsti southern North Sea. Water depths are typically
tuted and undisturbed specimens as reference. The between 15 m and 34 m relative to LAT.
methodology includes use of (1) net cone resistance The sites comprise Quaternary deposits with
values (qn , defined as qn ¼ qc þ ð1 - aÞu2 - σv , a predominantly sandy sedimentary profile with
where qc is cone resistance, α is net area ratio, u2 is occasional clay layers associated with internal
pore pressure at the cylindrical extension above the channelling (RVO, 2019; RVO, 2020). Sands are
base of the cone and σv is vertical total stress) nor mainly fine and medium with occasional coarse size
malised to vertical effective stress (σ0v ), (2) in some of the soil units. The sites have been sub
a procedure for pairing of CPT data with CU test ject to evolution throughout the Pleistocene and the
results, and (3) definition of a practical failure criter Holocene. Sediments and processes from these time
ion for deriving su from CU test data. periods dominate the geological framework. Geo
The approach outlined in this paper is particularly logical formations present at the two sites within
useful for application in offshore wind, where the the top 50 m below seafloor include (from older to
economics of wind farm development can dictate younger) Yarmouth Roads, Eem, Naaldwijk and
performing only a single cone penetration test (CPT) Southern Bight. These geological formations show
per wind turbine location. This setting drives devel no evidence of cementation. Figure 2 illustrates
opment of CPT-based methods for key geotechnical a microscopic photograph of a typical sand sample
parameters for foundation design. from the HKW site.
DOI: 10.1201/9781003308829-68
486
recognised that reconstituted specimens may give
lower shear strength than undisturbed specimens
(Hoeg et al., 2000).
The specimens were recompressed to the esti
mated in situ stress conditions, using conventional
back pressures for specimen saturation. No pre
cycling was applied. Recompression conditions were
either isotropic or anisotropic, depending on the esti
mated in situ stress state (K 0 ¼ 1 for isotropic stress
state and K 0 ≠1 for anisotropic stress state, where K0
is coefficient of earth pressure at rest).
Database screening was applied, considering
soil type and laboratory specimen homogeneity.
Soil type was assessed based on sample descrip
tion, review of particle size distribution and
Atterberg limits. Particularly for undisturbed test
specimens of transitional soil, specimens contain
ing interbedded or non-uniform material, distinct
strata/layer changes or gravel were excluded from
Figure 1. Locations of HKW and HKN sites. further analysis because they can adversely affect
undisturbed sample quality and test processing
results for a premise of a homogeneous labora
tory test specimen.
The screened database includes laboratory results
from 33 CU tests on reconstituted soil specimens (26
in sand and 7 in transitional soil) and 5 CU tests on
undisturbed soil specimens in transitional soil. The
specimen test depths ranged from 2 m to 38 m below
seafloor. Table 1 presents classification parameters for
the database used. Figure 3 presents results of two typ
ical triaxial tests from the database, one in sand and
one in transitional soil.
Table 1. Classification parameters.
Parameter Sand Transitional soil
Dr (%) 55-110 35-85

Figure 2. Microscopic photograph of a typical unwashed
FC (%) 1-8 20-80
sand sample from the HKW site.
CC (%) - 3-24
Cu (-) 1.5-3.8 5.3-80
D50 (mm) 0.17-0.35 0.02-0.15
2.2 CPT data quartz content 85-100 84-95
(%)
The available CPT data were acquired according to
particle shape subangular to well subangular to
ISO (2014). The data are available in digital tabular
rounded rounded
format and include piezocone CPTs and seismic
piezocone CPTs performed in both non-drilling Notes: Transitional soil = (very) silty sand, clayey sand,
mode (direct push from seafloor) and drilling mode low plasticity (clayey, sandy) silt; Dr = relative density; FC
(vessel drilling, downhole push) deployment. = fines content; CC = clay content; Cu = coefficient of uni
formity; D50 = particle diameter where 50 % of the dry
mass of soil has a smaller particle diameter
2.3 CU data
The database includes results of CU tests performed It is generally inconsistent and impractical to use
according to ISO 17892-9:2018, using reconstituted peak deviator stress as a criterion for deriving su for
specimens, prepared by moist reconstitution, and dense dilative soils such as many of the ones in the
undisturbed specimens. Reconstituted specimens database used for this study. In dilative specimens,
were prepared based on estimated in situ density. large negative pore pressures develop until the end of
Other specimen density considerations are described the test (to about 20 % axial strain; see blue line in
below (section titled ‘data pairing’). It is generally bottom plot of Figure 3) or until cavitation occurs.
487
– Selection of CPT values for comparison with
the laboratory data from reconstituted soil speci
mens focused on estimation of an equivalent
in situ relative density Dr of the reconstituted
soil specimen based on (1) specimen density
and (2) estimated values for minimum and max
imum (index) dry densities:
where emax is maximum index void ratio, emin is

minimum index void ratio and e is specimen
void ratio. Selection of values for emin and emax
included assessment of laboratory test results
per soil unit, per soil type and site-wide;
– Final selection of Dr involved some engineer
ing judgement, particularly for transitional soil
specimens, since the estimation of Dr inevit
ably involves significant uncertainty, which
increases with increase of percentage fines.
The uncertainty in the selected values for emin
and emax should also be noted, particularly
since there are various test methods commonly
used in the industry that can give significant
Figure 3. Example of typical triaxial test results from the
database (applied back pressures: CIUcBE11 – Sand: 1291
differences, especially for the maximum
kPa; CAUcBE09 – Transitional soil: 687 kPa). (index) dry density (Lunne et al., 2019). The
equivalent value of qn was then back-
calculated based on the following equation by
Kulhawy & Mayne (1990):
Cavitation depends on the back pressure applied to the
triaxial test specimen. Sufficiently high back pressure
should be applied to test specimens that are expected
to exhibit dilative behaviour while shearing. It should
be noted that large negative pore pressures can be sus where Pa is atmospheric pressure;
tained in a laboratory setting with controlled applica – Selection of CPT values for comparison with
tion of (high) back pressure, but are typically not the laboratory data from undisturbed soil speci
observed during cone penetration. All tests of the data mens focused on CPT-borehole proximity, use
base had a back pressure which was at least equal to of CPT data showing the lower qn values and
the hydrostatic pressure at the depth point of each test relatively high values of soil behaviour type
specimen. index Ic , thereby accounting for the expected
Common criteria were reviewed for deriving su bias in selection of the laboratory test speci
from the CU data, i.e. peak deviator stress, peak stress mens, and allowance for small (< 1 m) depth
ratio, peak pore pressure, zero excess pore pressure offsets between nearby CPT and sample bore-
and limiting strain (refer to Brandon et al. (2006) for hole locations.
background information on the various criteria for
interpretation of su ). Peak stress ratio was selected as
the most practically useful and most consistent criter
ion across the database. Therefore, this paper defines 3 CPT-BASED METHOD
su at max , where and are the effective
principal stresses. The approach to estimate continuous profiles of
su involved analysis of the relationship between
derived values of su and qn , normalised to
2.4 Data pairing effective vertical stress This led to a bi-linear
Pairing of CPT data (qn ) with CU test results con relationship that is presented in Figure 4 and
sidered the following: Equation 3:
– Laboratory test data were considered as primary,

because of single data points versus CPT profiling
data;
488
95 % of the values in transitional soil
across the wind farm sites;
– Derived values of su are in good agreement with
the scatter of derived values presented in
Andersen (2015).
4 DISCUSSION AND CONCLUSIONS
The method, in combination with an equivalent

method for clays, enables derivation of continuous su
profiles at any CPT location within a given site. This
is particularly useful for offshore wind farm develop
ments where one or multiple CPTs are performed per
wind turbine location without availability of location-
specific laboratory data. Soil behaviour type (i.e.
sand, transitional soil or clay behaviour) can be dis
tinguished directly from CPT with application of gen
eral or site-specific limits of soil behaviour type
Figure 4. Undrained shear strength derived from in CU indices such as Ic or IB (Robertson, 2016). Figure 5
tests on sand and transitional soil as a function of net cone presents an example profile from the HKW site.
resistance. Equation 3 can be applied with appropriate modi
fications to produce design profiles of characteristic
Comments on Equation 3 are as follows: values of su for use in foundation design calcula
– A best fit relationship based on linear least tions. To this purpose, the modifications would need
squares regression was considered for values of to consider at least the following (ISSMGE, 2021):
qn =σ0v ‡ 124 based on paired sets of data in sand – Calculation model and its specified principles;
(n ¼ 26, R2 = 0.87, S.E. = 0.59); – Limit state and mobilised zone of ground;
– A constant value of su =σ0v for values of – Loading regime and field drainage conditions;
qn =σ0v 5124 considering a mean value for su =σ0v – Transformation uncertainty of derived values to
within this range, the wide scatter and the characteristic values;
absence of a significant trend (R2 = 0.38) – Statistical evaluation accounting for statistical
between su and qn in transitional soil; fitting uncertainties within the given dataset.
– The method is robust and allows for develop
ment of continuous profiles of su in sand and For the horizontal portion of the bi-linear relation
transitional soil based solely on input from CPT ship (i.e. the cut-off value for qn =σ0v 5124, see Equa
data, though it is noted that for transitional soil tion 3), the following particular considerations also
engineering judgement has been applied; apply for selection of characteristic values:
– CPT parameter uncertainty for strongly layered – A probable low value for should be used
soil will be higher than for uniform soil (Peu for slightly overconsolidated soil that would be
chen and Terwindt, 2015). Note also that CPT in the order of magnitude for conventional
results are influenced by uncertainty related to clays;
undrained, partially drained or drained condi – Allowance should be made for overestimation of
tions during cone penetration, particularly in su derived from undisturbed transitional soil spe
transitional soil with drainage conditions influ cimens due to sample disturbance and subsequent
enced by factors such as soil constituents and reduction of water content during reconsolidation
(post-)depositional settings. Any of these condi that can lead to soil phase transformation from
tions may apply (DeJong and Randolph, 2012); contractive to dilative (Andersen, 2015).
– The method covers medium dense to very dense
normally consolidated to slightly overconsoli The method appears robust for two particular sites
dated silty to clean sands and sandy silts; at the North Sea. The two sites include multiple geo
– Derived values of su in sand correspond to values logical units and multiple soil types; further optimisa
for cone factor Nkt ranging between 85 and 176 tion should be feasible by differentiation on the basis
for the range which represents more of geological unit and soil type. Soil type differenti
than 95 % of the values in sand across the ation can consider CPT-based soil behaviour type indi
wind farm sites; ces, with confirmation by index sample data that can
– Derived values of su in transitional soil corres easily be acquired in an offshore laboratory, such as
pond to values for cone factor Nkt ranging particle size distribution and particle shape by image
between 28 and 107 for the range analysis (ISO, 2006). Further differentiation may also
20 � qn =σ0v � 75, which represents more than allow wider application of CPT-based methods.
489
Figure 5. Example profile of su from the HKW site comprising three soil types (i.e. sand, transitional soil and clay). Sup
plementary profiles of qn , Ic and Nkt are also displayed. Soil type is distinguished based on Ic (Ic 52:05: Sand,
2:055Ic 52:6: Transitional soil, Ic 42:6: Clay). Note that in clay a CPT-based correlation, similar to Equation 3, was used.
Various advanced regression algorithms can be triaxial compression test on water saturated soils.
trialled in order for the optimum results to be Geneva: ISO.
obtained in terms of statistical evaluation of datasets, International Society of Soil Mechanics and geotechnical
including opportunities for potentially making better engineering (ISSMGE) – Technical Committee TC304
use of the data by means of automated advanced ‘Engineering Practice of Risk Assessment and Manage
data analytics such as machine learning and big data. ment’. 2021. State-of-the-art review of inherent variabil
ity and uncertainty in geotechnical properties and
models.
Kulhawy, F.H. & Mayne, P.W. 1990. Manual on estimating
REFERENCES soil properties for foundation design. Electric Power
Research Institute (EPRI), Palo Alto, California, 1 vol.
Andersen, K.H. 2015. Cyclic soil parameters for offshore
(EPRI Report; EL-6800).
foundation design. The 3rd McClelland Lecture. Fron
Lunne, Knudsen, S., Blaker, Ø., Vestgården, T., Powell, J.J.
tiers in Offshore Geotechnics III, ISFOG’2015, Meyer
M., Wallace, C.F., Krogh, L., Thomsen, N.V.,
(Ed). Taylor & Francis Group, London. Proc., 5–82.
Yetginer, A.G. & Ghanekar, R.K. 2019. Methods used
Brandon, T.L., Duncan, J.M. & Rose, A.T. 2006. Drained
to determine maximum and minimum dry unit weights
and undrained strength interpretation for low-plasticity
of sand: Is there a need for a new standard?. Canadian
silts. Journal of Geotechnical and Geoenvironmental
Geotechnical Journal 56(4): 536–553.
Engineering 132(2): 250–257.
Peuchen, J. & Terwindt, J. 2015. Measurement uncertainty
Hoeg, K., Dyvik, R. & Sandbækken, G. 2000. Strength of
of offshore cone penetration tests. Frontiers in Offshore
undisturbed versus reconstituted silt and silty sand
Geotechnics III, ISFOG’2015, Meyer (Ed). Taylor &
specimens. Journal of Geotechnical & Geoenvironmen
Francis Group, London. Proc., 1209–1214.
tal Engineering 126 (7): 606–617.
Robertson, P.K. 2016. Cone penetration test (CPT)-based
INSPIRE Infrastructure for Spatial Information in Europe.
soil behaviour type (SBT) classification system – an
2018. Available from https://inspire.ec.europa.eu/.
update. Canadian Geotechnical Journal 53: 1910–1927.
International Organization for Standardization. 2006. ISO
Published at www.nrcresearchpress.com/cgj on
13322-2:2006 Particle size analysis - image analysis
14 July 2016.
methods - part 2: dynamic image analysis methods.
RVO Netherlands Enterprise Agency. 2019. Report - Geo
Geneva: ISO.
logical Ground Model HKN – Fugro. Available at
https://offshorewind.rvo.nl/file/view/55040046/Report
19901-8:2014 Petroleum and Natural Gas Industries –
+-+Geological+Ground+Model+HKN+-+Fugro.
Specific Requirements for Offshore Structures – Part 8:
RVO Netherlands Enterprise Agency. 2020. Report - Geo
Marine soil investigations. Geneva: ISO.
logical Ground Model HKW – Fugro. Available at
https://offshorewind.rvo.nl/file/view/55040628/Report
17892-9:2018 Geotechnical Investigation and Testing
+-+Geological+Ground+Model+HKW+-+Fugro.
Laboratory testing of Soil - Part 9: Consolidated
490
© 2022 the Author(s), ISBN 978-1-032-31259-0
Comparison of frozen soil strength characteristics by cone penetration

and triaxial compression testing
Daniil Lagosha
Lomonosov Moscow State University, Russian Federation
Ivan Sokolov & N.G. Volkov

Fugro, Russian Federation
ABSTRACT: The paper provides comparison of field and laboratory tests of permafrost soil. Electrical
resistivity and temperature piezocone penetration tests (RTCPTu) were carried out in the area of sporadic dis
tribution of permafrost. RTCPTu detected islands of permafrost at depth more than 10 meters. Based on this
information, undisturbed samples of warm permafrost soils with massive cryogenic structure and thick ice
lenses (up to 10 cm) were collected using CPT equipment with a direct push soil sampler (MOSTAP) mounted
instead of cone. The samples were kept in frozen condition and transported to the soil lab, where physical
properties and triaxial compression tests were carried out. The CPT results included the characteristics of
long-term frozen soil strength such as σc – the long-term cone resistance and σn – the net long-term cone
resistance. The results of triaxial tests provided the characteristics of peak strength: σ1 – the maximum vertical
stress and σ1-σ3 – deviator stress at failure. The frozen soil samples were cleared from ice lenses since it is
not possible to test them with thick ice lenses. The results of CPT and triaxial testing showed fairly similar
values, when the cone measured resistance in frozen soil with massive cryogenic structure. When the cone
was located near ice lenses, the net long-term cone resistance appeared to be much lower compared to devi
ator stress. This observation is in line with conventional theory on physics of ice which states that ice has
near-zero long-term strength. The obtained results clearly show the value and advantage of CPT in permafrost
(in-situ testing) compared with triaxial compression testing (laboratory). The properties of frozen soil in-situ
may significantly differ from properties of frozen soil in a sample. This must be considered for designing civil
structures on permafrost.
1 INTRODUCTION place. The surface relief is flat due to urban con

struction. Site soils are represented by clays (from
Conventional geotechnical drilling and sampling of sandy to silty) and sands, rarely peat. In some
permafrost is a challenging problem due its vulner areas, there are permafrost soils from a depth of
ability. Frozen soils with high ice content at 9-15 meters.
a temperature near freezing point may be easily dis The roof of permafrost soils is marked at dif
turbed due to frictional heating, drilling fluid, ice ferent depths under the structures on this site.
brittleness and other factors. CPT on permafrost is Moreover, there is the complete absence of
a maturing technology which has been successfully frozen rocks on some parts of the site. So, there
applied recently. A great benefit would be to use are uneven settlements of structures, which lead
CPT equipment together with direct push soil sam to theirs deformation and formation of cracks
pling which is commonly used for non-frozen soils. in them, causing the emergency condition of
This combination of CPT and frozen soil sampling structures.
was successfully tested in the city of Novy Urengoy, The site was located in the basements of two resi
West Siberia, Russia in 2019. dential buildings (Figures 1, 2). The height of the
The site is located in Novy Urengoy, Russia. The basements is less than 2 meters. There are many con
soils of the site were investigated on a depth of tinuous sewer and heating pipes inside, complicating
more than 25 m using field and laboratory tests. the tests. Thus, site conditions did not permit the
Retrospectively the site is located in the northern using of a heavy vehicle-based cone penetration unit
part of the West Siberian plain in the river basin, or other commonly used field methods using large-
where accumulation of lake-alluvial sediments took sized equipment.
DOI: 10.1201/9781003308829-69
491
Figure 1. Map location area – Novy Urengoy, Russia.
Figure 3. Map location area – Novy Urengoy, Russia.
- for the friction sleeve resistance: can be fsv ≈ ≈<

fss and fsv > fss, but fsv and fss are always more
than σs.
Cone penetration tests in velocity mode were exe
Figure 2. Site view. cuted at 2 cm/s for maximum depths up to 30 m,
including tests in cone stabilization mode. They
were performed to a depth of 25 m below the ground
surface. Temperature and stress relaxation measure
2 FIELD TESTS OF PERMAFROST - CPT ments were mostly made at the depth of permafrost
spreading: from 7-9 m to 25 m with intervals of
2.1 CPT performing 1 m at 9-11 m and 2-5 m at 11-25 m. Frozen soil
To perform cone penetration tests (CPT) with meas temperature measurements confirmed the presence
uring of conductivity, temperature and pore pressure of frozen soils at the site. The tests were carried out
RTCPTu, a special small-sized unit with a nominal in July-August 2019. Measured temperatures in the
penetration force of 150 kN was used (Figure 3). frozen soils ranged from -0.9 to -0.1°C.
CPT was carried out in two modes: continuous
(a constant rate-controlled penetration) and “with 2.2 MOSTAP soil sampling
stabilization” of the cone (similar to “dissipation”
test but rods are kept clamped). The second mode Soil sampling was performed with a MOSTAP direct
when qc is measured is also called stress relaxation push soil sampler (Robertson, 2014) from the CPT
test, which carried out in a frost layer of the soil to unit. A total of 13 soil sampling points were exe
investigate the relaxation behaviour of permafrost. cuted on the site. Sampling was carried out very
The stress relaxation test allows to obtain the soil close to the CPT points, at a distance of 1-2 m, for
strength parameters similar to those of the continu reliable correlation of the results of testing and sam
ous type test, but instead of the values of first type pling and minimal influence of soil variability factor.
test, stabilized second type test values of the cone In total, 146 samples were taken, of which 110 of
and at the friction sleeve resistance (qcs, MPa and fss, them were unfrozen and 36 of them were frozen
MPa) are obtained. Based on the results of the stress (Figure 4). It is worth to emphasize that the sampling
relaxation test, it is calculated the long-term strength of frozen soils using the MOSTAP technology was
of the soil under the cone (σc, MPa) and on the fric carried out for the first time.
tion sleeve (σs, MPa). In moving of a cone, the soil In this paper, frozen clay was considered as the
is loaded by a value more than the instantaneous object of researching. It was chosen due to the fact
strength. When a cone stops moving, it is fixed and that it was the cause of uneven sedimentation at the
the stress relaxation is recorded. The values are site. It was also of scientific interest to research the
recorded until the readout parameters stabilize. strength characteristics of the frozen soil areas with
Correlation of the listed parameters for frozen thick ice lenses.
soils is as follows: Considering frozen soil layer was investigated
well by both sampling and cone penetration testing.
- for the cone resistance: qcv > qcs > σc is always The temperature of the investigated soil varies from
observed, and qcv = qcs (0 sec) at the same CPT -0.3 to -0.11°С. It is deposited from a depth
point and at the same depth; of 10 m. Sampling was carried out at a depth of
492
10-15 m. 11 samples were taken. Within this layer,
16 points of CPT were executed in the velocity mode
and 45 stress relaxation tests in the stabilization
mode of the cone. Out of this number of tests, values
for analysis were selected from 8 points. All the
CPT points with stress relaxation tests were located
very close to the frozen soil sampling points - at
a distance of 1-2 m.
Figure 4. Part of investigated frozen soil, sampled by

MOSTAP.
3 LABORATORY TESTS OF PERMAFROST –

TRIAXIAL COMPRESSION TESTING
3.1 Specimens preparing

Various characteristics of the composition and proper
ties of investigated soils were determined in the field
laboratory. According to their results, the soil can be
characterized as frozen clay from silty to sandy, with
ice from well-bonded, no excess ice to thick ice inclu
sions, non-saline, from very stiff to soft.
To determine the strength characteristics of the
Figure 5. Example of a test soil specimen in cross-section
investigated soil by triaxial test, specimens were and longitudinal section.
made of selected samples of frozen clay. Soil sam
ples with a massive cryogenic structure were tested
in a triaxial compression apparatus to ensure that the
homogeneity condition of the specimen was met. previously reached and the drainage blocked. Kine
Specimen preparation for the test was carried out matic test mode was used: vertical loading is applied
in a special room with the maintenance of negative with a given constant rate of deformation of the speci
temperature (cryogenic chamber). The specimens cut men. The rate of specimen loading was 0.02 mm/
from the frozen samples had a cylinder shape with minute.
a height (h) to diameter (d) ratio of 2:1 (70 mm: Soil temperature during triaxial tests is -1°C. This
35 mm) (Figure 5). is the highest subzero temperature that can be main
tained in the apparatus to keep specimens in frozen
state. In the natural soil conditions, their temperature
3.2 Performing tests and theirs results is close to the freezing start temperature and to zero
The chamber and the counter-pressure system of the from -0.3 to -0.1°C, which is technically extremely
triaxial apparatus chamber supplying the fluid to the difficult to assign in the device today. The tests lasted
stamps, and the holes in the stamps were filled with from 11 to 21 hours. The consolidation time of the
a special liquid with a negative freezing temperature samples was 4 hours. The confining stress in the tri
by raising the pressure in the counter-pressure axial device was within 280-300 kPa.
system with complete displacement of air bubbles. Of the executed five tests, it can be highlighted
The triaxial tests were executed in the consolidated- two “paired” tests, i.e. performed on specimens
undrained mode without pore pressure measure taken from one sampling point and one depth inter
ments. The investigated soil was in the frozen state, val, and one “unpaired” test. Therefore, all speci
so it was not possible to measure pore pressure. mens were aged for 1 day and tested under identical
In this case to determine the strength of the soil in consolidation and loading conditions, so the “paired”
full stresses, it was measured by the peak of the stress tests could be compared to each other.
deviator. The specimen is broken by applying Figures 6 and 7 show the results of triaxial tests.
a vertical loading with the pressure in the chamber Specimens No. 2-5 exhibited a brittle behaviour,
493
specimen No. 1 exhibited a ductile behaviour 4 COMPARISON CPT AND TRIAXIAL TESTS
(Figure 7). “Paired” specimens No. 1 and No. 2, RESULTS
No. 3 and No. 4 showed quite close peak strength to
each other: 697 and 679 kPa; 989 and 935 kPa, 4.1 Mechanism of tests
respectively. The results are repeatable and confirm
the validity of the testing procedure. The triaxial tests simulated natural soil conditions
similar to the stress relaxation tests performed by
cone penetration testing.
The mechanism of these tests is quite similar. In
the triaxial compression device, the stamp transmits
vertical stress to the specimen, and the fluid filling
the chamber space makes all-round pressure on the
specimen, allowing the stress-strain state of the
tested soil in the natural conditions to be remake as
much as possible (Figure 8b).
When executing the stress relaxation test by CPT,
the cone is penetrated directly into the soil to the depth
of the investigation, stopping at that depth. As a result,
the soil mass makes all-round compression of the
investigated soil layer from all sides, and the cone,
similarly to the stamp in the triaxial apparatus, assigns
Figure 6. The graph of deviator stress dependence on ver
vertical stress at a constant rate of deformation. In add
tical axial deformation. ition, CPT is sensitive to transient zones of the soil
mass (changes in soil type) and soil variability, such as
ice lenses, due to the influence zone of the cone
It should be noted that fracturing time of specimens (Figure 8a).
No. 1 and No. 2 was quite close - 15 and 13 hours,
respectively. At the same time, specimens No. 3 and
No. 4 had two times more different fracturing times - 7
and 16 hours. However, in both cases the values of
stress peak deviator in “paired” specimens are quite
close.
The vertical stresses σ1 and deviators at fracture of
specimens σ1 - σ3 obviously have the same patterns.
The values of σ1 and σ1 - σ3 range from 0.96 to 1.28
MPa and from 0.68 to 0.94 MPa, respectively.
Figure 8. Mechanism of the main stresses in the test

method of: a) CPT and zone of influence of the cone; b) tri
axial compression.
The principle of the major stresses in both

test methods is similar, but it is necessary to
consider factors that will reflect the difference
between the results obtained by them. These fac
tors are:
• ground temperature;
• geometric shape of the stamp/cone;
• soil variability;
• stress-strain state of the soil.
As a result of the performed tests and their process
ing the following strength characteristics were
Figure 7. A frozen soil sample after the triaxial test and its
failure character.
obtained:
494
- by CPT method:
● cone resistance qc;
● long-term cone resistance σc;
● net long-term cone resistance σn;
- by triaxial compression method:
● vertical stress at specimen failure σ1;
● stress deviator at specimen failure (peak soil
strength) σ1 - σ3.
The cone resistance of the soil qc in this paper is
taken equal to the corrected cone resistance values
where a - base area coefficient of the cone,

u2 - pore pressure (Lunne et al., 1997),
i.e. pore pressure in frozen soils is not considered. Figure 9. The graph of long-term cone resistance and verti
Consequently, qt = qc. cal stress at triaxial test failure of specimen dependence on
From the values of qc obtained from direct meas depth.
urements, it was obtained the values of the long-term
cone resistance σc. For this purpose, it was used the
method of calculating this parameter based on the
dynamometric method of Vyalov (1986).
This method of measuring stress relaxation helps in
evaluating the long-term strength of the soils which
serve as the basement of the structure. Processing
application of stress relaxation curves obtained in the
field tests with the help of the described equation
makes it possible to obtain values of long-term
strength of soil at a given depth. The validity of appli
cation of this equation is confirmed by researches of
Vyalov (1986), Volkov and Sokolov (2018; 2019;
2020).
4.2 Comparison of the results

Based on the obtained values of the long-term cone
resistance σc, the values of the net long-term cone
resistance of the soil σn were obtained by:
Figure 10. The graph of net long-term cone resistance and

stress deviator at triaxial test failure of specimen depend
ence on depth.
where σvo - overburden stress (Lunne et al., 1997).
To compare the results of the strength characteris
tics investigation of the investigated soil, the values
3. The values of σn change from 30 to 920 kPa.
of σc and σ1 and as well σn and σ1 – σ3 were taken
A higher net long-term cone resistance with
and their dependence on the depth of the investiga
increasing depth is also noted.
tion was plotted (Figures 9, 10).
4. Minimum values of σn (30 kPa) are close to zero.
As a result of the analysis it should be paid atten
This indicates the presence of ice lenses in the
tion to the following:
zone of cone influence. Such frozen soils are
1. The pattern of σ1 and σ1 - σ3 values growth with characterized by low bearing capacity during
depth is clearly traced, similar to the pattern of long-term loading application.
the long-term cone resistance of the soil. 5. The maximum values of σn (0.56-0.92 MPa) and σc
2. The values of σc obtained at the depth of investi (0.82-1.15 MPa) obtained by CPT correlate well
gated soils vary from 0.25 to 1.15 MPa. It should with stress deviatiors (0.7-1.0 MPa) and vertical
be noted the increase of long-term cone resist stress (0.96-1.3 MPa) in failure of specimens tested
ance of the soil with increasing the depth of by triaxial compression. These values were
investigation. observed in frozen soils with massive cryogenic
495
structure, which was confirmed by sampling. In over time when the load is applied will tend to small
addition, frozen soil samples were tested in the tri values, close to zero.
axial apparatus with exactly this cryogenic structure
to meet the condition of sample homogeneity.
6. The temperature of the triaxial tests is -1℃. This REFERENCES
is the highest subzero temperature that can be
maintained for high-temperature frozen speci Lunne T., Robertson P.K., Powell J.J. 1997. Cone penetra
mens in triaxial testing. In the natural soil condi tion testing in geotechnical practice. Publishing house of
tions, their temperature is close to the freezing the Spon Press, London and New York.
start temperature and to zero - from -0.3 to Robertson, P.K., and Cabal, K.L. 2014. Guide to Cone
-0.1℃, which is technically extremely difficult to Penetration Testing for Geotechnical Engineering, 6th
Edition, Signal Hill, California: Gregg Drilling & Test
create in the device today. As it is known, when
ing, Inc.
the temperature of frozen soil decreases, its Sokolov I. 2020. Determination method for strength prop
strength increases. Perhaps that is why the values erties of frozen soils by cone penetration testing. PhD
of the results of triaxial tests are correlated with Thesis – Moscow State University, Moscow (in
the maximum values of CPT. Russian).
Volkov, N., Sokolov, I. & Jewell, R. 2018. CPT Testing in
Thus, the results of cone penetration and triaxial Permafrost. Proceedings 4th International Symposium
testing show fairly similar results in the case of on Cone Penetration Testing / N. Volkov [and etc.] // –
a massive cryogenic structure of soil. Determining CPT’18. – Netherlands, Delft – 2018. – рр. 1258–1268.
Volkov N.G., Sokolov I.S., 2019. Estimation of pile bearing
the strength in frozen soil with lensed cryogenic capacity in permafrost based on stress relaxation meas
structure is currently a important question. There is ured by cone penetration testing. Geotechnics, Vol. XI,
not yet a consistent method for such determination. No. 1, pp. 68–78.
In this case, CPT revealed that in frozen high- Vyalov, S.S. 1986. Rheological Fundamentals of Soil
temperature ice-rich clays, the long-term strength Mechanics, Volume 36, 1st Edition. Publisher: Elsevier.
496
© 2022 the Author(s), ISBN 978-1-032-31259-0
Fincone: A study on the use of CPT in soft sensitive clays
T. Länsivaara
B. Di Buò
Tampere University/Ramboll Finland Oy, Tampere, Finland
J. Selänpää
Tampere University/Destia, Tampere, Finland
M. Knuuti & M. Haikola

ABSTRACT: The paper presents the results of a recent study on the application of CPTU on soft sensi
tive Finnish clays. An extensive field and laboratory investigation program was carried out, including 9
test sites of which 5 were studied in more detail. In each of the test sites, a minimum of 4 CPTU sound
ings, 2-3 field vane test with a new type of down hole vane, sampling with a newly developed large diam
eter tube sampler and an extensive laboratory investigation program were carried out. Transformation
models were developed to estimate the undrained shear strength, preconsolidation stress, and constrained
modulus for the over consolidated region. In addition, new information about anisotropy of Finnish clays
were obtained. In general, the CPTU tests proved to be very reliable with very good repeatability. How
ever, the measurement of sleeve friction proved to be somewhat problematic for the very soft and sensitive
clays studied.
1 INTRODUCTION 2 TESTING PROGRAMME
Until 2010, the use of CPTU had not gained wide The CPTU tests included tests with two different
popularity in Finland. Attempts made in the 90’s to probes, i.e., a sensitive cone and a high-capacity cone
assess the undrained shear strength of Finnish soft with a maximum cone resistance of 7.5 MPa and 75
clays with the CPTU had not been very successful MPa respectively. Both cone types have a standard
mainly due to accuracy problems related to the low 60° apex tip, with a cross-sectional area (Ac) of
undrained shear strength (su)values for both the CPTU 10 cm2 and a sleeve area (As) of 150 cm2. The filter
and the field vane tests used as reference. As the prob element for measuring the pore-water pressure is
lems with the field vane test became clearer (Mansik located at the shoulder, above the cone tip (u2).
kamäki, 2015), a need to find better solutions became In addition, seismic and resistivity modules were
obvious. used. A minimum of 4 soundings were performed at
Based on this background Research Centre Terra each site. In addition to CPTU tests, field vane tests
at Tampere University started together with the were carried out with a new type of down hole vane,
Finnish Transport Agency a comprehensive study characterized by measuring system and torque motor
to establish reliable correlations to estimate su and located right above the vane, to overcome problems
preconsolidation stress (σ’p) of soft clays, and to related to rod friction (Selänpää et al. 2018). Com
promote the use of CPTU in Finland. To achieve parative test with standard type of up-hole devices
these goals, an extensive field and laboratory inves were carried out at some of the sites.
tigation program was carried out at 9 test sites. This To ensure high quality undisturbed samples for the
paper gives an overview of the project, describing laboratory tests, a new sampler was developed (Di
the test sites, field, and laboratory work executed, Buò et al. 2019). The sample resembles the SGI type
and the correlations and other results found during of Laval sampler, with a cutting wire and possibility to
the study. feed air/water to avoid suction during withdrawing.
DOI: 10.1201/9781003308829-70
497
Figure 1. Index properties and undrained shear strength for Paimio (upper figure) and Sipoo (lower figure) test sites.
The sample is though somewhat smaller is size, having crust, followed by an 8 m thick, soft clay layer. The
an internal diameter of 131 mm. Another difference is groundwater table is located at 1.2 m depth.
that the samples are stored in the sampling tubes that The Paimio test site is also located close to the
can be pressurized to keep the in-situ stress conditions. city of Turku. The stratigraphy consists of a 2 m thick
The laboratory tests included classification tests, clay crust overlaying an 8 m thick, soft clay layer.
CRS oedometer tests, triaxial compression and ex The groundwater table is located at 0.8 m depth.
tension tests and direct simple shear (DSS) tests. All The Sipoo site is situated 30 km north of the city
tests have been conducted in accordance with appro of Helsinki. The deposit consists of a homogeneous
priate standards when available. soft clay layer between 2 and 9 m depth and a
2 m thick dry crust layer. The groundwater table is
located at 1 m depth.
3 TEST SITES The Lempäälä test site is located close to the city of
Tampere. The soil stratigraphy consists of a 1‒
Field and laboratory investigations have been carried 1.5 m thick dry crust layer, followed by 1‒1.5 m of
out altogether in 9 different sites, i.e., in Perniö, organic soil underlain by a soft sensitive low-plastic
Masku, Paimio, Sipoo, Lempäälä, Murro, Kotka, clay layer. The groundwater table is located at
Joensuu and Pohja. The five foremost have been the 0.6 m depth.
main testing sites in calibration of cone factors and All sites are characterized by low su values and high
are presented here in more detail. water content (w), usually above the liquid limit (LL).
The Perniö test site is located on the southwestern The clays are thus very sensitive and their remolded
coast of Finland, about 140 km west of the city of shear strength values obtained by the fall cone test are
Helsinki. The soil stratigraphy of the deposit generally below 0.5 kPa for Perniö, Paimio and Lem
includes a 1‒1.5 m thick dry crust underlain by 8‒ päälä sites, and close to unity for Sipoo and Masku
9 m thick, soft clay layer; silt and stiff sandy layers sites.
can be found at a greater depth. The groundwater The index properties and the intact (black) and
table is located at 1 m depth. remolded (white) undrained shear strength obtained
The Masku test site is located at the southwestern from the fall cone test, of the clays for Paimio repre
coast of Finland, near the city of Turku. The soil senting highest sensitivity (St) and Sipoo represent
stratigraphy includes a 1.5 m thick weathered clay ing lowest St are presented in Figure 1. Typically for
498
the clays with lower plasticity index PI, the w is The new down-hole field vane device showed
much higher than the LL, resulting in high liquidity better consistency and repeatability than traditional
index (LI), which correlates strongly with St. up-hole devices. The COV values for the down
hole field vane test varied in between 0.04 and 0.23
between different sites, with an average of 0.14 for
4 QUALITY OF DATA 88 tests. When the tests are performed with care,
and casing is used for the up-hole device, the two
In general, the repeatability of the CPTU tests was methods gave very similar results for intact clay,
found very good. As an example, the coefficient of although there was a slight tendency to higher
variation (COV) values for corrected cone tip resist values for the down-hole devise. However, for the
ances and pore pressure measurements varied between remolded undrained shear strength the down-hole
0.026 – 0.060 and 0.023 – 0.106 respectively between device clearly yielded lower values, closer to
different sites (Knuuti & Länsivaara 2019 a, b). How values obtained with the fall-cone.
ever, for the sleeve friction measurements, the reso The quality of the undisturbed soil samples was
lution was too low for the very soft sensitive clay mainly evaluated based on the criteria proposed by
sites, in which the measured values generally were Lunne et al. (1997). Accordingly, sample quality is
below 5 kPa (COV values in the range 0.15 -0.30). In classified as “very good to excellent”, “good to fair”,
Figure 2 results from four tests carried out using the “poor”, and “very poor”. In Figure 3 the sample
sensitive cone and one with the high-capacity cone quality found for CRS oedometer tests are presented
are presented from Paimio test site. As can be seen, as change of void ratio versus depth. The same sam
the four corrected tip resistance measurements are pler types are indicated by equal marker shapes
very similar, while the one using the high-capacity while different shades are used for the various sites.
cone is giving somewhat lower values. This observa As can be seen, most of the samples fall into cat
tion was generally made in all test sites. It should be egories “very good to excellent” and “good to fair”.
noted that the high-capacity cone had not been cali The poor results for the TUT sampler at Lempäälä
brated especially to low values. The too low reso site is explained by it being the first site where the
lution is clearly visible on the sleeve friction graph. new sampler was tested, and all procedures were not
While the pore pressure measurements for the two in order at the beginning. However, as the proced
types of cones were performed with same kind on ures were optimized mostly “very good to excellent”
transducers having the same accuracy, the minor dif quality samples were obtained. Results from labora
ferences that can be seen between the different pore tory test obtained from samples classified as poor or
pressure measurements are related to the preparation very poor have not been used in the calibrations. As
works and actual measurements, not to the equipment. reported by Di Buò et al. (2019), the sample quality
did not suffer from a storage time of two years.
Figure 2. CPTU results from Paimio test site. Results from

four sensitive cone measurements are presented with solid Figure 3. Sample quality according to the Lunne criteria
lines, while the results from one high-capacity cone meas (1997), 1 – very good to excellent, 2 – good to fair, 3 –
urement are given by a black dotted line. poor, 4 -very poor.
499
5 TRANSFORMATION MODELS It could further be noted, that for Equation (1), the
low sensitivity clays indicated generally a lower
5.1 Evaluation based on SCE-CSSM multiplier than 0.28, while for high sensitivity clays
a higher multiplier were generally found. All clays
Di Buò et al. (2020) and Di Buò (2020) studied the had an OCR generally below 2.
possibilities of the hybrid spherical cavity expansion –
critical state soil mechanics (SCE-CSSM) framework
by Mayne (1991) and Chen and Mayne (1994), and 5.2 Evaluation based on index properties
the modified SCE-CSSM solution (Agaiby 2018) for After discussing varies theories Selänpää (2021)
the determination of σ’p /OCR. It can be concluded studied the influence of index properties and pore
that for low sensitivity (St) clays both solutions gave pressure ratio (Bq) to the cone factors. The main
a relatively good match, but as St increased the modi goal was to determine the best practical transform
fied solutions preformed much better. Another general ation models for determination of undrained shear
finding of the study is that the friction angel values strength. Cone factors were determined in relation
needed in the solutions can be seen more as curve fit to different shearing modes (tests) for undrained
ting values rather than true values. To get good correl (triaxial) compression (su comp), undrained (tri
ation to laboratory based σ’p values, the used friction axial) extension (su ext), direct simple shear (DSS)
angle values needed to be clearly higher than those (test) (su DSS), measured field vane (su FVmeas) and
observed in the laboratory. corrected field vane (su FVcorr) undrained shear
Based on the above theoretical framework Di Buò strength. Field vane measurements were corrected
et al. (2020) and Di Buò (2020) proposed simplified as a function of liquid limit according to Helene
transformation models based on average operational lund (1977). In general, it can be concluded that
values, calibrated those based on the entire data, and best correlations were found for su comp as the tri
finally suggested five equations for evaluating σ’p in axial compression test revealed to be the most reli
Finnish clays. Out of those five the following two able test with lowest scatter. For the su FVcorr
are herein suggested as the primary transformation Selänpää suggested that the Nkt and NΔu cone fac
models: tors would depend on PI and the Nke cone factor
on Bq. The two previous proposed su FV corr equa
tions are as follows:
The outcome of Eq. (1) and (2) are presented in

Figure 4 in comparison to CRS oedometer based
σ’p values for Paimio and Sipoo test sites.
Figure 5. Values of su evaluated using Equations (3) and (4)

and (1) and (2) applying the SHANSEP approach in com
Figure 4. Values of σ’p evaluated using Equations (1) and (2) parison to field vane and DSS test results for Paimio (left)
in comparison to CRS test results. and Sipoo (right) test sites.
500
practise never plotted on the zone for sensitive
clays. The SBT charts for Paimio site, with a high
sensitivity of the clay (St = 60…100%) is pre
sented in Figure 6 (Di Buò, 2020).
In Figure 5 the results of Equations (3) and (4)
are compared with su FVcorr and su DSS data for
Paimio and Sipoo test sites using average PI values. 6.2 Anisotropy of su
To compare the different approaches by Di Buò The anisotropy of su for Finnish clays was studied sys
(2020) and Selänpää (2021), values determined tematically for the first time. The anisotropy ratios
using Equations (1) and (2) applying the SHANSEP were determined by comparing the peak undrained
to estimate su is applied. Accordingly, su can be strength values of respective tests corresponding to the
determined from: different shear modes, irrespective of the shear strain
values (Selänpää 2021). The ratio su ext/su comp varied
between 0.52-0.70, with a slight tendency to increase
with plasticity and/or water content. The average value
for the ratio is 0.615. Similarly, the ratio su DSS/su comp
For the parameters S and m values reported varied between 0.56-0.72, with a slight tendency to
by D’Ignazio et al. (2016) with S = 0.244 and increase with water content. The average value for the
m = 0.763, are used. ratio is 0.64. What is notable is the very small differ
For the Sipoo test site, only one good quality ence between su ext and su DSS values.
field vane test was achieved, so the comparison is
primarily to DSS test results. As can be seen from 6.3 Constrained modulus for OC region
Figure 5, all equations performed quite well for the
Paimio site. For the Sipoo test sites, the estimation Di Buò et al. (2018) and Di Buò (2020) evaluated dif
is also good, although some minor underprediction ferent approaches to determine the constrained modu
might be indicated by the su DSS values in the upper lus M0 for the overconsolidated region. As the value
part. for M0 is quite sensitive to sample quality, the correl
ations generally showed a rather high scatter. However,
a simple approach relying on the relatively accurate
6 SOME OTHER FINDINGS determination of σ’p proved to provide a pragmatic
solution giving consistent values. The σ’p for high
quality samples of soft Finnish clays is reached
6.1 SBT charts
approximately at a vertical strain of 4%. Therefore, the
For all test sites the soil behaviour type (SBT) was value of M0 can be evaluated by firstly using trans
evaluated based on the normalized SBT charts by formation models as in Equations (1) and (2) to evalu
Robertson (1990). A general finding was that ate the σ’p, and then divide the σ’p value with 0.04 to
although the clays were in general sensitive, the obtain the M0. As example such an approach for the
classification charts indicated them with very few pore pressure based Equation (2) yielded a COV value
exceptions as clays, rather than sensitive clays. For of 0.21 between CRS and CPTU based M0 values.
the chart based on the normalized friction ratio this
can at least partly be explained by the low accur
acy of the sleeve friction measurements as dis 6.4 Additional modules
cussed before. However, neither the chart based on In all test sites investigations were also conducted
the normalized porewater ratio succeeded in identi using two additional modules connected directly
fying the sensitive clay layer while the values in behind the cone, namely the seismic (S) and resistivity
(R) modules. In general, two tests were performed for
each site using both modules, showing good
repeatability.
The shear wave velocity values varied between
40-110 m/s. The measured electrical conductivity
values ranged from about 50 up to 300 mS/m, correl
ating highly with pore water salinity.
7 CONCLUSIONS AND FUTURE

DEVELOPMENTS
In the FINCONE project the application of CPTU

was studied for soft sensitive Finnish clays. Compre
Figure 6. Soil behavior type (SBT) according to Robertson hensive testing, including high quality sampling,
(1990) for the sensitive Paimio clay (Di Buò, 2020). extensive laboratory program and down hole field
501
vane in addition to the CPTU was performed Di Buò, B., D’Ignazio, M., Selänpää, J., Länsivaara, T. &
altogether on 9 test sites, of which 5 were studied in Mayne, P. W., 2020. Yield stress evaluation of Finnish
more detail. Based on the achieved database trans clays based on analytical piezocone penetration test
formation models were developed to evaluate σ’p, su (CPTU) models. In: Canadian Geotechnical Journal. 57,
and M0. Moreover, information about the anisotropy 11, p. 1623–1638 16 p.
of Finnish clays were achieved for the first time. In Di Buò, B., Selänpää, J., Länsivaara, T., & D’Ignazio, M.
addition, shear wave velocity and resistivity data 2018. Evaluation of existing CPTu-based correlations
for the deformation properties of Finnish soft clays. In
were obtained from the S- and R-modules of the Cone Penetration Testing 2018 (pp. 185–191). CRC
used CPTU equipment. Press.
The CPTU data proved to be very reliable with Di Buò, B., Selänpää, J., Länsivaara, T., and D’Ignazio, M.
good repeatability. However, the sleeve friction 2019. Evaluation of sample quality from different sam
measurements suffered from too low resolution for pling methods in Finnish soft sensitive clays. Canadian
the very low values. Some doubts were also raised Geotechnical Journal, 56(8): 1154–1168.
about their accuracy. The resulting transformation D’Ignazio, M., Phoon, K.K., Tan, S.A., & Länsivaara, T.T.
models proved to work well for studied soft clays. It 2016. Correlations for undrained shear strength of Finn
ish soft clays. Canadian Geotechnical Journal, 53(10),
should though be noted that the clays are rather simi
1628–1645.
lar in nature, with low OCR and generally high Farhadi, M.S, Länsivaara, T. & Tonni, L. 2022 a). Applica
water and clay content. tion of integrated Game Theory-optimization subground
In the future, the aim is to broaden the range of stratification (-IGTOSS) model to Venetian Lagoon
studied soils to silty soils and fine sands and develop deposits. Submitted. CPT22. Bologna.
further transformation models for them. In addition, Farhadi, M.S, Länsivaara, T. L’Heureux, J.S & Lunne, T.
CPTU based soil characterization as well as deter 2022 b). Application of two novel CPTu subground
mination of deformation properties for the normally stratification models. Submitted. CPT22. Bologna.
consolidated will be studied. Some early results of Helenelund, K. V. 1977. Methods for reducing
the former are also presented in this conference (Far undrained shear strength of soft clay. In Swedish
Geotechnical Institute, Proceedings (No. Report
hadi et al. 2022). No. 3 Proceeding).
Knuuti, M. & Länsivaara, T. 2019 a). Variation of Meas
ured CPTu Data. Proceedings of the 7th International
ACKNOWLEDGEMENTS Symposium on Geotechnical Safety and Risk (ISGSR).
Ching, J., Li, D-Q. & Zhang, J. (eds.). Singapore: p.
The guidance of late Professor Rolf Sandven in the 164–169 6 p.
early phases of the project is highly acknowledged Knuuti, M. & Länsivaara, T. 2019 b). Variation of
and remembered with warmth. CPTu-based transformation models for undrained shear
The guidance of Professor Paul Mayne and his strength of Finnish clays. In: Georisk. 13, 4, p. 262–270
warm hospitality during the visiting periods for 9 p.
the second and third author contributed highly to Lunne, T., Berre, T., and Strandvik, S. 1997. Sample dis
turbance effects in soft low plastic Norwegian clay. In
their studies and is very much acknowledged. Symposium on Recent Developments in Soil and Pave
The financial support of the Finnish Transport ment Mechanics.
Agency and Tampere University is highly appreci Mansikkamäki, J. 2015. Effective stress finite element sta
ated in carrying out the study. bility analysis of an old railway embankment on soft
clay. PhD thesis, Department of Civil Engineering, Tam
pere University of Technology, Tampere, Finland.
Mayne, P.W. 1991. Determination of OCR in clays by
REFERENCES piezocone tests using cavity expansion and critical state
concepts. Soils and foundations, 31(2): 65–76.
Agaiby, S. 2018. Advancements in the interpretation of Robertson, P.K. 1990. Soil classification using the cone
seismic piezocone tests in clays and other geomaterials. penetration test. Canadian Geotechnical Journal, 27(1),
Doctoral dissertation, School of Civil & Environmental 151–158.
Engineering, Georgia Institute of Technology, Atlanta. Selänpää, J., Di Buò, B., Haikola, M., Länsivaara, T., &
Chen, B.S., and Mayne, P.W. 1994. Profiling the overcon D’Ignazio, M. 2018. Evaluation of existing CPTu-based
solidation ratio of clays by piezocone tests. Rep. No. correlations for the undrained shear strength of soft
GIT-CEEGEO-94, 1 Finnish clays. In Cone Penetration Testing 2018 (pp.
Di Buò, B. 2020. Evaluation of the Preconsolidation Stress xxx–xxx). CRC Press.
and Deformation Characteristics of Finnish Clays based Selänpää, J. 2021. Derivation of CPTu cone factors for
on Piezocone Testing. PhD Thesis, Tampere University, undrained shear strength and OCR in Finnish clays.
Tampere, Finland. ISBN 978-952-03-1468-2 PhD Thesis, Tampere University, Tampere, Finland.
502
© 2022 the Author(s), ISBN 978-1-032-31259-0
CPT-based unit weight estimation extended to soft organic clays

and peat: An update
H.J. Lengkeek
Delft University of Technology, Delft, The Netherlands
Witteveen+Bos, Deventer, The Netherlands
R.B.J. Brinkgreve
Bentley Systems, Delft, The Netherlands
ABSTRACT: Various CPT-based correlations exist for the unit weight of natural soils. One such correlation
includes organic soils Lengkeek et al. (2018). This correlation is presented as a framework where the coeffi
cients can be optimized and is based on predominantly Class 2 CPT records. This publication uses an
expanded database which includes additional pairs of predominantly Class 1 CPT records selected from Holo
cene deposits in the Netherlands, on mineral clays, organic clays and peats. This results in a more extensive
database and an improved CPT-based unit weight correlation for the whole range of soil types, which is pro
posed to replace the existing correlation. In addition, a specific unit weight correlation for peats is presented.
1 INTRODUCTION organic clays and peats well compared to other soils.

The properties of peats have been investigated and
1.1 Automated processing of CPTs extensively published, i.e. Den Haan and Kruse (2007),
(Mesri and Ajlouni, 2007). However, limited attention
Cone penetration testing (CPT) has become increas
has been devoted to the whole range of slightly organic
ingly popular as the preferred in-situ test method as
clay to peat, and how this relates to CPT measure
it can be used for soil classification, estimation of
ments. These organic soft soils are frequently present
geotechnical parameters and use in empirical
within the Holocene deposits in the Netherlands and in
methods. With the increase of automated processed
other deltaic areas worldwide. Organic soft soils are
CPT data in engineering (Brinkgreve, 2019), it is
characterized by a low unit weight and high compress
critical to have an accurate estimation of soil unit
ibility. Organic soft soils can be identified by a high
weight as this is the first and most important step in
organic content and high CPT friction ratio. In contrast
geotechnical parameter determination. This is par
to other soft soils, the strength is not necessarily low.
ticularly relevant for organic soils which are often
not included in existing CPT-based parameter deter
mination methods. 1.3 Aim of this publication
By combining soil properties obtained from laboratory
1.2 Organic soils testing with CPT results, layer-, site- or region-specific
Organic soils are formed during the decomposition of correlations can be obtained between CPT measure
dead organic substances i.e., remnants of plants and ment data and geotechnical properties of the soil.
animals. This process takes place in different ways, When automating the interpretation of CPTs, it is pref
mainly through bacterial activity, intensified by erable to have a direct and reliable relation between
oxygen and temperature. Another type of sediment the measurements and the soil unit weight. With
with a highly variable organic content are the flood a more reliable in-situ derived estimation of unit
plain sediments, which are deposited when streams at weight, the effect of human interference is limited to
high water overflowed natural embankments. The a minimum. Moreover, because many soil properties
peat areas and deposits of organic soils occur to (and thus the applicable correlations) depend on the
a large extent in the northern parts of the world. stress level, it is paramount to have an indication of the
To date, most published research on CPT application stress profile over the depth. For this purpose, the use
is on mineral soils. Existing CPT-based correlations for of lookup tables such as those found in textbooks is
mineral clays do not capture the behavior of soft not preferable. The aim of this publication is twofold:
DOI: 10.1201/9781003308829-71
503
– To validate and improve the CPT-based unit water.nl). The CPTs are standardized in GEF format.
weight correlation (Lengkeek et al., 2018) for the These standardized formats are very useful and effi
whole range of soils. cient to set up a comprehensive database. An over
– To present additional insight in relations between view of the locations and number of CPT-borehole
index properties and CPT measurements for pairs is presented in Figure 1. The total number of
organic soils. undisturbed samples is 464, the number of CPT pairs
is 233 of which 211 include the unit weight, 136
include organic content and 109 include specific
1.4 Research approach and databases
gravity. The data of this research is available in the
The 2018 database includes the sample unit weight Delft University of Technology repository and pub
and Class 2 CPTs (ISO22476-1, 2012) of Holocene lished in Lengkeek (2022).
and Pleistocene sedimentary deposits in the Nether
lands. This database is used for the initial unit
weight correlation and includes all soil types, how 2 UNIT WEIGHT CORRELATION
ever mainly mineral soils.
The 2021 database follows from soil investiga 2.1 Soil type categories
tions from various dike reinforcement projects
across the Netherlands. This database includes clas The selected classification method for organic fine-
sification laboratory tests and Class 1 CPTUs of grained soils is based on the FHWA system. Sand
mainly Holocene organic clays and peats. The CPT (coarse grained soils) are classified based on the
data is taken from the same depth as the samples, sample identification description. The FHWA classi
with a maximum allowable distance between bore- fication system, based on organic content measured
hole and CPT of 1 meter. These soil investigations by the Loss on ignition (N), consists of the following
are performed in the period 2010-2020. soil categories:
‒ mineral fine-grained soils: N≤3%.
‒ mineral fine-grained soils with organic matter:
3<N≤15%.
‒ organic fine-grained soils: 15<N≤30%.
‒ peats: N>30%.
The classification results for the 2021 database
with organic soils are presented in Table 1. The
names of the soil categories in the graphs are short
ened for practical reasons. For samples where the
organic content is unknown, the classification is
based on the unit weight; Peat: γsat � 12, Organic
clay: 125γsat � 14, Clay with organic matter:
145γsat � 17, Clay, mineral: γsat 417, all in kN/m³.
Table 1. Classification results for organic soil types: aver

age organic content, range of unit weight and specific grav
ity per soil type in the 2021 database.
Results: Nmean γsat Gs
Soil type (%) (kN/m³) (-)

Peat 79 10.1 - 13.1 1.4 - 2.0
Organic Clay 22 11.6 - 14.0 1.9 - 2.4
Clay (org.matter) 8 12.4 - 19.2 2.3 - 2.7
Clay (mineral) 2 15.6 - 20.0 2.6 - 2.7
2.2 Updated CPT-based unit weight correlation

The updated CPT-based unit weight correlation is
Figure 1. Overview of 57 CPT-borehole pair locations in based on the combined database. The 2018 database
the Netherlands. mainly consists of mineral soils whereas the 2021
database mainly consists of organic soils. The com
The Dutch Water Authorities requires that all new bined database allows for a validation and improve
soil investigations be performed according to ment of the correlation for unit weight. The CPT-
a dedicated protocol for dikes, summarized in based unit weight correlation of Lengkeek et al.
a standardized STOWA Excel sheet (www.helpdesk (2018) is shown in Equation (1). The correlation is
504
based on the corrected cone resistance qt and friction on regression (Sy) and the slope of the trendline
ratio Rf, which are both normalized by a reference through the origin [x=measured, y=predicted]. The
value. The reference unit weight, here 19.5 kN/m³, is comparison with other existing correlations Mayne
the value when qt equals qt,ref. The updated variables (2014), (Robertson and Cabal, 2010, Lengkeek et al.,
based on the combined database of 427 pairs are pre 2018) is presented in Table 3. The R² and Sy comply to
sented in Table 2. Ordinary Least Squares (OLS) regression with free
intercept. The slope complies to regression through the
origin and is a measure for the bias of the trend in
Figure 3. From this comparison it can be concluded
that the new correlation performs better for all statis
tical parameters. The 2018 correlation results in
slightly different values which validates the use it.
Herein:
γsat,ref is the reference unit weight at which the cone
resistance is constant regardless of Rf.
qt,ref is the reference cone resistance at which the
unit weight is constant regardless of friction ratio.
Rf,ref is the reference friction ratio at which the apex
of all lines of equal unit weight is located.
β is the fit factor, which is a measure for the inclin

ation of the equal unit weight contours.
Table 2. Updated parameters for unit weight correlation.
Parameter Value Unit
γsat,ref 19.5 kN/m3

qt,ref 9.0 MPa
Rf,ref 20 %
β 2.87
Figure 2 presents all measured data per soil type

combined with the lines of equal unit weight [10,
21] kN/m³. The results are plotted on the SBT tem
plate of Robertson (2010). From this figure it can be Figure 2. Unit weight measurements and lines of equal unit
concluded that the lines of equal unit weight are well weight of the improved correlation, presented on top of
aligned with the orientation of SBT zone boundaries. Robertson (2010) SBT template.
Coarse grained soils, SBT=5 and higher correspond
to a unit weight of 18 to 21 kN/m³. The variation in
unit weight for fine soils is much larger.
Figure 3 shows the measured unit weight versus
the predicted unit weight using the improved correl
ation. The points are subdivided in the database cat
egories [Peat; Organic Clay; Clay with organic
matter; Mineral Clay; Sand]. These database categor
ies are based on the laboratory classification. From
this graph it can be seen that the trend follows the 1:1
line very well. The scatter is larger for lower unit
weights and organic soils; however, for peats the
results are close to 10 kN/m3, which is also the min
imum value as applied.
Figure 4 presents an example of a CPT with clay,
peat and sand layers, including the unit weight
according to Equation 1. The unit weight from the
laboratory tests are respectively 12.6 to 15.0 kN/m³
in the upper 2m clay, 10.3 kN/m³ for the peat layer
and 19.5 kN/m³ for the underlain sand layer.
The performance of the improved correlation can be
expressed in statistical parameters such as the coeffi Figure 3. Measured versus predicted unit weight based on
cient of determination (R²) and the standard deviation the improved correlation.
505
Table 3. Comparison of statistical results of multiple cor Correlations for each soil category would result in
relations for the whole range of soils. a lower coefficient of determination and limit any
reliable correlation to an average value and standard
slope deviation per soil type.
OLS through
slope origin
Method R2 Sy [y:x] [y:x] 3.2 Correlations with organic content
Improved 0.80 1.32 0.84 1.00

Figure 5 presents the organic content versus the
correlation water content and was first published by Mitchell
Lengkeek (2022) and Soga (2005). This correlation provides a first
Lengkeek (2018) 0.79 1.33 0.80 1.00 estimate of the organic content for any soil which is
Robertson & 0.25 1.46 0.26 1.06 expected to be organic. The data shows an increase
Cabal (2010) of organic content with water content up to N=90
Mayne (2014) 0.12 1.68 0.20 1.03 which is considered as a physical upper bound. The
bi-linear fit performs better than the correlation by
Mitchell and Soga (2005), which is based on less
data.
Figure 6 presents the specific gravity versus the
organic content. The results confirm the empirical
relation as published by Den Haan and Kruse
(2007). Once the organic content is known, the spe
cific gravity and ultimately the unit weight can be
estimated.
Figure 4. CPT results and unit weight according to equa

tion 1 for a CPT from Eemdijk, the Netherlands.
3 CPT-BASED CORRELATIONS FOR

ORGANIC SOILS
Figure 5. Organic content versus natural water content, for
3.1 Introduction organic to mineral soils. Best bi-linear fit:
N ¼ min ½90%; 0:239 wnat - 7:08] with standard deviation
The pairwise established database of classification Sy=7.53.
test results and CPT measurements allows for com
parison of properties of organic soils and additional
insight in relations. In this section three graphs with
organic content and index properties are presented Figure 7 presents the unit weight versus the
as well as three graphs with CPT-based organic content. This figure is the basis for the sec
correlations. ondary criteria for classification of organic soils
For each graph the results and the confidence based on the unit weight. The variation is more than
intervals are plotted in the graphs and the statistical that for the specific gravity correlation as the unit
parameters are shown in the title. The subcategories weight is not just a unique soil property but also
are indicated in the legend. The regression is applied a state parameter depending on the preloading and
to all samples as one group and not per soil type. stress level.
506
Where γsat;peat is the saturated unit weight of peat in
(kN/m3) and qt is the corrected cone resistance in
(kN/m2).
Figure 9 presents the specific gravity versus the
friction ratio. Figure 10 presents the organic content
versus the friction ratio. Both correlations confirm
that the unique soil properties are reasonably correl
ated with the friction ratio with a high R². However,
the large variation Sy makes these correlations less
useful in practice.
Figure 6. Specific gravity versus organic content, for

organic to mineral soils. The standard error on regression
(Sy) is 0.082 and coefficient of determination (R2) is 0.97.
Figure 8. Saturated unit weight versus CPT corrected cone

resistance, for soils classified as peat.
Figure 7. Saturated unit weight versus organic content, for

organic to mineral soils. The best power function fit is:
γsat ¼ 20:8N -0:153 with standard deviation Sa=1.48 (for
y=a·xb).
3.3 CPT-based correlations

Figure 8 presents the unit weight of soils that are
classified as peat and the correlation is shown in
Equation (2). This figure illustrates a linear relation
where the range is [10, 12] kN/m³, the R² is moder
ate and the Sy=0.265 kN/m³. This correlation is only
applicable with prior knowledge of the soil type and
cannot be used for organic clays. The accuracy is
Figure 9. Specific gravity versus CPT friction ratio, for
however better than Equation (1).
organic to mineral soils. The subcategories are indicated in
the legend. The best linear fit is: Gs ¼ -0:147Rf þ 2:88
with standard deviation Sy=0.227.
507
The confidence interval and standard deviation are
provided to account for such bias. For final estimates
of soil parameters, it is recommended to combine
these correlations with sampling and testing of site-
specific geological units.
In general, it is highly recommended to perform
CPTs adjacent to boreholes, select pairs of high-quality
laboratory tests according to a standardized protocol
(STOWA). This will allow for new or improved correl
ations which will improve prior estimates.
ACKNOWLEDGEMENTS
The authors would like to thank the POVM, estab

lished in 2015 by Water Authorities in the Nether
lands, who initiated and financed “Eemdijkproef”.
This work is part of the “Perspectief” research
programme All-Risk with project number P15-21,
which is (partly) financed by NWO Domain Applied
Figure 10. Organic content versus CPT friction ratio, for and Engineering Sciences.
organic to mineral soils. The subcategories are indicated in
the legend. The best linear fit is: N ¼ 10:5Rf - 17:6with
standard deviation Sy=17.0. REFERENCES
Brinkgreve, R. B. J. 2019. Automated Model And Param
4 CONCLUSIONS AND RECOMMENDATIONS eter Selection: Incorporating Expert Input into Geotech
nical Analyses. Geo-Strata—Geo Institute of ASCE, 23,
The pairwise established database of classification 38–45.
test results and CPT measurements provides valuable Den Haan, E. J. & Kruse, G. A. M. 2007. Characterisation
insight into the properties of organic soils as well as and engineering properties of Dutch peats. Character
isation and engineering properties of natural soils, 3,
new and updated correlations.
2101–2133.
The existing unit weight correlation is validated ISO22476-1 2012. Geotechnical Investigation and Testing
and improved by the extension of the 2018 database Field Testing - Part 1: Electrical Cone and Piezocone
with organic soils resulting in 427 pairs. The statis Penetration Test. International Organization for
tical parameters of the improved CPT-based unit Standardization.
weight correlation are compared with existing correl Lengkeek, H. J. 2022. CPT-based classification and correl
ations and show better performance. The advantage ations for organic soils. 4TU.ResearchData.
of the improved correlation shown in equation 1 is Lengkeek, H. J., De Greef, J. & Joosten, S. 2018. CPT
that it can be applied for organic soils and mineral based unit weight estimation extended to soft organic
soils and peat. 4th International Symposium on Cone
sedimentary soils. This is useful for SBT classifica
Penetration Testing (CPT’18). Delft.
tions which include stress correction. Specifically, Mayne, P. W. 2014. Interpretation of geotechnical param
for soils which are classified as peat, equation 2 can eters from seismic piezocone tests. Proceedings, 3rd
be used with even higher accuracy. International Symposium on Cone Penetration Testing.
The 2021 database confirms existing relations Mesri, G. & Ajlouni, M. 2007. Engineering Properties of
between the organic content and other index param Fibrous Peats. Journal of Geotechnical and Geoenviron
eters. Furthermore, the organic content and specific mental Engineering, 133, 850–866.
gravity can be correlated to the CPT friction ratio. Mitchell, J. K. & Soga, K. 2005. Fundamentals of Soil
Both correlations confirm that the unique soil proper Behavior 3rd ed., JohnWiley &Sons. Inc. Foundation
failure.
ties are reasonably correlated with the friction ratio.
Robertson, P. K. 2010. Soil behaviour type from the CPT:
The correlations allow for establishing prior esti an update. 2nd international symposium on cone pene
mates where no laboratory tests are available. The tration testing, USA.
disadvantage of this approach is that it increases Robertson, P. K. & Cabal, K. L. 2010. Estimating soil unit
inherent variation along the trend and the possibility weight from CPT. 2nd International symposium on cone
that site specific units are biased to the trendline. penetration testing.
508
© 2022 the Author(s), ISBN 978-1-032-31259-0
CPT-based classification of soft organic clays and peat

H.J. Lengkeek
Witteveen+Bos, Deventer, The Netherlands
R.B.J. Brinkgreve
Bentley Systems, Delft, The Netherlands
ABSTRACT: An updated CPT-based classification system of organic clays and peat is proposed based on
an extensive pairwise established database of classification tests and CPT measurements. This new classifica
tion system is proposed to supplement the existing dimensionless qt/pa-Rf-chart of Robertson (2010). The
Robertson (2010) dimensionless classification system is selected for refinement because it appears to perform
better than normalized systems for peats with very low stresses (<20 kPa). A combination with Robertson
(2009 and 2016) is possible in cases where a stress normalization cut-off is used.
1 INTRODUCTION a Soil Behavior Type (SBT) classification. These

classification systems include pore pressure meas
1.1 Application of CPTs in dike projects urements from CPTU tests and the shear wave vel
ocity from SCPT tests, e.g. Robertson (2016).
To successfully plan, design and construct
Examples of CPT based empirical methods can be
a geotechnical project, various types of investigative
found in the Eurocode (EN1997-1, 2005, EN1997-2,
techniques to obtain sufficient geotechnical informa
2007), where the cone resistance is used for the esti
tion are required. Geotechnical field investigations
mation of soil strength. In addition, there is a wide
generally comprise boreholes with sampling and in-
range of publications on CPT based estimation of
situ cone penetration tests, performed with a friction
geotechnical parameters. A comprehensive overview
cone penetrometer (CPT) or with a piezocone pene
can be found in Kulhawy and Mayne (1990), (Lunne
trometer (CPTU).
et al., 2002, Mayne, 2014).
The use of CPTs in the design of dikes in the
To date, most published research in the field of CPT
Netherlands has increased over the years. On
application is on mineral soils. Existing CPT-based cor
a typical dike project, CPTs are performed typically
relations for mineral clays do not properly capture the
every 100m along the center line and supplemented
behavior of organic clays and peats compared to other
with 3 CPTs and 1 borehole along a cross section
soils. The properties of peats have been investigated
every 200m. The number of CPTs is typically 5
and extensively published, i.e., Den Haan (1997),
times greater than the number of boreholes. This is
(Mesri and Ajlouni, 2007). However, limited attention
due to the relative costs and the increased possibil
has been devoted to the whole range of slightly organic
ities associated with the use of CPTs.
clay to peat, and how this relates to CPT measure
ments. These organic soils are frequently present
1.2 CPTs in organic soils within the Holocene deposits in the Netherlands and in
other deltaic areas worldwide. Organic soils are charac
Cone penetration testing has become increasingly terized by a low unit weight and high compressibility.
popular as the preferred in-situ test method as it can Organic soils can be identified by a high organic con
be used for soil classification, estimation of geotech tent and high CPT friction ratio. In contrast to other
nical parameters and use in empirical methods. The soft soils, the shear strength is not necessarily low.
initial soil texture-based classifications were based
on direct measurement of cone resistance (qc) and
1.3 Aim of this publication
sleeve friction (fs) e.g. Begemann (1965). The cur
rent CPT-based classification systems are based on The aim of this publication is to improve the
behavior characteristics and are often referred to as applicability of CPTs for organic soils. To achieve
DOI: 10.1201/9781003308829-72
509
this, results from soil investigations from dike 2010-2020. Recently, the Dutch Water Authorities
reinforcement projects across the Netherlands required that all new soil investigations be performed
have been collected. CPTs and boreholes that were according to a dedicated protocol for dikes, summar
performed in proximity of each other have been ized in a standardized STOWA Excel sheet (www.help
selected. The laboratory tests results and CPT deskwater.nl). The CPTs are standardized in GEF
measurements were taken at the same level, format. These standardized formats are very useful and
paired, and processed into a regional database. In efficient to set up a comprehensive database. The data
this paper an improvement for CPT based classifi of this research is available in the Delft University of
cation systems for organic soils is proposed. Technology repository and published in Lengkeek
(2022).
1.4 Research approach and databases
This research combines an existing database (Leng 2 CLASSIFICATION OF ORGANIC SOILS
keek et al., 2018) and a new compiled database for
organic soil properties, referred to as the 2021 data 2.1 Laboratory classification
base. The 2018 database includes the sample unit
weight and Class 2 CPTs of Holocene and Pleisto Existing classification systems are based on geomorph
cene sedimentary deposits in the Netherlands. An ology, topography, chemical properties, botanical
overview of the locations and number of CPT origin, genetic processes, or physical characteristics.
borehole pairs is presented in Figure 1. From a geotechnical engineering perspective, the phys
ical characterization is the most relevant. Several clas
sification systems for organic soils are used in various
countries and are based on similar grounds. In many
cases, a certain degree of humification Von Post (1922)
is used for the classification of peat, together with the
normal geotechnical parameters, such as water content,
Atterberg limits, organic content, bulk density etc.
Understanding the stratification and properties in a soil
profile is made easier if the geological history and the
environmental conditions at deposition of the sedi
ments are known.
Examples of classifications for geotechnical
engineering can be found in Landva et al. (1983),
(NEN5104, 1989, Huang et al., 2009, ISO14688-1,
2017, ISO14688-2, 2017, Von Post, 1922). Both the
term ‘organic content’ and ‘ash content’ are used to
identify organic soils. The classification systems
differ, in particular for organic content in the range
[20,50] %, where peats and organic clays overlap.
2.2 CPT-based classification methods

CPT-based classification methods provide two-
dimensional charts for soil type classification based
on the CPT measurements. These charts were devel
oped through direct correlation between the CPT
data and the corresponding soil type determined
from adjacent borings. The initial soil texture-based
classifications were based on direct measurement of
cone resistance and sleeve friction (Begemann,
Figure 1. Overview of 57 CPT-borehole pair locations in 1965, Schmertmann, 1978).
the Netherlands.
Robertson et al. (1986) developed a non-
normalized soil behavior-based classification, ini
The 2021 database includes soil investigations from tially with 12 zones. In Robertson (2010) this is
various dike reinforcement projects across the Nether updated to 9 zones based on dimensionless cone
lands. The 2021 database includes classification labora parameters (qt/pa, Rf) and the non-normalized SBT-
tory tests and Class 1 CPTs of Holocene organic clays index ISBT. Robertson (1990) presented the normal
and peats. The CPT data is taken from the same ized soil behavior classification for 9 zones based on
level as the samples, with a maximum allowable dis the linear normalized cone parameters (Qt1, Fr, Bq).
tance between borehole and CPT of 1 meter. The soil behavior type index Ic1 is added to this in
These soil investigations were performed in the period Robertson and Wride (1998). In (Robertson, 2009,
510
Zhang et al., 2002) the classification system SBTn is tables, stresses are sometimes less than 20 kPa at 10m
adjusted with a variable stress exponent n and non depth. Therefore, care should be taken with CPT-
linear normalized cone resistance Qtn and nonlinear based classifications that include stress normalization,
SBT-index Icn. as illustrated in the following example.
Since 1990, more CPT soil behavior-type charts A peat layer below a dike with a high stress level
have been developed including (Been and Jefferies, of about 100 kPa is originally classified as SBT=2
1993, Eslami and Fellenius, 1997, Schneider et al., (Robertson, 2010) and SBTn=2 (Robertson, 2009),
2008). In Robertson (2016) a modified SBT classifi but the same peat layer beside the dike with a low
cation system is presented with 7 zones and charts stress of 20 kPa moves up to SBTn=3 and will be
based on Qtn versus the small-strain rigidity index IG classified as clay. This second classification is not
and versus the normalized pore pressure U2. Further correct as the soil type is the same, but only the
more, a new hyperbolic shaped modified SBT-index stress state is different. Consequently, the soil profil
IB is introduced. ing beside the dike can be incorrect, and the wrong
Existing CPT based classifications generally relate parameters will be appointed to this layer. In this
to mineral soils which are present worldwide. The example, the normalized cone resistance Qtn is 5
major disadvantage of existing CPT based classifica times higher than Qt. These high stress corrections
tion methods is that the classification of organic soils are not included in the international databases where
is inaccurate. In many cases a peat layer is classified most of the stresses are typically in the range of 50
as clay (SBT=3) instead of organic material (SBT=2). to 300 kPa.
Furthermore, it does not distinguish between peats Particularly for dike projects there is a second
and organic clays. Engineering of dike projects in the argument not to apply a large stress correction. The
Netherlands, where peat is often present, is therefore peat layers beside the dike are generally over-
mostly based on local experience or the non-stress consolidated by an OCR of 2, due to a combination
normalized qt/pa-Rf chart of Robertson (2010). of water level changes and aging. The same peat
CPTUs are generally performed; however, the pore layer below the dike, which has been raised periodic
pressure classification charts are not used due to the ally, is only slightly over-consolidated. As the cone
presence of gas in organic soils, which causes resistance is related to the preconsolidation stress
a reduced and unreliable pore pressure response. more than the vertical effective stress, the actual
stress correction should be about 2 to reflect the state
properties.
2.3 Organic soil type categories The proposed adjustments to the SBT charts, as
The 2021 database includes classification tests will be presented in the next paragraphs, are valid
according to different standards and systems (NEN, for the non-stress normalized SBT chart (Robertson,
EN, ISO).The organic content is measured for most 2010) and the stress-normalized SBT chart (Robert
samples. The fine grained soils are classified accord son, 2009) with the application of the stress normal
ing to one system: FHWA (Huang et al., 2009). The ization cut-off Cn≤2.
FHWA classification system is based on the organic
content measured by the loss on ignition (N) and 3.2 Proposed SBT adjustment
consists of the following soil categories:
This paragraph presents the adjustment to the SBT
– mineral fine-grained soils: N≤3%. classification for organic soils, such as those encoun
– mineral fine-grained soils with organic matter: tered in deltaic areas in the Netherlands. The results
3<N≤15%. from the combined database are plotted on the
– organic fine-grained soils: 15<N≤30%. (Robertson, 2010) template in Figure 2. The soil cat
– peats: N>30%. egories consist of the categories in Table 1, including
For samples where the organic content is one category for sand.
unknown, the classification is based on the unit The data coincides to a large extent with SBT
weight; Peat: γsat � 12, Org.Clay: 125γsat � 14, zones, which is expected for the mineral soils. A few
Clay (org.mat): 145γsat
� 17, Clay (mineral): datapoints coincide with SBT=1 (sensitive soils) and
γsat 417, all in kN/m³.
no points coincide with SBT=7, 8 and 9. Soils of
SBT=7 can be present in Pleistocene sand deposits
and gravelly deposits, which are present along the
3 UPDATED SBT ZONES FOR ORGANIC river Meuse in the South of the Netherlands. Soils of
SOILS SBT=8 and 9 are not expected in a deltaic area up to
15 m depth.
There are major differences in SBT=2 and 3, where
3.1 Stress normalization
a significant amount of organic soils plot in SBT=3.
The samples of the combined database (2018 and The performance results based on the existing Robert
2021) are taken from 0.5 to 15m depth and effective son (2010) classification of organic fine-grained soils
vertical stresses in the range of 5 to 150 kPa. For situ are presented in Table 2. It is concluded that most of
ations with the presence of peat layers and high-water the organic soils, including most of the peats, plot in
511
SBT=3 (Ic≤3.6). The performance is about the same Table 3. Performance results for organic soils based on
for Icn based on stress normalization including Cn≤2. proposed adjustments to Robertson (2010) SBT
Without the Cn cut-off almost all points plot outside of classification.
SBT=2.
SBT zone:
Table 1. CPT results of 2018 and 2021 database. Soil type SBT=2a SBT=2b SBT=2c SBT=3, 4
average range Peat 78% 15% 1% 4%
[N>30]
qt Rf qt Rf Org. Clay 16% 42% 40% 2%
[15<N≤30]
Soil type (MPa) (%) (MPa) (%) Clay (org.mat) 3% 22% 38% 37%
[3<N≤15]
Peat 0.5 7.8 0.1 - 1.8 3.2 - 11.0 Clay (mineral) 0% 11% 3% 86%
[N>30] [N≤3]
Org. Clay 0.4 3.8 0.1 - 0.9 1.0 - 9.6

[15<N≤30]
Clay (org.mat) 0.6 2.5 0.1 - 2.1 0.6 - 6.5 The proposed adjustment is that SBT=2 and
[3<N≤15] part of SBT=3 are redefined and split up into
Clay (mineral) 1.7 2.4 0.2 - 5.1 1.1 - 4.6 SBT=2a (Peat), 2b (Organic Clay) and 2c (Mineral
[N≤3]
Clay, with organic matter). No adjustments are
Sand 10.2 0.9 2.2 - 33.1 0.4 - 1.9
proposed to the boundaries between SBT=3, 4 and
higher. This is also not possible as this database
does not distinguish between silts and clays due to
the lack of Atterberg limits tests. Most of the clas
sified points plot in the correct SBT zone when
using the proposed adjustment, although there is
still some overlap with the adjacent SBT zones.
The selection of the boundaries is determined by
maximizing the group of positives and minimizing
the number of false positives and false negatives.
In addition, the boundaries are selected to separate
over-consolidated organic soils from over-
consolidated plastic clays, such as Pot clay (Pleisto
cene) and Boom clay (Oligocene) encountered in the
Netherlands. The maximum cone resistance occa
sionally measured in peats at high stress levels is
about 2 MPa. This results in a rather sharp transition
from SBT=2a to SBT=3. The new boundaries are
extended to a friction ratio of 20%, which is occa
sionally measured in peats at low stress levels.
The performance results are presented in Table 3.
It is concluded that majority (78%, 86%) of the clas
sified points in SBT=2a, 3, 4 are correct. For
SBT=2b and 2c, it is concluded that a significant
number of points plot in the adjacent SBT zone but
Figure 2. 2018 and 2021 database results and proposed
still the largest subgroup (38%, 42%) complies with
SBT adjustment for organic soils, presented on top of
Robertson (2010) SBT template. the proposed SBT zone. The number of false posi
tives outside of the adjacent SBT zones is less
than 5%.
Table 2. Performance results for organic soils based on
The formulation for the new proposed boundaries
existing Robertson (2010) SBT classification. Percentage is
number of samples per category plotted in a SBT zone. is shown in Equation (1). The parameter values are
shown in Table 4.
SBT zone: SBT2 SBT3,4
Soil type (Ic>3.6) (Ic≤3.6)
Peat [N>30] 35% 65%

Org. Clay [15<N≤30] 21% 79%
Figure 3 presents a CPT according to the adjusted
Clay (org.mat) 6% 94%
[3<N≤15]
classification system. From the borehole and sam
Clay (mineral) [N≤3] 0% 100%
ples, the following layers are identified: organic clay
from surface, soft clay with organic material
512
Table 4. Parameter values for boundaries of proposed
adjustments to Robertson SBT (2010) and SBTn (2009)
classification.
SBT & SBTn boundary:
Parameter SBT=2a SBT=2b SBT=2c
a (-) 8.0 5.2 4.7

b (-) 0.50 0.62 0.64
Rf,min (%) 5.2 2.3 0.60
(-1.5 m NAP), peat (-2.0 m NAP), sand

(-4.3 m NAP). The layers are well captured except
that based on the CPT classification an intermediate
layer is shown between the peat and sand layer,
which is likely a transition effect.
Figure 4. 2021 database results and proposed SBT adjust

ment for organic soils, presented on top of Robertson
(2009 & 2016) SBT template.
points plot in the CC category, contractive clays,

although quite some points plot in the CCS, contract
ive clays sensitive, category which is larger than the
SBT=1 (2009) category. A few points plot in the CD
category, dilative clays. Those points correspond to
the organic soils with high stresses or large over-
consolidation.
4 CONCLUSIONS AND RECOMMENDATIONS
This publication presents the challenges in CPT-

based classification of organic soils. One of the chal
lenges is that the identification and estimation of unit
weight for organic clays and peats from CPT data is
often in accurate using existing methods.
The coarse grained and fine-grained soils classi
fied as mineral soils correspond well with existing
Figure 3. CPT LKMP33 at Eemdijk the Netherlands classi SBT classifications. The organic soils, classified
fied with the proposed system to include soils. according to the FHWA method, do not match well
with the SBT classification. In the proposed adjust
ment to Robertson (2010), SBT=2 (Organic soils)
The proposed boundaries are optimized such that and SBT=3 are redefined and split up into SBT=2a
they can also be applied as adjustment to (Robertson, (Peat), SBT=2b (Organic Clay) and SBT=2c (Min
2009, Robertson, 2016), in combination with a stress eral Clay with organic matter). The classification is
normalization cut-off equal to Cn≤2. The parameters based on data pairs up to 15 m depths and 150 kPa
in Equation 1 are replaced by Qtn and Fr. These vertical effective stresses. The new SBT zones can
boundaries for organic soils do not apply if there is also be applied in the SBTn classifications by
no stress normalization cut-off applied. The results Robertson (2009) and (Robertson 2016) in combin
of the 2021 database are plotted in Figure 4 on top ation with a Cn=1.7 as stress normalization cut-off.
of the combined 2009 and 2016 template. In this In general, it is highly recommended to perform
figure Cn=1.7 is applied in line with recommended CPTs adjacent to boreholes, select pairs of high-
practice by (Boulanger and Idriss, 2016). The min quality laboratory tests according to a standardized
eral clays and sands are not included, as not all stres protocol (STOWA). The pairwise established 2021
ses required for normalization are known. Most database of classification test results and CPT
513
measurements provides valuable insight in the prop Landva, A. O., Korpijaakko, E. O. & Pheeney, P. E. 1983.
erties of organic soils and an improved classification Geotechnical Classification of Peats and Organic Soils.
system. In: Jarrett, P. M. (ed.) Testing of Peats and Organic
Soils. West Conshohocken, PA: ASTM International.
Lengkeek, H. J. 2022. CPT-based classification and correl
ACKNOWLEDGEMENTS ations for organic soils. 4TU.ResearchData.
Lengkeek, H. J., De Greef, J. & Joosten, S. 2018. CPT
The authors would like to thank the POVM, estab based unit weight estimation extended to soft organic
lished in 2015 by Water Authorities in the Netherlands, soils and peat. 4th International Symposium on Cone
who initiated and financed the “Eemdijkproef”. Penetration Testing (CPT’18). Delft.
This work is part of the “Perspectief” research Lunne, T., Powell, J. J. M. & Robertson, P. K. 2002. Cone
programme All-Risk with project number P15-21, Penetration Testing in Geotechnical Practice, CRC
which is (partly) financed by NWO Domain Applied Press.
and Engineering Sciences. Mayne, P. W. 2014. Interpretation of geotechnical param
eters from seismic piezocone tests. Proceedings, 3rd
International Symposium on Cone Penetration Testing.
REFERENCES Mesri, G. & Ajlouni, M. 2007. Engineering Properties of
Fibrous Peats. Journal of Geotechnical and Geoenviron
Been, K. & Jefferies, M. G. 1993. Towards systematic CPT mental Engineering, 133, 850–866.
interpretation, Thomas Telford Publishing. NEN5104 1989. Classificatie van onverharde grondmon
Begemann, H. K. 1965. The friction jacket cone as an aid sters (In Dutch), Classification of unconsolidated soil
in determining the soil profile. Proc. 6th Int. Conf. on samples. Nederlands Normalisatie-instituut.
SMFE, 1, 17–20. Robertson, P. K. 1990. Soil classification using the cone
Boulanger, R. W. & Idriss, I. M. 2016. CPT-Based penetration test. Canadian Geotechnical Journal, 27,
Liquefaction Triggering Procedure. Journal of Geo- 151–158.
technical and Geoenvironmental Engineering, 142, Robertson, P. K. 2009. Interpretation of cone penetration
04015065. tests — a unified approach. Canadian Geotechnical
Den Haan, E. J. 1997. An overview of he mechanical Journal, 46, 1337–1355.
behaviour of peats and organic soils and some appropri Robertson, P. K. 2010. Soil behaviour type from the CPT:
ate construction techniques. Conference on Recent an update. 2nd international symposium on cone pene
Advances in Soft Soil Engineering, 5-7 March 1997, tration testing, USA.
Kuching, Serawak. Geodelft. Robertson, P. K. 2016. Cone penetration test (CPT)-based
EN1997-1 2005. Eurocode 7: Geotechnical design - part 1: soil behaviour type (SBT) classification system — an
General rules. European Committee for Standardization. update. Canadian Geotechnical Journal, 53,
EN1997-2 2007. Eurocode 7: Geotechnical Design - Part 2: 1910–1927.
Ground investigation and testing. European Committee Robertson, P. K., Campanella, R. G., Gillespie, D. &
for Standardization. Greig, J. 1986. Use of piezometer cone data. Use of
Eslami, A. & Fellenius, B. H. 1997. Pile capacity by direct in situ tests in geotechnical engineering. ASCE.
CPT and CPTu methods applied to 102 case histories. Robertson, P. K. & Wride, C. E. 1998. Evaluating cyclic
Canadian Geotechnical Journal, 34, 886–904. liquefaction potential using the cone penetration test.
Huang, P.-T., Patel, M., Santagata, M. C. & Bobet, A. Canadian Geotechnical Journal, 35, 442–459.
2009. Classification of organic soils. Joint Transporta Schmertmann, J. H. 1978. Guidelines for cone penetration
tion Research Program, Indiana Department of Trans test: performance and design. United States. Federal
portation and Purdue University, West Lafayette, Highway Administration.
Indiana,. Schneider, J. A., Randolph, M. F., Mayne, P. W. &
ISO14688-1 2017. Geotechnical investigation and testing Ramsey, N. R. 2008. Analysis of Factors Influencing
Identification and classification of soil - Part 1: Identifica Soil Classification Using Normalized Piezocone Tip
tion and description. International Organization for Resistance and Pore Pressure Parameters. Journal of
Standardization. Geotechnical and Geoenvironmental Engineering, 134,
ISO14688-2 2017. Geotechnical investigation and testing 1569–1586.
Identification and classification of soil - Part 2: Prin Von Post, L. 1922. Sveriges Geologiska Undersöknings tor
ciples for a classification. International Organization for vinventering och några av dess hittills vunna resultat (In
Standardization. Swedish). SGU peat inventory and some preliminary
Kulhawy, F. H. & Mayne, P. W. 1990. Manual on estimat results. 36.
ing soil properties for foundation design.; Electric Zhang, G., Robertson, P. K. & Brachman, R. W. I. 2002.
Power Research Inst., Palo Alto, CA (USA); Cornell Estimating liquefaction-induced ground settlements
Univ., Ithaca, NY (USA). Geotechnical Engineering from CPT for level ground. Canadian Geotechnical
Group. Journal, 39, 1168–1180.
514
© 2022 the Author(s), ISBN 978-1-032-31259-0
Shear wave velocity – SCPTU correlations for sensitive marine clays

M. Long
School of Civil Engineering, University College Dublin (UCD), Ireland
J.-S. L’Heureux
Norwegian Geotechnical Institute (NGI), Trondheim, Norway
ABSTRACT: The purpose of this paper is to encourage the use of the seismic cone penetrometer (SCPTU)
in soil characterisation studies. There has been an increase in use of shear wave velocity (Vs) data in geotech
nical engineering. This has been prompted by improvements in measurement and analytical systems.
A significant advantage, as is confirmed here, is that Vs can be measured easily and repeatedly by several
different techniques in the sensitive marine clays under consideration here. Here the focus is on the derivation
of preconsolidation stress (pc') from Vs. A rational method of determining a Vs/pc' relationship is outlined
with resorting to empirical data analysis. The proposed relationship is shown to work well for Canadian sensi
tive clay data as has been shown previously for Norwegian and Swedish clays
1 INTRODUCTION 2 VS MEASUREMENTS
There has been increasing recent use of shear wave 2.1 Invasive methods
velocity (Vs) measurements in geotechnical engineer
ing practice. This has been driven by advances in cost Geophysical methods can be divided into two cat
effective and efficient methods of determination of Vs. egories: invasive and non-invasive. Common inva
Traditionally Vs measurements were used for seismic sive methods include down-hole logging, cross-hole
and dynamic analyses. However, they are being logging, suspension logging, seismic dilatometer
increasingly used for site characterisation studies, (SDMT) and the seismic cone penetration test
determination of soil parameters, foundation settlement (SCPTU). In Scandinavia and Canada most invasive
analyses, assessment of sample disturbance and in the testing is done with the SCPTU.
quality control of ground improvement schemes. A standard CPT is equipped with one or more seis
This paper focuses on the use of Vs values to pro mic sensors. The seismic signals are only recorded
vide first order estimates and quality control checking during pauses in penetration, commonly every 0.5 or
of some geotechnical properties of sensitive marine 1.0 m. A horizontal beam coupled to the ground sur
clays. Unfortunately, it has been shown that different face by the weight of the testing vehicle is the source
forms of the correlation equations have been devel of the seismic energy. The beam is struck on end with
oped in different areas. It appears that local correl a hammer to generate shear waves. Vs is determined
ations are necessary for satisfactory use of the from the travel-time differences along the assumed
technique as demonstrated for example by L’Heureux travel path length for receiver depth.
and Long (2017), Duan et al. (2019) or Elbeggo et al.
(2021). In this paper data for clays in eastern Canada 2.2 Non-invasive methods
will be examined and compared with similar clays in
Norway and southern Sweden. The marine clays of Of available non-invasive geophysical methods, per
these three countries have similar properties and share haps that most widely used in Scandinavia and
a comparable geological depositional environment. Canada is the multichannel analysis of surface waves
The Vs profiles and basic soil properties in these (MASW) technique. This technique was introduced
areas will be studied to investigate any systematic dif in the late 1990s by the Kansas Geological Survey
ferences and links between the Vs measurements. (Park et al., 1999). This method utilises the dispersion
Focus will be then placed on use of Vs to determine property of surface waves for the purpose of Vs pro
the important preconsolidation stress (pc') parameter. filing. Some further details on the use and validation
It is hoped that that this work can lead to a unification of the MASW technique in Norwegian clays can be
of these important practical relationships. found in L’Heureux and Long (2017).
DOI: 10.1201/9781003308829-73
515
Figure 1. (a) Vs profiles Onsøy site. Data from Gundersen et al. (2019) and Icelandic group MASW from Ólafsdóttir et al.
(2019) and (b) Tiller-Flotten. SDMT and SCPTU No TILC29 data from L’Heureux et al. (2019) and MASW from Icelandic
Group from Ólafsdóttir et al. (2019).
2.3 Comparison of Vs measurements using surface to about 225 m/s at 20 m depth. The Tiller-
different techniques Flotten data falls within this general trend. An excep
tion to the trend is the data from Onsøy where the
Between 2016 and 2019, NGI and its partners estab
values of Vs are significantly lower though they do
lished five National GeoTest Sites (NGTS) in
show a clear tendency for an increase with depth.
Norway for testing and verifying innovative soil
A similar set of data from Southern Sweden is
investigation methods and foundation solutions
shown on Figure 2b. Again all values are very simi
(L’Heureux et al., 2017). Two of the sites at Onsøy
lar but here they are much lower than the Norwe
and Tiller-Flotten are underlain by soft sensitive
gian measurements with Vs increasing from some
marine clays. The soils at Tiller-Flotten can be clas
50 m/s near ground level to 125 m/s at 20 m depth.
sified as quick below a depth of about 8 m using
In fact the Swedish data is very similar to the
laboratory Swedish fall cone data. Vs profiles have
Onsøy profile.
been made with several techniques at these two sites,
A compilation of available Eastern Canadian data
see Figures 1a and 1b.
is shown on Figure 3. Many of the profiles fall
The profiles from SCPTU, SDMT and MASW at
within the bounds of the Southern Sweden sites. An
the two sites are very similar. The Vs values at
exception is the profile from the Quyon Landslide
Onsøy are significantly less than those at Tiller-
site and perhaps the City of Ottawa data.
Flotten. The reasons for this will be explored below.
4 COMPARISON OF PROPERTIES OF CLAYS

3 VS PROFILES FROM NORWAY, SOUTHERN FROM THE THREE COUNTRIES
SWEDEN AND EASTERN CANADA
A summary of the key properties of the clays from
Several Vs profiles from a series of selected sites in the three countries is given on Table 1. For this pur
Norway are shown on Figure 2a. The sites are from pose typical sites have been chosen, namely Göteborg
several areas of the country including southern Central Station from Southern Sweden, St. Alban
Norway, the area around Trondheim and northern from Eastern Canada as well as the two NGTS sites
Norway. Measurements were made using a variety of at Onsøy and Tiller-Flotten from Norway.
techniques as discussed above. In general the Vs pro The Tiller-Flotten site (and generally many of the
files are very similar and show Vs increasing approxi Norwegian sites) are significantly different from the
mately linearly with depth from about 125 m/s at the other sites.
516
Figure 2. (a) Vs profiles for selected Norwegian sites. Data from L’Heureux and Long (2017) and this paper and (b) for
Swedish clays from Long et al. (2017) and Long and D’Ignazio (2020).
These Norwegian sites have relatively lower

water content and plasticity and higher density
(1.7 – 1.9 Mg/m3 compared to 1.6 – 1.7 Mg/
m3) than the Canadian and Swedish sites. Also
Tiller-Flotten has very low organic content com
pared to the other sites. The Onsøy site param
eters are much closer to those of the Swedish
and Canadian sites. All sites under consideration
have similar clay content and stress history.
Table 1. Summary of material properties for the

study sites: w = water content, Ip = plasticity index,
Org. = organic content OCR = overconsolidation ratio,
St = fall cone sensitivity. Main references Gundersen
et al. (2019), L’Heureux et al. (2019), Wood (2016)
and Trak et al. (1980).
W Clay Org.
Site (%) (%) Ip (%) (%) OCR St
Onsøy 40-80 50-70 25-50 2.5-4 1.1-2.0 5-8

Tiller 30-50 45-70 8-20 Very 1.5-2.0 up to
Figure 3. Vs profiles for Eastern Canadian sensitive clays Flotten low 350
Data from Bouchard et al. (2017), Lefebvre et al. (1994), Göteborg 60-90 70-90 27-40 2-5 1.5-2.0 12-30
Leroueil et al. (2003), Mayne et al. (2019), Fabien-Ouellet CS
et al. (2014), Motazedian et al. (2011), Elbeggo et al. St. Alban 60-90 45-81 5-30 0.9 2.2 14-22
(2021) and Agaiby (2018).
517
Figure 4. Normalised Vs profiles (a) by vertical effective stress and (b) preconsolidation stress.
5 NORMALISATION OF VS VALUES
5.1 Normalisation by vertical effective stress

According to Hight and Leroueil (2003) and Hardin
(1978) the controlling factors on Vs are primarily Unfortunately, as is well known, pc' can be heavily
functions of soil density, void ratio, and effective influenced by sample disturbance effects and by the
stress, with secondary influences including soil type, method used to determine pc' from the measured oed
age, depositional environment, cementation and stress ometer tests data. To deal with the issue of sample dis
history. It is logical then to attempt to harmonise the turbance the sites have been chosen where very high
Vs profiles by normalising them with respect to in situ quality samples are available. Data from Sherbrooke
vertical effective stress (σv0'). Here the normalised block or mini-block samples were available for all four
parameter Vsn is determined from as follows: sites. The Casagrande (1936) technique was used to
determine pc' at three of the sites with the Janbu
(1969) approach being used for the Göteborg Central
Station site. No correction has been applied to the pc'
data. The reported values have been used and com
pared directly to Vs measurements at the same depth.
Mayne et al. (1998), Robertson (2009) and others As can be seen on Figure 4b this form of normal
have chosen n = 0.25 based mostly on laboratory isation was very successful in harmonising the four
sets of data. All four profiles are very similar and
data on silica sands. Here a value of 0.5 has been
show an average Vsnp value of about 100 m/s.
chosen. Data from the four selected study sites nor
malised as above are plotted against depth on Taking this average Vsnp value the following equa
Figure 4a. Although the normalisation brings the tion can be obtained to relate Vs and pc'.
values from the four study sites closer together there
are still significant differences between the values
especially those of Tiller-Flotten and St. Alban
below about 4 m. The value of n was altered but no This form of power equation supports and justifies
improvements in the relationships were observed. some previous similar empirical equations that have
been developed. These include the general relationship
5.2 Normalisation by preconsolidation stress developed by Mayne et al. (1998) as shown on Equa
tion 4, that derived for Norwegian marine clays by
To take the stress history of the materials into L’Heureux and Long (2017) (Equation 5) and by Duan
account the measured Vs data have been normalised et al. (2019) for Jiangsu clays in China (Equation 6)
by the preconsolidation stress (pc') on Figure 4b.
A form of normalisation very similar to that
expressed in Equation 3 has been used as follows:
518
Figure 5. Relationship between Vs and pc' for Eastern Canada clays. Data from Lefebvre et al. (1994), Tavenas and Ler
oueil (1979), Leroueil et al. (1983b), Landon (2007), Leroueil et al. (1983a), Lo et al. (1976), Bozozuk (1972), Hamouche
et al. (1995), Leroueil et al. (2003), Tanaka et al. (2003), Bouchard et al. (2017) and Burnotte et al. (2004).
SCPTU device. The particular focus here was the deriv

ation of pc' from Vs data for sensitive marine clays. It
was shown that despite Vs profiles for marine clays
from different countries being often different, normalisa
tion by pc' harmonises the different profiles. A rational
method, without resorting to empirical correlations, is
6 VS-PC' RELATIONSHIP FOR EASTERN outline for the derivation of a formula which relates Vs
CANADA CLAYS and pc'. The derived formula is tested successfully on
a set of data for Canadian marine clays. More sites
Available Vs and parallel pc' data for Eastern Canada should be included in the relationship shown on
clays is shown on Figure 5. This data should be trea Figure 4b to study the likely variation in the values
ted with caution as it comes from several sources and determined.
it is possible different sampling techniques, testing
techniques and methods of deriving pc' may have
been used. The data is shown merely to illustrate the REFERENCES
application and usefulness of Equation 5 and suggests
further work on this approach is well warranted. Agaiby, S. S. (2018) Advancements in the interpretation of
seismic piezocone tests in clays and other geomaterials.
PhD Thesis, School of Civil and Enironmental Engin
7 CONCLUSIONS eering, Georgia Institute of Technology.
Bouchard, S., Ali, H., Leboeuf, D., Leroueil, S. &
Cascante, G. (2017) Dynamic properties of sensitive
The purpose of this paper was to highlight some advan
clay deposits. IN Thakur, V., L’Heureux, J.-S. &
tages in the use of Vs values in soil characterisation Locat, A. (Eds.) 2nd International Workshop on Land
studies and to therefore encourage the use of the slides in Sensitive Clays (IWLSC).
519
Bozozuk, M. (1972) The Gloucester test fill. PhD Disserta a structured Champlain clay. Canadian Geotechnical
tion, Department of Civil Engineering, Purdue Univer Journal, 31, 61–70.
sity, West Layfayette, IN, 184. Leroueil, S., Hamouche, K., Tavenas, F., Boudali, M., Locat, J.,
Burnotte, F., Lefebvre, G. & Grondin, G. (2004) A case Virely, D., Roy, M., La Rochelle, P. & Leblond, P. (2003)
record of electrosmotic consolidation of soft clay with Geotechnical characterisation and properties of a sensitive
improved soil-electrode contact. Canadian Geotechnical clay from Québec. IN Tan, T. S., Phoon, K. K., Hight, D. W.
Journal, 41, 1038–1053. & Leroueil, S. (Eds.) Proceedings International Workshop
Casagrande, A. (1936) The determination of the on Charaterisation and Engineering Properties of Natural
pre-consolidation load and its practical significance. Pro Soils. Singapore, Balkema, Rotterdam.
ceedings of the 1st International Soil Mechanics and Foun Leroueil, S., Samson, L. & Bozozuk, M. (1983a) Laboratory
dation Engineering Conference. Cambridge, Massachusetts. and field determination of preconsolidation pressures at
Duan, W., Cai, G., Liu, S. & Puppala, A. J. (2019) Correl Gloucester. Canadian Geotechnical Journal, 20, 477–490.
ations between shear wave velocity and geotechnical Leroueil, S., Tavenas, F., Samson, L. & Morin, P. (1983b)
parameters for Jiangsu clays of China. Pure and Applied Preconsolidation pressure of Champlain clays, Part 2,
Geophysics, 176, 669–684. Laboratory determination. Canadian Geotechnical Jour
Elbeggo, D., Ethier, Y., Dubé, J. S. & Karray, M. (2021) nal, 20, 803–816.
Critical Insights in laboratory shear wave velocity cor Lo, K. Y., Bozozuk, M. & Law, K. T. (1976) Settlement
relations of clays. Canadian Geotechnical Journal, analysis of the Gloucester test fill. Canadian Geotech
Accepted manuscript. nical Journal, 13, 339–354.
Fabien-Ouellet, G., Fortier, R. & Giroux, B. (2014) Joint Long, M. & D’ignazio, M. (2020) Shear wave velocity as
acquisition and processing of seismic reflections and sur a tool for characterising undrained shear strength of Nordic
face waves in a sensitive clay deposit in the Outaouais clays. 18th Nordic Geotechnical Meeting. Helsinki, Finland
region (Québec), Canada. 1st International Workshop on (Conferece held virtually January 2021), IOP Conf. Series:
Landslides in Sensitve Clays (IWLSC). Québec, Springer. Earth and Environmental Science 710 (2021) 012008.
Gundersen, A. S., Hansen, R. C., Lunne, T., L’heureux, J. Long, M., Wood, T. & L’heureux, J.-S. (2017) Relationship
S. & Strandvik, S. O. (2019) Characterization and between shear wave velocity and geotechnical parameters for
engineering properties of the NGTS Onsøy soft clay Norwegian and Swedish sensitive clays. 2nd International
site. AIMS Geosciences, 5, 665–703. Workshop on Landslides in Sensitive Clays (IWLSC).
Hamouche, K., Leroueil, S., Roy, M. & Luteneger, A. J. Mayne, P. W., Cargill, E. & Miller, B. (2019) Geotechnical
(1995) In situ evaluation of K0 in eastern Canada clays. characteristics of sensitive Leda clay at Canada test site
Canadian Geotechnical Journal, 32, 677–688. in Gloucester, Ontario. AIMS Geosciences, 5, 390–411.
Hardin, B. O. (1978) The nature of stress – strain behaviour Mayne, P. W., Robertson, P. K. & Lunne, T. (1998) Clay
for soils. In Proceedings ASCE Speciality Conference stress history evaluated from seismic piezocone tests. IN
on Earthquake Engineering and Soil Dynamics, Pasa Robertson, P. K. & Mayne, P. W. (Eds.) Proceedings
dena, California. 1st. International Conference on Geotechnical Site
Hight, D. W. & Leroueil, S. (2003) Characterisation of soils for Characterisation. Atlanta, Georgia, Balkema.
engineering purposes. IN Tan, T. S., Phoon, K. K., Hight, Motazedian, D., Hunter, J. A., Pugin, A. & Crow, H.
D. W. & Leroueil, S. (Eds.) Proceedings International (2011) Development of a Vs30 (NEHRP) map for the
Workshop on Charaterisation and Engineering Properties city of Ottawa, Ontario, Canada. Canadian Geotech
of Natural Soils. Singapore, Balkema, Rotterdam. nical Journal, 48, 458–472.
Janbu, N. (1969) The resistance concept applied to deform Ólafsdóttir, E. A., Bessason, B., Erlingsson, S.,
ations of soils. Proceedings of the 7th International Soil L’heureux, J.-S. & Bazin, S. (2019) Benchmarking of an
Mechanics and Foundation Engineering Conference. open-source MASW software using data from three
Mexico City, A.A. Balkema, Rotterdam. Norwegian GeoTest Sites. 17th European Conference
L’heureux, J.-S., Lindgård, A. & Emdal, A. (2019) The on Soil Mechanics and Geotechnical Engineering
Tiller-Flotten research site: Geotechnical characterisa (ECSMGE). Reykjavik, Iceland.
tion of a sensitive clay deposit. AIMS Geosciences, 5, Park, C. B., Miller, D. M. & Xia, J. (1999) Multichannel
831–867. analysis of surface waves. Geophysics, 64, 800–808.
L’heureux, J.-S., Lunne, T., Lacasse, S., Carroll, R., Robertson, P. K. (2009) Interpretation of cone penetration
Strandvik, S. O., Ozkul, Z., Instanes, A., Sinitsyn, A., tests - a unified approach. Canadian Geotechnical Jour
Degago, S. A. & Nordal, S. (2017) Norway’s National nal, 46, 1337–1355.
GeoTest Site Research Infrastructure (NGTS). 19th Tanaka, H., Shiwakoti, D. R. & Tanaka, M. (2003) Applic
International Conference on Soil Mechanics and Geo- ability of SHANSEP method to six different natural
technical Engineering (ICSMGE). Seoul. clays using triaxial and direct shear tests. Soils and
L’heureux, J.-S. & Long, M. (2017) Relationship between Foundations, 45, 43–55.
shear wave velocity and geotechnical parameters for Nor Tavenas, F. & Leroueil, S. (1979) Clay behaviour and the
wegian clays. Journal of Geotechnical and Geoenviron selection of design parameters. 7th European Confer
mental Engineering ASCE, 04017013-1 – 04017013–20. ence on Soil Mechanics and Foundation Engineering
Landon, M. E. (2007) Development of a non destructive (ECSMFE). Brighton, UK.
sample quality assessment method for soft clays. PhD Trak, B., La Rochelle, P., Tavenas, F., Leroueil, S. &
Thesis, Department of Civil and Environmental Engin Roy, M. (1980) A new appeoach to the stability analysis
eering, University of Massachusetts, Amherst. PhD of embankments on sensitive clays. Canadian Geotech
Thesis, Department of Civil and Environmental Engin nical Journal, 17, 526–544.
eering, University of Massachusetts, Amherst. Wood, T. (2016) On the small strain stiffness of some Scandi
Lefebvre, G., Beloueuf, D., Rahhal, M. E., Lacroix, A., navian soft clays and impact on deep excavations, PhD
Warde, J. & Stokoe, K. H. (1994) Laboratory and field thesis Department of Civil and Environmental Engineering,
determinations of small-strain shear modulus for Chalmers University of Technology, Göteborg, Sweden.
520
© 2022 the Author(s), ISBN 978-1-032-31259-0
A simplified method to incorporate the benefits of microstructure for cyclic

liquefaction analyses using the SCPT
K. Lontzetidis
CMW Geosciences, New Zealand
P.K. Robertson
Gregg Drilling Inc., USA
D.J. Morton
CMW Geosciences, New Zealand
ABSTRACT: Cyclic liquefaction resistance of sand deposits can increase due to microstructure that results
from several factors, such as aging and bonding. Available empirical correlations, using either SPT, CPT or shear
wave velocity (VS) data, were derived from cyclic liquefaction case histories that considered very young Holo
cene-age, essentially normally consolidated, unbonded, silica-based soils that may not apply to soils with signifi
cant microstructure. Seismic CPT (SCPT) data can be used to identify soils with significant microstructure, since
both aging and bonding tend to increase the small-strain stiffness (reflected in the measured VS) significantly more
than they increase the large-strain strength of a soil (reflected in the CPT penetration resistance). Hence, for
a given soil, both age and bonding tend to increase VS more than the larger-strain cone resistance, all other factors
(such as, in situ stress state, density, etc.) being constant. The normalized rigidity index (KG) has been proposed
(Robertson, 2016) as a parameter, that combines VS and normalized CPT tip resistance (Qtn), to detect and quan
tify the presence of microstructure. This paper presents and discusses a suggested method to quantify the increased
resistance to cyclic liquefaction due to microstructure by utilizing KG. Results will be presented from older soil
deposits in New Zealand well as compare the results with existing correlations that account for “aging”.
1 INTRODUCTION are considered to have a low risk of cyclic liquefac

tion (Youd and Perkins, 1978). Several researchers
Over the last 40 years, since the first simplified have worked to quantify the influence of the age on
method to assess soil liquefaction potential was pro liquefaction potential through laboratory and in-situ
posed by Seed and Idriss (1971), empirical methods testing (e.g., Seed 1979; Troncoso et al. 1988,
have evolved to improve the reliability of predicting Arango and Migues 1996; Arango et al. 2000;
liquefaction triggering. Empirical correlations are Robertson et al. 2000; Lewis et al. 2004, Lewis et al.
based on either standard penetration test (SPT), cone 1999; Hayati and Andrus 2008).
penetration test (CPT) or shear wave velocity (Vs) The objective of this paper is to investigate an
data. The CPT is now the most commonly used in- alternative method to estimate the potential positive
situ test to evaluate liquefaction potential in liquefac effects of either age or bonding of a soil by utilizing
tion-prone areas. The main advantages of the CPT the normalized rigidity index (KG) proposed by
are the continuous and repeatable measurements that Robertson (2016). Seismic CPT (SCPT) data from
provide a detailed profile of the soil, as well as early to late Pleistocene-age soils in New Zealand
major developments in CPT-based liquefaction pre (NZ) are used to evaluate the proposed method. The
diction methods that have occurred over recent years results using the proposed method are compared
(e.g., Boulanger and Idriss 2014). A limitation of all with another method also utilising seismic CPT data.
liquefaction prediction methods is that they have
been derived from cyclic liquefaction case histories
that consider very young Holocene-age, essentially 2 PREVIOUS STUDIES
normally consolidated, unbonded, silica-based soils.
Correlation of results from those case histories can Several researchers have established correlations
therefore be conservative when applied to older or between the age of a soil and the increase in lique
bonded soils. Pleistocene-aged soils (>12,000 years) faction resistance (e.g., Seed (1979); Troncoso et. al
DOI: 10.1201/9781003308829-74
521
(1988); Lewis et al. (1999) Arango et al. (2000) soil microstructure. This was done by extending the
Lewis et al. (2004) Leon et al. (2006) Hayati and link between the CPT and the VS using the empirical
Andrus (2008). Results show that there is an increase parameter, KG defined by:
in liquefaction resistance with increasing soil age.
Andrus et al. (2009) proposed a ratio of the meas
ured to estimated Vs (MEVR) to quantify the
increase in resistance to cyclic loading of older soils.
Saftner et al. (2015) presented a summary of past
relationships, in terms of a Strength Gain Factor where Go is in same units as qn, Qtn is dimensionless,
KDR, for resistance to cyclic loading and age and KG is essentially a normalized rigidity index.
(Figure 1) and proposed a new relationship to Most of the existing empirical correlations devel
account for geologic age by evaluating data from oped for the interpretation of CPT results are predom
explosive compaction tests. inately based on experience in silica-based soils with
little or no microstructure (e.g., Robertson 2009;
Mayne 2014). Hence, if soils have KG < 330, the soils
are likely young and unbonded (i.e., have little or no
microstructure) and can be classified as ‘ideal’ soils
(unstructured) where most traditional CPT-based
empirical correlations for liquefaction prediction likely
apply. Soils with KG > 330 tend to have significant
microstructure, and the higher the value of KG, the
more microstructure is likely present. Hence, if a soil
has KG > 330, the soils can be classified as ‘structured’
soils where traditional generalised CPT-based empir
ical liquefaction prediction correlations may have less
reliability and where local modification may be
needed. The influence of increasing microstructure on
in situ soil behavior is often gradual, and any separat
ing criteria can be somewhat arbitrary. Data suggests
that very young unbonded soils tend to have KG values
closer to 100, whereas soils with some microstructure
Figure 1. Relationship between strength gain factor (KDR) (e.g., early Pleistocene-age) tend to have KG values
and soil age based on data from explosive compaction tests closer to 330. As will be shown later, soils with KG <
and previous studies (After Saftner et al. 2015). 330 tend to have little or no microstructure where
empirical CPT-based liquefaction prediction correl
ations tend to provide good estimates of soil behavior.
One limitation of a correction based on geologic
age, it that past liquefaction events can modify the
behavior of a soil so that its ‘behavior age’ is not the 4 STUDY AREA
same as its geologic age. Hence, a correction based on
behavior characteristics may have wider application.
4.1 Site description
The Ruakura site comprises an area of approximately
3 MICROSTRUCTURE 177 ha and is located at the intersection of Silverdale
Road and Ruakura Road, in Hamilton, New Zealand.
Eslaamizaad and Robertson (1996) showed that it is The land use has generally remained unchanged from
possible to identify soils with significant microstruc agricultural practices since at least 1974 (circa).
ture (due to age, bonding or unusual mineralogy) There are a few buildings/structures, which have
using the SCPT based on a link between the ratio of occupied the land to service the agricultural oper
small strain shear modulus (Go), net cone resistance ations (i.e., sheds, water tanks). The natural topog
(qn) and the normalized cone resistance (Qtn), since raphy of the site is typically flat with minor
both aging and bonding tend to increase the small- undulations from historic stream channels. The eleva
strain stiffness (Go) significantly more than they tion of the site ranges from RL 40.0 m to RL 42.0 m.
increase the large-strain strength of a soil (reflected
in Qtn). Hence, for a given soil, both age and bond
4.2 Geology
ing tend to increase the small-strain shear wave vel
ocity (Vs) more than the larger-strain cone The published geological map (Edbrooke, 2005,
resistance, all other factors (in situ stress state, etc.) Figure 2) indicates that most of the site is underlain by
being constant. late Pleistocene-age river deposits of the Hinuera Forma
Robertson (2015) modified slightly the equation tion (12,000-27,000yr) consisting of cross-bedded
proposed by Schneider and Moss (2011) to identify pumice sand, silt, and gravel with interbedded peat
522
(2Qa). Early to mid-Pleistocene-age river and igneous 5 CORRELATION BETWEEN THE STRENGTH
deposits of the Walton Group (27,000-2,000,000y) GAIN FACTOR AND KG
extend into the western part of the site and comprise
a sequence of ignimbrites and tephra from several Andrus et al. (2009) and Hayati and Andrus (2009)
sources and fine-grained volcanoclastic alluvium (eQa). suggested a method using MEVR, to account for soil
The north-eastern part of the site is indicated to be aging on the resistance to cyclic loading (CRR)
underlain by Holocene-age swamp deposits and peat of using CPT and Vs results. Based on laboratory and
the Tauranga Group (0-12,000y), consisting of soft, dark field cases, they proposed a deposit resistance factor
brown to black, organic mud, muddy peat and woody (strength gain factor, KDR) to correct for age using:
peat with minor overbank sand, silt, and mud swamp
deposits (Q1a).
Based on site specific CPT data, the mean Ic
values for each geological formation are 1.9 for the
Tauranga Group, 1.8 for the Hinuera Formation and to correct the cyclic resistance ratio, CRRK due to
2.7 for the Walton Group. the age of the deposit:
4.3 Geotechnical investigation

Several phases of geotechnical investigation have been
undertaken within the site over recent years. It was rec
ognised that the soils at Ruakura are unlikely to fit the where CRRCPT is the CPT-based CRR from case his
historical liquefaction database used to develop most tories for young Holocene-age deposits. The meas
empirical liquefaction prediction techniques due to: ured VS is computed from in-situ distance and travel
time measurements using the SCPT and the esti
1. The geologic age of the deposits found in parts of mated VS can be obtained for sands by the following
the site are older (Pleistocene-age) than the data equation:
base (Holocene-age).
2. The Hinuera deposits are composed of unusual
mineralogy (pumice) compared to the historical
data base (silica-based soils).
The latest investigation included twenty-two (22) where Vs1,E is the estimated shear-wave velocity cor
SCPTs. Factual data from those tests are presented in rected for overburden pressure and Qtn,cs is the nor
various geotechnical reports prepared by CMW. The malized clean sand equivalent cone resistance.
SCPTs were undertaken to support the liquefaction Hayati and Andrus (2009) showed that the MEVR
analysis as well as to check for increased resistance approach is based on a reference age of about 23
to cyclic liquefaction of soils due to their age. years for Vs1,E which was considered a representa
tive average age for the liquefaction case history
database. Many of the liquefaction cases are associ
ated with deposits that have a “behavioral age” (the
time since the last critical disturbance or liquefaction
event) not more than 100 years. Previous liquefac
tion (or even ground improvement implementation)
is assumed to be a critical disturbance event when
the grain-to-grain contacts were damaged and
reformed.
A similar approach has been proposed by Robert
son (2015) using the Schneider and Moss (2011)
empirical parameter KG following a similar meas
ured to estimated KG ratio defined by
where KG,M = KG based on measured values of Vs

and qt, and KG,E = 200 is the estimated average
value for very young (~23 years), unbonded soils.
Based on the definition of KG [Eq. (1)], the MEKG
ratio is insensitive to changes in CPT qt and Qtn,cs
Figure 2. Geological map of the Ruakura site. due to aging. Since qt has been shown to be
523
relatively insensitive to aging and/or light bonding,
it is reasonable to assume that,
The advantage of using MEKG is that it does not

require the calculation of an estimated Vs, which has
an associated uncertainty, especially in soils that are
not clean sands. Hence, a similar approach can be
applied using MEKG instead of MEVR and apply
Eqs. (2) and (6) to estimate the strength gain factor,
KDR.
6 DATA USED TO ESTIMATE SOIL

MICROSTRUCTURE
The locations of the most recent twenty-two SCPTs

undertaken as part of the project investigation are Figure 3. Site plan showing the locations of the SCPTs.
shown in Figure 3.
At each test location, shear wave velocity (Vs)
measurements were undertaken at 1m intervals verti
cally. The CPT cone resistance data was then averaged
over a depth of 0.5m above and below the relevant Vs where ρ is the soil density in kgr/m3.
measurement depth. From the Vs data and soil density Compiling the CPT data, the net cone resistance
(averaged within 1m zone), the maximum shear modu (qn) and the normalized cone resistance (Qtn), were
lus (Go) was calculated using the following. calculated using the following two equations.
Figure 4. Typical examples of in-situ test results and the calculated values of KGM and KDR.
524
rigidity index (KGM) was then calculated. Finally,
by combining equations 2, 5 and 6, the strength
gain factor (KDR) was calculated using the fol
lowing equation.
And since KG,E = 200 for very young (~23 years),

unbonded soils, this is simplified to:
Equation 11 presents the strength gain factor KDR, as

a function of the measured normalised rigidity index
Figure 5. Comparison of the KDR calculated using MEVR
KG,M. without the need to calculate a reference Vs
and KG,M. value.
Figure 4 presents the results from three typical
SCPTs, one for each geological formation encoun
tered in the area, together with the calculated profiles
of measured normalized rigidity index (KGM) and
the strength gain factor (KDR). Figure 4 illustrates
that the average KDR value increases with the geo
logical age of the soil deposits.
Figure 5 presents a comparison of KDR values cal
where qt is the corrected cone resistance, σvο the ver culated by Andrus et al., 2009 using MEVR with
tical total stress, Pa and Pa2 are reference pressures those using KG,M as proposed in this study. This
(100kPa), σ’vο is the effective vertical stress and n is shows that the KG,M method estimates a KDR value
the stress exponent that varies with Ic (Robertson & that is on average 10% lower than that using the
Cabal 2015). MEVR method. It is also noted that at shallow
Based on the above methodology, a set of depths, the KDR values calculated from KG,M are
small strain shear modulus (Go), net cone resist higher than those calculated using MEVR, while this
ance (qn) and normalized cone resistance (Qtn) trend decreases with depth.
data points were calculated at 1m depth intervals. Figure 6 shows the cumulative frequency distribu
By utilizing equation 1, the measured normalized tion of KDR from each method (KG,M and MEVR) for
Figure 6. Comparison of the results of the KDR, calculated using the proposed KG,M, and MEVR for each of the three geo
logical formations.
525
each of the three geological formations. The results Earthquake Geotechnical Engineering for Protection
indicate that the Holocene-age deposits of the Tau and Development of Environment and Constructions.
ranga Group (Q1a), indicate little or no microstructure Clayton, P. J., and Johnson, J. T., 2013, Liquefaction resist
with 50th percentile KDR(KG,M) = 1.0, and KDR ance and possible aging effects in selected Pleistocene
(MEVR) = 1.1. The slightly older Hinuera Formation soils of the Upper North Island, Proc. 19th NZGS Geot.
(2Qa) indicates some microstructure with 50th per Symposium.
centile KDR(KG,M) = 1.2 and KDR(MEVR) = 1.3. The Clayton, P. J., Tilsley, S. C., Bastin, S. H., and
Green, R. A., 2017, Case study in the use of paleolique
older Walton Group (eQa), indicates more microstruc faction techniques to investigate liquefaction potential
ture with 50th percentile KDR(KG,M) = KDR(MEVR) = of Waikato soils for the Hamilton section of the Waikato
1.5. These results plot at the low range of the values expressway, Proc. 20th NZGS Geotechnical Symposium.
presented in Saftner et al. 2015 (Figure 1) for the age Clayton, P. J., Yong, I., and Wotherspoon, L., 2017, Case
of the Hinuera (2Qa) and Walton Group (eQa). study in the use of shear wave velocity techniques to
investigate liquefaction potential of Waikato soils for
the Hamilton section of the Waikato expressway, Proc.
7 CONCLUSIONS 20th NZGS Geotechnical Symposium
Edbrooke, S. W., 2005, Geology of the Waikato area, Insti
As part of the geotechnical investigation for the Rua tute of Geological and Nuclear Sciences, Lower Hutt,
kura Development in New Zealand a number of geo New Zealand.
Eslaamizaad, S., and Robertson, P. K. 1996, Seismic cone
technical investigation campaigns have been penetration test to identify cemented sands, Proc. 49th
undertaken that included 22 SCPTs to support lique Canadian Geotechnical Conf., St John’s, Newfound
faction analysis and to evaluate the increased resist land, Canadian Geotechnical Society (CGS), 1,
ance of older soils to cyclic loading. 352–360.
This paper presents an alternate method to quan Hayati, H., and Andrus, R. D. 2008, Liquefaction potential
tify the increased resistance to cyclic loading of soils map of Charleston, South Carolina based on the 1886
due to microstructure, which can be readily imple earthquake, Journal of Geotechnical and Geoenviron
mented utilising SCPT data. mental Eng., 134(6) DOI:10.1061/(ASCE)1090
The results of the strength gain factor KDR, by the 0241(2008)134:6(815).
Hayati, H., Andrus, R.D., 2009, Updated liquefaction
proposed method using the normalised rigidity resistance correction factors for aged sands, J. Geotech.
index, compare well with results using the MEVR. Geoenviron. Eng. 135 (11), 1683–1692.
The proposed method returns a rather low Leon, E., Gassman, S.L., and Talwani, P., 2006, Account
increase in resistance to liquefaction compared to ing for Soil Aging when Assessing Liquefaction
other methods (see Figure 1). Potential, J. of Geotechnical and Geoenvironmental
Eng., 132 (3), 363–377.
Lewis, M. R., Arango, I., Kimball, J. K., and Ross, T. E.
REFERENCES 1999, Liquefaction resistance of old sand deposits,
Proc., 11th Panamerican Conf. on Soil Mechanics and
Andrus, R. D. and Hayati, H. 2009. Updated Liquefaction Geotechnical Engineering, Foz do Iguassu, Brazil:
Resistance Correction Factors for Aged Sands, 821–829.
J. Geotech. Geoenviron. Eng., 135(11),1683–1692. Lewis, M. R., McHood, M. D., and Arango, I. 2004, Lique
Andrus, R. D., Hayati, H., and Mohanan, N. P. 2009. Cor faction evaluations at the Savannah River Site, A Case
recting liquefaction resistance of aged sands using meas History, Proc., Fifth Int. Conf. on Case Histories in
ured to estimated velocity ratio J. Geot. Geoenv. Eng., Geotechnical Engineering, New York, NY.
135(6),735–744. Mayne, P. W. 2014, Interpretation of geotechnical param
Arango, I., Lewis, M. R., and Kramer, C. 2000, Updated eters from seismic piezocone tests, 3rd Int. Symp. on
liquefaction potential analysis eliminates foundation Cone Penetration Testing, CPT14, Gregg Drilling &
retrofitting of two critical structures, Soil Dynamics Testing, CA.
Earthquake Eng., Elsevier Limited, 20: 17–25. Robertson, P. K. 2009, Interpretation of Cone Penetration
Arango, I., and Migues, R. E. 1996, Investigation of the Tests—A unified approach, Can. Geotech. J., 46
seismic liquefaction of old sand deposits, Report on (11),1337–1355.
Research, Bechtel Corporation, National Science Foun Robertson, P. K. 2015. Comparing CPT and Vs Liquefac
dation Grant No. CMS-94-16169, San Francisco, CA. tion Triggering Methods, Journal of Geot. and Geoen
Boulanger, R. W., and Idriss, I. M., 2014, CPT and SPT vironmental Engineering, ASCE, 141 (9):04015
based liquefaction triggering procedures, Rep. No. 037,10.1061.
UCD/CGM-14/01, Center for Geotechnical Modeling, Robertson, P. K. 2016. Cone penetration test (CPT)-based
Dept. of Civil and Environmental Engineering, College soil behavior type (SBT) classification system —an
of Engineering, Univ. of California, Davis, CA, 138. update, Can. Geotech. J. 53: 1910–1927.
Bwambale, B., Andrus R. D., and Cubrinovski, M., 2017, Robertson, P. K. and Cabal, K. L. 2015, Guide to Cone
Influence of age on liquefaction resistance of Holocene Penetration Testing for Geotechnical Engineering, 6th
alluvial and marine soils in Christchurch and Kaiapoi, edition, Gregg Drilling & Testing, Inc.
New Zealand, PBD III Earthquake Geotechnical Engin Robertson, P. K, Wride, C. E. (Fear), List, B. R.,
eering, Vancouver. Atukorala, U., Biggar, K. W., Byrne, P. M.,
Clayton, P. J., de Graaf K. L., Yong I., and Green, R. A., Campanella, R. G., Cathro, D. C., Chan, D. H.,
2019, Comparison of assessments of liquefaction poten Czajewski, K., Finn, W. D. L., Gu, W. H., Hammamji, Y.,
tial in selected New Zealand pumiceous soils, Hofmann, B. A., Howie, J. A., Hughes, J., Imrie, A. S.,
526
Konrad, J.-M., Küpper, A., Law, T., Lord, E. R. F., Seed, H. B. 1979, Soil liquefaction and cyclic mobility
Monahan, P. A., Morgenstern, N. R., Phillips, R., evaluation for level ground during earthquakes, Journal
Piché, R., Plewes, H. D., Scott, D., Sego, D. C., of the Geotechnical Eng. Div., 105(2): 201–255
Sobkowicz, J., Stewart, R. A., Watts, B. D., Woeller, D. J., Seed, H. B., and Idriss, I. M. 1971, Simplified procedure
Youd, T. L., and Zavodni, Z. 2000, The CANLEX project: for evaluating soil liquefaction potential, J. Soil Mech.
summary and conclusions, Canadian Geotechnical Jour and Found. Div., 97(9),1249–1273.
nal, 37: 563–591. Troncoso, J., Ishihara, K., and Verdugo, R. 1988, Aging
Saftner, D.A., Green R.A., and Hryciw, R. D., 2015, Use of effects on cyclic shear strength of tailing materials,
explosives to investigate liquefaction resistance of aged Proc., Ninth World Conf. on Earthquake Engineering,
sand deposits, Engineering Geology 199: 140–147. Vol. III: 121–126.
Schneider, J. A., and Moss, R. E. S. 2011, Linking cyclic stress Youd, T. L., and Perkins, D. M., 1978, Mapping of lique
and cyclic strain-based methods for assessment of cyclic faction induced ground failure potential, J. Geotech.
liquefaction triggering in sands, Geotech. Lett., 1, 31–36. Engrg. Div.
527
© 2022 the Author(s), ISBN 978-1-032-31259-0
Classification of Miocene deposits using CPT data

A. Makra
GR8 GEO, Athens, Greece
H. Kim
DL E&C Co. Ltd, Seoul, South Korea
ABSTRACT: Soil classification based on CPT data is commonly carried out using classification charts
which link the type of soil behaviour to the measured tip resistance, sleeve friction and pore water pressure.
These charts have been successful in interpreting soil behaviour in young, saturated, uncemented soil
deposits but often diverge from observations in samples in older deposits. The geotechnical investigation
conducted for the 1915 Çanakkale bridge included 20 offshore boreholes with 18 adjacent CPT soundings
conducted downhole that penetrated several meters into Miocene-age sediments. This paper investigates
those data sets and proposes a modified chart for the classification of the Miocene deposits accounting for
their microstructure. The proposed modification is useful for classifying Miocene deposits in similar geo
logical settings through the performance of CPTs.
1 INTRODUCTION Miocene sediments because the high stiffness and

strength of the deposit typically precludes significant
Cone Penetration Testing (CPT) is a very common CPT penetration. This lack of data has limited the
in situ sounding tool used to collect geotechnical development of a globally accepted approach for inter
data at sites. The CPT provides a cost-efficient, fast pretation of CPT data in Miocene sediments. However,
and reliable method to obtain continuous soil pro evolutions in CPT techniques increasingly allow for
files. The basic idea of the test is that a standardized achieving greater penetration in stiffer soils. An early
cone is pushed into the ground at a constant penetra interpretation of the ground profile in these stiffer soils
tion rate while the tip resistance, sleeve friction and using CPT based SBT classification systems can be
pore water pressure are measured. very useful, especially at an early stage of project
Several methods of soil classification have been development.
developed to correlate CPT data to soil characteristics In the 1915 Çanakkale bridge project area,
(Schmertmann 1978, Robertson 1990, 2009, Schneider a significant number of CPT soundings were con
2008, Robertson 2016). Unlike the most common soil ducted onshore, nearshore and offshore in combin
classification systems [ASTM D2487 (USCS), BSCS ation with geophysical surveys, boreholes and other
BS 5930] that are based on the physical properties of geotechnical in situ tests (SPT, wireline logging,
soils such as grain size distribution and plasticity etc.) and extensive laboratory testing.
index, CPT classification methods are based on behav In addition to seabed CPTs, the down-hole mode
iour characteristics and are generally described as Soil was adopted for offshore CPTs to be able to collect
Behaviour Type (SBT) classification systems (Robert data from relatively stiff and competent materials. In
son 2016). contrast to onshore and seabed CPT testing where the
These CPT based classification charts have proven cone is continuously pushed from the surface to
to be very efficient in predicting soil type in young refusal, down-hole CPTs are conducted within
(Quaternary) soil deposits. However, older soils with a borehole with typically 1.5 to 3.0m stroke lengths.
significant microstructure often require site specific Upon completion of the CPT, the portion penetrated
modifications to the classification charts so that the by the cone is drilled out and the borehole is continued
effect of the microstructure on the CPT data is properly with either a subsequent CPT test or sampling
reflected in the SBT classification system (Robertson (Figure 1).
2016). In the 1915 Çanakkale Bridge project, down-hole
Although the existing CPT database is extensive for CPTs were continuously advanced with drill outs to
younger and softer deposits, fewer data exist for the depth where the material became too stiff for
DOI: 10.1201/9781003308829-75
528
Figure 2. CPT Classification Chart from Robertson (2009).
Figure 1. Procedures for drilling, push sampling and CPT Figure 3. CPT classification chart from Robertson (2016).
testing (modified after ISSMGE 2005).
CPT testing. Thus, a significant CPT data set of Mio 2 FRAMEWORK

cene soils was collected from the site.
The identification of the Miocene deposits was pri The 1915 Çanakkale bridge is located between Geli
marily based on the interpretation of the geophysical bolu in East Thrace and Lapseki on the Biga Penin
survey data, which were also confirmed by borehole sula (Troas). The site investigation was conducted by
data. When CPT data associated with the Miocene Fugro (Fugro 2018a, b) and consisted of extensive
deposits were interpreted, the classification based on geological, geophysical and geotechnical investiga
the Robertson (2009, 2016) charts shown on Figures 2 tions. CPTs were performed onshore, nearshore and
and 3 was inconsistent with the observations and offshore. Although the onshore CPTs penetrated only
laboratory data obtained from borehole core samples a few meters into the upper (typically weaker) Mio
probably due to the significant microstructure of the cene layers before experiencing refusal, offshore CPT
Miocene. For instance, the CPT data obtained from using down-hole mode allowed for the continuation
Miocene mudstone were expected to fall within Zone of testing beyond initial refusal and the collection of
9: Very Stiff Fine Grained and CD: Clay like Dilative data deeper within harder strata. Although down-hole
based on Robertson (2009, 2016) (Figures 2 and 3, CPTs are discontinuous and require additional care in
respectively), but many CPT data points plotted out the execution and processing of the test results, they
side of Zone 9, displaying a different trend. Thus, site- are an efficient method to collect CPT data in deeper
specific modifications of the chart were required for strata or below a very stiff/dense layer where seabed
the Miocene sediments. This paper proposes CPTs would have refused. Offshore CPTs using
a modified Robertson classification chart that can be down-hole mode are the main source of CPT data in
used for the classification of fine grained heavily over- Miocene units presented in this paper. Boreholes were
consolidated Miocene mudstones using CPT data drilled adjacent to the CPT locations to facilitate the
based on the data set from the 1915 Çanakkale bridge interpretation and integration of the geotechnical data.
project. The boreholes progressed with continuous sampling
529
in the soil layers and rotary core drilling in the rock coarse grained (Zone 4: Silt mixtures, Zone 5: Sand
and the samples were immediately processed at the mixtures, refer to Figure 2 and Figure 6a). The TD
site and onboard laboratories to minimize sample dis and CD are described as dilative and mainly transi
turbance and to expedite data availability. In several tional material with smaller portion in the clay-like
of these explorations, after the completion of drilling, and sand-like range. Note that these descriptions
P-S wireline logging was conducted (Fugro 2018a). showing the soil behaviour do not properly reflect
the expected behaviour of the Miocene deposit.
A small fraction of the data (about 20%) are
3 RESULTS AND DISCUSSION plotted within Zone 3 which corresponds to
clays. CPT data plotted within Zone 3 are mostly
The bedrock of the project site consists of Miocene located close to the boundary between Quaternary
sedimentary rock of the Çanakkale formation, com and Miocene deposits where the stiffness of mud
posed primarily of mudstones and sandstones with few stones/clays is low (i.e., less than a few hundred
siltstone, marl and limestone layers of limited thick kilopascals).
ness. The top of the Miocene geologic unit was ini
tially interpreted from geophysical data and then
refined locally by borehole and CPT data (Figure 4).
The Miocene deposits are of the same origin as the
overlying younger (Holocene and Pleistocene) layers,
complicating the visual distinction between the strati
graphic units. Thus, the top of Miocene bedrock in the
boreholes was identified based on the degree of
cementation, structure (bedding, fissures) and strength
of the material.
Figure 4. Geophysical and geotechnical cross section, Figure 5. Typical CPT trace of Holocene and Miocene
Çanakkale Strait, Offshore, (Fugro 2018a). deposits from the European Tower explorations (CPT ID:
ET-5).
CPT data were acquired only in Miocene mud-
stones with occasional thin sandstone interlayers.
The CPT equipment was not able to penetrate Laboratory tests performed on mudstone samples
through thick sandstone layers. The top of Miocene showed fines contents exceeding 80% and an aver
was characterized by a sudden increase in tip resist age plasticity index (PI) of 45, which is classified as
ance, followed by increases in friction ratio and pore fat clay (CH) per ASTM D2487. Based on the geo
water pressure. Figure 5 presents typical tip, sleeve technical descriptions of the boreholes with consoli
friction and pore water pressure traces of CPT data dation and strength tests (uniaxial compressive and
in Miocene mudstones. The tip resistance generally UU triaxial strength values), the Miocene mudstones
ranges between 5MPa and 25MPa and the friction are lightly lithified, fresh, overconsolidated, with
ratio from 1.5 to 6%, with the lower friction ratio strengths ranging from extremely weak (<1MPa) to
(less than 2%) corresponding to the thin sandstone very weak (1-5MPa) rock per ISRM (2007) recom
interlayers. mendations or hard clay (>0.2MPa) per the ASTM
Figure 6 shows CPT data collected from the Mio soil classification system.
cene deposit plotted on the Robertson (2009, 2016) Although the Miocene mudstone is classified as
classification charts. Most of the CPT data are plot CH per USCS, the CPT classification chart incor
ted within Zones 4 and 5 of the chart as per Robert rectly suggests that the material displays sandlike
son (2009) and TD and CD as per Robertson (2016). behaviour due to its relatively high normalized tip
However, the Zones 4 and 5 are described as having resistance caused by the effects of cementation and
a mainly drained behaviour and being primarily aging as well as overconsolidation.
530
Figure 7 shows a comparison of the CPT data
from Holocene clayey deposits with the data
from Miocene mudstones in the offshore Euro
pean side of the Çanakkale bridge project. The
CPT data from two deposits show that whilst the
normalized friction ratios are similar, the normal
ized tip resistances are much higher in the Mio
cene deposits. As shown on Figure 7a, b, the
CPT data obtained from Holocene deposits fall
within Zone 3 (clay and clay-like contractive),
and CC (Clay-like Contractive) and CCS (Clay
like Contractive Sensitive). Although the classifi
cation of Miocene deposits using the CPT data
were inconsistent with sample descriptions and
laboratory tests, the SBT classification using CPT
data for the Holocene clayey deposits is consist
ent with the borehole and laboratory data, both of
which indicate the Holocene deposit to be a soft
to firm, near-normally consolidated clay.
Figure 7. CPT data from Miocene and Holocene deposits

on (a) Robertson (2009) and (b) Robertson (2016) Classifi
cation Charts, (Data obtained from Çanakkale Strait, Off
shore European and Anatolian Towers).
Since Miocene age deposits are expected to have

microstructure related to aging and cementation or
other post-depositional processes, the microstructure
of the sediments around the 1915 Çanakkale bridge
was assessed following the recommendations of
Robertson (2016) which relate microstructure to the
normalized tip resistance (Qtn), the small strain stiff
ness (G0) and the net cone tip resistance (qn=qt-σv).
Robertson (2016) introduces a modified normalized
small strain rigidity index KG* defined as:
Figure 6. CPT data from the Miocene deposit on (a)

Robertson (2009) and (b) Robertson (2016) Classification Young, uncemented soils tend to have KG* values
Charts, (Data obtained from Çanakkale Strait, Offshore ranging between 100 and 330, while soils with sig
European and Anatolian Towers). nificant microstructure tend to have KG* values
531
exceeding 330 (Robertson 2016). The small strain mudstones extends Zone 9 into Zones 4 and 5 with
stiffness of the Miocene deposits was measured a normalized friction ratio of 2% and higher. In add
using seismic CPT and wireline logging data. ition, the zone also covers small portions of Zones 6
Figure 8 presents the normalized tip resistance and 8 bounded by Fr (%) of 2% and Qtn of 200. The
(Qtn) versus small-strain rigidity index (IG) plot for newly proposed Miocene Zone represents the CPT
Holocene, Pleistocene and Miocene deposits in the data collected from the Miocene mudstones at the
Çanakkale bridge area. The plotted data in the figure site.
are from the European and Anatolian Tower explor
ations and cover a range of depths between 5 and
50m to represent the various ground units (Holocene,
Pleistocene and Miocene) and changes in stiffness.
The plot was made using representative small strain
stiffness values corresponding to measured shear
wave velocities and associated normalized tip resist
ances taken from the representative value of meas
ured tip resistances within the same sublayer where
the shear wave velocity was measured (about 1m
above and below the depth where the shear wave
velocity was measured).
Based on Figure 8, Holocene soils appear to have
little or no microstructure while the Pleistocene and
Miocene deposits likely have significant microstruc
ture. Hence, classifying these sediments requires
adjustments to the SBT classification systems pro
posed by Robertson (2009, 2016) to account for
microstructure.
Figure 9. Proposed Modification of the (a) Robertson

(2009) and (b) Robertson (2016) Classification Charts for
Miocene Deposits.
Figure 8. Qtn – IG chart to identify soils with microstruc 4 SUMMARY AND CONCLUSION
ture per Robertson (2016).
A Modified Soil Behaviour Type (SBT) classifica
tion is proposed to identify Miocene fine-grained
The CPT data obtained from the Miocene deposit deposits on the Qtn versus Fr (%) chart based on the
are characterized by: normalized tip resistances CPT data obtained from the 1915 Çanakkale bridge
between 25 and 100MPa and normalized friction project. The Miocene deposits were identified
ratios between 2% and 5% (Figure 6). Thus, a new through the performance of geophysical surveys, and
Zone associated with the Miocene mudstones on the boreholes drilled adjacent to CPT soundings. The
Qtn versus Fr (%) plot is proposed by modifying subsurface stratigraphy interpreted from the geo
Robertson (2009) chart. Basically, the modified zone physical survey is in good agreement with borehole
from SBT classification that corresponds to Miocene data, which enabled delineation of CPT data within
532
the Miocene deposits. The use of down-hole mode British Standards Institution 2015 BS 5930:2015 Code of
CPT allowed for the collection of significant data set practice for ground investigations. London: BSI.
from the Miocene deposits. International Society of Soil Mechanics and Geotechnical
While the Robertson charts have proven to be Engineering Technical Committee 1 (ISSMGE TC1)
effective for young uncemented soils, modifications 2005. Geotechnical & geophysical investigations for
appear to be warranted for sediments with significant offshore and nearshore developments. International
microstructure. Society for Soil Mechanics and Geotechnical
Engineering.
Within Miocene mudstones, inconsistencies were Fugro 2018a. 1915 Çanakkale bridge site characterization
noted between the soil classification of samples and report. Prepared for DLSY JV. Çanakkale: Fugro
those interpreted from CPT charts typically where Fugro 2018b. 1915 Çanakkale bridge engineering param
the CPT data plot in zones of the soil classification eters report. Prepared for DLSY JV. Çanakkale: Fugro
charts that are described as primarily coarse-grained. International Society of Rock Mechanics (ISRM) 2007,
Based on a comparison of the CPT data with the The complete ISRM Suggested Methods for rock char
cores and samples collected within the Miocene acterization, testing and monitoring: 1974–2006. In:
deposit, the materials characterized by friction ratios Ulusay R, Hudson JA (eds) Suggested Methods pre
pared by the Commission on Testing Methods, Inter
greater than about 2% that plot within these zones
national Society for Rock Mechanics, compilation
are mostly fine-grained and appear to display mainly arranged by the ISRM Turkish National Group. Ankara:
undrained responses during cone penetration. For Kozan Ofset.
this study, the Robertson (2009) chart has been Robertson, P.K. 1990. Soil classification using the cone
modified to delineate Miocene deposit with friction penetration test. Canadian Geotechnical Journal, 27(1):
ratios greater than 2 percent as fine-grained. Similar 151–158
modifications are expected to be applicable in simi Robertson, P.K. 2009. Interpretation of cone penetration
lar geological settings where fine-grained intermedi tests – a unified approach. Canadian Geotechnical Jour
ate geomaterials are present and are weak enough to nal, 46(11): 1337–1355.
permit meaningful penetration by commonly used Robertson, P.K. 2016 Cone Penetration Test (CPT)-Based
Soil Behaviour Type (SBT) Classification System—An
CPT equipment. The finding from this study should Update. Canadian Geotechnical Journal, 53(12):
facilitate the classification of similar deposits espe 1910–1927
cially when CPT soundings are performed at an Schmertmann, J.H. 1978. Guidelines for CPT: performance
early project stage prior to conducting boreholes. and design. Report FHWA-TS-78-209. Washington DC:
Federal Highway Administration.
Schneider, J.A., Randolph, M.F., Mayne, P.W., and
REFERENCES Ramsey, N.R. 2008. Analysis of factors influencing soil
classification using normalized piezocone tip resistance
ASTM. 2011. D2487-11. Standard practice for classifica and pore pressure parameters. Journal of Geotechnical
tion of soils for engineering purposes (Unified Soil Clas and Geoenvironmental Engineering, ASCE, 134(11):
sification System). ASTM International 1569–1586
533
© 2022 the Author(s), ISBN 978-1-032-31259-0
On the interpretation of piezocone dissipation testing data in clay

F.M. Mántaras
Geoforma Engenharia Ltda, Brazil
F.S. Pereira
University of Western Australia, Australia
E. Odebrecht
State University of Santa Catarina, Brazil
Geoforma Engenharia Ltda, Brazil
F. Schnaid
Federal University of Rio Grande do Sul, Brazil
ABSTRACT: Piezocone dissipation tests are routinely used to estimate the in situ coefficient of consolida
tion (ch). This is achieved either by extracting a single point of the dissipation curve (usually, the time for 50%
dissipation, t50) or by adjusting a theoretical solution to the experimental data. Results from a piezocone inves
tigation carried out in the Holocene clay deposit of Tubarão are reported and interpreted to evaluate the accur
acy of existing procedures to estimate ch. Explicit recommendations on how to curve fit dissipation curves and
use the first and second derivates to estimate t50 are presented. Validation of the proposed method is provided
from field tests reaching up to 75% dissipation. The proposed method does not require accurate measurements
of equilibrium pressure and relies less on engineering judgment when compared to reference procedures. Fur
thermore, the method performed reasonably well with a shorter dissipation time (40% dissipation).
1 INTRODUCTION are uncertainties in assessing both the initial pore

pressure ui and the equilibrium in situ pore pres
The coefficient of consolidation (ch) can be measured sure u0. Reliable measurement of the equilibrium
in laboratory from oedometer tests or assessed from in situ pore pressure, essential to the interpretation
in situ test results, preferably from piezocone dissipa of ch, is not always straightforward to obtain and
tion tests. In the case of piezocone data, evaluation of requires considerable engineering judgment in
ch is based on the variation in pore pressure with time many applications. In clays, the pore pressure gen
and interpretation rely either on one-dimensional erated in a piezocone test can be allowed to dissi
cavity expansion (Vésic, 1972; Torstensson, 1977; pate to the equilibrium value to assess the in situ
Randolph & Wroth, 1979; Burns & Mayne, 1998) or equilibrium pore pressure but this can take hours
two-dimensional strain path method (Levadoux & and makes the use of the piezocone economically
Baligh, 1986; Baligh & Levadoux, 1986; Teh & unfeasible for routine engineering applications. In
Houlsby, 1991). Both monotonic and dilatory soil cases such costal and nearshore investigation, tests
response can be modelled in this type of approach, are affected by tide variations and piezometer
being the Teh & Houlsby method (1991) the standard measurements have to be interpreted to estimate
for monotonic pore pressure dissipation response and average equilibirum pressures. In man-made struc
the Burns & Mayne method (1998) the alternative for tures such as tailing-retention dams the determin
dilatory response. ation of phreatic surface location is influenced by
The procedure for assessing ch from dissipation pond location, anisotropic permeabiity of deposits,
tests is based on a series of straightforward recom boundary flow conditions, permeable layers,
mendations that requires plotting normalized among other factors (e.g. Vick, 1983). Further
excess pore pressure (U = (ut – u0)/(ui – u0), more, equilibrium pore pressure will be higher than
against time on a log and/or √t scale, where ut is hydrostatic pressure in the presence of artesian
the pore pressure at time t and u0 is the in situ equi flow or in soil layers undergoing consolidation, and
librium pore pressure (Lunne et al, 1997). Albeit lower than hydrostatic pressure in the presence of
the procedure is apparently simple to apply, there donward flow (Robertson & Cabal, 2008).
DOI: 10.1201/9781003308829-76
534
Due to uncertainties in defining u0, the time for dissipation tests. The method’s mathematical reason
50% dissipation of excess pore pressure (t50) may be ing was also verified against a theoretical solution
in error and so is the estimated the value of ch. To for dilatory response (Burns & Mayne, 1998), which
partially overcome this problem, a recommendation is available in Mántaras et al., 2014. As dilatory
is made to adjust the complete dissipation curve to response is not in the scope of this paper, the deriv
obtain the best global value of the horizontal coeffi ation is not presented here.
cient of consolidation, instead of matching just Teh & Houlsby (1991) method: From a theoretical
a single point (usually 50%) of the recorded dissipa perspective, the strain path method was used to demon
tion (e.g. Lunne et al. 1997). The present paper strate that dissipation curves are not unique because the
extends this view by deriving the equation used to initial pore pressure distribution is highly sensitive to
curve fitting the measured pore pressure data. Once the value of Ir (Baligh, 1985). This prompted the nor
this is accomplished, the slope of the derivate at the malization of test data by means of dimensionless time
minimum point is zero and corresponds to t50. The factor T�, as introduced by Teh & Houlsby (1991):
time t50 can then be used in the interpretation of the
value of ch using standard methods.
A polynomial equation was shown to best-fit the
theoretical solutions of Teh & Houlsby (1991) and
Burns & Mayne method (1998), and the derivative
of the function is easily computed. Since the equilib where r is the probe radius, t the dissipation time (nor
rium in situ pore pressure is no longer required to mally adopted as t50), Ir the rigidity index (= G/Su)
calculate the percentage of dissipation, one of the and G the shear modulus. A constant Ir is used in the
uncertainties in deriving ch is eliminated. solution although in fact the value of the shear modu
These concepts have been applied to two long lus depends on the shear strain amplitude, which is
duration dissipation tests carried out in the Holocene shown by strain path calculations to vary in
clay deposit of Tubarão. Results show excellent a complex manner around a 60° penetrometer.
agreement between the reference and polynomial Excess pore pressure in the soil generated during
curve fitting methods in assessing ch. The polyno penetration of the piezocone can be expressed as
mial fitting method was shown to provide reasonable a negative function distribution for the probe radial dis
estimates of ch even for lower degree of dissipation. tance, with no excess pore pressure at an infinite dis
tance from the cone face. The rate of decay can be
conveniently represented using approximate numerical
2 THEORETICAL BACKGROUND algorithms, thus offering a means of implementing data
on a spreadsheet. Exponential and logarithm equations
The proposed method was first presented in Mántaras for matching the data have been proposed by Mayne
et al. (2014) and consists on curve fitting the field pore (2001), Chung (2014) and others (Fig. 1). The loga
pressure measurements to estimate t50. The fitting gen rithm equation proposed by Mayne (2001) to fit the
erates a standard curve that can be useful in interpolat data is:
ing a set of pore pressure values to define a function
that is smooth and comes close to the measured data.
Once the function is selected and the curve fitting is
completed, the function’s first derivative is calculated
and the minimum point of the function where the slope In a logarithmic time scale, the variable is
is zero is identified. In the second derivate this point is replaced by 10 (x-3):
zero. These two points (zero slope in the first derivate
and zero in the second) correspond to the point of
inflection of the normalized dissipation curve and
define the theoretical value of t50.
This simple mathematical procedure captures the
essence of physical models developed for the inter The derivative of the function with respect to the
pretation of piezocone dissipation data. Verification, variable x is expressed as:
justification, and reasoning is given through math
ematical derivation of the solutions proposed by Teh
& Houlsby (1991). A polynomial equation is used to
curve fit the data, with the actual mathematical
expression (degree) defined by the minimum r2. First
and second derivates correspond to the point of
inflection of the normalized dissipation curve and Alternatively, a polynomial expression is used to
define the theoretical value of t50. facilitate the mathematical treatment of the differenti
Basic arguments introduced by Mántaras et al. ation. Expression (4) could be accurately represented
(2014) are re-stated here before discussing the by an xth degree polynomial, providing means of fit
applicability of the proposed approach to piezocone ting both monotonic and dilatory soil response.
535
In the analysis, the program is instructed to graph
the curve to check the compatibility of the measured
data with the superimposed curve, as well to plot the
first and second derivates to calculate t50. The min
imum value of the polynomial first derivate corres
The first and the second derivates from equation ponds to a time factor T� =0.245 which is the
(5) can be easily obtained: theoretical value calculated by Teh & Houlsby (1991).
It is then concluded that the mathematical approach
of adjusting the complete dissipation curve, followed
by the determination of t50 from the first and second
derivates, is the simplest and most consistent way to
obtain the best global value of the coefficient of con
solidation, ch.
3 TESTING PROGRAM
The testing program was performed at the Experi

Equation 6 represents the slope of pore pressure dissi mental Testing Site located in the delta of Tubarão
pation in log scale and can be differentiated to obtain river in the Southern Coast of Brazil. Comprehensive
the inflection point of the u�2 xT� curve (Fig. 2). site investigation comprising seismic piezocone,
field vane shear tests, dilatometer tests and standard
penetration tests has been carried out to identify soil
type and stratigraphy (e.g. Mántaras et al., 2015).
Sediments in the upper 20m are predominantly nor
mally consolidated, formed during the last 8000
years in the Holocene period after the most recent
glaciation (Odebrecht & Schnaid, 2018; Schnaid &
Odebrecht, 2015; Lunne et al. 2014). Below a depth
of 6m the clay layer has an average plasticity index
(PI) of 55% (Schnaid et al., 2016).
A standard 10cm² piezocone (CPTu) with pore
pressure measurement on the shoulder just above the
cone face (u2) was used in the testing program. Tests
were carried out according to ASTM D5778 (2012)
standards. Figure 3 shows the penetration test results
obtained from the piezocone penetration test at the
Tubarão Site. The clay layer from 7 to 13m is rela
Figure 1. Teh & Houlsby solution (1991) and approxima tively homogeneous, with occasional lenses of silty
tion with logarithmic equation from Mayne (2001).
sand. After tests measurements show that the water
level fluctuates around 1m below the ground surface.
The penetration test was halted at the target
depths of 10 and 13m, where the soil is normally
consolidated. Up to 1.5h was allowed for pore pres
sure dissipation to obtain at least 75% of dissipation.
Table 1 summarizes the piezocone test program.
4 RESULTS
The normalized field excess pore pressure dissipation

curves for the tests CPTu_10m and CPTu_13m are pre
sented in Figure 4. The clay is normally consolidated,
exhibiting monotonically decreasing dissipation curves.
The monotonic excess pore pressure dissipations
were evaluated by best fitting the complete dissipation
curves using the Teh & Houlsby (1991) method. The
theoretical fitting provides good matching with the
experimental dissipation over the complete pore pres
Figure 2. 1st and 2nd derivates of Teh & Houlsby (1991) sure monotonic decay for both tests, and consequently,
solution. the interpreted ch values are considered as reference
536
The polynomial fitting method was used to com
pute t50 from tests CPTu_10m and CPTu_13m. First,
the measured piezocone pore pressure dissipation
data were best fitted with a polynomial equation of
8th degree in the logarithm of time plot. The first
and second derivate were then computed to obtain
the inflection point, which correspond to t50, as dis
cussed in section 2. Figure 5 illustrates the applica
tion of the polynomial fitting method to obtain t50
from tests CPTu_10m and CPTu_13m.
The ch values evaluated fromt50 extracted from
the polynomial method fall within the range of
values calculated with the reference method. For the
tests performed at 10m depth, ch was 7.32E-7m2/s,
whereas for the test carried out at 13m, ch was
4.55E-7m2/s. This represents a difference of less than
~10% between methods for both tests.
This set of complete dissipation tests shows that
the proposed mathematical fit provides estimates of ch
that compares well with values from the existing the
oretical solutions, enhancing the confidence in the
proposed method. The polynomial fitting method has
the advantage of being independent of an accurate
measurement of the equilibrium pressure and relying
less in engineering judgement in fitting the data with
Figure 3. Results from piezocone penetration test at theoretical solutions, which introduces uncertainties in
Tuburão Experimental Testing Site: (a) corrected cone tip the estimation of ch.
resistance, qt (b) pore pressure, u0 and u2 and (c) pore pres
sure parameter, Bq.
Table 1. Test program at the Tubarão clay site
Degree of dissipa
Dissipation Test time tion at end of test
test name Depth (m) (s)
CPTu_10m 10.00 3700 0.78

CPTu_13m 13.00 5470 0.75
Figure 4. Dissipation test at 10 and 13m depth.
for the Tubarão Site. The horizontal coefficient of con

solidation evaluated by the model was 7.79E-7m2/s and
4.10E-7m2/s for 10m and 13m depth respectively,
assuming the rigidity index was 100 (Odebrecht & Figure 5. Application of the polynomial fitting method for
Schnaid, 2018). tests CPTu_10m and CPTu_13m.
537
5 INCOMPLETE PIEZOCONE DISSIPATION reference ch (ch,ref), normalized by ch,ref,and
TESTS expressed as percent, [(ch,est – ch,ref)/ ch,ref]. The ch,ref
is the one obtained from the complete available data
Previous interpretation was centered in complete dissi interpreted using the reference method. Figure 7
pation tests where the degree of dissipation was greater shows the percentage error ratio against truncated
than 75%. It is then necessary and economically con data sets at different degrees of dissipation.
venient to evaluate incomplete CPTu dissipation tests
where ch values are estimated for early degrees of dis
sipation. Previous researchers have already considered
the alternative of interpreting incomplete dissipation
test (e.g. Schnaid et al, 1997; Krage et al, 2015),
because significant time savings allow for more pore
pressure dissipation tests in a site investigation plan.
The predictive approach proposed to estimate t50,
and therefore ch, from measured cone dissipation tests
was similar to the one described by Krage et al.
(2015). The polynomial function was fitted to each dis
sipation test using truncated data sets from 40 to 70%
dissipation. The optimal fitting equation was then
derived to estimate t50 for each set of partial dissipation
data, which was then used to estimate ch (ch,est). For Figure 7. Percentage error ratio against truncated data sets
data set truncated at dissipation degree less than 50%, from 40 to 70% dissipation for tests CPTu_10m and
a point simulating equilibrium pore pressure in a very CPTu_13m.
long time (t=100,000s; u2=u0) was added to obtain
a polynomial equation with a realistic dissipation like
The percentage error ratio to ch,ref is less than 25%
shape. An example is provided in Figure 6, where the
in all cases and less than 20% in most cases, even
fitted equation for test CPTu_13m truncated at 40% of
with only 40% dissipation of excess pore pressure. It
dissipation is shown. The polynomial equation fits well
is worth recalling that the Ir value has a direct effect
the dissipation field data up to 40% dissipation but
on the estimated absolute value of ch (e.g. Keaveny
diverges after, as expected, since the data for dissipa
& Mitchell 1986, Houlsby & Wroth 1991, Robertson
tion higher than 40% was not used to derive the poly
et al. 1992, Schnaid et al. 1997, Mayne, 2001) and
nomial equation. Despite this, ch,est was 3.21E-7m2/s,
the uncertainty in selecting an appropriate Ir value
representing a difference of only 22% when compared
exceeds the uncertainty in predicting t50 results once
to the reference value.
dissipation test exceeds 40%. The method seems to
have potential to reduce the required time in field to
estimate ch, although further validation is required.
6 CONCLUSIONS
This paper presents an alternative method for t50

computation by fitting the piezocone pore pressure
dissipation curve with a polynomial equation. Expli
cit recommendations are given for the fitting tech
nique, including using the first and second derivates
to calculate t50. The proposed method has the advan
tage of being independent of the equilibrium pore
pressure measurement and requiring less engineering
judgement. Furthermore, the proposed method
seems to provide ch estimates within 25% accuracy
from incomplete dissipation test halted after only
40% dissipation (in respect to the ch using
a complete dissipation test and reference method).
Figure 6. Application of the polynomial fitting method for

tests CPTu_13m truncated at 40% of dissipation.
REFERENCES
ASTM. (2012). “Standard Test Method for Electronic Fric
The percentage error ratio for ch,est was calculated tion Cone and Piezocone Penetration Testing of Soils.”
as the numerical difference between predicted and ASTM D5778-12, West Conshohocken, PA.
538
Baligh, M. M.; Levadoux, J-N. (1986). “Consolidation Mántaras, F. M.; Odebrecht, E.; Schnaid, F. (2015). “Using
after undrained piezocone penetration II: Interpretation.” piezocone dissipation test to estimate the undrained
J. of Geotechnical Engineering, 10.1061/(ASCE)0733 shear strength in cohesive soil.” Canadian Geotechnical
9410(1986)112:7(727) Journal, 52(3),318–325.
Burns, S. E.; Mayne P. W. (1998). “Monotonic and Dilatory Mayne, P.W. (2001). “Stress-Strain-Strength-Flow Param
pore-pressure decay during piezocone tests in clay.” eters from Enhanced In-Situ Tests.” Proc., International
Canadian Geotechnical Journal, 35(6),1063–1073. Conference on In-Situ Measurement of Soil Properties
Chung, S. G., Kweon, H. J., and Jang, W. Y. (2014). “Hyper and Case Histories, Bali, Indonesia, 27–48.
bolic fit method for interpretation of piezocone dissipation Odebrecht, E; Schnaid, F. (2018). “Assessment of the stress
test.” J. of Geotechnical and Geoenvironmental Engineer history of Quaternay clay from piezocone tests.” Soils
ing, 10.1061/(ASCE)GT.1943-5606.0000967, 251–254. and Rocks, 41(2),179–189.
Houlsby, G.T., and Wroth, C.P. (1991). “The variation of Randolph, M.F.; Wroth, C.P. (1979). “An analytical solu
shear modulus of a clay with pressure and overconsoli tion for the consolidation around a driven pile”. Inter
dation ratio.” Soils and Foundations, 31(3),138–143. national Journal for Numerical and Analytical Methods
Keaveny, J., and Mitchell, J. K. (1986). “Strength of in Geomechanics, 3(3),217–229.
fine-grained soils using the piezocone.” Proc., Use of in Robertson, P.K.; Cabal, K.L. (2008). “Guide to cone pene
Situ Tests in Geotechinical Engineering, ASCE, Reston, tration testing for geo-environmental engineering”.
VA, 668–685. Gregg Drilling & Testing Inc.
Krage, C. P., DeJong, J. T., and Schnaid, F. (2015). “Esti Schnaid, F.; Sills, G.C.; Soares, J.M.; Nyirenda, Z. (1997).
mation of the Coefficient of Consolidation from Incom “Predictions of the coefficient of consolidation from
plete Cone Penetration Test Dissipation Tests“. J. of piezocone tests.” Canadian Geotechnical Journal, 34
Geotechnical and Geoenvitonmental Engineering. (2),143–159.
10.1061/(ASCE)GT.1943-5606.0001218. Schnaid, F.; Odebrecht, E. (2015). “Challenges in the Inter
Levadoux, J-N.; Baligh, M. M. (1986). “Consolidation pretation of the DMT in Tailings.” Proc., 3th Inter
after undrained piezocone penetration I: Prediction”. national Conference on the Flat Dilatometer, Rome.
J. of Geotechnical Engineering, 10.1061/(ASCE)0733 Schnaid, F.; Odebrecht, E.; Sosnoski, J.; Robertson, P.K.
9410(1986)112:7(707). “Effects of test procedure on flat dilatometer test (DMT)
Lunne, T.; Robertson, P. K.; Powell, J. J. M. (1997). Cone results in intermediate soils.” Canadian Geotechnical
Penetration Testing in Geotechnical Practice. Blackie, Journal, 53(8),1270–1280.
Melbourne, Australia. Teh, C. I.; Houlsby, G. T. (1991). “An analytical study of
Lunne, T.; Yang, S.; Schnaid, F. (2014). “CPT the cone penetration test in clay.” Géotechnique, 41
Interpretation.” Proc., 3th International Symposium on (1),17–34.
Cone Penetration Testing, Las Vegas, 145–164. Torstensson, B. A. (1977). “Time-dependent effects in the
Mántaras, F. M.; Odebrecht, E.; Schnaid, F. (2014). “On field vane test”. Proc., International symposium of soft
the Interpretation of Piezocone Dissipation Testing clay, Bangkok, 387–397.
Data.” Proc., 3th International Symposium on Cone Vick, S. G. (1983). Planning, Design, and Analysis of Tail
Penetration Testing, Las Vegas, 315–322. ings Dams. Wiley, New York.
539
© 2022 the Author(s), ISBN 978-1-032-31259-0
Determination of fine-grained soil parameters using an automated system

I. Marzouk & F. Tschuchnigg
Graz University of Technology, Institute of Soil Mechanics, Foundation Engineering and Computational Geotech
nics, Graz, Austria
F. Paduli, H.J. Lengkeek & R.B.J. Brinkgreve

ABSTRACT: Performing numerical analysis successfully depends on several factors. One of the most import
ant factors is determining the constitutive model parameters correctly. It is often the case that these parameters
are determined based on limited soil data. Using in-situ tests for determining these parameters has several advan
tages such as minimal disturbance of the soil and lower cost compared to laboratory tests. However, it is not
possible to determine soil parameters directly from in-situ tests results. Thus, empirical correlations are required
for interpreting soil parameters. Generally, several correlations exist for the same parameter, which will lead to
calculating several values for the same parameter. An ongoing research project focuses on formulating an auto
mated parameter determination (APD) framework that uses a graph-based approach to identify constitutive model
parameters based on in-situ tests. This is achieved by using two spreadsheets as an input, one for parameters and
the other for equations (correlations used to calculate parameters). Based on these two spreadsheets, the system
generates paths between the parameters and calculates the value(s) for each individual parameter. So far, the
research project focused on determining the parameters for coarse-grained soil based on cone penetration test
(CPT) results. Due to the fact that the system was set up in a modular and adaptable way, it is possible to expand
the system to accommodate more soil types and in-situ tests. It is the aim of the research project to increase the
reliability of the parameters values (required to perform numerical analysis) determined from in-situ tests. This
paper focuses on expanding the current framework to determine parameters for fine-grained soil. By using the
two spreadsheets as an input, the system successfully calculates the value(s) for fine-grained parameters. Further
validation, dealing with several values for each parameter, determining the accuracy of derived parameters and
expanding the system to accommodate other in-situ tests and types of soils are part of ongoing research.
1 INTRODUCTION disturbance of the soil and lower cost compared to

laboratory tests. The main disadvantage of the inter-
There are several reasons that make the use of numer- pretation of in-situ tests is that parameters cannot be
ical analysis preferable compared to the traditional determined directly from the results of the tests as the
methods. One of the main advantages is the level of laboratory tests. However, a number of empirical rela
detail that can be obtained in several geotechnical tionships exist that link soil parameters to in-situ tests
engineering problems such as soil-structure interaction results, it is often the case that several relationships
Brinkgreve (2019). Several factors influence the suc- exist to determine the same parameter, which lead to
cess of the numerical analysis. One of the most import- a wide range of values for the parameter of interest.
ant factors is determining the constitutive model The reason for this variation is mainly related to the
parameters properly. The main challenge in determin- fact that these relationships are not applicable for all
ing these parameters is the limited available soil data. situations (e.g., specific soil types). In literature, sev-
It is often the case that these parameters need to be eral guides exist dealing with the interpretation of CPT
defined based on experimental tests (e.g., triaxial and such as Kulhawy & Mayne (1990), Lunne, Robertson,
oedometer tests) which are not always available in all & Powell (1997), Mayne (2014) and Robertson
projects. (2015).
On the other hand there are in-situ investigations, An ongoing research project focuses on creating an
where the cone penetration test (CPT) is one of the automated parameter determination (APD) system to
most popular in-situ tests as it is quick and often used determine constitutive model parameters based on in-
in soil profiling and estimating soil parameters. More- situ tests. The framework relies on a graph-based
over, CPT has other advantages such as minimal approach that uses some of the characteristics of graph
DOI: 10.1201/9781003308829-77
540
theory. The project aims to create a transparent and an The paper focuses on the output of module 4 at
adaptable parameters determination framework. Trans a specific depth. The layering process is not considered
parency is achieved by illustrating how the available (modules 2 and 3) in this contribution. Moreover, the
information is used to compute parameters and adapt paper only presents the determination of fine-grained
ability is achieved by allowing the users of the system soil parameters (output of module 4) without the tran
to incorporate their knowledge and experience into the sition to constitutive model parameters (module 5).
system. Van Berkom et al. (2022) illustrated the deter
mination of parameters for coarse-grained soils based
on CPT data. This paper extends the framework pre Table 1. SBT zones according to Robertson (2010).
sented in Van Berkom et al. (2022) by including
parameters for fine-grained soils. Zone Soil Behaviour Type (SBT)
The 2nd section briefly describes the APD frame
1 Sensitive fine-grained
work, while the 3rd section presents selected empir
2 Clays – organic soil
ical relationships used to determine parameters for
3 Clays: clay to silty clay
fine-grained soils. In the 4th section, the output of the
4 Silt mixtures: clayey silt & silty clay
APD for a simple example is illustrated. In the final
5 Sand mixtures: silty sand to sandy silt
section the conclusions of this study are summarized.
6 Sands: clean sands to silty sands
7 Dense sand to gravelly sand
8 Stiff sand to clayey sand (overconsolidated)
2 AUTOMATED PARAMETER
9 Stiff fine-grained (overconsolidated)
DETERMINATION (APD) FRAMEWORK
2.1 Framework 2.2 SBT interpretation

The framework consists of several modules that are Robertson (2010) modified non-normalized SBT chart
connected together. A schematic representation of is used to classify the CPT profile. This SBT chart is
the modules is shown in Figure 1. CPT raw data, in based on dimensionless cone resistance, (qc =pa ),
Geotechnical Exchange Format (GEF) are imported where pa is the atmospheric pressure and friction ratio
to the first module (GEF Reader). Afterwards the (Rf in percent, Rf ¼ fs =qc 100%). The chart consists
CPT measurements (cone resistance qc , sleeve fric of 9 different zones, each corresponding to a different
tion fs & porewater pressure readings u2 ) are passed soil behaviour type (Table 1). At each depth, qc and
to the second module (CPT layer interpretation). Rf are used to access the chart and determine the
The second module determines the SBT based on SBT for this depth. As a result, this module is used to
Robertson (2010) modified non-normalized SBT distinguish between fine and coarse-grained soils.
chart and stratifies the CPT profile into several
layers sharing the same SBT. For each layer, the
average of the CPT measurements (qc , fs & u2 ) 2.3 Graph-based approach
within this layer is computed. The averaged CPT The graph-based approach used in APD is described
measurements are used by module 3 (Layer state), to in detail in Van Berkom et al. (2022) and illustrated
determine the state of all layers (overconsolidation in Figure 2. The idea is to create links between
ratio OCR and coefficient of earth pressure K0 ). The source parameters (CPT raw data) via intermediate
output of modules 2 and 3 is transferred to module parameters to destination parameters (final soil or
4, where the parameters are connected with the equa model parameters). Based on a given set of correl
tions (correlations) and the parameters of interest are ations, the system will create all the paths (chains of
calculated. In the final module, parameters calculated correlations) that provide the link from the source
in module 4 are converted to constitutive model parameters all the way to the destination parameters
parameters. The system is built in the programming and the system will calculate the destination param
language Python. eter values from the input values of the source
parameters (CPT data).
In the APD framework, the terms ‘correlation’,
‘formula’, ‘equation’, ‘rule of thumb’ is replaced by
the term ‘method’. This general term is used as
parameters could be determined based on several
ways (e.g., tables and charts) (Van Berkom et al.
(2022)). The system must link the methods and
parameters that share a relationship. As an example,
a method to compute the coefficient of earth pressure
at rest according to Jaky (1944) is defined as follows,
K0 ¼ 1 - sinð0 Þ, where K0 is the coefficient of
Figure 1. Schematic representation of the parameter deter earth pressure at rest and 0 , is the effective internal
mination modules.
friction angle of the soil. The system must identify
541
the input and output for this method (the output is K0 method is only valid for silt, the validity would
and the input is 0 ). Consequently, links connecting be SBT(4). Regarding the method of coefficient
these parameters should be generated. of earth pressure at rest, the validity would be
SBT(1234567). The reference field is an optional
argument, where the user could state the author
of the method (e.g., Jaky_1944).
The parameters CSV file requires the following
properties, symbol, value, unit, constraints, and
description. All of the parameters that have been used
in the methods CSV files (in the fields of formula,
parameters_in and parameters_out) must be defined in
the parameters CSV file. The notation of the parameter
(which was used in the methods CSV file) is stated in
the symbol field (e.g., u for porewater pressure). In
case the user wants to fix a value for a parameter (e.g.,
unit weight of water), the value field is used for this
purpose. The unit field is an optional argument where
the user could specify the unit of the parameter. It is
highly recommended to provide the unit for all param
eters to avoid unit conversion mistakes (e.g., using qc
in MPa in a method that requires qc in kPa). Lower
Figure 2. Graph-based approach implemented in APD. and upper bounds could be applied to parameters
through the constraints field. Any computed value
lower than the lower bound or higher than the upper
2.4 Generating the graph bound would be discarded for the given parameter. The
As shown in the previous subsection, the relation description field is an optional argument, where the
ships between methods and parameters are defined user could define the parameter (e.g., OCR is the over-
by the output and input(s) of different methods. The consolidation ratio).
parameters and methods are considered as external By formulating the two CSV files (methods and
inputs to the system. The system requires two input parameters) as described, the system imports the two
files: methods and parameters. Users of the system files and forms links between the methods and param
may extend the standard database of methods and eters (parameters_in & parameters_out) that are
parameters provided with the system. The system related together. The output of this procedure is a graph
connects the methods and parameters together, and showing the links between all the defined parameters
computes the intermediate and destination param and methods. Moreover, the computed values for dif
eters. Two different spreadsheets in comma-separate ferent parameters are shown on the graph. The current
values (CSV) format corresponding to parameters version of APD contains more than 100 methods.
and methods are used to generate the graph.
Each of the two CSV files has special proper
ties. The methods CSV file requires the following 3 SELECTED CPT FINE-GRAINED SOIL
properties, method_to, formula, parameters_in, CORRELATIONS
parameters_out, validity and reference. Each of
these unique properties need to be provided by A standard validated database for methods and
the user in a CSV file. Taking the coefficient of parameters has been compiled and is continuously
earth pressure at rest method presented in the pre updated and improved. However, users are respon
vious subsection as an example, method_to would sible for validating the outcome of the system, even
present the name of the method, in this case it if they used the provided standard database. Users
might be method_to_K0. In the field of formula, still need to apply their geotechnical experience and
the equation should be defined, 1 - sinð0 Þ. Para knowledge to the outcome. Nevertheless with
meters_in implicitly states the input for this limited geoetechnical knowledge, the system should
method, 0 . Similar to parameters_in, the output result in reasonable values for different parameters.
of the method is stated in the field of parame In this section, some methods for different fine-
ters_out, K0 . The validity field specifies the grained soil parameters are presented. These
applicability of different methods. Some methods methods and parameters are used to generate the
are applicable for all types of soils, other methods graph in the following section.
are only valid for coarse-grained soils and others
are only suitable for fine-grained soils. As shown 3.1 Unit weight
in Table 1, the SBT is based on Robertson (2010)
modified non-normalized SBT chart. In that sense, The calculation of the total unit weight (γt ) is
the validity is defined in terms of SBT. If the required to compute the total and effective vertical
542
stress, that are important in many correlations
between CPT results and soil parameters. The
selected correlations for estimating the unit weight
in the APD system are:
by Robertson & Cabal (2010), where γw is the unit

weight of water and qt is the corrected cone resist 3.3 Strength parameters
ance (defined as qt ¼ qc þ ð1 - aÞ x u2 , where a is
the cone tip net area ratio). The following correlation0 is used to determine the
effective friction angle ( ) in the APD system:
by Mayne et al. (2009), where Bq is the normalized

by Lengkeek, de Greef, & Joosten (2018). porewater pressure (Bq ¼ ðu2 - u0 Þ=ðqt - σv Þ) and
Qt is the normalized cone resistance
The valid range for this correlation is
0
0:1 � Bq � 1:0 and 20 � � 45.
3.4 Stiffness parameters
by Mayne (2014). The 1-D constrained tangent modulus, M is used to

estimate settlements. The following correlation is
used to determine the constrained modulus in the
3.2 Stress history APD system:
The stress history is often represented by the over
0
σ
consolidation ratio (OCR ¼ σp0 , where is the pre
v
consolidation stress and is the effective vertical

stress). The selected correlations for estimating OCR Robertson (2009) suggested an approach based on Ic
in the APD system are: to determine αM as follows:
by Mayne et al. (2009), where σv is the total vertical

stress and is the yield stress exponent that
increases with fines content and decreases with mean
grain size. Mayne (2017) proposed determining
from CPT material index Ic as follows: 4 DETERMINING FINE-GRAINED SOIL
, where Ic is determined by an PARAMETERS
iterative process Robertson (2009) based on nor In this section, an example of the output of the
malized cone parameter (Qtn ) with variable stress system is presented. The methods CSV file used for
exponent (n) that varies with Ic ). this example, contains the correlations presented in
the previous section, as well as other formulas used
to compute some intermediate parameters (e.g.,
methods to calculate qt ,Bq ,Rf ,Qt ). The parameters
CSV file includes all the parameters defined in the
methods CSV file.
by Kulhawy & Mayne (1990) and Robertson (2009), The system imported a CPT GEF file and deter
where: mined the SBT at each depth. The interpreted SBT at
543
each depth is shown in Figure 3. For generating The unit weight of water (gamma_w) is defined as
the graph, a CPT measurement at a depth of 10 kN=m3 . The atmospheric pressure (pa ) corresponds
10 m (z ¼ 10 m) was chosen (Figure 3). This to 100 kPa. The interpreted SBT is 3, therefore, the
measurement has the following properties, soil type at this depth is clay (according to Table 1).
qc ¼ 1015:5 kPa, fs ¼ 31:5 kPa and u2 ¼ 351:6 kPa. The generated graph is shown in Figure 4.
The ground water level (GWL) is located at The graph consists of green and blue nodes. The
6 m below the ground level. The cone tip net area ratio green nodes correspond to parameters, while the
is provided in the CPT GEF file as 0.85 (a ¼ 0:85). blue nodes correspond to methods. The arrows
between different nodes, show the link between dif
ferent entities (parameters and methods) within the
system. The arrows have a defined direction (going
from a parameter to a method or from a method to
a parameter).
Focusing on the unit weight of the soil (gam
ma_sat located at the lower left corner in
Figure 4), it is clear that three methods contribute
to gamma_sat. The methods correspond to the
three correlations presented in the previous sec
tion, where method_to_gamma_sat_1 is Equa
tion 1, method_to_gamma_sat_2 is Equation 2
and method_to_gamma_sat_3 is Equation 3.
Three values were computed respectively as,
17:33, 16:25 and 17:09 kN=m3 . Moving to OCR
(located at the lower right corner in Figure 4),
two methods contribute to OCR, where method_
to_OCR_1 corresponds to Equation (4) and meth
od_to_OCR_2 corresponds to Equation (5). Two
values were computed respectively as 2:14 and
2:18. The friction angle (phip located at the lower
part in Figure 4) is obtained by only one method
(method_to_phip) corresponding to Equation (6).
The friction angle was computed as 30.99. Simi
lar to the friction angle, the constrained modulus
(M_CPT located at the right-hand side of the
graph in Figure 4) is obtained by only one
method (method_to_MCPT) corresponding to
Equation 7. The constrained modulus was com
puted as 5903 kPa.
As discussed in Equation 4, Ic , Qtn and n are
determined through an iterative process. This itera
tive process requires the knowledge of the total and
Figure 3. Interpreted SBT at each depth. effective vertical stress. As a result, an initial
Figure 4. An example of a graph.
544
estimate for the unit weight is required to compute between soil parameters and Plaxis Hardening Soil
these parameters. In that sense, Equation 1 is used to model with small-strain stiffness (HSsmall) (Benz
compute an initial value for the unit weight, which (2007)). This is one of the main aspects of the
in turn, is used to calculate the total stress (sigma_tot research project as it will allow the transition from
in Figure 4), effective stress, Ic , Qtn and n. Conse the CPT measurements to constitutive model param
quently, it might be noticed from Figure 4 that Qtn eters that could be used directly for numerical
and Ic (located at the top right corner in 4) are used analysis.
directly as source parameters because they were cal
culated in a previous step internally before the graph
was generated. Therefore, Equations 2 and 3 are REFERENCES
only used to compute gamma_sat for comparison
purposes and they do not influence the calculation of Benz, Thomas (2007). Small-strain stiffness of soils and its
numerical consequences. Ph. D. thesis, University of
the total and effective stress. As the system is formu
Stuttgart, Germany.
lated in an adaptable way, the user can decide which Brinkgreve, R.B.J. (2019). Automated model and param
correlation for the unit weight to be used for the ini eter selection. Geostrata, 41–47.
tial estimate for the total, effective stress and for the Jaky, J. (1944). The coefficient of earth pressure at rest. In
calculation of Ic , Qtn and n. Hungarian (a nyugalmi nyomas tenyezoje). J. Soc.
Hung. Eng. Arch. (Magyar Mernok es Epitesz-Egylet
Kozlonye)), 355–358.
5 CONCLUSIONS Kulhawy, F.H. & Mayne, Paul (1990). Manual on Estimat
ing Soil Properties for Foundation Design.
Lengkeek, H. J., de Greef, J., & Joosten, Stan (2018). CPT
This paper is an extension to the automated param
based unit weight estimation extended to soft organic
eter determination system presented in Van Berkom soils and peat. Cone Penetration Testing 2018.
et al. (2022). The previous section presented proof Lunne, Tom, Robertson, P., & Powell, John (1997). Cone
of concept where a graph-based approach was used Penetration Testing in Geotechnical Practice. Soil Mech
to calculate parameters for fine-grained soil. The pre anics and Foundation Engineering 46.
sented system is transparent, flexible, and adaptable Mayne, Paul (2014). Interpretation of geotechnical param
where the users can incorporate their experience and eters from seismic piezocone tests. Proceedings, 3rd
knowledge into the system by extending the standard International Symposium on Cone Penetration Testing
database of methods and parameters provided with (CPT14, Las Vegas) 102, 47–73.
Mayne, Paul (2017). Stress History of Soils from Cone
the system. The research project aims to increase the
Penetration Tests. Soils and Rocks 40, 203–216.
confidence in the parameters values (required to per Mayne, Paul, Coop, M, Springman, Sarah, Huang, An-Bin,
form numerical analysis) determined from in-situ & Zornberg, Jorge (2009). State-of-the-art paper (SOA
tests. 1): geomaterial behavior and testing.
Figure 4 presented a simple example where Robertson, P. (2009). Interpretation of cone penetration
a limited number of methods were used. In case of tests – A unified approach. Canadian Geotechnical
using several methods, this will lead to a scatter for Journal 46, 1337–1355.
the computed parameters. Dealing with this scatter Robertson, P.K. (2010). Soil Behaviour Type from the
and determining which approach is more suitable for CPT: An Update. 2nd International Symposium on Cone
Penetration Testing, Huntington Beach 2, 575–583.
choosing a specific value from the range of the com
Robertson, P.K. (2015). Guide to cone penetration testing
puted values is part of an ongoing research. In add for geotechnical engineering. Proceedings, 3rd Inter
ition, other SBT charts (e.g., Robertson (2009)) national Symposium on Cone Penetration Testing
normalized SBTn chart and Robertson (2016) SBT (CPT14, Las Vegas).
chart) are added to the system. Moreover, the com Robertson, P.K. & Cabal, K.L. (2010). Estimating soil unit
piled correlations database is continuously validated, weight from CPT. 2nd International Symposium on
updated and the output of different correlations is Cone Penetration Testing, Huntington Beach.
compared to laboratory tests results whenever they Robertson, P. K. (2016). Cone penetration test (CPT)-based
are available. Correlations for calculating typical soil behaviour type (SBT) classification system — an
update. Canadian Geotechnical Journal 53(12).
fine-grained soil parameters (e.g., plasticity index,
Van Berkom, I.E., Brinkgreve, R.B.J., Lengkeek, H.J., &
PI, liquid limit, LL, compression index, Cc and De Jong, A.K. (2022). An automated system to deter
swelling index, Cs ) were also added to the database. mine constitutive model parameters from in situ tests.
Furthermore, the connection between soil parameters Proceedings of the 20th International Conference on
and constitutive model parameters is to be estab Soil Mechanics and Geotechnical Engineering, Sydney
lished. The database includes several correlations 2022. To be published.
545
© 2022 the Author(s), ISBN 978-1-032-31259-0
Undrained shear strength of clays from piezocone tests: A database

approach
P.W. Mayne
Georgia Institute of Technology, Atlanta, GA, USA
J. Peuchen
ABSTRACT: Deriving undrained shear strength of clays from piezocone tests (CPTU) suits the use of
a database approach, particularly because good correlations are expected on the basis of theoretical corres
pondence. Benefits of a database approach include a minimal environmental footprint and field schedule
shortening, primarily because of reduced borehole sampling and laboratory testing focused on verification
rather than development of comprehensive site-specific data sets. This paper expands on the classic expression
su = qnet/Nkt where su is a reference undrained shear strength obtained by laboratory testing, qnet is the net
cone resistance and Nkt is a cone bearing factor that is noted to decrease with increasing values of CPTU pore
pressure ratio Bq. The database includes CPTU results and high-quality laboratory triaxial compression tests
from 70 different clay deposits, of which 8 represent new case studies. The clays are allocated to 5 main cat
egories: (a) soft-firm offshore; (b) soft-firm onshore; (c) soft sensitive; (d) stiff overconsolidated intact; and
(e) stiff fissured clays. Organic clays and cemented clays are excluded.
1 INTRODUCTION Here, an empirical database approach is con

sidered for Nkt. The database approach uses data
When a geotechnical exploration discovers that clay from high quality piezocone penetration test (CPTU)
forms all or a portion of the subsurface environment, results matched at the same elevations as high qual
the magnitude of the undrained shear strength (su) ity samples subjected to laboratory testing.
parameter is generally sought for input into calcula
tion models involving ground stability, particularly
related to shallow foundations, pilings, and slopes. 2 DATABASE PARAMETERS
Undrained shear strength can also be important for
transitional soils and sands, where combinations of 2.1 Triaxial compression tests
geometry and loading rate can lead to undrained soil
response. The reference undrained shear strength was defined
Undrained shear strength is not a unique prop as a triaxial compression mode, designated suc:
erty of clays but affected by many variables, • Derived value (Eurocode 7) of suc from laboratory
including mode of shearing, rate of loading, shear tests on Class 1 samples (ISO 22475-1:2006);
direction, initial stress state, failure criterion and • Anisotropically-consolidated undrained triaxial
other factors (Mayne 2008). This paper expands compression tests (CAUC or CK0UC) according
on the most common expression su = qnet/Nkt to ISO 17892-9:2018 or equivalent; Note for fis
where su is a reference undrained shear strength, sured clays, often only CIUC tests were
qnet is net cone resistance and Nkt is a cone bear available;
ing factor according to classical bearing capacity • Recompression to the estimated in-situ stress con
theory. Values of Nkt can be obtained from ana ditions, using conventional back pressures for
lytical, theoretical, or numerical solutions, such specimen re-saturation;
as those based on limit plasticity, cavity expan • suc = ½(σ1-σ3)max defined as failure criterion or
sion, finite elements, and strain path method. (σ10 =σ30 )max as failure criterion when (σ1-σ3) pro
Well over 50 solutions are available for Nkt (e.g., vides no distinct maximum (Ladd & DeGroot
Lunne, et al. 1997; Yu & Mitchell 1998; Col 2003; Lade 2016), where σ10 and σ30 are the effect
reavy 2016; Agaiby 2018). ive principal stresses.
DOI: 10.1201/9781003308829-78
546
2.2 Piezocone penetration tests and expressed by Equation 1 (Mayne & Peuchen
2018) for values of Bq > -0.1:
For the CPTU (ISO 22476-1:2012), three separate
measurements are obtained: (a) corrected cone resist
ance, qt; (b) sleeve friction, fs; and (c) pore pressure,
u2. These measurements are acquired at depth inter
vals of between 10 mm and 50 mm during a constant
vertical push rate of 20 mm/s. Comments are as follows:
For each elevation, values of qnet and pore pres
• An advantage of this approach over many other
sure ratio Bq were derived:
solutions for Nkt is that the CPTU provides all the
• qnet = qt - σvo, where σvo = total vertical overbur necessary input;
den stress; • The methodology covers a wide range of clays
• Bq = Δu2/qnet, where Δu2 = u2 - u0 and u0 = showing Bq > -0.1, including soft to firm to stiff
hydrostatic pressure. clays which vary from sensitive to insensitive,
and intact to fissured;
Of further note, fs can be used to provide an • The approach does not apply to organic or
evaluation of the soil unit weight (γ) needed in the cemented clays.
calculation of σvo (Mayne et al. 2010a, 2010b;
Mayne & Peuchen 2012). Consequently, all three Since the initial database findings, a number of
readings (qt, fs, u2) are utilized in the assessment of new case studies have become available that permit
suc of clays. a validation of Equation 1. Herein, triaxial and CPTU
data from 8 clays from Europe, Asia, and North
America are presented. Three sites are from offshore
3 TRIAXIAL - PIEZOCONE DATABASE locations and five clays are onshore deposits.
A total of 62 natural clays that were subjected to

CAUC lab testing (n = 407) provided the initial basis 4 NEW CASE STUDIES
for this study (Mayne 2014). The majority of clays
were deposited in a marine environment, although 4.1 Luva, Norwegian sea
a few were lacustrine or alluvial or deltaic in origin.
A few sensitive clays were originally formed as sedi Luva is an offshore gas reserve located in 1300 water
ments in salt-water and later exposed to leaching by depth of the Norwegian sea. The site consists of very
freshwater. soft plastic clays having sensitivities in the range of 2
The clay sites were classified into 5 separate groups to 5. Index testing indicates a natural water content wn
(Mayne & Peuchen 2018), including: (a) 17 offshore ≈ 65 % to 75 %, liquid limit wL ≈ 70 %, plastic limit
clays that were normally-consolidated (NC) to lightly wP ≈ 29 %, and plasticity index Ip ≈ 41 % (Lunne
overconsolidated (LOC); (b) 29 onshore clays that et al. 2014). Series of consolidation tests indicate that
were also NC - LOC, (c) 6 soft sensitive clays; (d) 5 the site has not been mechanically overconsolidated,
intact overconsolidated clays (OC); and (e) 5 fissured showing yield stress ratios (YSR = σp0 /σvo0
) in the gen
OC clays. eral range of 1.2 to 1.7, primarily due to ageing.
From the database approach, the trend of the rela Figure 2 shows qt and u2 with depth below the
tionship for evaluating Nkt is presented in Figure 1 seafloor, pore pressure ratio Bq, and derived profile
of suc from the Nkt relationship. In the last graph, the
CPTU results are shown in comparison with 42
CAUC triaxial tests with good agreement.
4.2 Sipoo, Finland

Sipoo is located about 30 km north of Helsinki. The
site consists of a homogeneous soft clay deposit
between 2 and 9 m depth and water table near the
surface (DiBuò et al. 2019). Index tests on the clay
include: wn = 101 ± 12 %, wL = 79 ± 10 %, wP = 30
± 1 %, Ip = 50 ± 9 %, and clay fraction CF = 79 ±
11 %. Laboratory CRS consolidation tests give
a mean YSR = 1.76 and sensitivity by laboratory fall
cone ranges from 15 to 44 with a mean St = 25 ± 8.
Results from a representative CPTU at Sipoo are
shown in Figure 3. The CPTU-evaluated profiles of
suc compare well with the four laboratory CAUC test
Figure 1. Trend of Nkt with Bq (Mayne & Peuchen 2018).
values on high-quality samples taken from the site.
547
YSR = 1.83. Mean values of laboratory index param
eters include: wn = 107 ± 20 %, wL = 88 ± 20 %, wP
= 38 ± 4 %, and Ip = 54 ± 13 %.
Using the aforementioned consolidation data and
normalized undrained strength ratios from triaxial
compression tests on Saga clay reported by Samang
and Miura (2005), a profile of undrained shear
strength was developed using the SHANSEP method
(Ladd & DeGroot 2003). Figure 4 shows the com
parison of the laboratory reference profile of suc in
very good agreement with the CPTU derived values.
Figure 2. Profiles in clay at offshore Luva site, Norwegian 4.4 Tiller-Flotten, Norway
Sea (data from Lunne et al. 2014). Tiller-Flotten near Trondheim, Norway serves as the
experimental grounds involving quick clay research
(L’Heureux et al. 2019). Groundwater is subjected to
drawdown so that the hydrostatic pressure is consid
erably lower than normal.
Figure 5 shows a representative piezocone sound
ing (Mayne et al. 2019). The uppermost 2.5 m of soil
is interpreted as a dry/desiccated layer of stiff over-
consolidated sandy clay. Beneath this crust, through
out the sounding depths up to 30 m lies a soft fine-
grained soil. The clay is extremely sensitive to quick
from approximately 7.5 m below the surface (sensitiv
ity > 100). In the quick clay zone below 7 m depth,
typical index parameters are wn = 45 %, wL = 35 %,
wP = 20 %, and Ip = 15 %, with CF ≈ 45 %.
Figure 3. Profiles at Sipoo soft clay test site, Finland (data Figure 5 shows very good agreement from the
from DiBuò et al. 2019).
CPTU-derived suc values with those obtained from 7
benchmark CAUC series on high-quality block
samples.
Figure 4. Profiles in soft clay at Saga, Japan (data from

Hossain & Chai 2014). Figure 5. Profiles in highly sensitive clay at Tiller-Flotten
research site, Norway (data from L’Heureux et al. 2019).
4.3 Saga, Japan

4.5 Sainte-Monique, Canada
The Saga TB site is located at the south of Japan next
to the Ariake Sea. The upper 4 to 5 m is comprised of Locat et al. (2015) detail the results of ground investi
sandy silty clay while the underlying 12 m is a silty gations for a major landslide involving firm to stiff
clay. Groundwater is encountered at a depth of 0.6 m. sensitive nearly normally-consolidated plastic grey
Figure 4 shows profiles from a representative silty clay. The site is located some 130 km northeast of
CPTU at the site (Hossain & Chai 2014). Nine con Montreal near the Nicolette River. Index tests show 56
solidation tests on this clay indicated a mean ≤ wn ≤ 77%, 53 ≤ wL ≤ 65 %, 28 ≤ IP ≤ 40%, 72 < CF
548
< 85%, and liquidity index 1.1 ≤ LI ≤ 1.4. Sensitivities 4.7 Martin’s Point Bridge, United States
derived from fall cone indicate 39 ≤ St ≤ 55. Labora
A field case study with CPTU soundings in natural
tory consolidation tests gave YSR between 0.9
overconsolidated intact clays of the Presumpscot For
and 1.2.
mation is presented using data from Martin’s Point
Figure 6 shows the piezocone profile in an undis
Bridge, near Portland, Maine (Hardison & Landon
turbed area outside of the limits of the landslide. The
2015). The general stratification of the site (Figure 8)
high value of Bq gives a low value of Nkt which in
consists of a shallow organic silt layer underlain by
turn compares well with two CAUC and three CIUC
stiff OC Presumpscot clay that extends from depths to
triaxial tests performed on undisturbed samples.
2 to 14 m and overlies glacial outwash sand and
bedrock.
Shelby tube samples of the clay were collected at
the site from different elevations and tested for index
properties, consolidation parameters, and triaxial
strength characteristics. The results of laboratory index
testing on the clay gave an average unit weight γ =
16.5 kN/m3, natural water contents wn ≈ 30 to 40 %,
liquid limits in the range of 20 ≤ wL ≤ 45 %, and plas
ticity indices (Ip) between 10 and 20 %. Sensitivities
derived from field vane and fall cone were generally in
the range of 2 to 9.
The groundwater table is located at a depth of
2 m. The stress history profile was determined by
Figure 6. Profiles in sensitive clay at Sainte Monique,
a series of 13 constant-rate-of-strain (CRS) consoli
Quebec (data from Locat et al. 2015). dation tests that showed YSRs decreasing from 11 at
2 m depth to YSR = 3 at 14 m depth.
A total of 15 CAUC triaxial tests were performed,
including both recompression and SHANSEP type
methods (Ladd & DeGroot 2003). The derived pro
file of suc from the CPTU is presented in Figure 8
with values increasing from about 35 to 100 kPa in
the deposit and shown to be in reasonably good
agreement with the triaxial series.
Figure 7. Profiles in soft plastic clay at Liwan offshore site,

South China Sea (data from Palix et al. 2013).
4.6 Liwan, South China Sea

A deepwater geotechnical investigation was performed
at the Liwan 3-1 offshore site in the Pearl River
Mouth Basin of the South China Sea involving geo
physics, sampling, piezocones, T-bars, vane, dissipa Figure 8. Profiles in firm OC clay at Martin’s Point Bridge,
tion tests, and advanced laboratory testing (Palix et al. Maine (data from Hardison & Landon 2015).
2013). The site is underlain by soft highly-plastic
clays. Liquid limits decrease from 130 % at seafloor
to 80 % at 10 m below seafloor, with corresponding Ip 4.8 Offshore Denmark
going from 85 % to 50 % over the same depth inter
val. Clay fractions range between 25 to 50 %. Natural This case study considers the Danish sector of the
water contents vary from over 200 % at seafloor to North Sea. The site has a water depth of about
about 90 % at 20 m depth. Measurements of calcium 45 m and includes about 3 m of Holocene sands
carbonate content range from 6 % to 25 %. which are underlain by hard Pleistocene age clays of
Figure 7 shows a piezocone profile and the corres the Doggerbank formation. Laboratory index tests
ponding and reasonable profiles of suc from the repre on the clay indicate mean values: wn = 34 %, wL =
sentative CPTU sounding and 6 CAUC triaxial tests. 49 %, Ip = 27 % and γ = 19.4 kN/m3. Calcium
549
carbonate contents average 19 % for the clay. Con DeGroot, D., Lunne, T. and Tjelta, T.I. (2011). Recom
solidation tests indicate the clay to be overconsoli mended best practice for geotechnical site characteriza
dated with YSR decreasing from about 9 to 4 in the tion of cohesive offshore sediments. Frontiers in
depth interval from 3 to 15 m below the seafloor. Offshore Geotechnics II (Proc. ISFOG, Perth), Taylor &
Figure 9 shows low Bq values averaging 0.08, the Francis Group, London: 33–57.
Nkt-Bq algorithm gave a high mean value of Nkt = 21 DiBuö, B., D’Ignazio, M., Selãnpaã, J., Länsivaara, T. and
for the clay and the corresponding profile of suc com Mayne, P.W. (2019). Yield stress evaluation of Finnish
clays based on analytical CPTu models. Canadian Geo-
pares reasonably with the values from only two tech. Jour. 57 (11): 1623–1638; doi:10.1139/cgj-2019
CAUC tests on undisturbed samples from the site. 0427.
Hardison, M.A. and Landon, M.L. (2015). Correlation of
engineering parameters of the Presumpscot formation to
SCPTU. Report No. ME 15-12, prepared by Univ. of
Maine for Maine DOT, Augusta, 394 p.
Hossain, J. and Chai, J. (2014). Estimating coef. of consoli
dation and hydraulic conductivity from piezocone test
results. Geomechanics and Engineering 6 (6): 577–592.
Hong, S.J., Lee, M.J., Kim, J.J. and Lee, W.J. (2010).
Evaluation of undrained shear strength of Busan clay
using CPT. Proc. 2nd Intl. Symp. on Cone Penetration
Testing, Vol. 2 (Huntington Beach, CA), Omnipress:
313–320.
ISO (International Organization for Standardization) 2012.
Geotechnical investigation and testing – Field testing –
Part 1: Electrical cone and piezocone penetration tests,
International Standard ISO 22476-1:2012. (with tech
Figure 9. Profiles in stiff fissured OC Doggerbank clay at nical corrigendum 1, January 2013). Geneva.
Danish offshore site. Karlsrud, K., Lunne, T. and Brattlien, K. (1996). Improved
CPTU interpretations based on block samples. Proc.
Nordic Geotechnical Meeting, Reykajavik, Paper 3.4:
195–201.
5 CONCLUSIONS Knappett, J.A. and Craig, R.F. (2012). Craig’s Soil Mech
anics, 8th Edition, Spon Press, Taylor & Francis Group,
The empirical methodology considers cone bearing London: 570 pages.
factor (Nkt = qnet/suc) as a function of pore pressure L’Heureux, J-S., Lindgård, A. and Emdal, A. (2019). The
ratio (Bq = Δu2/qnet). The underlying database Tiller-Flotten research site: geotechnical characteriza
includes statistical analyses of 407 CAUC triaxial tion of a very sensitive clay deposit. AIMS Geosciences
tests performed on 62 different clay deposits that Volume 5, 4: 831–867
were also field tested by CPTU. Ladd, C.C. and DeGroot, D.J. (2003). Recommended prac
tice for soft ground site characterization. Soil & Rock
The general trend shows Nkt varying from as high America 2003, Vol. 1 (Proc. 12 PCSMGE, MIT), Vertag
as 30+ for stiff fissured overconsolidated clays to Glückauf, Essen: 3–57.
low values of around 6 for soft sensitive and quick Lade, P.V. (2016). Triaxial Testing of Soils, John Wiley &
clay deposits. Generally, Nkt decreases with Bq. Sons, Ltd, Chichester, UK: 402 p.
Eight new case studies are presented showing the Locat, A., Leroueil, S., Fortin, A., Demers, D. and
reasonableness and reliability of the earlier-derived Jostad, H.P. (2015). The 1994 landslide at
methodology for assessing undrained strength of Sainte-Monique, Quebec: geotechnical investigation.
clays from piezocone penetration tests. Triaxial com Canadian Geotechnical Journal 52 (4): 490–504.
pression tests (CAUC, CK0UC, and occasionally Low, H.E., Lunne, T., Andersen, K.H., Sjursen, M.A., and
Randolph, M.F. (2010). Estimation of intact and
CIUC) on high-quality undisturbed samples were remoulded undrained shear strengths from penetration
used as the benchmark reference tests. The final tally tests in soft clays. Geotechnique 60 (11): 843–859.
for the database now includes CPTU data from 70 Lunne, T., Christoffersen, H.P. and Tjelta, T.I. (1985).
clay sites with a total of 497 triaxial compression Engineering use of piezocone data in North Sea clays.
tests that provide the benchmark values of suc. Proc. ICSMFE, Vol. 2, San Francisco, 907–912.
Lunne, T., Robertson, P.K. and Powell, J.J.M. (1997). Cone
Penetration Testing in Geotechnical Practice, Routledge
REFERENCES -Taylor & Francis Group, London: 352 p.
Lunne, T. (2010). The CPT in offshore soil investigations
Agaiby, S.S. (2018). Advancements in the interpretation of a historic perspective. Geomechanics & Geoengineer
seismic piezocone tests in clays and other geomaterials. ing, Vol. 7 (2): 75–101.
PhD dissertation, Civil & Environmental Engineering, Lunne, T., Andersen, K.H. and Yang, S.L. (2014).
Georgia Institute of Technology, Atlanta, GA: 923 p. Undrained shear strength for deep water field develop
Colreavy, C. (2016). Use of piezoball penetrometers for meas ment in the Norwegian Sea. Proceedings. 5th Inter
uring shear strength and consolidation characteristics of national Workshop on CPTU and DMT in soft clays and
soft soil. PhD dissertation, Centre for Offshore Foundation organic soils, Polish Committee of Geotechnics,
Systems, Univ. of Western Australia, Perth: 304 pages. Poznan: 213–228.
550
Mayne, P.W. (2008). Piezocone profiling of clays for mari sensitive clay at Tiller-Flotten, Norway. Proc. XVII
time site investigations. Geotechnics in Maritime Engin European Conf on Soil Mechanics & Geotech Engrg.
eering, Vol. 1 (Proc. 11th Baltic Sea Geotechnical (ECSMGE): Paper 0153, Reykjavik, Icelandic Geotech
Conference, Gdansk), Polish Committee on Geotech nical Society: www.issmge.org.
nics: 333–350. Palix, E., Chan, N., Yangrui, Z. and Haijin, W. (2013).
Mayne, P.W, Peuchen, J., and Bouwmeester, D. (2010a). Liwan 3-1: How deep water sediments from South
Unit weight evaluation from CPT. Proceedings, 2nd Intl. China Sea compare with Gulf of Guinea sediments. Pro
Symposium on Cone Penetration Testing (CPT’10), ceedings, Offshore Technology Conference, Houston,
Vol. 2, Huntington Beach, California: 169–176. Texas, OTC Paper 24010, Vol 2, 1101–1108.
Mayne, P.W, Peuchen, J., and Bouwmeester, D. (2010b). Paniagua, P., L’Heureux, J-S., Carroll, R., Kåsin, K. and
Soil unit weight estimated from CPTu in offshore soils. Sjursen, M. (2017). Evaluation of sample disturbance of
Frontiers in Offshore Geotechnics II (Proc. ISFOG, three Norwegian clays. Proc. Intl. Conf. on Soil Mech
Perth), Taylor & Francis Group, London: 371–376. anics and Geotechnical Engineering, Seoul, 481–484.
Mayne, P.W. and Peuchen, J. (2012). Unit weight trends www.issmge.org
with cone resistance in soft to firm clays. Geotechnical Powell, J.J.M. & Quarterman, R.S.T. (1988). The interpret
and Geophysical Site Characterization 4, Vol. 1 (Proc. ation of cone penetration tests in clays with particular
ISC-4, Pernambuco), CRC Press, London: 903–910. reference to rate effects. Penetration Testing 1988, Vol.
Mayne, P.W. (2014). Development of an automated meth 2 (Proc. ISOPT-1, Orlando), Balkema, Rotterdam:
odology for evaluation of undrained shear strength of 903–909.
offshore clays from piezocone penetration tests. Report Robertson, P.K. 2009. Interpretation of cone penetration
122931 by Georgia Tech Research Corp. (Project tests: a unified approach. Canadian Geotechnical Journal
No. 2006U94) to Fugro, The Netherlands: 192 p. 46 (11): 1337–1355.
Mayne, P.W., Peuchen, J. & Baltoukas, D. (2015). Piezo Samang, L., Miura, N. and Sakai, A. (2005). Geotechnical
cone evaluation of undrained strength in soft to firm off properties of soft cohesive lowland soils deposited in
shore clays. Frontiers in Offshore Geotechnnics III, Vol. Saga Airport Highway, Japan. Media Komunikasi
2 (ISFOG, Oslo), Taylor & Francis, London: Teknik Sipil, Vol. 13 (3): EDISI XXXIII: 19–35
1091–1096. Schnaid, F. (2009). In Situ Testing in Geomechanics: The
Mayne, P.W. & Peuchen, J. (2018). Evaluation of CPTU Main Tests. CRC Press, Taylor & Francis Group,
Nkt cone factor for undrained strength of clays. Cone London, 352 pages.
Penetration Testing 2018 (Delft), CRC: 423–429. Yu, H.S. and Mitchell, J.K. (1998). Analysis of cone resist
Mayne, P.W., Paniagua, P., L’Heureux, J-S., Lindgård, A., ance: review of methods. Journal of Geotechnical and
and Emdal, A. (2019). Analytical CPTu model for Geoenvironmental Engineering 124 (2): 140–149.
551
© 2022 the Author(s), ISBN 978-1-032-31259-0
Evaluating geoparameters of Maine sensitive clay by CPTU

P.W. Mayne
Georgia Institute of Technology, Atlanta, GA USA
P. Paniagua
Norwegian Geotechnical Institute and Norwegian University of Science & Technology, Trondheim, Norway
B. Di Buò
S.S. Agaiby
Cairo University, Giza, Egypt
ABSTRACT: Using two sets of analytical solutions for CPTU in clays, a suite of theoretically-consistent
geoparameters is shown to be in good agreement with independent laboratory and field test results obtained
on sensitive Presumpscot clay in Portland, Maine, USA. Fall cone tests indicate a mean sensitivity of St ≈
37. Values of undrained rigidity index (IR), undrained shear strength (su), and yield stress ratio (YSR) are
provided by a modified spherical cavity expansion-critical state hybrid model while an effective stress limit
plasticity solution is utilized to assess the effective friction angle of the sensitive clay at both peak strength
[ϕ' at qmax] and also at maximum obliquity [ϕ' at (σ1'/σ3')max]. A CPTU screening method that uses three
simplified equations for YSR helps to identify that the clay is sensitive.
1 INTRODUCTION this paper, two sets of analytical closed-form solu

tions are utilized so that a consistent and rational
1.1 Falmouth Bridge, Maine assessment is made for stress history and shear
strength, both in terms of effective stress parameters
A geotechnical investigation for a new bridge
(i.e., friction angle, f’) and total stress analysis (i.e.
along state highway route 26/100 at the northern
undrained shear strength, su). � �
portion of Portland, Maine, USA was performed
The yield stress, or preconsolidation stress σp ’
by the University of Maine for the Maine Depart
is presented in terms of the normalized yield stress
ment of Transportation (Hardison & Landon
ratio: YSR ¼σ0p =σ0vo . Conventionally, results from
2015). The site is located to the east of Interstate
one-dimensional consolidation tests are taken at
I-95 and underlain by sensitive clays of the Pre
various elevations to develop the profile of YSR
sumpscot Formation and in close proximity to
with depth in clays. Herein, a modified spherical
prior landslides along the Presumpscot River in
cavity expansion-critical state soil mechanics
the community of Westbrook (Devin & Sandford
(SCE-CSSM) hybrid model for CPTU in clays pro
2000), as shown by Figure 1.
vides three YSR profiles, as well as a measure of
The subsurface exploration for the Portland-
undrained rigidity index (IR = G/su), where G is the
Maine bridge included soil test borings, drive sam
shear modulus.
pling, undisturbed sampling, seismic piezocone
penetration test (SCPTU) soundings, and various
series of laboratory tests (Langlais 2011). A detailed
2 CONE PENETRATION TESTS
summary of the geotechnical data and results is pro
vided by Hardison & Landon (2015).
2.1 Piezocone soundings
1.2 Interpretation of CPTU in clays The CPTU provides three continuous readings with
depth: (a) cone tip resistance, qt; (b) sleeve friction, fs;
The interpretation of CPTU in clays often relies on and (c) penetration porewater pressure, u2. The stand
empirical correlations and simple statistical trends, ard rate of advancement is 20 mm/s, although the use
although theoretical formulations also play a role. In of variable rate CPTU soundings have been revealed
DOI: 10.1201/9781003308829-79
552
Figure 1. Locations of bridge site and landslide in soft sen
sitive Presumpscot clay, Portland, Maine, USA.
Figure 2. Profiles of qt, fs, u2, and Vs from SCPTU at

to provide an effective method to characterize silts, Route 26/100 bridge site in Portland, Maine (data from
mixed soils, and mine tailings that exhibit partially- Hardison & Landon 2015).
drained behavior.
2.2 Seismic piezocone testing 3 GEOPARAMETERS FROM CPTU

The addition of a set of geophones to the standard
The evaluation of soil engineering parameters from
penetrometer allows for downhole geophysical test
CPTU is often addressed using empirical correl
ing, most often at the 1-m rod breaks. A horizontal
ations and/or statistical trends derived from prior
seismic source or autoseis unit is used to generate
databases. As a result, some of the obtained values
horizontally-polarized shear waves that are propa
of the geoparameters are inconsistent with each
gated vertically with depth. The profile of shear
other, or not well-matched well amongst each other,
wave velocity (Vs) is used to obtain the small-strain
since they are assessed independently.
shear modulus (G0 = Gmax) via elastic theory:
In this paper, the following geoparameters are
assessed theoretically using two analytical solutions:
(a) effective stress friction angle ðf’Þ at both qmax and
ðσ1 ’=σ3 ’Þmax ; (b) rigidity index, IR; (c) undrained
shear strength, su; and (d) yield stress ratio
where ρt ¼γt =ga = total soil mass density, γt = total YSR ¼σp ’=σvo ’ Thus, their values are obtained in
soil unit weight, and ga = acceleration constant. a consistent and rational manner. Moreover, independ
A representative SCPTU at the Portland-Maine ent laboratory reference data on recovered soil samples
Bridge site showing qt, fs, u2, and Vs in sensitive soft are shown to be comparable with the CPTU
clay is presented in Figure 2. evaluations.
2.3 CPTU parameters 3.1 Index parameters of clay

It is convenient to express the CPTU results as net Laboratory index tests on the sensitive Presumpscot
readings: (1) net cone resistance: qnet ¼qt -σVo ; (2) clay at the Portland-Maine site indicated: natural
excess porewater pressure: Δu ¼ u2 -u0 ; and (3) water content: wn = 43.6 ± 7.3%, liquid limit: LL =
effective cone resistance: qE = qt-u2; where σvo = total 42.1 ± 6.9%, plasticity index: PI = 17.5 ± 5.6%,
vertical overburden stress, u0 = equilibrium porewater liquidity limit: LI = 1.13 ± 0.34; and specific gravity
pressure; and σvo ’ ¼ σvo - u0 = effective overburden of solids: Gs = 2.78. Measured unit weights gave
stress. a mean of γt ¼ 1:74 kN/m3, while natural water con
Furthermore, several normalized and dimension tents using Gs and S = 1 indicated a value of around
less CPTU parameters can be defined: Q ¼qnet =σvo ’, 18.4 kN/m3.
¼ Δu =qnet , U ¼Δu =σvo ’, and Fr (%) = 100∙fs/qnet.
Note that the first three of these parameters are inter
related via: U = Q∙Bq. 3.2 Soil behavior type
An update to the normalized Q is now pro For soil classification by CPTU, it is common to util
vided with a variable exponent that depends ize soil behavior type (SBT) charts. Hardison &
upon soil type, termed Qtn. Details are provided Landon (2015) discuss the use of SBT that rely on
by Robertson (2009) and Robertson & Cabal Q-F and Q-Bq diagrams (Robertson & Cabal 2015).
(2015). For the CPTU data at Portland-Maine Bridge, the
553
Q-F charts primarily indicate a zone 3 soil type
(clays to silty clays), with an intermingling of zone 1
(sensitive soils), as shown in Figure 3. For the Port
land CPTU data, the Q-Bq chart fails to find sensitive
clays, as presented in Figure 4.
A SBT chart by Schneider et al. (2012) uses Q and
U to identify soil types. This approach seems to
better recognize sensitive clays, as shown in Figure 5.
Figure 5. Portland-Maine CPTU data in Q-U soil behavior

chart.
Figure 3. Portland-Maine CPTU data in Q-F soil behavior

chart.
Figure 6. Profiles of SBT zone number from Q-F chart and

clay sensitivity for Presumpscot clay at Portland-Maine
site.
3.3 Screening for sensitive clays by CPTU

In addition to SBT charts for identification of sensi
tive soils, a simple CPTU screening can be used, as
detailed elsewhere (Agaiby & Mayne 2018, 2021;
Mayne et al. 2019).
For “regular” clays that are inorganic and insensi
Figure 4. Portland-Maine CPTU data in Q-Bq soil behavior tive, the following applies:
chart.
In fact, clay sensitivities (St) measured by lab fall

cone range from 9 to 268 at the Portland bridge site.
Based on the guidelines discussed by Holtz et al. For sensitive clays, the following hierarchy applies:
(2011), the clay classifies as medium to highly sensi

tive below depths of 6 m where St > 8, as evidenced
by Figure 6.
554
For the CPTU at Portland-Maine Bridge,
Figure 7 shows that the hierarchy from (3) applies,
thus identifying sensitive soft clay at depths
below 6 m.
Figure 8. Triaxial stress path for Presumpscot clay at land

slide site in Portland, Maine (data from Devin & Sandford
2000).
Figure 7. CPTU screening hierarchy to identify sensitive

clays at Portland-Maine site.
3.4 Friction angle of sensitive Presumpscot clay

The effective stress friction angle ðf’Þ is a funda
mental property of soil and an important parameter
for stability analysis, foundation design, and numer
ical FEM simulations.
For the landslide investigation near Route 26/100,
Devin & Sandford (2000) presented triaxial com
pression test data on soft sensitive Presumpscot clay.
Figure 8 shows CK0UC triaxial stress paths for a NC
specimen, indicating a value of f’1 ¼ 30� at peak
strength, while at later stages of shearing, a value of
f’2 ¼ 30� is obtained
� at maximum obliquity (M.O.),
defined when σ01 =σ03 Þmax occurs.
To obtain f’ in sensitive clays from CPTU, the
Norwegian Institute of Technology (NTH, now
NTNU) developed an effective stress limit plasticity
solution for assessing f’ in all soil types (Janbu &
Senneset 1974; Senneset et al. 1989; Sandven et al
2016). The expression for the case where c' = 0 and
undrained penetration ðβ ¼ 0Þ can be expressed:
Figure 9. Profiles of f’1 and f’2 from CPTU using NTH
solutions.
Since iteration is required, an approximate inver

sion to express f’ directly in terms of CPTU param
eters Q and Bq has been devised for the following
ranges: 0.05 ≤ Bq < 1.0 and 18° < f’ < 45° (Mayne The value of f’corresponding to large strains or
2007): maximum obliquity is obtained with this original
555
NTH solution. Figure 9 shows the profile of f’2 at 50% strength. In this case, IR50 is obtained
with depth and that a value of f’2 ≈33� . from:
To obtain the value of f’1 at qmax, a modified
NTH solution is implemented (Sandven et al. 2016;
Ouyang & Mayne 2019). In this case, Q in (4) is
replaced with Q’ that includes stress history:
where Gmax, qnet, and σvo ’ are all in same units.

At the Portland, Maine site, this approach gives
a range: 148 < IR < 260.
where YSR ¼σp ’=σvo ’ = yield stress ratio. The A spherical cavity expansion - critical state soil
exponent Λ can be theoretically calculated as mechanics (SCE-CSSM) model for CPTU in sensi
Λ ¼1- Cs =Cc where Cc = compression index and Cs tive clays provides the direct assessment of IR
= swelling or recompression index, however, more (Agaiby & Mayne 2018):
often is assigned as a value Λ ≈ 0:7 to 0:8 for
insensitive clays and Λ ≈ 0:95 to 1:0 for sensitive
and quick clays (Ouyang & Mayne 2019).
At the Route 26/100 Falmouth Bridge site in Port
land, Maine, the trend of YSR with depth from CRS
consolidation tests indicates:
where z = depth (meters). The CRS results will be

presented later in the paper.
Adopting Λ ¼ 0:95 for soft sensitive Presumpscot
clay and using Q’ from (6) in (5), the profile of f’1
with depth is shown in Figure 9. From depths
between 8 to 20 m, the CPTU value more or less
agrees with the CK0UC value f’1 ¼ 30� .
3.5 Rigidity index

The undrained rigidity index is defined as IR =
G/su where G = shear modulus and su = undrained
Figure 10. Plot of U-1 versus Q to obtain slope parameter
shear strength. The difficulty here is that the mag aq for evaluating rigidity index of Presumpscot clay.
nitude of G ranges greatly, from a very high value
at the nondestructive range at Gmax to a low value
at failure (Gf) corresponding to peak strength.
In many instances, empirical correlations for
estimating IR are used. A UC-Berkeley method
developed from triaxial tests on clays by Keaveny
& Mitchell (1986) relates IR with plasticity index
(PI) and YSR, for which an approximate expres where Mc = 6∙sinφ’/(3-sinφ’) is the frictional param
sion is available (Mayne 2007). eter in Cambridge q-p’ space and aq is obtained as
the slope of (U-1) versus Q. The value of Mc1 cor
responds to f’1 at qmax while Mc2 is associated with
f’2 at large strains.
For the Portland site, a slope parameter aq = 0.581
is obtained, as shown in Figure 10. Using the corres
ponding values of Mc1 = 1.20 and Mc2 = 1.33 gives
which applies when: 10 < PI < 50 and YSR < 10. a calculated IR = 266.
For the range of PI at Portland-Maine Bridge (9 < PI The three approaches for IR are presented in
< 28), this indicates an 120 < IR < 250. Figure 11. The upper bounds for the UCB and UCD
A UC-Davis approach uses results from SCPTU methods imply that IR = 266 from SCE-CSSM solu
to obtain a mobilized stress level based value of IR tion is reasonable.
556
Figure 11. Undrained rigidity index profiles at Portland
Figure 12. Profiles of yield stress and undrained shear
site.
strength for sensitive Presumpscot clay in Portland, Maine
(reference lab data from Hardison & Landon 2015).
3.6 Yield stress ratio at Portland-Maine site

Alternatively, a SHANSEP type approach can be
The SCE-CSSM provides three expressions for YSR used where (Hardison & Landon 2015):
in clays that are functions of either Q or U, as well
as both Q and U (Di Buò, et al. 2019):
where S = 0.33 and Λ ¼ 0:95. Here, the results of

the CRS consolidation tests provide the YSR.
For the CPTU, the undrained shear strength is
a triaxial compression mode such that:
where Nkt is a cone bearing factor obtained from

SCE theory (Vesić 1977):
The input parameters of Q and U, together with
Mc1 = 1.20, Mc2 = 133, Λ = 0.95, and IR = 266 are
used to generate three profiles of YSR. The associated
yield stresses are consistent and shown to be in
good agreement with CRS consolidation tests on For a value of IR = 266, the calculated bearing
undisturbed samples from the site, as seen in factor is Nkt = 11.35.
Figure 12a. All three evaluations of suc for soft sensitive clay
at Portland are presented in Figure 12b with reason
3.7 Undrained shear strength able agreement shown for all profiles.
For the Portland, Maine site, a series of constant
volume direct simple shear (DSS) tests were per 4 ADDITIONAL CASE STUDIES
formed (Langlais 2011). These results can be con
verted to an equivalent triaxial compression mode These analytical CPTU solutions have also been
(CK0UC) via a simple relationship that depends on successfully applied to a number of other sensitive
f0 (Mayne 2008): clays, including Finland (Di Buò, et al. 2019),
Norway (D’Ignazio, et al. 2019; Mayne et al.
2019), USA (Mayne & Benoît 2020), and Canada
(Agaiby & Mayne 2018; Agaiby et al. 2021).
557
5 CONCLUSIONS Janbu, N., and Senneset. K. 1974. Effective stress interpret
ation of in situ static penetration tests. Proceedings of
Two sets of analytical solutions for CPTU in soft the 1st European Symposium on Penetration Testing,
sensitive clays are applied to a case study involv Vol. 2, Swedish Geotechnical Society, Stockholm:
ing the Presumpscot Formation in Portland, 181–193.
Maine. A consistent and theoretical assessment is Keaveny, J. and Mitchell, J.K. 1986. Strength of
made for the geoparameters of the clay, including fine-grained soils using the piezocone. Use of In-Situ
Tests in Geotechnical Engineering (Proc. In-Situ’86,
effective stress friction angle (f’ at qmax and f’ Blacksburg), GSP 6, ASCE, Reston, Virginia: 668–685.
at M.O.), rigidity index (IR), undrained shear Krage, C.P., Broussard, N.S., & DeJong, J.T. 2014. Esti
strength (suc), and yield stress ratio (YSR). The mating rigidity index based on CPT measurements. Pro
values from these closed-form solutions are in ceedings of the 3rd International Symposium on Cone
general agreement with results from laboratory Penetration Testing, Las Vegas: 727–735. www.usucger.
testing on undisturbed samples, including CRS- org
type consolidation tests, triaxial compression Langlais, N.D. 2011. Site characterization using the seis
tests, and direct simple shear. A simple means mic piezocone in Presumpscot clay and correlations to
engineering parameters, MS Thesis, ID 1569, Civil
for screening to identify sensitive clays from
Engineering Department, The University of Maine:
“regular” insensitive and inorganic clays is also 249 pages.
presented. Mayne, P.W. 2007. NCHRP Synthesis 368: Cone Penetra
tion Testing, Transportation Research Board, Washing
ton, DC: 118 p. www.trb.org
ACKNOWLEDGMENTS Mayne, P.W. 2008. Piezocone profiling of clays for mari
time site investigations. Geotechnics in Maritime Engin
The first author extends appreciation to ConeTec eering, Vol. 1 (Proceedings, 11th Baltic Sea
Group of Burnaby, BC for providing funds to Geor Geotechnical Conference, Gdansk), Polish Committee
gia Tech on in-situ research activities. on Geotechnics: 333–350.
Mayne, P.W., Paniagua, P., L’heureux, J-S., Lindgård, A.,
and Emdal, A. 2019. Analytical CPTu model for sensi
REFERENCES tive clay at Tiller-Flotten site, Norway. Proc. XVII
ECSMGE: Geotechnical Engineering Foundation of the
Agaiby, S.S. and Mayne, P.W. 2018. Interpretation of Future, Paper 0153, Reykjavik, Icelandic Geot. Society:
piezocone penetration and dissipation tests in sensitive www.issmge.org
Leda Clay at Gloucester Test Site. Canadian Geotech. Mayne, P.W. and Benoît, J. 2020. Analytical CPTU models
Journal, 55(12): 1781-1794: http://dx.doi.org/10.1139/ applied to sensitive clay at Dover, New Hampshire.
cgj-2017-0388. Journal of Geotechnical & Geoenvironmental Engineer
Agaiby, S.S. and Mayne, P.W. 2021. CPTU identification ing 146 (12) DOI: 10.1061/(ASCE)12GT.1943
of regular, sensitive, and organic clays towards evaluat 5606.0002378
ing preconsolidation stress profiles. AIMS GeoSciences Ouyang, Z. and Mayne, P.W. 2019. Modified NTH method
7 (4): 553–573. for assessing effective friction angle of normally con
Agaiby, S.S., Mayne, P.W., and Greig, J. 2021. CPTU solidated and overconsolidated clays from piezocone
screening method to identify soft sensitive clays in tests. Journal of Geotechnical & Geoenvironmental
Canada. Proceedings GeoNiagara: 74th Canadian Geo- Engineering 145(10), doi.org/10.1061/(ASCE)GT.1943
technical Conference, Paper ID 118, Canadian Geotech 5606.0002112
Society. Robertson, P.K. 2009. Interpretation of cone penetration
Di Buò, B., D’Ignazio, M., Selãnpaã, J., Länsivaara, T. and tests: a unified approach. Canadian Geot. J. 46 (11):
Mayne, P.W. 2019. Yield stress evaluation of Finnish 1137–1355.
clays based on analytical CPTu models. Canadian Geo- Robertson, P.K. and Cabal, K. 2015. Guide to Cone Pene
tech. Jour. 57 (11): 1623–1638; doi.org/10.1139/cgj tration Testing for Geotechnical Engineering, 6th edi
2019-0427 tion, Gregg Drilling, Signal Hill, CA: 140 pages.
Devin, S.C. and Sandford, T.C. 2000. Shear strength of sen Sandven, R., A. Gylland, A. Montafia, K. Kåsin,
sitive clay slopes in Maine. Proceedings Slope Stability A. A. Pfaffhuber, and M. Long. 2016. In-situ detection
2000, GSP 101 (GeoDenver) ASCE, Reston, Virginia: of sensitive clays, Part II: Results. Proc., 17th Nordic
114–128. Geotechnical Meeting: Challenges in Nordic Geotech
D’Ignazio, M., Lindgård, A. and Paniagua, P. 2019. Soil nic, Icelandic Geotechnical Society, Reykjavik:
Parameters in Geotechnical Design, Report SP8, pre 113–123.
pared by Norwegian Geotechnical Institute (NGI) for Schneider, J.A., Hotstream, J.N., Mayne, P.W. and
the Norwegian Research Council, Oslo: 70 p. Randolph, M.F. (2012). Comparing CPTu Q-F and Q
Hardison, M.A. and Landon, M.L. 2015. Correlation of Δu2/σvo’ soil classification charts. Geotechnique Letters,
engi-neering parameters of the Presumpscot formation Vol. 2 (4): 209–215.
to the seismic cone penetration test. Technical Report Senneset, K., Sandven, R., and Janbu, N. 1989. Evaluation
15-12 pre-pared by University of Maine, Advanced of soil parameters from piezocone tests. Transportation
Structures and Composites Center, submitted to Maine Research Record 1235, National Academy Press, Wash
Dept. of Transportation, Augusta: 394 pages. ington, DC: 24–37.
Holtz, R.D., Kovacs, W.D. and Sheahan, T.C. 2011. An Vesić, A.S. 1977. NCHRP Synthesis 42: Design of Pile
Introduction to Geotechnical Engineering, 2nd Edition, Foundations, Transportation Research Board, National
Pearson, Upper Saddle River, NJ: 853 p. Academies Press, Washington, DC: 68 p.
558
© 2022 the Author(s), ISBN 978-1-032-31259-0
CPTU evaluations in Appalachian Piedmont residual sandy silts

P.W. Mayne
Georgia Institute of Technology, Atlanta, GA, USA
E. Cargill
ConeTec Group, Richmond, VA, USA
ABSTRACT: Results from recent piezocone testing are compared with a statistical set of 22 prior CPTU
soundings at the Opelika National Geotechnical Experimentation Site in Alabama. The site is underlain by
residua comprised of fine sandy silts of the Appalachian Piedmont geologic province in the eastern USA. The
cone resistance and sleeve friction from a 2016 CPTU compare well with the mean values from previous
soundings while the porewater pressure readings are similar yet different because of a drop in the groundwater
table from 2 m to 10 m over the period of study. Evaluations of the effective friction angle and yield stress
from CPTU in residual soils are presented and compared with benchmark values obtained from laboratory
triaxial and consolidation test results. Most interestingly, as the soil is intermediate with about 50-50 silt and
sand, both undrained and drained penetration give more or less the same results.
1 INTRODUCTION extent, the Piedmont lies beneath younger sediments of

the Atlantic Coastal Plain deposits. Moreover, the Pied
Since 70% of the Planet Earth is covered by oceans, mont serves as an important source of crushed stone,
most soils are formed originally as marine sediment. aggregate, and sands from quarries, as well as provides
As a consequence, the majority of geotextbooks and the natural foundation bearing material for buildings,
research studies have focused on the interpretation of bridges, and highway pavements for major urbanized
in-situ and laboratory tests involving water-borne centers, including Atlanta/GA, Greenville/SC, Colum
deposits. It is estimated that approximately 5% of soils bia/SC, Raleigh/NC, Charlotte/NC, Richmond/VA,
globally are found to be residual type, formed by the Washington/DC, Baltimore/MD, and Philadelphia/PA.
inplace disintegration and weathering of parent bedrock
(USDA 2021), thus the evaluation of geoparameters in
residua has been less well understood and quantified.
In this paper, results from piezocone penetration
tests (CPTU) in residual sandy silts at a national test
site in the southeastern USA are presented and inter
preted. Of specific interest, the interpretation of
effective stress parameters (c' and ϕ') and yield stress
profile (σp') with depth provide the main focus.
When standard CPT soundings are performed at
20 mm/s, the results are considered undrained in
clays, whereas in sands the response is taken as
drained (Lunne et al. 1997). For silts, however, it is
unclear whether the data are undrained or drained, or Figure 1. Extent of Appalachian Piedmont in eastern USA.
more likely in the regime of partially-drained behav
ior (DeJong, et al. 2012; Holmsgaard et. al. 2016; Primary rock types include gneiss and schist of Pre
Blaker et al. 2019). cambrian Z-age that were later intruded by granitic rocks
of Paleozoic age. Residual soils commonly form as very
2 APPALACHIAN PIEDMONT RESIDUUM fine sandy silts (ML, MH) to very silty fine sands (SM)
and a dual system (ML-SM) has been used in a modified
The Appalachian Piedmont geologic province extends form of the Unified Soil Classification System.
along the eastern USA ranging from Alabama to New A generalized profile of the residual and saprolitic
Jersey, as shown in Figure 1. In addition to the surficial soil and rock types is presented in Figure 2. In the
DOI: 10.1201/9781003308829-80
559
is approximately 150 hectare, owned by the Alabama
Dept. of Transportation, and managed by Auburn
University.
2.2 CPTU soundings at Opelika NGES

During the period from around 1995 to approxi
mately 2000, many CPTU soundings were con
ducted at the Opelika NGES by several research
groups and commercial testing firms (Mayne &
Brown 2003). A statistical summary of some 22
CPTUs at the site are reported by Mayne et al.
(2009), as shown by Figure 3. At the time of those
series of soundings, the groundwater table was gen
erally found to be around 2 to 3 m deep (Anderson
Figure 2. Generalized soil-rock profile in Piedmont geology. et al. 2019).
In 2016, two new CPTUs were conducted by Cone-
Tec Group as part of a new research program on
energy piles (Atalay 2019). Results from one of these
soundings is superimposed on the qt and fs profiles in
Figure 3, showing very good overall agreement in
these profiles in magnitudes while also displaying
some local variations within the residual soil profile
due to differential weathering.
Penetration porewater pressures at the shoulder
position (u2) in the Piedmont residuum is often nega
tive below the groundwater table (Finke et al. 2001),
as evident in Figure 4. The 2016 CPTU reading is
also shown and differs in that the water table was
considerably lower (zw ≈ 10 m), as detailed by
Anderson et al. (2019).
Figure 3. Comparison of mean qt and fs profiles from 22 CPTU

series (2009 vintage) with 2016 sounding at Opelika test site.
southern Piedmont, saprolitic residuum is called par

tially-weathered rock (PWR) when the standard
penetration test (SPT) values exceed 100 blows per
foot (bpf), whereas in the northern Piedmont, the
term decomposed rock is used and defined when
SPT values exceed 60 bpf.
2.1 Opelika test site, Alabama

Some three decades ago, six national geotest sites
were established with federal funding in the contin
ental USA (Benoît & Lutenegger 2000).
The Opelika test site in Alabama is situated in the
Piedmont geology and serves as research grounds for
laboratory, in-situ, geophysical, and full-scale foundation
studies (Vinson & Brown 1997; Mayne et al. 2000; Figure 4. CPTU porewater pressure readings in fine sandy
Mayne & Brown 2003; Anderson et al. 2019). The site silt at Opelika test site.
560
2.3 Groundwater effect on CPTU lab program included: index testing, grain size, one-
dimensional consolidation, triaxial compression, reson
The depth to groundwater governs the equilibrium
ant column, direct shear, and permeability (Vinson &
porewater pressure (u0) and may affect the CPTU read
Brown 1997).
ings. Of interest here too is that the soils may be either
Results from mechanical analyses using sieves of
dry, partially- or fully-saturated due to capillarity, since
recovered samples are presented in Figure 5. The
partially saturated soils may occur in the vadose zone
measured fines content (FC) and mean grain size
between the ground surface and groundwater.
(D50) are shown with depth to 16 m. It is evident
In some reported studies involving groundwater
that the soil particle sizes are at the threshold demar
tables and partially saturated soils, the CPTU read
cation between fine-grained soils and coarse-grained
ings can show differences at seasonal changes due to
soils, i.e. D50 = 0.075 mm corresponding to the US
matrix suction, partial or full capillarity, desiccation,
No. 200 sieve.
and rainfall (e.g., Lehane et al. 2004; Huffman et al.
The fines content has a mean value FC = 44 %
2015; Giaceti et al. 2019). In fact, for CPT in
(n = 63), thus the dual symbol ML-SM is seen
residual clayey sands derived from sandstone, Gia
appropriate for the fine sandy silts to silty fine
ceti et al. (2019) showed changes in qt and fs in the
sands. Average liquid limits and plasticity indices
upper 4 m while less differences at greater depths.
were 46% and 8%, respectively, although many
Lehane et al. (2004) had two test areas at the same
specimens test as non-plastic (Mayne & Brown
site, one with euclalyptus trees and one in an open
2003). Natural water contents typically range
area. The CPTU soundings in the open area did not
between 20 and 40% in the upper 16 m, yet specif
show seasonal changes, while those in the treed area
ically for the earlier set of data, the mean wn = 29.9
did. Huffman showed some seasonal changes in
± 6.2 % (n = 26) that dropped to wn = 25.7 ± 7.0%
CPTU readings at a silty site in Oregon.
(n = 37) in 2016, presumably due to the ground
However, the qt and fs profiles at Opelika do not
water drop.
show significant differences in the 2000 and 2016 pro
Due to the closure of a nearby marble quarry
files of cone and sleeve resistances, despite the large
some 4 km from the site in 2014, the groundwater
changes in groundwater levels.
has now begun a recovery toward its former regime
(Anderson et al. 2019).
3.2 Yield stress profiles in Piedmont residuum

Consolidation tests on undisturbed samples from
the site are reported by Hoyos & Macari (1999).
Figure 6 shows the interpreted profile of yield
stress (σp') and yield stress ratio (YSR = σp'/σvo')
from this test series. Assuming that full capillarity
occurs in the overburden, a drop in the groundwater
table to 20 m and subsequent rise to a depth of
3 m could explain the apparent preconsolidation
stress caused by changes in effective stress at the
Opelika site (Mayne 2013).
For CPTU, a generalized first-order evaluation of
σp' is made from (Mayne et al. 2009):
Figure 5. Fines content and mean grain size at Opelika.

where m' is an exponent that varies with soil
3 GEOCHARACTERIZATION OF PIEDMONT type: m' = 1.0 (clays), 0.9 (organic soils), 0.85
(silts), 0.80 (silty sands), and 0.72 (clean quartzitic
In this section, selected CPTU interpretations in sands). For the fine sandy silts of the Piedmont geol
Piedmont residuum at Opelika will be compared ogy, a value of 0.83 has been found suitable (Mayne
with laboratory results. 2013). The profile agrees well with the values from
one-dimensional consolidation tests, as evident from
Figure 6. The value of m' has also been related to
3.1 Laboratory testing of residuum mean grain size (D50), fines content (FC), and mater
An extensive set of drive samples and undisturbed ial index (Ic), as detailed by Agaiby & Mayne
tube samples were collected for laboratory testing. The (2019).
561
Figure 6. Yield stress and YSR at Opelika test site.
3.3 Triaxial friction angle of Piedmont soils

A total of 23 triaxial compression tests were per
formed on tube samples taken from the site (Vinson
& Brown 1997; Brown & Vinson 1998). A summary
of these tests is presented in Figure 7 giving an over
all effective stress envelope represented by the
Mohr-Coulomb parameters: c' = 0 and ϕ' = 35.5°.
At each sample depth, several CIUC type triaxials
were conducted where the confining stresses were
applied either at the in-situ overburden, or approxi
mately half or about double the effective overburden.
Figure 8. Triaxial results for various depths at Opelika.
2012; Holmsgaard et al. 2016; Bihs et al. 2018).

This is a major conundrum for CPTU interpretation
in Piedmont fine sandy silts.
4.1 CPTU evaluation of ϕ' in sands

At the standard rate of 20 mm/s, CPTU in clean
sands is considered drained response. Various
methods for evaluating ϕ' from CPTU in sands are
available (Ching et al. 2017) including an approach
from Robertson & Campanella (1983):
Figure 7. Summary triaxial tests at Opelika test site.
Consequently, an evaluation of the secant effect A method based on corrected CPT chamber tests
ive friction angle at six depths is made in Figure 8. from Kulhawy & Mayne (1990):
Values of ϕ' ranged from 33.5° to 37.1°.
4 CPTU EVALUATION OF FRICTION ANGLE
For CPTU in silts, it is not initially clear whether the

evaluation should be drained, undrained, or inter where σatm ≈ 1 bar = 100 kPa. A modified form of
mediate, such as partially-drained (DeJong et al. this is given by Robertson & Cabal (2015):
562
where Qtn = (qnet/σatm)/(σvo'/σatm)n is a normalized
net cone tip resistance that has a variable exponent In fact, data on over 105 different clays tested under
that ranges from about 1 in clays to 0.75 in silts to both triaxial compression and CPTU have been cali
around 0.5 in sands. brated to show they give comparable ϕ' values
For the Opelika CPTU, the various normalized (Ouyang & Mayne 2018).
cone resistance parameters (qt/σvo', qt1, Q, Qtn) are
shown in Figure 9a. It can be stated that these pro
files are quite similar. 4.3 CPTU evaluation of ϕ' in fissured geomaterials
For the case where Bq < +0.05, eqn (6) is not valid
and results from CPTUs at Opelika show Bq ≈ 0, in
fact, technically the Bq values are negative and aver
age around -0.02 to -0.05, as evidenced by
Figure 9b. Negative porewater pressures are often
recorded in fissured geomaterials, such as stiff over-
consolidated clays (Mayne et al. 1990), but also
observed in residual soils (Schneider et al. 2001;
Finke et al. 2001).
From a measurement viewpoint, Campanella &
Robertson (1988) showed the u2 readings in a stiff
clay were either slightly negative or slightly positive
depending on the specific filter element, and thus u2
can be affected by the thickness, width, and actual
location of the porous element. Furthermore, studies
by DeJong et al. (2007) show that type of fluid
(water, oil, glycerine, silicone), its viscosity, and
Figure 9. Normalized CPTU parameters at Opelika: (a) degree of saturation play a role in porewater pressure
cone resistance; (b) porewater pressure ratio. measurements during CPTU. Of final note, the NTH
solution cannot handle negative Bq (Sandven 1990).
In fact, a recent study of 27 sands and silty sands As such, a value of Bq = 0 is assumed at Opelika
that were sampled undisturbed using special freezing for CPTU at standard rates of 20 mm/s. An approxi
methods and/or gel samplers confirmed the relation mation for (5) when Bq = 0 can be expressed
ships given by both (3) and (4) by comparison with (Ouyang & Mayne 2019):
laboratory triaxial compression tests (n = 63) and
field CPTU (Uzielli & Mayne 2019).
4.2 CPTU evaluation of ϕ' in clays 4.4 Effective ϕ' from Ic relationship
At the standard CPTU rate of 20 mm/s in clays, Using the database on 27 undisturbed sands and silty
response is taken to be undrained, corresponding to sands, a relationship was also found between ϕ' and
no volumetric strains (DeJong et al. 2012). Clays are CPT material index, Ic (Mayne 2020):
identified when the CPT index Ic ≥ 2.95 (Robertson &
Cabal 2015). In consideration of clayey silts, a value
of Ic > 2.6 is often taken to be “undrained” response.
A limit plasticity solution developed at the Trond
heim Institute of Technology (NTH) for CPTU evalu
ation of ϕ' under undrained conditions is available for values of Ic ≤ 2.6.
from Senneset et al. (1989) that relates Q to ϕ' and Bq: The profile of Ic at Opelika is presented in
Figure 10 showing the intermediate geomaterial
more or less follows the threshold value of Ic = 2.60
to 11 m depth.
4.5 CPTU evaluation of ϕ' in Piedmont silts

An approximation for ϕ' is expressed directly as
The above 4 drained equations [i.e., (2), (3), and (4)]
a function of Q and Bq for the following ranges: 0.05
one undrained method [i.e., eqn (7)], and CPT index
≤Bq < 1.0 and 18° < ϕ' < 45° (Mayne 2007):
563
expression [i.e., eqn (8)] are all applied to the mean 5 CONCLUSIONS
CPTU data at Opelika, as shown in Figure 11. Inter
estingly, all 5 methods approximately agree with Interpretation of CPTU in residual soils is compli
each other. Moreover, all CPT expressions agree cated by their mixed constituency of clay, silt, sand,
well and provide comparable profiles to the effective and rock particles, as well as considerations of par
friction angles obtained from the lab triaxial com tially drained, fully-drained, and undrained behavior,
pression test series. This likely only occurs for this particularly at the standard rate of 20 mm/s. Never
very silty fine sand to very sandy silt because of its theless, success was shown for CPTU in residual
high fines content (average FC = 44%) and material silts and sands (ML-SM) at the Opelika national test
index at the undrained-drained border of Ic = 2.60, site, located within the Appalachian Piedmont geo
since the mean value of index Ic ≈ 2.7 for this site. logic province in eastern USA. Specifically, a means
to profile the yield stress ratio (YSR) from
a generalized approach that uses net cone resistance
and an exponent m' = 0.82. Moreover, a surprising
agreement is found in the assessment of effective
friction angle ϕ' by use of both drained and
undrained CPTU equations that compare well with
series of triaxial compression tests.
ACKNOWLEDGMENTS
Thanks to Dr. Fikret Atalay for sharing recent lab

data from Opelika NGES, Alabama.
REFERENCES
Agaiby, S.S. and Mayne, P.W. 2019. CPT evaluation of
yield stress in soils. Journal of Geotechnical & Geoen
vironmental Engineering 145 (12): doi:10.1061/(ASCE)
GT.1943-5606.0002164.
Anderson, J.B., Montgomery, J., Jackson, D., Kiernan, M.
and Shi, C. 2019. Auburn University national geotech
nical experimentation site in Piedmont residuum. AIMS
GeoSciences 5(3): 645–664. DOI: 10.3934/
geosci.2019.3.645
Figure 10. Profile of CPT material index at Opelika. Atalay, F. 2019. Engineered transition zone systems for
enhanced heat transfer in thermo-active foundations.
PhD Dissertation, Civil & Env. Engineering, Georgia
Institute of Technology, Atlanta GA: 221 pages.
Benoît, J. and Lutenegger, A.J., editors 2000. National
Geotechnical Experimentation Sites. GSP 93, ASCE,
Reston, Virginia: 397 pages.
Bihs, A., Nordal, S., Long, M., Paniagua, P. and
Gylland, A. 2018. Effect of piezocone penetration rate
on the classification of Norwegian silt. Cone Penetra
tion Testing 2018, (Proc. CPT’18, Delft), CRC Press,
Taylor & Francis, London: 143–149.
Blaker, Ø., Carroll, R., Paniagua, P., DeGroot, D.J. and
L’Heureux, J-S. 2019. Halden site: geotechnical charac
terization of a post glacial silt AIMS Geosciences 5 (2):
184–234.
Brown, D.A. and Vinson, J. 1998. Comparison of strength
and stiffness parameters for a Piedmont residual soil.
Geotechnical Site Characterization, Vol. 2 (Proc. ISC-1,
Atlanta), Balkema, Rotterdam: 1229–1234.
Campanella, R.G. and Robertson, P.K. 1988. Current status
of the piezocone test. Penetration Testing 1988, Vol. 1
(Proc. ISOPT, Orlando), Balkema, Rotterdam: 93–116.
Ching, J., Lin, G-H., Chen, J-R. and Phoon, K-K. 2017.
Transformation models for effective friction angle and
Figure 11. Profile of effective friction angle from five relative density calibrated based on generic database of
CPTU expressions in comparison with triaxial results on coarse-grained soils. Canadian Geot. J. 54 (4):
natural Opelika sand-silt mixture. 481–501.
564
DeJong, J.T., Yafrate, N.J. and DeGroot, D.J. 2007. Design Soil Mechanics & Geotechnical Engineering, Vol. 4
of a miniature piezoprobe for high resolution stratigraphic (ICSMGE, Alexandria), Millpress/IOS Press Rotterdam:
profiling, ASTM Geotech. Testing J. 30 (4): 1–13. 2777–2872.
DeJong, J.T., Jaeger, R.A., Boulanger, R.W., Randolph, M. Mayne, P.W. 2013. Evaluating yield stress of soils from
F. and Wahl, D. 2012. Variable penetration rate cone laboratory consolidation and in-situ cone penetration
testing for characterization of intermediate soils. Geo- tests. Sound Geotechnical Research to Practice, (Proc.
technical and Geophysical Site Characterization 4, GeoCongess, San Diego, GSP 230), ASCE, Reston/
Vol. 1 (Proc. ISC-4, Recife), Taylor & Francis, London: VA: 406–420.
25–42. Mayne, P.W. 2020. The 26th Széchy Lecture: Use of in-situ
Finke, K.A., Mayne, P.W. & Klopp, R.A. 2001. Piezocone geotechnical tests for foundation systems. Proceedings
penetration testing in Atlantic Piedmont residuum. Journal of the Széchy Károly Emlékkonferencia, published by
of Geotechnical & Geoenv.Engineering 127 (1): 48–54. the Hungarian Geotechnical Society, Budapest: 12–73.
Giacheti, H.L., Bezerra, R.C., Rocha, B.P. and Ouyang, Z. and Mayne, P.W. 2018. Effective friction angle
Rodrigues, R.A. 2019. Seasonal influence on cone pene of clays and silts from piezocone penetration tests. Can
tration test: an unsaturated soil site example. Journal of adian Geotechnical Journal 55 (9): 1230–1247.
Rock Mechanics & Geotechnical Engineering 11: Ouyang, Z. and Mayne, P.W. 2019. Modified NTH method
361–368. for assessing effective friction angle of normally con
Holmsgaard, R., Nielsen, B.N. and Ibsen, L.B. 2016. Inter solidated and overconsolidated clays from piezocone
pretation of cone penetration testing in silty soils con tests. Journal of Geotechnical & Geoenvironmental
ducted under partially drained conditions. Journal of Engineering 145(10), doi:10.1061/(ASCE)GT.1943
Geotechnical and Geoenvironmental Engineering 142 5606.0002112
(1) ASCE: 04015064. https://doi.org/10.1061/(ASCE) Robertson, P.K. and Cabal, K. 2015. Guide to Cone Pene
GT.1943-5606.0001386 tration Testing for Geotechnical Engineering, 6th edi
Hoyos, L.R. & Macari, E.J. 1999. Influence of in-situ fac tion, Gregg Drilling, Signal Hill, CA: 140 pages.
tors on dynamic response of Piedmont residual soils. Sandven, R. 1990. Strength and deformation properties of
ASCE Journal of Geotech. & Geoenvironmental Engrg fine-grained soils obtained from piezocone tests. PhD
125 (4), 271–279. Dissertation, Dept. of Civil Engrg., Norwegian Institute
Huffman, J.C., Martin, J.P. and Stuedlein, A.W. 2015. of Technology, Trondheim: 605 pages.
Assessment of reliability-based serviceability: Limit Schneider, J.A., Peuchen, J., Mayne, P.W., and
state procedures and full-scale loading tests. Proceed McGillivray, A.V. 2001. Piezocone profiling of residual
ings, Fifth International Symposium on Geotechnical soils. Proc. Intl. Conf. on In-Situ Measurement of Soil
Safety & Risk (5th ISGSR, The Netherlands), IOS Press, Properties & Case Studies, Bali, Indonesia: 593–598.
Rotterdam: 313–319. www.usucger.org
Lehane, B.M., Ismail, M.A. and Fahey, M. 2004. Seasonal Senneset, K., Sandven, R., and Janbu, N. 1989. Evaluation
dependence of in-situ test parameters in sand above the of soil parameters from piezocone tests. Transportation
water table. Geotechnique 54 (3): 215–218. Research Record 1235, National Academy Press, Wash
Lunne, T., Robertson, P.K. and Powell, J.J.M. 1997. Cone ington, DC: 24–37.
Penetration Testing in Geotechnical Practice, EF Spon/ United States Department of Agriculture (USDA) 2021,
Routledge, Taylor & Francis Group, London: 418 p. Natural Resources Conservation Service, Washington,
Mayne, P.W., Kulhawy, F.H., and Kay, J.N. 1990. Observa DC: https://www.nrcs.usda.gov/wps/portal/nrcs/detail/
tions on the development of pore water pressures during wa/soils/ ?cid=nrcs144p2_036333
piezocone penetration in clays, Canadian Geotechnical Uzielli, M. and Mayne, P.W. (2019). Probabilistic assign
Journal 27 (4): 418–428. ment of effective friction angle of sands and silty sands
Mayne, P.W. & Brown, D.A. 2003. Site characterization of from CPT using quantile regression. Georisk: Assess
Piedmont residuum of North America. Characterization ment & Management of Risk for Engineered Systems
and Engineering Properties of Natural Soils, Vol. 2, and Geohazards 13(4): Taylor & Francis, London:
Swets and Zeitlinger, Lisse: 1323–1339. 272–275. DOI: 10.1080/17499518.2019.1663388
Mayne, P.W. 2007. NCHRP Synthesis 368: Cone Penetra Vinson, J.L & Brown, D.A. 1997. Site characterization of
tion Testing, Transportation Research Board, Washing the Spring Villa geotechnical test site and a comparison
ton, DC: 118 p. www.trb.org of strength and stiffness parameters for a Piedmont
Mayne, P.W., Coop, M.R., Springman, S., Huang, A-B., residual soil, Report No. IR-97-04, Highway Research
and Zornberg, J. 2009. State-of-the-Art Paper (SOA-1): Center, Harbert Engineering Center, Auburn University,
Geomaterial behavior and testing. Proc. 17th Intl. Conf. AL: 385 p.
565
© 2022 the Author(s), ISBN 978-1-032-31259-0
Soil unit weight prediction from CPTs for soils and mining tailings
T. Menegaz & E. Odebrecht

University of Santa Catarina State, Joinville, Santa Catarina, Brazil
H.P. Nierwinski
Federal University of Santa Catarina, Joinville, Santa Catarina, Brazil
F. Schnaid
Federal University of Rio Grande do Sul, Porto Alegre, Rio Grande do Sul, Brazil
ABSTRACT: An accurate estimation of the total unit weight (γt) of soils and mining tailings is
a requirement for several geotechnical engineering applications. In the interpretation of CPT testing data, the
value of γt is needed for the evaluation of both the total and effective stresses, as well as in the accurate deter
mination of subsequent parameters. Current methods for determining γt based CPT data have been developed
for a wide range of natural soils where the specific gravity of solids (G) generally ranges from 2.5 to 2.7.
However, a large gap exists for soils or mining tailings where G is often higher, such as in the case of bauxite
and zinc. This paper presents a compilation of results of CPT tests, and laboratory measurements of specific
gravity of solids and total unit weight on bauxite and zinc mine tailings. The application of the literature
methodologies was evaluated for this database and the statistical coefficient of determination (r) was pre
sented, which is determined between the measured value and the value estimated by the correlations
1 INTRODUCTION resistance (qt), the lateral friction (fs), and also the
value of the shear wave velocity (vs) through the soil.
An adequate geotechnical investigation is necessary These methods consist of empirical formulations
for the understanding of soil behavior. Laboratory and developed from natural soil databases, encompassing
field data are generated, which allow the designer to a range of specific gravity of solids (G) values, usually
predict the behavior of the studied soil, offering more ranging from 2.5 to 2.7 (MAYNE, 2007). Some spe
security for project development. According to cific proposals seek to address organic soils, for
Lunne, Robertson & Powell (1997), field and labora G values ranging from 1.45 to 2.33 (LENGKEEK et al
tory tests complement each other; however, field tests 2018). However, considering the extensive applicabil
have some attractions. One of these is the cone pene ity of cone tests, encompassing both soft soil deposits
tration test (CPT) or piezocone test (CPTu), which is and mining tailings, there is a need to evaluate correl
conducted by driving a set of steel rods and a conical ations for estimating soil unit weight for wider ranges
point into the ground with the aid of a reaction of specific gravity of solids values, for which the cor
system mounted on the ground surface. The following relations available in the literature may present
quantities are continuously measured in this test: tip limitations.
resistance (qc), lateral friction (fs), and generated pore In this context, the present study compiles and
pressure (u). Combining these values also makes it organizes a database of cone tests and laboratory
possible to determine two parameters used to evaluate tests, encompassing bauxite and zinc mining tail
the soil behavior, namely the friction ratio (Rf) and ings soils with a specific gravity of solids values
the pore pressure parameter (Bq). The literature pre above the average of natural soils. The aim is to
sents several correlations for obtaining geotechnical verify the applicability of existing specific weight
parameters from CPTu test results (Kulhawy & estimation relationships and verify which method
Mayne (1990); Schmertmann (1978); Teh & Houlsby best adapts to these data. The study will be based
(1991); Larsson & Mulabdic (1991)). However, one on statistical analysis with the help of the statis
of the essential information in the interpretation of tical software R Studio®, where the value of
the test is the value of the natural soil unit weight. Pearson correlation coefficient (r), determined
The methods presented in the literature for deter between the measured value and the estimated
mining γt from the results of the cone test use the tip value of specific weight, will be evaluated. The
DOI: 10.1201/9781003308829-81
566
Pearson correlation coefficient (r) represents the identified a relationship between the plasticity index
linear association between two random variables. and the ratio between the tip resistance and the
Usually, the analysis is made by graphical repre depth, denominated mq (mq= qt/z). Analyses were
sentation of the relation of data, using a scatter performed and two equations were obtained (Equa
diagram. This coefficient varies from -1 to +1 and tions 3 and 4).
indicate the tendency of one variable to increase
or decrease with the other variable, respectively
(PATTEN & NEWHART, 2017; ZHOU et al
2016). In this paper, the Pearson correlation coef
ficient (r) is used to verify the relation between
the measured soil unit weight and the estimated
value by literature correlations.
A differentiated methodology is proposed by Mayne
(2014) to estimate the soil unit weight through CPTu
1.1 Methods for estimating the natural soil unit test results, relying only on the relationship of spe
weight cific weight with the lateral cone friction ðfsÞ meas
There are several proposals in the literature for deter urements. The proposed equation is presented by
mining the natural soil unit weight (γt); among them, Equations 5 and 6.
we can mention Robertson & Cabal (2010), Mayne
& Peuchen (2012) and Mayne (2014). To develop
the correlation proposed by Robertson & Cabal
(2010), as can be seen in Equation 1, experiments
and correlations between wave velocity and soil unit
weight were combined (MAYNE, 2007), along with
relationships between unit weight and DMT test
results (MARCHETTI, 1980). Thus, Robertson &
Cabal (2010) developed approximate contours of
unit weight values as a function of dimensionless 2 DATABASE
parameters of resistance (qt /σatm ) and lateral fric
tion (Rf ¼ fs=σatm ). In order to achieve the research objective,
a database of geotechnical investigation results was
analyzed. The database is formed by 197 observa
� � tions resulting from investigations in 4 different
γt =γ w ¼ 0:27 log Rf þ 0:36 ½logðqt=σatm Þ Þ] þ 1:236 Brazilian bauxite and zinc deposits. The items sur
ð1Þ veyed in the database were based on the definition
of the main parameters to be used in the statistical
analysis to apply existing correlations, thus making
where Rf = friction ratio x 100%; qt = corrected tip it possible to verify their performance with the
resistance; γw = water unit weight (kPa); and σatm = compiled data. The data for bauxite and zinc
atmospheric pressure (kPa). mining tailings are presented, by depth range, in
Robertson & Cabal (2010) also point out that the Tables 1 and 2, respectively.
vast majority of soils present a specific gravity of
solids (G) in the range of 2.5 to 2.7. For soils with
G outside this range, some interference may occur in
the proposed correlation given by the previous equa
Table 1. Bauxite mining tailings database – parameters
tion, so they proposed Equation 2, where the value variation.
of G is introduced.
Depth γt qt
(m) (kN/m³) (MPa) fs (kPa) G Rf
γt ==γ w ¼ ½0:27½log Rf ] þ 0:36½logðqt =σatm Þ] þ 1:236]G=2:65
1 – 4.75 13.48 – 0.03 – 0.5 – 2.39 – 0.26 –
ð2Þ 20.43 0.99 13.36 3.36 14.48
5 – 10 14.48 – 0.03– 0.79 – 2.85 – 0.48 –

As for Mayne & Peuchen (2012), the authors’ first 21.45 2.0 17.49 3.22 3.58
evaluation consisted of relating the unit weight of
the materials with the respective plasticity index, 11 – 15 16.44 – 0.39 – 4.48 2.88 – 0.68 –
20.44 0.95 -23.29 3.14 3.63
verifying a tendency of reduction of the specific
weight with the increase of the plasticity of the 16 – 20 15.58 – 0.73 – 6.82 – 2.98 – 0.93 –
material. As the plasticity index is not obtained 20.26 1.95 19.11 3.14 2.74
using the cone test, the authors investigated and
567
Table 2. Zinc mining tailings database – parameters The use of Pearson correlation coefficient (r) was
variation. chosen to evaluate if those two variables, the meas
ured specific weight and, the estimated specific
Depth γt qt weight, are related to each other. Furthermore, it is
(m) (kN/m³) (MPa) fs (kPa) G Rf intended to analyze which method demonstrates to
have more compatibility with each soil type evalu
1–5 11.27 – 0.03 – 0.14 – 3.28 – 0.28 – ated. For each type of studied soil and correlations
13.58 0.22 2.57 3.35 1.56
graphs were generated. The results of the analyses
6 – 12 12.6 – 0.22 – 1.77 – 3.29 – 0.5 – are presented by soil type in Tables 3 and 4, being
14.3 1.10 16.99 3.37 1.62 bauxite tailings and zinc, respectively.
The bauxite mining tailings were the first to be
analyzed, bringing the strongest correlation value,
the method of Robertson & Cabal (2010), in Equa
tion 2, with an r = 0.42. This equation presents in its
3 VERIFICATION OF EXISTING METHODS IN
formulation the coefficient of specific gravity of
LITERATURE
solids (G). And the weakest correlation occurred in
two equations of different methods, the first by
In this step of the study, the methods present in the
Robertson & Cabal, Equation 1, and the second by
literature for natural soils as presented in step 1.1
Mayne & Peuchen (2012), Equation 4, presenting an
were used to estimate the soil unit weight, comparing
r = 0.2. The graphs with the strongest and weakest
the result obtained by applying the formulations with
correlations are illustrated in Figures 1 and 2.
the values defined in the results of geotechnical tests.
For zinc mining tailings, the results were more
Together with the database formed with the param
positive when compared to bauxite mining tail
eters for the analysis of correlations between CPT and
ings. For the Mayne (2014) method, the strongest
CPTu tests, the equations were entered into the Excel
correlation was obtained, being r = 0.85. How
software spreadsheet. Each of the correlations was
ever, not all results were above r = 0.7. Mayne
entered in a column, sorted by author and year of
development, starting from the oldest to the most
recent.
In order to make an evaluation between the values
calculated by the correlations of the methods and the
unit weight values present in the results of the geo
technical tests, the software R Studio® was used as
a statistical tool.
Table 3. Verification of methods with bauxite mining

tailings.
Method Equation r value
Robertson & Cabal (2010) Equation 1 0.2 Figure 1. The strongest correlation in bauxite mining
Robertson & Cabal (2010) Equation 2 0.42 tailings.
Mayne & Peuchen (2012) Equation 3 0.2
Mayne (2014) Equation 6 0.3
Table 4. Verification of methods with zinc mining

tailings.
Method Equation r value
Robertson & Cabal (2010) Equation 1 0.79

Robertson & Cabal (2010) Equation 2 0.78
Mayne (2014) Equation 6 0.85 Figure 2. The weakest correlation in bauxite mining
tailings.
568
and Peuchen (2012), in Equation 4, presents the Through this study, it was possible to define
weakest result among all, where r = 0.43. The which method in the literature for natural soils best
graphs with the strongest and weakest correl suited the mining tailings data in the database.
ations are presented in Figures 3 and 4. The Pearson correlation coefficient (r) values
between the measured data and the data calculated
by the methods indicated that some of the methods
offered a positive result when using the database, but
in most cases the results showed that there is a need
to generate a new equation that encompasses a range
of soils with actual specific gravity of solids values
varying more than just the range for normal soils
(2.5 to 2.7).
REFERENCES
Lengkeek, H. J.; De Greef, J.; Joosten, S. (ed) 2018. CPT
based unit weight estimation extended to soft organic
Figure 3. The strongest correlation in bauxite mining soils and peat. Cone Penetration Testing 2018:
tailings. 389–394.
Lunne, T; Robertson P. K.; Powell, J.J.M. (ed) 1997. Cone
Penetration Testing. London: Blackie Academic &
Professional.
Mayne, P. W. (ed) 2007. Cone Penetration Testing. Wash
ington, DC: The Nacional Academies Press.
Mayne, P. W.; Peuchen, J.; Bowmeester, D. (ed) 2010. Unit
weight trends with cone resistance in soft to firm clays.
Geotechnical and Geophysical Site Characterization:
Proceedings of the 4th International Conference on Site
Characterization ISC-4: 903–910.
Mayne, P.W. (ed) 2014. Interpretation of geotechnical
parameters from seismic piezocone tests. Proceedings,
3rd International Symposium on Cone Penetration Test
ing: 47–73.
Robertson, P. K.; Cabal, K.L. (ed) 2010. Estimating soil
unit weight from CPT. 2nd International Symposium on
Figure 4. The weakest correlation in bauxite mining Cone Penetration Testing: 2 – 40.
tailings. Robertson, P. K.; Cabal, K.L. (ed) 2014. Guide to Cone
Penetration Testing for Geotechnical Engineering. Cali
fornia: Gregg Drilling & Tests Inc.
Patten, M.L.; Newhart, M. (eds) 2017. Understanding
4 CONCLUSIONS Research Methods. London: Routledge.
Zhou, H.; Zhihong, D.; Xia, Y.; Fu, M. (ed) 2016. A new
This paper sought to assemble a database of Brazil sampling method in particle filter based on Pearson
ian mining tailings soils for verification of methods correlation coefficient. Neurocomputing 216: 208–
present in the literature for natural soils. 215.
569
© 2022 the Author(s), ISBN 978-1-032-31259-0
Use of DMT and CPTU to assess the G0 profile in the subsoil

Z. Młynarek
University of Life Sciences, Poznań, Poland
J. Wierzbicki
Institute of Geology, Adam Mickiewicz University, Poznań, Poland
P. Monaco
ABSTRACT: The subject of this article is the analysis of the relationship between G0/MDMT and KD, where
G0 is the small strain shear modulus, while MDMT and KD are respectively the constrained modulus and the
horizontal stress index determined from DMT tests. This relationship allows to determine a profile with depth
of G0 from standard DMT test results, useful when data from non-seismic DMT investigations are available.
The analysis was based on a large amount of data for a wide range of soils of different origins in Poland. The
dataset included OC and NC loams, silts, medium sands, silty sands and fine sands. The overconsolidation
ratio (OCR) was estimated from CPTU and DMT tests. The obtained empirical G0/MDMT vs. KD relationships
were compared with the correlations established by Marchetti et al. (2008) for different soil types. To account
for the significant influence of overconsolidation, an original empirical relationship between G0/σ’p and KD,
where σ’p is the preconsolidation stress, was determined based on data from all investigated fine-grained soils.
1 INTRODUCTION the quality of the testing technique (Młynarek 2010),

and (2) the parameter variability, related to subsoil
Recent achievements of research on in-situ testing properties (Lacasse & Nadim 1994). The analysis of
allow to determine a wide range of soil parameters the influence of the latter group of factors on the correl
necessary for geotechnical design of a planned invest ation between G0 and the horizontal stress index KD
ment. The piezocone (CPTU) and flat dilatometer from DMT, with particular emphasis on the preconsoli
(DMT) tests take up a special position in these investi dation effect, is the subject of this paper.
gations. In the group of geotechnical parameters sought
for design of foundations of many engineering struc
tures, the parameters related to soil stiffness, such as 2 LOCATION AND BASIC GEOLOGICAL
the small strain shear modulus G0 and the constrained SETTING OF THE TEST SITES
modulus M, play an important role. The CPTU and
DMT tests are particularly convenient for determining The research was carried out at five test sites located
the profiles with depth of these moduli in the subsoil. in Poland (Figure 1), four in the area covered by the
In particular, the seismic piezocone (SCPTU) and the Late Pleistocene glaciation and one in the impact
seismic dilatometer (SDMT) tests permit to determine zone of periglacial processes in the Pleistocene. Two
the in-situ depth profile of G0 from the measured shear sites, Kaźmierz and Lipno, are located within the
wave velocity Vs. The cost of these techniques, how Weichsel glaciation, ended about 15,000 years ago.
ever, is higher than the standard non-seismic CPTU The near-surface geological formations are dominated
and DMT tests, which are preferred in many practical by loam and glacial sands over 12 m thick. In the
cases. In absence of Vs measurements, G0 can be esti subsurface zone loams are mainly originated from
mated from the results of a standard DMT using the melt-out facies and characterized by low preconsoli
empirical correlations proposed by Marchetti et al. dation. They rest on loams of the lodgement facies
(2008). The use of such correlations must be accom with a higher degree of preconsolidation (Kaźmierz
panied by the awareness that many independent vari site). The clay fraction in glacial clays does not
ables affect the quality of the assessment of the G0 exceed 25%. They are dominated by the sand frac
modulus. It is generally accepted that the quality of tion, even up to 70%. The sites of Jarocin and
a geotechnical parameter depends on two factors: (1) Koźmin are located in the Riss glaciation zone, in the
DOI: 10.1201/9781003308829-82
570
3 ESTIMATE OF OCR FROM CPTU AND DMT
The stress history of soil deposits is commonly

assessed based on the overconsolidation ratio (OCR)
or the preconsolidation stress (σ’p). Profiles with
depth of these parameters at the investigated sites
were obtained from CPTU and DMT data.
Based on CPTU test results, the OCR was esti
mated using the relationship (Wierzbicki 2010):
where Qt = normalized cone resistance.

Figure 1. Location of the test sites. The original formula by Marchetti (1980) was
used to estimate the OCR from DMT:
foreground of the Weichsel glaciation line. Here the
dominant soils are loams of the lodgement facies,
characterized by a high (> 10%) calcium carbonate
content, a sand fraction lower than the loams of the
Weichsel glaciation and an increase (up to 40%) in where KD = horizontal stress index.
the silt fraction. These clays are often called ‘grey The OCR profiles obtained from CPTU and DMT
loams’ for their characteristic grey-brown color. The were used to qualify the investigated soil deposits as
Łańcut site is located in the zone of loess covers and normally consolidated (NC) or overconsolidated
loess-like silts. These soils were originated in the (OC). The OCR was found to vary between 1 to 11.
Middle and Late Pleistocene as extensive covers of The comparison of OCR depth profiles estimated
aeolian silty sediments, deposited in the foreland of from CPTU and DMT at two representative sites is
the Pleistocene ice sheet. The particle size distribu shown in Figures 2 and 3, along with the profiles of
tion is very homogeneous and characterized by 60 measured and interpreted parameters obtained from
70% of silt fraction and about 20% of sand fraction. CPTU (corrected cone resistance qt, friction ratio Rf,
A characteristic feature of the loess covers is the fre pore pressure u2) and from DMT (corrected pressure
quent occurrence of carbonate cementation, which readings p0 and p1, material index ID). A relatively
causes a quasi-preconsolidation effect (Młynarek large difference between the OCR estimated from
et al. 2013). CPTU and DMT is found at the Kaźmierz site in loam
Figure 2. CPTU and DMT test results and estimated OCR profiles in loam at Kaźmierz test site (LI = liquidity index, saCL
= sandy loam, saclSi = fluvioglacial silt, MSa = fluvioglacial sand).
571
Figure 3. CPTU and DMT test results and estimated OCR profiles in loess at Łańcut test site (LI = liquidity index, Si-siCl
= loess, siSa = fluvial sand).
and sandy loam (Figure 2), particularly in correspond The correlations formulated by Marchetti et al.
ence of thin sand layers. However the mean values of (2008) were taken as a reference for the comparative
the OCR from DMT and CPTU for the entire subsoil analysis with the experimental data obtained in this
profile do not differ significantly (level of confidence study. A total number of 989 SDMT test results,
interval of statistical analysis α = 0.05). The average obtained from soil deposits of different origin,
values over the 9 m investigated depth show a limited macrostructure and OCR, were analyzed. The ana
standard deviation. The loess deposit at the Łańcut site lysis of the relationship written in the form of Equa
(Figure 3) is composed of silty clay and silt with homo tion 3 was carried out considering the following soil
geneous macrostructure. The OCR estimated from groups, corresponding to the soil classification
CPTU and DMT are similar, and their average values adopted by the Polish Standards (1986): clay (clay
over the 12 m investigated depth do not differ signifi fraction > 30%); sandy loam (clay fraction = 10
cantly from the statistical point of view. 20%, sand fraction 50-90%); loam; silt (clay fraction
10-20%, silt fraction 30-60%); fine/medium sand;
fine/silty sand.
4 ASSESSMENT OF THE CORRELATION Figure 5 shows the datapoints G0(m)/MDMT vs.
BETWEEN G0 AND DMT PARAMETERS KD obtained for all soil groups, where G0(m) is
intended as the G0 determined from VS measured
Marchetti et al. (2008) investigated the experi by SDMT. Figure 5 confirms the conclusions for
mental interrelationship between small strain and mulated by Marchetti et al. (2008): (i) a functional
working strain stiffness using SDMT. They iden correlation exists between the ratio G0/MDMT and
tified a relationship between the ratio G0/MDMT KD; (ii) a correlation between these variables
and KD having the functional form shown by should be constructed for specific groups of soils, at
Equation 3: least distinguishing between fineand coarse-grained
soils; (iii) these correlations are affected by the pre
consolidation effect, which can be defined by σ’p
or OCR.
These findings prove that it is necessary to ana
where G0 = small strain shear modulus obtained lyze the relationship between G0(m)/MDMT and KD in
from the measured shear wave velocity VS as G0 = selected soil groups. Figures from 6 to 11 illustrate,
ρ·VS2 (ρ = soil density), MDMT = constrained modu for each soil group: (a) the relationship between
lus and KD = horizontal stress index obtained from G0(m)/MDMT and KD, compared with the equations
standard DMT interpretation (Marchetti 1980). proposed by Marchetti et al. (2008) for similar soil
Based on data obtained from 34 sites, in a variety of types (Figure 4); (b) the comparison between G0(m)
soil types, Marchetti et al. (2008) proposed distinct obtained from measured VS and G0(c) calculated
correlations for clay, silt and sand (Figure 4). according to Marchetti et al. (2008).
572
Figure 7 refers to sandy loam. The G0(m)/MDMT
vs. KD datapoints (Figure 7a) plot relatively close
to the correlations proposed by Marchetti et al.
(2008) for clay and silt. The moduli G0(m)
(Figure 7b) are very similar to the moduli G0(c) cal
culated according to Marchetti et al. (2008) for
clay. Again, the lower values of G0(m) and G0(c) lie
near the 1-1 correlation line. The best-fit relation
ship for the G0(m)/MDMT vs. KD datapoints in
Figure 7a is given by Equation 5:
Figure 8 refers to loam. The G0(m)/MDMT vs.

Figure 4. Relationship between the ratio G0/MDMT and KD KD datapoints (Figure 8a) plot slightly higher
according to Marchetti et al. (2008) (from Monaco et al. than the correlations proposed by Marchetti et al.
2009).
(2008) for clay and silt. Accordingly, the moduli
G0(m) (Figure 8b) slightly differ from the moduli
G0(c) estimated according to Marchetti et al.
(2008) for clay. The best-fit relationship for the
G0(m)/MDMT vs. KD datapoints in Figure 8a is
given by Equation 6:
Figure 9 refers to silt. The G0(m)/MDMT vs. KD

datapoints (Figure 9a) plot rather close to the correl
ation proposed by Marchetti et al. (2008) for silt.
The moduli G0(m) (Figure 9b) are similar to the
moduli G0(c) calculated according to Marchetti et al.
(2008) for silt. The best-fit relationship for the
G0(m)/MDMT vs. KD datapoints in Figure 9a is given
Figure 5. Relationship between the ratio G0(m)/MDMT and by Equation 7:
KD in different soil types from all investigated sites.
Figure 6 refers to clay. The G0(m)/MDMT vs. KD

datapoints (Figure 6a) from the investigated sites
(OC clay) plot higher than the correlation proposed
by Marchetti et al. (2008) for clay. The moduli G0(m) Figure 10 refers to fine/medium sand. The G0(m)
obtained from the measured VS differ significantly /MDMT vs. KD datapoints (Figure 10a) plot very
from the moduli G0(c) calculated according to close to the correlation proposed by Marchetti et al.
Marchetti et al. (2008), particularly for the higher (2008) for sand. The moduli G0(m) (Figure 10b) are
values of G0(m), while for low G0(m) the values of very similar to the moduli G0(c) calculated according
G0(c) and G0(m) are closer to the 1:1 correlation line to Marchetti et al. (2008) for sand. The best-fit rela
(Figure 6b). The same effect was found by Młynarek tionship for the G0(m)/MDMT vs. KD datapoints in
et al. (2013) comparing the constrained moduli Figure 10a is given by Equation 8:
M from DMT and CPTU tests for Pliocene OC
clays. The best-fit relationship for the G0(m)/MDMT
vs. KD datapoints in Figure 6a is given by
Equation 4:
Figure 11 refers to fine/silty sand. The G0(m)/MDMT

vs. KD datapoints (Figure 11a) plot very close to the
573
Figure 6. (a) Relationship between G0(m)/MDMT and KD in clay. (b) Comparison between G0(m) obtained from measured VS
and G0(c) calculated according to Marchetti et al. (2008) for clay.
Figure 7. (a) Relationship between G0(m)/MDMT and KD in sandy loam. (b) Comparison between G0(m) obtained from meas
ured VS and G0(c) calculated according to Marchetti et al. (2008) for clay.
Figure 8. (a) Relationship between G0(m)/MDMT and KD in loam. (b) Comparison between G0(m) obtained from measured
VS and G0(c) calculated according to Marchetti et al. (2008) for clay.
574
Figure 9. (a) Relationship between G0(m)/MDMT and KD in silt. (b) Comparison between G0(m) obtained from measured VS
and G0(c) calculated according to Marchetti et al. (2008) for silt.
Figure 10. (a) Relationship between G0(m)/MDMT and KD in fine/medium sand. (b) Comparison between G0(m) obtained
from measured VS and G0(c) calculated according to Marchetti et al. (2008) for sand.
correlations proposed by Marchetti et al. (2008) for KD in cohesive soils, a new correlation was estab
silt and for sand. The moduli G0(m) (Figure 11b) are lished in which G0(m) was normalized by the precon
very similar to the moduli G0(c) calculated according solidation stress σ’p. The value of σ’p was
to Marchetti et al. (2008) for sand. The best-fit rela determined from the OCR estimated from DMT.
tionship for the G0(m)/MDMT vs. KD datapoints in Figure 12 shows the G0(m)/σ’p vs. KD datapoints
Figure 11a is given by Equation 9: obtained from the tested soils belonging to the group
of cohesive soils (sandy loam, loam, clay and silt).
The two areas denoted as A and B, that plot far away
from the general G0(m)/σ’p – KD trend line, are
related to fissured clays and cemented silts. These
soils need a separate interpretation. The best-fit rela
The lowest values of the coefficient R2, obtained tionship for the G0(m)/σ’p vs. KD datapoints in
for clay (R2 = 0.6102) and for fine/silty sand (R2 = Figure 12 is given by Equation 10:
0.6582), are very similar to those found by Marchetti
et al. (2008) in clay and sand (Figure 4). Possibly in
these soils the assessment of the G0(m)/MDMT – KD
relation is affected by local variability in grain size
distribution and macrostructure at the various sites. This correlation has a satisfactory statistical evalu
Due to the significant impact of overconsolidation ation and, in principle, could be applied to all cohe
on the functional shape of the relationship G0(m) vs. sive soils to estimate G0 based on KD.
575
Figure 11. (a) Relationship between G0(m)/MDMT and KD in fine/silty sand. (b) Comparison between G0(m) obtained from
measured VS and G0(c) calculated according to Marchetti et al. (2008) for sand.
depth of G0 in soils of different origin based on

standard DMT test results.
d. Fissured clays and cemented soils require
a separate extensive analysis and interpretation.
e. The relationship G0(m)/σ’p = f (KD) can be con
sidered very promising for practical purposes,
because it allows to determine G0 for the whole
set of OC cohesive soils.
REFERENCES
Lacasse, S. & Nadim, F. 1994. Reliability issues and future
challenges in geotechnical engineering for offshore
structures. Proc. 7th Int. Conf. on Behaviour of Offshore
Structures, Cambridge MA, USA, 1: 1–48. Amsterdam:
Figure 12. Relationship between G0(m)/σ’p and KD for Elsevier.
sandy loam, loam, clay and silt. Marchetti, S. 1980. In Situ Tests by Flat Dilatometer.
J. Geotech. Eng. Div. 106(3): 299–321.
Marchetti, S., Monaco, P., Totani, G. & Marchetti, D. 2008.
5 CONCLUSIONS In Situ Tests by Seismic Dilatometer (SDMT). Geotech.
Spec. Publ. GSP 180, From Research to Practice in
The obtained research results allow to formulate the Geotechnical Engineering: 292–311.
following general conclusions. Młynarek, Z. 2010. Quality of In-Situ and Laboratory Tests
Contribution to Risk Management. Proc. 14th Danube
a. In the construction of the correlation G0/MDMT = European Conf. on Geotech. Eng., Bratislava, Slovakia.
f (KD) it is necessary to take into account soil Młynarek, Z., Wierzbicki, J. & Stefaniak, K. 2013.
type, with a general distinction between cohesive Deformation characteristics of overconsolidated subsoil
from CPTU and SDMT tests. Proc. 4th Int. Conf. on
and non-cohesive soils. This fully confirms the
Geotechnical and Geophysical Site Characterization,
concept by Marchetti et al. (2008). Porto de Galinhas, Brazil, 18-21 September 2012, 2:
b. Overconsolidation has a significant influence on 1189–1193. London: Taylor & Francis Group.
the functional form of the G0/MDMT = f (KD) rela Monaco, P., Marchetti, S., Totani, G. & Marchetti, D. 2009.
tionship. For the examined OC loams, especially Interrelationship between small strain modulus G0 and
for clays, the Marchetti et al. (2008) correlations operative modulus. Performance-Based Design in
require some correction. Grain size of the soil and Earthquake Geotechnical Engineering – from Case His
overconsolidation are significant. On the other tory to Practice, Proc. IS-Tokyo 2009, Tsukuba, Japan,
hand, a satisfactory agreement between the G0 15-17 June 2009, 1315–1323. London: Taylor & Francis
Group.
measured and calculated according to Marchetti
Polish standards. 1986. Building soils, Nomenclature, clas
et al. (2008) was found for medium/fine sands and sification and description. PN-86/B-02480.
silts, and in general for NC soils (low KD). Wierzbicki, J. 2010. Evaluation of subsoil overconsolidation
c. The proposed G0(m)/MDMT = f (KD) correlations by means of in situ tests at the aspect of its origin. Scien
have a high statistical significance and may prove tific dissertations No. 410, University of Life Sciences in
useful to determine a continuous profile with Poznań Publishing, Poland, 181–182 (in Polish).
576
© 2022 the Author(s), ISBN 978-1-032-31259-0
Physical and numerical modelling of T-CPT for mechanisms of penetration

and heat transfer
P.Q. Mo & L. Gao
China University of Mining and Technology, Xuzhou, China
H.S. Yu
University of Leeds, Leeds, UK
X.L. Tao
Jiangsu Vocational Institute of Architectural Technology, Xuzhou, China
Q.Z. Ma
CCCC-FHDI Engineering Co., Ltd. GuangZhou, China
ABSTRACT: Thermal-Cone Penetration Test (T-CPT) is proposed as a new type of in-situ soil testing method
for estimations of both mechanical and thermal properties of soil layers. Physical modelling of T-CPT is con
ducted in this study to obtained the effects of soil density, penetration depth and heating duration on the penetra
tion resistance and thermal responses. Discrete Element Method (DEM) is then adopted to simulate the
processes of penetration, heating and cooling, for investigations of penetration induced stress and temperature
distributions in the surrounding soil. The measurements of thermal responses are analyzed to validate the simula
tion and to evaluate the interpretation method for thermal properties. The results could be used as a benchmark
for further analyses of mechanisms of penetration and heat transfer. The interpretation method for T-CPT data is
to be developed to improve the performance, with implications to utilization of geothermal energy.
1 INTRODUCTION and thermal responses of the probe during pene

tration (Akrouch et al. 2016; Vardon et al. 2019;
Utilization of geothermal energy has become one Mo et al. 2021). As a convenient in-situ testing
of the most important and practicable approaches technique, the T-CPT avoids the sampling disturb
for solving energy shortage with rapidly growing ance for laboratory tests and overcomes the
demand and environmental problems (EIA 2009; defects of high cost and long testing time of in-
Capareda 2019). Geothermal energy pile, serving situ thermal response test. This tool is useful and
as a new type of pile foundations to extract and/ convenient in practical applications by inserting
or inject heat into ground at shallow depths using the probe into certain depths and conducting ther
heat carrier medium in heat exchanger piles, can mal tests during the pause of penetration. It is of
provide majority of required heating/cooling great significance to the development and utiliza
energy, save energy costs, and reduce fossil fuel tion of geothermal energy.
demand and carbon footprint (Rotta Loria and In order to look insight into the macro-micro pene
Laloui 2017; Sani et al. 2019). However, the tration and thermal mechanisms of T-CPT, a type of
measurements and determinations of thermal T-CPT probe and its testing method are proposed in
properties are crucial to the evaluation of geo this study to measure the thermal responses of both
thermal energy in the soil profile and the estima heating and cooling stages after certain penetration
tion of heat amount to be extracted or injected depths. The penetration and heat conduction of
into the soil layers (Loveridge et al. 2015). Ther T-CPT are then simulated using discrete element
mal cone penetration test (T-CPT) is a new in- method, with investigation of soil movement, force
situ testing method based on the traditional cone chain and penetration resistance during probe push
penetration test, by introducing additional tem ing. The thermal responses during heating and cooling
perature sensors and/or heating elements, which are also examined to evaluate the heat transfer mech
aims to simultaneously measure both mechanical anisms and the interpretation methods.
DOI: 10.1201/9781003308829-83
577
2 PHYSICAL MODELLING OF T-CPT
2.1 Physical model and penetrometer of T-CPT

This study attempts to conduct physical model-
ling of T-CPT, to develop a new type of in-situ
CPT based testing method and to investigate the
relations between soil properties and penetration
and temperature responses. The testing system is
assembled by the container with soil, penetrom
eter, actuator and data acquisition, as shown in
the schematic of Figure 1a. This model with
1 m cubic size is small enough to conduct full-
size (36 mm diameter) penetration tests without
introducing significant boundary effects. A T-CPT
penetrometer is proposed in this study, by install
ing a section with heating elements behinds the
traditional standard CPTu module (Figure 1b).
The heat section is embedded between two insu
lation sections, and three temperature sensors are
equipped at different locations in the heat and
insulation sections.
The soil used in the experiments is typical
sand in Xuzhou, China, with an average diameter
of d50 = 0.5 mm, uniformity coefficient of cu =
2.44, and curvature coefficient of cc = 1.05. The
thermal properties are measured by Hot Disk
Analyser at CUMT, giving that thermal conduct
ivity k ¼ 1:15W=m=K and specific heat cap Figure 1. Physical model for T-CPT and design of
acity c ¼ 850J=kg=K. penetrometer.
The probe is pushed at various depths (i.e.
z ¼ 180; 360; 540mm) for heating and cooling tests,
and the heating periods are designed with 60, 120,
300, and 600 s, to investigate the effects of stress con
dition and heating time. Temperature sensors are also
embedded in the soil at various locations to obtain the
temperature distributions during heating and cooling.
Note that the physical model is a full-scale model with
limited boundary effects, and thus the scale effect for
mechanical and thermal measurements should be
negligible.
2.2 Results of penetration

Th results of cone tip resistance in soils with dif
ferent relative density are presented in Figure 2.
The two tests in dense sand with relative density
of 80 % provide comparable developments of
penetration resistance, validating the repeatability
of the testing method. The maximum qc is
reached at a magnitude of 1 MPa after a depth of
about 180 mm, which is 5 times of probe diam
eter. This surface influence zone is relatively
larger, owing to the stress gradient and the low
stress condition without surcharge. The changes
of tip resistance after 360 mm of penetration are
presumably attributed to the inhomogeneous soil
layers and the approximating bottom boundary.
The loose sample, with relative density around
50 %, generally shows 40 % smaller cone tip
resistance. Figure 2. Cone tip resistance against penetration depth.
578
2.3 Results of heating and cooling a given moment during heating and cooling, contrib
uting as an important benchmark for numerical and
The heating and cooling tests are conducted following
theoretical calculations.
the pause of penetration. The temperature responses
The results of Sensor 1 in dense sand, with vari
of three sensors installed in the probe during the heat
ous penetration depth (i.e. z ¼ 10; 20; 30R, R is
ing for 600 s in dense sand and the subsequent cool
probe radius) and different heating duration
ing are described in Figure 3a. The temperature of
(theat ¼ 60; 120; 300; 600s) are systematically ana
Sensor 1, which is located at the center of heating
lyzed. Similar trends are obtained, and the traditional
elements, rises to 58 ºC at 600 s, whereas the increase
interpretation method of ASTM (2014) and Vardon
for Sensor 2 (at side of heating elements) is less than
et al. (2019) gives an average estimation of thermal
25 ºC. The Sensor 3 is mounted in the insulation sec
conductivity with k pred ¼ 1:27W=m=K, which is
tion, which is less affected by the heating elements.
10% overestimation comparing to the calibrated
However, the gradual increase of 8 ºC is attributed to
1:15W=m=K from elementary tests. This indicates
heat conduction through soil. Both measurements
that the proposed T-CPT penetrometer and testing
from Sensor 1 and Sensor 2 can be directly used for
procedure with limited heating and cooling durations
interpretation of soil thermal properties.
need to be applied with an improved interpretation
The temperature response of soil is depicted in
method based on further investigation of mechan
Figure 3b, based on the three embedded temperature
isms of penetration and heat transfer.
sensors at a similar depth with the heating elements
of the probe. The radial distances to the center of
probe is 20, 40, and 60 mm, respectively. It is clear
3 DEM SIMULATION OF T-CPT
to notice that the trends of temperature change are
similar to the probe sensors, whereas the curves are
more gradual, and the peak appears later with longer 3.1 Mechanical and thermal properties of soil
distance for heat transfer. It also shows the magni sample
tude of temperature distribution around the probe at To investigate the mechanisms of penetration and
heat transfer of T-CPT, the testing processes with
penetration, heating and cooling are simulated using
three-dimensional Discrete Element Method (DEM).
The soil is prepared to mimic the Ottawa 20-30
sand, with an average diameter of d50 = 0.725 mm,
uniformity coefficient of cu = 1.13, and curvature
coefficient of cc = 1.01. Cubic samples with side
length of 13 mm are created using about 5800 spher
ical particles for triaxial tests, to verify its mechan
ical behaviour. In order to maintain a similar initial
void ratio with e0 = 0.650 before shearing, isotropic
consolidation tests under different confining pres
sures are conducted by setting corresponding fric
tional coefficient between particle contacts.
Considering the calculation efficiency and accuracy,
the finish of consolidation is defined by the follow
ing criterions according to Zhao et al. (2020):
where βf = unbalanced force ratio, fjc and fjb =

total contact force and body force of the j th particle,
f c = contact force; γf = error of mean stress, σ0 =
designed confining pressure, p0 = mean stress of par
Figure 3. Temperature responses of sensors in probe and ticle assembly; and δf = ratio of deviatoric and mean
soil. stresses, q = deviatoric stress of particle assembly.
579
The consolidated samples under confining pres
sure of 0.1 MPa, 2.0 MPa, and 3.5 MPa are adopted
to conduct undrained triaxial shear tests with
a constant strain rate of 0.08 %/s. The particle-
particle friction coefficient is 0.3, and both normal
and shear contact stiffnesses are set as 3 × 102 kPa,
using linear elastic contact model. The results of tri
axial shear tests show that the critical states of sam
ples are reached with large shear strain for ε1 430%,
and the ultimate stress ratios lead to the estimation
of internal friction angle with cs ¼ 21:5� . The soft
ening phenomenon after peak stress ratio is more
obvious for sample with lower confining pressure,
and the dilation is restrained under high stress condi
tion, which indicating the reasonability of the mech
anical properties of the granular material.
The thermal properties of soil samples in this
study are calibrated by elementary tests for one-
dimensional heat conduction. The unit thermal
resistance Tc in DEM model is defined by the fol
lowing heat transfer relationship between two con
tacted particles:
where ΔT = temperature difference between par

ticles, ni = direction of heat transfer, L = distance
between particle centers, qi = heat flux of unit length.
The thermal dilation is also considered to include the
thermal effects on stress-strain behaviour of granular
material, by defining a linear relation with:
Figure 4. Results of one-dimensional thermal conduction

for calibration of thermal properties.
where α = coefficient of linear expansion, R =
radius of particle, and ΔR = temperature induced
change of particle radius. sections, whereas the outer particles (i.e. distance
Taking α ¼ 2 x 10-7 K-1 for silicon sand, soil to center larger than 21 mm) are expanded uni
samples with over 10,000 particles and identical par formly to reduce the total particle number to
ticle size distributions are calibrated to examine the around 55,000. The soil samples are initially con
variation of heat conductivity against samples with solidated by gravity and then compressed with an
different unit heat resistance and void ratio. overburden pressure of 200 kPa, resulting an
Figure 4a presents the stabilized temperature field average initial void ratio of e0 = 0.614. The par
after 600 s of heat conduction with a constant tem ticle density is 2650 kg/m3, and frictional coeffi
perature gradient of 1 K/cm. The results of heat con cient between particles is again set as 0.3.
ductivity varying with unit heat resistance and void A penetrometer with diameter of 6 mm and apex
ratio are shown in Figure 4b, which also serve as angle of 60º is modelled for T-CPT tests, with
a benchmark for the design of thermal properties of a proportion of 1/6. The frictional shaft is set behind
granular material. the cone shoulder with a length of 5 mm, and the
heating section is in coincidence with the frictional
3.2 Numerical model of T-CPT shaft. To evaluated the scale effect, the ratio of probe
diameter and average particle size is approximately
According to the axisymmetric model, a quarter 8.3, and the ratio of sample size to probe size is
model with 90º sector is adopted in this study, larger than 14 to reduce the boundary effect. The
while the soil parameters are calibrated through penetration is conducted at a constant velocity of
the aforementioned elementary tests. To minimize 20 mm/s, and the total penetration depth is about
the calculation time for a physical model of 36 mm. The initial temperature field is set as 20 ºC,
T-CPT, the soil samples with radius of 42 mm and the heating and cooling processes are simulated
and height of 45 mm are prepared as previous after the pause of penetration. According to the
580
calibration chart of Figure 4b, the unit thermal resist
ance is set as 2.5×106 K=W=m, giving an equivalent
thermal conductivity of k ¼ 1:1W=m=K. The spe
cific heat capacity is 960 and 460 J=kg=K for probe
and soil, respectively. The heat power is 0.04 W, due
to the scaled heating element in the miniatured
probe. The heating period lasts for 60 s and the sub
sequent cooling stage is longer than 420 s. The
numerical results of penetration and heat transfer are
described in the following sections. With these
results, the soil distortion, mechanical developments
and heat transfer during penetration, heating and
cooling are clear to understand the T-CPT mechan
isms, which could contribute to the interpretation of
T-CPT data in practical applications.
3.3 Results of penetration

0
The normalized cone tip resistance qc =σv 0 against
normalized penetration depth z=B is shown in
Figure 5. The penetration resistance increases with
depth, and tends to reach a relative constant magni
tude with 18 times of initial vertical stress after
a transition zone with a depth of 2.5 times probe Figure 6. Results of one-dimensional thermal conduction
diameter. for calibration of thermal properties.
3.4 Results of heat transfer

During the heating and cooling stages after the pause
of penetration at z=B ¼ 6, the heat flux is transferred
from heating element embedded in the probe to the
surround soil through heat conduction. The tempera
ture fields at various heating and cooling periods are
depicted in Figure 7, showing the evolution of heat
conduction. It is obvious to see that the heat transfer in
soil is mainly in a spherical profile, owing to the rela
tively small size of heating element. The contours
shown in Figure 7 also describe the influence zone of
heating. After 60 s of heating, the surrounding soil
with a size of probe diameter is affected by notable
temperature change (i.e. 1 ºC). The size of this influ
ence zone increases with time through heat conduction,
Figure 5. Normalized cone tip resistance during penetration.
while the magnitude of temperature dissipates during
the cooling stage.
The penetration induced soil stress distributions at

various penetration depths (i.e. z=B ¼ 2; 4; 6) are
described in Figure 6 for both radial and vertical
stress fields. It is clear to notice that the stress distri
bution around the penetrometer is mainly developed
adjacent to the cone tip. Comparing to the develop
ment of radial stress, the degradation of vertical
stress is generally concentrated in vertical directions
below the cone face. Apart from the growing pene
tration induced stresses, the influence zone is also
increased with penetration depth. The results of
stress fields provide the meso-scale information of
penetrometer-soil interaction, along with the force Figure 7. Results of one-dimensional thermal conduction
chain distributions and evolution during penetration. for calibration of thermal properties.
581
The measurements of temperature in the probe The interpretation of temperature response is
are obtained at both heat section and insulation sec shown in Figure 8b and 8c, based on the methods of
tion, as presented in Figure 8a. The blue curve indi Vardon et al. (2019) for both heating and cooling
cates the temperature at the center of heating stages. Although the heat power is relatively small
element, which rises to 34.9 ºC at 60 s of heating and the heat transfer in granular material is complex,
and reduces significantly in the following cooling the back-calculated heat conductivity is not far from
stage. The element at the heat insulation section is the input value of k ¼ 1:1W=m=K. The heating data
hardly affected, while limited temperature change gives the predicted magnitude of heat conductivity
(<0.8 ºC) is induced by the heat transfer from kh ¼ 1:149W=m=K, with 4.5 % of overestimation.
heated soil. Additionally, the error is larger for cooling data,
resulting in kc ¼ 1:19W=m=K, even though the cool
ing stage lasts for over 400 s. It may be contributed
to the boundary effect, since the side and bottom
boundaries are not far enough from the cone and the
they are set as adiabatic. Note that these differences
may be attributed to the material properties, espe
cially from the mechanical aspects. Detailed investi
gation is to be conducted to promote the simulation
of penetration and heat transfer and thus to improve
the interpretation method with limited heating and
cooling time in further studies.
4 CONCLUSIONS
Physical modelling and DEM simulation of T-CPT

are conducted in this study to investigate the mech
anisms of penetration and heat transfer. A type of
T-CPT probe with both temperature sensors and
heating elements is proposed, together with its test
ing method for heating and cooling after a pause of
penetration. Apart from the measurement of penetra
tion resistance, the thermal responses of sensors in
the probe and soil are obtained to examine the tem
perature changes and heat conduction during heating
and cooling stages. The existing interpretation
method is shown to overestimate the thermal con
ductivity of soil. DEM model is later validated to
simulate the processes of T-CPT, after calibration of
thermal properties of granular material. The distribu
tions and evolutions of stresses and temperature in
soil are presented to illustrate the meso-scale mech
anisms of penetration and heat conduction. The
simulated thermal responses are analyzed to validate
the numerical method, and to evaluate the existing
interpretation method. The results of this study are
important to understand the mechanisms of probe-
soil interactions, which could contribute to the fur
ther developments of T-CPT and its testing and inter
pretation methods.
ACKNOWLEDGEMENT
The authors would like to acknowledge financial sup

ports from National Natural Science Foundation of
China (Grant No. 51908546, Grant No. 52178374),
China Postdoctoral Science Foundation (Grant
No. 2020T130699), and Key University Science
Figure 8. Results of one-dimensional thermal conduction Research Project of Jiangsu Province (Grant
for calibration of thermal properties. No. 20KJA560003).
582
REFERENCES Mo, P.Q., Ma, D.Y., Zhu, Q.Y. & Hu, Y.C. 2021. Inter
pretation of heating and cooling data from thermal
Akrouch, G.A., Briaud, J.L., Sanchez, M. & Yilmaz, R. cone penetration test using a 1D numerical model
2016. Thermal cone test to determine soil thermal and a PSO algorithm. Computers and Geotechnics
properties. J. Geotech. Geoenviron. Engng. 142(3): 130: 103908.
04015085. Rotta Loria, A.F. & Laloui, L. 2017. Thermally induced
ASTM. 2014. D5334-14: Standard test method for deter group effects among energy piles. Géotechnique 67(5):
mination of thermal conductivity of soil and soft rock by 374–393.
thermal needle probe procedure. West Conshohocken, Sani, A.K., Singh, R.M., Amis, T. & Cavarretta, I. 2019.
PA, USA: ASTM International. A review on the performance of geothermal energy pile
Capareda, S.C. 2019. Introduction to Renewable Energy foundation, its design process and applications. Renew
Conversions. CRC Press. DOI: 10.1201/ able and Sustainable Energy Reviews 106: 54–78.
9780429199103. Vardon, P.J., Baltoukas, D. & Peuchen, J. (2019). Interpret
EIA. 2009. Renewable Energy Consumption and Electricity ing and validating the thermal cone penetration test
Preliminary Statistics 2008. (T-CPT). Géotechnique 69(7): 580–592.
Loveridge, F., Olgun, C.G., Brettmann, T. & Powrie, W. Zhao, S.W., Zhao, J.D. & Guo, N. 2020. Universality of
2015. Group thermal response testing for energy piles. internal structure characteristics in granular media under
Eur. Conf. Soil Mech. Geotech. Eng., vol. XVI. shear. Physical review. E 101(1–1).
583
© 2022 the Author(s), ISBN 978-1-032-31259-0
Soil stratigraphy from seismic piezocone data and multivariate clustering in

alluvial soil deposits: Experience in the Lower Tagus Valley region
F. Molina-Gómez, D. Cordeiro, C. Ferreira & A. Viana da Fonseca
CONSTRUCT-GEO, Faculdade de Engenharia da Universidade do Porto (FEUP), University of Porto, Portugal
ABSTRACT: The identification of soil stratigraphy at a given site is crucial for geotechnical analysis and
design. This study addresses the delineation of soil stratigraphy from seismic piezocone penetration test
(SCPTu) measurements. For this purpose, SCPTu data from an experimental site in the Lower Tagus Valley
region (close to Lisbon) were compiled. Soil stratigraphy was obtained by applying a multivariate clustering
approach using the direct measurements reported by SCPTu. The hierarchical clustering results were compared
against the soil behaviour index profile. Moreover, the differences/similitudes obtained from multivariate clus
tering were validated by contrasting statistical results against the visual description of samples collected in the
experimental site at Lower Tagus Valley using advanced sampling techniques; namely, Gel-Push sampler. The
main findings showed that the clustering approach implemented herein detects and groups soil layers with
similar soil behaviour types, allowing delineating the soil stratigraphy in multilayer alluvial soil deposits.
1 INTRODUCTION located at Lower Tagus Valley, south of Portugal.

Embedded layers of sandy and clayey soils with an
The identification of soil stratigraphy is one of the alluvial origin from the Tagus River compose such
main concerns for the characterisation of soils deposits. an experimental site (Viana da Fonseca, Ferreira,
A realistic identification of the soil variability in depth Ramos, & Molina-Gómez 2019). Besides, in such
reduces the failure probability of geotechnical struc a site, high-quality undisturbed samples were col
tures. The cone penetration test (CPT) and its enhanced lected using an advanced sampling technique—the
versions provide reliable identification of soil multi- Gel-Push (GP) sampler. The visual observation of
layering due to its simultaneous measurements. Hence, these samples confirmed the detected stratigraphy
soil stratigraphy can be assessed using CPT data by from the multivariate clustering.
correlating in situ measurements with parameters
inferred by applying a strong theoretical background.
Although no samples are recovered in CPT sound 2 DESCRIPTION OF THE CLUSTERING
ing, the large number of data from this type of testing APPROACH
allows identifying the soil stratigraphy. Therefore, such
data can be interpreted by statistical methods. Hegazy In this study, the agglomerative hierarchical clustering
& Mayne (2002); Liao & Mayne (2007); Molina- approach proposed by Molina-Gómez, Viana da Fon
Gómez, Viana da Fonseca, Ferreira, Sousa, & Bulla- seca, Ferreira, Sousa, & Bulla-Cruz (2021) was
Cruz (2021) applied cluster analysis approaches to applied. Such an approach comprises the following
delineate the stratigraphy in different types of soil phases: (i) selection of the data and identification of
deposits. In these studies, the authors compared their variables; (ii) standardisation of the variables; (iii)
statistical analyses against soil behaviour index profiles, computation of the distance (or similarity) matrix; (iv)
originally proposed by (Robertson 1990). Such application of clustering technique; (v) selection of
a comparison showed a good fitting between results, the cut level of the dendrogram, associated with the
validating the applicability of clustering for delineating best partition, for defining the number of clusters; (vi)
the soil stratigraphy. interpretation of cluster results; and (vii) validation by
This study addresses the delineation of soil stratig comparison against soil behaviour type profiles.
raphy from seismic piezocone (SCPTu) measure Besides, it includes a comparison against visual
ments by applying the clustering approach proposed examination of high-quality undisturbed samples col
by Molina-Gómez, Viana da Fonseca, Ferreira, lected using the GP sampler, as a novel validation in
Sousa, & Bulla-Cruz (2021). The SCPTu measure such a clustering approach. The clustering procedure
ments were carried out in an experimental site adopted for this study is described following.
DOI: 10.1201/9781003308829-84
584
For the data standarisation, the Zscore method was where P þ Q are clustered measurements of SCPTu
adopted, as suggested by Hegazy & Mayne (2002): variables and R is a possible new group.
The best-known result of a Hierarchical Classifi
cation is a graphical representation—dendrogram.
The quality of SCPTu data grouping (i.e. the number
of clusters) can be assessed by comparing the
number of soil classes identified by the soil behav
where Zij is the standardised value, xij is the SCPTu iour type (SBT) chart proposed by (Robertson
measurement at a certain depth of the SCPTu vari 1990). Afterwards, such a number is validated by the
able, Xj is the SCPTu variable of each profile, BxðXj Þ Average Silhouette Method (Rousseeuw 1987). The
is the average of the SCPTu variable and sðXj Þ is the Average Silhouette Method computes the average
standard deviation of the SCPTu variable. silhouette of observations for different numbers of
The distance among pairs of objects is represented clusters (K), indicating Kopt value as the maximum
by a matrix DðnxnÞ (Härdle & Simar 2015). Equation 2 average silhouette over a range of possible values
describes the matrix distance, D, that contains measures for a different number of clusters.
of similarity among the n objects, where di;j are the dis
tances between the measurements of SCPTu variables.
Nowadays, there are different similarity measure

ments, including Euclidean distance, Manhattan dis
tance, Pearson, cosine and power (Johnson &
Wichern 2007). Härdle & Simar (2015) claimed that
no specific rule or theory is available on which to Figure 1. Experimental site map and location of the inves
base the choice of similarity measurements. In this tigation points.
study, Euclidean distance was used to compute the
matrix D of piezocone data, as recommended by As before noticed, the underlying mathematics of
(Liao & Mayne 2007). Such a method estimates the cluster analysis is simple to implement. However, it
geometric distance—based on the Pythagorean the requires extensive calculations because of the large
orem—of two observations in a multidimensional amount of data. Therefore, all computations were con
space. Equation 3 describes the Euclidean distance ducted in RStudio, which is a free integrated environ
of two observations, where k is the number of vari ment for statistical programming and graph plotting
ables and i is the variable index. (Molina-Gómez, Bulla-Cruz, & Darghan 2019).
3 EXPERIMENTAL SITE
Within the scope of two research projects on soil lique

faction developed in the CONSTRUCT-GEO research
centre of FEUP, a vast site characterisation campaign
Starting with the matrix D (see Equation 2) an iterative was carried in an experimental site in the Lower Tagus
procedure takes place to join/grouping the most similar Valley (south of Portugal). The site characterisation
classes and update matrix D, which ends when all data comprised the definition of liquefaction risk by micro
is grouped in a particular cluster (Bulla-Cruz, Lyons, & zonation maps (Saldanha, Viana da Fonseca, & Fer
Darghan 2021). The iterations are carried out according reira 2018). This geotechnical test site is located in the
to a specific aggregation criterion. In this study, the municipalities of Vila Franca de Xira and Benavente,
complete linkage method was applied. The algorithm covering a zone with 14683 ha area and 50.8 km per
of complete linkage distance is: imeter. In situ tests for geotechnical characterisation
were performed at specific locations, referenced as site
investigation points (SI). Figure 1 shows the map and
SI locations within the experimental site.
585
For this study, SI15 was selected since there Figure 3 presents the interpretation of the soil pro
advanced sampling techniques were implemented for files in terms of soil type chart. Classification results
collecting high-quality samples for the first time in indicate that the soil profile at SI-15 has five differ
Portugal (Molina-Gómez, Viana da Fonseca, Ferreira, ent soil types, which are characteristic of soil
Ramos, & Cordeiro 2021). The groundwater level in deposits in the Lower Tagus Valley from the Quater
SI15 was identified at about 2.5 m depth. Moreover, in nary (Ferreira, Viana da Fonseca, Ramos, Saldanha,
this point, an SCPTu was carried out. Shear wave vel Amoroso, & Rodrigues 2020).
ocities (Vs) were measured at each 0.50 m depth. The The soils of the experimental site appear in zones
values of cone resistance (qc ), sleeve friction ( fs ) and 3, 4, 5 and 6, thus typically behave as `clay to silty
pore-water pressure generated during cone penetration clay’, `clayey silt to silty clay’, `silty sand to sandy
(u2 ) were averaged each ±0:25 m depth (ranges of silt’ and `clean sand to sand to silty sand’, respect
0.5 m), considering the depth of the VS measurements, ively. Points located in Zone 1 were not considered
to compare the SCPTu data profiles. Figure 2 presents for defining the number of clusters because there is
the SCPTu profiles and the averaged values in the spe not a well-defined characterisation for these soil
cific depths of VS measurements at SI15. types; that is, out-of-range zones with Ic ¼ N/A.
Based on the above consideration, the clustering of
SI15 can be performed considering four groups.
1000
7 8
Normalised cone resistance, Qtn
100
9
6
5
10 4 3
Figure 2. SCPTu profiles at SI15.

1 2
1
0.1 1 10
The SCPTu data were analysed using the software Normalised friction ratio, F r (%)
Cliq v.3.0. The analyses were carried out based on
the unified approach proposed by Robertson (1990), Figure 3. Results of soil classification type.
which includes the classification of soil behaviour
type (SBT) zones according to the soil behaviour
type index (Ic). Ic computation considers the follow
ing procedure: 4 ADVANCED SAMPLING
The collection of high-quality samples is critical to

obtain representative test results for soil characterisa
tion in the laboratory (Ferreira, Viana da Fonseca, &
Nash 2011). A typical sampling campaign involves
the following stages: drilling, sampler insertion, sam
pler retrieval, tube sealing, transport, soil extrusion,
sample storage and preparation for element testing
(Hight 2000). Nowadays, sampling processes are
conducted using diverse techniques, such as hydraul
ically activated push tubes, rotary devices and freeze
sampling, where the freezing method is the most
where qt is the cone resistance corrected for pore expensive (Viana da Fonseca & Pineda 2017).
water effects (qt ¼ qc þ u2 ð1 - aÞ); a is the cone In this research, an advanced sampling technique
area ratio; σv0 is the current in situ total vertical stress; developed by the Japanese geotechnical company
σ0v0 is the current in situ effective vertical stress. Kiso-Jiban Consultants, to develop a method able to
586
replace freezing techniques. There are four different
variations of GP: GP-Rotary, GP-Drilling, GP-
Triple, and GP-Static. In this study, the GP-Static
(GP-S) was implemented. This sampling device uses
a hydraulic activated fixed-piston and follows the
same principle of the Osterberg-type sampler.
On the other hand, the GP is an advanced sam
pling technique, which uses a viscous polymer gel
to collect high-quality undisturbed samples. The
gel use, as the main innovation for soil sampling,
afforded the name for this technique. The gel must
be prepared at a 1 vol% concentration ratio of the
viscous polymer in clean water (Mori & Sakai
2016). The purpose of using the viscous gel is to
significantly reduce the friction between the sample
and walls liner during both the insertion of the sam
pler into the ground and during the sample extru
sion in the laboratory (Viana da Fonseca, Ferreira,
Molina-Gómez, & Ramos 2019). Besides, the rheo
logical properties of the polymer gel allow preserv
ing the soil structure of the collected samples.
Figure 4. Schematic of GP sampler components (Molina-

Gómez, Ferreira, Ramos, & Viana da Fonseca 2020).
Figure 5. Sampling depths for each advanced technique
and averaged soil behaviour index profile at SI15.
Unlike conventional hydraulic activated fixed-
piston, the GP-static device includes three pistons: (i)
the stationary piston, (ii) the sampling tube-advancing
piston, and (iii) the core-catcher activating piston. At
the maximum performance, the GP sampler can col 5 ANALYSIS OF RESULTS
lect samples with 71 mm diameter and 1 m length.
However, according to the soil type to be sampled, the A cluster analysis in terms of the depth (Z) and
device can suffer damages in the cutting shoe using the standardised data recorded at SI15 by the
(Molina-Gómez, Viana da Fonseca, Ferreira, Ramos, four channels of SCPTu (e.g. qc , fs , u2 VS ) was car
& Cordeiro 2021). Details about the performance and ried out. Liao & Mayne (2007) used normalised data
quality sampling using the GP sampler are provided in to define the soil stratigraphy by applying
Molina-Gómez, Ferreira, & Ramos (2020). Figure 4 a multivariate approach (e.g. Qt and Fr parameters).
schematises the components of the GP-Static sampler. However, Molina-Gómez, Viana da Fonseca, Fer
The GP sampling was carried out in four bore reira, Sousa, & Bulla-Cruz (2021) obtained a better
holes located at about 5 m distance to the SCPTu groping using standardised raw data instead of stand
sounding. During the experimental campaign, ardised normalised data. These authors showed that
a total of 14 GP samples were collected at the SI15 the differences between clustering were because of
location. The sampling depths were selected con the uncertainties associated with the in situ stress-
sidering suitable layers for liquefaction assessment state assessment, which depends on the unit weight
by element testing in the laboratory. The liquefi (γ). Nevertheless, γ is a physical property that differs
able layers were defined by detecting zones with according to the CPT correlation applied, although
Ic52:6. Figure 5 illustrates the sampling depths the literature reports several correlations to estimate
for each technique. Moreover, in Figure 5, the such a physical parameter, such as Mayne, Peuchen,
sampling depths are contrasted against the aver & Bouwmeester (2010, Robertson & Cabal (2010,
aged soil behaviour index profile of SI15. Lengkeek, de Greef, & Joosten (2018).
587
In this study, only the clusters with at least two
continuous measurements are considered soil layers,
as suggested by Liao & Mayne (2007). Hence,
a minimum layer thickness of 1.0 m was defined.
Clusters with points without a continuous group of
data (single points) were considered as lenses or
transition zones (Hegazy & Mayne 2002). The clus
ter analysis was conducted for obtaining four groups,
which correspond to different soil behaviour types or
layers previously identified in Figure 3. Such
a number of groups was compared against the results
of the Average Silhouette Method, which allows esti
mating the optimum number of clusters, Kopt .
Figure 6 shows Kopt ¼ 4, validating the selection of
the number of clusters based on the SBT criterion.
Figure 6. Estimation of the optimum number of clusters.
Figure 7 presents the dendrogram obtained from
the cluster analysis of SCPTu data. This dendrogram
shows the clustering of all depths and the distances
between the aggregate clusters. A distance of about 5 In addition, Figure 7 indicate the SBT zone of
defined four clusters, which correspond to the layers each cluster through a comparison against the Ic pro
composing the soil stratigraphy of the studied zone. file SI15 sounding. This grouping represents the soil
Figure 7. Soil stratrography of SI15 inferred from cluster analysis and comparisons between dendrogram against Ic profile
and high-quality samples.
588
layers that behave as ‘silty sand to sandy silt’ ACKNOWLEDGEMENTS
(1 m to 1.5 m depth), ‘clean sand to sand to silty
sand’ (3 m to 7 m depth), ‘clayey silt to silty clay’ This work was also financially supported by UIDB/
(2 m to 3 m depth and 8 m 14.5 m depth) and ‘clay 04708/2020 and UIDP/04708/2020 of CON
to silty clay’ (12.5 m to 14.5 m depth and 15.5 m to STRUCT – Institute of R&D in Structures and Con
20 m depth). Besides, two transitional layers com struction funded by the national funds through the
posed of soils that behave as ‘silty sand to sandy silt’ FCT/MCTES (PIDDAC). The first author acknow
were identified at 7.5 and 15 m depth. ledges the Portuguese Foundation for Science and
On the other hand, Figure 7 shows photographs of Technology (FCT) for the support through SFRH/
GP samples collected at representative depths of the BD/146265/2019 grant.
layers identified in the cluster analysis (e.g. 4.5 m,
7 m, 11 m and 16 m depth). A visual examination in
the laboratory of GP samples supported cluster REFERENCES
results; that is, the sample collected at 4.5 m depth is
a sand, the sample collected at 7 m depth corres Bulla-Cruz, L. A., L. Lyons, & E. Darghan (2021). Com
ponds to a silt mixture, and the samples collected at plete-Linkage Clustering Analysis of Surrogate Meas
11 m and 16 m depth are clay with different content ures for Road Safety Assessment in Roundabouts.
Revista Colombiana de Estadística 44(1), 91–121.
of granular particles. Ferreira, C., A. Viana da Fonseca, & D. F. T. Nash (2011).
The simultaneous comparisons between dendro Shear wave velocities for sample quality assessment on
gram, Ic profile and high-quality GP validated the a residual soil. SOILS AND FOUNDATIONS 51(4),
implementation of the clustering approach to define 683–692.
the soil stratigraphy and detect interlayers from Ferreira, C., A. Viana da Fonseca, C. Ramos,
SCPTu data. Therefore, the clustering approach pro A. S. Saldanha, S. Amoroso, & C. Rodrigues (2020).
posed herein provided a reliable data grouping for Comparative analysis of liquefaction susceptibility
defining soil stratigraphy of alluvial soil deposits of assessment methods based on the investigation on
the Lower Tagus Valley—a region with high inter a pilot site in the greater Lisbon area. Bulletin of Earth
quake Engineering 18, 109–138.
layering. Moreover, this clustering approach can be Härdle, W. K. & L. Simar (2015). Applied Multivariate
applied to other datasets from other experimental Statistical Analysis (4 ed.). Springer Berlin Heidelberg.
sites using raw data from the four parameters of the Hegazy, Y. & P. W. Mayne (2002). Objective site character
SCPTu. This information provides a method that ization using clustering of piezocone data. Journal of
allows defining soil layers with similar properties Geotechnical and Geoenvironmental Engineering 128,
and behaviour, useful in the design of geotechnical 986–996.
structures. Hight, D. W. (2000). Sampling Methods: Evaluation of
Disturbance and New Practical Techniques for High
Quality Sampling in Soils. In 7 Congresso Nacional de
Geotecnia, Porto.
6 CONCLUSION Johnson, R. A. & D. W. Wichern (2007). Applied Multi
variate Statistical Analysis (6 ed.). Pearson Education
In this paper, a hierarchical clustering approach International.
for defining the soil stratigraphy from SCPTu Lengkeek, A., J. de Greef, & S. Joosten (2018). CPT based
data at a site investigation point located in the unit weight estimation extended to soft organic soils and
Lower Tagus Valley (south of Portugal) has been peat. In Cone Penetration Testing IV: Proceedings of the
applied. The approach comprises a complete- 4th International Symposium on Cone Penetration Test
linkage algorithm, formulated in RStudio, which ing (CPT’18), Delft, The Netherlands, pp. 389–394.
used the Euclidean distance of standardised Taylor & Francis Group.
Liao, T. & P. W. Mayne (2007). Stratigraphic delineation
SCPTu measurements (qc , fs, u2 and Vs). The by three-dimensional clustering of piezocone data.
method is objectively capable of classifying simi Georisk 1, 102–119.
lar groups of data in the soil profile, delineating Mayne, P., J. Peuchen, & D. Bouwmeester (2010). Soil unit
different layer boundaries and soil transitions. The weight estimation from CPTs. In 2nd International Sym
statistical results were validated by comparing the posium on Cone Penetration Testing.
dendrogram against the Ic profile and visual Molina-Gómez, F., L. Bulla-Cruz, & E. Darghan (2019).
examination of high-quality samples collected Profiles analysis as a modality of repeated measures for
using the GP sampler. These comparisons are the comparing grain size distributions in granular bases.
Measurement: Journal of the International Measure
main contribution of this study since they con
ment Confederation 146, 930–937.
firmed the correspondence between the defined Molina-Gómez, F., A. Viana da Fonseca, C. Ferreira,
stratigraphy (clusters) with the soil behaviour type C. Ramos, & D. Cordeiro (2021). Novel sampling tech
and soil composing the soil layers. Therefore, this niques for collecting high-quality samples: Portuguese
clustering approach can be applied to other data experience in liquefiable soils. In 6th International Con
sets from other experimental sites using raw data ference on Geotechnical and Geophysical Site Charac
from the four parameters of the SCPTu. terization, ISC’6, Budapest, pp. paper 115.
589
Molina-Gómez, F., C. Ferreira, & A. Ramos, C. Viana da Rousseeuw, P. J. (1987). Silhouettes: A graphical aid to the
Fonseca (2020). Performance of gel-push sampling in interpretation and validation of cluster analysis. Journal
liquefiable soils. Géotechnique Letters 10, 256–261. of Computational and Applied Mathematics 20, 53–65.
Molina-Gómez, F., A. Viana da Fonseca, C. Ferreira, Saldanha, A. S., A. Viana da Fonseca, & C. Ferreira (2018).
F. Sousa, & L. A. Bulla-Cruz (2021). Defining the soil Microzonation of the liquefaction susceptibility: case study
stratigraphy from seismic piezocone data: A clustering in the lower Tagus valley. Geotecnia 142, 07–34.
approach. Engineering Geology 287, 106111. Viana da Fonseca, A., C. Ferreira, F. Molina-Gómez, &
Mori, K. & K. Sakai (2016). The gp sampler: a new innov C. Ramos (2019). Collection of high-quality samples in
ation in core sampling. In Lehane, Acosta-Martínez, and liquefiable soils using new sampling techniques. In Pro
Kelly (Eds.), Geotechnical and Geophysical Site Char ceedings of the XVII ECSMGE-2019, pp. paper 014.
acterisation 5, pp. 99–124. Australian Geomechanics Viana da Fonseca, A., C. Ferreira, C. Ramos, & F. Molina-
Society. Gómez (2019). The geotechnical test site in the greater
Robertson, P. K. (1990). Soil classification using the cone lisbon area for liquefaction characterisation and sample
penetration test. Canadian Geotechnical Journal 27, quality control of cohesionless soils. AIMS Geosciences
151–158. 5, 325–343.
Robertson, P. K. & K. L. Cabal (2010). Estimating soil unit Viana da Fonseca, A. & J. Pineda (2017). Getting high-
weight from CPT. In 2nd International Symposium on quality samples in ‘sensitive’ soils for advanced labora
Cone Penetration Testing, Huntington Beach, CA, USA. tory tests. Innovative Infrastructure Solutions 2, 34.
590
© 2022 the Author(s), ISBN 978-1-032-31259-0
Undrained strength from CPTu in brittle soils: A numerical perspective

L. Monforte, M. Arroyo & A. Gens
CIMNE-UPC, Barcelona, Spain
ABSTRACT: Static liquefaction of soils that have a brittle undrained response (hydraulic fills, mine tailings
or sensitive clays) may lead to sudden failures of large consequence. Given the importance of undrained failure,
obtaining precise estimates of peak and residual yield strength is important. The CPTu plays a major role in the
geotechnical characterization of these geomaterials and so do CPTu-based estimates of undrained strength. Most
of the methods available for CPTu-based estimation of undrained strength are empirical, based on correlation
with other laboratory or field tests. When such correlations are established difficulties appear due to variable
disturbance affecting the reference laboratory samples and parasitic effects, such as unaccounted for partial
drainage during penetration or unknown side friction, affecting the cone results. Such difficulties are not present
when using numerical simulation. The paper builds upon a series of CPTu simulations using a model able to
represent brittle undrained failure. Confounding factors such as partial drainage and cone side friction are sys
tematically varied to examine their effect on the results. The results are then employed to examine the perform
ance of several empirical methods frequently employed to obtain peak and residual strength from CPTu.
1 INTRODUCTION work we exploit the simulation database of Monforte

et al (2021) to assess several cone interpretation tech
Some geomaterials (hydraulic fills, mine tailings or niques frequently employed in current practice to
sensitive clays) may exhibit undrained softening infer undrained strength.
upon shearing (static liquefaction): a drastic decrease
of the effective mean stress and mobilized undrained
shear strength. Static liquefaction is a brittle 2 NUMERICAL DATABASE
response that may lead to catastrophic sudden fail
ures, often associated with loss of life and major 2.1 Simulation method
environmental impact (Gens, 2019). The identifica
tion and characterization of these materials is there Numerical simulations (Monforte et al, 2021) have
fore of paramount importance. been carried out by means of G-PFEM (Geotech
The extraction of good-quality samples of these nical Particle Finite Element Method), specially
loose/soft soils is challenging and in situ tests are designed for the analysis of problems involving the
essential for their characterization (Been, 2016; penetration of rigid structures into fluid-saturated
Gens 2019). The interpretation of CPTu soundings soil masses (Monforte et al, 2017, 2018).
in these brittle materials is not always straightfor The cone has standard dimensions and is assumed
ward, particularly as their permeability does fre rigid. CPT symmetry allows for an axisymmetric
quently result in partly drained conditions (Schnaid, model. The stress state is set up by prescribing
2021). Nevertheless, several correlations are in use a vertical stress of 96 kPa at the top boundary and of
to relate cone metrics and peak and residual 56 kPa at the radial boundary, i.e. a K0 value of 0.58.
undrained shear strength in brittle soils. The CPT is advanced by 30 diameters into the soil
Numerical simulation of CPT may offer new attaining a steady state in all the simulated records.
insights of the mechanisms during insertion, leading
to more reliable interpretation techniques. To this end, 2.2 Materials simulated
Monforte et al (2021) reported a parametric analysis
employing the Particle Finite Element method The soil is described by a version of the Clay and
(PFEM) in which several materials with different Sand Model (CASM) (Yu, 1998; Manica et al. 2021;
idealized materials of variable undrained brittleness Arroyo & Gens, 2021). Common constitutive param
were considered in the simulation of CPTu. The ana eters for all cases examined are reported in Table 1.
lysis also assessed the effect of drainage conditions Materials with different brittleness index are obtained
and interface friction angle on cone metrics. In this by modifying the geometry of the yield surface (shape
DOI: 10.1201/9781003308829-85
591
parameter, n; and spacing ratio, r), see Table 2. The interface friction angles employed
Because of these choices all the simulated materials (δ ¼ 0; 7� ; 12� and 19� ) correspond to interface effi
share the same normal compression line, but their crit ciencies, tan (δ)/tan (), between 0 and 0.74, since
ical state line has different positions on the compres the soil friction angle is 24º.
sion plane. In undrained conditions, all materials share Currently, manufactured average roughness, Ra, for
the same peak undrained shear strength, whereas the CPTu friction sleeves may lie between 0.65μm and
residual undrained shear strength varies (Table 2). 0.15μm (EN ISO 22476-1), in a range that aims to
Example triaxial responses after anisotropic (K0) approach the roughness that may be later acquired
consolidation are reported in Figure 1. In undrained upon use. For fine grained soils the resulting normal
conditions, after reaching the peak, deviatoric stres ized roughness (Ra/D50) will typically vary in the range
ses reduce due to strain softening; material A is the (10-3 to 10-1) that has been experimentally shown
most brittle whereas material H is almost insensitive. (Subba Rao et al. 2000; Eid et al. 2014) to result on
interface efficiency between 0.3 and 0.9. On this basis
the higher values of interface friction in the parametric
study would be representative of testing on clays and
fine silts, whereas the lower values would be more rep
resentative for testing on coarser silts and sands.
Table 1. Constitutive parameters common for all materials.
κ λ M G K0 e0 OCR
(kPa)
0.016 0.053 0.98 3000 0.58 1 1
Table 2. Constitutive parameters varied.
Material n r S peak
u S res
u �
(kPa) (kPa)
A 10 12 23.5 6.9 0.093

E 8 4 23.8 13.2 0.052
H 4 2 24 22.1 0.026
3 SIMULATION RESULTS AND

CLASSIFICATION CHARTS
Figure 2 reports the results in two classification

charts proposed by Robertson (1991; 2016). The
1991 chart is expressed using the following normal
ized parameters
Figure 1. K0 triaxial compression behavior of materials A,
E and H: undrained loading, top, and drained loading, bottom.

whereas the 2016 chart is expressed using the fol
lowing normalized parameters
2.3 Parametric study

For every CASM material, the parametric analysis
presented below examines the effect of drainage
conditions and soil-cone interface friction angle. where n is an index, which is ultimately dependent
Drainage conditions are controlled through perme on Qtn and Fr, and is established by iteration.
ability values, which have been varied between 10-9 In the Bq vs Q1 classification graph the dominant
m/s and 10-3 m/s. To avoid clutter most results effect is that of permeability. The undrained simula
below are only presented for three values: 10-8 m/s tions have always Bq > 0.5, those drained have Bq = 0
(fully undrained), 10-6 m/s (partly drained) and 10-3 and the partly drained ones lie in between. For imper
m/s (fully drained). meable materials brittleness plays also a significant
592
role, with material A (most brittle) generating the classify as “clay-like” and be expected to behave in
larger pore pressures. Interface friction has no system an undrained manner under CPTu. Finally, it is
atic effect. According to this chart Material A will clas noticeable how changes on interface friction result in
sify as sensitive, fine grained, if impermeable, as clay significant shifts along the Fr axis for the same brittle
to silty clay if partly drained and as a silt mixture if ness and permeability.
fully drained.
4 PEAK UNDRAINED SHEAR STRENGTH
4.1 Cone factors

One of the most frequent and important applications
of the CPTu is to obtain estimates of peak undrained
strength. The dominant approach uses empirically
determined cone factors, that relate cone measure
ments and undrained strengths. Cone factors based on
tip resistance and excess pore pressure are defined as
It has been stressed many times (e.g. Mayne & Peu

chen, 2018) that for this approach to make sense the
target undrained strength of the correlation should be
clearly specified. Peak undrained strengths measured
on triaxial compression after anisotropic (K0) consoli
dation are adopted here as a target, in line with previous
work (e.g. Karlsrud et al. 2005). Another important
factor (Lunne et al. 1997) is sample disturbance, as it
will affect the laboratory values employed in correl
ations. Poor sample quality will typically decrease the
undrained peak strength measured on the laboratory
and thus increase empirically determined cone factors.
Figure 2. Results of the analyses of CPTu test with differ

ent permeabilities and interface friction angles plotted on
classification charts. Above: Robertson 1991. Below:
Robertson 2016. The marker color indicates undrained
brittleness; marker shape indicates side friction values;
marker hue indicates pemeability. PD = Partly drained.
In the Fr vs Qtn classification graph all numerical

results plot in the “contractive” zone, which is correct
Figure 3. Cone factors for peak undrained strength in
as none of the CASM materials were dilatant in terms of the brittleness for smooth, undrained simulations.
shear. For material A most simulated results fall into
the “sensitive” zone (CCS), which seems correct.
However, this is also the case for material E if partly
or fully drained, which seems excessive. There is also Figure 3 reports the cone factors in terms of the
some disagreement with the graph when the effect of brittleness index for smooth, undrained simulations.
permeability is considered. Results for all materials The tip resistance cone factor Nkt is far more variable
and interface frictions shift upwards as permeability than the pore pressure cone factor NΔu, something
is increased. Although this is in line with the trend in well in line with empirical observations (Lunne et al.
the graph, even the fully drained material would 1997). Furthermore, the values of NΔu obtained from
593
these undrained smooth simulations cluster around
the value 7.5, well supported by experimental work
on Norwegian clays (Paniagua et al. 2019).
The advantage of NΔu as a more stable factor disap
pears when partial drainage is introduced in the picture
Which is closely correct for the undrained results
(Figure 4). Excess pore pressure reduces and the cone
and far off the mark for partially drained cases
factor required to recover peak undrained strength
(Figure 6).
reduces accordingly. The effect of interface friction on
this trend appears to be relatively small. The more real
istic frictional simulations result on slightly smaller
NΔu than the empirical average, something that may be
explained by a small amount of sampling disturbance.
Figure 5. Simulation results and relations from literature

between cone factor Nkt and pore pressure parameter Bq.
Figure 4. Simulation results and relations from literature

between cone factor NΔu and pore pressure parameter Bq.
The variability of Nkt is sometimes accounted for

using Bq as an auxiliary variable (Lunne et al. 1997;
Mayne & Peuchen, 2018). Figure 5 presents the
results for frictional undrained and partly drained
simulations in that format. The results suggest that
the higher cone factors that are sometimes obtained
at low Bq may be related to partial drainage. For
fully undrained cases, it is also noteworthy that the
cone factors for the less brittle material H are only
slightly below the value (Nkt = 12) that empirical
studies assign to clays of moderate to low sensitivity
(Low et al. 2010). As material brittleness is Figure 6. Simulation results and relations from literature
increased the simulated Nkt cone factors reduce. The between cone factor Nkt and pore pressure parameter Bq.
reduction observed is steeper than what data from
general databases, such as that of Mayne & Peuchen
(2018) would suggest. This difference is likely to be 4.2 Case-history based correlations
the result of increased sampling damage for the
more brittle soils, something also borne out by the Olson & Stark (2003) back-analyzed case history
steeper decrease that was observed by Lunne et al data of flow liquefaction failures to obtain estimates
of stress normalized undrained peak (yield) strength
(1997), using a more restricted dataset of higher
sampling quality. which were then correlated with the results of CPT.
As noted by Karlsrud et al (2005) a relation This correlation reads
between Nkt and Bq mediated by NΔu is just implied
by their definition. According to the results just pre
sented the undrained CASM simulations should then
be fitted by
594
5.2 Case-history based correlations
Case-history back-analysis by Olson & Stark (2002) in
which post-liquefaction geometry was considered led
to another correlation of residual undrained strength
with corrected tip resistance qc1. Again, the correlation
where the tip and corrected tip resistance qc1 are obtained has very little sensitivity to qc1 in the range
expressed in MPa. The simulated CPT results lie in covered by our simulations and the predicted stress nor
the range 0.14 to 1.14 qc1 for which the correlation malized residual (liquefied) undrained strength is prac
predicted upper bound of normalized peak strength tically constant. This is interesting, as the stress
is 0.21, only slightly below the input value. normalized residual strength is controlled by CASM
material parameters and also constant for a given
material type. The simulation results are above the
5 RESIDUAL UNDRAINED SHEAR recommended limits, but, for the most brittle material
STRENGTH A, well within the values supporting the correlation
(Figure 8). This comparison is unaffected by partial
5.1 Sleeve friction drainage or sleeve friction values. It is also worth men
Empirical observations of the close similitude of tioning that although material A is brittle, CASM has
sleeve friction to residual undrained strength in clay been fitted to represent even more brittleness when
as measured by vane tests or laboratory test on back-analyzing liquefaction failures (Arroyo & Gens,
remolded soil are frequent (Lunne et al. 1997; 2021; Mánica et al. 2021). Despite that Olson & Stark
Robertson, 2010). This can be simply expressed as (2002) upper bound might still be somewhat
conservative.
Figure 7 reports the normalized friction sleeve resist

ance of PFEM simulations in terms of the normalized
residual undrained shear strength predicted by CASM;
results are plotted for undrained and partly drained con
ditions. For high interface friction angles and undrained
conditions, Equation (7) holds, and the friction sleeve
is approximately equal to the residual undrained shear
strength. For lower interface friction angles the tangen
tial stress acting on the friction sleeve is lower than the
residual undrained shear strength and Equation (7)
becomes highly conservative. The importance of sleeve
roughness for this correlation is thus clearly demon
strated. Interestingly, the effect of partial drainage is
small for the more brittle material A, but large for the
less brittle H.
Figure 8. Comparison of the numerical results with case-

history based correlation of Olson & Stark (2002).
Robertson (2010, 2021) also proposed correlations

between residual undrained shear strength and cone
metrics based on case history data. For materials with
soil behavior type (SBT) index, Ic > 3 it is recom
mended to use the relation -discussed abovewith sleeve
friction. For materials with SBT Ic < 3 a different rela
tion was proposed. The soil behavior type index, Ic is
defined as
Figure 7. Residual undrained strength and side friction. All

results correspond to practically undrained tests.
595
The relation with undrained strength is based on (2002) proposals are more robust than Robertson
Qtn;cs a corrected, “clean sand equivalent”, value of (2010, 2021), as these are more sensitive to the possi
normalized cone resistance Qtn . This correction is bility of partial drainage and -for fine grained soilsto
given by: the state of the cone-soil interface. Still, these results
are based on a limited set of simulations and would
require confirmation from more systematic parametric
analyses.
The correction factor Kc is itself also dependent

(trough fifth-order polynomials) on Ic . ACKNOWLEDGMENTS
Robertson relations (2010, 2021) are plotted in
The authors acknowledge financial support from the
Figure 9 alongside the results for the fully drained
Spanish Ministry of Economy and Competitiveness,
simulations, which were the only ones resulting on Ic
through the “Severo Ochoa Programme for Centres
< 3. The relations are postulated as lower bounds and
of Excellence in R&D” (CEX2018-000797-S).
the simulation results plot above them, except for
a datapoint that represents high mobilized interface
friction for soil H, which is almost non-brittle. For the REFERENCES
higher brittleness case A, which is likely more similar
to the soils in the database, the degree of conservatism Arroyo, M. & Gens, A. 2021. Computational analyses of Dam
implied by the relation does not appear excessive. I failure at the Corrego de Feijao mine in Brumadinho,
Final Report; available at http://www.mpf.mp.br/mg/sala
de-imprensa/docs/2021/relatorio-final-cinme-upc-1
Been, K. 2016. Characterizing mine tailings for geotech
nical design. Australian Geomechanics, 51(4), 59–78.
Gens, A. 2019. Hydraulic fills with special focus on
liquefaction. Proceedings of the XVIII European Confer
ence on Soil Mechanics and Geotechnical Engineering.
Eid, H. T., Amarasinghe, R. S., Rabie, K. H. &
Wijewickreme, D. 2014. Residual shear strength of fine-
grained soils and soil–solid interfaces at low effective
normal stresses. Canadian Geotechnical Journal, 52
(2),198 –210.
EN ISO 22476-1. 2012. Geotechnical investigation and
testing. Field testing. Electrical cone and piezocone
penetration test
Karlsrud, K., Lunne, T., Kort, D. A. & Strandvik, S. 2005.
CPTU correlations for clays. In Proceedings of the 16th
international conference on soil mechanics and geotech
nical engineering (pp. 693–702). IOS Press.
Lunne, T., Robertson, P. K. & Powell, J. J. 1997. Cone
Figure 9. Comparison of numerical results with Roberston penetration testing in geotechnical practice. CRC Press.
(2010, 2021) correlations for normalized residual undrained Low, H. E., Lunne, T., Andersen, K. H., Sjursen, M. A.,
strength. The marker color indicates undrained brittleness; Li, X., & Randolph, M. F. 2010. Estimation of intact
marker shape indicates side friction values; marker hue indi and remoulded undrained shear strengths from penetra
cates permeability. tion tests in soft clays. Géotechnique, 60(11), 843–859.
Mánica, M. A., Arroyo, M., Gens, A. & Monforte, L. 2021.
Application of a critical state model to the Merriespruit
tailings dam failure. Proceedings of the Institution of
6 CONCLUSIONS Civil Engineers - Geotechnical Engineering, 1–15.
Mayne, P. W. & Peuchen, J. 2018. Evaluation of CPTU Nkt
The ability to simulate CPTu response using realistic cone factor for undrained strength of clays. In Cone
geometry and constitutive models is still relatively Penetration Testing 2018 (pp. 423–429). CRC Press.
recent. If exploited systematically it will open a way Monforte, L., Arroyo, M., Carbonell, J.M. & Gens, A. 2017.
to reduce the relatively large uncertainty that hinders Numerical simulation of undrained insertion problems in
empirically based correlations, by identifying which geotechnical engineering with the particle finite element
method (PFEM). Computers and Geotechnics.
parasitic factors may be playing a role in some data 82:144–156.
sets and/or suggesting ways to adapt general correl Monforte, L., Arroyo, M., Carbonell, J.M. & Gens, A.
ations to the local soil characteristics. 2018. Coupled effective stress analysis of insertion
This has been illustrated here for the case of soils problems in geotechnics with the particle finite element
with a brittle undrained response. The results indicate method. Computers and Geotechnics. 101:114–129.
that partial drainage is a major factor when selecting Monforte, L., Gens, A., Arroyo, M., Mánica, M. &
cone factors for su_pk. They also suggest that, when Carbonell, J.M. 2021. Analysis of cone penetration in
aiming for undrained residual strength, Olson & Stark brittle liquefiable soils. Computers and Geotechnics.
134:104123.
596
Olson, S. M. & Stark, T. D. 2002. Liquefied strength ratio Robertson, P. K. 2016. Cone penetration test (CPT)-based
from liquefaction flow failure case histories. Canadian soil behaviour type (SBT) classification system—an
Geotechnical Journal, 39(3), 629–647. update. Canadian Geotechnical Journal, 53(12),
Olson, S. M. & Stark, T. D. 2003. Yield strength ratio and 1910–1927.
liquefaction analysis of slopes and embankments. Jour Robertson, P.K. 2021. Evaluation of flow liquefaction and
nal of Geotechnical and Geoenvironmental Engineering, liquefied strength using cone penetration test: an update.
129(8), 727–737. Canadian Geotechnical Journal. In press.
Paniagua, P., D’Ignazio, M., L’Heureux, J.-S., Lunne, T., Schnaid, F. 2021. The Ninth James K. Mitchell Lecture: On
Karlsrud, K. 2019. CPTU correlations for Norwegian the Geomechanics and Geocharacterization of Tailings.
clays: an update. AIMS Geosciences 2019, 5(2): 82–103 Int. Conf. On Site Charact., Budapest.
Robertson, P. K. 1991. Soil classification using the cone Subba Rao, K. S., Allam, M. M., & Robinson, R. G. 2000.
penetration test: Reply. Canadian geotechnical journal, Drained shear strength of fine-grained soil–solid surface
28(1), 176–178. interfaces. Proceedings of the Institution of Civil Engin
Robertson, P.K. 2010. Evaluation of flow liquefaction and eers-Geotechnical Engineering, 143(2), 75–81.
liquefied strength using cone penetration test. Journal of Yu, H.S. 1998. CASM: a unified state parameter model for
Geotechnical and Geoenvironmental Engineering.136 clay and sand. International Journal for Numerical and
(6): 842–853. Analytical methods in Geomechanics. 22: 621–653.
597
© 2022 the Author(s), ISBN 978-1-032-31259-0
Assessment of deltaic soil behavior classification using AUT: GMD

database regarding CPTu records
Mohammad H. Naghibi, Abolfazl Eslami & Sara Heidarie Golafzani
Department of Civil and Environmental Engineering, Amirkabir University of Technology (AUT), Tehran, Iran
ABSTRACT: Cone penetration tests (CPT and CPTu) are unique geotechnical in-situ tools for soil behavior
classification (SBC). Identification and classification of subsurface marine layers help geotechnical engineers
analyze, design, and monitor. Thus, a significant database of 398 cases was collected from 58 sites in 18 coun
tries with CPTu soundings, and soil profiling obtained by direct boring and laboratory testing, namely AUT:
CPTu&GMD (Amirkabir University of Technology): Geo-Marine CPTu Database. Furthermore, diverse essential
parameters such as normalized cone resistance qt, friction ratio Rf, soil behavior type index ISBT are available at
0.1 m intervals. CPTu-based classification methods were subdivided into three generations, and several methods
for marine deposits were examined utilizing 57 case studies from the AUT: CPTu&GMD. The results indicated
that directly incorporating all three parameters of qc, fs, and u2 is recognized to be more accurate than utilizing
mathematical relationships, which reduces uncertainties and increases accuracy in identifying subsurface layers.
1 INTRODUCTION environments. NDT tests lack the necessary preci

sion to determine soil profiles (Eslami et al., 2019,
Projects constructed in marine environments such as Randolph and Gourvenec, 2017).
nearshore, offshore and onshore are mainly accompan Furthermore, the development of geotechnical
ied by particular foundations suitable for the super engineering has led to widespread in-situ testing by
structure regarding the available site soil deposits. standardized penetrometers. The cone and piezocone
Furthermore, characterizing the subsurface layers penetration tests (CPT and CPTu) are some of the most
in such environments is accompanied by limitations remarkable geotechnical in-situ tools for soil behavior
and difficulties heightening the effect of the embedded classification (SBC) due to rapid application, accuracy,
uncertainties in geotechnical designs. Accordingly, the providing continuous records in-depth, and the possi
projects in these areas are highly important, and care bility of implementation in marine areas. Due to inher
should be taken in their designs. Otherwise, irrevers ent soil uncertainty, measurement, and modeling errors,
ible damages will happen, leading to a decrease in the classifying soils in marine environments by the CPTu
infrastructure’s stability, limited or prohibited utiliza test becomes more important. Thus, the application of
tion, and an increase in the projects’ preservation CPT and CPTu databases can significantly improve the
costs due to the ambiguities in subsoil depiction. validation and evaluation process and increase the reli
Therefore, identifying and classifying the subsurface ability in geotechnical engineering (Eslami et al., 2017,
marine layers is crucial and assists geotechnical engin Eslami & Fellenius, 2004, Robertson, 2010). Accord
eers in analyzing, designing, monitoring, and main ingly, a significant database of 398 cases was collected
taining civil projects. Geotechnical studies are the first from 58 sites in 18 countries with adjacent CPTu
and most critical step in construction because they soundings, namely Amirkabir University of Technol
enable the geotechnical engineer to make appropriate ogy: CPTu and Geo-Marine Database, or AUT:
decisions based on this information while minimizing CPTu&GM database. This database consists of off
existing uncertainties (Eslami et al., 2019). shore, onshore, nearshore, and riverine areas, with the
Identification of subsurface layers is frequen most recorded cases being from Europe, the United
tly accomplished through laboratory, nondestructive States of America, and China. Soil profiling was
(NDT), and in-situ tests. Laboratory tests have limi obtained via direct boring and laboratory testing in this
tations, including intact sample preparation, sample database. Moreover, for all cases, several parameters
size, sample transport and storage, accurate site such as normalized cone resistance Qt, friction ratio Rf,
stress modeling for laboratory conditions and infor soil behavior type index ISBT are available at
mation discontinuity, and in-depth measurements, 0.1 m intervals.
and their effects will be more crucial in marine
DOI: 10.1201/9781003308829-86
598
Researchers have presented numerous diagrams mentioned parameters without mathematical equa
to identify and classify the behavior of subsurface tions to identify soil, is a third-generation
layers using CPT and CPTu test results. Considering (advanced-generation) diagram since using no
that each diagram is developed based on different mathematical relationships reduces the uncertain
databases, the accuracy of each diagram changes. As ties. The use of mathematical relationships is the
a result, this research attempted to measure the main difference between the second-generation and
accuracy of several most common diagrams. The the third-generation diagrams. This study aims to
diagrams are compared proportionally to the type of assess the performance of seven common SBC dia
soil, based on 57 case studies of AUT: CPTu&GMD, grams belonging to the three mentioned gener
which will lead to the presentation of a model, ations, including 1-Douglas and Olsen (1981), 2
allowing geotechnical engineers to select diagrams Campanella et al. (1985), 3-Robertson (1990)
appropriate to the type of soil available the site to (Qt-Bq), Robertson (2010), 4-Jefferies and Davies
identify the behavior of subsurface layers. (1993), 5-Eslami & Fellenius (1997), 6-Eslami
et al. (2015) 7-Eslami (2019), using 57 cases from
the AUT: CPTu&GM Database. The investigated
2 AMIRKABIR UNIVERSITY OF cases are from 16 countries worldwide as follows:
TECHNOLOGY: GEO-MARINE-CPTU USA: 10 logs, China: 5 logs, Canada: 3 logs, New
DATABASE; AUT: CPTU&GMD Zealand: 3 logs, Italy: 3 logs, Norway: 3 logs,
Poland: 3 logs, Finland: 3 logs, Sweden: 3 logs,
The compiled marine database comprises CPTu test Ireland: 3 logs, Brazil: 3 logs, England: 3 logs,
records and soil profiles from boreholes drilled Japan: 3 logs, Portugal: 3 logs, Australia: 3 logs,
within their vicinity. The records include 398 cases Turkey: 3 logs.
from 58 sites and 18 countries, mainly in the United Additionally, all four types of sites in the database
States, Europe, China, and New Zealand. The gath AUT: CPTu&GMD have been evaluated, including 15
ered sites in the database belong to onshore sites Onshore, 20 Nearshore, 4 Offshore, and 18 Riverine
near the shoreline about 21%, nearshore sites located cases.
in waters less than 30 meters deep around 26%, off Fifty-seven mentioned cases include 1420
shore sites located in waters deeper than 30 meters meters of CPTu soundings and soil descriptions
nearly 15%, and riverine sites roughly 38%. This from drilling boreholes. Their frequency is such
database contains approximately 10000 meters of that it consists of 117 meters of sensitive soil, 98
soil profiles and covers a wide range of marine meters of clay, 1135 meters of mixed or deltaic
deposits, including clay, sensitive, sand-gravel, over- soil, 54 meters of sand and gravel, and 16 meters
consolidated clay, and mixed or deltaic soils. of over-consolidated soil. Figure 2 demonstrates
In addition to the CPTu records digitized each the distribution of investigated deltaic soils in
10 cm in-depth and the soil profiles from boreholes, different SBC charts.
other parameters used via soil behavior classification The accuracy of each diagram in predicting each
(SBC) diagrams are also determined using available soil type forms the evaluated areas, i.e., clay, sensi
correlations. Figure 1 shows the details of the AUT: tive, deltaic (mixed), overconsolidated, and sand-
CPTu&GM database in a flowchart. gravel, and has been calculated in percentage accord
ing to Equation 1.
3 ANALYSES AND ASSESSMENT
In this study, soil behavioral classification diagrams

are classified into three generations based on the
results of the CPTu test. First-generation (basic Where SR is the success rate; Lʹ is the total length of
generation) diagrams, such as diagrams of Douglas correctly identified soil profiles regarding a specific
and Olsen (1981), Campanella et al. (1985) and, soil type; L is the total length of investigated profiles
Robertson (2010), use only two parameters out of concerning a particular soil type.
the three parameters obtained from the piezocone According to Figure 3, each diagram’s statistical
test, along with mathematical equations, for soil evaluation results are displayed separately on a radar
classification. The cone tip resistance qc and sleeve chart. The radar diagram contains five axes regarding
friction fs are the two parameters usually used. the considered soil types. Axes are calibrated from
Second-generation (middle-generation) diagrams, 0% to 100%. These values indicate the desired dia
such as diagrams of Robertson (1990), Jefferies gram’s accuracy in identifying the soil associated
and Davies (1993), Eslami & Fellenius (1997), and with each axis.
Eslami (2019), apply all three parameters of cone As shown in Figure 3, identifying sensitive
tip resistance, sleeve friction, and excess pore soils has improved with the advent of second-
water pressure, as well as mathematical equations, generation diagrams that use all three parameters
to identify subsurface layers. Moreover, Eslami qc, fs, and u2 and mathematical equations for soil
et al. (2015) diagram, which directly uses the three classification.
599
Figure 1. The details of AUT: CPTu&GMD database.
600
Figure 2. Deltaic soils of AUT: CPTu&GMD on different charts.
601
Figure 3. The performance assessment of each SBC diagram for the investigated database.
602
The Douglas and Olsen (1981) and Campanella The investigated diagrams were 1-Douglas and
et al. (1985) diagrams, i.e., the first-generation of the Olsen (1981), 2-Campanella et al. (1985), 3-Robert
charts, showed a low accuracy in identifying sensi son (1990) (Qt - Bq), Robertson (2010), 4-Jefferies
tive soils. However, this accuracy has reached more and Davies (1993), 5-Eslami & Fellenius (1997), 6
than 80 percent by applying second-generation dia Eslami et al. (2015), 7-Eslami (2019). Five soil
grams. Although this trend continued until the types, i.e., sensitive soils, clay, mixed or deltaic
Eslami & Fellenius (1997) diagram showed more soils, overconsolidated clay, and sand-gravel, were
than 90% success in identifying sensitive soils, the considered to assess the performance of SBC dia
Robertson (2010) chart has shown a low accuracy in grams in predicting the marine soil types.
identifying sensitive layers due to using a pattern Results indicated that the accuracy of the dia
similar to the first-generation diagrams and not using grams changes with the variation of soil type.
the excess pore water pressure parameter. That Also, assessed diagrams are adequately accurate
shows the essential role of the excess pore water in identifying deltaic soils, especially diagrams
pressure parameter in identifying sensitive soil. The that use all three parameters of qc, fs and, u2 for
Eslami & Fellenius (1997), Eslami et al. (2015) and, soil classification. Eventually, Eslami et al.
Eslami (2019) diagrams were able to identify sensi (2015) chart directly implements all three CPTu
tive soils with more than 90% accuracy. records, i.e., qc, fs, and u2, which reduces uncer
Due to the low cone tip resistance of silts and tainties in soil behavioral classification instead of
normally consolidated clays, it seems this param applying mathematical correlations and was more
eter plays a minor role in distinguishing clay suitable for classifying deltaic soils for the inves
from silt. The diagrams now mainly use sleeve tigated database.
friction and excess pore water pressure param
eters for this purpose. Therefore, it is more chal
lenging to identify clay than the other four soil REFERENCES
types for diagrams. However, the maximum
accuracy is about 60%, which belongs to Jefferies Douglas, B. Soil classificaion using electric cone
and Davies (1993), Eslami & Fellenius (1997), penetrometer. Symp. on Cone Penetration Testing and
and Eslami (2019) charts. Except for Eslami Experience, Geotech. Engrg. Div., 1981. ASCE,
209–227.
(2019) diagram, other diagrams of the first
Eslami, A., Alimirzaei, M., Aflaki, E. & Molaabasi, H.
and second generations identify overconsolidated 2017. Deltaic soil behavior classification using CPTu
clay with less than 60% accuracy. The Eslami records—Proposed approach and applied to fifty-four
(2019) diagram has characterized overconsoli case histories. Marine Georesources & Geotechnology,
dated clay with about 77% accuracy. Although 35, 62–79.
the highest accuracy is for the third-generation Eslami, A. & Fellenius, B. H. 1997. Pile capacity by direct
chart of Eslami et al. (2015), it has identified CPT and CPTu methods applied to 102 case histories.
a success rate of 85% overconsolidated clay. Canadian Geotechnical Journal, 34, 886–904.
Douglas and Olsen (1981), Campanella et al. Eslami, A., Moshfeghi, S., Molaabasi, H. & Eslami, M. M.
2019. Piezocone and Cone Penetration Test (CPTu and
(1985), Robertson (1990) (Qt - Bq), Robertson
CPT) Applications in Foundation Engineering, Butter
(2010), and Jefferies and Davies (1993) diagrams worth-Heinemann.
identify mixed or deltaic soils with an accuracy Heidarie Golafzani, S., Eslami, A. & Jamshidi Chenari, R.
of 80 to 90 percent. While the Eslami et al. 2020. Probabilistic Assessment of Model Uncertainty
(2015) diagram detects these soils with less than for Prediction of Pile Foundation Bearing Capacity;
20% success, Eslami & Fellenius (1997) and Static Analysis, SPT and CPT-Based Methods. Geotech
Eslami (2019) diagrams have an accuracy of over nical and Geological Engineering, 38, 5023–5041.
90% for this purpose. Firstand second-generation Jefferies, M. & Davies, M. 1991. Soil classification by the
diagrams identify layers of sand gravel with 80 to cone penetration test: Discussion. Canadian Geotech
nical Journal, 28, 173–176.
90 percent accuracy, and only the third-generation
Randolph, M. & Gourvenec, S. 2017. Offshore geotech
chart does so with over 90% success. nical engineering, CRC press.
Robertson, P. K. 1990. Soil classification using the cone
penetration test. Canadian geotechnical journal, 27,
4 CONCLUSIONS 151–158.
Robertson, P. K. Soil behaviour type from the CPT: an
Marine deposits are mainly problematic soils due to update. 2nd International symposium on cone penetra
their formation process, and their characterization tion testing, 2010. Cone Penetration Testing Organizing
increases the reliability of geotechnical designs in Committee, 575–583.
Robertson, P. K. & Cabal, K. Estimating soil unit weight
onshore and offshore structures. Accordingly, the
from CPT. 2nd International Symposium on Cone Pene
AUT: CPTu&GMD was compiled, and 57 cases tration Testing, 2010. 2–40.
were selected, including offshore, onshore, and river Robertson, P. K., Campanella, R. G., Gillespie, D. &
ine sites, to assess the performance of SBC diagrams Greig, J. Use of piezometer cone data. Use of in situ
classified as first, second, and third generations tests in geotechnical engineering, 1986. ASCE,
according to their simplifications and assumptions. 1263–1280.
603
© 2022 the Author(s), ISBN 978-1-032-31259-0
A review of methods for estimating undrained brittleness index from the CPT
Y. Narainsamy & S.W. Jacobsz
University of Pretoria, Pretoria, South Africa
ABSTRACT: The Cone Penetration Test (CPT) is a useful tool for soil profiling due to its near continuous data
measurements, low cost and repeatability. When conducting designs or safety evaluations of slopes, it is often of
interest to understand the undrained response of the soil at large strains. This behaviour can be defined in terms of
the Undrained Brittleness Index (IB) which relates the yield undrained shear strength to the steady state undrained
shear strength. Over time, a number of field case histories where strain softening during undrained shear was
deemed to have occurred have been assessed, and a relationship between corrected tip resistance and IB was pro
posed (Sadrekarimi, 2014). A quick clay test site and a silt test site in Norway included in a recently published
open access geotechnical database were assessed. It was found that the method proposed by Sadrekarimi (2014)
underestimated the IB for both the quick clay and silt test sites. It was also noted that there are some limitations
with a popular screening method for identifying soils susceptible to strength loss during undrained shear.
1 INTRODUCTION soils at two test sites in Norway: a quick clay test

site and a silt test site.
1.1 Cone Penetration Testing
The Cone Penetration Test (CPT) is a useful tool for 1.2 Undrained Brittleness Index (IB)
soil profiling. It provides near continuous data meas
The Undrained Brittleness Index (IB) was initially
urements, repeatable results, is cost effective and
proposed by Bishop (1971), and is used to define
tests can be conducted relatively quickly. When
a relationship between the yield and steady state
designing and conducting safety evaluations of slopes
undrained shear strength of a soil and is shown in
and embankments, it is often useful to understand the
Equation 1.
undrained response of the soil at large strains. This is
important as soils which exhibit strain softening (con
tractive) behaviour will have a lower strength than
soils which exhibit strain hardening (dilative) behav
iour. This state is sometimes represented using the
state parameter ψ as suggested by Been & Jefferies
(1985). The state parameter is the difference in void where su (yield) is the yield undrained shear strength
ratio between the current void ratio and the void ratio and su (steady state) is the steady state undrained
at steady state, at the same effective confining stress. shear strength. These parameters are shown visually
However, the state of the soil simply describes the in Figure 2. IB ranges from 0 to 1, with IB = 1 indi
tendency for contraction or dilation at large strains. cating complete loss of shear strength, i.e. a highly
For many problems it is of more value to know the contractive response, while IB = 0 indicates no loss
extent of the expected contraction, if any. Soils of strength with strain. Note that the undrained
which exhibit a rapid reduction in undrained shear brittleness index should not be confused with the
strength are referred to as brittle soils and an modified soil behaviour type index IB which is used
example of such behaviour can be seen in Figure 2. by some researchers (e.g. Robertson, 2016).
Brittle failure mechanisms are of particular concern
as failure can initiate rapidly as was seen in the 2019 2 REVIEW OF EXISTING RELATIONSHIPS
Feijão tailings dam failure (Robertson et al., 2019).
One method of quantifying the brittleness of a soil is
2.1 CPT correlations
to use the Undrained Brittleness Index. The objective
of this paper is to investigate the performance of the Due to the large deformations and stresses induced
CPT-based method proposed by Sadrekarimi (2014) during CPT probing, it is common to pursue empir
on predicting the Undrained Brittleness Index for ical relationships between geotechnical parameters
DOI: 10.1201/9781003308829-87
604
and measured responses from the CPT. In one such 3 ANALYSIS
study, a large database of 600 laboratory shear tests
were reviewed. An attempt was then made to relate 3.1 NGTS test sites
the yield and steady state undrained shear strengths
to the corrected tip resistance (qc1), based on field In 2017 the Norwegian Geotechnical Institute (NGI)
case histories where undrained shearing was established the Norwegian GeoTest Sites (NGTS).
believed to have occurred (Sadrekarimi, 2014). As The objective of the NGTS was to develop test sites
part of this study, a relationship between IB and qc1 that could be used for testing and verifying innova
was also identified. tive soil investigation methods, as well as facilitate
When analysing the field case histories, the research activities relating to soil behaviour and
yield and undrained shear strengths were deter foundation design (L’Heureux et al., 2017). The
mined based on results of static limit equilibrium NGTS comprise five test sites in different soils, all
back analyses (Olson, 2001; Muhammad, 2012). located in Norway as shown in Figure 1.
The characteristic yield undrained shear strength A unique aspect of the NGTS is that the field and
was determined by varying the shear strength in laboratory test data have been made publicly avail
the soil zones deemed susceptible to strength loss able through the Datamap web based application
during undrained shear until a factor of safety (Doherty et al., 2018). In addition to the NGTS data,
against failure of 1.0 was calculated using Spen data from the Australian National Field Testing
cer’s method (Spencer, 1967). Drained shear Facility (NFTF) is also included on the Datamap
strengths were assumed to be mobilized in the application. Although six test sites were included in
zones above the phreatic surface (i.e. material the combined NGTS and NFTF database, only two
zones deemed not to be susceptible to strength contained sufficient information for use in this study.
loss during undrained shear). The liquefied or These two sites are discussed in further detail below.
steady state undrained shear strength was either
determined using a simplified force diagram
approach with kinematic considerations or
a rigorous limit equilibrium approach considering
the final failure geometry, depending on the avail
able information. The corrected tip resistance
measured in the field was then related to IB in
the form of a screening method as shown in
Equation 2.
where qc1 is the corrected cone tip resistance as

determined using the equations described by
Kayen et al. (1992). Upon review of the field
case history database used to develop the rela
tionship, it becomes clear that a biased dataset
was used (i.e. the dataset only includes data from Figure 1. Location of the NGTS test sites.
sites where strain softening was deemed to have
occurred and therefore does not include material
that strain hardens). 3.2 The Tiller-Flotten test site
The Tiller-Flotten geotechnical test site is located in
2.2 Open access geotechnical databases Trondheim in Norway. The site consists of a 50 m thick
With the rapid advancement of technology and marine deposit of sensitive clay (L’Heureux et al.,
the ease of global communication, there has been 2019). Due to the glacial history of the area, the clays
a recent development to populate and maintain are overconsolidated, with an OCR between 1.5 and
open access geotechnical databases. Two such 3.0. A rigorous CPTu field testing regime was con
databases are: the Premstellar Geotechnik and ducted and Anisotropically Consolidated Undrained
Norwegian GeoTest Site databases. The former Compression (CAUC) triaxial tests were conducted
includes hundreds of CPT tests results and asso on high quality Sherbrooke block samples obtained
ciated indicator tests (Oberhollenzer et al., 2021), from boreholes adjacent to the CPTu test locations.
and the latter is described in more detail below. The water table is located between 1 and 2 m below
It is expected that more open access geotechnical ground level and the pressure distribution is approxi
databases will be developed and maintained in mately hydrostatic to 5.5 m whereafter the pressure
the future. build-up is sub-hydrostatic.
605
The clays have a bulk unit weight of 18 kN/m3 The silts have a bulk unit weight of 19 kN/m3 and
and are split into two sub-profiles as shown in are split into two sub-profiles as shown in Figure 5:
Figure 4. Unit IIA extends from 2 m below surface Unit II which extends from 5 m below surface to
to 7.5 m below surface and comprises clay of 12 m below surface; and Unit III which extends from
medium sensitivity and Unit IIB extends from 12 m below surface to 16 m below surface. Units II
depths greater than 7.5 m and comprises clay of and III are regarded as the same material with the
extreme sensitivity. Results from the CAUC triaxial same geologic origin and were separated simply
tests, shown in Figure 2, indicate brittle behaviour based on the results from the indicator tests which
which is expected of a sensitive clay. indicated that the silt becomes sandier in the lower
Unit III. Results from the CAUC triaxial tests are
shown in Figure 3. With the exception of the sample
obtained at 9 m depth from borehole 1 (B01_9m)
which showed strain softening behaviour, all the
other samples showed strain hardening behaviour.
Figure 2. Triaxial test results on the Sherbrooke samples

from the Tiller-Flotten test site.
3.3 The Halden test site

The Halden geotechnical test site is located approxi Figure 3. Triaxial test results on the Sherbrooke samples
mately 120 km south of Oslo in Norway. The site for the Halden test site.
consists of a 10 to 12 m thick deposit of fjord-marine,
low plasticity clayey silt (Blaker et al., 2019). The
3.4 Analysis of CPT data
silts are normally consolidated and the water table is
located 2 m below surface with a hydrostatic profile The raw CPT data obtained from the database were
in the silt layer, changing to sub-hydrostatic in the processed using version 3.6.2.6 of the CPeT-IT soft
underlying clays. As with the Tiller-Flotten site, ware package developed by Geologismiki (Geologis
a rigorous CPTu field testing regime was conducted miki, 2021). When analysing CPT data, it is important
and CAUC triaxial tests were conducted on high qual to note that the load cells which measure the sleeve
ity Sherbrooke block samples obtained from bore friction and tip resistance are geometrically offset on
holes adjacent to the CPTu test locations. the cone. Therefore, to correct the data such that the
606
Figure 4. CPT data for selected probes at the Tiller-Flotten quick clay reference site (after L’Heureux et al., 2019).
Figure 5. CPT data for selected probes at the Halden silt reference site (after Blaker et al., 2019).
607
measured cone resistance and sleeve friction can be Table 1. Summary of data analysed.
compared at the same depth, a vertical shift is
applied. This shift distance is a function of the Test site Tiller-Flotten Halden
physical location of the sleeve friction and load cell
as well as the rigidity of the soil. The Cross Correl Material Marine clay Clayey silt
ation Function (CCF), as proposed by Jaksa et al. No of triaxial tests 10 9
(2002) was used for this assessment. The inter Type of test CAUC CAUC
preted CPT data for the Tiller-Flotten and Halden Undrained response Strain softening Strain hardening
test sites are shown in Figure 4 and Figure 5, No of CPT soundings 5 5
respectively. Based on the soil behaviour index Ic Susceptible to strength Yes Yes
as proposed by Robertson & Wride (1998) and the loss at large strains?*
pore pressure ratio Bq, it was deemed that the prob Average qc1 (MPa) 0.554 0.633
ing through the Tiller-Flotten clay and Halden silt Average IB 0.713 0.050
was predominantly undrained. It is therefore appro
* according to the Plewes et al. (1992) screening method
priate to assess the CPT data in terms of undrained
hydraulic conditions.
3.5 Susceptibility to strength loss during undrained 4 RESULTS

shear
The IB determined from the CAUC triaxial test results,
Several methods have been proposed to estimate the and the average qc1 determined from the CPT tests for
state of the soil, specifically with regard to the the Tiller-Flotten clay and Halden silt are shown in
undrained behaviour at large strains. One such Figure 6. Also plotted on Figure 6 are the case histor
method is a screening method proposed by Plewes ies investigated by Olson (2001) and Muhammad
et al. (1992) where the state parameter (ψ) of the soil (2012), as well as the relationship proposed by Sadre
is estimated. The method is referred to as karimi (2014). It was found that for the Tiller-Flotten
a screening method as the state parameter is esti clay, the proposed relationship provided a reasonable
mated based only on CPT data and some limitations estimate of the IB. However, the Halden silt plots
have been identified (e.g. Torres-Cruz, 2021; Narain much lower. This is due to the fact that the proposed
samy et al., 2022). For this assessment, the original relationship is simply a function of the corrected tip
implementation of the Plewes method was used. resistance and not any laboratory test data. Since the
Material with ψ > -0.05 was assumed to be suscep Halden silt exhibited strain hardening behaviour
tible to strength loss at large strains during undrained during undrained shear in triaxial compression, the
shear, material with ψ < -0.08 was considered not material shows no brittleness.
susceptible to strength loss at large strains.
A transition zone was defined where -0.08 < ψ <
-0.05. These limits are commonly used in research
and practice (e.g. Jefferies & Been, 2015).
Applying the screening method to the NGTS site
results in the ψ with depth charts shown in Figure 4
and Figure 5. Figure 4 clearly shows the Tiller-
Flotten clay as susceptible to strength loss at large
strains and this correlates well with the observed
response from the triaxial data (Figure 2). Figure 5
also shows the Halden silt as susceptible to strength
loss at large strains. However this does not correlate
well with the observed response from the triaxial
data (Figure 3). This emphasises the screening
nature of the Plewes method and that more
advanced methods may need to be considered in
practice.
3.6 Summary of test data analysed Figure 6. Results from this study compared to previously
published data.
A summary of the test data assessed is shown in
Table 1. For the Tiller-Flotten clay site, a total of 10
triaxial tests and 5 CPT tests were assessed. An aver This over-estimation of the IB for a strain-
age qc1 of 0.55 MPa and IB of 0.71 was determined. hardening material is expected as the empirical rela
For the Halden silt site, a total of 9 triaxial tests and tionship was derived based on field case histories
5 CPT tests were assessed, and an average qc1 of where only strain softening material was involved
0.63 MPa and IB of 0.05 was determined. (i.e. it was a biased database). Care therefore needs
608
to be taken when using the Plewes screening field testing facility in Ballina, Australia. Computers
method, as well as when using the relationship pro and Geotechnics, 93, 3–8.
posed by Sadrekarimi (2014), ensuring that they are Geologismiki. 2021. CPET-IT. Retrieved from http://www.
applied to materials similar to those from which they geologismiki.gr/Products/CPeT-IT.html
were derived and validated against. Jaksa, M. B., Kaggwa, W. S., & Brooker, P. I. 2002. An
improved statistically based technique for evaluating the
CPT friction ratio. Geotechnical Testing Journal, 25(1),
61–69.
5 CONCLUSIONS Jefferies, M., & Been, K. 2015. Soil liquefaction: a critical
state approach, 2nd Ed. CRC press.
CPT and triaxial data from two NGTS research test Kayen, R. E., Mitchell, J. K., Seed, R. B., Lodge, A.,
sites were assessed: the Tiller-Flotten quick clay site Nishio, S., Coutinho, R., & others. 1992. Evaluation of
and the Halden silt site. A strain softening screening SPT-, CPT-, and shear wave-based methods for liquefac
assessment was performed, and the undrained brittle tion potential assessment using Loma Prieta data. Proc.,
ness index was determined for both sites. The fol 4th Japan-US Workshop on Earthquake-Resistant Des.
lowing was found: of Lifeline Fac. and Countermeasures for Soil Liquefac
tion, 1, 177–204.
1. The Plewes et al. (1992) screening method identi L’Heureux, J.-S., Lindgård, A., & Emdal, A. 2019. The
fied that the Halden silt material is contractive at Tiller–Flotten research site: Geotechnical characteriza
large strains. However, with the exception of one tion of a very sensitive clay deposit, AIMS Geosciences,
test, the results from the triaxial tests indicated dila 5(4), 831–867.
L’Heureux, J. S., Carroll, R., Lacasse, S., Lunne, T.,
tive behaviour. Care therefore needs to be taken
Strandvik, S. O., Degago, S., & others. 2017. New
when using this method to assess soils susceptible Research Benchmark Test Sites in Norway. In Geotech
to strain softening behaviour during undrained nical Frontiers 2017, 631–640.
shear. Muhammad, K. 2012. Case history-based analysis of lique
2. The relationship proposed by Sadrekarimi (2014) faction in sloping ground. PhD Thesis, University of
between corrected tip resistance (qc1) and Illinois.
undrained brittleness index (IB) was found to Narainsamy, Y., Jacobsz, S., Geldenhuys, L., &
overestimate the undrained brittleness index for Hörtkorn, F. 2022. A review of liquefaction potential
both the Halden silt and the Tiller-Flotten clay. screening in engineering practice. Proc. 20th Int. Conf.
on Soil Mechanics and Geotech. Eng. Sydney.
This method should therefore only be used as
Oberhollenzer, S., Premstaller, M., Marte, R.,
a screening method and a more rigorous assess Tschuchnigg, F., Erharter, G. H., & Marcher, T. 2021.
ment is recommended when assessing the Cone penetration test dataset Premstaller Geotechnik.
undrained brittleness index using the CPT. Data in Brief, 34, 106618.
Olson, S. M. 2001. Liquefaction analysis of level and slop
ing ground using field case histories and penetration
ACKNOWLEDGEMENTS resistance. PhD Thesis, University of Illinois.
Plewes, H. D., Davies, M. P., & Jefferies, M. G. 1992. CPT
based screening procedure for evaluating liquefaction
The authors gratefully acknowledge GeoLogismiki susceptibility. Proc. of the 45th Canadian Geotech.
for providing an educational licence for the CPeT-IT Conf., Toronto, 4, 1–9.
CPT interpretation software. Robertson, P. K., & Wride, C. E. 1998. Evaluating cyclic
liquefaction potential using the cone penetration test.
Canadian Geotechnical Journal, 35(3), 442–459.
REFERENCES Robertson, P K, de Melo, L., Williams, D. J., &
Wilson, G. W. 2019. Report on the expert panel on the
Been, K., & Jefferies, M. G. 1985. A state parameter for technical causes of the failure of the feijão dam 1.
sands. Géotechnique, 35(2), 99–112. Robertson, P. K. 2016. Cone penetration test (CPT)-based
Bishop, A. W. 1971. Shear strength parameters for undis soil behaviour type (SBT) classification system—an
turbed and remolded soil specimens. Roscoe Memorial update. Canadian Geotechnical Journal, 53(12),
Symp., 3–58. 1910–1927.
Blaker, Ø., Carroll, R., Paniagua Lopez, A. P., Sadrekarimi, A. 2014. Effect of the mode of shear on static
DeGroot, D. J., & L Heureux, J.-S. 2019. Halden liquefaction analysis. Journal of Geotechnical and
research site: geotechnical characterization of a post gla Geoenvironmental Engineering, 140(12), 4014069.
cial silt, AIMS Geosciences, 5(2), 184–234. Spencer, E. 1967. A method of analysis of the stability of
Doherty, J. P., Gourvenec, S., Gaone, F. M., Pineda, J. A., embankments assuming parallel inter-slice forces. Géo
Kelly, R., O’Loughlin, C. D., & others. 2018. A novel technique, 17(1), 11–26.
web based application for storing, managing and sharing Torres-Cruz, L. A. 2021. The Plewes Method: a Word of
geotechnical data, illustrated using the national soft soil Caution. Mining, Metallurgy & Exploration, 1–1
609
© 2022 the Author(s), ISBN 978-1-032-31259-0
A comparative study on CPTu-based soil classification methods: Case

studies
T.D. Nguyen & P.S. Khin
VNU Vietnam Japan University, Hanoi, Vietnam
Q.N. Pham
Vietnam Petroleum Institute, Hanoi, Vietnam
A.T. Vu
Le Quy Don Technical University, Hanoi, Vietnam
ABSTRACT: This paper presents a comparative study on the applicability of three CTPu-based soil classifi
cation charts, namely, Robertson’s normalized (SBTn) chart (Robertson 2009), Eslami-Fellenius (EF)’s non-
normalized chart (Eslami and Fellenius 1997) and Robertson’s non-normalized (SBT) chart (Robertson 2010).
A well-monitored database of CPTu data at six study sites of different geological conditions was used for the
analyses. It is found from the study that when the effective stress (σʹv0) is small, typically smaller than 150
kPa, the EF’s and Robertson’s SBTn charts results in similar soil types obtained from visual classification
procedure (VCP) whereas the Robertson’s SBT chart results in similar or slightly coarser soil types. When the
σʹv0 becomes larger, the Robertson’s SBTn chart still results in soil types well matched with those obtained
from the VCP but both the non-normalized charts tend to result in coarser soil types. This feature of the non-
normalized charts is more pronounced in stiff clayey or dense sandy soils.
1 INTRODUCTION Robertson (2010) and Robertson and Cabal (2015)

recommended that in general the normalized chart
The CPTu is one of the most versatile and useful (Robertson 1990, 2009) provide more reliable identi
field tests in soil investigation and one of its main fication of SPT than the non-normalized chart.
applications from the test results is soil classification. The CPTu has been an indispensable field test in
In the literature, the Robertson’s normalized soil soil investigation for nearshore and offshore wind
behaviour type (SBTn) chart (Robertson 1990, 2009) farms projects along the coast of Vietnam these
is the most popularly recommended one. Besides, days. The applicability of the charts to new geo
Eslami-Fellenius (E-F)’s method (Eslami and Felle logical conditions, especially to onshore and off
nius 1997, Fellenius 2021) is also a good alternative. shore sites, is merit to be investigated.
The key difference in input parameters for Robert This paper presents a comparative study on the
son’s SBTn and E-F’s charts is that only basic meas applicability of the Robertson’s SBTn chart (Robert
urements of the test are required for the E-F’s chart son 2009), Robertson’s non-normalized chart
whereas effective vertical stress (σʹv0) is additionally (Robertson 2010) and E-F’s chart (Elami-Fellenius
required for the Robertson’s STBn chart. In many 1997, Fellenius 2021). For this, CPTu data from four
cases, exact value of soil density and in-situ pore onshore sites in Vietnam and Korea and two offshore
pressure (u0) (and therefore the effective stress) are sites in the East Sea (Vietnam) are used for analyses.
not available, the Roberson’s SBTn chart is therefore
less effective in real-time classification (Fellenius
2021). 2 EXPEIMENTAL TEST SITES
To facilitate the real-time application to soil clas
sification, Robertson (2010) introduced an updated 2.1 Test site location
non-normalized chart (SBT) initially proposed by
Robertson et al. (1986) that requires only basic In this study, CPTu test results from six study sites of
measurements but stressed that this updated non- different geological conditions are brought into ana
normalized chat could give reasonable result in the lyses. The abbreviated names and locations of the sites
range of effective stress from 50 to 150 kPa. are given in Table 1, which are graphically illustrated
DOI: 10.1201/9781003308829-88
610
Table 1. Name and location of the test sites.
No. Site name Condition Longitude Latitude Location
1 VSIP Onshore 106° 42ʹ 8.98ʺ E 20° 54ʹ 52.21ʺ N Hai Phong Province, Vietnam
2 KC Onshore 105° 43ʹ 44ʺ E 21° 3ʹ 22.92ʺ N Hanoi, Vietnam
3 TPP Onshore 106° 12ʹ 55.54ʺ E 20° 5ʹ 55.29ʺ N Nam Dinh Province, Vietnam
4 MOC Onshore 128° 54ʹ 14.25ʺ E 35° 5ʹ 5.8ʺ N Busan City, Korea
5 PVN1 Offshore 108° 20ʹ 56.63ʺ E 10° 12ʹ 41.36ʺ N East sea, Vietnam
6 PVN2 Offshore 108° 48ʹ 09.69ʺ E 07° 55ʹ 24.40ʺ N East sea, Vietnam
in Figure 1. The first three test sites, namely, Vietnam measured using a combination of electronic sensor
Singapore Industrial Park (VSIP), Kim Chung Resi and echo sounder and the average value at the PVN1
dential Complex (KC), and Nam Dinh Thermal Power and PVN2 sites was 56.3 m and 119.9 m, respectively.
Plant (TPP), are located in the Red River delta, Viet
nam. A research program on consolidation characteris
2.2 CPTu tests
tics of clayey soils has been conducted recently by the
first author at the sites in which CPTu and elaborate For the first three test sites (VSIP, KC, TPP), the CPTu
soil sampling for laboratory tests were conducted. was carried out using a piezocone of 10 cm2 cross-
The fourth site, namely, Myeongji Ocean City sectional area with a filter mounted at the cone shoul
(MOC), is located in the Nakdong River Estuary of der (u2 position). The test was performed following
Nakdong River delta, West of Busan city, Korea. procedures recommended in the ASTM 5778 – 20
The CPTu and soil sampling at the site were con (2020) standard. The key research objective at the sites
ducted by the first author some years ago in was on consolidation characteristics of the clayey soils
a research program on bearing capacity of driven thus the CPTu was carried out in upper clayey and
pile foundation (Kim et al. 2012). silty soil layers only. The CPTu at MOC site was car
ried out using a piezocone of 15 cm2 with also the
measurement of pore pressure at the cone shoulder
(u2position).
At the offshore test sites, drilling operations
were carried out through a motion compensated
Furgo Offshore Drilling Rig over a moon pool in
the centre of the vessel. On the other hand, sam
pling was conducted through the open-centre bit
using Wison BHA and collars, which facilitated
umbilical piston, push and hammer sampling oper
ations. The CPTu was conducted using Fugro’s
Downhole Wison system with a cone of 10 cm2
base area, 60° apex, 150 cm2 cylindrical sleeve,
and a filter at cone shoulder to measure u2. The rate
of penetration during testing was kept between
20 mm to ± 5 mm per second throughout the
3.0 m continuous stroke.
Although analyses were fully carried out for the
six test sites, typical analysis results from only four
sites, namely, VSIP, MOC, PVN1, and PVN2, will be
presented in detail due to limited allowance of space.
2.3 Soil profiles

Soil layers at each site were carefully classified
Figure 1. Location of the test sites. based on results from physical tests in the lab with
the support of visual classification procedures (VCP)
The last two sites, namely, PVN1 and PVN2 are from boring records (e.g., soil sample from the SPT
located in the East Sea of Vietnam (Figure 1). The sampler) at the site. Figure 2 shows soil layers and
CPTu with soil sampling and other field tests were basic soil properties at the four sites. To facilitate the
elaborately conducted by Fugro for PetroVietnam evaluation later, soil layers in each site were simply
(PVN) in an investigation program for building drill named from L1 to Ln downwards as shown in the
rigs at the sites. The water depth at the two sites was figure.
611
Figure 2. Soil profiles at the study sites.
612
At the the VSIP site, the CPTu was carried out in 3 SOIL CLASSIFICATION RESULTS
soft to firm clay layers up to the depth of
17 m (Figure 2(a). At the MOC site, the CPTu was 3.1 Influence of effective stress
carried out up the sandy gravel layer (at 55.0 m)
(Figure 2(b)), and thus both sandy and clayey soil The influence of effective stress to soil behaviour
types are available for analyses. type has been discussed extensively in many studies
At the PVN1 site, the soil profile (up to the (e.g., Robertson 1990, 2009) and it is examined
depth of 140 m) consists of many alternate sub- through the case studies herein. It is found from the
layers of silty clay, clayey silt, sandy silt and cases that when the effectives stress is less than
silty sand; however, the soil profile might be about 150 kPa the E-F’s and Robertson’s normalized
broadly divided into four main layers as shown (SBTn) charts provide rather similar soil types for
in Figure 2(c). At the PVN2 site, the soil profile both clayey soil and sandy soil, however the Rober
(up to the depth of 30.0 m) consists of 6 layers son’s non-normalized (SBT) chart tends to indicate
of both clayey and sandy soil types as shown in similar or slightly coarser grained soil types com
Figure 2(d). pared with those from the normalized chart. As an
Figure 2 also shows the cone resistance (qc), example, Figure 3 shows the three charts applied to
sleeve friction (fs), pore water pressure (u2) and layer L4 (silty clay) at the VSIP site. As shown, both
excess pore water pressure (Δu2) diagrams, and soil E-F’s and Robertson’s SBTn charts indicate that soil
behavior type index (Ic) (Roberson 2009) from the in this layer is clay to silty clay and is well matched
CPTu at the study sites. It is interesting to note from with results from lab test results and VCP but the
qc and u2 diagrams that the soil layers can easily be Robertson’s SBT chart indicates that the soil is more
identified and are well matched with the classifica likely silt mixtures (clayey silt & silty clay). Similar
tion from physical parameters profile shown in finding is found for the upper layers at the six sites
Figure 2. Note that the Δu2 value of some sandy (i.e., the effective stress is relatively small). Note
layers was negative (e.g., 24.0 – 48.0 m at PVN1 that the parameters of the charts are as follows:
and 15.0 – 23.0 m at PVN2). This unusual character the effective cone stress qE = qt – u2, where qt =
istic is typically found in sandy deposits in offshore qc + (1-a)u2; the normalized cone resistance Qtn =
environment due to the cavitation phenomenon [(qt-σv0)/pa] (pa/σʹv0)n, the normalized friction ratio
(Lunne et al. 1997). Fr = [fs/(qt-σv0)]100%, the sleeve fiction ratio: Rf =
The soil behavior type index (Ic) also helps to 100(fs/qc).
classify the soil profiles well. For example, at the When the effective stress becomes larger, the
depths of 0.4 to 17.6 m (VSIP site), 13.8 to influence of effective stress becomes more pro
32.8 m (MOC site), 48.0 to 100.0 m (PVN1 site) nounced and especially the E-F’s chart also tends to
the Ic is pronouncedly larger than 2.6, indicating indicate slightly coarser soil type compared with that
that the layers are clayey soils. The index value of obtained from the lab test results and VCP, and from
other layers is mostly equal or less than this bound Robertson’s SPTn chart. For instant, Figure 4 shows
ary value, indicating the layers are silty to sandy a comparison of classified soil types from the charts
soils. applied to layer L4 (stiff to very stiff silty clay – silt)
Figure 3. Soil classification charts for layer 4 at VSIP site.
613
Figure 4. Soil classification charts for layer 2 at PVN1 site.
at PVN1 site. The figure indicates that the Robert typically increases linearly with the increase in
son’s SBTn chart results in similar soil types vertical effective stress (σʹv0). However, in
obtained from the lab test results and VCP whereas a homogeneous sand layer, the resistances increase
the EF’s and Robertson’s SBT charts indicate soil nonlinearly with the increase in the effective stress.
types that are likely silt to sandy silt. Besides the This is why the stress exponent (n) is applied to the
influence of effective stress, the discrepancy in this normalized cone resistance (Qtn) and n is often 1.0
case may also be attributed to the stiffness of this for clayey soils and less than 1.0 for sandy soils. The
soil layer since the soil layer is stiff to very stiff. The non-normalized classification charts (e.g., E-F’s and
discrepancy is found less pronounced in soft to firm Robertson’s SBT ones) do not take into account this
silty clayey soil even at relatively large effective characteristic.
stress (e.g., for the layers L3 and L5 at MOC site). Figure. 5 shows an example of the charts applied
to layer L4 (medium dense fine sandy) at PVN2 site.
The figure indicates that at relatively small vertical
3.2 Influence soil types
effective stress and medium dense condition, the
It is known that in a normally consolidated, homoge three charts result in similar soil types and they well
neous clay layer, the cone resistances (qc or qt) match with the soil type obtained from the VCP.
614
Similar finding is obtained from layers L1 & L2 at influenced by large effective stress and sandy soil
PVN2 site, L2 at MOC site, L1 at PVN1 site. type as discussed above.
When the depth (i.e., the effective stress) becomes
larger the influence sandy soil type to the effective
ness of the charts becomes more pronounced. For 4 CONCLUSIONS
example, Figure 6 shows a comparison of soil types
obtained from the three charts applied to layer L3 This paper presents a comparative study on the
(dense, fine to medium sand with sandy silt & silt) at applicability of the three CTPu-based soil classifica
PVN1 site. It is very clear from the figure that the tion charts, namely, Robertson’s normalized (SBTn)
Robertson’s SBTn chart results in soil types matched chart, Eslami-Fellenius (E-F)’s non-normalized chart
relatively well with the results from the VCP and Robertson’s non-normalized (SBT) chart. CPTu
whereas the two non-normalized charts indicate data in association with soil sampling at six sites
rather coarser soil types. Similar finding was found from different geological conditions were used for
from layers L4, L6, L7, L8 at MOC site. the analyses. The following key conclusions are
drawn from the study.
1) when the effectives stress is less than about 150
kPa the E-F’s and Robertson’s normalized
(SBTn) charts provide rather similar soil types
for both clayey soil and sandy soil, however the
Roberson’s non-normalized (SBT) chart tends to
indicate similar or slightly coarser grained soil
types compared with those from the normalized
chart. When the effective stress becomes larger
the influence of the stress becomes more pro
nounced, and especially the E-F’s chart also
tends to indicate slightly coarser soil type com
pared with that obtained from the lab test results
and VCP and from Robertson’s SPTn chart. This
Figure 7. Comparison of soil profiles of soil types from the indicates that the non-normalized charts tend to
charts at PVN1 site. result coarser soil types than the actual soil types
when the effective stress is large enough.
2) The characteristic of indicating coarser soil types
Figure 7 shows a typical comparison of soil types from the non-normalized charts becomes more
obtained from the lab test results and VCP and from pronounced in sandy soils, especially dense sands
the three classification charts. For simplicity, in each at large depths.
main layer, soil types from the charts are simply 3) From the analysis results in this study, it might be
noted by zone numbers that data points fall on. It is concluded that the Robertson’s normalized chart
clear from the figure classified soil types are is most reliable among the three charts.
615
ACKNOWLEDGEMENT the unified design method – A case history. Geotech
nical Special Publication No. 227, ASCE, pp.
The authors would like to thank PVI and PVN for 545–567.
their kind permission of use of data at the two off Lunne, T., Robertson, P.K. and Powell, J.J.M. (1997). Cone
shore sites. penetration testing. Blackie Academic & Professional.
Robertson, P. K (2009). Interpretation of cone penetration
tests - a unified approach. Canadian Geotechnical Jour
REFERENCES nal, 46(11),1337–1355.
Robertson, P.K. (2010). Soil behavior type from the CPT:
ASTM 5778 – 20 (2020). Standard test method for elec An update. The 2nd International symposium on Cone
tronic friction cone and piezocone penetration testing of Penetration Testing, CPT’10, Hungtington Beach, CA,
soil. ASTM International. USA.
Eslami, A., Fellenius, B. H. (1997). Pile capacity by direct Robertson, P.K. and Cabal, K.L. (2015). Guide to cone
CPT and CPTu methods applied to 102 case histories. penetration testing for geotechnical engineering (6th
Canadian Geotechnical Journal, 34(6),886–904. ed). Gregg Drilling & Testing, Inc.
Fellenius, B.H. (2021). Basics of foundation design Robertson, P.K., Campanella, R.G., Gillespie, D. and
(Jan 2021 ed.). Electronic edition. Greig, J. (1986). Use of piezometer cone data. In
Kim,S.R., Chung, S.G., Nguyen, T.D. and Fellenius, B. Situ’86 – Use of In-situ testing in Geotechnical Engin
H. (2012). Design for settlement of pile groups by eering, GSP 6, ASCE, pp. 1263–1280.
616
© 2022 the Author(s), ISBN 978-1-032-31259-0
Characterization of young sediments using CPTu and Medusa SDMT

S. Oberhollenzer, L. Hauser, F. Brand, R. Marte & H.F. Schweiger
Institute of Soil Mechanics, Foundation Engineering & Computational Geotechnics, Graz University of Technology,
Graz, Austria
D. Marchetti
S. Pfeifer
illwerke vkw AG, Bregenz, Austria
ABSTRACT: Piezocone penetration tests (CPTu) and seismic flat dilatometer tests (SDMT) present costand
time-efficient insitu investigation techniques for onshore as well as offshore projects. Since soil sampling is often
related to a strong soil disturbance in fine-grained sediments, parameter identification is frequently based on
insitu measurements in combination with correlations. As shown in previous studies, correlations are more diffi
cult to develop in silt-dominated sediments. To overcome this problem, the research project PITS (parameter
identification using insitu tests in silts), was launched by Graz University of Technology in cooperation with the
Federal Chamber of Architects and Chartered Engineering Consultants. To investigate the influence of time
effects (age) and microstructure on the load-settlement behaviour of normallyto slightly underconsolidated sedi
ments, deposits younger than 50 years have been investigated at the water storage reservoir Raggal (Austria)
using CPTu as well as Medusa SDMT. Both probes were pushed first through the water and subsequently into
the sediments by means of a stand-alone pushing device, where the testing setup was installed on a floating pon
toon. In order to prevent buckling of the penetration rods, additional casing tubes (along the water) were used. In
a last step, soil sampling was executed using the CPT-Ranger system by Geomil. To characterize and quantify
the sediments microstructure, shear wave velocities determined insitu by means of SDMT (VS,SDMT) are com
pared with measurements on reconstituted soil samples using bender elements (VS,BE). Ratios VS,BE/VS,SDMT ≈ 1
indicate the presence of no or moderate soil microstructure. On the other hand, postglacial deposits of similar
grain size distribution are characterized by smaller VS,BE/VS,SDMT ratios, indicating a higher microstructure.
1 INTRODUCTION within the grain-to-grain matrix, which lead to an

increase in strength and stiffness (Leroueil 1992). At
1.1 Motivation the same time, they can easily be destroyed due to
heavy construction measures or non-adequate soil sam
Fine-grained sediments define the soil layering of vari pling techniques. On the other hand, insitu tests and
ous basins and valleys within Alpine regions. Many of especially seismic measurements (e.g. seismic piezo
these basins and valleys were formed during several cone penetration test SCPTu or seismic flat dilatometer
glacial periods and remained as lakes after the melting. test SDMT) are becoming increasingly popular to quan
Over thousands of years they have been filled by tify and characterize microstructure (Robertson 2016).
mainly fine-grained sediments. Therefore, such (geo To gain an improved understanding of structure
logically) young soils are often characterized by a high in postglacial deposits by means of insitu tests Graz
groundwater table and are generally in a normally con University of Technology in cooperation with the
solidated or slightly under consolidated state. In the Federal Chamber of Architects and Chartered
area of Salzburg or Bregenz it was observed that such Engineering Consultants initiated the research pro
sediments often present unexpected low settlements ject PITS (parameter identification using insitu tests
under static loading on shallow foundations. On the in silts).
other hand, it was observed that dynamic loads intro
duced by heavy construction measures (e.g. soil
1.2 Aim
improvement measures, jet grouting) can lead to signifi
cant settlements. One possible explanation for this The present paper tries to investigate how aging can
observation might be related to microstructural bonds influence the development of microstructure.
DOI: 10.1201/9781003308829-89
617
Therefore, young sediments with an age of approxi A motorized syringe, driven by an electronic board
mately 50 years deposited in the water storage reser powered with rechargeable batteries, hydraulically
voir Raggal (Austria) have been investigated using expands the membrane to obtain the A, B and
CPTu, SDMT and laboratory tests (e.g. oedometer, C pressure readings, which are acquired and stored
bender element). automatically at each test depth (Marchetti 2014).
Based on the comparison of shear wave velocities, Since the test execution is performed automatic
determined insitu (Medusa SDMT) and in the ally, the influence of operators can be reduced sig
laboratory on reconstituted samples (bender elem nificantly, alternative timing of measurements
ent), the degree of microstructure could be quantified become feasible and repeated A-pressure readings
and is further compared with measurements of post- can be carried out accurately.
glacial (older) sediments. The seismic flat dilatometer is a combination of the
flat dilatometer (or Medusa DMT) with the seismic
module for measuring the shear wave velocity VS
2 INSITU TESTS behind the blade. The seismic module is a cylindrical
element situated above the DMT blade and equipped
2.1 Piezocone penetration test - CPTu with two receivers fixed at a vertical distance of
0.50 m. The measurements are commonly performed
The piezocone penetration test (CPTu) is a widely at depth intervals equal to 50 cm while stopping the
used insitu test for soil classification and parameter penetration procedure. The shear wave source –
identification. During test execution, a cone with located at the ground surface or in the present case at
a cross-section area equal to 10 or 15 cm2 is pushed the lake floor bottom - generally consists of
under constant penetration rate (2 cm/s) into the soil a S-hammer which strikes horizontally a rectangular
using a pushing device (e.g. truck, rig or demount steel plate pressed against the soil. The generated
able systems). Simultaneously, the tip resistance qc, shear wave first reaches the upper receiver, then, after
sleeve friction fs and dynamic pore water pressure ui a delay, the lower receiver. The seismograms acquired
are measured continuously over depth. The measure by the two receivers, amplified and digitized at depth,
ment of the pore water pressure is usually performed are transmitted to a computer at the surface for real-
above the cone at position u2. Alternatively, the pore time interpretation of VS (Marchetti et al. 2008).
water pressure can be measured directly at the cone
(position u1). Normalized parameters can be calcu
lated based on insitu measurements and further used 3 TEST SITE WATER STORAGE RAGGAL
for soil classification (using soil behavior type chart)
or in combination with correlations to identify soil
3.1 General information
parameters.
The water storage reservoir Raggal (see Figure 1),
located in the western part of Austria, is operated by
2.2 Seismic Medusa flat dilatometer test - Medusa
the energy operator illwerke vkw and is part of the
SDMT
hydropower plant Oberstufe Lutz. Due to natural
The flat dilatometer is an insitu soil testing equip sedimentation, about 50,000 m³ of material are
ment developed by Professor Silvano Marchetti in deposited within the storage every year. To prevent
the late 1970s (Marchetti 1980). A steel blade – con blockage of important water intake points and to
taining a thin, expandable, circular steel membrane pass the annual inflow of sediments through the
mounted on one side – is pushed into the soil on dam, these sediments are regularly removed near the
a constant penetration rate equal to 2 cm/s. The dam using a dredger (see Figure 1).
blade is connected to a pneumatic electrical cable
running through the penetration rods, up to a control
3.2 Investigation
unit at surface. In the standard testing procedure, the
penetration is stopped every 20 cm. When perform In a first step, the thickness of the sediments was
ing a classic DMT, the membrane is inflated with investigated by means of echo soundings. The exca
gas to obtain two pressure readings at defined vation works on the one hand and the location of
deformations of the membrane: the A-pressure insitu tests on the other hand have been designed/
(center of the membrane deforms 0.05 mm) and defined based on the latter. With increasing distance
B-pressure (center of the membrane deforms from the dam, the flow velocity rises and the depos
1.10 mm). A third pressure reading, the C-pressure ited sediments become coarser. This relationship
(closing pressure), can optionally be taken by slowly could be confirmed based on aerometer and sieve
deflating the membrane soon after B until it returns analyses on soil samples, recovered from the reser
to position A. voir. In the front third of the water reservoir - where
The Medusa dilatometer (Medusa DMT) is a self- the finest sediments are deposited - illwerke vkw car
contained, fully automated version of the flat dilat ried out excavation works during our investigation
ometer, able to autonomously perform dilatometer campaign. In order not to hinder the excavation work
tests without the pneumatic cable and gas tank. and to ensure all safety regulations, the insitu tests
618
were carried out approximately 300m behind the con pontoon was moved to the desired position using
crete-dam (see Figure 1). Based on echo soundings it a motorboat and additionally fixed at the shore
could be ensured to investigate sediments of approxi using four steel cables (see Figure 2g).
mately 20 m thickness. All insitu tests were carried
out starting from a floating pontoon within
4.2 Piezocone penetration test and seismic Medusa
a rectangular area of approximately 10 x 5 m to
flat dilatometer test
ensure the comparability of test results (see Figure 1).
The insitu campaign consisted of piezocone pene The water depth and the thickness of the sediment
tration tests (CPTu) with pore water pressure measure deposits were approximately 7 m and 20 m respect
ments at position u1 and u2, seismic Medusa flat ively during test execution. In order to prevent any
dilatometer tests (Medusa SDMT) and soil sampling buckling of penetration rods along the water sec
by means of CPT-Ranger. All tests were executed in tion, additional casing tubes (with a slightly larger
collaboration with the companies mjp ZT GmbH and inner-diameter than the outer diameter of the push
Studio Prof. Marchetti. ing rods) were used to increase the cross-section
and moment of resistance. In a first step, CPTu or
SDMT probes were lowered to the lake floor
bottom ensuring an embedment depth of approxi
mately 2 m. Afterwards, casing tubes were lowered
by using the stand-alone system. Once the casing
tubes reached the lake bottom (sediment top sur
face) the penetration process was continued by
means of pushing rods. Consequently, casing tubes
were installed along the water section only. Due to
the buckling problem and the limited weight of
the pontoon, the sediments were investigated
down to a depth of approximately 15 m. After
reaching the final testing depth, first the casing
tubes and subsequently the penetration rods were
pulled back.
To determine the shear wave velocity, the
S-hammer shown in Figure 3a was designed and built
in cooperation with Studio Prof. Marchetti and
Figure 1. Water reservoir Raggal: Overview and location test
Behensky. The shear wave is triggered using a 45kg
site.
drop-weight. Since water does not allow the transmis
sion of shear waves, the S-hammer was lowered from
4 TEST EXECUTION the floating pontoon to the lake bottom using two
winches (see Figures 3b and 3c). During the lowering
4.1 Floating pontoon and pushing device procedure, additional casing tubes were continuously
attached to the head of the structure to verify its pos
Since the reservoir was filled with water during test ition. A rope was used to lift and release the 45kg drop
execution, all insitu tests were executed from the weight from the floating pontoon (see Figure 3c). The
floating pontoon presented in Figure 2a (which is shear wave velocity was determined at 50 cm
usually used for sediment transportation). The con intervals.
struction is composed of four air-filled steel boxes
(two boxes each side), connected by three 2.8m
long IPE300 profiles. The external dimensions and 4.3 Soil sampling and laboratory testing
weight of the barge amount to 10 m x 8 m and 13 In a final step, soil sampling for laboratory testing
tons respectively. The mobile stand-alone system was executed in 5 depth levels using the CPT-
by Geomil (Fox-150) was used as pushing device Ranger system. All recovered samples (length =
for all insitu tests. Thereby, a separate power pack, 50 cm, diameter = 7 cm) were carefully transported
driven by a petrol motor, powers the Fox-150 to Graz University of Technology and further inves
hydraulically. In order to avoid any tilting during tigated at the geotechnical laboratory. The sedi
test execution, it was tried to fix the penetration ments were characterized with respect to their
device on the floating-pontoon center. In a first particle size distribution, Atterberg limits, natural
step, HEA profiles were clamped onto the IPE300 density, natural water content and oedometer stiff
profiles. Subsequently, the “stand-alone system” ness. Furthermore, the shear wave velocity of
was attached to the (longitudinal assigned) HEA reconstituted soil samples was determined at differ
profiles using two GEWI bars (see Figures 2b and ent insitu stress levels using bender element tests
2d). Free areas were covered by wooden construc within a triaxial cell. All reconstituted soil samples
tions to ensure work safety during test execution were artificially mixed considering the insitu dens
(see Figures 2e and 2f). Finally, the floating ity and water content.
619
Figure 2. Test execution: Floating pontoon and pushing device.
Figure 3. Test execution: Seismic measurements.
5 RESULTS sand to sandy silt” and “clay to silty clay” (see

Figure 4d). The DMT classification system accord
5.1 Insitu and laboratory results ing to Marchetti (1980) is based on the material
behaviour index ID and leads to a similar result
The sedimentation history of the water reservoir (“silty sand to sand” and “clayey to sandy silts”).
Raggal is composed of a sequence of fineand coarse- In a second step, all recovered soil samples have
grained layers. As shown in Figure 4, these sand-silt been classified based on their particle size distribu
alterations are characterized by an erratic distribution tion and Atterberg limits according to EN ISO
of CPTu measurements over depth (lithology L1). 14688-1 and ASTM 2487-11 (USCS). The particle
The tip resistance qc and sleeve friction fs vary size distribution of five depths is presented as a bar
between 0.7 - 3.5 MPa and 5 - 50 kPa respectively. chart in Figure 5a (red = clay, green = silt, blue =
The measured pore water pressure u2 corresponds to sand). It is evident that sections with increased sand
the hydrostatic (insitu) pore water pressure u0 within content are represented by higher ID (DMT) and
sand-dominated layers and rises (u2 > u0) with lower IC (CPTu) values. On the other hand, four soil
higher fines content. This varying trend is also samples - classified as clayey, sandy silts (cl’ sa’ Si)
reflected by DMT intermediate parameters ID (= 0.4 according to EN ISO 14688-1 - have been recovered
- 5), KD (= 0.5 - 6) and ED (= 0.5 – 12 MPa) in within fine-grained sections. These sediments pre
Figure 4. sent a fines content (<0,075mm) larger than 90 %
The soil behaviour type index IC – based on and are classified as organic silts (OL, OH) based on
Robertson 1990 – classifies the sediments as “silty USCS (see Figure 5c). Due to their high organic
620
Figure 4. CPTu and SDMT results: (a) tip resistance qc, (b) sleeve friction fs, (c) dynamic pore water pressures u1 and u2,
(d) soil behaviour type index IC, (e) shear wave velocity VS, (f) horizontal stress index KD, (g) dilatometer modulus ED and
(h) material behaviour index ID.
Figure 5. Laboratory results: (a) particle size distribution PSD, (b) natural water content wnat and (c) Atterberg limits.
content (2-3 %), all points are situated below the determined within fine-grained (under
A-line within the Casagrande diagram. consolidated) layers. Nevertheless, it should be
A small insitu density (ρd ≈ 1.2 g/cm³, ρsat ≈ noted that natural water contents decrease and
1.75 g/cm³) and a high natural water content insitu densities increase with increasing particle
wnat > 40 % was expected and furthermore size distribution.
621
5.2 Microstructure 6 CONCLUSION
The insitu shear wave velocity VS,SDMT determined
Alpine regions are often characterized by basin land
by means of Medusa SDMT is presented in
scapes, which were filled by fine-grained sediments
Figure 4e. As described earlier in section 4.1 for
after the last glacial period. These (normally to
CPTu and DMT measurements and intermediate
slightly under-consolidated) sediments often show
parameters, also the shear wave velocity is strongly small settlements under static loading, leading to the
influenced by the particle size distribution. Thereby, assumption that they might be characterized by
layers of higher sand-content are characterized by a microstructure. Various authors (e.g. Robertson
higher VS values. 2016) showed that structured soils present a higher
In order to quantify the degree of microstructure, shear wave velocity compared to unstructured (ideal)
bender element tests were performed in a second soils. The present paper investigated whether and to
step on reconstituted soil samples using a triaxial what extent young sediments, deposited in the water
device. It is important to note that material of soil storage reservoir Raggal, are characterized by micro-
samples recovered at -6.8 m, -8.1 m, -13.3 m and structure. In a first step CPTu, Medusa SDMT and
-16.1 m (all characterized by a similar particle size soil sampling was performed from a floating pontoon
distribution as shown in Figure 5a) was used for this using a mobile pushing device. The investigated sedi
experimental approach. After a K0-consolidation, the ments are composed of sand-silt alternations and are
shear wave velocity was determined for different characterized by a small insitu density and a high
insitu stress levels. The bender element results are water content wnat > 40 %. In a second step bender
shown in Figure 6a by blue triangles and are com element tests were performed on reconstituted soil
pared with SDMT results. Thereby, a regression samples and compared with the SDMT results. It was
(grey dotted) line - which goes through fine-grained shown that shear wave velocities determined on
sections where soil sampling was executed - is used reconstituted soil samples are in good agreement with
for comparison. Based on the ratio VS,BE/VS,SDMT it SDMT results (VS,BE/VS,SDMT ≈ 1), indicating no or
is evident that shear wave velocities determined in little microstructure. Within the research project PITS,
the laboratory by means of bender elements (VS,BE) similar investigations have been performed in older
and in situ using Medusa SDMT (VS,SDMT) are in sediments (e.g. rhine valley, basin of Salzburg) too. In
good agreement. Since VS,SDMT and VS,BE differ these cases VS,BE/VS,SDMT ratios were significantly
only slightly, it can be assumed that the investigated smaller than 1, indicating a stronger microstructure in
sediments within the reservoir Raggal are character postglacial deposits. The present contribution showed
ized by no or moderate microstructure. that VS,BE/VS,SDMT ratios are a potential indicator to
quantify and characterize soil microstructure.
REFERENCES
ASTM D2487-11. 2011. Standard Practice for Classifica
tion of Soils for Engineering Purposes (Unified Soil
Classification System).
EN ISO 14688-1. 2019. Geotechnical investigation and
testing – Identification and classification of soil.
Leroueil, S. 1992. A framework for the mechanical behav
ior of structured soils, from soft clays to weak rocks. In
Proceedings, US-Brazil NSF Geotechnical Workshop on
Applicability of Classical Soil Mechanics Principles to
Structured Soils, Belo Horizonte, pp. 107–128.
Marchetti, S. 1980. In Situ Tests by Flat Dilatometer.
J. Geotech. Eng. Div. 106(GT3): 299–321.
Marchetti, D. 2014. Device comprising an automated cable-
less dilatometer. U.S. Patent 8,776,583, filed July 29,
2011, issued July 15, 2014.
Marchetti, S., Monaco, P., Totani, G., and Marchetti, D.
2008. In Situ Tests by Seismic Dilatometer (SDMT). In
From Research to Practice in Geotechnical Engineer
ing, Geotech. Spec. Publ. GSP 180, 292–311. American
Society of Civil Engineers, Reston, VA, USA.
Robertson, P.K. 1990. Soil classification using the cone pene
tration test. Canadian Geotechnical Journal, 27(1):
151–158.
Robertson, P.K. 2016. Cone penetration test (CPT)-based
soil behaviour type (SBT) classification system — an
Figure 6. Shear wave velocity VS: Comparison of SDMT update. Canadian Geotechnical Journal, 53:
and bender element tests. 1910–1927.
622
© 2022 the Author(s), ISBN 978-1-032-31259-0
PFEM modeling of CPTu tests in saturated structured soils

Kateryna Oliynyk
School of Science and Engineering, University of Dundee, Dundee, UK
Department of Civil and Environmental Engineering, University of Perugia, Perugia, Italy
Matteo O. Ciantia
School of Science and Engineering, University of Dundee, Dundee, UK
Claudio Tamagnini
Department of Civil and Environmental Engineering, University of Perugia, Perugia, Italy
ABSTRACT: The conventional interpretation of CPTu tests is typically based on empirical and semi-empirical
correlations based on very crude descriptions of soil behavior, such as the total stress approach coupled with the
Terzaghi-Rendulic pseudo-3d consolidation theory for modeling the time evolution of excess pore water pressure.
The aim of this work is to show that a more rational interpretation of the coupled deformation and flow processes
occurring in the soil during a CPTu test is possible by resorting to the numerical solution of the relevant governing
equations, incorporating a realistic constitutive model for the soil. In order to deal with the large displacements
and deformations induced by the cone penetration, the Particle Finite Element Method code G-PFEM, recently
developed for geomechanical applications, has been used for this purpose. A key feature of the present work is the
use of a finite deformation version of a non-associative isotropic hardening plasticity model for structured geoma
terials - the FD_Milan model. The model is equipped with a structure-related internal variable which provide
a macroscopic description of the effects of structure in natural, fine-grained soils. In order to deal with strain local
ization, typically observed in structured geomaterials upon yielding, the model has been equipped with a non-local
version of the hardening laws, which has demonstrated capable of regularizing the pathological mesh dependence
of classical FE solutions in the post-localization regime. A number of PFEM simulations of CPTu tests on a soft
structured natural clay has been performed in order to assess the effects of the initial bond strength and permeabil
ity on the predicted results of the test, as well as on the spatial distributions of accumulated plastic strains, internal
variables and excess pore water pressures. The results obtained represent a promising step towards a more rational
interpretation of the CPTu tests in structured geomaterials and for their use in the calibration of advanced soil
models.
1 INTRODUCTION identify fine-grained soils (where lpw are large)

from coarse-grained soils (where lpw are small
The Cone Penetration Test (CPT) is a widely used or negligible), or to measure the soil coefficient
investigation tool for the characterization of both of consolidation from pore pressure dissipation
coarse-and fine-grained soils, for its simplicity, reli stages, where the tip advancement is stopped.
ability and its relatively low cost. The development The aim of this work is to show that a more
of the piezocone, featuring piezometers for the meas rational interpretation of the coupled deformation and
urement of pore water pressure pw at different posi flow processes occurring in the soil during a CPTu
tions on the cone tip, has significantly improved the test is possible by resorting to the numerical solution
capabilities of this site investigation tool - now of the relevant governing equations for the penetra
referred to as CPTu test - and opened the way for tion process, and by adopting a realistic constitutive
a more rational interpretation of the test results. model for the soil. In early attempts to simulate the
The conventional interpretation of CPT tests is CPT test with the ALE-FE method, the extreme
currently based on empirical and semi-empirical deformations imposed to the soil by the penetration
correlations, the last based on very crude descrip of the piezocone have represented a substantial diffi
tions of soil behavior, such as the adoption of culty, due to strong mesh distortions which resulted
a total stress approach. The excess pore pressure in significant loss of accuracy or loss of convergence
measurements in CPTu tests are typically used to at relatively small penetration depths.
DOI: 10.1201/9781003308829-90
623
An effective alternative to the FEM which has evolution equations in the spatial setting are briefly
proven to be quite efficient in simulating the cone summarized here:
penetration process is the Particle Finite Element
Method (PFEM, Oñate et al. 2011). The PFEM
shares many similarities with the updated Lagran
gian approach of non-linear FEM, but it is capable
of handling the problem of severe mesh distortion by
a frequent mesh re-triangulation and h-adaptive
refinement, using very efficient algorithms based on
extended Delaunay tesselation. The nodes of the spa
tial discretization - performed with low-order linear
triangles or tetrahedra - are treated as material par
ticles, the motion of which is tracked during the
numerical simulation. Applications of PFEM to the
modeling of CPTu tests have been reported, e.g., by
subjected to the Kuhn-Tucker complementarity
Monforte et al. (2017), Monforte et al. (2018), Mon-
conditions:
forte et al. (2021), Hauser and Schweiger (2021) and
Carbonell et al. (2022).
As far as modeling CPTu tests in clays is con
cerned, most of the works cited have been carried
out adopting the classical MCC model for the soil. r
Although this critical state model is capable of cap In the above equations, τ and τ are the Kirchhoff
turing the essential features of soft, lightly overcon stress and its Jaumann objective rate; d is the rate of
solidated clays, it fails to reproduce the behavior of deformation tensor; d p is the plastic rate of deform
natural structured clays, characterized by the pres ation tensor; ae is the spatial hyperelastic tangent
ence of intergranular bonds of various origin. To stiffness of the material; γ_ is the plastic multiplier; f
investigate the effects of bonding on the soil and g are the yield function and the plastic potential,
response to the piezocone advancement, in this work respectively (see Oliynyk et al. 2021 for details); Ps
the isotropic hardening elastoplastic model for nat and Pt are internal variables; the functions
ural, structured soils proposed by Nova and co
workers (Tamagnini et al. 2002, Nova et al. 2003)
has been extended to finite deformations adopting
a multiplicative decomposition of the deformation
gradient into elastic and plastic parts (see, e.g., Borja
& Tamagnini 1998). The finite deformation plasticity are the plastic volumetric and deviatoric rates of
model thus obtained - referred to in the following as deformation; and ρs , ξ s , ρt and ξ t are material
the FD_Milan model - has been implemented in constants.
a geomechanics-oriented PFEM code (G–PFEM, The first internal variables Ps (preconsolidation
Monforte et al. 2017) and has been used in this work pressure) accounts for the hardening/softening
to simulate CPTu tests in a relatively soft natural effects due to volumetric and deviatoric plastic
clay. strains. The second, Pt (bond strength), quantifies
The remainder of the paper is as follows. A brief the effects of material structure (fabric and bonding).
overview of the FD_Milan model is provided in A representation of the yield surface in the Kirchhoff
Sect. 2, while the details of the CPTu simulations stress invariants space P:Q, highlighting the role of
program are given in Sect. 3. A selection of the the internal variables, is given in Figure 1.
results obtained in the PFEM simulations is pre For ξ s ¼ 0, eq. (3) reduces to the classical volu
sented in Sect. 4. Sect. 5 provides the main conclud metric hardening law of critical state soil mechanics.
ing remarks and suggestions for further studies. Plastic volumetric compaction produces an increase
of Ps while plastic dilation is accompanied by
a reduction of Ps .
2 THE FD_MILAN MODEL A positive value of the bond strength Pt results in
an expansion of the elastic domain in stress space.
2.1 Local version On the positive part of the P axis, the isotropic yield
stress in compression is increased by a quantity
The FD_Milan model is a non-associative, isotropic Pm ¼ kPt , with k a material constant. On the nega
hardening finite-deformation plasticity model for tive axis, the vertex of the yield surface is displaced
structured soils and weak rocks, developed by from the origin of the stress space to P ¼ -Pt .
Oliynyk et al. (2021) based on the multiplicative Therefore, the structured material possesses a true
decomposition of the deformation gradient and on cohesion and a non-negligible tensile strength. The
the adoption of a suitable free energy function to region in the stress space contained between the
describe the elastic response of the material. Its actual yield surface (full black line in Figure 1) and
624
and Ps0 and Pt0 represent the initial values of the
internal variables. Once Evp , Esp and Nvp are chosen as
alternative state variable, their values can be spa
tially averaged in a neighborhood O of each material
point x ¼ ðX ; tÞ at time t. The spatial averages are
computed numerically by the following expressions:
Figure 1. Yield surface and plastic potential of the FD_Mi

lan model.
where the symbol A stands for �one of the � variables

Epv , Esp or Nvp. In eq. (9), rij ¼ �xj - xi � is the dis
the intrinsic yield surface of the unstructured
tance between points located at xj and xi , and w is
material (with Pt ¼ Pm ¼ 0, black dashed line in
a suitable weighting function, for which the expres-
Figure 1), represents the so-called structure
sion proposed by (Galavi & Schweiger 2010):
permitted space, which is accessible only by the
combined effects of fabric and intergranular bond
ing. As a consequence of the microstructural
rearrangement of the soil fabric and the concurrent
breakage of intergranular bonds, the structure of
the material is progressively destroyed. From
a macroscopic point of view, this process induces
a progressive reduction of Pt with increasing accu has been adopted. The scalar quantity ‘c appearing
mulated plastic deformations. This process is in eq. (10) is a material constant providing the
described by the hardening law of eq. (4), in length scale sought after. Once the non-local quan-
p p p
which the RHS is always negative. The final tities Ev , Es and N v are known, the stress-like
asymptotic value for Pt is zero (full destructura internal variables Ps and Pt are computed by eqs.
tion): the bond-permitted space reduces to zero (6) and (7).
and the yield surface coincides with the intrinsic
yield surface.
3 SIMULATION PROGRAM
2.2 Non-local version In the PFEM simulations of CPTu tests, a standard
In order to provide a characteristic length scale to piezocone, with radius R = 1.78 cm and a cone tip
the constitutive equation, to regularize the numerical angle of 60� is inserted in a calibration chamber with
solution in presence of strain localization phenom radius B = 0.45 m and height H = 1.05 m, filled with
ena, the integral non-local approach of Oliynyk et al. a fully saturated clay. The problem has been
(2021) has been adopted, in which the internal vari assumed as axisymmetric. A sketch of the problem
ables Ps and Pt are treated as non-local, spatially geometry along with some snapshots of the spatial
averaged quantities over a neighborhood O of the adaptive PFEM discretization at 3 different tip
material point. The size of this neighborhood is con advancement depths is given in Figure 2.
trolled by a material constant ‘c , called characteris The piezocone is wished-in-place at an initial depth
tic length. z0 = 0.25 m and then displaced downwards at
Let us rewrite the hardening laws (3) and (4) in a constant penetration speed of 2.0 cm/s, up to a depth
the following equivalent integrated form: z ¼ z0 þ 20R. The piezocone tip and its lateral surface
are modeled as rigid, impervious surfaces, and
a smooth contact interface with the soil is adopted.
Given the relatively small dimensions of the calibra
tion chamber, the self weights of the pore water and of
the soil have been ignored. The initial pore water pres
sure has been assumed uniform and equal to zero.
The material constants adopted in the simulations
where Epv , Eps and Nsp are three strain-like internal have been obtained by calibrating the model on the
variables, whose evolution equations are pro available experimental data for the Osaka clay
vided by: (Adachi et al. 1995) and are reported in Table 1. The
625
Figure 2. Problem geometry and spatial discretizations at 3 different tip advancement depths.
Table 1. Sets of material constants adopted in the CPTu the bond strength, Pt0 , from 0 to 60 kPa. With these
test simulations (see Oliynyk et al. 2021 for details on the initial conditions and the material properties listed in
meaning of each constant). Tab. 1 the apparent peak undrained strength cu;p of
^κ G0 α Pref M f ;c αf μf M g;c
the soil ranges from 41 to 106 kPa.
(−) (MPa) (−) (MPa) (−) (−) (−) (−)
The effects of the initial degree of bonding are
shown in Figures 3 and 4. The simulations reported
1.82 3.0 0.0 5.0 1.1 0.75 1.50 1.1 have been performed with kh = 1.0e-9 m/s. For such
a low permeability value, the penetration process
αg μg ρs ρt ξs ξt k ‘c occurs in almost undrained conditions. Figure 3
(−) (−) (−) (−) (−) (−) (−) (mm) shows the profiles, relative to the cases of Pt0 equal
0.75 1.5 8.33 15.0 0.0 3.0 5.0 5.0 to 0, 30 and 60 kPa, of net cone resistance, qn and
excess pore water pressures computed at the cone
base (lpw2 ) and at the cone tip (lpw1 as a function
flow rule has been assumed as associative, so the of the normalized penetration depth Z ¼ ðz - z0 Þ=R.
plastic potential coincides with the yield function.
All the PFEM simulations have been performed
as fully coupled hydromechanical problems, adopt
ing the mixed u-�-pw formulation of Monforte et al.
(2017). The bottom and lateral surfaces of the cali
bration chamber have been assumed as rigid, imper
vious and perfectly rough boundaries. At the top
surface of the soil body a uniform normal pressure
q0 = 100 kPa and a constant pore water pressure
pw ¼ 0 have been imposed. Consistently with these
boundary conditions, the initial Cauchy effective
stress in the soil mass has been assumed axisymmet
ric, with components σz = 100 kPa and σr ¼ K 0 σz
and K 0 = 0.5. The initial value of Ps has been set to
120 kPa.
4 SELECTED RESULTS
Six simulations have been performed considering 4 Figure 3. Profiles of net cone resistance, excess pore pres
different hydraulic permeabilities, kh , in the range sure lpw;2 and lpw;1 for different initial values of bond
1.0e-9 to 1.0e6 m/s, and 3 different initial values for strength Pt0 and kh = 1.0e-9 m/s.
626
After an initial transient stage up to Z ¼ 5, all the The effects of soil permeability on the predicted
profiles reach a stationary state up to the final value CPTu results are shown in Figure 5, reporting the pro
of the penetration Z ¼ 20. It can be observed that the files of qn , lpw;1 and lpw;2 with Z for the four kh
stationary values of qn , lpw1 and lpw2 increase values considered. From the figure, the impact of the
with increasing Pt0 : the higher is the initial structure soil permeability on the computed excess pore pres
of the soil, the higher are the cone resistance and the sures at the two piezometers locations is immediately
excess pore pressures developed during the cone apparent. In the simulation with the highest permeabil
advancement. ity value the soil deforms in almost drained conditions,
Figure. 4 shows a comparison of the contour with a maximum lpw of about 50 kPa at the position
maps of Eps , Ps , Pt and lpw1 for the two extreme of the piezometer no. 1. The maximum lpw are
cases of Pt0 = 0 and 60 kPa. In the unstructured obtained at both piezometers for the lowest Kh value
soil, the plastic zone around the piezocone extends (1.0e-9 m/s), and only slightly smaller values of
by about 3R around the shaft and the cone, with excess pore water pressure are registered at piezometer
contour lines following the shape of the penetrating no. 2 for the next to smallest permeability. This value
device. In the structured soil, on the other hand, the of kh (1.0e-8 m/s) appears to mark the transition from
observed pattern of plastic shear deformations is fully undrained to partially drained soil response.
much more irregular and show the presence of
localized shear zones which originate at about 3R
below the cone tip and bend upwards as the cone
advances. The presence of shear localization is
most likely due to the softening behavior associated
to soil destructuration. In both cases, the preconso
lidation pressure Ps is only slightly affected by the
penetration process, due to the virtually undrained
nature of the soil deformation. In the structured
soil, the contour map of Ps highlights the presence
of a region of decreasing Ps , located at the bound
ary of the plastic region, where plastic dilatancy is
occurring. The contour map of the bond strength
for the structured soil clearly indicates that, in most
of the plastic region around the piezocone, the
destructuration process is almost complete. Yet, the
presence of structure maintains a non-negligible
effect on both the cone tip resistance and the spatial Figure 5. Profiles of net cone resistance, excess pore pres
distribution of the pore water pressures, as shown sure lpw;2 and lpw;1 for different soil permeabilities and
in the contour maps of pw . Pt0 = 30 kPa.
As expected, the qn profiles show a larger cone

resistance in the ‘‘drained’’ case, as compared to the
‘‘undrained’’ one, with the two intermediate cases
profiles remaining close to the two extremes. Sur
prisingly, the difference between the different pro
files of qn is not very large, being on average
around 13%.
The contour maps of Epv in Figure 6 and of Esp , Ps ,
Pt and pw in Figure 7, for the two limiting cases of
the smallest and highest permeability values adopted,
indicate that, apart from the large differences
observed in the final pore water pressure fields, the
collapse mechanisms around the advancing cone tip
is fundamentally different in the two cases. In the
‘‘undrained’’ case, the plastic region around the
advancing piezocone is quite large, extending for
about 6R both vertically, below the cone tip, and
horizontally from the piezocone axis. The plastic
zone in characterized by a complex pattern of accu
Figure 4. CPTu tests results: contour maps of Eps , Ps , Pt mulated volumetric and deviatoric plastic deform
and pw at the maximum penetration depth, for Pt0 = 0 kPa ations, with clearly visible bands of localized shear
(top) and Pt0 = 60 kPa (bottom). and volumetric deformations. On the other hand, in
627
5 CONCLUDING REMARKS
The results of this preliminary study appear consistent

with the observed CPTu profiles in similar soils and
demonstrate that the soil response to the penetration
of the piezocone in CPTu tests can actually be mod
eled quantitatively in a rational and computationally
efficient way. The quality of the results obtained
depends crucially on the capability of the constitutive
model adopted to capture the essential features of the
soil response during the penetration process. The
FD_MILAN model presented in this work appears
particularly well suited in this respect for a wide
range of natural geomaterials, ranging from natural
clays to porous soft rocks. The multiplicative plasti
city model recently proposed by Oliynyk & Tamag
nini (2020), based on the breakage mechanics, can
represent a possible alternative for coarse-grained
soils with crushable grains and collapsible weak
rocks. The numerical simulations, performed with dif
ferent initial bond strengths and soil permeabilities,
show that the coupled non-linear PFEM model is cap
Figure 6. Contour maps of Epv for the maximum and min- able of capturing: a) the development of plastic
imum permeabilities considered. deformations induced by the advancement of the cone
tip; b) the destructuration associated with plastic
deformations; c) the space and time evolution of pore
water pressure during the test. A characteristic of the
plastic deformations pattern in presence of structure is
the appearance of shear bands below the cone tip
which propagate laterally as the tip advances and
remain stationary thereafter. This feature is most
likely associated to the softening mechanism induced
by soil destructuration, as described by the FD_MI
LAN model. These results represent a promising step
towards a more rational interpretation of the CPTu
tests in structured geomaterials and for their use in the
calibration of advanced soil models. The extension of
the proposed model to incorporate thermal coupling
effects (see, e.g., Tamagnini and Ciantia 2016) will
allow the interpretation of T-CPTu tests, which repre
sent a promising development for the thermal charac
terization of soils in the design of low-enthalpy
energy geostructures.
REFERENCES
Figure 7. CPTu tests results: contour maps of Eps , Ps , Pt
and pw at the maximum penetration depth, for the minimum Adachi, T., F. Oka, T. Hirata, T. Hashimoto, J. Nagaya,
(top) and maximum (bottom) permeabilities considered. M. Mimura, & T. B. S. Pradhan (1995). Stress–strain
behavior and yielding characteristics of Eastern Osaka
clay. Soils and Foundations 35(3), 1–13.
Borja, R. I. & C. Tamagnini (1998). Cam–clay plasticity,
the ‘‘drained’’ case, the penetration mechanism is part III: Extension of the infinitesimal model to include
more similar to a ‘‘punching’’ process, with a much finite strains. Comp. Meth. Appl. Mech. Engng. 155,
73–95.
smaller and homogeneous plastic zone around the Carbonell, J. M., L. Monforte, M. O. Ciantia, M. Arroyo,
cone tip and shaft. This certainly has a major effect & A. Gens (2022). Geotechnical particle finite element
on the stress distribution in the soil, which could method for modeling of soil-structure interaction under
explain the relatively modest differences registered in large deformation conditions. J. of Rock Mech. Geotech.
the stationary values of qn in the two cases. Engng., http://dx.doi.org/10.1016/j.jrmge.2021.12.006.
628
Galavi, V. & H. F. Schweiger (2010). Nonlocal multilami Oliynyk, K., M. O. Ciantia, & C. Tamagnini (2021).
nate model for strain softening analysis. International A finite deformation multiplicative plasticity model with
Journal of Geomechanics 10(1), 30–44. non–local hardening for bonded geomaterials. Comp. &
Hauser, L. & H. F. Schweiger (2021). Numerical study on Geotechnics 137.
undrained cone penetration in structured soil using Oliynyk, K. & C. Tamagnini (2020). Finite deformation
gpfem. Computers & Geotechnics 133, 104061. hyperplasticity theory for crushable, cemented granular
Monforte, L., M. Arroyo, J. M. Carbonell, & A. Gens materials. Open Geomechanics 2, 1–33.
(2018). Coupled effective stress analysis of insertion Oñate, E., S. R. Idelsohn, M. A. Celigueta, R. Rossi,
problems in geotechnics with the particle finite element J. Marti, J. M. Carbonell, P. Ryzhakov, & B. Suárez
method. Computers & Geotechnics 101, 114–129. (2011). Advances in the particle finite element method
Monforte, L., J. M. Carbonell, M. Arroyo, & A. Gens (pfem) for solving coupled problems in engineering. In
(2017). Performance of mixed formulations for the par Particle–Based Methods, pp. 1–49. Springer.
ticle finite element method in soil mechanics problems. Tamagnini, C., R. Castellanza,&R. Nova (2002).
Computational Particle Mechanics 4(3), 269–284. A Generalized Backward Euler algorithm for the numer
Monforte, L., A. Gens, M. Arroyo, M. Manica, & ical integration of an isotropic hardening elastoplastic
J. M. Carbonell (2021). Analysis of cone penetration in model for mechanical and chemical degradation of
brittle liquefiable soils. Computers and Geotechnics 134. bonded geomaterials. Int. J. Num. Anal. Meth. Geomech.
Nova, R., R. Castellanza, & C. Tamagnini (2003). 26, 963–1004.
A constitutive model for bonded geomaterials subject to Tamagnini, C. & M. O. Ciantia (2016). Plasticity with gen
mechanical and/or chemical degradation. Int. J. Num. eralized hardening: constitutive modeling and computa
Anal. Meth. Geomech. 27(9), 705–732. tional aspects. Acta Geotechnica 11(3), 595–623.
629
© 2022 the Author(s), ISBN 978-1-032-31259-0
Effect of the scatter between CPTU measured parameters in soil

classification
P. Paniagua
Norwegian Geotechnical Institute & Norwegian University of Science & Technology, Trondheim, Norway
J.-S. L’Heureux
ABSTRACT: Using piezocones from different manufacturers may yield different results even if the equip
ment complies with international standards. This causes problems when soil investigation contractors, using
different cones, operate in the same area, and especially on the same project. Studies done in soft clay, sand,
silt, and quick clays from different Norwegian sites show that repeatability for the cone resistance measure
ments and penetration pore pressure is good and that it has improved from one cone type to another. However,
the scatter in the measured sleeve friction, and hence the friction ratio, is still very significant. Here, an
attempt is made to qualitatively describe the impact of the scatter in the soil classification based on CPTU
parameters. The most common soil classification charts are used to illustrate this scatter.
1 INTRODUCTION piezocones. Further analyses regarding the statistical dis

tribution of scatter with depth and between piezocones
Performing piezocone tests (CPTU) with cone pene for each site is done in Lindgård et al. (2018), Lunne
trometers from different manufacturers may give differ et al. (2018), Paniagua et al. (2020) and NGI (2020).
ent results even though the equipment is aligned with
international standards. Previous studies (Lunne et al.
1986, Gauer et al. 2002, Powell & Lunne 2005, Tiggle 2 METHODOLOGY
mann & Beukema 2008, Lunne 2010, Cabal & Robert
son 2014 Lunne et al. 2018) have shown that all three Two relevant CPTU data sets selected at the NGTS
measured parameters, cone resistance qc, friction sleeve study sites in Norway have been chosen for this study.
fs and pore pressures u2, could vary significantly, and in The data sets selected correspond to the ones giving the
particular for fs, depending on the equipment used. largest scatter between them, i.e., the maximum (data in
Recent advances in electronics and cone design have red in the figures) and minimum (data in blue in the fig
been incorporated in the design of the new cones and ures) values. In other words, the selected data sets come
motivated further field testing. The establishment of the from the cones which provide extreme profile results.
Norwegian GeoTest Sites (NGTS) (L’Heureux & Lunne The difference between the two data sets presented here
2019) has given the opportunity to different companies are measured in terms of its CPTU parameters qt, u2 and
to do testing at the silt (Paniagua et al. 2021), sand (NGI fs, and then the data was plotted in terms of the classifi
2020), quick clay (Lindgård et al. 2019) and soft clay cation systems presented in this paper.
(Lunne et al. 2018) sites. In general, after these tests, it The data sets are then plotted in the soil classifica
was observed that the measured u2 and qc (or corrected tion charts proposed by Robertson (1990), Schneider
cone resistance, qt) showed little variation between dif et al. (2008) and Senneset et al. (1989). The data
ferent cone types, while the fs measurement gave rela plotted in the soil classification charts has been fil
tively large scatter between different cone types. Then, tered to represent just the layer where the main soil
the question that arises is how much this scatter in the material for the site is to be found.
CPTU parameters influence further treatment of the data It should be clarified that the discussion about the
like for example soil classification based on CPTU. validity of the soil classification charts mentioned
The present article investigates the impact of varying above for Norwegian soil conditions is out of the
CPTU equipment and its scatter on the most common scope of this study. Therefore, it is assumed that the
CPTU soil behaviour charts. The paper focuses on the soil classification charts may represent fairly well
effect of the scatter between different types of the soil types studied in the present paper.
DOI: 10.1201/9781003308829-91
630
The observations presented here assume that the the formula qt = qc + a (1-u2), where the parameter
requirements and recommendations given in ISO a relates the cross-sectional area of the shaft and pro
22476-1:2012 (Geotechnical investigation and testing jected area of the cone.
Field testing - Part 1: Electrical cone and piezocone test Robertson (1990) relates the normalized param
ing) and Norwegian Geotechnical Society (NGF) Guide eters Qt, Bq and Fr, noted as normalized cone resist
line No. 5 (2010) are followed. Some of these are: ance, pore pressure parameter and normalized
friction ratio, respectively.
– Zero readings to be taken before and after each Schneider et al. (2008) and Senneset et al. (1989)
test with the cone penetrometer at a temperature
make use of some of the normalized parameters (i.e.,
as close as possible to ground temperature. Qt and Bq) and relate them to either Δu/σvo’ for
– It is important to wait until the readings have sta
Schneider et al (2008) or qt for Senneset et al. (1989).
bilized before taking zero readings.
– The thrust machine shall push the rods so that the
axis of the pushing force is as close to vertical as 3 RESULTS
possible.
– The pore pressure measurement system shall be The main observations for all sites are given and dis
saturated to give good pore pressure response cussed below. However, due to space limitation focus
during penetration. will mainly be given to the results obtained at the
– For deep CPTUs, it is important to correct the Halden and Onsøy sites, and some specific observations
penetration length for inclination effects. for Tiller-Flotten and Øysand. The complete figures for
– Recommended minimum distance between Tiller-Flotten and Øysand can be sent to the interested
a CPT and adjacent boreholes is 2 m. readers upon request to [email protected].
3.1 Tiller-Flotten quick clay site

2.1 Test sites
The testing of eight different cones from five CPTU
The NGTS sites selected for the present study are manufacturers) showed less variability for the u2 param
well portrayed in the following publications and are eter. The qt value showed a larger variability than u2 but
shortly described as follows: lower than fs (Lindgård et al. 2018). In fact, there are
– Tiller-Flotten quick clay (L’Heureux et al. 2019): more uncertainties associated to the fs value since some
The site consists of thick marine clay deposit of of the cone types gave good repeatability for fs readings,
a low to medium sensitivity (1-7,5 m) deposit over while some show relatively large variation.
a high sensitivity (quick) clay (7,5-20 m depth). The following observations were made at Tiller-
– Onsøy soft clay (Gundersen et al. 2019): The site Flotten:
consists of a high plasticity marine clay (1- – For the Robertson (1990) Qt-Bq plot, the data
8 m depth), a medium plasticity clay (8- sets with largest variability all fall in the same
13 m depth) and a high plasticity clay (13-20m). soil behaviour type “clay-sensitive” clays. This
– Halden silt (Blaker et al. 2019): The silt layer observation is valid no matter where the largest
varies between 5-15 m and consists of a uniform scatter is in the CPTU parameters.
marine natural silt. – For Robertson (1990) Qt-Fr plot, when the variabil
– Øysand sand (Quinteros et al. 2019): The site ity focuses in fs, the data plots in two neighbouring
consists of fluvial and deltaic gravelly-sandy-silt areas (sensitive clay and the limit sensitive clay-
sediments with a gravelly sand layer from clay). When the variability appears in the qt or u2
0-14 m, a silt layer from 14-17,5 m depth and value, then the data plots in the same soil type,
a sand layer down to 20 m depth. however, with a tendency to cover opposite areas in
the same or neighbouring soil types.
– For Schneider et al. (2008) (Qt-Δu/σvo’) chart
2.2 Derived CPTU parameters studied and Senneset et al. (1989) (qt-Bq) chart, no clear
The soil classification charts are a combined repre difference is observed between the data sets
sentation of the in-situ behaviour characteristics of showing largest scatter, for qt and u2 parameters.
the soil under a CPTU. The CPTU might be able to The quick clay is classified as sensitive clay (1c)
measure up to seven independent parameters like in Qt-Δu/σvo’ plot and over the soft to very soft
cone resistance, qc, friction sleeve, fs, pore pressure, clay area in the qt-Bq plot. However, the variation
u2, shear wave velocity, Vs, and when performing in fs for Senneset et al. (1989) shows that the soil
dissipation tests, the time for 50% consolidation, t50, can be classified as stiff clay-silt.
in situ pore pressure, uo, and hydraulic gradient, i.
Usually the most common parameters (i.e., qc, fs
and u2) are normalized to be represented in soil clas 3.2 Onsøy soft clay site
sification charts. Before normalization, the parameter Seven different cone penetrometers from five manufac
qc is corrected for unequal pore pressure effects by turers were used in the comparative testing program and
631
it was concluded that u2 was the parameter that showed varying soil conditions for the site. Quinteros et al.
better repeatability, followed by qt which generally (2019) explained this as consequence of the depos
varies somewhat more (Lunne et al. 2018). Some of the itional history of the site and the influence of the del
cone types give good repeatability for fs readings, while taic foresee beds dipping at an angle of 20-25
some show relatively large variation. degrees. By adjusting the CPTU results in depth, the
When looking at the derived CPTU parameters in sand layers appeared to be more homogenous. The
the soil classification charts, the following aspects relative variation in qt and u2 was small and for all
are observed (see Figure 1): practical purposes negligible. However, sleeve fric
tion results showed a large scatter between the dif
– In Qt-Bq and Qt-Fr charts from Robertson (1990), the ferent cone types. The variation in sleeve friction
soil type is classified in soil types 1 and 3 (sensitive also seemed to increase with depth.
clay-clay) independently of the scatter in any of the Comparison of CPTU data from Øysand on the
CPTU parameters. However, for Qt-Fr chart, when soil classification charts leads to the following
the scatter is in the u2 and fs parameter, the data observations:
moves also towards a soil type 2 (organic soils).
– Regarding the charts of Schneider et al. (2008) (Qt – In Qt-Bq and Qt-Fr charts from Robertson (1990), the
-Δu/σvo’) and Senneset et al. (1989) (qt-Bq), no dif soil type is mainly classified from soil types 4 to 7
ference is observed between the data sets showing (silt mixtures to gravelly sand) independently of the
largest scatter, for qt and u2 parameters. The clay is scatter in any of the measured CPTU parameters.
classified as clay (1b)-sensitive clay (1c) in Qt – A similar observation applies for the charts of
-Δu/σvo’ plot and soft-very soft clay in qt-Bq plot. Schneider et al. (2008) (Qt-Δu/σvo’) and Senneset
However, the variation in fs for Senneset et al. et al. (1989) (qt-Bq), where no difference is
(1989) shows that the soil can be classified as fine observed between the data sets showing largest
silt-medium clay. scatter. The deposit is classified as transitional
soils (type 3)-sands (type 2) in Qt-Δu/σvo’ plot and
from silt to sand/hard stiff soil in the qt-Bq plot.
3.3 Halden silt site
After testing five different cone penetrometers on
this site, it was concluded that u2 and qt showed
good repeatability between the measurements, while
fs gave the largest variation (Paniagua et al. 2020).
Figure 2 presents the CPTU data sets that give the
largest scatter for each of the CPTU parameters in
the soil classification charts. Observations made are
as follow:
– In the chart Qt-Bq from Robertson (1990), the soil
type is classified in soil types 3 and 4 (silty clay-
silt mixtures) independently of the scatter in any
of the measured CPTU parameters. However, for
Qt-Fr chart, when the scatter is in the fs parameter,
the data mainly plots in two different soil types
(either type 3 or 4). A similar observation applies
when the scatter is in the qt parameter but does
not apply when the scatter is in the u2 parameter.
– For the other charts relating Qt-Δu/σvo’ (Schnei
der et al. 2008) and qt-Bq (Senneset et al. 1989),
Figure 1. Two CPTU data sets from Tiller-Flotten plotted
no difference is observed (i.e., both CPTU data in two classification charts for two CPTU parameters show
sets plot on top of each other in roughly the same ing extreme values (min-blue and max-red).
soil types: silts (1a)-transitional soils (3) and silt-
fine silt, respectively. An exception is observed
when the scatter is in u2 for the Schneider et al. 4 DISCUSSION AND CONCLUSION
(1998) plot, where the data sets tend to move to
neighbouring soil types (one to soil type 3 and Generally, previous research om NGTS sites has
the other one to soil type 1a). shown that the measured u2 shows less variation for
one cone type to another while, while qt shows
somewhat larger variation and fs the largest variation.
3.4 Øysand sand site Fr shows much larger variation compared to Bq.
Therefore, this study has investigated the impact
The CPTU parameters from nine cone penetrometers of using CPTU data from different cone types or
types tested showed results more dependent on the manufacturers and its influence on the interpretation
632
Figure 2. Two CPTU data sets from Onsøy plotted in the different classification charts for each of the CPTU parameters
showing extreme values (min and max).
633
Figure 3. Two CPTU data sets from Halden plotted in the different classification charts for each of the CPTU parameters
showing extreme values (min and max).
634
Gauer P, Lunne T, Mlynarek Z, Wolynski W & Croll M
2002 Quality of CPTu -Statistical analyses of CPTu data
from Onsøy NGI Report No.: 20001099-2 Oslo
Gundersen AS, Hansen RC, Lunne T, L’Heureux JS,
Strandvik SO 2019 Characterization and engineering
properties of the NGTS Onsøy soft clay site. AIMS
Geosciences 5 665–703.
ISO 2012 Geotechnical investigation and testing – Field
testing – Part 1: Electrical cone and piezocone pene
tration tests International Standard ISO 22476-1:
2012
L’Heureux J-S & Lunne T 2019 Characterization and
Figure 4. Two CPTU data sets from Øysand plotted in two
Engineering properties of Natural Soils used for Geo-
classification charts for two CPTU parameters showing
testing AIMS Geosciences 5 940–59.
extreme values (min-blue and max-red).
L’Heureux JS, Lindgård A, Emdal A 2019 The
Tiller-Flotten research site: Geotechnical characteriza
of soil behavior. The most common soil classification tion of a sensitive clay deposit. AIMS Geosciences 5
charts have been used in four different soil types. 831–867
In general, soil behavior charts using pore pressure Lindgård A, Gundersen A, Lunne T, L’Heureux J-S,
measurement and/or derived parameters involving u2 Kåsin K, Haugen E, Emdal A, Carlson M, Veldhuizen A
seem to show less variability in soil classification & Uruci E 2019 Effect of cone type on measured CPTU
between the different charts and all soil types. results from Tiller-Flotten quick clay site Proceedings
Geoteknikkdagen 2018 Oslo 38.1–15
A similar trend is also observed for the cone resistance.
Lunne T 2010 The CPT in Offshore Soil Investigation –
In other words, the variation of u2 and qt between data a historic perspective Proc. CPT’10 Los Angeles
sets for a defined soil type do not seem to give 71–113.
a different soil classification in the charts studied here. Lunne T, Eidsmoen T, Gillespie D & Howland J D 1986
The previous observation does not apply when Laboratory and field evaluation on cone penetrometers
using charts based on sleeve friction measure Proceedings of ASCE Specialty Conference In Situ’86:
ments. This was highlighted for the Halden silt Use of In Situ Tests in Geotechnical Engineering
when the classification chart involved a parameter Blacksburg ASCE 714–29
derived from fs. It seems that the variation in fs Lunne T, Strandvik S, Kåsin K, L’Heureux J-S, Haugen E,
for Tiller-Flotten clay and Onsøy clay does not Uruci E and Kassner M 2018 Effect of cone penetrom
eter type on CPTU results at a soft clay test site in
have such a strong impact in the classification of
Norway. Cone Penetration Testing IV: Proc CPT 2018
the material using the different charts. This seems Delft 417–22
to also apply for Øysand sand, however, the NGF 2010 Guideline for execution of CPTU tests Guide
observations for this site could be influenced by line No. 5 Norwegian Geotechnical Society [in
the natural variability of the deposit and therefore Norwegian]
it is difficult to conclude. NGI 2020 Impact of cone penetrometer type on measured
Finally, due to the large uncertainties with the fs CPTU parameters at 4 NGTS sites: silt, soft clay, sand,
readings, once should be careful using this param and quick clay. NGI report 20160154-21-R.
eter, and Fr, when interpretating soil parameters for Paniagua P, Lunne T, Gundersen A., L’Heureux JS,
design. Since the measured u2 appear to frequently Kåsin K 2021 CPTU results at a silt test site in Norway:
effect of cone penetrometer type. in IOP Conference
be the most reliable parameters it should be used in
Series: Earth and Environmental Science. IOP
addition to qt for deriving soil parameters. Publishing,
Powell J & Lunne T 2005 A comparison of different
piezocones in UK clays Proc. 16th International
ACKNOWLEDGEMENTS Conf on Soil Mechanics and Geotechnical Engineer
ing Osaka 729–34
-To the Research Council of Norway for funding the Quinteros S, Gundersen A, L’Heureux JS, Carraro A,
Norwegian GeoTest Site project (No. 245650/F50) Jardine R 2019 Øysand research site: Geotechnical char
and the master student Erika Solem for her assist acterization of deltaic sandy-silty soils. AIMS Geo-
ance in plotting the data during the summer 2021. sciences 5 750–783.
Robertson PK 1990 Soil classification using the cone penetra
tion test. Canadian Geotechnical Journal, 27: 151–158.
Schneider JA, Randolph MF, Mayne PW & Ramsey N
REFERENCES
2008. Analysis of factors influencing soil classification
Blaker Ø, Carroll R, Paniagua P, DeGroot D & using normalized piezocone tip resistance and pore pres
L’Heureux J-S 2019 Halden research site: geotechnical sure parameters. J. Geotech. Geoenv. Eng 134 (11)
characterization of a post glacial silt AIMS Geosciences 1569–1586.
5 184–234. Senneset K, Sandven R & Janbu N. 1989. Evaluation of
Cabal K & Robertson P 2014 Accuracy and Repeatability soil parameters from piezocone tests. Transp Research
of CPT Sleeve Friction Measurements Proc. CPT’14 Record 1235, Nat Acad Press, Washington D.C 24–37
Las Vegas 271–79 Tigglemann L & Beukema H 2008 Sounding Ring Investi
gation Proc. ISC-3 Taiwan 757–62
635
© 2022 the Author(s), ISBN 978-1-032-31259-0
Evaluation of shear wave velocity profiles in alluvial and deltaic soils using
a CPT database
Subterra, Ecuador
ABSTRACT: A seismic piezocone SCPTu database from Guayaquil (EC), consisting of Holocene alluvial
and estuarine deltaic soil deposits, is evaluated in this document through Vs mapping when comparing more
than 600 CPTu profiles and 45 seismic downhole testing locations. Shear wave velocity downhole measure
ments were used to adjust CPT-based Vs estimates and compared with cone resistance values. A suitable fit is
observed between Robertson (2009a) calculation and downhole measurements, but it tends to over-estimate
by 20-30% in soft clayey soils. An age correction factor as a function of normalized su/σ’v ratio has been
determined for clayey soils while a Vs-qt correlation is provided for clays and all soils. A Vs30 map has been
developed based on proposed shear-wave velocity calibration.
1 INTRODUCTION practice was to execute geophysical exploration (e.g.,

multichannel analysis of surface wave and microtre
The alluvial plain and estuarine deltaic deposit of mors array measurement). However, SCPTu has also
Guayaquil, located on the Ecuadorian coast, correspond been performed in the last few years, thus, it is con
geologically to the Holocene age. The stratigraphy usu venient to analyze an updated database that includes
ally includes layers of soft to firm clayey soils interbed these CPTs and downhole measurements.
ded with loose sands susceptible to liquefaction in the Therefore, this paper synthesizes an existing Guaya
top 30 meters. It is common to find organic soil, peat quil metro area soundings database with emphasis on
and sensitive clay in the estuarine deltaic area. Neigh CPT and downhole tests, carried out in an effective
boring the city, stands the Chongón-Colonche Moun area of 450 km2. By comparing existing CPT-based Vs
tain Range that favors the existence of colluvial soils in estimates with downhole measurements or correlating
the north and west area, so there are also soil profiles cone tip resistance with shear-wave velocity, it is pos
with high impedance contrast. sible to develop equations that allow estimating Vs
In most of these special cases of geotechnical values more accurately than with conventional estima
behavior, local standards recommend site-specific tions, while capturing local behavior of Guayaquil allu
response analysis (SRA), where measurement of vial and estuarine deltaic soil deposits. This analysis is
shear-wave velocity (Vs) profiles results relevant. In appropriate because in geotechnical research of uncon
addition, Vs30 criterion is used for site seismic classi ventional soils, it is always advisable to calibrate the
fication, thus Vs profile constitutes an important geotechnical parameters estimated from the CPT, with
input for most Guayaquil construction projects. measurements obtained in other in situ and laboratory
In 2014, Vera presented a detailed geotechnical tests (Mayne, 2007b; Robertson, 2012).
characterization of the Guayaquil soils, studying their Therefore, this paper proposes analyze: 1)
static and dynamic response. In that research, a correlation between downhole Vs measured values
a geotechnical zoning map was implemented, in and those estimated according to Robertson (2009a),
which estuarine deltaic geotechnical zones and allu as a function of normalized undrained shear strength
vial zones were identified, according to the geological (su ratio), 2) a Vs - qt correlation for clays and 3) a sim
context and the geotechnical framework. The data plified geotechnical zoning of Guayaquil and the sur
base used in that study, includes mainly geotechnical rounding area through Vs30 using Vs calibrated values.
information from boreholes, laboratory tests, geo This research goes in the same line of the theoret
physics estimates and CPT measurements in at least ical framework proposed by Andrus et al. (2007),
14 sites locations. The applied methodology was where an age correction factor was encountered.
plausible, because until recent, the common practice Besides, follows the sense of previous investigations,
in Guayaquil for geotechnical research consisted in which correlations have proposed between CPT
mainly of boreholes, and for measuring Vs, the usual measurements and Vs in different types of soils
DOI: 10.1201/9781003308829-92
636
(Robertson & Campanella, 1983; Baldi et al., 1989; 27.7 m and σ = 12.0 m). In addition, 63% of soundings
Mayne & Rix, 1995; Hegazy & Mayne, 1995; Robert have a depth greater than 24 m, while 83% have
son, 2009a). a depth greater than 16.5 m.
With the obtained results, it will be possible to Geotechnical zoning has been evaluated consider
have a more accurate and cost-efficient tool for seis ing several geotechnical parameters from such data
mic classification and site response analysis, than base (e.g., average Ic and undrained shear strength
traditional geophysical methods. every 5 meters, liquefaction potential), in order to
have a complete and true understanding of Guayaquil
soils. The work of Paredes (2020) can be reviewed for
2 DATABASE CHARACTERISTICS a thorough analysis of Guayaquil geotechnical zoning.
Considering only the 615 CPTs, average total depth
All CPTu soundings conducted by Subterra between is 23.5 m, with a standard deviation of 8.2 m (95% of
2012 and 2020 were registered into an ArcGIS map. CPTs underwent refusal conditions). The rigid stratum
For each test, the following characteristics were noted: has been detected at an average depth of 22 m.
coordinates, year, total depth, rigid stratum depth,
among others. Then, soundings from several geotech
nical companies were compiled and a geotechnical 3 SHEAR-WAVE VELOCITY EVALUATION
exploration campaign was implemented based exclu
sively on CPTu tests distributed in different sectors of In situ testing was executed with a standard 10 cm2
Guayaquil, especially towards the northeast and south cone, compression type, and 150 cm2 sleeve friction.
of the city, where information available was scarce at The pore pressure filter (u2 position) is made out of
the moment. bronze, and saturated with silicone oil. The piezo
The total database consisted of 1432 soundings cone is pushed at 2 cm/s with a 15-ton Pagani TG-63
which includes 615 (42.9%) CPTs and 817 (57.1%) equipment recording measurements every cm.
boreholes. From the 615 CPTs, 38 were exclusively A seismic module with two triaxial series of geo
carried out for this study. Downhole Vs profiles have phones is adapted behind the cone for recording
been measured in 45 CPTs (SCPTu) at 36 different shear-wave time arrival every meter.
sites, as illustrated in Figure 1. Regarding timeline, An initial database filter of shear-wave measure
77% of the soundings were carried out after year 2010, ments was applied to discard signals with poor amp
while 91% after year 2000. Explored depths range litude or profiles located in unfavorable testing
between 2.3 m (shallow rock) and 145.2 m (avg. = conditions like partially flooded ground surface,
where contact area between seismic source and
ground surface deemed inappropriate for obtaining
a clear Vs profile. Thus, 733 downhole seismic read
ings between 2 and 25 m deep were compiled for
this study. Shear-wave velocity values ranged
between 50 and 390 m/s (avg. 120 m/s).
3.1 Calibration factor

Seismic CPT has been performed in Guayaquil and the
surrounding area since 2014 and it was notable
throughout the years that Robertson (2009a) Vs esti
mate, which is based on vast experience of normalized
SBTn chart considering uncemented Holoceneand
Pleistocene-age soils, tends to over-estimate shear-
wave velocity by 20-30% in soft clayey soils, both in
estuarine and alluvial zones. However, a good overall
trend in the soil profile was always observed if
a correction factor was applied to the estimate. As an
example, Figure 2a to Figure 2c illustrate several
Vs downhole profiles (red dots) and corresponding
CPT-based Vs.
There appears to be some scattering when analyz
ing CPT-Vs pairs, primarily due to the difference
between depth interval measurements where read
ings are taken (e.g., 1 cm for CPT and 1 m for Vs).
Thus, data has been filtered based on standard devi
ation of CPT index Ic (σIc), limiting random variabil
Figure 1. Guayaquil database sounding locations. Rock ity. Also, shallow downhole readings may lack some
outcrop (Benítez & Vera, 2006). quality at times due to short time arrivals and thus,
637
Figure 2b. SBTn (Robertson), su-rat and Vs profiles, in allu
vial zone of Samborondón and Guayaquil.
Figure 2a. SBTn (Robertson), su-rat, Vs downhole and esti

mated profiles, from two Guayaquil soundings located in
estuarine zone.
greater percentage errors. Reason why only data

pairs at depths greater than 5 m have been con
sidered in every filtered trend analysis presented in
this paper.
A calibration coefficient was calculated for each
one-meter avg. CPT-based Vs and downhole Vs
measurement. Given a strong relation was initially
observed with estimated normalized shear strength
ratio (su-rat = su/σ’v), measurements were filtered by
σIc < 0.10, Ic > 2.6 and su-rat < 2, the latter two cor
responding to clayey soils. Undrained shear strength
was calculated from Fr-based Nkt factor (Robertson
2012).
Figure 3 shows a simplified regression analysis
performed on 324 pairs to calculate correction factor Figure 2c. SBTn (Robertson), su-rat, Vs downhole and esti
ðμvs Þ in clays and transitional soils given by: mated profile, from Durán sounding.
638
deposits are generally homogeneous compared to
sandy layers which are rather intercalated throughout
the profile, being apparent a better trend when filter
ing for Ic > 2.6. Considering both datasets, a general
For sandy soils with Ic < 2.6, μvs ranged between 0.8 Vs-qt trend is proposed in the following expression:
and 1.2, with an average value of 0.98. More research
is necessary in order to evaluate an accurate calibration
factor in these soils. Thus, shown database correction
from Figure 3 does not apply for sandy soils.
Robertson (2009a) database included Holocene and where, cone resistance qt is in MPa, Vs in m/s and
Pleistocene soils while 25% higher estimates have regression variables are
been reported for Pleistocene soils. Andrus et al.,
(2007) even proposed a correction factor as function of
deposit age. Therefore, it would be reasonable for
Guayaquil soils deposits, where shallow layers date
mainly from the Holocene, that estimated values are alldata:
slightly higher than those measured. However, for
deeper soil layers with a higher degree of consolidation
(older soils with signs of cementation), Robertson’s A similar equation has been presented by Mayne and
estimate fits better, and in some cases an underestima Rix (1995) from 31 natural clay sites. Clays varied
tion is also observed. This apparent underestimation from intact to fissured materials with wide ranges of
could be related to cementation that tends to increase plasticity, sensitivity and OCR. Variables provided in
Vs (Schneider, et al 2004), while resistance to penetra this paper, corresponding to Equation (2) are a =
tion captures to a lesser extent the effects of age and 133.06, b = 0.63.
cementation (Robertson, 2012). The correction factor The filtered proposed Vs - qt correlation shows an
applies best for a ratio su/σv´ between 0.2 and 1.5, excellent correspondence in clays, especially for
yielding a coefficient that varies between 0.6 and 1.18. qt < 2 MPa. When including all data, at qt > 2 MPa,
Vs-qt trend is much lower than that filtered.
Figure 3. Vs calibration factor as a function of normalized Figure 4. Correlation between shear-wave velocity and
cone tip resistance in Guayaquil.
3.2 Vs-qt regression 3.3 Vs30 mapping

Two statistical regression analyzes were performed On an effort to consolidate previous findings, a Vs30
to further evaluate the existing database: one with map has been developed for Guayaquil based on 615
733 CPT-Vs pairs (all data) including different geo CPTs. It is plausible to apply Equation (2) for clay in
materials, and another with 388 resulting from filter soft homogeneous clayey soils with Ic > 2.4 and qt < 2
ing by Ic > 2.6 and qt < 8 MPa with low MPa, while correction factor from Equation (1),
heterogeneity average intervals (σIc < 0.15), at limited by 1.18, for the rest of clayey soils with
depths greater than 5 m. The observed trend between Ic > 2.4.
Vs and qt for Guayaquil is presented in Figure 4. With regards to Vs values for sandy material,
As expected, a higher R2 value is obtained from since the database is relatively limited, it would
filtered data by avoiding values with considerable seem appropriate to, conservatively, correct Robert
lenses across one-meter intervals. Guayaquil clay son (2009a) equation by a coefficient μvs of 0.85 for
639
contractive shallow sands and 0.95 for dilating correlation is proposed for Guayaquil soft homoge
sands, although the latter could certainly be 20% neous clays with qt < 2 MPa.
higher. Yet a correction value of 0.90 for sandy In Guayaquil, local standards require SRA in most
material Ic < 2.4 was employed. Figure 5 shows projects where soil profiles with special features are
Guayaquil Vs30 map by applying iterative finite dif encountered. One of the most important input param
ference interpolation methodology in ArcGIS. eters in this type of analysis is Vs profile. Tradition
Given that 25% and 66% of CPTs underwent ally, geophysics is used to determine this seismic
refusal conditions at more than 30 and 20 meters parameter. However, the accumulated experience and
deep, respectively, Vs estimate was extrapolated proposed equations will allow using a more efficient
from those estimated in deepest layers, provided that and accurate methodology when downhole testing or
refusal conditions were effectively present. So, Vs30 geophysical methods are not available.
values are rather conservative in these cases.
REFERENCES
4 CONCLUSIONS
Andrus, R., Mohanan, N., Piratheepan, P., Ellis, B., &
Holzer, T. (2007). (s.f.). Predicting shear-wave velocity
The presence of geotechnical profiles with special
from cone penetration resistance. Proceedings of Fourth
characteristics is very common in the alluvial plain International Conference on Earthquake Geotechnical
and estuarine deltaic deposit of Guayaquil. For Engineering. Thessaloniki, Greece, 25–28, June 2007.
a better understanding of the variation of geotech Paper No. 1454.
nical parameters, a database has been compiled with Baldi, G., Bellotti, R., Ghionna, V., Jamiolkowski, M. &
geotechnical information obtained in recent years, Lo Presti, D. (1989). Modulus of sands from CPTs and
consisting of 1432 soundings, with an average depth DMTs, Proc. 12th International Conference on Soil
of 27.7 m. This database has 615 CPTs, from which Mechanics and Foundation Engineering, Vol. 1,
45 Vs downhole profile have been obtained while the 165–170.
Benítez, S., & Vera, X. (2006). Estudio Geológico de la
relationship with CPT has been analyzed, in order to
ciudad de la ciudad de Guayaquil. Informe final de
consolidate results through a map of Vs30, and for its Investigación y Estudio del comportamiento dinámico
extended use in the construction projects of the city. del subsuelo. IIFIUC, Universidad Católica Santiago de
A Vs correction factor for calibrating Robertson Guayaquil.
(2009a) estimate has been obtained through a direct Hegazy, Y. & Mayne, P. (1995). Statistical correlations
proportionality with su ratio in soils with index Ic between VS and cone penetration data for different soil
(SBTn) > 2.4. The proposed correction ranges between types, Proc., International Symposium on Cone Penetra
0.6 and 1.18, for su ratio < 1.5. Furthermore, a Vs-qt tion Testing, CPT ’95, Linkoping, Sweden, 2, Swedish
Geotechnical Society, 173–178.
Mayne, P., & Rix, G. (1995). Correlations Between Shear
Wave Velocity and Cone Tip Resistance in Clays. Soils
& Foundations Vol. 35, No. 2, 107–110.
Mayne, P. (2007b). Invited Overview Paper: In-situ test cali
brations for evaluating soil parameters, Characterization
& Engineering Properties of Natural Soils, Vol. 3 (Proc.
IS-Singapore), Taylor & Francis Group, London:
1602–1652.
Paredes, J. (2020). Evaluación de parámetros geotécnicos
de los depósitos de suelos ubicados en la llanura aluvial
y en el complejo deltaico estuarino de Guayaquil. Mas
ter’s Thesis. Escuela Superior Politécnica del Litoral
(ESPOL), Guayaquil, Ecuador.
Robertson, P., & Campanella, R. (1983). Interpretation of
Cone Penetration Tests. Part I: sand, Canadian Geotech
nical Journal, 20(4), 477–486.
Robertson, P. (2009a). Interpretation of cone penetration
tests – a unified approach. Canadian Geotechnical Jour
nal 46, 1337–1355.
Robertson, P. (2012). Interpretation of in situ tests - some
insights. J.K. Mitchell lecture, Proceedings of ISC’4,
(pp. 3–24). Recife, Brazil.
Schneider, J., McGillivray, A., & Mayne, P. (2004). Evalu
ation of SCPTu intra-correlations at sand sites in the
Lower Mississippi River valley, USA, Geotechnical &
Geophysical Site Characterization, Vol. 1, (Proc. ISC-2,
Porto), Millpress, Rotterdam, 1003–1010.
Vera, X. (2014). Seismic Response of a Soft, High Plasti
city, Diatomaceous Naturally Cemented Clay Deposit.
Figure 5. Guayaquil Vs30 Map. Doctoral Thesis. University of California, Berkeley.
640
© 2022 the Author(s), ISBN 978-1-032-31259-0
Upscaling 1 500 000 synthetic CPTs to voxel CPT models of offshore sites
J. Peuchen & W. van Kesteren
V. Vandeweijer & S. Carpentier

TNO, Utrecht, The Netherlands
F. van Erp
RVO – Netherlands Enterprise Agency, Utrecht, The Netherlands
ABSTRACT: In 2020, the Dutch government published 1 500 000 synthetic CPT profiles for use in develop
ment of the Hollandse Kust (west) Wind Farm Zone, offshore Netherlands. The scale of this approach was novel
at that time and possibly first-ever. The synthetic CPT profiles were derived from a training data set of 122 actual
CPTs and ultra-high-resolution (UHR) seismic reflection traces, using machine learning by a convolutional neural
network. The synthetic CPT profiles were limited to positions along the 162 UHR survey track lines (2D) and
were limited to cone resistance to a depth of 50 m below seafloor. The UHR track lines were spaced at about
400 m. This paper explores upscaling the synthetic CPT approach to voxel (3D) models and adding (S)CPT
based parameters such as shear modulus at small strain Gmax. Future added-value is expected from post-2020
improvements seen in seismic reflection data resolution, attribute extraction and neural network architecture.
1 INTRODUCTION synthetic CPT approach. The status quo is presented

by an example in the public domain (www.offshore
Reducing ground risk is important for the development wind.nl): the Hollandse Kust (west) Wind Farm Zone,
of an offshore wind farm. This requires understanding HKW WFZ (Figure 1), offshore Netherlands (Fugro
of the geological and geotechnical conditions to 2020a and 2020b). HKW WFZ data acquisition, data
depths in the order of 30 m to 100 m below seafloor analysis and advice were largely completed in 2019
(BSF), depending of type of support structure for the and 2020. The 1 500 000 synthetic CPT profiles were
wind turbines. generated at no schedule impact. DNV GL (2020)
The understanding of ground risk is typically sees this cutting edge development as ‘a huge step for
expressed by a ground model or multiple ground ward in terms of project area overview with respect to
models (ISO 2021). These models typically rely on geotechnical site conditions and also as a valuable
integrated interpretation of geological information, tool to improve and understand the correlation
geophysical (UHR and UUHR multichannel seismic between future geological, geophysical and geotech
reflection) data and geotechnical data (particularly nical investigations.’
cone penetration tests, CPTs.
Since the 1990’s, there has been increasing focus on
deriving geotechnical properties directly from geo 2 HOLLANDSE KUST (WEST) WIND FARM
physical data using methodologies developed in the ZONE
oil and gas industry (e.g. Nauroy et al. 1998). More
recently, trials were made with synthetic CPTs and The site for the HKW WFZ is located approximately
geo technical properties generated by interpolating 53 km from the Dutch coast, covering an area of
CPT data between investigated locations (Forsberg roughly 176 km2 in water depths ranging from
et al., 2017) and using statistical methods and multi- 18 m to 36 m LAT.
attribute regression through an artificial neural net
work (Sauvin et al. 2019). The general approach is Ground model input mainly included:
covered by ISO 19901-10 Marine Geophysical Inves – Geological information;
tigations (ISO 2021). – Geophysical data: multibeam echosounder, side
This paper describes the status quo for 2020 and scan sonar, magnetometer, sub-bottom profiler
explores future opportunities for upscaling the and 2D-UHR single channel and multi-channel
DOI: 10.1201/9781003308829-93
641
The acquired geophysical and geotechnical data
were integrated into a traditional quasi-3D ground
model, with seismic reflections tied to geotechnical
boundaries derived from seafloor CPT and borehole
information (Figure 2). The ground model comprised
eight geological soil units, each having a distinct
seismic character and spatial distribution.
Figure 2 provides cross sections that illustrate
how the traditional ground model was enhanced by 1
500 000 synthetic CPT (cone resistance) profiles and
associated error predictions to 50 m below seafloor
(Carpentier et al. 2021). The presented cross section
has a length of 11 500 m with 4 actual CPTs and
about 9000 synthetic CPTs to a depth of 50 m.
The HKW WFZ synthetic CPT profiles were
derived from a training data set of 122 actual CPTs
and ultra-high-resolution seismic reflection traces,
using machine learning by a convolutional neural net
work. The synthetic CPT profiles were limited to
positions along the 162 UHR survey track lines (2D).
The UHR track lines were spaced at about 400 m.
Figure 3 shows example checks on predictions. In
general, the predicted and measured net cone resist
Figure 1. Location of the future Hollandse Kust (west) ance values showed reasonably good agreement, par
Wind Farm Zone. ticularly in the upper 20 m BSF. Below 20 m,
prediction was more trend-type. In addition, a trend-
type prediction also applies to transitional and strongly
seismic reflection data. The data were acquired layered (<1 metre scale) soil. It can be concluded that
according to a draft version of ISO (2021); the prediction quality for the synthetic CPTs is such
– Geotechnical data acquired from 57 boreholes that added value can be derived to enhance the gen
with sampling and cone penetration testing (CPT) eral ground model. The HKW WFZ prediction quality
to a maximum depth of 90 m below seafloor is inadequate for geotechnical design.
(BSF), 122 seafloor CPTs to a maximum depth of Reasons for the observed trend-type predictions
56 m BSF, 30 seafloor seismic cone penetration include data conditioning, resolution of 2D-UHR
test to maximum depth of 56 m BSF, and labora seismic reflection data and limitations in refinement
tory testing. The data were acquired according to of the interpreted geological units. Data conditioning
ISO (2014). was applied by down sampling the CPT data to align
Figure 2a. Example of integrated interpretation of 2D-UHR multi-channel seismic line, aligning geophysical horizon inter
pretation to identified geotechnical boundaries from seafloor CPT and borehole data.
642
Figure 2b. Predictions of cone resistance (qn*) from 2D-UHR seismics (colors ranging from orange to purple) plotted
together with measured cone resistance (black traces) and sleeve friction (red traces). See Figure 2a for comparison.
with vertical resolution of the 2D-UHR seismic data

(i.e. sample rate of 0.1 m for prediction input versus
0.02 m as measured). Down sampling reduces net
cone resistance effects of soil layering. The reso
lution of the 2D-UHR seismic data decreases with
depth. This also affects the ability to identify add
itional geological units. This effect becomes more
pronounced below approximately 20 m BSF.
As expected, training of the convolutional neural
network showed decreasing prediction accuracy with
increasing lateral distance between the seafloor CPT
location and the nearest seismic trace along the 2D
UHR line. This is particularly significant where the
correlation distance for spatial soil variability is less
than the distance between the seafloor CPT location
seismic trace selected for training.
Figures 3 and 4 illustrate prediction quality by
means of a quality indicator per geological unit. It
can be seen that the lower limits of the quality indi
cator can provide statistical values for qn � that fall
outside credible ranges for these specific soils.
3 UPSCALING
3.1 Opportunities for future added value

The following opportunities for upscaling are
discussed:
– Enhanced geophysical interpretation
– Improved data pairing for training Figure 3. Comparison of actual net cone resistance (black
– Impact of technology developments line) versus synthetic net cone resistance (red line). The red
– Predictions for multiple parameters halo represents the interval in which predictions are likely
– Voxel model by geo-statistics to fall (5th and 95th percentiles of the error misfit). Other
– Voxel model by 3D geophysics colour infills indicate geological units.
643
closer the better. Particularly, logging of spatial
trajectories of deep CPTs should be considered,
compared to conventional assumptions for a vertical
CPT. Seismic reflection survey should consider spe
cific positioning of the source(s) and specific position
ing of multiple points along the streamer(s). This is
particularly important for situations where correlation
distance for soil spatial variability is limited.
Mitigation options for pairing of spatially distant
data can include point-specific matching checks and
adjustments, using marker points in the profiles.
3.4 Impact of technology developments

For marine geophysics, significant technology
developments are taking place, with high potential
for added value in de-risking for ground condi
tions. For geophysical data acquisition, these
include improvements in acoustic sources and
streamer control. For processing methodology, not
able improvements include de-ghosting algorithms,
Figure 4. Statistical values for qn � that fall outside credible multiple removal algorithms and velocity models).
ranges (green box). These improvements will result in opportunities
for very high resolution data and high quality seis
mic attributes.
The opportunities can be considered individually. Fast developments are taking place in neural net
However, combinations are expected to lead to work architecture. Technology developments for
a step change in added value. marine soil investigation (CPTs, other in situ testing
The following sections discusses potential oppor and laboratory testing) are expected to be ‘incremen
tunities for a time window of 2020 to 2025. tal’, i.e. at a slow pace compared to marine geophys
ics and neural network architecture.
3.2 Enhanced geophysical interpretation
Figure 2b Indicates that the synthetic CPTs can 3.5 Predictions for multiple parameters
allow for further refinement of the geological model. The HKW WFZ choice for synthetic profiles for
This, in turn, helps steering future, turbine-specific cone resistance is obvious: input cone resistance
geotechnical data acquisition, increasing safety and data are accurate (Peuchen & Terwindt 2015) and
cost-efficiency. typically show good correlation with geological
For example, light green colours can be seen units and other geotechnical parameters. Shear
between approximately 40 m and 50 m LAT. These modulus at small strain Gmax is another candidate
strength insights (i.e. trends in net cone resistance) for synthetic predictions. Comments for Gmax are
may justify refining the geological model at this depth, as follows:
by assigning a new geological soil unit to encompass
these geotechnical conditions. It should be noted that – Gmax is an important parameter for geotechnical
this potential additional geological soil unit could not design of monopiles used for support of offshore
be identified from ‘regular’ seismic data (e.g. seismic wind turbines;
amplitudes) alone. The example also indicates that – Gmax is a low-strain soil parameter. Seismic reflec
synthetic CPT results (1) can give further insight into tion data are also low-strain and good predictive
potential soil heterogeneity, (2) may aid in identifying capability would seem obvious;
areas of higher geotechnical uncertainty and (3) can – Good predictive capability may be impeded by
identify where geotechnical conditions deviate from higher uncertainties (compared to CPT cone
regional trends. resistance) for actual Gmax profiles (Parasie et al.
2022) required for training a neural network.
Actual Gmax values are typically derived from
3.3 Improved data pairing for training seismic cone penetration tests. These tests rely on
Quality of input data is important for success of time and distance measurements. Data processing
machine learning. requires estimates of input soil density. Premises
Attention should be given to accuracy of spatial include theories on acoustic wave propagation
positions of paired data derived for actual CPT and assumptions about heterogenous soil behav
locations and seismic reflection data points: the ing as an isotropic elastic medium.
644
3.6 Voxel model by geo-statistics marine seismic reflection data using a neural network
Commonly (2022), visualisation of a ground model approach. In Second EAGE Workshop on Machine
Learning, 8-9 March 2021.
is by means of 2D cross sections and 2D charts, i.e. DNV GL (2020). Hollandse Kust (west) Wind Farm Zone,
in pixels. In some cases, 3D visualisation (e.g. Review of Geotechnical Report, Synthetic CPT Profiles
Figure 5) is implemented, allowing interpretation in (Document No. CR-SC-DNVGL-SE-0190-05500-0
terms of voxels. Voxel data are typically generated Synthetic CPT Profiles Rev.0, date 6 November 2020 to
by geo-statistics on a soil-unit basis. Synthetic CPTs Rijksdienst voor Ondernemend Nederland).
can enhance this approach. Forsberg, C.F., Lunne, T., Vanneste, M., James, L.,
An important consideration is the volume of data. Tjelta, T.I., Barwise, A., Duffy, C. (2017) Synthetic
Can the information be made available within tight CPTs from Intelligent Ground Models based on the Inte
schedules required for energy transition? Can it be gration of Geology, Geotechnics and Geophysics as
a Tool for Conceptual Foundation Design and Soil
easily assessed for decision making? Investigation Planning. Offshore Site Investigation and
Geotechnics 2017 conference proceedings – smarter
3.7 Voxel model by 3D geophysics solutions for future offshore developments, London, UK,
12-14 September 2017 (Volume 1, pp. 1254–1259).
ISO 19901-10 Marine Geophysical Investigations Society of Underwater Technology.
(ISO, 2021) covers acquisition of 3D UHR seismic Fugro. (2020a). Geological Ground Model – Hollandse
reflection data. Currently (2022), acquisition of these Kust (west) Wind Farm Zone – Dutch Sector North Sea
data in the foundation zone (upper 100 m BSF) is (Document No. P904711/06, issue 3, dated
performed only for occasional offshore wind sites, 12 May 2020 to Rijksdienst voor Ondernemend
Nederland).
with some indications of growth in applications of Fugro (2020b). Geotechnical Report – Synthetic CPT pro
this technology. The availability of 3D UHR seismic files - Hollandse Kust (west) Wind Farm Zone – Dutch
reflection data in combination with generation of Sector North Sea (Document No. P904711/08, issue 3,
synthetic geotechnical parameters has the potential dated 14 October 2020 to Rijksdienst voor Onderne
for a step-change in voxel ground models and associ mend Nederland).
ated added value to offshore developments. International Organization for Standardization. (2014). Pet
roleum and natural gas industries – specific require
ments for offshore structures – part 8: marine soil
investigations. (ISO 19901-8:2014). https://www.iso.
org/standard/61145.html.
International Organization for Standardization. (2021). Pet
roleum and natural gas industries - specific require
ments for offshore structures – part 10: marine
geophysical investigations. (ISO 19901-10:2021).
https://www.iso.org/standard/77017.html.
Leapfrog. (2021).[Software]. Seequent. https://seequent.
com/products-solutions/leapfrog-geo/.
Nauroy, J.F., Colliat, J.L., Puech, A., Kervadec, J.P. and
Meunier, J. (1998). GEOSIS: integrated approach of
geotechnical and seismic data for offshore site investiga
tions. In P.K. Robertson & P.W. Mayne (eds.). Geotech
nical site characterization: proceedings of the first
Figure 5. Excerpt of 3D HKW WFZ model in Leapfrog international conference on site characterization –
(Seequent, 2021) software. ISC’98, Atlanta, Georgia, USA, 19-22 April 1998
(Vol.1, pp. 497–502). Balkema.
Parasie, N., Franken, T and Peuchen J. (2022). Assessment
of seismic cone penetration testing for small strain
ACKNOWLEDGEMENTS
modulus. In 5th International Symposium on Cone Pene
tration Testing, CPT ’22, Bologna, Italy, 8-10 June 2022.
We acknowledge the valuable assistance of and fruit Peuchen, J. and Terwindt, J. (2015). Measurement uncer
ful discussions with the Netherlands Enterprise tainty of offshore Cone Penetration Tests. In Meyer, V.
Agency (RVO) during the HKW WFZ project. The (ed.) Frontiers in Offshore Geotechnics III: proceed
primary data used and presented in this paper can be ings of the Third International Symposium on Fron
retrieved through www.offshorewind.rvo.nl. tiers in Offshore Geotechnics (ISFOG 2015), Oslo,
Norway, 10-12 June 2015. Boca Raton: CRC Press,
pp. 1209–1214.
REFERENCES Sauvin, G., Vanneste, M., Vardy, M.E., Klinkvort, R.T. and
Forsberg, C.F. (2019). Machine Learning and Quantita
Carpentier, S., Peuchen, J., Paap, B., Boullenger, B., tive Ground Models for Improving Offshore Wind Site
Meijninger, B., Vandeweijer, V., Van Kesteren, W. & Characterization. Offshore Technology Conference,
Van Erp, F. (2021). Generating synthetic CPTs from Houston, USA, 6-9 May 2019, OTC Paper 29351.
645
© 2022 the Author(s), ISBN 978-1-032-31259-0
Automated CPT interpretation with a Convolutional Neural Network

M. Pippi, R. Vink & J. Haasnoot
CEMS Crux Engineering MicroServices BV, Amsterdam, The Netherlands
S. Bersan
CRUX Engineering, Amsterdam, The Netherlands
ABSTRACT: Cone Penetration Tests are widely used in the Netherlands, due to their ease of execution in
the Dutch delta deposits and their relatively low cost. As the amount of performed CPTs increases, an auto
mated soil interpretation becomes more and more relevant. Attempts to automate soil classification have been
done in the past, but the empirical formulas commonly used do not always provide a satisfactory interpret
ation for engineering purposes. Besides that the soil type is often not interpreted correctly, there is also the
problem that the classification is provided for each measurement (every 2 centimeters) and no strategy is pro
vided to aggregate those tiny layers. This paper shows how a data driven approach can yield better results
than the traditional empirical methods. A machine learning model is presented which is based on
a Convolutional Neural Network. The Neural Network has been trained on 1800 pairs of CPTs and boreholes
that met the condition of being less than 6 meters apart. An algorithm based on the theory of signals is applied
to the classification given by the model to group measurements into soil layers. The paper explains the theory
behind the model, shows a comparison with the soil classification given by the Robertson correlation and
shows how the model can be used in the geotechnical practice.
1 THE DATA in the format they asked for. However, the result was
that there was still a number of formats, which were
1.1 Open source data hard to be interchanged thus being a hindering for
a sustainable storage of these data.
In the past decades the Netherlands have been able
The variation in formats lead to waste of time and
to perform and store a large amount of soil investiga
the related costs are estimated to amount to at least
tion. Part of this data have been made available to
NLG 600000 or Euro ≈ 270000 a year (price level
everyone by the no-profit organization TNO through
1999).
the website DINOloket (TNO, 2021).
This situation induced the development of
Since 2020, the organization BRO (Basis Regis
a standard format: the GEF (Geotechnical Exchange
tratie Ondergrond) is responsible for the manage
Format). The characteristic and requirements of the
ment of the National Data Repository. Every type of
format are specified in the document “Geotechnical
soil investigation that is performed for a project
Exchange Format for CPT-data” (CUR, 2006).
financed by the public administration must be regis
tered and stored in the national database.
The reasons behind this approach is given by the 1.3 Parsing of the data
fact that making soil investigation available to every
In order to be able to read and use the data in
one will improve the estimation of the subsurface and
a program, it has first to be parsed. For this task an
therefore reduce the risks related to incorrect model-
open source Python library, pygef, has been developed.
ling of the soil. Figure 1 give an overview of the CPTs
The library is able to read gef formats of CPTs and
that are available on DINOloket (these are indicated as
boreholes and store their content into a DataFrame
brown dots).
object. The library provides also a traditional way to
classify a CPT: via the Robertson (Robertson, 2010) or
1.2 Data format Been & Jefferies (Been, K., & Jefferies, M. G. 1992)
Before the year 2000 the output of the cone penetra classification methods. The Roberson classification
tion test was available in a digital form according to method provided by the Python library has been used
a customer specific format. This seemed to be an for the validation metrics of the model presented in this
ideal situation, as the customers would get the output paper.
DOI: 10.1201/9781003308829-94
646
2.3 Model architecture
The model is based on a Neural Network architec
ture with convolutional layers that can apply feature
extraction on the input signal (Sassi, A. et al., 2019).
The model is enhanced with location-based embed
ding, in this way the model could learn its own loca
tion embedding and could learn the probabilities of
soil type conditional on a certain location. Further
more, most of the bore-hole data show that layers
consist of multiple soil components in variable per
centages. Therefore we should predict the total soil
distribution per layer.
In Figure 2 an example of the result of the model
is shown. The model predicts the soil distribution
over the depth. Qualitatively the predictions of the
model seem very reasonable and align with
a geotechnical mapping.
Figure 1. Map of CPTs available in Dinoloket (TNO,

2021).
2 THE CEMS MODEL
2.1 Data-driven iteration

A data-driven approach, in a country where so much
open source data is available, seems to be the best
way to proceed. In order to train the model roughly
49000 CPTs and 40000 boreholes were checked,
from which 1800 pairs met the condition of being
less than 6 meters apart. These have been used as
labeled data for the first model training. The model
is retrained periodically whenever new data are
Figure 2. Example result of the CPT interpretation model.
available that meet the condition of being not more
than 6 metres apart.
2.4 Location embedding
2.2 Decision factors
From the 1800 pairs, 48 location clusters were cre
Machine learning is often beneficial with high ated by applying K-means algorithm on the location
dimensional data. In this case, the amount of data data.
dimensions is relatively low. When training powerful In Figure 3 the locations of the clusters are indi
ML models, like Neural Networks (Multi Layer Per cated on the map of the Netherlands. The colors rep
ceptron architecture) or Gradient Boosting Trees, on resent a similarity measure between the clusters
this data, we see comparable results with each other. based on the cosine similarity. Clusters close to each
The results are quite reasonable, but they do not other on the color scale are likely to have similar soil
reflect the decision factor of a geotechnical engineer. distributions.
An essential part of the decision making is based on Besides comparing the clusters by similarity,
the location where the CPT is taken. And when a CPT, we can also run inference for the embedding
for instance, is taken at sea, they can tell by the curve by nullifying the features. Intuitively this can
of the line that a certain layer consists of seashells. be regarded as the soil classification you would
If we want a model to be able to take the same expect if you only know the location and not
decision factors into account as a human does, the considering the information in the CPT.
model needs to be able to make decision based on Figure 4 shows the biases that can be applied per
the same information. location.
647
3.2 Comparison with Robertson
Comparing the outputs of the model with Robert
son requires some translation. Robertson does not
predict a soil distribution, but assigns soil types,
e.g. ‘Silt mixtures: clayey silt & silty clay’ and
‘Sand mixtures: silty sand & sandy silt’. To be
able to make a comparison these classification
were transformed into a soil class {Gravel, Sand,
Loam, Clay, Peat}.
3.3 A simple validation

A simple way of doing a validation is by looking at
precision and recall scores (F1 score) by transforming
the probability distributions to main soil classes. In
Table 1 the F1 scores are shown, where 0 is a bad pre
diction and 1 would be a perfect prediction. The
column “Support” indicates the number of soil layers
with that type of soil as the main type.
Figure 3. Location clusters. Table 1. F1 scores.
Soil KL-
type Support Robertson divergence Wasserstein
Gravel 3731 0.15 0.15 0.1

Sand 137998 0.86 0.83 0.85
Loam 0 0 0 0
Clay 91523 0.64 0.7 0.67
Peat 21398 0.37 0.57 0.74
This comparison is not accurate since a lot of infor

mation is lost if we reduce a distribution to a number
by taking the mode. If we look at the loam soil type,
we can observe that this validation does not consider
its prediction. Loam is almost always a subtype in
soil layers so it will never come out as the main soil,
making it seem that no model can predict this class.
3.4 A better validation: Error distribution

Since we are talking about probabilities, the error
Figure 4. Biases per location. distributions show the error margins. In the tables
below the Mean Absolute Error (MAE) per soil type
is reported. We consider the absolute errors on two
3 VALIDATION METRICS subsets of the test set.
3.1 Training metrics

Table 2. Precision based absolute errors (lower is better).
When training a machine learning model usually we
have “ground truth labels” being binary True or MAE MAE MAE
False, i.e. y ∈ (0,1). However, the soil labels are Soil type Robertson KL-divergence Wasserstein
never 0 or 1 but they are distributions. Every layer
contains different volumetric percentages of soil com Gravel 0.77 0.29 0.22
ponents, but the sum of all volumes adds up to 1. Sand 0.23 0.31 0.22
For this reason the metrics used to train the model Loam 0.93 0.05 0.06
are the Kullback-Leibler divergence (Van Erven, T., Clay 0.35 0.29 0.3
& Harremos, P., 2014) and the Wasserstein distance Peat 0.56 0.28 0.37
(Hou, L et al., 2016).
648
Table 3. Recall based absolute errors.
MAE MAE MAE

Soil type Robertson KL-divergence Wasserstein
Gravel 0.18 0.24 0.17
Sand 0.22 0.29 0.2
Loam 0.14 0.04 0.05
Clay 0.44 0.34 0.35 Figure 6. Two Gaussian signals with changepoint τ.
Peat 0.3 0.23 0.22
the likelihood of the whole signal coming from

one Gaussian.
A subset where the true distribution has non-zero This observation can be turned into an opti
probability for that class can intuitively be regarded mization problem. We search for the minimal
as recall. A subset where the models assign negative likelihood by adding new change points
a significant probability to that class P(y) > 0.025 τi for every point
can intuitively be regarded as precision. K in the signal. To prevent having change
points at every data point, we introduce
a penalty λ.
4 THE GROUPING ALGORITHM
4.1 The motivation

The output of the model is a prediction given for each
measurement of the CPT. Although this is probably
the most accurate representation of the soil stratig
raphy, it is impossible to use 1000+ layers in the trad
itional software, hence we need to group these tiny 4.3 Final result
layers into macro layers. This is done using an algo Applying the grouping algorithm to the predic
rithm that is based on considering the CPT measure tion of the model results in a manageable
ment a signal. amount of soil layers. By changing the penalty λ
we can choose to have more or less layers.
4.2 The theory A higher penalty will generate less layers,
a lower value will generate more layers, as illus
The measured cone resistance and friction ratio are trated in Figure 7.
signals and as such we can assume that they follow
a Gaussian distribution, as shown in Figure 6.
Figure 5. Gaussian signal.
The likelihood of this signal is determined by:
If we consider the signal in Figure 6, the likeli Figure 7. Influence of the penalty parameter on the discret
hood of two sub-signals coming from two Gaus ization into soil layers.Fig (a): penalty=2; Fig (b):
sians, separated by change point τ, is higher than penalty=1.
649
5 USAGE OF THE CEMS MODEL
5.1 Web API

The model can be accessed with a HTTPS call via
a web REST API, if using the language Python the
library requests could be used to make the call. In
this way the users can incorporate the prediction of
the model in their own workflow. The documenta
tion of the REST API can be found on the website
(CEMS, 2022). Please note that authentication
Figure 9. Example of input for a levee stability calculation.
is needed when using the model, visit the CEMS
website at https://cemsbv.nl/ to learn how to get
access.
of time when using it in an automated process for
geotechnical design is considerable.
5.2 Examples The model has been trained only on Dutch soil data.
As the soil interpretation is often the first step of Even though it would probably still give a reasonable
a geotechnical design, this building block can be prediction outside of the Netherlands, it cannot apply
used in many geotechnical applications. So far the any biases for the location. Future development could
CPT interpretation tool has been used in a number of include the expansion of the geographical limits of the
applications including: pile design, levee assessment model, by including CPTs and boreholes from other
and improvement, prediction of vibrations induced countries and the definition of location biases and clus
by the installation of sheet piles. ters based on expert knowledge.
In case of levees, the model is used to automatic
ally generate longitudinal lithological profiles along
the levee as the one shown in Figure 8. These pro REFERENCES
jects often involve a lot of soil investigation which is
TNO Geologische Dienst Nederland, DINOloket, https://
otherwise difficult to visualize.
www.dinoloket.nl/, last visited 19-11-2021
CUR, 2006. GEOTECHNICAL EXCHANGE FORMAT
FOR CPT-DATA Version: 1, 1, 2
Hou, L., Yu, C. P., & Samaras, D. 2016. Squared earth
mover’s distance-based loss for training deep neural net
works. arXiv preprint arXiv:1611.05916.
Van Erven, T., & Harremos, P. (2014). Rényi divergence
and Kullback-Leibler divergence. IEEE Transactions on
Figure 8. Example of geotechnical longitudinal profile. Information Theory, 60(7), 3797–3820.
Robertson, P. K. 2010. Soil behaviour type from the CPT:
an update. In 2nd International symposium on cone
Moreover, the automated soil classification can be penetration testing (Vol.2, pp. 575–583). Cone Penetra
combined with the geometry of the dike to build cal tion Testing Organizing Committee.
culation files for stability calculations. In the Been, K., & Jefferies, M. G. 1992. Towards systematic
CPT interpretation. In Predictive soil mechanics: Pro
example in Figure 9 the results of soil investigation ceedings of the Wroth Memorial Symposium held at St
on the crest and at the toe of the dike have been auto Catherine’s College, Oxford, 27-29 July 1992 (pp.
matically combined with Lidar measurements of the 121–134). Thomas Telford Publishing
ground surface and bathimetry data. Sassi, A., Brahimi, M., Bechkit, W., & Bachir, A. 2019.
Location embedding and deep convolutional neural net
works for next location prediction. In 2019 IEEE 44th
5.3 Conclusions LCN symposium on emerging topics in networking
(LCN symposium) (pp. 149–157). IEEE.
A CPT interpretation based on data proved to be CEMS, Nuclei website, https://crux-nuclei.com/api/gef
better than empirical formulations. The gain in terms model/ui/, last visited 14-01-2022
650
© 2022 the Author(s), ISBN 978-1-032-31259-0
Watch out for the use of global correlations and “black box” interpretation
of CPTU data
J.J.M. Powell
Geolabs Limited, Watford, UK
L. Dhimitri
In Situ Site Investigations Ltd, UK
ABSTRACT: Many people interpret soil properties from CPTU measurements based on correlations embed
ded in software packages without ever questioning the validity of those correlations. This could be termed the
‘black box’ approach! This paper aims to show how dangerous this can be but also the power of the CPTU in
helping to show the variations in soil properties within profiles. Based on correlations properties can be both
over and under-estimated, which can of course result in both unsafe design and over design. Too often one
correlation must be used to derive a soil property required in another correlation, further compounding the
potential for errors. A range of sites will be examined with a range of soils varying from very soft clays and
silts to stiff clays, sands and soft rocks.
1 INTRODUCTION all possible soils properties. This is an easy electronic

process even without proper user input.
The Cone Penetration Test (CPT) or Cone Pene This paper aims to show how dangerous data pro
tration Test with pore water pressure measure cessing without the proper input can be and how
ment (CPTU) is almost certainly the most widely knowledge is required on reliable interpretations.
used in-situ test both onshore and offshore. Its But it also shows how if done with care, the power
equipment and operation are well standardised of the CPTU in determining the variations of soil
(ISO 22476-1 and ASTM D5778). If these stand properties is still so very worthwhile. Based on cor
ards are followed and the necessary quality relations, properties can be both over and under
checks are performed, then reliable measured estimated, which can of course result in both unsafe
results should be easily obtained and can then be design and over-design. Too often one correlation
used to derive estimates of the geotechnical must be used to derive a soil parameter required in
parameters to assist the geotechnical design. This another correlation. To further confuse the situation,
is when problems start. Most of the interpret some of the derived parameters are obtained by
ations are semi-empirical in nature and over the using values from other derived/ guessed parameters,
years many correlations have been published, adding more errors. It is not intended to mention cor
linking the measured CPT/U data to the required relations by name, but to point out the problems that
soil properties. Many people interpret soils from can occur if outputs from software are taken without
CPTU based on these correlations, which have any user interaction. The packages used are commer
also been embedded in various software packages cially available. No pre-selection of correlations has
without ever questioning their validity. Unfortu been made.
nately, this is being increasingly done and could For this purpose, 18 commercial sites and 14 test
be termed the ‘black box’ approach! bed sites, which include 100s of CPTUs with
Clients are forcing/encouraging CPT contractors to a maximum depth of 40m have been studied. Dif
derive soils properties without providing any add ferent soils examined vary from very soft to stiff
itional input. It should be acknowledged that these clays, silts, sands and soft rocks. Data cannot be
derivations might be guesses, not even best estimates. presented for all the sites, but the conclusions are
Processing of CPT/U raw data starts from convert drawn based on all the data that the authors have
ing voltages to engineering units, plotting the results reviewed. Guidance and advise will be given where
on soil behaviour type charts and ends on generating possible.
DOI: 10.1201/9781003308829-95
651
2 SOFTWARE DERIVATION only. At this phase of data processing Soil Behaviour
Type, SBT can be plotted on one of the charts avail
2.1 Parameters able based on qc/ qt and Rf.
To derive more soil properties, it is nearly always
Some of the geotechnical parameters that can be necessary to have information on groundwater con
derived from CPTU test results by using various cor ditions, GWL and density/unit weight, γ to establish
relations published in the literature and incorporated total and effective vertical stresses, σv0 and σ’v, to
into various software packages are listed in Table 1. derive pore pressure ratio, Bq and other normalized
parameters, Qt and FR.
Information about GWL from the CPTU can
Table 1. List of possible parameters to derive from correl only be obtained if full dissipation tests are run and
ations available in the literature. this is seldom done. Hence, guessed GWL must be
Parameters Symbols Unit
used as an input to the software if the client cannot
supply any information from monitoring it on site.
Relative Density Dr % Regarding γ more details are given in the following
Undrained Shear Strength su kPa section.
Water Content wc %
SPT number N60 - 2.2 Unit weight/ density (γ) and water content (wc)
Shear Wave Velocity vs m/s
Unit weight γ kN/m When γ is derived from the equations available in
Small Strain Shear Modulus G0 MPa the literature and found in many software packages
Small Strain Youngs Modulus E0 MPa qt, fs, and specific gravity of solids, Gs are required
Constrained Modulus M MPa to run these calculations. One of the correlations
Coefficient of volume change mv m2/MN require also shear wave velocity, Vs which is one of
Compression Index Cc - the parameters that will be discussed later in the
Overconsolidation Ratio OCR - paper.
Friction Angle φ’ ° In Figure 1 are shown some examples of derived γ
Effective Cohesion c’ kPa which are compared with laboratory γ. A significant
Sensitivity St - and consistent underestimation of γ for the glacial till
Coefficient of Lateral at Cowden (Powell & Butcher, 2002) in England and
Earth Pressure K0 - the silt at Lierstranda (Lunne, 2002) in Norway can
Rigidity Index Ir - be observed. The same behaviour was present at 2
Hydraulic Conductivity k (kh and kv) m/s other till sites and 2 more silty clay sites.
In the same figure results from the London clay
site at Canons Park (Powell et al., 2003) are also
included. Again, derived γ profiles are on the low
The process of deriving all geotechnical param side of the measured profile. One of the correlations
eters from CPTU results after gathering the meas is giving results on the opposite direction of the
ured data from site, which consists of cone measured. This pattern of results was apparent for all
resistance qc, sleeve friction fs, and porewater pres 5 London clay sites investigated by the authors and
sure u2 starts with generating the corrected cone was similar in other heavily overconsolidated and
resistance, qt and friction ratio, Rf through very aged clay sites. Underestimation is not always the
simple calculations which involve measured results case. γ results for the soft clay at Cran (Shields et al.,
Figure 1. Examples of underestimated and overestimated γ derived from various correlations in use.
652
1996), in France are highly overestimated, as can be available in the specialist software are used to derive
seen from the last graph in Figure 1. Dr, noted in the figure as C1, C2, C3 and C4 Pre and
Some packages give also wc as an output parameter. Post, respectively. Once again, the wide range of
As might be expected, these results tend to be mirror results, especially those being highly overestimated
images of the density profiles using a selected Gs. after compaction is worrying.
2.3 Relative Density (Dr) 2.4 Overconsolidation Ratio (OCR)

When considering sand then Dr is often a desired When considering OCR or alternatively yield stress,
parameter. In order to derive Dr from CPTU results, CPTU is known to be a powerful tool to profile it.
qc, qt, σv0 and σatm are typically necessary input. Some correlations can estimate OCR based on qt, u2,
Figure 2 shows results from the sand site at Dunkirk Δu, σv0, σ’v and σatm. Some others require φ’ and G0,
(McAdam et al., 2020), which is partially placed and which on the other hand require qc, qt and/or Qt and
partially natural deposit. The wide range of derived γ. The long list of measured and derived parameters
Dr results after using some of the correlations avail to estimate OCR makes its calculation prone to errors.
able in the software is worrying. Generally, under It is worth mentioning, that OCR is the parameter
estimation is seen at shallower depths and with the largest number of correlations available.
overestimation at deeper depths. Incorporating the Figure 3 shows results from some of the sites stud
correct γ values and the correct GWL, will signifi ied and it can be observed that the shape of the profiles
cantly improve the results and better define the two generally follows the sites best estimate. However, the
layers that comprise this site. The second graph in absolute values vary considerably. From all the results
Figure 2, shows derived Dr before and after compac we have processed, one of the correlations consistently
tion of a sand fill. Four different correlations gives the same results of OCR around 1 for all ground
conditions and soil types!
Looking into more details, for the glacial till at
Cowden some of the profiles match well with the
site best estimates, especially below 22m, where
OCR increases. There are results from correlations
that highly overestimate OCR, which are not correct.
The first attempt for Cowden was done using the
default value of γ set up in the specialist software.
The second analysis of the results was carried out
used the true γ of 22kN/m3. Using the correct γ
rather than the software generated values for this
site, improves the derived OCR results, especially
for correlation 6 which is now closer to the best esti
mate site characteristic values. Although, it is worth
mentioning that two of the correlations that seem to
not agree are both based on porewater pressure.
At a London Clay site in Canons Park the results
Figure 2. Examples of derived Dr in sands showing under presented in this Figure show significant underesti
and overestimation of site best estimate values. mation compared to the site best estimated values,
Figure 3. Examples of derived and site best estimate OCR results for 4 sites comprising glacial till, London clay and
clayey sand.
653
although in most cases generated graphs have the This pattern has been observed on many other clay
correct shape, with values being reduced with depth. sites studied, including the soft clays in Glasgow.
Things can go very wrong indeed, a clayey sand However, on two sites presented in Figure 5
site in Jossingny (Shields et al., 1996), shows unreal results are over and underestimated, in sandy layers
istic overestimation as seen in Figure 3. and in clayey layers, respectively. Regarding soft
Using these correlations without questioning the rocks, in the second graph below are shown the
applicability of them in certain soil conditions is seen Chalk results, where in the upper meters in the very
to give wildly wrong results. Uniform sites (like weathered Chalk, all SPT profiles seem to agree
Cowden) have been examined to show that some cor well. When penetrating through competent Chalk
relations respond to changes in OCR caused by ero there is a greater difference, with derived results
sion. However, what happens if the geology changes, being higher than the measured ones.
as well? Turning back to the examples above and
many more results that authors reviewed for the pur
pose of this paper, it is not uncommon that a value of
around 1 is derived for the full profile-raising con
cerns on applicability of this specific correlation.
OCR is one of the parameters most influenced by
the GWL. Even 1m of GWL change can affect the
predicted OCR results from as little as 5% for stiff
clays to as high as 45% for soft clays. Therefore, cor
relations involving GWL as input parameter should
be avoided when this information is not accurate.
2.5 Standard Penetration Test (SPT) N Value

The SPT, love it or hate is still one of the widely used
in-situ tests worldwide. It has been said that “the best
way to get reliable N values is from correlations with Figure 5. Examples of derived SPT profiles in other soils.
CPTU”. The correlations available use qc, Soil Behav
iour Type Index, Ic and σatm to calculate N. Figures 4 An interesting observation for silt layers from all the
and 5 show CPTU derived SPT profiles and compare studied sites (two also included in Figures 5, the silt
them to the in-situ measured SPTs. Four different layer 6-9.5m in Hull results and 2-6m in Hemel Hemp-
sites in England are chosen as representative stead results) indicates that measured and derived
examples, which include the London Clay site of results show a better agreement in this soil type.
Heathrow Terminal 5 (Hight et al., 2002b) and three
commercial sites in Glasgow, Hull and Hemel Hemp-
2.6 Shear wave velocity (Vs) and Small Strain
stead of mixed glacial deposits, silt mixtures underlain
Stiffness (G0)
soft clay and silt underlain Chalk, respectively.
For London clay at T5, it can be seen that although When Vs and G0 are considered, it is needed to
the generated scattered CPTU profile has almost the ensure that anisotropy in the ground is not ignored.
same shape, it is below the in-situ measured profile. This affects the strength and the stiffness of the soils.
Results generated from standard geophysical tests,
which are typically referred to as downhole and cross-
hole can be different because of anisotropy. There is
a less common crosshole test that allows a third orien
tation of stiffness to be considered (Butcher &
Powell, 2004). As a result, is not enough to know
what is being measured, but also which measured dir
ection (vertical or lateral) is the most appropriate for
design. For clarity, the following subscripts have been
added to V and G to define the orientation of the
values: vh for downhole tests and seismic cone, hv
for standard crosshole tests and hh for results in a true
horizontal plane. The importance of Vs and G0 means
that these parameters are widely desired and therefore
derived from measured CPTU results.
Their derivation involves qt, fs, σv0 and γ. In the
following graphs, derived results are compared with
the measured ones from seismic cone tests and are
Figure 4. Examples of derived SPT profiles in London and
soft clay showing under and overestimation of measured
noted as Vvh. Figure 6 presents results from four
values. sites, with different geology.
654
Figure 6. Comparison between measured and derived Vvh results in clay sit, sand, glacial till and stiff clay.
Pentre (Lambson et al., 1993) represents a clayey

silt site, where the measured Vvh profile fall in
between the derived one.
For the sands at Dunkirk, which show the largest
scatter of derived results the measured Vvh profile
falls towards the upper bound of the derived values.
For the glacial till deposits at Cowden the measured
Vvh falls towards the lower bound of derived profiles.
Although the scatter here is not as large as the one
observed for Dunkirk, below 16m derived Vvh are
overestimated. Furthermore, big differences between
are also seen for the stiff clays in Madingley (Butcher
& Lord, 1993). Surprisingly, the greatest differences
noticed are when deriving Vvh from the correlation
which is suggested to work best in clays. This indi
cates that even if some correlations are recommended
for one soil type, big differences can still be present. Figure 7. Examples of underestimated derived Vvh results.
Does this mean that soil type is not the only limita
tion to the applicability of some correlations?
However, overestimation is not always the case.
Figure 7 shows results from very stiff clays in Ban-
bury and dense sands in Machynlleth, where derived
Vvh are significantly lower than the measured one.
The behaviours seen in Figures 6 and 7 for Vvh are
also noticed for derived and measured Gvh, Figure 8.
The results patterns for Gvh match those of Powell
(2017) using a wider range of correlations, with the
measured values for normally and lightly overconso
lidated soils resulting in between derived ones. Mean
while, the measured Gvh for heavily overconsolidated
clays fall in the lower bound of derived results.
Derivations of Gvh is generally linked to CPTU
data. However, derived Vvh and γ can be used to cal
culate Gvh. In this case, even if Vvh is realistic Gvh
Figure 8. Examples of derived and measured Gvh profiles.
could have errors for incorrect γ. Results from Dun
kirk support this. The Gvh profile generated from the
software is closer to the site measured profile when Long (2022) who, using the same and additional data
correct γ is used. This again shows the importance of suggests that Vhh correlates better to qt than Vvh does.
accurate input parameters to derive more parameters.
Powell and Butcher (2004) and later Powell et al. 2.7 Other parameters
(2016) suggested that it was the horizontal stresses
and stiffnesses that influence qt and that qt correlates Undrained shear strength, su is highly affected by the
better with Ghh. This idea is further supported by choice of cone factor, Nkt. This is the principal
655
problem, although γ and GWL should ideally be Correlations can be very useful but should only be used
known. Some of the software can generate Nkt values. when all influences in their derivation are understood.
Great care must be taken when this is done. It has been Finally, always revisit processing when site specific
found on the studied sites that it can lead to under and information becomes available and site-specific correl
overestimation of su. Values of Nkt should ideally be ations are established. They are far from foolproof.
selected by the designers based on past experiences Soils properties should be treated with caution if they
and the shear test types required, for example compres are derived without basic input information.
sion, extension or simple shear.
Another problem is the assessment of coefficient of
lateral earth pressure, K0. Correlations exist and are REFERENCES
incorporated into software, as well. But, realistic
assessment of K0 from CPTU is still extremely weak ASTM D5778-12. 2012. Standard Test Method for Per
forming Electronic Friction Cone and Piezocone Pene
and should be treated with great caution.
tration Testing of Soils. ASTM International. West
Another parameter that needs to be treated with Conshohocken, PA.
growing awareness is permeability (hydraulic con Butcher, A.P. & Powell, J.J.M. 1995. The effects of geo
ductivity), k, which is widely required for settlement logical history on the dynamic stiffness in soils. Proc,
calculations as an input parameter in design soft 11th ECSMFE. 1: 27–36.
ware. k from CPTU is a rough guide of possible Butcher, A.P. & Lord, J.A. 1993. Engineering properties of
values for each SBT, giving lower and upper bounds. Gault clay in and around Cambridge, UK. Geotechnical
The importance of dissipation tests to estimate coef Engineering of Hard rocks – Soft soils. 1: 405–416.
ficient of consolidation, cv is crucial. cv is acceptable Balkema.
Hight, D.W., McMillan, F., Powell, J.J.M., Jardine, R.J. &
when dissipations are carried out below GWL.
Allenou, C.P. 2002a. Some characteristics of London
When GWL is unknown, dissipation tests need to clay. Characterisation and Engineering Properties of
run to equilibrium. Natural soils, eds Tan et al. 2: 851–908.
Hight, D.W., Paul, M.A., Barras, B.F., Powell, J.J.M.,
Nash, D.F.T., Smith, P.R., Jardine, R.J. & Edwards, D.
3 DISCUSSIONS AND CONCLUSIONS H. 2002b. The characterization of the Bothkennar clay.
Characterisation and Engineering Properties of Natural
In this paper has been suggested that the use of soft soils, eds Tan et al. 1: 543–598.
ware to process CPTU results into geotechnical infor ISO 22476-1:2012. Geotechnical investigation and testing:
mation should be done with care. Although these Field testing Part 1: Electrical cone and piezocone pene
packages perform the mathematical calculations cor tration test.
Lambson, M.D., Clare, D.G., Senner, D.W.F. & Semple, R.M.
rectly, they are often used with too little basic informa
1993. Investigation and interpretation of Pentre and Til
tion, experience and/ or knowledge. brook Grange soil conditions. In Large scale pile tests in
It has been shown that many of the embedded cor clay (editor J Clarke). Thomas Telford: 134–196.
relations are not suitable for ALL soil types. Lacasse, S. & Lunne, T. 2002. Engineering Properties of Lean
In addition, the following points can be made: Lierstranda Clay. Proc. Intl Symp Coastal Engineering in
Practice, Yokohama September 2000. 2: 177–186.
• Correct γ and GWL can improve the results of Long, M., 2022. Practical use of shear wave velocity measure
many derived geotechnical parameters; ments from SCPTU in clays. Proc. CPT’22, June 2022.
• Derived parameters obtained based on other pre McAdam, R.A., Byrne, B.W., Housley, G.T., Burd, H.J.,
viously derived parameters should be avoided; Gavin, K.G., Igoe, D.J.P., Jardine, R.J., Martin, C.M.,
• Different correlations can give very different Potts, D.M., & Zdravković, L. 2020. Monotonic lat
derived parameters for the same soil type; erally loaded pile testing in a dense marine sand at
• Sometimes agreements between measured and Dunkirk. Géotechnique. 70 (11): 986–998
derived parameters have been found to be purely Mokkelbost, K.H., Lunne, T. & Powell, J.J.M.2000. Semi
empirical design procedures for foundation design. Proc
fortuitous. Two incorrectly derived parameters Nordic Geotechnical Conference
appearing to give a correct answer, this is worrying! Powell, J.J.M. & Butcher, A.P. 2002. Characterisation of
Correlations used in similar soil types are valu a glacial clay till at Cowden, Humberside. Characterisa
tion and Engineering Properties of Natural soils, eds
able, as are site specific ones. Parameter profile
Tan et al. 2: 983–1020.
shape is often promising, but the absolute values Powell, J.J.M., & Butcher, A.P. 2004. Small Strain Stiffness
between correlation vary wildly. assessments from in situ tests. Proc. ISC2 Porto. 17171722.
Can we find correlations that use soil type in Powell, J.J.M., Butcher, A.P. and Pellew, A. 2003, Capacity
terms of Ic as input parameter? It looks unlikely. of driven piles with time – implications for re-use. Proc
Why derive parameters that can be easily measured, XIIIth ECSMFE. Prague August 2003. 2: 335–340.
like Vvh. However, site specific correlation can be Powell, J.J.M., Dhimitri, L., Ward, D. & Butcher, A.P.
useful for example, between qt and measured Vvh from 2016. Small strain stiffness assessments from in situ
SCPTU, when used with other CPTUs on a site. tests – revisited. Proc.5th Intl Conf on Geotechnical and
Geophysical Site Characterization ISC’5 Gold Coast.
The importance of linking CPTU parameters with
Shields, C.H., Frank, R., Mokkelbost, K.H. & Denver, H. 1996.
results from quality laboratory tests cannot be over Design fourfold. Ground Engineering. 29 (2): 22–23.
stressed if efficiency in design is to be achieved.
656
© 2022 the Author(s), ISBN 978-1-032-31259-0
New methods for assessing Plasticity Index and Low-strain Shear Modulus
in fine-grained offshore soils
N. Ramsey
Fugro Australia Marine Pty. Ltd., Perth, Western Australia, Australia
K.K. Tho
Fugro Singapore Marine Pte. Ltd., Singapore
ABSTRACT: This paper introduces a new method for predicting Plasticity Index, PI, directly from CPT data.
Plasticity Index values predicted using the new method have been designated “PICPT” to emphasise their CPT-
based origin. The new PICPT method has, in turn, enabled a new method of assessing Low Strain Shear Modu
lus, Gmax, to be proposed, based on a well-established equation published by Viggiani & Atkinson (1995). As
an additional benefit, the predicted PICPT profiles have proven very useful as a simple means of identifying soil-
type variations and layer boundaries. This paper presents details of the new formulae that have been developed
to estimate PICPT and Gmax, the database that has been used to calibrate the formulae, and the results that have
been obtained. Some limitations of the method are presented and discussed, as well as some of the benefits.
1 INTRODUCTION 10 cm2 or 15 cm2 cones; but, at one site a cone with

a projected tip area of 5 cm2 cone was used.
This paper introduces new methods for estimating PI Several sites in the database included Gmax meas
and Gmax directly from CPT data, in fine-grained off urements from Resonant Column (RC), Bender
shore soils. The new method for estimating PI is also Element (BE) or Seismic Cone (SCPT) tests.
shown to be a simple means of differentiating soil- Four sites have been selected for more detailed
type variations and layer boundaries. presentation and discussion of PI and Gmax predic
This paper presents details of the formulae that tions. These sites were chosen because, together, they
have been developed, the database that has been covered a wide range of plasticity index and liquidity
used to calibrate the formulae and the results that index. At all four sites, the CPTs and sampling bore
have been obtained. Some limitations of the method holes were performed in close proximity, and Gmax
and some discussion on the additional benefits are measurements had been made via RC and/or BE test
also presented. ing. One site also included SCPT measurements.
General details of the soil conditions at the four
selected sites are presented in Table 1.
2 DATABASE AND SELECTED SITE DETAILS
Table 1. General details of site conditions.
2.1 General
The PI database used for this paper comprises ~PI ~OCR
approximately 200 PI measurements taken at more
Site Name Plasticity Symbols % - Notes
than a dozen sites. The sites were purposely chosen
to be from a variety of geological environments, and A CI-CH 15-40 5-10
to provide data with a range of plasticity. The major B CI 10-20 1.5-10
ity of the sites comprised predominantly silica soils, C CVH-CEH 55-75 1-2
but one site comprised high plasticity carbonate silt. D MH 15-25 1-3 Carbonate
The chosen sites were also widely distributed geo
graphically, and from a range of water depths.
The CPT data were gathered by four different geo
technical contractors, using cones made by three differ The ranges of soil conditions at the selected sites,
ent manufacturers. Most of the tests were performed are also considered representative of the ranges in the
with standard piezocones with projected tip areas of database.
DOI: 10.1201/9781003308829-96
657
3 ASSESSING PI FROM CPT DATA and C3 coefficients that gave consistently acceptable
agreement with the measured data. Therefore, an
3.1 General iterative approach was used to optimise the C1, C2
and C3 coefficients, using the following criteria:
The general form of the proposed formula for pre
dicting Plasticity Index from CPT (PICPT) data is • When viewed on an overall basis, a clear majority
presented in Equation 1: of measured PI values, in the database, should be
reasonably close to corresponding PICPT values.
• When viewed on an individual-site basis, a clear
majority of measured PI values, should be reason
ably close to, and follow the trends of, corres
ponding PICPT values.
Based on the two criteria above, a best fit was
where Rft is the ratio of sleeve friction, fs, to total achieved using C1, C2 and C3 coefficients of 17.5,
cone resistance, qt; Bq is the ratio of excess pore- 0.31 and 1.2 respectively, as shown in Equation 3.
water pressure, Δu, to net cone resistance, qnet; and
Qt is the ratio of qnet to in-situ vertical effective
stress, po’.
Although, Equation 1 is essentially an empirical
formula, it is based on some well-established con
cepts. The basic premise of the formula is that fric
tion ratio is primarily affected by the soil’s where both Rft and PICPT are percentage-values, in
mineralogy, stress history and lateral effective stress. accordance with standard industry practice. Conse
So, by removing the effect of stress history and lat quently, from now on, PICPT will be presented, and
eral effective stress, the friction ratio becomes pri used, as a percentage.
marily affected only by the soil’s mineralogy, which
is manifested in terms of the measured Plasticity
Index. 3.2 Direct comparisons
As both Bq and Qt are strongly influenced by Figure 1 presents a combined plot of measured PI
stress history, it would seem reasonable that both values plotted directly against automatically picked
these parameters could be used to try and reverse the corresponding PICPT values, from all the sites. Auto
effects of stress history on the measured skin fric matic picking was used to avoid unconscious human
tion. Reversing the effects of effective lateral stress bias influencing the picked values. Only values that
is more challenging, as the Bq parameter is not sig were clearly inconsistent were removed – for
nificantly influenced by lateral effective stress – as example where a measured PI value was close to
may be observed in normally consolidated clay, a soil-type boundary, but where the CPT data clearly
where Bq can be almost constant over several tens of indicated a different boundary elevation.
metres. At first glance, Qt might also seem to fall
into this category. However, it may be by remem
bered that mechanically induced Overconsolidation
Ratio, OCR, may be written in terms of the current
effective vertical stress and an additional vertical
effective stress, Δp that has been removed at some
point in time (Equation 2). Hence, as the Qt param
eter is normalised by the current vertical effective
stress, it will tend to reduce gradually with depth, if
Δp in the soil-layer is constant and greater than zero.
Initial attempts to assess appropriate coefficients

were encouraging, as it was clear that the general
trends of measured PI values were being correctly
predicted. However, when low values of the C1 coef
ficient were chosen, it was necessary to use rela
tively high C2 and C3 coefficients, which made the
predicted PI very sensitive to small changes in Bq
and Qt. Whereas when high values of the C1 coeffi
cient were chosen, it was not possible to choose C2 Figure 1. Comparison of all measured PI and PICPT values.
658
Figure 1 confirms that, overall, measured PI
values, in the database, are reasonably close to cor
responding predicted PICPT values – and when the
error is greater than the ranges shown, there is often
a clear and identifiable reason. For example:
a) As shown on Figure 1, extremely high plasticity
active clays, often including soil descriptors such
as “slickensided” or “organic” tend to be
underpredicted.
b) Strongly structured soils tend to be more difficult
to predict. In the example shown on Figure 1, the
soils were described as “fissured”, and PICPT
values were greater than measured PI values.
However, it is considered possible that PICPT
could over-predict or under-predict measured PI
in strongly structured soils.
c) It is well known that a coarse inclusion, such as
a piece of coarse gravel in a fine-grained matrix, can
cause temporary variations in both pore-water pres
sure and sleeve friction – and these variations can
lead to erroneous PICPT predictions. The example
highlighted in Figure 1 was caused by an inclusion
that was observed in the corresponding soil sample.
3.3 Example sites
3.3.1 General
The section presents comparisons of measured PI
and PICPT at four selected sites. It may be noted, Figure 2. Site A – Measured and inferred PI versus depth.
however, the consistency of agreement shown for
these four sites is considered typical of the consist
ency observed at other sites in the database. Even at
the “slickensided” site, highlighted in Figure 1,
a reasonably consistent difference between predicted
and measured values was observed, suggesting the
error was systematic rather than random.
3.3.2 Site A
Figure 2 presents a simplified soil profile (based on
the adjacent sampling borehole log) to illustrate the
interbedded nature of this site. Even so, it may be
seen that the predicted PICPT profile is in good agree
ment with measured PI values. Furthermore, if PICPT
≥7% is used to define fine-grained soils, the PICPT
profile can be seen to provide a simple means of dif
ferentiating fine-grained and coarse-grained soil
and soil-type variations and layer boundaries.
3.3.3 Site B
This site was chosen because of the relatively low Plas
ticity Index and the relatively large number of measure
ments. It may be seen from Figure 3 that the predicted
PI profile is in good agreement with the measured
values. For example, where the measured PI values are
uniform in the upper 20 metres, the predicted PI profile
is uniform. Below 20 metres there is more variability
in the measured PI values, and this is also seen in the
predicted profile. Finally, the decrease in measured PI
values below 40 metres is also observed in the pre
dicted PI profile (although there is some variability). Figure 3. Site B – Measured and inferred PI versus depth.
659
3.3.4 Site C
This site was chosen because of its very high to
extremely high plasticity. As shown in Figure 4,
the PICPT profile generally slightly underpredicts
the measured PI values; but, at some locations,
the difference was greater, for example in the
range 30-40 metres. It is notable, however, that
in this depth range, “organic staining” was
included in the corresponding soil descriptions.
This might suggest the greater differences were
caused by locally higher organic content.
Although, no direct organic content measurements
were made on the corresponding samples – it is
notable that organic contents in the same soil
unit, at the same site, did indicate local variations
in organic content.
Figure 5. Site D – Measured and inferred PI versus depth.
3.4 Discussion
Generally, the predicted PICPT values are in good
agreement with measured data, at the sites within
the database, provided the soils are not organic or
significantly structured. However, obviously, the
ability to make reliable predictions is highly
dependent on the reliability of the input data. Con
sequently, the following potential sources of error
must be avoided or taken into account when inter
preting the results:
• Errors in cone zero measurements – particularly
the friction sleeve zero measurement.
• Errors in friction sleeve measurements caused by
temperature effects – for example when the cone
Figure 4. Site C – Measured and inferred PI versus depth.
has passed through very dense sand.
• Cone dimensions – particularly friction sleeve
dimensions - that are outside the allowable toler
ances given in ISO 19901-8 (2014).
3.3.5 Site D • Cone measurements that do not achieve ISO
This site was chosen because it comprised high 19901-8 (2014) Application Class 1 limits.
plasticity carbonate silt. The agreement between • Incorrect calculation of qt, qnet, Rft and Qt.
predicted and measured PI values at this site was • Sleeve friction measurements not suitably corrected
better than expected. However, we would not con for friction sleeve area effects (Ramsey, 2022).
sider it wise to assume that a similar level of agree
ment would be achieved at all carbonate silt/clay An additional benefit of the PICPT formula was
sites. that it proved useful for defining the boundaries
660
between fine-grained and coarse-grained soils. Based
on the sites investigated, it also appeared to predict
the thickness and plasticity of interbedded layers rea
sonably reliably.
A further benefit of being able to predict PI, rea
sonably reliably, is that it enables PI to be explicitly Equations 7 and 8 generate ‘n’ and ‘m’ coefficients
included in Gmax assessments based on CPT data – that are similar to the “general trends” lines shown by
and this aspect is discussed in the next section. V&A (1995). Whereas Equation 6 generates ‘A’ coef
ficients that lie slightly below V&A’s “general trends”
line, because doing so produced a better overall fit
4 ASSESSING LOW-STRAIN SHEAR with the measured Gmax values in our database.
MODULUS USING CPT DATA
4.1 General 4.3 Assessing OCR from CPT data

The new method for assessing Gmax directly is an OCR is one of the key input parameters in Equa
adaptation of the well-established Gmax formula pub tion 5. Various formulae for predicting OCR from
lished by Viggiani & Atkinson (V&A) in 1995, and CPT data were assessed, including a new formula
shown in Equation 4: that accounted for plasticity. However, it was
concluded that the formula presented in Equa
tion 9 (Kulhawy & Mayne, 1990) and illustrated
in Figure 6, provided reasonable agreement
between predicted and measured OCR values. It
is notable that the four data points lying signifi
cantly below the “1 to 1” line, in Figure 6, were
where, Go is equivalent to Gmax (kPa), Ro is equivalent measured in soils that were considered to have
to OCR, pr is a reference stress (equal to 1 kPa), p’ is been desiccated by sub-aerial exposure. This is
the mean effective stress (kPa) and ‘A’, ‘n’ and ‘m’ are considered to suggest that the yield stress ratio
coefficients that depend on Plasticity Index (%). may be underpredicted when the yield stress has
The advantage of the V&A (1995) formula over increased due to desiccation rather than being
most popular CPT-based methods of assessing Gmax, mechanically induced – and is, therefore,
is that it explicitly accounts for the principal factors a limitation on the accuracy of Gmax predictions
that influence Gmax, which are OCR, p’ and PI. So, as made using the new method.
the PICPT formula can now be used to estimate PI, all
the required V&A (1995) input factors can now be
directly estimated from CPT data, using Equation 5:
4.4 Assessing mean effective stress from CPT data

Mean effective stress has been calculated using the
Equation 10:
Further details of the methods used to calculate

the input factors are discussed in the next sections;
followed by profiles comparing predicted and meas
ured Gmax values, at the four selected sites. where Ko is the lateral earth pressure at rest, and
po’ is the current in-situ vertical effective stress.
4.2 Estimating the A, n and m coefficients Ko has been calculated using Equation 11, using
the approach proposed by Mayne & Kulhawy (1982)
In the Viggiani & Atkinson (1995) paper, the and assuming an angle of internal friction of 30°
coefficients ‘A’, ‘n’ and ‘m’ are graphically (which is considered sufficiently accurate for the
related to PI (%), and lines that “indicate only intended purpose):
the general trends of the data” are annotated on
explanatory graphs. Nevertheless, V&A’s general
trend lines have been used as the basis for the
formulae proposed in this paper, which are pre
sented in Equations 6 to 8:
The preferred method for estimating current in-
situ vertical effective stress, po’, is to use laboratory
data from nearby sampling boreholes. However, if
661
no local data are available, then there are published 4.5.5 Site D – Gmax predictions
relationships for assessing submerged unit weight The predicted Gmax profile is in good agreement
from CPT data, such as Robertson & Cabal (2010). with the three of the four measured values at this
high plasticity carbonate silt site. The reasons for the
discrepancy with the lowest laboratory measured
value is not completely clear. However, local zones
of higher cementation were observed at this site, so
it is considered possible that the laboratory test was
affected by a locally higher degree of cementation.
Figure 6. Comparison of predicted and interpreted OCR.
4.5 Comparisons between predicted and measured

Gmax values
4.5.1 General
The four examples that follow are comparisons
between predicted and measured Gmax values at the
same four sites, as the PI comparisons.
4.5.2 Site A – Gmax predictions

Figure 7a) illustrates that the laboratory measure
ments and SCPT measurements were in good agree
ment with the predicted Gmax profile, at this
interbedded sand, silt and clay site.
4.5.3 Site B – Gmax predictions

Figure 7b) presents predicted and measured Gmax
values from Site B. At this, there was initially
poor agreement between predicted and measured
values. However, on closer inspection, the labora
tory test specimens had been consolidated to the
Figure 7. Comparison of predicted and measured Gmax
estimated vertical effective stress, rather than the values.
mean effective stress. After accounting for this
discrepancy, the predicted values are much closer
to the measured values, albeit systematically 4.6 Gmax discussion
higher. The difference is considered likely to be
due to correction errors. • Generally, the new method of predicting Gmax
directly from CPT data, is considered to provide
4.5.4 Site C – Gmax predictions good predictions of Gmax values at the sites
Figure 7c) indicates very good agreement within our database. However, the initial lack of
between predicted and measured Gmax values, agreement between predicted and measured
which might seem surprising given that the PICPT values at Site B, highlights the need to ensure that
profile at this site was consistently underestimated any corresponding laboratory tests are consoli
(see Figure 4). However, at high PI, the A(PICPT) dated appropriately. Also, as the method is
and n(PICPT) coefficients become less sensitive to dependent on reliable PICPT predictions, the same
changes in PI, which probably explains the good potential sources of error, discussed in Sec
agreement. tion 3.4, need to be avoided, or taken into
662
account, when processing the results. Finally, • The limitations of the proposed new method
potential errors caused by inaccurate OCR predic for predicting Gmax, are generally similar to
tions in sub-aerially exposed soils, should be the limitations of the new PICPT method - with
taken into account. the addition of potential errors caused by
inaccurate OCR predictions in sub-aerially
exposed soils.
5 CONCLUSIONS AND RECOMMENDATIONS
• A new method of prediction PI values directly REFERENCES

from CPT data is proposed. Generally, the pre
dicted PICPT values are in good agreement with ISO. 2014. Petroleum and natural gas industries - specific
measured PI values in the database. requirements for offshore structures - Part 8: marine
• An additional benefit of the predicted PICPT pro soil investigations, International Standard ISO 19901-8:
files is that the profiles can be used as a simple 2014.
Kulhawy, F.H. & Mayne, P.W. 1990. Manual on estimating
means of differentiating soil-type variations and soil properties for foundation design. EPRI Report
layer boundaries. EL6800, Electric Power Research Institute, Palo Alto.
• A limitation of the proposed new method for Mayne, P.W. & Kulhawy, F.H. 1982. Ko-OCR relationships
predicting PICPT is that the quality of the pre in soil. Journal of the Geotechnical Engineering Div
dictions is dependent on high quality CPT ision, 108(6), pp.851–872.
measurements. Fortunately, this limitation can Ramsey, N. 2022. The effects of cone geometry and testing
be mitigated by ensuring all tests are per deployment on CPT measurements - practical examples
formed with cones that comply with the toler in a normally consolidated deep-water clay. Proceedings
of the 20th International Conference on Soil Mechanics
ances and accuracies specified in ISO 19901-8
and Geotechnical Engineering, Sydney, Australia.
(2014) for Class 1 testing. Robertson, P.K. & Cabal, K.L. 2010. Estimating unit
• The new PICPT formula has enabled a new weight from CPT. Proceedings of the 2nd International
method for predicting Gmax to be proposed, based Symposium on Cone Penetration Testing, CPT’10, Hun
on a well-established equation proposed by Vig tington Beach, CA, USA.
giani & Atkinson (1995). Viggiani, G. & Atkinson, J.H. 1995. Stiffness of
• The predicted Gmax values are in good agreement fine-grained soil at very small strains. Géotechnique, 45
with measured Gmax values in the database. (2), pp.249–265.
663
© 2022 the Author(s), ISBN 978-1-032-31259-0
Cone penetration testing and interpretation in the holds of two ore-carrying

vessels
N. Ramsey
Fugro, Australia
ABSTRACT: This paper provides details of two in-situ testing programmes performed to assess ore proper
ties in the holds of ore carrying vessels. The first case comprised CPTs in the hold of an iron-ore carrying
vessel - at both the ports of loading and discharge – enabling assessment of the ore properties after loading
and the changes caused by the voyage. The second case comprised two CPTs and one Seismic Dilatometer
Test (SDMT) in the hold of a coal carrying vessel shortly after loading – enabling assessment of the coal
properties after loading, and direct comparison of CPT and SDMT interpretations.
1 INTRODUCTION
Ore-liquefaction can be caused by cyclic loading

induced by ship-motions. If liquefaction occurs, the
ore may lose most, if not all, its inter-particle frictional
strength and may flow like a viscous fluid. If flow
does occur, one potential consequence is an irredeem
able lateral shift in the ship’s centre of gravity – a shift
that could have potentially catastrophic consequents.
This paper presents details of two high-quality in-situ
testing programmes that were performed in the holds of
ore carrying vessels, to provide reference information
for ore liquefaction and stability assessments. Full
details of these assessments are beyond the remit of this
paper. However, opportunities to perform in-situ tests Figure 1. CPT equipment at iron-ore port of loading.
on ore mounds in ships’ holds are rare – so, it is hoped
that the data and interpretations, presented herein, will
be of interest to the geotechnical community.
2 PROJECT DETAILS
2.1 Case 1 – iron-ore
2.1.1 Fieldwork
The fieldwork was performed in two phases. Phase 1
comprised, three CPTs at the port of loading and Phase
2 comprised three CPTs at the port of discharge. Each
test was performed from the top of the ore mound to
the base of the mound. The tests were performed on an Figure 2. CPT equipment at iron-ore port of discharge.
approximate one-metre grid, so the maximum separ
ation between any test was approximately two metres.
As may be seen in Figures 1 and 2, different CPT 2.1.2 Iron-ore CPT results
rigs were used at the ports of loading and discharge. Reference CPT data, for the iron ore mound, are pre-
However, similar Class 1 piezocones were used at sented on Figure 3 (port of loading) and Figure 4 (port
both sites. of discharge). The test data are generally consistent at
DOI: 10.1201/9781003308829-97
664
the port of loading (except for some variability in CPT 2.2 Case 2 – coal
NP2 in the upper 2.5 m). In contrast, significantly
more variability is apparent at the port of discharge – 2.2.1 Fieldwork
in particular, the sleeve-friction measurements are The fieldwork was performed shortly after the ship
significantly higher. The measured pore-water pressure was loaded. The fieldwork comprised CPTs at two
values were negligible, indicating fully drained behav locations, approximately 7.6 m apart. CPT PK1 was
iour at both ports. performed on the top-centre of the coal mound, whilst
CPT PK2 was performed closer to the edge of the
coal mound. A Seismic Dilatometer (SDMT) test was
performed approximately 1.6 m from CPT PK1. Dilat
ometer measurements were made at 2 m intervals and
shear-wave measurements at 4 m intervals.
Figure 5 Illustrates that some preparation of the
coal mound was considered necessary to enable the
testing rig to be installed in a safe and stable manner.
For operational reasons, the test programme was per
formed at night (Figure 6).
Figure 3. Case 1 –CPT data – Iron-ore port of loading.
Figure 5. Coal mound at port of loading.
Figure 6. CPT and SDMT equipment on coal mound.
2.3 Coal – CPT results

Reference CPT data are presented on Figure 7. CPT
PK1 indicates high variability of cone resistance
within the upper seven metres of testing. This is con
Figure 4. Case 1 – Basic CPT data – Iron-ore port of sidered likely to be a product of the initial mound
discharge. levelling works, as the results for CPT PK2 (where
less preparation was performed) were highly
665
variable only in the upper two metres. Beneath the could then be inferred by comparing correspond
highly variable zones, both CPTs indicated ing results.
a steady trend of increasing cone tip resistance.
Localised peaks of cone resistance below the 3.2.1 Iron-ore reference data
upper variable zone, are considered likely to be Average dry density, ρdry = 1.94 Mg/m3
caused by the presence of larger coal fragments. Maximum dry density, ρd max = 2.66 Mg/m3
Pore-water pressure values were generally negli Minimum index dry density, ρd min = 1.76 Mg/m3
gible, except between 6-7 m in CPT PK1. This Particle density, PD = 3.96 Mg/ m3
zone of measurable pore pressure was considered
likely to be due to a localised zone of higher water 3.2.2 Methodology for inferring relative and dry
and fines contents. density profiles
The average dry density profile in the ore-mound
was inferred using a three-step iterative process.
Step 1 – infer nominal relative density profiles from
the CPT data
Step 2 - use the maximum and minimum dry density
data to convert the nominal relative density profiles
to equivalent dry density profiles.
Step 3 – iterate until the inferred relative density
and dry density profiles are plausible and the aver
age inferred dry density is the same as the average
dry density (calculated from survey measurements
of the mound volume and the measured weight of
the loaded ore).
The formulae used to perform these calculations are
discussed in the following section.
3.2.3 Density and void ratio calculations

Relative Density, Dr, is an estimate of the state of
material packing, relative to reference maximum and
minimum values. Relative density is defined as:
Figure 7. Case 2 - Basic CPT data – Coal port of loading.
3 INTERPRETATIONS Based on testing in a large calibration chamber,

Baldi et al. (1986) published the generic formula
3.1 General presented in Equation 1, for estimating the nominal
relative density of silica sands.
In each case, the average bulk density of the ore’, in
the ships hold, was calculated as the ratio of the
volume of the ore mound (surveyed using a laser-
scanning technique), and the mass of ore deposited
into the hold during loading.
In each case, maximum and minimum dry density
values were provided by the Client, so exact details
of the methods are unavailable for this paper. How The C0, C1 and C2 coefficients that provided the
ever, it is understood that maximum dry density best agreement for Ticino sand are summarised in
assessments were based on a standard “Proctor” Table 1. These values were used as a starting point
approach, and minimum dry density assessments for assessments of the iron-ore mound at the port of
were made using a funnel-type approach.
Table 1. Summary of coefficients used to estimate relative
density profiles in the iron-ore mound.
3.2 Case 1 – iron-ore
Material C0 C1 C2 K0
The CPT results were used to infer density and
void ratio profiles at the port of loading and the Ticino Sand 181 0.55 2.61 0.4
port of discharge. Changes in these profiles, Iron-ore 118 0.55 2.61 0.4
caused by the ship’s movements during sailing,
666
loading. It was concluded that plausible profiles volumetric strain, from greater than 4 % in the upper
could be obtained simply by changing the C0 coeffi metre, reducing to less than 1% below 4 m. Of
cient, as shown in Table 1. course, the magnitude of these inferred changes was
Dry density was calculated using Equation 3: dependent on the metocean conditions encountered
during the ship’s voyage. Discussion on these
aspects is, however, beyond the remit of this paper.
Although the profiles presented on Figures 8 and
9 are inferred, it is interesting to note that the aver
age inferred volumetric strain is in excellent agree
Void ratio and volumetric strain profiles were cal ment with the average volumetric strain calculated
culated using Equations 4 and 5, respectively. The from detailed surveys of the ore-mound geometries
results are presented on Figure 9, together with the at the ports of loading and discharge.
change in volumetric strain between the ports of
loading and discharge.
Figure 8 presents composite profiles of inferred

nominal relative density, dry density, and void ratio.
Although the profiles tend to be relatively consistent
below approximately 8 m depth, there is much more
variability in the upper material. To enable clearer
identification of trends, the void ratio data were aver
aged over 1 m increments. Figure 9 presents the
results of this averaging process, in terms of pre
dicted void ratio, change in void ration and volumet
ric strain. It may be seen that the trends are much
clearer in Figure 9. In particular the average volu
metric strain profile indicates a steady reduction in Figure 9. Changes in void ratio and volumetric strain
incurred between ports of loading and discharge.
3.3 Case 2 – coal
3.31 General approach

The objectives of the testing at the port of loading
were to assess density and stiffness variations in the
coal, shortly after loading in the ship’s hold. The
CPT data were used to infer dry density profiles
using the same three step approach used for the iron-
ore. The CPT data were also used to infer low-strain
shear stiffness, Gmax, and constrained modulus pro
files, to enable for comparison with profiles esti
mated using the SDMT data.
3.3.2 Coal reference data

Average dry density, rdry = 0.95 Mg/m3
Maximum dry density, rd max = 1.05 Mg/m3
Minimum dry density, rd min = 0.71 Mg/m3
Particle density, PD = 1.37 Mg/ m3
3.3.3 Density and void ratio calculations

Figure 8. Inferred relative density, dry density, and void As for the iron-ore, it was found that the general
ratio profiles at the ports of loading and discharge. form, given in Equation 2, could be adapted for use
667
in the coal mound. However, in the case of the coal
mound, all three coefficients were changed in order
to infer plausible profiles. The C0, C1 and C2 coeffi
cients that provided the best agreement are summar
ised in Table 2.
Table 2. Summary of coefficients used to estimate relative The original coefficients published by Baldi
density profiles in the coal mound. et al. (1986) are presented in Table 3, together
with the coefficients that provided better agree
C0 C1 C2 Ko ment with M values inferred from the SDMT
measurements.
17 0.4 3.9 0.4
Table 3. Summary of coefficients used to estimate con
strained modulus in the coal mound.
Inferred profiles of relative density, dry density
and void ratio are presented on Figure 10. Material C0 C1 C2 C3
Baldi et al (1986) 14.48 -0.116 0.313 -1.123

Coal mound 5.4 -0.116 0.313 -0.4
The right-hand graph on Figure 11 presents M,

profiles results of both CPT and SDMT interpret
ations – it is considered that there is acceptable
agreement over whole profile.
Figure 10. Inferred relative density, dry density, and void

ratio profiles at the port of loading.
3.3.4 Constrained modulus calculations

Constrained Modulus, M, was inferred from the
SDMT measurements using Equation 6 (Marchetti
et al., 2001).
Figure 11. Inferred Gmax and M profiles at the port of
loading.
M ¼ 0:85 � 34:7 � ðP1 -P0 Þ=1000Þ ð6Þ
3.3.5 Low-strain shear modulus calculations
Baldi et al (1986) developed a correlation for esti Direct measurements of shear wave velocity were
mating M in silica sands (for a vertical strain performed during the SDMT testing adjacent to Test
of approximately 0.1%) by comparing one- PK1. Low-strain shear modulus, Gmax, values
dimensional compression results with cone resistance between individual test locations were then calcu
data measured in a calibration chamber (Baldi, 1985). lated using Equation 8:
The correlation, which is presented in Equation 7, con
siders cone resistance, qc, in-situ stress conditions,
σm 0 , and over-consolidation ratio, OCR.:
668
Equations for estimating Gmax from cone data to suggest that there had been no significant ten
were also published by Baldi et al. (1986). However, dency of the ores to saturate at the base immedi
in the case of the coal mound, it was found that ately after loading.
acceptable agreement with the SDMT Gmax measure 7) Good agreement between stiffness properties
ments could be obtained simply by increasing the C0 inferred from CPT data and equivalent properties
coefficient in the Baldi et al. (1986) Constrained inferred from SDMT results could be obtained by
Modulus equation, by a factor of 4.5 - as shown in making ore-specific amendments to published
Equation 9. correlations.
REFERENCES
Baldi et al. (1985) Laboratory validation of in-situ tests.
Published by the Italian Geotechnical Society (AGI) on
the occasion of the ISSMFE Golden Jubilee
The results of the Gmax interpretations, which are Baldi, G., Bellotti, R, Ghionna, V. N., Jamiolkowski, M. &
presented on the middle graph on Figure 11, indicate Pasqualini, E. 1986. Interpretation of CPTs and CPTUs.
good agreement between the CPT interpretations and Part II: Drained Penetration of Sands, Proc. 4th Inter
the SDMT measurements. national Geotechnical Seminar on Field Instrumentation
and In-situ Measurements, Nanyang Tech Inst. Singapore.
Marchetti S., Monaco P., Totani G. & Calabrese M. 2001.
The Flat Dilatometer Test DMT. in soil investigations
4 DISCUSSION AND CONCLUSIONS A Report by the ISSMGE Committee TC16, IN SITU
2001, International. Conference on in-situ Measurement
of Soil Properties, Bali, Indonesia
1) As expected, ore properties were inferred to vary
much more in the upper part of the ore-mound,
with variability gradually reducing with depth.
2) A general trend of increasing nominal relative
SYMBOLS AND ABBREVIATIONS
density with depth was inferred in all tests. This
would be expected, as the standard “high-energy CPT Cone Penetration Test (with pore-pressure
air-pluviation” method used to load the ore ves measurements)
sels would inevitably lead to more compactive Dr nominal relative density (%)
effort being applied to materials at greater depths. e void ratio (-)
3) It was possible to obtain good agreement between e0 initial void ratio (-)
material properties inferred from CPT data, and fs CPT measured sleeve friction (MPa)
material properties inferred from survey data and Gmax low-strain shear modulus (MPa)
the weight of material loaded into the ships’ Ic Soil behaviour type index (-)
holds. Ko Coefficient of earth pressure at rest (-)
4) As expected, the CPT measurements implied that M secant constrained modulus (MPa)
vessel-movement induced densification occurred OCR overconsolidation ratio
from the top of the iron-ore mound downwards. pa reference stress equal to 98.1 kPa
5) The consistent increase in sleeve friction noted at PD particle density of ore (Mg/m3)
in the iron-ore mound at the port of discharge, P0 corrected dilatometer pressure reading when the
was considered to suggest that there had been membrane first breaks away from the seating.
a wave-induced increase in Ko. However, as no P1 corrected dilatometer reading when the membrane
testing was performed to quantify any change in movement is 1.1 mm away from the seating.
Ko, a constant value of Ko was assumed in the po0 effective overburden pressure (kPa)
calculations. This is considered a potential source qc CPT measured cone resistance (MPa)
of error in the post-voyage calculations of rela SDMT Dilatometer test with shear-wave measurements
tive density (and hence dry density and void u CPT measured pore-water pressure (MPa)
ratio). vs measured shear wave velocity in soil (m/s)
6) No significant zones of pore pressure were meas ρdry dry density of soil (Mg/m3)
ured at the ports of loading. This was interpreted ρbulk bulk density of soil (Mg/m3)
ρd max maximum dry density (Mg/m3)
ρd min minimum dry density (Mg/m3)
σm 0 mean effective stress at test depth (kPa)
669
© 2022 the Author(s), ISBN 978-1-032-31259-0
Suction influence on CPT and DMT for some Brazilian tropical soils
B.P. Rocha
Federal Institute of São Paulo (IFSP), Advanced Campus of Ilha Solteira, Brazil
R.A. Rodrigues & H.L. Giacheti

Department of Civil and Environmental Engineering, São Paulo State University (Unesp), Brazil
ABSTRACT: Suction plays an important role in geotechnical engineering practice since it influences the
mechanical behavior of unsaturated soils. Interpretation of in situ test data at unsaturated soil sites must con
sider both spatial and seasonal variability, especially in strength and stiffness parameters. This paper presents
and discusses CPT and DMT performed at different seasons at two sites with tropical sandy soils. CPT and
DMT data were interpreted considering the influence of soil suction. Suction was estimated from water con
tent profiles and soil water retention curves (SWRC). Failure to consider the effect of soil suction in the site
investigation of unsaturated soils can lead to inappropriate soil classification as well as significant and uncon
servative overestimates of geotechnical parameters.
1 INTRODUCTION different periods of the year. The tests data are pre
sented and interpreted considering the soil-water
Site characterization can be defined as the process of retention curves (SWRC). Soil suction influence was
identifying the geometry of relatively homogeneous incorporated to the effective stress following Bishop’s
zones and developing index, strength, and stiffness (1959) equation to interpret the in-situ tests data.
properties for the soils within these zones. Some in-situ
testing methods can be used as an alternative to the
traditional approach of drilling, sampling, and labora 2 BACKGROUND
tory testing.
Many geotechnical works (e.g., shallow founda Interpretation of in-situ tests usually assumes pore pres
tions, embankments, and slopes) are performed on sure equal to zero in unsaturated soils. Therefore, the
unsaturated deposits. Suction increases the shear total stress (σv) is assumed to be equal to the effective
strength and stiffness of these soils (Alonso et al. stress (σvʹ). Moreover, the interpretation does not con
1990). Such influence on the geotechnical soil behav sider the suction contribution to the effective stress.
ior is well understood based on laboratory tests, how Soil suction should be incorporated into the inter
ever they are expensive and time-consuming since it pretation of in-situ tests on unsaturated soils. Suction
involves undisturbed soil sampling. can be incorporated into effective stress following
In-situ tests, such as the cone penetration (CPT) Bishop (1959), when σvʹ is defined by Equation 1:
and the flat dilatometer (DMT) tests allow combining
stratigraphic logging with specific measurements,
which is a modern approach available for site charac where χ is the effective stress parameter, ua is the pore
terization and could be used to improve the site char air pressure and uw is the pore water pressure. Soil
acterization in unsaturated soils. Only few studies deal suction is the difference between ua and uw. χ is equal
with in situ test interpretation in unsaturated soils, as to 1 for saturated soils and 0 for dry soils, and it can
well as the influence of soil suction in the interpret be assumed to be equal to the degree of saturation (Sr)
ation (Lehane et al. 2004; Pournaghiazar et al. 2013; (Öberg & Sällfors 1997; Robertson et al. 2017).
Lo Presti et al. 2018; Giacheti et al. 2019; Rocha The interpretation of CPT data starts by calculat
et al. 2021). ing the normalized cone resistance (Qtn) (Eq. 2).
This paper presents and discusses the influence of
soil suction on CPT and DMT carried out in two
unsaturated tropical soil sites from Brazil. CPTs,
DMTs and water content profiles were determined in
DOI: 10.1201/9781003308829-98
670
where σv is the total stress, σvʹ is the effective provided a better CPT and DMT data interpretation for
stress, pa is the atmospheric reference pressure in the unsaturated soils.
same unit as qc and σv, (qc e σv)/pa is the dimension
less net cone resistance, (pa/σvʹ)n is the stress nor
malization factor, and n is the stress exponent that 3 STUDY SITES
relates qc to σ’v and can be defined in power law
relationships defined according to Robertson (2009). 3.1 Unesp and USP research sites
The stress exponent (n) is 1.0 for most fine-grained
soils, and ranges from 0.5 to 0.9 for coarse-grained The Unesp research site are in the city of Bauru,
soils. state of São Paulo. The soil at the site is an unsatur
Similar approach can be used to interpret DMT in ated red clayey fine sand with high porosity. It is
unsaturated soils. DMT data interpretation begins by classified as a soil from SM Group in the Unified
determining three intermediate parameters: material Soil Classification System (USCS). The top 13 m is
index (ID – Eq. 3), horizontal stress index (KD – a colluvium soil overlaying a residual soil derived
Eq. 4), and dilatometer modulus (ED – Eq. 5). from the weathering of sandstone. The colluvium
soil has lateritic soil behavior (LA’). The ground
water level was not found up to 30 m deep.
The USP research site are also located on the cen
tral part of São Paulo state in the São Carlos city,
which are around 150 km apart from Unesp - Bauru
research site. The site profile is a saprolitic sandstone
residual soil layer covered by a lateritic clayey sand
(LA’) layer (6 m thick colluvial soil - Cenozoic sedi
ment) (Machado and Vilar 1998). A 0.2 to
where p0 is the corrected first reading, p1 is the 0.5 m thick layer of pebbles separates the Cenozoic
corrected second reading, and u0 represents the pre Sediment layer from the residual soil. Both layers
insertion in-situ equilibrium pore pressure. are classified as a soil from SC Group by USCS. The
Lutenegger (1988) reported little change in the ID groundwater level varies seasonally between 9 and
profile and a decrease in the KD and ED profiles with 12 m below the ground surface. Table 1 shows typ
a reduction in soil suction. So, since the soil suction ical values for the grain size distribution, void ratio
influences DMT interpretation, it should be incorpor (e), dry unit weight (γd), unit weight of solids (γs)
ated into KD and ED values. Soil suction can be and consistency limits (wLL and wLP) for both sites.
incorporated in the σv’ to calculate the KD value.
However, effective stress is not included in Equa
tion 5. In this way, Equation 6 (Janbu 1963) was Table 1. Physical indexes, grain size distribution and
USCS classification for the soils from each site.
used to incorporate soil suction into the ED value.
Unesp USP
e 0.65 – 0.81 0.60 – 1.05

γd (kN/m³) 14.8 – 16.5 12.5 – 16.6
γs (kN/m³) 26.5 – 26.9 27.1 – 27.5
Sand (%) 78 – 83 60 – 69
where n is a stress exponent (0.5 for coarse-grained
Silt (%) 4–8 5 – 14
soils), KE is the modulus number and pa is the refer
Clay (%) 12 – 17 17 – 28
ence pressure assumed equal to 100 kPa.
wLL (%) 17 – 23 39 – 45
Two steps were used to determine the measured
wLP (%) non-plastic 19 – 25
ED values without and with considering the soils
USCS SM SC
suction:
• KE was calculated by Equation 6 considering the
soil suction value in the σvʹ by Equation 1.
• The normalized ED value was calculated by
3.1.1 Previously performed laboratory tests
Equation 6 considering KE previously defined
Drying soil-water retention curves (SWRCs), oed
and, assigning soil suction equal to zero to calcu
ometer and triaxial tests were carried out on undis
late σvʹ.
turbed soil samples collected from each site. Samples
Such approaches allow soil suction to be incorpor were collected at 1, 3 and 5 m depth for Unesp, and
ated into the interpretation of CPT and DMT data. Suc 2, 5 and 8 m depth for USP. The laboratory test cam
tion can be incorporated in σv’ by using χ and soil paigns were carried out to study the influence of soil
suction based on the water content profiles and soil- suction on soil strength and stiffness parameters. The
water retention curves (SWRC). That was used by main purpose of the laboratory investigation was to
Giacheti et al. (2019) and Rocha et al. (2021) and assess the influence of suction on the soil behavior.
671
The SWRCs were obtained using the suction- For the Unesp site, water content values higher
plate, pressure-chamber, and filter-paper. The results than 6 to 7% (Region A) the cohesion intercept and
for the tested samples were adjusted according to the preconsolidation stress values were slightly affected.
van Genuchten (1980) equation. Triaxial compression However, the opposite occurs for water content
and oedometer tests were performed on saturated and lower than 6 to 7% (Region B). It is noted that in
non-saturated samples. Controlled suction tests were this region of the curves, a slight variation in water
carried out by imposing the desired suction according content can substantially affect soil suction.
to axial translation technique (Hilf 1956). Figure 1
and 2 presents the SWRCs for Unesp and USP sites,
respectively. The suction effects in the cohesion Table 2. Cohesion (c) and preconsolidation stress (σp) for
intercept (c) and preconsolidation stress (σʹp) are pre different depths and soil suctions for Unesp site (adapted
sented in Table 2 for Unesp site and in Table 3 for from Fernandes et al. 2017).
USP site. The air entry value is very low for the soils
from both sites as can be seen in the SWRCS. They c (kPa) σp (kPa)
are typical sandy soils curves in that most of the
Suction
water is extracted by a small change in soil suction. (kPa)
1m 3m 5m 1m 3m 5m
The interpretation of soil suction influence on pre
consolidation and cohesion intercept combined with 0 0 1.2 5.3 30 36 69
SWRC allows to define two regions in SWRCs for 50 3 6.5 10.3 77 135 176
Unesp (Figure 1) and for USP sites (Figure 2). 100 - - - 108 163 205
200 11 13.4 24.2 140 200 221
400 16 21.5 28.3 176 221 229
Similar behavior can be observed for USP site.

c and σp values are slightly influenced by soil suction
for water content higher than 15 to 16%, i.e., Region
A of SWRCs (Figure 2). In Region B, both c and σp
vary markedly with suction, for water content values
below 15 to 16% (Figure 2). In this region a slight
variation in water content substantially modifies soil
suction.
Table 3. Cohesion intercept and preconsolidation stress

for different depths and soil suctions for USP site (adapted
from Machado 1998).
Figure 1. Soil-water retention curves for the soil samples
collected at 1.0, 3.0 and 5.0 m depths for Unesp site Cohesion intercept Preconsolidation
(adapted from Fernandes et al. 2017). (kPa) stress (kPa)
Suction
2m 5m 8m 2m 5m 8m
(kPa)
0 0 10.5 26.9 41 122 171
40 14.9 26.7 44.6 - - -
50 - - - 150 181 291
80 21.1 29.6 57.6 - - -
100 - - - 189 207 310
120 23.1 36.1 51.9 - - -
160 30.1 43.9 53.4 - - -
300 - - - 316 331 351
400 - - - 430 - 461
3.1.2 Previously performed in-situ tests

CPTs and DMTs were previously performed in both
sites to study the effect of the unsaturated condition on
Figure 2. Soil-water retention curves for the soil samples the interpretation of these tests. Water content profiles
collected at 2.0, 5.0 and 8.0 m depths for USP site (adapted (Figure 3) and soil water retention curves (Figure 1
from Machado 1998).
and Figure 2) were employed to evaluate such effects.
672
Figure 3a presents the average water content profiles and one soil sampling were conducted for each cam
obtained for the wet and the dry seasons at the Unesp paign. A multi-function penetrometer (150 kN thrust
site, while Figure 3b shows moisture content profiles capacity) was used to perform the in situ tests. Soil
measured between 2014 and 2017 for the USP site. It sampling was carried out from ground surface up to
can be seen in Figure 3a (Unesp site) that there is 8.0 m depth by using a helical auger to collect samples
a significant variation in water content up to 4 m depth, every 0.75 m to determine the water content profile.
where the soil interacts with the atmosphere, and
decreases with increasing depth. The water content in
3.3 CPT and DMT at the Unesp site
the soil down to 4 m depth is lower than 7% in the dry
season and vary from 7% to 10% in the wet season. It Figure 4 and Figure 5 presents the difference
can be observed in Figure 3b (USP site) that March/ between CPTs and DMTs carried out under differ
2016 was the period with higher water content and Octo ent water content conditions in terms of average qc
ber/2017 was the lowest at the end of the dry season. and fs (Figure 4) and average ID, KD and ED
It is important to point out that there are trees in (Figure 5). The data show the soil suction influ
both sites. The extraction of water via the roots of ence on CPT and DMT up to approximately
the trees during the dry season brought the degree of 5 m depth in the, mainly on the qc, KD and ED.
saturation to approximately 30-40% and, conse Such behavior can be explained from SWRCs
quently, high values of soil suction (Lehane et al. (Figure 1). The water content values determined in
2004; Giacheti et al. 2019; Rocha et al. 2021). April/2019 are greater than 6% and tend to be in
Giacheti et al. (2019) presented and discussed Region A in the SWRCs (Figure 1), while the
CPT data to illustrate the seasonal variability for the water content values determined in September/
Unesp site. Soil suction significantly influenced CPT 2019 are lower than 6% and tend to be in Region
data up to a depth of 4 m at the Unesp site. Rocha B of the SWRCs (Figure 1). In Region A of the
et al. (2021) presented and discussed the influence of curve, water content varies greatly with slight vari
soil suction on DMTs carried out at the USP site. ation in suction, and the opposite trend occurs in
Both papers incorporated soil suction into σvʹ by Region B, where suction values vary significantly
using χ and soil suction (Eq. 1), as estimated from with little variation in water content.
the degree of saturation and SWRCs, respectively, Soil suction was incorporated in σvʹ by Equa
for better interpretation of CPT and DMT data. tion 1 to improve CPT and DMT interpretation.
The estimated soil suction was defined from water
content profiles (Figure 4a and Figure 5a) and the
3.2 CPT, DMT and soil sampling
SWRCs (Figure 1). It was found that soil suction
Two campaigns were performed in 2019 at the Unesp values can be higher than 1 MPa up to 5 m depth
site, one in the wet season (April) and one in the dry in the September/2019 campaign. Khalili & Khab
season (September). Two CPTs, two DMTs and one baz (1998) highlighted that soil suction varies
soil sampling were carried out for each campaign. Two from 0 to 600 kPa for most practical applications
in situ testing campaigns were also carried out at the in geotechnical engineering. An average suction
USP site (March/2016 and October/2017). Three CPTs value equal to 300 kPa was used in this analysis. χ
was assumed to vary linearly with the degree of
saturation (Sr) for the Unesp site (Robertson et al.
2017; Giacheti et al. 2019; Rocha et al. 2021).
Table 4 presents the average χ parameter for each
test campaign and the assumed soil suction value
obtained from Figure 1 and Table 1.
Figure 3. Water content profiles for (a) Unesp and (b) USP Figure 4. Average qc, fs and water content profiles from
sites (adapted from Rocha et al. 2021; Giacheti et al. 2019). each test campaign for the Unesp site.
673
Table 4. Assumed values for χ and soil suction for each test
campaign - The Unesp site.
χ Soil suction (kPa)
April 2019 0.38 8
Sept/2019 0.25 300
Figure 7. (a) KD and (b) ED, considering soil suction influ

ence on σʹv for the Unesp site.
Figure 5. Average ID, KD, ED and water content profiles
from each test campaign for the Unesp site.
Figure 6 and Figure 7 showing, respectively, the 15.6% - Region A) and higher for October/2017
profiles of Qtn and of KD and ED were elaborated with (water content values lower than 15.3% - Region B).
out considering the soil suction influence, following The CPT interpretation starts from normalized
item 2 presented earlier in this paper. The stress expo cone resistance (Qtn) calculation by Equation 2. The
nent (n) was assumed equal to 0.70 for both saturated soil suction was incorporated into σvʹ following
and unsaturated conditions to determine Qtn for the Equation 1. Table 5 presents the average χ parameter
Unesp site. When the suction values from each test for both campaigns and the assumed soil suction
campaign were incorporated into the effective stresses, value determined by Figure 2 and Table 1. The stress
the average profiles of Qtn, KD and ED were similar exponent (n) was assumed equal to 0.97 to calculate
and equivalent to the profiles without soil suction. Qtn for both saturated and unsaturated conditions.
All Qtn profiles from each campaign are similar
(Figure 9) when the soil suction is included in effect
3.4 CPT at the USP site
ive stresses (Eq. 1 and Table 5).
Figure 8 shows the water content profiles
and average qc and fs determined for March/2016
and October/2017 campaigns. It is interesting to note
Table 5. Assumed values for χ and soil suction for each
in this figure that the variation in soil suction was test campaign - The USP site.
significant up to approximately 6 m depth.
These differences can be explained by the χ Soil suction (kPa)
SWRCs and water content profiles (Figure 2 and
Figure 8a). CPTs carried out in March/2016 tend to March/2016 0.65 10
be in Region A, and October/2017 CPTs in Region October/2017 0.42 150
B. The estimated soil suction values are lower for

March/2016 (water content values higher than
Figure 6. Average Qtn profiles (a) without and (b) with Figure 8. Water content and average CPT profiles from
incorporating soil suction for the Unesp site. each campaign for the USP site.
674
average CPT data to the depth of 5 m (active zone)
plotted on the CPT classification chart (Robertson
2009) with and without incorporating soil suction
into σvʹ. It can be seen in this figure that all CPT
data are plotted in the drained penetration zone,
which is an appropriate classification for this soil
type in unsaturated condition (Yang & Russell 2016).
It can be seen in Figure 10a that when soil suction
is not included in σvʹ, the soil is classified predomin
antly as sands (zone 6) during the dry season (high
suction) and as sand mixtures (zone 5) during the wet
season (low suction). Figure 10a also indicates that the
soil is more contractive during the wet season, while it
is more dilative during the dry season. When a suction
value equal to 300 kPa was incorporated into σvʹ, the
Figure 9. Average Qtn profiles incorporating soil suction
soil behaves in the same way as during the wet season
for the USP site.
(low suction), i.e., contractive sand mixtures
(Figure 10b). This agrees with the dispersant-free par
ticle size distribution curve (Table 1), as well as the
change in soil volume during shear (Fernandes et al.
2017).
5 CONCLUSIONS
The data presented show the influence of seasonal

variation in water content and soil suction profiles
and how they influence soil behavior CPT and
DMT data are significantly influenced by soil suc
tion. In general, the effects of suction are more
pronounced for lower soil densities and confining
stresses. It was observed that soil suction signifi
cantly influenced CPT data up to a depth of 6 m at
the USP site and up to 5 m depth for the Unesp
site The same behavior was also observed for the
DMT for the Unesp site.
The effective stress approach allows soil suc
tion to be considered into CPT and DMT data
interpretation. The influence can be incorporated
in σvʹ by using χ and soil suction based on the
water content profiles and SWRC. Such approach
provides a better CPT and DMT interpretation for
unsaturated soils.
ACKNOWLEDGMENTS
The authors thank the São Paulo Research Founda
tion, FAPESP (2015/17260-0 and 2017/23174-5),
the National Council for Scientific and Techno
logical Development, CNPq (308895/2015-0 and
436478/2018-8) and the Coordination for the
Figure 10. Average CPT data from both wet and dry sea Improvement of Higher Education Personnel,
sons plotted on the Robertson (2009) classification chart (a) CAPES, for the first author scholarship.
without and (b) with consideration of soil suction into σvʹ.
4 CPT CLASSIFICATION AND SOIL SUCTION REFERENCES

Alonso EE, Gens A, Josa A (1990) A constitutive model
The influence of suction on CPT classification was for partially saturated soils. Geotechnique 40: 405–430.
evaluated for the Unesp site. Figure 10 presents the http://dx.doi.org/10.1680/geot.1990.40.3.405
675
Bishop AW (1959) The principle of effective stress. Teknisk Marchetti S (1980) In situ tests by flat dilatometer. Journal
Ukeblad 106:859–863 of Geotechnical Engineering 106: 299–321
Fernandes JB, Rocha BP, Rodrigues RA, Giacheti HL (2017) Marchetti S, Monaco P, Totani G, Calabrese M (2001)
Strength and Stiffness Parameters of an Unsaturated Trop The Flat Dilatometer Test (DMT) in soil investigations:
ical Soil. 2nd PanAmerican Conference on Unsaturated a report by the ISSMGE TC Committee 16. In: Proceed
Soils. https://dx.doi.org/10.1061/9780784481707.044 ings of the international conference on in situ
Giacheti HL, Bezerra RC, Rocha BP, Rodrigues RA (2019) measurement of soil properties and case histories. Para
Seasonal influence in CPT: an unsaturated soil site hyangan Catholic University, Bandung, Indonesia, pp
example. Journal of Rock Mechanics and Geotechnical 95–132.
Engineering 11(2):361–368. https://dx.doi.org/10.1016/ Morais TSO, Tsuha CHC, Bandeira No LA, Singh RM
j.jrmge.2018.10.005 (2020) Effects of seasonal variations on the thermal
Hilf JW (1956) An investigation of pore-water pressure in response of energy piles in an unsaturated Brazilian
compacted cohesive soils. Ph.D. thesis, Faculty of the tropical soil. Energy Build 216:109971. https://dx.doi.
Graduate, School of the University of Colorado org/10.1016/j.enbuild.2020.109971
Janbu N (1963) Soil compressibility as determined by oed Öberg AL, Sällfors G (1997) Determination of shear
ometer and triaxial tests. In: Proc. of European confer strength parameters of unsaturated silts and sands
ence on soil mechanics and foundation engineering, based on the water retention curve. Geotechnical Test
vol 1. pp 19–25. Deutsche Gesellschaft fu¨r Erd-und ing Journal 20(1): 40–48. https://dx.doi.org/10.1520/
Grundbau, Wiesbaden, Germany. GTJ11419J
Khalili N, Khabbaz MH (1998) A unique relationship for Pournaghiazar M, Russell AR, Khalili N (2013) The cone
the determination of the shear strength of unsaturated penetration test in unsaturated sands. Geotechnique 63
soils. Géotechnique 48(5): 681–687. (14):1209–1220. https://dx.doi.org/10.1680/geot.12.P.083
Lehane BM, Ismail MA, Fahey M (2004) Seasonal depend Robertson, PK (2009) Interpretation of cone penetration
ence of in situ test parameters in sand above the water tests e a unified approach. Canadian Geotechnical Jour
table. Geotechnique 54(3): 215–218. https://dx.doi.org/ nal 46(11): 1337–1355.
10.1680/geot.2004.54.3.215 Robertson PK, Fonseca AV, Ulrich B, Coffin J (2017) Char
Lo Presti D, Stacul S, Meisina C, Bordoni M, Bittelli M acterization of unsaturated mine waste: A case history.
(2018) Preliminary validation of a novel method for the Canadian Geotechnical Journal 54(12): 1752–1761.
assessment of effective stress state in partially saturated https://dx.doi.org/10.1139/cgj2017-0129
soils by cone penetration tests. Geosciences 8(1):1–13. Rocha BP, Rodrigues RA, Giacheti HL (2021) The Flat
https://dx.doi.org/10.3390/geosciences8010030 Dilatometer Test in an Unsaturated Tropical Soil Site.
Lutenegger AJ (1988) Current status of the Marchetti dilat Geotechnical and Geological Engineering. https://dx.
ometer test. In: Proc. of ISOPT-1 penetration testing: doi.org/10.1007/s10706-021-01849-1
special lecture, vol 1, pp 137–155. Balkema, van Genuchten MT (1980) A closed form equation for pre
Rotherdam. dicting the hydraulic conductivity of unsaturated soils.
Machado SL, Vilar OM (1998) Unsaturated soils shear Soil Science Society of America Journal 44(5):892–898.
strength: laboratory tests and expedite determination. https://dx.doi.org/10.2136/
Solos e Rochas 21(2):65–78 (in Portuguese) sssaj1980.03615995004400050002x
676
© 2022 the Author(s), ISBN 978-1-032-31259-0
Study of SPT-CPT and DP-CPT correlations for sandy soils

M.D. Santos
Department of Civil Engineering, Federal University of Espirito Santo, Vitória, Brazil
Department of Civil Engineering, COPPE, Federal University of Rio de Janeiro, Rio de Janeiro, Brazil
K.V. Bicalho
Department of Civil Engineering, Federal University of Espirito Santo, Vitória, Brazil
ABSTRACT: Field penetration tests are widely used in geotechnical engineering for site investigation. The
most common in-situ penetration tests are: the Standard Penetration Test (SPT), the Cone Penetration Test
(CPT), and the Dynamic Probing (DP). Since there are limitations on their test methodologies and equipment,
equations that can describe the relationship between their data are important in geotechnical engineering. This
study presents a review of SPT-CPT and CPT-DP correlations for sandy soils, and discusses the uncertainties
involved in some previously published correlations in literature. It is also recommended adjustments account
ing for variations in soil properties input. Linear regression and residual analysis were carried out in several
experimental data sets from different sandy soils and locations. The observed sandy soils’ heterogeneity
implies in the need of correlations obtained at a regional level. Thus, this study does not aim to exhaust the
subject about in situ penetration tests correlations, but rather to recommend and stimulate the critical evalu
ation and use of previously published correlations and those that will be formulated in the future
1 INTRODUCTION pushing of a 60° conical tip with a 20mm/s velocity.

Although the CPT does not recover samples in depth
Site investigation is one of the most important like the SPT, CPT data are near-continuous with the
aspects of the preliminary design or viability studies oretical background (Robertson 2009). In the piezo
of the geotechnical structures’ safety evaluation. cone test (CPTu) is also possible to obtain the pore
Field penetration tests are widely used in geotech pressure in depth and dissipation tests can be per
nical engineering for site investigation and predic formed do obtain the permeability in situ of materials.
tion of bearing capacity and settlement for The Dynamic Probing test (DP) determines indir
foundation design. The most common in-situ pene ectly the soil resistance by the blow count necessary
tration tests are: the Standard Penetration Test (SPT), to prospect the equipment. Therefore, the execution
the Cone Penetration Test (CPT), and the Dynamic of DP test is similar to the SPT test, in which
Probing (DP). Since there are limitations on their a hammer is loose in free fall and the blows are
test methodologies and equipment, equations that counted during the prospection. The DP is usually
can describe the relationship between their data are used as a complementary test to the SPT test, in
important in geotechnical engineering. verification of compacted fills. The blow count can
The Standard Penetration Test (SPT) is a simple be transformed to a tip resistance (qd) using the
and relatively low-cost test routinely used to subsoil Dutch Formula, which considers the Law of conser
investigation in a variety of ground conditions. SPT vation of energy: the soil resistance must be propor
execution involve the driving of a sampler in the soil, tional to the initial potential energy, before the free
by the repetitive hit of a standard hammer loose in free falling. The ISSMFE (1989) recognizes four types
fall. The SPT test result is known as N-value (NSPT or of DP equipment, which differs in the hammer
N60 when corrected to a hammer energy efficiency of mass, drop height, stem and tip characteristics.
60% as recommended by the ISSMFE (1989). In add Thus, there are the Dynamic Probing Light (DPL),
ition to energy efficiency calculation, other corrections the Dynamic Probing Medium (DPM), the Dynamic
are recommended in the literature in order to reduce Probing Heavy (DPH) and the Dynamic Probing
the spatial and depth variability of the N-Values Super Heavy (DPSH).
(Sitharam & Samui 2007). Empirical correlations between CPT and SPT
The Cone Penetration Test (CPT) obtains the soil results commonly relate the ratio of the CPT tip
tip resistance (qc) and sleeve friction (fs) by the resistance (qc, qt) and SPT blow count N60, to
DOI: 10.1201/9781003308829-99
677
particle/grain size information (mean particle size The influence of depth was also investigated in
D50 or fines content) (Robertson 1990, Santos & SPT-CPT correlations, as soils in greater depth are
Bicalho 2017), soil behavior (Jefferies & Davies more confined and may affect the blown count. This
1993, Mayne 2006), or a constant value for certain correction transforms N60 in to (N60)1, in which the
soil types (Robertson 1990). effective stress is corrected to 1 atm by:
This paper investigates correlations between the
SPT-CPT and DP-CPT results obtained in Brazilian
sandy soils. These correlations allow the evaluation
and expansion of data, minimizing complementary
campaigns which could be onerous and long stand where CN is the correction coefficient for the
ing. Thus, correlations of this type are useful in civil confinement, in this paper adopted equal to Liao
and geotechnical engineering and were studied by & Whitman (1986) method, valid for normal con
many researchers, as Schmertmann (1970), Martins solidated sands and presented in Equation 3. It
& Miranda (2003) and Souza et al. (2012). Nonethe was considered a sand’s bulk density of 20 kN/
less, as correlations are empirical it is important to m³ and water table level at 1.5m.
highlight some of their limitations: they are applic
able only in similar conditions in which they were
elaborated, as test standard and soil type; and statis
tical treatment must be consistent. In this paper,
empirical correlations were obtained using linear The blow count obtained in DPL and DPSH tests
regression with the verification of the residue’s nor were transformed to tip resistance (qd) using the
mality. In SPT-CPT correlation the influence of vari Dutch Formula presented in Equation 4, with M the
ation of relative density (DR) and depth were also hammer’s mass, H the fall height, N the blow count
investigated. obtained in DP test, A the tip probe area, s the pene
trated length, g the gravity acceleration and M’ the
stems’ mass.
2 MATERIALS AND METHODS
SPT-CPT and CPT-DP results empirical correlations

were obtained for sandy soils in the metropolitan
region of Grande Vitória (GV-ES), located in the state
of Espírito Santo (ES), Brazil. The region is geologic SPT-CPT data correlation was written as qc = Kc
ally characterized by a sedimentary formation attrib N60, while DP-CPT correlation was written qc = Ke qd
uted to the action of transgression and regression of (i.e., Kc = qc/N60 and Ke = qc/qd. N60ʹ unit is given in
the sea level during the Quaternary Period. Therefore, blows/300mm while qd and qc are given in MPa).
soils of the area are composed of marine heteroge To form qc x N60 data pairs, qc was taken between
neous sediments, with variable relative density. (Cas the depth of z + 150 and z + 450 mm, to be com
tello & Polido 1988, Souza et al. 2005). pared to the last 300mm which NSPT represents. DP
Tests were driven in shallow subsoil stratum, with data were registered in situ for each 20 cm, while the
depth up to 6m in the saturated zone. The soil was CPT results were registered for every 25 cm. Thus,
identified as clean quartz sands, presenting angular data pairs were taken considering the medium of
and sub-angular grain shape, with less than 5% of results of a tested meter.
fines (< 0.074mm), mean particle size between 0.3 To verify if the correlations are consistent, with
and 0.7mm, and relative density varying from loose a normal distribution of the residue’s as required in
to dense. the Quadratic Leart Squares, graphical and hypoth
Correlations were developed applying linear esis tests were carried out using the R software. Res
regression in N60, qc and qd data. SPT and CPT tests idues indicates data’s natural variation, analytically
were carried out by Brazilian standard proceeding’s: defined in Equation 5:
ABNT NBR 6484:2020 and ABNT NBR
12069:1991 (mechanical CPT were used). There is
not a standard to the DP test in Brazil. So, it was
used a modified DPL and DPSH recommended by
Where ri is the ordinary residue, yi is the observed
ISSMFE (1989).
data and μi is the estimated value in a regression.
NSPT were corrected to a 60% energy (N60) con
To investigate the desired normal distribution
sidering a Brazilian SPT 75% efficiency as follows:
of residues, a graphical approach using histogram
was carried out. Results are satisfactory if the
histogram is symmetrical as illustrated in
Figure 1.
678
density values. D50 in this study ranges from
0.2 mm to 0.6 mm. Kc values tend to increase
with the D50 (Santos & Bicalho 2017).
Figure 1. – Histogram example of data presenting normal

distribution.
The hypothesis tests were carried out using the

Kolmorogov-Smirnov test, which considers the fol
lowing hypothesis:
• H0: data follow a normal distribution
• H1: data does not follow a normal distribution Figure 2. CPT–SPT data correlation and qc x N60 disper
sion graphic.
Hypothesis H0 can be validated if the test result
(DKS) is less than the critical value D’KS given in
Equation 6, for n number of observations. On con
trary, if DKS is higher than D’KS, the normal distri Figure 3 presents the graphical approach to
bution of residues can be rejected with 1α verify the residues’ distribution. Although the histo
confidence. In this study, α was adopted equal gram is not perfectly symmetrical as required, DKS
a 0,05 (95% confidence, which means that the was found to be 0.19, once to a sample size of 34
results are in the interval of 95 of 100 D’KS = 0.23, DKS < D’KS and the normal distribu
observations). tion of residues could be verified. This means that
Kc = 0.44 can be classified as a strong coefficient
with 95% confidence by the Kolmorogov-Smirnov
test.
3.1 SPT-CPT data correlations

Considering 34 data pairs of N60 and qc
obtained in GV-ES sands, the linear regression
was determined and presented in Figure 2, the
coefficient of determination value (R²) was
found to be 85%. The Kc = 0.44 for GV-ES
sandy soils is in reasonable agreement with the
coefficient values found by Folque (1976) where
Kc varies from 0.4 to 0.45; Ajayi & Balogun Figure 3. – Residues’ histogram for qc x N60 correlation.
(1988) where Kc = 0.44; Acka (2002) where Kc
= 0.47; and Mayne (2006) where Kc = 0.438.
However, note that Acka (2002) studied soils Note that in Figure 2 N60 were classified by the
with strong cementation, characteristic not relative density (DR) using the intervals from ABNT
observed in GV-ES sands. The sand relative NBR 6484:2020, in which GV-ES sands were found
density influence could not be observed in the varying from loose to very dense. Table 1 presents
correlations between CPT and SPT results for the variation of Kc in terms of DR values suggested
the investigated sand in the tested site by the Brazilian standards ABNT NBR 6484 (2020).
(Figure 2). Cubrinovski & Ishihara (1999) added Souza et al. (2012) observed that Kc is higher in
the mean particle size D50 in their empirical cor very loose sands and decreases in very dense sands.
relation between N-value and sand relative The tendency of decreasing Kc with the increasing DR
679
Table 1. Variation of Kc values in terms of DR values. Table 3. Variation of Kc with depth (m).
DR NSPT interval* Kc Depth Kc (N60) Kc (N60)1
Blows Blows/300mm m Blows/300mm Blows/300mm

Very loose ≤4 0.84 1.3 0.58 0.29
Loose 5–8 0.49 2.3 0.20 0.14
Medium 9 – 18 0.42 3.3 0.30 0.23
Dense 19 – 40 0.44 4.3 0.16 0.13
Very Dense > 40 0.43 5.3 0.37 0.33
6.3 0.56 0.53
* ABNT NBR 6484 (2020).
could not be observed in GV-ES sands, as Kc slightly 3.2 DP-CPT data correlations
varies from loose do very dense sands. It was worth to For correlations between 66 data pairs of qc and qd,
mention that Kc = 0.84 for very loose sands was found obtained using CPT and DP tests, respectively, the
in a sample of 4 data pairs, and the coefficient must be coefficient Ke = qc/qd was found to be 2.34 (R² =
used with caution, once the sample size is small. 93%). Dispersion graph and histogram are presented in
Table 2 shows the result of the nonparametric Figure 4 and 5 bellow, in which normality could not
Kolmorogov-Smirnov hypothesis test for each DR be verified:
interval investigated herein. As DKS < D’KS in all
cases, all coefficients shown in Table 1 can be classi
fied as strong with 95% confidence. Exception is for
dense soils, where DKS is higher than D’KS and
a normal distribution of residues can be rejected
with a 95% confidence. For very dense soil, DKS is
very close to D’KS and the normal distribution of res
idues can be classified as questionable. Thus, spe
cially in this two cases, the indicated Kc in Table 1
should be used with wariness.
Table 2. Kc x DR – Residue’s analysis.
DR DKS D’KS
Very loose 0.25 0.68

Loose 0.22 0.61 Figure 4. – Dispersion graph for qc x qd correlation.
Medium 0.13 0.35
Dense 0.97 0.45
Very Dense 0.75 0.78
Table 3 shows results of Kc for each tested meter. In

both data treatments (N60 and (N60)1) it was not pos
sible to confirm a clear tendency of decreasing or
increasing Kc with depth. For this purpose, authors
suggest that depths greater than 6.3 m must be
investigated.
Worth to mention that Schmertmann (1970) and
Sanglerat (1972) suggest that the SPT-CPT correl
ation should be independent of depth. Naime & Fiori
(2002) suggest that the independence with depth is
valid only in young soils, but Kc decreases with
depth in mature soils. Figure 5. – Residues’ histogram for qc x qd correlation.
680
The Kolmorogov-Smirnov hypothesis test returned counts considering that are many sources of energy
a DKS of 0.18. For a 66 sample size, D’KS is 0.17. As loss during the fall of the hammer, as losses by fric
DKS ≈ D’KS and the normality of residues is question tion, sound, and thermic energy.
able. Therefore, the correlation must be used with The correlations presented in this study are valid
caution. for sandy soils located in the region of Grande
Gadeikis et al. (2010) recommend a Ke between Vitória-ES, Southwest Brazil. It is important to rec
1.3 to 2.25 using the DPSH, while Czado & Pietras ognize that empirical correlations are limited for the
(2012) found this coefficient between 0.5 to 2.3 using soil type, tests conditions and data treatment. It is
the DPH. Thus, although there is a doubt in the con always recommended a critical evaluation of previ
sistency of GV-ES found Ke, its value is similar to ously published correlations.
those observed in the published literature on this topic.
Waschkowski (1983) suggests that Ke = 1.0 (qd =
qc), which was not verified in this study. Similar ACKNOWLEDGEMENTS
results were reported by Viana da Fonseca (1996),
and the equivalency must be valid only in homoge The authors would like to acknowledge the Brazilian
neous soils, which is not the case of GV-ES sands. agency CAPES for financial support. The second
On the other hand, the researchers believe that the author acknowledges support from the Brazilian
blow counts from the DP tests should be corrected to agency CNPq for the research grant called scientific
an international reference like is done in SPT, as some productivity fellowship.
energy losses can occur during the test, like in friction,
sonorous and thermic energy. This correction should be
made before applying the Dutch Formula, as it con REFERENCES
siders the potential energy before the free fall of the
hammer. In this way, the effect of different types of DP ABNT – ASSOCIAÇÃO BRASILEIRA DE NORMAS
and standards would be minimized in the DP results. TÉCNICAS. NBR 12069. 1991. Solo – Ensaio de pene
tração de cone in situ (CPT). Rio de Janeiro, ABNT. (in
Portuguese)
4 CONCLUSIONS ABNT – ASSOCIAÇÃO BRASILEIRA DE NORMAS
TÉCNICAS. NBR 6484. 2020. Solo – Sondagens de
simples reconhecimento com SPT – Método de Ensaio.
This paper presented SPT-CPT and CPT-DP correl Rio de Janeiro, ABNT (in Portuguese)
ations for sandy soils located in the region of Grande Acka, N. 2002. Correlation of CPT-SPT data from the
Vitória – ES, composed of marine heterogeneous United Arab Emirates. Engineering Geology 67,
sediments, with variable relative density values. pp. 219–231.
SPT-CPT correlations were in form of Kc = qc Ajayi, L.A.; Balogun, L.A. 1988. Penetration testing in
/N60. The GV-ES Kc was found to be 0.44. This tropical lateritic and residual soils - Nigerian experience.
value is similar to those found in literature (Folque Penetration Testing - 1988. Proceedings ISOPT-1,
(1976), Ajayi & Balogum (1988), Mayne (2006)). Orlando, Vol. 1, pp. 315–328. Ed. De Ruiter. A.A.
Rotterdam: Balkema.
The linear regression was validated once the residues
Castello, R.R., Polido, U.F. 1988. Sistematização geotécnica
could be classified in a normal distribution. dos solos quaternários de Vitória, ES. Anais do Simpósio
It was not possible to find a clear tendency of sobre Depósitos Quaternários das Baixadas Litorâneas
variation of Kc with relative density as suggested in Brasileiras: Origem, Características Geotécnicas
Souza et al. (2012), as well as it was found that Kc e Experiências de Obras, Associação Brasileira de Mecâ
does not have a clear dependency with depth vari nica dos Solos, Rio de Janeiro. Tema 3. (In Portuguese).
ation. This last observation is also found in Schmert Czado, B., Pietras, J.S. 2012. Comparison of the cone pene
mann (1970) and Sanglerat (1972). tration resistance obtained in static and dynamic fields
CPT-DP results correlation was presented in form tests. AGH Journal of Mining and Geoengineering.
v.36, n.1.
of Ke = qc/qd. The value of Ke around 2.34 was
Cubrinovski, M. & Ishihara, K. 1999. Empirical correlation
observed for the investigated sands from GV-ES. between SPT N-value and relative density for sandy
Although this value is similar to those found in lit soils. Soils and Foundations, 39(5): 61–71
erature (Gadeikis et al. 2010, Czado & Pietras Folque, J. 1988. Modernas tendências nas técnicas de
2012), by means of the histogram and Kolmorogov- ensaio em Mecânica dos Solos. Informação Técnica
Smirnov hypothesis test the normal distribution of ICT, Geotecnia ITG15. LNEC, Lisboa.
residues are questionable, and the correlation must Gadeikis, S. Zarzoujus, G. Urbaitis, D. 2010. Comparing
be used with caution. CPT and DPSH in Lithuanian soils. In: 2nd International
Waschkowski’s (1983) suggests that Ke = 1.0, but Symposium on Cone Penetration Testing, Huntington
Beach, CA, USA. Volume 2ee: Technical Papers, Ses
it was not verified to GV-ES sands. It can be
sion 3: Applications, 3–22.
observed that supposedly Ke = 1.0 is valid only in ISSMFE - International Society for Soil Mechanics and
homogeneous soils. And, it may be recommended Foundation Engineering. 1989. Report of the ISSMFE
that the blow count from the DP data be corrected to Technical Committee on Penetration Testing of Soils
an international reference value as applied to the TC 16. Reference to Test Procedures CPT-SPT-DP
SPT data. This correction must adjust the blow WST. Swedish Geotechnical Institute Information, n. 7.
681
Jefferies, M., Davies, M. 1993. Use of CPTu to Estimate Santos, MD & Bicalho, KV. 2017. Proposals of SPT-CPT
Equivalent SPT N60, Geotechnical Testing Journal 16, and DPL-CPT correlations for sandy soils in Brazil.
no. 4 pp. 458–468. Journal of Rock Mechanics and Geotechnical. V. 9. n.6.
Liao, S. S. C., & Whitman, R. V. 1986. Overburden Correc pp. 1152-1158
tion Factors for SPT in Sand. Journal of Geotechnical Schmertmann, J.H. 1970. Static cone to compute settlement
Engineering, 112(3),373–377. over sand. Journal Soil Mechanics and Foundations Div
Martins, J.B., Miranda, T.F.S. 2003. Ensaios de Penetração ision, ASCE, v. 96, n. SM3, pp. 1011–1043.
nos Solos Graníticos da Região Norte de Portugal. Sitharam, T.G., Samui, P. 2007. Geostatistical modelling of
Algu1mas Correlações. Portugal. (in Portuguese). spatial and depth variability of SPT data for Bangalore.
Mayne, P.W. 2006. In situ test calibrations for evaluating Geomechanics and Geoengineering. 2 (4). 307–316.
soil parameters. Overview paper, characterization and Souza, C.R. de G., Suguio, K., Oliveira, A. M. dos S.,
engineering properties of natural soils II (proc. Singa Oliveira, P.E. de. 2005. Quaternário do Brasil. Associa
pore Workshop). ção Brasileira de Estudos do Quaternário, ed.
Naime, R., Fiori, A.P. 2002. Variações nas razões qc/N na Holos, São Paulo, Brasil, p. 102–104. (In Portuguese).
região de Passo Fundo, devido a fatores geológicos Souza, J.M.S., Danziger, B.R., Danziger, F.A.B. 2012. The
ambientais. Acta Scientiarum Maringá, v. 24, n. 6, pp. Influence of the Relative Density of Sands in SPT and
1819–1824. (in Portuguese) CPT Correlations. Soils and Rocks, São Paulo, v. 35,
Robertson, P.K.1990. Soil classification using the cone pp. 99–113.
penetration test. Canadian Geotechnical Journal, 27 (1), Viana da Fonseca, A.J.P. 1996. Geomecânica dos solos
pp. 151–158. residuais do granito do Porto. Critérios para dimensiona
Robertson P.K 2009. Interpretation of cone penetration mento de fundações directas. Tese de Doutorado. Facul
tests - A unified approach. Canadian Geotechnical Jour dade de Engenharia da Universidade do Porto, Portugal.
nal 46 (11), pp. 1337–1355N. (in Portuguese).
Sanglerat, G. 1972. The Penetrometer and Soil Exploration, Waschkowski, E. 1983. Le Pénétromètre Dynamique. Bul
Amsterdam, London, New York: Elsevier Publishing letin de Liaison des Laboratoires des Ponts et Chaus
Company. sées, n.125, pp. 95–103.
682
© 2022 the Author(s), ISBN 978-1-032-31259-0
Correlation between SPT and CPT tests in liquefiable deposits
Rashid Shahgholian, Cristiana Ferreira & António Viana da Fonseca

CONSTRUCT-GEO, Faculty of Engineering of University of Porto, Porto, Portugal
ABSTRACT: Cone penetration test (CPT) started being used for soil investigation almost four decades after
the standard penetration test (SPT). Despite their differences, both tests estimate the resistance of the soil against
penetration of a probe. Although it is expected that the results of these tests are comparable and directly correl
ated, different driving mechanisms (static versus dynamic), sensitivity, type and frequency of measured param
eters, operational errors, among others, compromise such correlations. In addition, the diverse nature of soils in
terms of grain size, packing and fabric, shear strength, permeability, and other soil characteristics also limit the
derivation of a unique correlation between CPT and SPT. Many authors have studied these in situ tests, recog
nizing that such correlation would be valuable, as it would enable the use of design methods and other libraries
of correlations available for each individual test. A variety of formulas and charts has been proposed in the
literature to describe this relationship, which demonstrates that this is a complex, not a global and simply pre
dictable correlation. In this research, some of the most usual CPT-SPT relations are evaluated for a large dataset
of field measurements from a pilot site on liquefiable soils, near Lisbon in Portugal. The extensive analysis of
these data showed that the application of previously established, tested, and published correlations, available in
the literature for specific sites, requires great care. One of the requirements is the verification of applicability, by
comparison between the type of soils under study and those in the literature. Specific correlations for each geo
logical condition must be defined before making engineering estimations based on correlated parameters.
1 INTRODUCTION with previous geotechnical studies, namely for the con

struction of a motorway at the same region, providing
Comparison between the results of different geotech a valuable geotechnical data bank. The site is located
nical tests in a site is essential for a correct estimate in the Greater Lisbon region, in the municipalities of
of the geotechnical parameters of the soil profile. The Vila Franca de Xira and Benavente, on the left bank of
Standard Penetration Test (SPT) and the Cone Pene the river Tagus. Boreholes with SPT adjacent to CPT
tration Test (CPT) are two of the most common site tests were carried out along 9 km of the motorway
characterization techniques, which have advantages alignment as shown in Figure 1. More details are avail
and weaknesses over each other. Besides, the litera able in Ferreira et al. (2020).
ture shows that these tests have different libraries of In order to correlate the SPT and CPT results, the
formulas and relations behind. Correlation of these closest (in plan) CPT point to each SPT borehole was
two tests has a history over seventy years and exten selected, defining a SPT-CPT pair (Figure 1). As
sive research indicates there is not a unique relation a result, techniques such as the inverse distance
between them, which may result from the fact that weighted average method (Zhao & Cai, 2015) is not
these penetration tests are differently influenced by used. In the present case of a heterogeneous environ
various factors, including geological conditions, ment, such as recent sedimentary deposits, condensing
available machinery and tools, operator skills, etc. the CPT records to define the soil profile can be par
The application of the empirical equations and charts ticularly challenging.
available in the literature needs to be performed with In fact, the choice of the size of the averaging
great care; a certain level of similarity should be window and its adjustment in terms of the most
assured between the site and the region from which appropriate depth leads to considerable variation in
the empirical correlations were developed. the results. Even relying on just “averaging” may
have negative influence over correlation because it
neglects the fact that CPT records are influenced by
2 METHODOLOGY the stiff layers in front of the tip, even before its
penetration. As Figure 2 schematically shows, tip
Recent investigations on the assessment of liquefaction resistance and sleeve resistance at the same record
hazard around Lisbon, in Portugal, were combined ing of the CPT correspond to adjacent (not identical)
DOI: 10.1201/9781003308829-100
683
The tested soils have been classified according to
the soil behavior type index (SBT) (Robertson &
Wride, 1998); while samples mainly classified as
“clay”, nothing belongs to the “gravel”. The classes
suggested by Cubrinovski et al., (2019) are adopted
in this research. Unified soil classifications, based on
laboratory grading tests are only available for two
boreholes (named SI1 and SI7).
In this paper, some of the correlations between
CPT and SPT tests available in the literature are pre
sented and discussed (Table 1). Selected SPT-CPT
correlations were applied to the existing database to
Figure 1. Selected SPT and CPT locations for this study. evaluate their performance. Finally, the overall trend
of the available data bank is presented and discussed.
depths. On the other hand, the influence of transition

zones is considered to be the same for both the CPT
and SPT tests. In addition, most researchers believe Table 1. Some categories of correlation between CPT and
that the presence of interbedded thin layers will SPT.
decrease the correlation of the N values and qc (after
Reference SPT CPT Other Eq.
Zhao & Cai, 2015).
Jefferies & Davies 1993 N60 qc Ic 3
Lunne et al. 1997 N60 qc/pa Ic 4
Robertson 2012 N60 qc/pa Ic 5
Shahien & Albatal 2014 N60 qc/pa Ic 6
Robertson et al. 1983 N60 qc/pa D50 7
Shahri et al., 2014 N60 qc/pa D50 8
Kulhawy & Mayne 1990 N qc/pa D50 9
Kulhawy & Mayne 1990 N qc/pa FC 10
After Ahmed et al., 2014 N55 qc/pa FC 11
After Shahien & Albatal 2014 N qc/pa FC 12
* N q -
* Numerous authors (see Shahri et al., 2014 & Zhao et al.,

Figure 2. Mismatch of the measuring zone of SPT, with tip 2015)
resistance and sleeve resistance of CPT.
In this research, CPT values (qc and fs) are aver 3 BACKGROUND ON SPT-CPT
aged over the same 30 cm interval of each SPT CORRELATIONS
record, following Jefferies & Davies (1993).
Although CPT probe senses ahead of its tip, it is not 3.1 Ratio methods
a clear mechanism to be involved in calculation of
equivalent CPT exactly over the same 30 cm depth Shahri et al. (2014) cited forty formulas that were
of the SPT. developed between 1959 and 2003. Similar set of
To eliminate the influence of non-equal end formulas is cited by Zhao & Cai (2015). The
effects, qt is used instead of qc in this study (Campa appearance of these formulas is either n = qc/N or
nella et al., 1982). n = (qc+fs)/N, which do not differ considerably
SPT test results, NSPT, are assumed to correspond since sleeve friction (fs) rarely reaches more than
to N60, since the energy ratio measured in Portu 10% of the tip resistance (qc). Depending on the
guese SPT testing campaigns is typically 60%. The soil type, it is shown that usually 0.1<n<1 MPa,
corrections to the SPT records including rod length, although higher values have also been reported or
borehole diameter and effective overburden stress shown in the literature. By normalizing qc with
have been applied to the measured SPT values, for atmospheric pressure (Pa = 101 kPa), the dimen
obtaining the normalized SPT values, (N1)60. sionless domain of n would be 1<n<10.
A total of 285 SPT-CPT pairs of data have been Schmertmann (1978) believed that n=4 is generally
compiled out of the 21 boreholes and their adjacent acceptable for most soils; sensitive clays are an excep
CPT profiles. The maximum measurement depth tion, because SPT in sensitive clays approaches to zero
reaches 26.5 meters, while the majority of the meas and n increases significantly. The range of n is also
urements correspond to depths lower than 21 meters. approved by many recent researches, namely Akca
684
(2003), Shahri et al. (2014), Jarushi et al. (2015) and
Aral & Gunes (2017) among others. Akca (2003)
reported occasionally very high n-values in his
research that may be caused by specific (and less fre
quent) soil conditions, such as cementation, higher
density layers, shell fragments, gravels, and other
heterogeneities.
Recent research confirms the domain of variation
of n but there is considerable scatter in the measure
ments (Aral & Gunes, 2017). Asci et al. (2015) also
Although all these formulas show the correct
reported a trend of n for SPT values below 13, but as
trend of the measurements, scatter of the real meas
SPT values increase, the trend completely differs. urements is considerable. It demonstrates that other
Besides the ratio method, Zhao & Cai (2015) pro
factors also play a relevant role in the correlation.
posed two other correlation categories between SPT
One of those factors may be the difference between
and CPT measurements: “function methods” which
crushable and non-crushable grains (Ahmed et al.,
are more advanced expressions considering more 2014). In natural deposits with different mixture of
than just the ratio, usually based on statistical ana
these components, the measurements would not
lyses but ignoring soil properties; and, “soil param
easily obey a single formula. This shows the
eter methods” which directly address these importance of geotechnical similarity between sites
properties in the correlations. when using the literature correlations. To apply
these properties in correlation equations, various
3.2 Function methods attempts have already been made leading to the
third group of correlations, based on geotechnical
Referring to the investigation by Jefferies & Davies indexes.
(1993), SPT is related to the tip resistance and soil
behavior type index (IC). Equations 1 and 2 show the
expressions for IC according to Robertson & Wride 3.3 Soil parameter methods
(1998) and Jefferies & Davies (1993), respectively.
3.3.1 Grain size
Grain size distribution plays a significant role in
both SPT and CPT test results. Robertson et al.,
(1983) presented a graph showing that n=qc/N
increases as D50 increase. These kind of correlations
are mathematically formulated, as in Equations 7-9
(Shahri et al., 2014). Other similar formulas are cited
by Shahien & Albatal (2014) and Ahmed et al.
(2014). Non-unique formulas for n as a function of
In these equations, Q and Qn are normalized tip
D50 indicates that n is influenced by more than just
resistance. Fr is normalized friction resistance and
D50.
Bq is pore pressure ratio in CPT test.
The correlation proposed by Jefferies & Davies
(1993) is shown in Equation 3. Lunne et al. (1997)
proposed a small modification to the same for
mula (Eq. 4).
Robertson (2012) found that the above listed

equations under-predict N 60 in some clays. He pro 3.3.2 Fines content (FC)
posed the modified formula as in Equation 5. Sha In some cases, SPT and CPT results can be best cor
hien & Albatal (2014) found these equations do not related according to the fines content (Shahri et al.
fit their data bank so they developed Equation 6 for (2014) and Shahien & Albatal (2014)). Equations 10
silty sand deposits in Nile Delta. 12 are samples from the collection.
685
3.3.3 Correlation based on other parameters
Other correlations regarding Dr, Vs, both the D50 and
Ic, etc., have been studied as well. Available data
bank does not cover all the required parameters with
sufficient number of data points so they are not
covered in this text.
Figure 3. Relation between qt and N60.
4.2 Correlation between SPT and Ic (Soil behavior

type index from CPT)
Figure 4 presents the scatter plot of measurements
versus Ic. The distribution of the measurements do
not follow even complicated relations as
Equations 3-5.
4.3 Correlation between SPT and CPT based on

D50 (grain size)
Grain size distribution is only available for bore
holes SI1 and SI7. It refers to 48 cases out of 285 in
this database. Figure 5 shows the measurements,
which do not follow the trends proposed in the
literature.
4 RESULTS AND DISCUSSIONS ON THE

PERFORMANCE OF SPT-CPT
CORRELATIONS
4.1 Ratio of the CPT/SPT values

Figure 3 Shows the scatter plot of the arithmetic
average over CPT records versus SPT values. The
records of the borehole SPT S208 and its relevant
CPT measurements, CPT2, appeared as outliers so
these were ommited here. In order to organise the
database and narrow down the correlation, Cubri
novski et al., (2019) ranges over SBT were used as
the boundaries of different groups of soil. Critical
SBT boundaries from their point of view are: 1.3,
1.8, 2.1 and 2.6, also adopted by Ferreira et al. Figure 4. Scatter of the data using Equations 3-5.
(2020).
Scatter of the measurements seems reasonable
as coarse materials (low SBT values) have higher 4.4 Relation of SPT and CPT based on fines
qt and vice versa but the trend, even in each sub content
group, is not so clear. Frequent measurements Based on the fines content (FC), different correl
have fallen outside the proposed boundaries as ations of SPT and CPT tests are suggested in the lit
well. erature. Figure 6 shows that the measurements from
686
implemented in the normalized parameters as q1N,cs.
An important aspect of these graphs is that none of
the boreholes has been removed, even the ones that
appeared as outliers in previous plots. Figure 7a
shows the best fitted curve for sandy soils while silty
soils (Figure 7b) show an acceptable trend as well.
Figure 8 shows for any specific value of SPT, clean
sand has the maximum value of q1N,cs /(N1)60; as the
fines content increases, this index decreases. Similar
trends can be observed for different soil classes. It is
clear that for finer non-liquefiable soils (Ic>2.6), the
correlation scatter increases considerably. However,
the number of available data points in some of the
best-fitted groups is not very large; more populated
Figure 5. Measured data and published curves in literature data banks should be checked to show how general
(mean grain size, D50, obtained from sieve analysis). such trends are. Other soil classifications may lead to
different trends.
In sum, based on the current data, this approach
should only be applied to granular non-plastic soils.
this database do not follow the trend suggested by
Kulhawy & Mayne (1990).
5 CONCLUSIONS
4.5 The best-fit correlation between SPT and CPT Different scatter plots of the available SPT-CPT data
for the current data bank set were prepared, and trends were compared with
The most suitable graphs to represent the correlation the most common correlations published in the lit
between CPT and SPT in this site are presented in erature. In most cases, the current data does not
Figure 7. In this figure, the clean sand equivalent of follow the published correlations. The characteristics
CPT records is used, which is normalized for the of each site, which are embedded in its correspond
effective overburden pressure (q1N,cs) (Boulanger & ing correlation equation, restrict its validity to the
Idriss 2014). The figure shows q1N,cs/(N1)60 versus most similar sites. The existence of a wide range of
(N1)60 for different groups of soils according to SBT different correlation proposals is demonstrative that
index and Cubrinovski et al., (2019) classification. a universal correlation cannot yet be formulated.
Even though the current data bank compiles
results by different geotechnical companies over
a large period of time (since 2004), the scatter in the
graphs shown in Figure 7 evidence clear trends of
the measurements, providing an adequate level of
confidence in the available data.
Some of the difficulties of correlation between
CPT and SPT profiles may be caused by the proced
ure adopted to condense CPT values over SPT
depths. Non-negligible distances in plan between the
CPT locations and the nearby SPT were found to
have negative influence over the correlation.
All the samples from different boreholes generic
ally follow the trend defined for the relevant soil
class, which means that the proposed trends include
all data points, since none has been considered as
outlier. The trends in coarse granular materials are
Figure 6. CPT-SPT ratio vs. laboratory fines content and clearer than at higher IC classes. However, since this
comparison with a literature proposal. research focused on the characterization of liquefi
able deposits, the obtained trends are within the rele
The trends show a variance, which may be related vant range of applicability. The standard rule of
to the mismatch of the CPT and SPT profiles, due to thumb that granular soils show higher values of
the occasional large distances between the CPT and n (=qn/N) is confirmed in the proposed correlations.
SPT locations in plan. Variation of the soil properties These correlation trends should be controlled with
including grain size distribution, relative density, measurements in other liquefiable sites to assess
fines content, variations in cementation etc., may also whether these correspond to general relations or are
have some effect; however, parts of those effects are just a result of the geological and geotechnical
687
characteristics of this specific site. However, based
on this research, it can be concluded that this
approach to the SPT-CPT correlation is applicable in
granular non-plastic soils, such as liquefiable soil
deposits.
ACKNOWLEDGEMENTS
The first author has received funding from FCT (Por

tuguese Foundation for Science and Technology) in
the form of the 2021.07363.BD. This work was also
financially supported by UIDB/04708/2020 and
UIDP/04708/2020 of CONSTRUCT - Institute of
R&D in Structures and Construction, funded through
FCT/MCTES (PIDDAC).
REFERENCES
Ahmed, S. M., Agaiby, S. W. and Abdel-Rahman, A. H.
(2014) ‘A unified CPT-SPT correlation for non-
crushable and crushable cohesionless soils’, Ain Shams
Engineering Journal. Faculty of Engineering, Ain
Shams University, 5(1), pp. 63–73. doi: 10.1016/j.
asej.2013.09.009.
Akca, N. (2003) ‘Correlation of SPT-CPT data from the
United Arab Emirates’, Engineering Geology, 67(3–4),
pp. 219–231. doi:10.1016/S0013-7952(02)00181-3.
Aral, I. F. and Gunes, E. (2017) ‘Correlation of Standard
and Cone Penetration Tests: Case Study from Tekirdag
(Turkey)’, in IOP Conference Series: Materials Science
and Engineering, p. 9. doi: 10.1088/1757-899X/245/3/
032028.
Asci, M. et al. (2015) ‘Correlation of SPT-CPT Data from
the Subsidence Area in Gölcük, Turkey’, Soil Mechan
ics and Foundation Engineering, 51(6), pp. 268–272.
doi: 10.1007/s11204-015-9288-x.
Boulanger, RW, Idriss, IM (2014). CPT and SPT based
liquefaction triggering procedures. Report No. UCD/
CGM-14/01. Center for Geotechnical Modeling, Univer
sity of California, Davis. 134 pp.
Campanella, R. G., Gillespie, D. and Robertson, P. K.
Figure 7. SPT-CPT correlation for different classes of soil. (1982) ‘Pore pressures during cone penetration testing.’,
Penetration testing. Proc. 2nd European symposium,
Amsterdam, 1982, Vol 2, (May), pp. 507–512.
Cubrinovski, M. et al. (2019) ‘System response of liquefi
able deposits’, Soil Dynamics and Earthquake Engineer
ing, 124(May 2018), pp. 212–229. doi: 10.1016/j.
soildyn.2018.05.013.
Ferreira, C., Viana da Fonseca, A., Ramos, C., Saldanha,
Ana Sofia, Amoroso,S., Rodrigues, C. (2020) ‘Com
parative analysis of liquefaction susceptibility assess
ment methods based on the investigation on a pilot site
in the greater Lisbon area’, Bulletin of Earthquake
Engineering. Springer Netherlands, 18(1), pp. 109–138.
doi: 10.1007/s10518-019-00721-1.
Jarushi, F., Alkaabim, S. and Cosentino, P. (2015) ‘A new
correlation between SPT and CPT for various soils’,
International Journal of Environmental, Chemical, Eco
logical and Geophysical Engineering, 9(2), pp. 101–107.
Jefferies, M. G. and Davies, M. P. (1993) ‘Use of CPTu to
Estimate Equivalent SPT N60’, Geotechnical Testing
Figure 8. SPT-CPT correlation for different soil classes (in Journal, GTJODJ, 16(4), pp. 458–468. doi: 10.1520/
log scale). gtj10286j.
688
Kulhawy, F. H. and Mayne, P. W. (1990) Manual on Esti penetration test: Discussion’, Canadian Geotechnical
mating Soil Properties for Foundation Design, Ostigov. Journal, 35, pp. 442–459.http://www.nrcresearchpress.
http://www.osti.gov/energycitations/product.biblio.jsp? com/doi/10.1139/t99-102.
osti_id=6653074. Schmertmann, J. H. (1978) Guidlines for cone penetration
Lunne, T., Robertson, P. K. and Powell, J. J. M. (1997) test performance and design. FHWA-TS-78-209.
Cone Penetration Testing in Geotechnical Partice, Shahien, M. M. and Albatal, A. H. (2014) ‘SPT-CPT Cor
BLACKIE ACADEMIC & PROFESSIONAL. relations for Nile Delta Silty Sand Deposits in Egypt’,
Robertson, P. K. (2012) ‘Interpretation of In-situ Tests in 3rd International Symposium on Cone Penetration
Some Insights’, in Proceedings of the Fourth Inter Testing. Las Vegas, Nevada, USA, pp. 699–708.
national Conference on Site Characterization. Recife, Shahri, A. A., Juhlin, C. and Malemir, A. (2014)
Brazil, pp. 1–22. ‘A reliable correlation of SPT-CPT data for southwest
Robertson, P. K., Campanella, R. G. and Wightman, A. of Sweden’, Electronic Journal of Geotechnical Engin
(1983) ‘SPT-CPT correlations’, ASCE Journal of Geo- eering, 19 E, pp. 1013–1032.
technical Engineering, 109(11), pp. 1449–1459. doi: Zhao, X. and Cai, G. (2015) ‘SPT-CPT Correlation and Its
DOI: 10.1061/(ASCE)0733-9410(1983)109:11(1449). Application for Liquefaction Evaluation in China’,
Robertson, P. K. and Wride, C. E. (Fear) (1998) ‘Evaluat Marine Georesources and Geotechnology, 33(3), pp.
ing cyclic liquefaction potential using the cone 272–281. doi: 10.1080/1064119X.2013.872740.
689
© 2022 the Author(s), ISBN 978-1-032-31259-0
CPTU-detection of thin clay layers in sand: Results from calibration

chamber tests
H. Skrede
Norwegian Geotechnical Institute (NGI), Norwegian University of Science and Technology (NTNU), Norway
H.B. Hammer
Dr.techn. Olav Olsen AS, Norway
S. Nordal
Norwegian University of Science and Technology (NTNU), Norway
J.-S. L’Heureux
Norwegian Geotechnical Institute (NGI), Norway
ABSTRACT: The detection of thin clay layers (i.e. ≲ 20 cm) is challenging for all conventional geotech
nical field investigations techniques, including high quality CPTU tests. During the last two years, a research
program has been carried out in the geotechnical laboratory at NTNU in Trondheim. The work aims to iden
tify possibilities and limitations in detecting thin clay layers and assess their properties using the CPTU tool.
Tests were run in a pressurized chamber where thin horizontal clay layers of both pottery clay and natural,
sensitive clay were embedded in a homogenous, medium dense sand. Both a standard piezocone (10 cm2) and
a mini-piezocone (5 cm2) have been utilized. The results show to what degree the CPTU response in thin
layers is influenced by the surrounding sand, and how this influence in practice may lead to serious overesti
mation of shear strength in thin layers. In addition, the effect of depth-offset of measurements in connection
with soil type characterization was evaluated.
1 INTRODUCTION et al. (2018), with main focus on liquefaction poten

tial and pile resistance.
Analyses of past landslides along the coast of
Norway have shown that thin clay layers in sandy
shoreline deposits often act as a sliding 2 METHODOLOGY
plane (L’Heureux et al. 2010). An example of
a landslide that caused fatalities with these character In the research program at NTNU, six experiments
istics is the Finneidfjord landslide, which occurred have so far been conducted (E1-E6). Each experi
in 1996 (Longva et al. 2003). The clay layers may be ment consists of constructing a chamber sample,
so thin (i.e. < 20 cm) that even high quality survey pressurizing it, performing soundings on the cham
techniques, alike the CPTU, struggle to detect the ber sample followed up by excavation with supple
layers. mentary laboratory testing of soil properties. Clay
This is the background of an ongoing research type, layer thicknesses and stress levels have been
program at NTNU. A large-scale pressurized model systematically varied during the experimental pro
testing facility has been set up and samples are built- gram. This chapter describes the methodology in
in with clay layers of various thicknesses embedded general, for further details it is referred to MSc
in sand. CPTU-soundings are conducted on chamber theses of Skrede (2021) and Hammer (2020).
samples with the aim of determining if thin layers Chamber samples were built into a testing chamber
can be properly identified, and to which degree it is of about 1.5 meter height, using reinforced concrete
possible to determine the properties of these layers. cylinders (sewage manhole rings) with an internal
This article will mainly focus on the executed labora diameter of 1.2 meter. In the base of the chamber an
tory experiments and the most important results. outlet was installed to allow for regulation of the water
Similar research on the topic has been conducted level. A pore pressure sensor was installed on the
by e.g. van der Linden et al. (2018) and de Lange outlet.
DOI: 10.1201/9781003308829-101
690
The bottom 10 cm of the chamber was filled with Table 1. Properties of pottery clay and clay from Flotten.
gravel, covered by a filter cloth. This allowed for an
evenly distributed in and outflow of water across the Pottery Natural
chamber cross-section during the building procedure, Parameter Symbol clay clay* Unit
preventing interior channelling in the sand. Above the
filter cloth a first sand layer of 10 cm was built in. Pres Water content w ~24 ~45 %
sure cells of the Geokon 3500-3 model were placed on Liquid Limit wL 32.7 ~28-33 %
top, four horizontally and one vertically (Figure 1). The Plastic limit wP 18.8 ~20 %
cells were calibrated by inserting them into a special Plasticity index IP 13.9 ~8-13 %
designed casing and imposing air pressure within the Liquidity index IL 0.4 ~2.4 -
casing. Unfortunately, the cells suffered from some Sensitivity St ~2.5 50-350 -
sensitivity drift. This was dealt with by correcting the Undrained shear su,CAUC 27.5 ~75 kPa
strength
data linearly from start to end.
The sand used in the chamber came from Kvål, * Rough, values based on range of results around 8-10 m
with a d50 of 0.492 mm; Cu of 4.24; and emin and emax depth.
of respectively 0.502 and 0.873. Two types of clays
were used in the tests. The first clay was an industrially
made kaolin pottery clay (K148 by Sibelco) delivered
in bricks (31x11x12cm3). The second clay type was
a natural sensitive clay extracted from 8 to 10,5 meters
depth at Flotten, a Norwegian Geo-Test Site, sampled
by using a mini-block sampler providing cylindrical
samples with 16 cm diameter (Emdal, et al. 2016).
All the pottery clay bricks had almost identical
properties. The very sensitive, quick clay extracted
from Flotten, is well-documented based on-site
investigations reported by L’Heureux et al. (2019).
The key properties of the clays are presented in
Table 1.
The procedure of constructing sand layers aimed at
a homogeneous, medium densely packed sand (relative
density Dr �40-60%) in order to resemble a natural
sand deposit. Details about the procedures are found in Figure 1. Pressure cell arrangement.
the MSc theses (Hammer, 2020; Skrede, 2021).
In all experiments except the last, pottery clay
was utilized. From each brick of pottery clay
a specimen was cut to a preferred dimension with
a thread saw. Several specimens were then laid
together side by side and carefully clamped to form
a clay “unit”. The purpose of uniting several speci
mens was to make the clay layer wide enough to act
as if it was a continuous horizontal layer. It was still
made with limited horizontal extension and did not
cover the entire cross section of the chamber partly
to save material, but primarily to allow water to flow
vertically in the chamber during sample construction
and testing.
In the last experiment (E6) mini-blocks (diameter
of 16 cm) of natural sensitive clay from Flotten were
utilized. Assemblance of units proved to be feasible
also for the quick clay but required considerable pre
cision and careful execution. Each clay layer was
designed and placed so that the CPTU soundings
would be close to the centre of the clay unit as illus Figure 2. Cross-section of chamber with dimensions of
trated in Figure 2 (layer set-up for tests E1-3 both the clay set-up from experiment 1-3 and 5; and of the
and E5). overlaying metal disc with its 10 soundings holes (1S,
Two penetrometers were utilized, firstly a standard 2a-4c). Each of the 3 clay units shown here in grey are built
sized (10 cm2) NOVA-probe from GeoTech AB with from 4 specimens cut from bricks of pottery clay.
a measurement frequency of about 1Hz. Secondly
a mini electrical probe (5 cm2) with an approximate available from Geomil Equipment B.V. This probe was
measurement frequency of 8Hz that was kindly made exclusively used in the last experiment (E6). The rate
691
of penetration was set to 15 mm/s for all experiments u2 with respect to in-situ stress. This provides the
to increase the spatial resolution, thus providing more normalized tip resistance, Qt, the normalized friction
continuous curves, still operating in accordance with ratio, Fr, and the pore pressure ratio, Bq; In Equa
European standard (EN ISO 22476-1:2012). tions (2)-(4) qn is the net cone resistance, see Equa
Further details about the chamber sample set-ups, tion (5).
and about how the signal processing was done to
increase precision are found in the master’s thesis
(Skrede, 2021).
To simulate stress states at larger depths, the sur
face of the chamber samples was subjected to an
over-burden load. The vertical stresses were imposed
by a circular metal disc pushed down by three air-
bellows fastened to an upper supporting metal
framework, see Figure 3.
The chamber was designed with ten possible posi
tions for CPTU testing. Three in-line holes were
placed in 3 radial sectors 120 degrees apart, denoted
sectors a, b, c. The centre hole is referred to as the
S position. The holes were in addition numbered
from the centre as 1,2,3 and 4. Thus position 1S is in
centre while 4c is closest to the concrete wall in
sector c. The reference system is shown in Figure 2.
Multiple test positions were used for each chamber
sample, these were divided into two rounds consist
ing of “primary soundings” and “secondary sound
ings”. The primary soundings consisted of tests with
significant distance to walls and previously tested
positions. These were thus considered to reflect
undisturbed soil. Secondary soundings were run
after the primary soundings in neighbouring posi
tions as reserve for validation, though the measure
ments reflect disturbed soil.
For each chamber sample, after CPTU testing,
a meticulous excavation phase followed, where lab
tests were conducted on both the clay and the sand to
analyse the soil profile for density and strength.
3 THEORETICAL BACKGROUND
3.1 Stress state in chamber – The silo effect

The stress level in the chamber was influenced by
the silo effect or the arching effect caused by vertical
shear stresses on the wall. The formula by Janssen Figure 3. The overburden loading framework on the cham
(1895) was used with effective stress parameters, ber. Above: the actuator and the probe fixed to
Equation (1), to quantify the effect. The decay a framework.
length, l’, was estimated using the given surcharge,
q; and the earth pressure cells at the base, to deter To estimate the undrained shear strength, su,
mine the stress σ’v(z) over the height of the chamber. in the clay layers, the empirical relations sug
gested by Karlsrud et al. (2005) have been used,
see Equations (6)-(8). The cone factors, Ni, in
these formulas are based on OCR, St and IP.
These parameters were estimated for each speci
men by using interpolated data from the Flotten
3.2 CPTU-parametrizations site report (L’Heureux et al. 2019).
The conventional procedure to classify soils from
CPTU results is to normalize the corrected tip resist
ance, qt, the sleeve friction, fs, and the pore pressure,
692
Figure 4. Experiment 5. Left: CPTU-measurements. Mid: Shear strength estimations for the clay layers, note that su,Δu is
approximately zero for all layers. Right: Sand density samples results and estimated stress distributions.
explanation. The sample preparation was quite

successful in experiment 5 with a fairly constant
relative density, Dr, in the sand over the depth of
the chamber. Comparing various tests, it is con
firmed that qt varies considerably with relative
density and stress level.
4.2 Normalized profiles
4 RESULTS To study the variation of qt through layers of varying

thicknesses, a compilation of qt-profiles is shown in
Figure 5. The depth has been normalized with respect
4.1 Soundings from experiment 5 to the cone diameter, dc. The qt-profiles in the
The results of experiment 5 (Figure 4) illustrates graphs are indexed X-Y-Z for respectively experi
typical trends that unveiled upon penetrating clay ment number; surcharge on top of chamber sample
layers interbedded in sand. The clay layer thick in kPa; and sounding hole. The colour of the pro
nesses were respectively 2, 4 and 8 cm from the files is according to the surcharge level, with pri
top. The surcharge was set to 40 kPa for soundings mary and secondary soundings respectively solid
in holes 2a and 2b, 80 kPa for 2c. These were the and dashed. The jagged appearance of the mini-
primary soundings, while secondary soundings cone profiles is due to the different approach used
were performed in the number 3 holes. The latter in post data treatment for the mini cone. The pro
are not presented in this figure. files include a penetration through a 36 cm high
The u2-measurements were depth-shifted to pottery clay unit as a reference for the characteris
account for the distance between the position of tic tip resistance, qtchar, of the pottery clay (E4-40
the pore pressure sensor and the cone tip (the 3b). Here qtchar is defined as the expected tip
depth of qt), while fs was corrected to the end of resistance in an infinitely thick layer for the spe
the sleeve closest the tip, see ch. 5.2 for cific stress level.
693
5 DISCUSSIONS 5.2 Depth offset
During penetration through thin layers of clay,
5.1 The thin layering effect all CPTU-parameters starts to approach their
To what degree the measured qt in the thin layers characteristic value, yet these approaches are
is higher than qtchar depends on the ratio of the “cancelled” upon closing in on the second inter
layer thickness over the cone diameter, H/dc, and face. As measurements are saved as data points
on the contrast between qtchar in the clay and the versus time, the measurements from the pore
tip resistance in the sand. The effect H/dc is shown pressure sensor are actually made about 1 cone
in Figure 5 where the mini-cone profiles from E6: diameter behind the cone tip. This means that
4b and 4c have equal sensing- and developing the extremal value of u2 is saved at a different
depths compared to the standard-cone profile 1S, depth than the extremal value of qt, conse
when normalized. The “depth terms”, sensing- and quently lowering the extremal value of the pore
developing depths, denote respectively the dis pressure ratio, Bq, see Figure 6. For the sound
tance ahead of an interface where qt is influenced ings on the natural sensitive clay (E6-1S;4b;4c),
by the next layer; and the distance after the inter the effect of correcting the depth-offset altered
face qt retain influence from the previous layer. the Bq extremal value up to 0.15. Moreover, due
The test results show that the qt-profiles to the short distances of excess pore pressure
through interbedded layers are not symmetric, as build-up in thin layers, Bq never approach levels
the developing- and sensing distances are differ which are close to the reference values, requir
ent for the upper and lower interface of each ing extra care upon interpreting CPTU.
clay layer, as apparent in Figure 4 and Figure 5. Accounting for the depth-offset is more compli
The characteristics of qt-profiles during transi cated for the sleeve friction measurements, as each
tions thus have some degree of uniqueness, measurement is the product of stresses working on
which then might be used in an attempt to back- the entire friction sleeve. Therefore, it can be stated
calculate layer thicknesses and qtchar. that the sleeve’s length entails a smoothening
The tip resistance approaches the characteristic effect, making it a component of less depth accur
value asymptotically with increasing interbedded acy and maybe relevance regarding transitions.
layer thicknesses. This implies that calculated su However, from the experiments, a consistent pat
based on Equation (6) and (8) provide overestim tern is apparent: As the probe transits from the sand
ations of the “true su” (i.e. su based on triaxial to the clay the sleeve friction increases drastically
CAUC-test) whenever the developing depth is not when the front end of the sleeve hits the original
surpassed. As to illustrate, overestimation magni level of the interface (prior to deformation), where
tudes (with reference to calculated extremal value) upon it is decreasing (ref. E4-40-3b). Due to this
are listed for all experiments on pottery clay in phenomenon fs may turn out to be quite useful for
Table 2. The table does in addition include the identifying thin layers of clay embedded in sand.
magnitude of underestimation when calculated su However, this part is associated with great uncer
is based on Equation (7), which is a consequence tainty as field measurements of fs can be rather high
of the excess pore pressure build-up never in sand, while low in clay, and in addition, fs varies
approaching the characteristic value. From the ref a lot with probe types. Consequently, this aspect
erence test (E4-40-3b), the developing depth of qt requires further research.
was about 5-5.3 dc (18-19 cm) after the interface. With regards to the previous paragraph, the
choice of depth correction for fs is not obvious.
Upon correcting the data points upwards on the
sleeve, the extremal magnitude of Fr is smooth
Table 2. The ratios between the extremal value of the ened with respect to both lower and upper
CPTU-su-parametrizations within a clay layer compared to bounds, making a distinction between materials
the “true su” of the clay, presented for different clay layer more difficult (Figure 7). Ironically and concep
thicknesses. For su,kt and su,ke the ratio of overestimation is tually incorrect, the greatest magnitude of Fr is
presented, while for su,Δu the ratio of underestimation is reached if no corrections are made at all.
presented. In conclusion, depth-offset has a big impact on
Layer thickness 2 cm 4cm 8cm 12 cm
the classification parametrizations during transitions,
greatly impacting the identification of the materials
su,kt,min/su 6.3-10.5 2.3-4.6 1.4-2.1 1.1 and the layer thicknesses. It should however be
su,ke,min/su 5.0-7.7 2.0-3.5 1.3-1.7 0.9-1.0 underlined that offset corrections do not influence su
su/su,Δu,max 80-125 35-65 4-25 1.8-3.4 -estimations except for su,ke with a negligible
magnitude.
694
Figure 6. Relative difference of Bq when depth-offset is
corrected (left), and not corrected (right) (E6-1S;4b;4c).
Figure 7. Left: Profile with fs-readings (E6-1S) not cor

rected with respect to depth offset (blue) and profiles
adjusted for different positions on the friction sleeve. Right:
Resulting Fr-graphs based on data from the graphs in the
left profile.
6 CONCLUSIONS
Figure 5. Rotated tip resistance profiles of penetration of
clay layers (surface towards left, chamber base towards
right), normalized with respect to the cone diameters. The Thin, weak and possibly sensitive interbedded clay
thicknesses of the layers penetrated by respectively the layers are not easily detected as characteristic reference
standard cone and the mini-cone were 2, 4, 8 and 12 cm; values are not reached before the cone leaves the layer.
and 2.8, 5.7 and 8.5 cm. Soundings in natural sensitive clay In this respect, depth correction proves to be very
are marked with orange (E6). important for the magnitude of the normalized ratios Fr
695
and Bq. Furthermore, calculations of su based on the issmge.org/publications/publication/cptu-correlations-for
correlations su,kt and su,ke will lead to overestimation of clays
su as the recorded tip resistance does not approach the de Lange, D.A., Terwindt, J. and van der Linden, T.I. 2018.
characteristic tip resistance of the thin layer. Likewise, CPT in thinly inter-layered soils. Paper presented at the
calculation of su based on the correlation su,Δu will lead Cone Penetration Testing 2018: Proceedings of the 4th
to severe underestimation due to the small excess pore International Symposium on Cone Penetration Testing
pressure build-up in thin layers. (CPT’18), 21-22 June, 2018, Delft, The Netherlands.
https://www.icevirtuallibrary.com/doi/10.1680/jgeen.17.
00061
L’Heureux, J. S., Hansen. L., Longva, O., Emdal, A., &
ACKNOWLEDGEMENTS Grande, L. O. 2010. A multidisciplinary study of submar
ine landslides at the Nidelva fjord delta, Central Norway,
The authors would like to acknowledge everyone Implications for the assessment of geohazards. Norwegian
who have contributed to this research program, Journal of Geology, Vol 90: 1–20.Trondheim, ISSN 029
mainly lab staff at NTNU. We are grateful to NGI 196X. https://www.researchgate.net/publication/234101
for financially aiding the program and to Geomil 112_A_multidisciplinary_study_of_submarine_landslide
Equipment for generously lending the mini-cone to s_at_the_Nidelva_fjord_delta_Central_Norway_-_Implic
the research. ations_for_geohazard_assessment
L’Heureux, J.S., Lindgård, A., Emdal, A. 2019. The
Tiller-Flotten research site: Geotechnical characteriza
tion of a very sensitive clay deposit. AIMS Geosciences,
REFERENCES 5(4):831–867. https://ngi.brage.unit.no/ngi-xmlui/hand
le/11250/2630879
Emdal, A., Gylland, A., Amundsen, H.A., Kåsin, K., Longva O., Janbu N., Blikra L.H., Bøe R. 2003. The 1996
Long, M. 2016. Mini-block sampler. Canadian geo Finneidfjord Slide: Seafloor Failure and Slide Dynamics.
technical journal, 53(8):1235–1245. https:// In: Locat J., Mienert J., Boisvert L. (eds.), Submarine
cdnsciencepub.com/doi/10.1139/cgj-2015-0628 Mass Movements and Their Consequences. Advances in
Hammer, H.B. 2020 Physical experiments on CPTU thin- Natural and Technological Hazards Research, vol 19.
layer effects of thin clay layers embedded in sand: With Springer, Dordrecht. https://link.springer.com/chapter/
analysis-and possible correction of cone resistance in 10.1007/978-94-010-0093-2_58
layered profiles. M.Eng. Master Thesis. Faculty of Civil Skrede, H. 2021 CPTU-detection of thin clay layers in sand:
Engineering, Norwegian University of Science and Results from calibration chamber testing. M.Eng. Master
Technology (NTNU). https://ntnuopen.ntnu.no/ntnu Thesis. Faculty of Civil Engineering, NTNU. https://
xmlui/handle/11250/2689484 ntnuopen.ntnu.no/ntnu-xmlui/handle/11250/2977001
Janssen, H.A. 1895. Versuche über Getreidedruck in Silozel van der Linden, T.I., De Lange, D.A., & Korff, M. 2018.
len. Zeiteschrift des Vereins deutscher Ingenieure. 39 Cone Penetration Testing in Thinly Inter-Layered Soils.
(35):1045–1049. https://web.archive.org/web/2011030322 Geotechnical Engineering. https://www.taylorfrancis.com/
3406/http:/www.phy.duke.edu/~msperl/Janssen/Janssen chapters/oa-edit/10.1201/9780429505980-51/cpt-thinly-int
1895.pdf er-layered-soils-de-lange-terwindt-van-der-linden
Karlsrud, K., Lunne, T., Kort, D.A., Strandvik, S. 2005. European committee for standardization. 2012. Geotech
CPTU correlations for clays. In Proceedings of the inter nical investigation and testing: Field testing: Part 1:
national conference on soil mechanics and geotechnical Electrical cone and piezocone penetration test. EN ISO
engineering, 2005, Osaka,16:693–702. https://www. 22476-1: 2012.
696
© 2022 the Author(s), ISBN 978-1-032-31259-0
Determination of hydraulic conductivity using HPT & CPTu

M. Slowiok, S. Oberhollenzer & R. Marte
Institute of Soil Mechanics, Foundation Engineering & Computational Geotechnics, Graz University of Technology,
Graz, Austria
T. Freudenthaler
mjp ZT GmbH, Saalfelden, Austria
ABSTRACT: In geotechnical engineering insitu tests are becoming increasingly popular to determine the
hydraulic conductivity of soils. Direct push methods such as piezocone penetration test CPTu or the hydraulic
profiling tool HPT allow characterizations along a linear profile. The determination of hydraulic conductivity
with HPT is an alternative to CPTu for drained and partially drained conditions. Thereby, water is continu
ously injected from a small screened port at the probe into the soil. Based on the required pressure to inject
the water into the soil and the constant flow rate, the hydraulic permeability can be determined for a certain
range of drainage conditions. The present article compares HPT and CPTu results with slug tests, executed at
different test sites in a wide range of grain size distributions. Based on this comparison, a new correlation was
elaborated for HPT which enables an improved characterization of hydraulic conductivity.
1 INTRODUCTION the subsoil. Based on the required pressure to ensure

constant flow, K can be determined based on correl
The hydraulic conductivity K of soils is an important ations, which were often calibrated based on slug-
soil parameter for a wide range of geotechnical prob tests (and might rely therefore on site-specific data).
lems (e.g. excavations in combination with dewater In the present article different insitu & laboratory
ing works). In order to determine K, the fluid tests are compared with reference to the hydraulic con
velocity v and the hydraulic gradient i can be used ductivity. The main focus is related to direct push
based on Darcy’s law. Hydraulic conductivity can methods (HPT, CPTu) in order to characterize their
vary between approximately ten orders of magnitude application limits. In a final step, a new correlation to
(K = 10-2 – 10-12 m/s) and is strongly influenced by determine K is elaborated for HPT based on additional
the grain size distribution, density, void ratio and slug tests.
anisotropy. Furthermore, it was shown earlier that Four test sites, carried out within Austria (state
the horizontal hydraulic conductivity can become up Salzburg), represent the basis of the present work.
to ten times larger compared to the vertical one in The investigation program includes CPTu, HPT,
sediment deposits (Boley 2019). slug-tests and additional laboratory tests. All applied
In the laboratory, the vertical hydraulic conductiv methods (insitu & laboratory tests) are described in
ity Kv is determined based on permeameter tests with the following section.
constant or variable pressure head. In general, the
quality of sampling has a great influence on the
determined permeability. On the other hand, direct 2 APPLIED METHODS
push (DP) methods are time- and cost-efficient alter
natives, which became more popular over the last 2.1 Piezocone penetration test (CPTu)
decades. Especially piezocone penetration tests
(CPTu) enable a determination of K based on the soil The piezocone penetration test (CPTu) represents
behavior type index Ic or dissipation tests. Thereby, a widely used direct push technique for soil charac
the determined permeability - in contrast to the terization. Thereby, a penetrometer is pushed at con
laboratory tests - is a combination of the horizontal stant rate (equal to 2 cm/s) into the subsurface.
and vertical one. The hydraulic profiling tool (HPT) Simultaneously, the tip resistance qc, sleeve friction
represents an alternative investigation technique and fs & dynamic pore water pressure ui are measured at
enables a continuous characterization over depth. the cone continuously over depth. The friction ratio
Thereby, water is injected under constant rate into Rf = fs/qt (%) can be used as a first indicator for
DOI: 10.1201/9781003308829-102
697
characterizing the soil behavior type. Thereby, the where Kh represents the horizontal hydraulic con
corrected cone resistance qt is defined as ductivity, M the 1D constrained modulus at the insitu
stress level and γw the unit weight of water. In order
to back-calculate Kh, ch & M must be calculated
using existing correlations. Based on Teh & Houlsby
(1991) or Robertson et al. (1992) ch (m²/s) can be
where a is the area ratio of the cone and u2 is calculated based on t50 (time for 50 % of excess pore
the dynamic pore water pressure measured behind water dissipation).
the tip. The cones used for all field investigations
present a cross-section area equal to 15 cm² and
the porewater pressure was measured at pos
ition u2.
The hydraulic conductivity can (mainly) be
determined based on two approaches: The first for 15 cm² piezocones.
method is based on the soil behavior type classifica The constrained modulus M can also be deter
tions and provides a continuous K-profile over mined based on Robertson (2009) using:
depth. These classifications can be estimated based
on the soil behavior type index Ic (Robertson &
Wride 1998).
when Ic > 2.20:
where Qtn is the normalized cone resistance
2.2 Hydaulic Profiling Tool (HPT)

with the total vertical stress σvo and the effective ver
tical stress σ’vo and the normalized friction ratio Fr The Hydraulic Profiling Tool belongs to the direct
push methods too. A probe is pushed into soil under
constant penetration rate equal to 2 cm/s. Simultan
eously, water is injected at constant flow rate via
a small port at the side of the probe. A pressure
The soil behavior type index Ic is defined by the sensor inside the probe (near the outlet) records the
radius of concentric circles within the soil behavior required injection pressure while flow remains con
type diagram according to Robertson (2010). Higher stant. Based on the ratio between flow rate and pres
Ic values represent soils with higher fines content & sure the hydraulic conductivity can be estimated in
lower hydraulic conductivity. The empirical correl combination with correlations. In a first step the
ation to characterize the hydraulic conductivity in m/s measured pressure ptot is corrected by the atmos
based on Ic is shown in equations 5 and 6: pheric pressure patm and the piezometric pressure
ppiezo. The piezometric pressure can be calculated
based on dissipation tests for selected depths.
In this study two approaches (McCall & Christy

The second approach to determine the hydraulic con (2010), Borden et al. (2021)) are used to calculate
ductivity is based on CPTu-dissipation tests, measur the hydraulic conductivity.
ing the pore water pressure decay after stopping the The first (empirical) correlation according to
penetration process. The dissipation process is McCall & Christy (2010) was calibrated based on
mainly governed by the horizontal coefficient of slug tests:
consolidation, defined as:
698
It should be noted that the approach according to
McCall & Christy (2010) is limited to K-values
between 2.65⋅10-4 and 3.53⋅10-7 m/s.
The enhanced correlation according to Borden
et al. (2021) considers Q (mL/min), Pcorr (kPa) and Furthermore, the grain-size distribution was ana
additionally the penetration rate v (mm/s) as well as lyzed for each soil sample in order to enable classifi
probe diameter d (mm), cations based on Chapuis (2012).
3 TEST SITES
Insitu tests and additional soil sampling was carried

out at four test sites within the state of Salzburg, Aus
In addition, the HPT tool offers the measurement of
tria. The investigated soils at Flachau (TS1), Sigger
electrical conductivity (EC) using a four-point
wiesen (TS2), Bruck/Großglocknerstraße (TS3) and
Wenner array near the tip of the probe. Based on this
Salzburg (TS4) are characterized by a wide range of
measurement qualitative estimations regarding the
different grain size distributions (see Figure 1).
permeability are possible. An increased electrical
Test site 1 (TS1) mainly consists of coarse-
conductivity is an indicator for a higher fines content
grained deposits which are interrupted by silty
and lower hydraulic conductivity.
layers. For TS2 Siggerwiesen (SIG) the subsoil is
composed of a heterogeneous aquifer which is fol
2.3 Direct Push Slug Test (DPST) lowed by a low permeable silty fine sand. The top
layer (0-6m) at TS3 is classified as peat and is under
Slug tests are a reliable and cost-efficient technique
laid by sand-silt mixtures. The fourth testing site is
to determine the permeability of soils insitu. First, an
situated close to the main train station of Salzburg.
excess porewater pressure is created artificially,
The subsoil conditions are characterized by 3 main
before measuring the drop in porewater pressure
lithologies: sandy gravel (0-4m), sand-silt alterations
over time.
(4-9m) and clayey silts (9-25m).
The change in water head can be created by adding
CPTu and HPT-soundings were carried out at all
or removing water within the rods. In our case, the
four test sites. Additional slug tests were carried out
slug test was performed using direct push equipment.
at three depth levels on each site. At TS3 and TS4
After pushing the penetration rod into the desired
additional soil sampling (for permeameter tests and
depth, the penetration procedure is stopped and a filter
determination of grain size distribution) was executed.
body (4/8 mm gravel) is created by pulling the rods
about 50 cm. While pulling the rods, a lost tip remains
in the soil and the filter material is poured through the 4 RESULTS
rod to create a filter length equal to 50 cm.
When the filter section is created, a water head is CPTu and HPT measurements (qt, Rf, EC, Pcorr) are
applied by filling a defined volume of water into the presented for the different test sites in Figure 1.
penetration rods. Afterwards, the pressure drop is A low corrected tip resistance qt and high friction
recorded over time by a sensor. In each depth at least ratio Rf represent fine-grained layers. As mentioned
two slug tests were performed and analyzed in com earlier, fine grained sections, presenting a low per
bination with the software-package SlugIn (IGME) meability, are characterized by a high electrical con
by using the interpretation methods according to ductivity EC and high Pcorr values.
Cooper et al. (1967), Hvorslev (1951) and Bouwer- TS1 – Flachau consists of a fine-grained layer
Rice (1976). between 2 and 3 meters, which can be detected
based on qt, Rf, EC and Pcorr. A second silt-domin
2.4 Laboratory test with constant pressure head ated lens follows between 5 and 6 meters. The lower
and grain-size analysis aquifer (6-10m) presents low friction ratios Rf and
small Pcorr values, indicating a high permeability.
At two test sites additional soil sampling was exe At TS2 – Siggerwiesen backfill material is present
cuted using the CPT-Ranger system. The recovered between 0 and 4 meters which is followed by a silty
samples present a diameter d and height h equal to fine sand layer with increasing fines content.
6 cm and 50 cm respectively. Each sample was The subsoil at TS3 – Bruck/Großglocknerstraße
trimmed down to d = 5 cm and h = 12 cm to enable consists of a peat layer within the first 5 meters, fol
permeameter tests. The tests were carried out with lowed by sand mixtures. Within the top layer, the fric
a constant pressure head of Δh = 3.6 m. Conse tion ratio Rf amounts to approximately 10 %. On the
quently, the vertical hydraulic conductivity can be cal other hand, HPT pressures Pcorr indicate rather high
culated based on the measured discharge Q over time permeabilities within this layer. The following sandy
t, the hydraulic gradient i = Δh/Δl and cross-section section (7 – 20m) presents increased qt, and small Rf
area A. values. Pcorr & EC present a rising trend over depth.
699
Figure 1. Results from Direct Push stratigraphic profiling over depth (m) – tip resistance qt (MPa), friction ratio Rf (%),
electrical conductivity EC (mS/m) & corrected HPT pressure Pcorr (kPa).
The subsoil at TS4 – Salzburg is composed of For TS1 – Flachau hydraulic conductivity based on
backfill material within the upper 4 meters. The fine- Ic yields to the lowest results. Both HPT-correlations
grained “Salzburger Seeton” (9 to 25m) is character lead to similar results within the same order of magni
ized by a small qt and increased Rf, EC- as well as tude. Slug tests are evaluated based on Hvorslev
Pcorr-values. (1951). The latter results are situated between CPTu
In a second step the hydraulic conductivity was (Ic, dissipation test) and HPT results. The slug tests
evaluated based on CPTu (Ic, dissipation test), HPT, identify the impermeable lens in 5 meter depth. How
slug tests and laboratory tests (permeameter test, grain- ever, the sensitivity of the different tests differ
size distribution). Figure 2 presents the determined strongly. Slug test data verify the soil layering and are
K values based on different interpretation techniques as situated between CPTu and HPT results.
well as test sites. A continuous K profile over depth At TS2 – Siggerwiesen the high permeable aquifer
can be evaluated from CPTu (Ic) and HPT data. Select in a depth of 4 meters is recognized by all methods.
ive permeability tests in certain depths complement the The fine grained layer from 4 to 20 m indicates low
test program and are used for comparison. permeablilty for both CPTu methods. The slug tests
700
Figure 2. Hydraulic conductivity K (m/s) from different approaches over depth (m) – cone penetration test soil behavior
type index CPTu-Ic, Hydraulic Profiling Tool HPT, Direct Push Slug Test DPST, cone penetration dissipation test CPTu
Diss, Laboratory permeameter test Lab K & Grain Size Distribution correlation GSD K.
suggest a higher permeability whereas McCall & CPTu-Ic and laboratory results indicate rather low
Christy (2010) HPT approach reach the lower K limit. hydraulic conductivity within this section.
The continuous readings of TS3 – Bruck/ In the last step, existing HPT correlations where
Großglocknerstraße yields to similar trends: CPTu- evaluated. McCall & Christy (2010) suggested for
Ic results lead to the lowest and both HPT-correl their correlation a lower K limit of 3.5⋅10-7 and
ations lead to the highest hydraulic conductivity a higher K limit equal to 2.7⋅10-4 m/s. Their loga
values. The organic layer within the upper 5 meters rithmic approach (see Figure 4) was calibrated
present a high scatter in K (up to five orders of mag
nitude). For the lower silty sand layer, the scatter
becomes smaller.
At TS4 – Salzburg the same trend (between CPTu
and HPT results) is given. Again, the difference in
K rises in fine-grained layers. Within the so-called
“Salzburger Seeton” slug tests indicate a higher per
meability, whereas laboratory and CPTu (dissipation,
Ic) results lead to smaller conductivities.
In order to investigate the scatter in K, all insitu
and laboratory results are compared within Figure 3.
The scatter was evaluated considering the soil
behavior type index Ic. To enable a comparison, the
data from CPTu and HPT was assigned over an inter
val equal to 50 cm meters using the median. Vice
versa “punctual” K measurements (e.g. dissipation
test) were assigned to a median Ic.
As shown in Figure 3, with increasing fine-content
the difference in K rises strongly. For small Ic values
(indicating coarse-grained material) the variation of the
different interpretation techniques is approximately one
order of magnitude. As mentioned above, HPT-correl
ations yield the upper K-end, whereas K values based
on CPTu-Ic mark the lower K-end. For sand-silt mix
tures the discrepancy increases. This trend continues
with increasing fines-content, where HPT-correlations Figure 3. Hydraulic conductivity K (m/s) vs. Soil Behavior
reach the end of applicability. CPTu dissipation, Type Index Ic (-).
701
5 SUMMARY
The objective of this study was to compare and evalu

ate different insitu and laboratory tests in order to
estimate the hydraulic conductivity. Based on investi
gations at 4 test sites, it was shown that direct push
techniques (CPTu, HPT and Slug test) lead to realistic
K-results and are reproducible. In addition, it was
shown that different investigation techniques present
a stronger scatter with increasing fines content. Here,
K-values from CPTu-Ic mark the lower boundary,
whereas K-values from HPT-correlations mark the
upper boundary. It was proven, that existing HPT cor
relations (McCall & Christy 2010, Borden et al. 2021)
should only be used between K = 1⋅10-6 and 5⋅10-4
m/s, which is for many geotechnical projects (e.g.
dikes, dewatering) sufficient. Both HPT correlations
are realistic representations to determine K by Q/Pcorr.
In addition a new HPT correlation was presented.
REFERENCES
Boley, Conrad. 2019. Handbuch Geotechnik. Grundlagen –
Anwendungen - Praxiserfahrungen. Springer Fachme
Figure 4. HPT correlations in comparison with slug test dien Wiesbaden: Springer Vieweg
and CPTu-Ic in K (m/s) vs. Pcorr (kPa) space. Borden, Robert C., Cha, Ki Young, Liu, Gaisheng. 2021.
A Physically Based Approach for Estimating Hydraulic
Conductivity from HPT Pressure and Flowrate. In
Ground water 59 (2).
based on slug tests and exhibits an asymptote for Bouwer, Herman & Rice, R. C. 1976. A slug test for deter
high Pcorr. At the low K end permeability changes mining hydraulic conductivity of unconfined aquifers
strongly along a small range of Pcorr values. On with completely or partially penetrating wells. In Water
Resour. Res. 12 (3): 423–428.
the other hand, Borden et al. (2021) defines Chapuis, Robert P. 2012. Predicting the saturated hydraulic
a straight line within the K-Pcorr space. In com conductivity of soils: a review. In Bull Eng Geol
parison with the available slug test and CPTu-Ic Environ 71 (3): 401–434.
data, the correlation according to McCall & Cooper, Hilton H., Bredehoeft, John D.,
Christy (2010) is in good agreement with the Papadopulos, Istavros S. 1967. Response of a
results. It should be noted, that for Pcorr values finite-diameter well to an instantaneous charge of water.
between 10 and 100, the correlation leads to rather In Water Resour. Res. 3 (1): 263–269.
high permeabilities. On the other hand, Borden Hvorslev, M. J. 1951. Time lag and soil permeability in
et al. (2021) leads to significantly (unrealistic) ground water observations. Vol. 36.
IGME. SlugIn 1.0. Slug Test Analysis Software. Version
high results. 1.0: Instituto Geologico y Minero de Espana (IGME).
In order to achieve a better agreement with McCall, W. & Christy, T. M. 2010. Estimating Formation
our test results two best-fit curves were defined. Hydraulic Conductivity (K) from HPT Q/P Ratios
The new correlation is based on a potential func (abstract). Proceedings of the 2010 North American
tion which is further subdivided: Environmental Field Conference. Socorro.
Robertson, P. K., Sully, J. P., Woeller, D. J., Lunne, T.,
Powell, J. J. M., Gillespie, D. G. 1992. Estimating coef
ficient of consolidation from piezocone tests. In Can.
Geotech. J. 29 (4): 539–550.
Robertson, P. K. & Wride, C. E. 1998. Evaluating cyclic
liquefaction potential using the cone penetration test. In
Can. Geotech. J. 35 (3): 442–459.
Robertson, P. K. 2009. Interpretation of cone penetration
tests — a unified approach. In Can. Geotech. J. 46 (11):
1337–1355.
Robertson, P. K. 2010. Estimating in-situ soil permeability
from CPT & CPTu. 2nd International Symposium on
Cone Penetration Testing.
Teh, C. I. & Houlsby, G. T. 1991. An analytical study of the
cone penetration test in clay. In Géotechnique 41 (1): 17–34.
702
© 2022 the Author(s), ISBN 978-1-032-31259-0
DEM-FDM coupling simulation of cone penetration tests in a virtual

calibration chamber
Y. Song, X.Q. Gu & J. Hu
Tongji University, Shanghai, China
ABSTRACT: We performed numerical simulation of the cone penetrating process with discrete element
method (DEM) coupled with finite difference method (FDM) to investigate the mechanics of granular soil
interacted with a cone penetrometer. The efficacy of particle size amplification method to estimate soil phys
ical indexes has yet to be established and it was validated in this work. The effects of cone-particle chamber
size, confining stress and sand modulus on cone tip resistance are analyzed. The results show that the effect of
confining stress is consistent with the empirical relationship and the cone tip resistance correlates positively
with the number of particles contacting with the cone penetrometer. Furthermore, the distribution of radial
stress agrees well with the cavity expansion theory that a nearly linear relation between cone tip resistance
and radial stress near the cone penetrometer can be observed.
1 INTRODUCTION models to simulate the stress-strain characteristics of

sand, and the large distortion of finite element mesh
Cone penetration testing can be used for soil classifi of the soil near the cone-tip does not guarantee legit
cation, determination of soil physical and mechan imate numerical results. The indoor calibration
ical parameters, evaluation of pile bearing capacity chamber test may present different results due to the
and liquefaction site discrimination (Yu & Mitchell, size effect than those from the in-situ test
1998). Till now, the relationship between tip resist (Wesley,1998). Therefore, the model size effect
ance and soil parameters is established mainly needs to be investigated cautiously in this regard.
through bearing capacity theory, cavity expansion In this work, the discrete element method coupled
theory, finite element simulation or cone penetration with finite difference method is used to simulate the
calibration chamber test. Chen & Juang (1996), cone penetration test, which can effectively analyze
Sokolovskii (2016) proposed a formula to calculate cone penetrator-soil interaction mechanics while
the cone probe coefficient based on the bearing cap considering the large deformation characteristics of
acity theory and developed an empirical relationship soil near the cone-tip. In order to consider the shape
to determine the volumetric strain. The bearing cap effect of sand particles, the rolling-resistant linear
acity theory fulfill the equilibrium equation and yield contact model is used in the simulation. The contact
criterion, however the kinematics and initial stress parameters were calibrated in accordance with the
state around the penetrometer are disregarded, and triaxial drainage testing result. Effects of different
the stiffness and compressibility of soil are not care contact parameters on soil properties were briefly
fully accounted. The resulting prediction of cone tip discussed and pertinent suggestions were given for
resistance could be spurious. Compared with the the choice of model parameters.
bearing capacity theory, the cavity expansion theory
not only considers the elastic deformation of soil
during penetration, but also considers the plastic 2 COUPLED DEM-FDM SIMULATION
deformation, which leads to a simple and accurate SCHEME
solution (Chen & Abousleiman, 2012 and Chen
et al., 2019). However, the cavity expansion theory 2.1 Cone penetration test model
suffers from several shortcomings. For example, the
influence of penetration rate and pore water pressure Discrete element method (DEM) is a numerical
is not considered. Yi (2012) used finite element method calculating the behavior of a large number of
method to analyze soil deformation around pene particles by computing the trajectories of each particle
trometer with different penetration rates. Finite with Newton’s law of motion (Cundall &
element simulation requires complex constitutive Strack,1979). It is widely used for conducting down
scaled numerical experiments on virtual sand samples
DOI: 10.1201/9781003308829-103
703
(Arroyo et al., 2011; Butlanska et al, 2014; Gu et al.,
2015). It is advantageous in studying geomechanics by
offering both macroscopic and microscopic informa
tion of sands compared with traditional numerical
methods based on continuum mechanics. While due to
the demanding requirement of computational
resources, its application on full scale analysis is unin
viting. In contrast, finite difference method (FDM) is
able to solve the governing equation in a speedy
manner while the large deformation of the soil near the
cone-tip cannot be handled and the fabric evolution of
microstructures can hardly be obtained. Therefore,
a DEM and FDM coupling scheme, which can effect
ively take advantage of the merits of both methods, is
appealing for simulating cone penetration testing. In
this study, we adopt well recognized software PFC3D Figure 2. Grain size distribution.
for DEM modeling of the near field sand and
FLAC3D for FDM modeling of both the cone pene
trometer and the far field domain. The coupling of 2.2 DEM parameter calibration
these two is realized via the equilibrium of the inter In order to realize the faithful constitutive behavior
face of the neighboring domains of DEM and FDM. of sands modeled by the discrete particle assembly,
The schematic of the cone penetration test model the contact parameters were calibrated by comparing
is shown in Figure 1a, where the cylindrical ground them with the triaxial compression test. Note that we
consists of the inner cylinder and the outer hollow adopt the rolling-resistant linear contact model in the
cylinder. The inner cylinder is modeled by DEM DEM simulation to account for the irregular shape of
containing 79,590 spheres and the hollow cylinder sand particles. The rolling-resistant linear contact
is modeled by FDM containing 50,000 elements. model consists of three parts: linear normal and tan
The adopted dimensions of the model are as fol gential force-displacement relations and a linear rota
lows: the diameter of the model domain (Dc) is tional constitutive relation. The particle assembly was
752 mm, the height of the soil model (H) is generated by the radius expansion method to emulate
432 mm, the radius of the inner cylinder (D1) is the particle size distribution of Fontainbleau sand
288 mm. As shown in the Figure 1b, the rigid cone measured in the physical test (Bolton et al., 1999), as
penetrometer is modeled by FDM consisting of shown in Figure 2. First, 16,160 proportionally smal
a frictionless cone rod of diameter (dc) 35.6 mm, ler non-overlapping particles were generated in a 4.5
a frictional cone with an inclination of 60° and × 4.5 × 9 mm3 cuboid, and then multiplied by the
a frictional sleeve of length (hc) 32.6 mm. In the radius expansion coefficient to gradually expand to
simulation, we prescribe a penetrating speed of the target particle size. The DEM model and FDM
0.1 m/s during the entire penetration. In order to model of the triaxial compression simulation for cali
ensure that the boundary conditions are consistent brating DEM physical parameters are shown in the
with that in the in-situ test, we adopted the bound Figure 3. The FDM model consists 50,000 elements
ary conditions suggested by Salgado et al., 1998: and the Mohr-Coulomb model is adopted as the soil
the horizontal boundary is prescribed by constant constitutive relation with its parameters listed in
stress, the upper boundary is stress free and the Table 1. The strain-controlled loading with a constant
bottom is displacement free. speed of 10-6 m/s is applied in the axial direction and
the uniform confining pressure of 100 kPa is applied
to the circumferential direction for both PFC3D and
FLAC3D models.
In the DEM simulation, the particle assembly was
generated to attain the relative density Dr=70% which
is comparable to physical samples. Typical ranges of
contact law parameters suggested by Gu et al. (2020)
were first adopted, e.g. contact stiffness, rolling resist
ance and contact friction. The contact parameters were
further fine-tuned to reconcile the DEM simulation
result with that from the FDM simulation, which is
deemed to be the true constitutive relation of the soil.
Figure 4 Shows the deviatoric stress-strain relations
and volumetric strain-axial strain relations for the triaxial
Figure 1. The numerical calculation model of the cone drained test with confining stress of 100 kPa from both
penetration test. DEM and FDM simulations. It can be seen that the
704
Figure 3. 3D triaxial compression test model for calibrating
DEM parameters: (a) PFC3D model and (b) FLAC3D model.
DEM and FDM simulation results (Seif et al., 2010)

agree to a satisfactory extent for the dense sample. The
calibrated parameters are presented in Table 1.
Table 1. Physical parameters in simulations.
Parameters Value
Normal and tangential stiffness of cone rod 1.0×1010 N/m

Friction coefficient of cone rod 0.2
Normal and tangential stiffnesses of side 1.0×1010 N/m
walls of calibration chamber
Side wall friction coefficient of calibration 0
chamber Figure 4. 3D triaxial compression test result: (a) deviator
Particle stiffness coefficient 6.0×108 N/m2 stress versus axial strain and (b) volumetric strain of the
Stiffness ratio of particle 0.25 ks/kn sample versus axial strain.
Particle friction coefficient 0.33
Rolling resistance coefficient of particles 0.8
number of cone tip particles increases continuously in
Particle density 2650 kg/m3
parallel with the cone tip resistance. When the normal
Young’s modulus in FDM 60.6 MPa
ized penetration h/dc depth reaches 2, the cone tip
Internal angle of friction in FDM 37.04
resistance tends to be stable and the average value is
Dilation angle in FDM 6.94
about 8.5 MPa, where h is the penetration depth. The
results imply that the microscale and macroscale fea
tures are highly correlated and the tip resistance could
serve as an indicator of the microstructural characteris
3 NUMERICAL SIMULATION AND RESULT tics of granular soil.
ANALYSIS Particle size amplification method increases the
simulation efficiency substantially by reducing the
In order to improve the efficiency of discrete element number of particles contacting with the cone tip, while
simulation, the particle size amplification method was the number of contacting particles and the soil near the
used to reduce the number of particles, and the cali cone-tip structure may change. In the following, we
brated parameters for the DEM simulation were show the simulation results considering different sizes
adopted to honor the true stress-strain relationship (Yan of the cone penetrometer to validate the efficacy of the
& Dong, 2011). In this study, the particle expansion particle size amplification method. Figure 6a shows the
ratio was chosen to be 35. The simulation result of the evolution of the tip resistance with the normalized
cone penetration calibration chamber test with confin penetration depth, with two cone penetrometer sizes
ing stress of 100 kPa is shown in Figure 5. As shown and the fixed dimension of the coupling model, i.e.
in the figure, the cone tip resistance as well as the dc/D50=4.89 and 3.81 with D50 denoting the mean par
number of particles contacting the cone penetrometer ticle diameter determined from the amplified particle
clearly progresses along two stages: growth stage and size distribution. It shows that the average tip resistance
stable stage. During the initial growth stage, the contact for dc/D50=4.89 and 3.81 are 7.45 MPa and 7.23 MPa
705
that there is a threshold of Dc/dc under which the tip
resistance increases with the size of the DEM domain
and beyond which no obvious size effect can be
observed.
The soil constitutive behavior depends on the
inter particle contact stiffness. To investigate its
effect, simulations with two inter particle stiffness
were conducted while keeping other parameters
the same as in Table 1. Figure 7a shows the vari
ation of radial stress along with the normalized
distance r/R, where r is the distance from the
center of the cone and R is the radius of the
cone. As shown in Figure 7a, the radial stress
decreases monotonically to the confining stress,
i.e. 100 kPa, as the normalized distance increases.
With the increase of the inter particle stiffness,
Figure 5. Simulation results of cone penetration test with the cone tip resistance increases due to the stiffer
confining stress of 100 kPa. behavior of soil.
Figure 7b shows the variation of radial stress with
respectively. The results indicate that the particle size the normalized distance under three different confining
amplification method with a careful selection of the stresses 50 kPa, 100 kPa and 200 kPa. As shown in
amplification ratio would not hinder the prediction of Figure 7b, the radial stress decreases monotonically
tip resistance. from the cone neighborhood to the outer boundary with
To investigate the size effect of the DEM domain, all three confining stresses. It is shown that the differ
we conducted simulations with different D1 and con ences of the radial stresses under different confining
stant R1. The results are shown in Figure 6b. It shows stresses are significant especially at the neighborhood
of the cone penetrometer. To reconcile the DEM-FDM
simulation with in situ CPT testing, attention should be
paid on the confining stress prescribed at the outer
boundary.
The relationship between the stable tip resistance
and the radial stress of the cone tip at r/R = 2.5 was
also investigated and shown in Figure 8. The results
show that there is a good linear relationship between
the tip resistance and the tip radial stress (r/R = 2.5)
under different simulation conditions, which agrees
well with the cavity expansion theory (Randolph et al.
1994).
4 DISCUSSION AND OUTLOOK
In this work, we modeled the cone penetrate

testing with a DEM-FDM coupling approach.
The micro-contact parameters in DEM are cali
brated by comparing the triaxial compression
simulation results based on DEM and FDM. The
size effect of the particle amplification method
was analyzed, and the influence of contact stiff
ness and confining stress was investigated. The
main findings of this study can be summarized
as follows:
(a) The DEM-FDM coupling method is a promising
and efficient simulation method investigating
cone penetration mechanics from both macro
scopic and microscopic perspectives.
(b) The particle size amplification method is capable
of further increasing simulation efficiency sub
Figure 6. Analysis of the size effect of the model: (a) dif stantially without undermining the accuracy of
ferent cone penetrometer sizes and (b) different coupling the prediction of cone tip resistance, with
model sizes. a cautious selection of the amplification ratio.
706
(d) The cone tip resistance and the radial stress
around the cone from DEM-FDM simulation
manifests a linear relation, which agrees well with
the cavity expansion theory.
REFERENCES
Arroyo, M., Butlanska, J., Gens, A., Calvetti, F. &
Jamiolkowski, M. 2011. Cone penetration tests in a virtual
calibration chamber, Géotechnique 61(6): 525–531.
Bolton, M.D., Gui, M.W., Garnier, J., Corte, J.F., Bagge, G.,
Laue, J. & Renzi, R. 1999. Centrifuge cone penetration
tests in sand, Géotechnique 49(4): 543–552.
Butlanska, J., Arroyo, M., Gens, A. & O’Sullivan, C. 2014.
Multi-scale analysis of cone penetration test (CPT) in
a virtual calibration chamber, Canadian Geotechnical
Journal 51(1): 51–66.
Chen, H., Li, L., Li, J. & Wang, H. 2019. Stress transform
method to undrained and drained expansion of
a cylindrical cavity in anisotropic modified cam-clay soils,
Computers and Geotechnics 106: 128–142.
Chen, J.W. & Juang, C.H. 1996. Determination of drained
friction angle of sands from CPT, Journal of Geotech
nical Engineering 122(5): 374–381.
Chen, S.L. & Abousleiman, Y.N. 2012. Exact undrained
elasto-plastic solution for cylindrical cavity expansion
in modified Cam Clay soil, Géotechnique 62(5):
447–456.
Cundall, P.A. & Strack, O.D.L. 1980. Discussion:
a discrete numerical model for granular assemblies,
Géotechnique 30(3): 331–336.
Gu, X., Hu, J. & Huang, M. 2015. K0 of granular soils:
Figure 7. Factors influencing radial stress distribution of a particulate approach, Granular Matter 17(6): 703–715.
cone tip: (a) soil modulus, (b) confining pressure. Gu X., Zhang J. & Huang X. 2020. DEM analysis of mono
tonic and cyclic behaviors of sand based on critical state
soil mechanics framework. Computers and Geotechnics
2020, 128: 103787.
Randolph, M.F., Dolwin, R. & Beck, R. 1994. Design of
driven piles in sand, Géotechnique 44(3): 427–448.
Salgado, R. 1998. Calibration chamber size effects on
penetration resistance in sand, Journal of Geotechnical
and Geoenvironmental Engineering 124(9).
Seif El Dine, B., Dupla, J.C., Frank, R., Canou, J. &
Kazan, Y. 2010. Mechanical characterization of matrix
coarse-grained soils with a large-sized triaxial device, Can
adian Geotechnical Journal 47(4): 425–438.
Sokolovskii, V.V. 2006. Statics of granular media. London:
Pergamon Press.
Wang, D., Bienen, B., Nazem, M., Tian, Y., Zheng, J.,
Pucker, T. & Randolph, M.F. 2015. Large deformation
finite element analyses in geotechnical engineering,
Computers and Geotechnics 65: 104–114.
Wesley, L.D. 2002. Interpretation of calibration chamber
tests involving cone penetrometers in sands, Géotechni
que 52(4): 289–293.
Figure 8. The relationship between cone tip resistance and Yan, W.M. & Dong, J. 2011. Effect of particle grading on
cone tip radial stress (r/R = 2.5). the response of an idealized granular assemblage, Inter
national Journal of Geomechanics 11(4): 276–285.
Yi, J.T., Goh, S.H., Lee, F.H. & Randolph, M.F. 2012.
(c) The effects of contact stiffness and confining A numerical study of cone penetration in fine-grained
stress on the stress field in the soil and cone tip soils allowing for consolidation effects, Géotechnique
resistance are considerable and attention should 62(8): 707–719.
be paid on them for emulating the in-situ cone Yu, H.S. & Mitchell, J.K. 1998. Analysis of cone resist
penetration testing. ance: review of methods, Journal of Geotechnical and
Geoenvironmental Engineering 124(2).
707
© 2022 the Author(s), ISBN 978-1-032-31259-0
Data-driven soil profile characterization using statistical methods

and artificial intelligence algorithms
R.L. Spacagna
CNR IGAG Area della Ricerca Roma 1, Montelibretti (Roma), Italy
A. Baris, L. Paolella & G. Modoni

University of Cassino and Southern Lazio, Cassino, Italy
ABSTRACT: CPT soil profile interpretation represents a fundamental aspect for subsoil stratigraphic recon
struction of complex geological contexts. In some situations, the soil profile may not exhibit evident boundary
changes, making the interpretation more difficult. This crucial aspect plays a key role in the layers boundaries
discontinuities identification and the construction of bi-dimensional and three-dimensional geotechnical models.
In this paper, CPT and boreholes are used to calibrate and validate a massive and automated site characterization
by combining statistical tools and artificial intelligence algorithms (AI). The procedure is applied in the complex
stratigraphic context of Terre del Reno (Italy). The proposed data-driven analysis allows to combine the geological
and geotechnical knowledge of the subsoil in an efficient and automatic way based on site-specific data, obtaining
reliable and indispensable results for the construction of a robust and coherent geotechnical model of the subsoil.
1 INTRODUCTION stratigraphic context of Terre del Reno (Italy), exten

sively affected by seismic liquefaction phenomenon
One of the subsoil stratigraphic reconstruction chal during the May 2012 earthquake sequence (Fioravante
lenges is to identify homogeneous soil layers charac et al. 2013). From the geodatabase of the Emila
terized by a different behaviour response. The Romagna Region, 102 pairs of CPT and boreholes
detection of the strata boundaries represents a crucial were selected to calibrate and validate the method. The
aspect in the construction of two-dimensional and CPT profiles are analyzed, and the homogeneous
three-dimensional geotechnical models. The develop layers are identified following statistical methods
ment of the data-driven analysis and the increasing based on the spatial variability analysis of the meas
availability of digitized geo-databases allow to ured parameters. Each statistically homogeneous layer
improve the current methodologies to obtain an effect is associated with the stratigraphic information
ive and extensive site characterization, providing a less extrapolated from closer boreholes, according to the
subjective interpretation of the subsoil data (Phoon Unified Soil Classification System (USCS). From the
et al., 2021). previous results, an artificial classifier is trained, and
The main studies rely on statistic and probabilistic the automatic classification procedure of soil profile is
approaches based on the cone penetration test (CPT) tested on a complementary subset of data.
data (Phoon et al. 2003, Facciorusso & Uzielli 2004, The proposed data-driven analysis combines the
Uzielli 2008, Wang et al. 2013, Paolella et al. 2019, geological and geotechnical knowledge of the sub
Shuku et al. 2020). The continuous measurements of soil in an efficient and automatic way, based on site-
soil parameters allow a statistical treatment for the specific data, to obtain reliable and indispensable
identification of lithological discontinuities and the results for the construction of a robust and coherent
reconstruction of the stratigraphic profiles (Lo Presti geotechnical model of the subsoil.
et al. 2009). For this purpose, the CPT-based soil
behaviour type index (Ic), and the correspondent soil
behaviour type (SBT), are used to delineate soil 2 METHODOLOGY
boundaries (Robertson 2016).
This work developed a data-driven methodology for The proposed approach aims to define a methodology
the subsoil stratigraphic reconstruction combining stat for the subsoil stratigraphic recognition, applicable at
istical tools and artificial intelligence algorithm (AI). the urban scale. In practice, this operation is based on
The proposed procedure is applied to the complex the interpretation of a considerable amount of data
DOI: 10.1201/9781003308829-104
708
coming from different sources. The most diffused 2.2 Soil boundaries discontinuities
approach is to combine the borehole logs stratigraphy
The homogeneous soil layers within the CPT profiles
and Cone Penetration Test results in a deterministic
are identified based on the soil behaviour type index
way with a consequent subjectivity of the interpret
Ic (Robertson, 2016). The automatic procedure pro
ation of available data.
vides an accurate interpretation of the CPT tests con
The increasing availability of site-specific data col
sidering the spatial correlation of the considered
lected, digitized and stored in the geodatabase allows
values along with the vertical profile.
to improve, even automatically, the consolidated
This statistical test verifies the equality of the
methodologies for interpretation of subsoil data.
means and the variance of two subsets of data,
This procedure, summarized in Figure 1, provides
according to the procedure shown in Figure 2.
a stratigraphic interpretation model applicable to all
The two subsets of data (namely Ω1 and Ω2 , with
available surveys in the analyzed area, combining
size respectively equal to n1 and n2, average Q1
information from different investigation tests. In par
and Q2 , and variance σ1 2 and σ2 2 ) are identified
ticular, the proposed methodology identifies the
along with the vertical CPT profile with
main lithologies from borehole log stratigraphies,
a moving window Wd0 dived by d0. The T ratio
the layer discontinuities from the sectioning statis
(Equation 1) and the intra-class correlation coeffi
tical test of the CPT profiles and the automized
cient ρI (Equation 2) are calculated along with
stratigraphic classification of each statistically homo
the vertical CPT profile.
geneous layer from AI.
Figure 2. Definition of the two subsets of relevant param

eters along the vertical axis of CPT test (from Spacagna
et al, 2015).
Figure 1. Proposed methodology for the subsoil strati
graphic recognition.
2.1 Definition of stratigraphic units

The definition of the main stratigraphic units detect
able in the study area requires a detailed geological, Where:
hydrogeological, geomorphological and sedimento

logical framework.
The stratigraphic reference units for the selected

case study are defined by combining the information
from the geological framework and the borehole log
stratigraphy. The detail of the stratigraphic model, its
extension and stratigraphic classification are strictly
related to the typology of the performed analysis (i.e.
subsidence, liquefaction, design of structures and infra
structures). Therefore, the model (depth and thickness
of the strata) and the stratigraphic units are defined
according to the analyzed phenomenon (i.e., for
a liquefaction risk assessment, the liquefiable layers).
709
Where: The amplitude of wd0 used for the statistical test
is defined as a ratio of the range a. The T ratio and ρI
are calculated for each point d0, implementing two
new vertical profiles. The higher values of T ratio
and ρI correspond to a change of behaviour of the
CPT-based parameters Ic.
where Q is the average of the data Qi belonging to The critical value of the parameter T ratio (tc) is
the window wd0, with i=1,2, …, (n1+n2). evaluated considering the 90% confidence interval of
To define the window Wd0, the geostatistical the T ratio distribution and is calculated following
approach proposed by Spacagna et al. (2015) Equation 9.
suggests to calculate the one-dimensional experi
mental variogram of the variable (Chilès & Del
finet 2012), with a lag equal to the minimum
distance of measured point, following the
Equation 7. where μTratio and σTratio are respectively mean and
standard deviation of the distribution of the T ratio
values along the vertical profile. The depth to which
T ratio values fall outside the confidence interval
represents a change of behaviour along the CPT
profile.
The critical value of ρIc is calculated following
where z(i) is the value of the considered vari
the Equation (10) proposed by Herzagy, Mayne, and
able at a location, z(i+h) is the value of the vari
Rouhani (1996).
able at the distance h, and N(h) is the number of
couples of points with a distance equal to h. This
spatial correlation of the variable is modelled
with a theoretical function. In the present study,
the spherical model (Chilès & Delfinet 2012) is
adopted to interpolate the spatial correlation where μρI and σρI are respectively the mean and
(Equation 8). standard deviation of the distribution of the ρI values
along the vertical profile. The depth to which ρI
values are higher than ρIc represents a change of
behaviour along the CPT profile.
The transition between two different homoge
neous layers is assumed at the d0 depth points which
both critical conditions occur simultaneously.
The described algorithm is implemented with
open-source R software (R Core Team, 2021).
2.3 Stratigraphic recognition with AI

In the proposed methodology, the stratigraphic rec
ognition, basically the attribution at each statistically
homogeneous layers the correspondent stratigraphic
unit, is automatically carried out with artificial
intelligence (AI), studying the underlying relation
between cone penetration tests information and
stratigraphic properties. Reale et al. (2018) proposed
an automatic classification of fine-grained soils using
CPT measurements and Artificial Neural Network.
Ching, Wu & Phoon (2020) and Ching et al. (2020)
Figure 3. Spherical variogram.
proposed a Hierarchical Bayesian Model (HBM) to
construct transformation models for soil and rock
In general, the variogram γðhÞ is an increasing properties.
function with the distance h. The model introduces The algorithms are trained comparing the outcomes
two parameters. The sill C describes the level of spa derived from couples of CPT-boreholes considered
tial variability, and the range a represents the max spatially correlated, providing a complementary
imum distance at which spatial correlation is description of the subsoil. The CPT-borehole distance
observed. has been chosen based on the analysis of the spatial
710
structure of cone tip resistance in the horizontal direc
tion. A measure of the spatial correlation is the scale of
fluctuation, representing the maximum distance over
which the points are significantly related (Chilès &
Delfinet, 2012).
The calibration procedure is structured as follows:
first, the couples of complementary CPT and bore
holes are extracted from the geodatabase; once CPT
profiles are processed, to each statistically homoge
neous layer is associated the stratigraphic unit
reported in the complementary borehole; last, the
CPT output parameters and their correspondent
stratigraphic recognition are used to train, test and
validate various algorithms.
The artificial classifier characterized by the high
est efficiency is selected and applied to automatically
assign the stratigraphic units to the remained sec
tioned cone penetration test profiles distributed over
the studied area.
3 CASE STUDY
The Terre del Reno municipality is located in the

southern Po river plain in the Emilia Romagna
region (Italy). The area was struck by an intense
seismic sequence associated with compression fault
ruptures in 2012. The main shock of May 20th 2012
(Mw6.1) produces extensive liquefaction phenomena
due to subsoil composition, geologic history and the
shallow depth of the groundwater table (Fioravante
et al. 2013). The municipality covers an area of
51 km2, along a former branch of Reno River and is
divided into three main districts: Sant’Agostino, San
Carlo and Mirabello (Figure 4a).
Figure 4. A) Terre del Reno municipality b) San Carlo dis
The San Carlo area is characterized by a complex trict, c) Geological sections (adapted from Calabrese et al.,
stratigraphic context due to the depositional history 2012).
of the Reno river, highlighted by the significant
number of liquefactions induced damages
(Figure 4b). The calibration of the methodology has
been carried out by analyzing the entire available The deepest layers are constituted of sediments
geodatabase on the territory. However, in this work, deposited by Apennine and Alpine rivers, inter
the stratigraphic recognition is limited to the San spersed with marine accretion. These alternated pro
Carlo area. cesses are ruled by the regression and progression of
the Adriatic Sea in the upper Pleistocene-Holocene
(Romeo 2012).
3.1 Definition of stratigraphic units in the The territory of Terre del Reno is covered by
municipality of the Terre del Reno around 1700 geognostic surveys, highlighted in
The subsoil of San Carlo is characterized by Figure 4a. The definition of the stratigraphic units
a relatively recent geologic history, an intensive has been carried out analyzing the description
depositional sequence of the Reno river and reported in 252 boreholes logs distributed all over
a shallow groundwater table (Figure 4c). The the municipality. From the geological and sedimen
depositional activity of the river left superficial tological characterization, four stratigraphic units
layers of loose sand with a thickness of a few have been identified in the upper 20 m of the
meters. In the same period, the population con subsoil:
structed artificial levees mixing sand with silt to - Unit #1: silt, silty clay, clayey silt;
regulate the fluvial regime and limit flooding. The - Unit #2: sandy silt, silty sand;
urban area is mainly built near those paleo-channel - Unit #3: fine and medium clean sand;
and paleo-levees. - Unit #4: clay, organic clay, organic material.
711
3.2 CPT profiles processing a correlation length equal to Θ = 16.5 m in the hori
zontal direction and a scale of fluctuation δ = 33 m,
The soil boundaries discontinuities recognition has
representing the maximum distance over which the
been applied to the district of San Carlo. In the
points are significantly related.
selected area, there is a high density of surveys
All boreholes located in the defined distance are
(Figure 4b). For 369 investigation tests distributed
classified as complementary of the cone penetration
over an area of 3 km2, the available boreholes and
tests. The total number of couples CPT-boreholes is
cone penetration tests reach 82.5%. In particular, the
equal to 132, but 30 have been deleted due to the
mechanical CPT correspond to 35.5% and the elec
poor quality and reliability of the subsoil description.
trical ones to 32% of the total.
The remaining 102 pairs constitute the starting point
The implemented tool provides automatic section
for the calibration of the algorithm. The sectioned
ing of CPT profiles into statistically homogeneous
CPT and the complementary boreholes have been
layers.
compared.
An example of sectioned CPT profile is shown in
Besides the output sectioning CPT parameters, at
Figure 5. From the CPT parameters, the Ic profile is
each statically homogeneous layer has been attrib
computed. The statistical parameters Tratio and ρ are
uted the stratigraphic unit recognized in the comple
calculated, and the critical thresholds are estimated.
mentary boreholes. These parameters are the input
The horizontal dashed lines identify the transition
values for the calibration of a classifier capable to
between two different homogeneous layers.
identify the corresponding stratigraphic unit.
The choice of the algorithm involves the estima
tion of the efficiency between 32 different artificial
classifiers, implemented in the Classification
Learner APP available in MATLAB R2021b, with
a 10 folds cross-validation procedure (Stone 1974).
This procedure is suggested to avoid overfitting and
to estimate the accuracy obtained within the 10 iter
ations. Cross-validation divides the dataset into 10
folders of the same size: 9 folders are used to train
the classifier while one is used to validate it. This
procedure is iterated 10 times, training and testing
each folder. The best classifier is a linear discrimin
ant, characterized by an efficiency equal to 81.6%.
Linear Discriminant Analysis (LDA) has been pro
Figure 5. Example of sectioned CPT profile. posed by Fischer in 1936. It consists in finding the
hyperplane projection that minimizes the interclass
variance and maximizes the distance between the
projected means of the classes.
For each stratum, the following parameters are In Figure 6a is shown the efficiency of the algo
computed: rithm to classify the four units, attributing the per
- the Soil Behaviour Type SBT, centage of success and misclassification. The
- the mean and standard deviation of the Ic values, positive prediction is expressed as Positive Predicted
- the mean and standard deviation of the friction Values (PPV), and the negative classification is
ratio, FR, defined as ratio between sleeve fric expressed as False Discovery Rates (FDR).
tion and cone resistance. Figure 6b consist in a report showing a summary
of prediction results of the mutual classification
between classes, the grey cells correspond to the
3.3 Training and application of the artificial PPV and the sum of the remaining elements along
intelligence algorithm the columns is equal to the FDR. For example, the
first column summarizes the detail of the layers
At this stage, information reported in boreholes and automatically classified as Unit #1: truly classified
sectioned CPT must be cross-correlated to perform in 75.4% of occurrences; Unit #2 has been mistaken
an automatic stratigraphic recognition. The calibra with Unit #1 in 15.8% of manifestations; Unit #1
tion of the artificial intelligence algorithm has been has been attributed to the real Unit #3 in 1.2% and
performed comparing the outcomes derived from in the remaining 7.6%, Unit #4 has been classified
couples of CPT-borehole considered spatially correl as Unit #1. In particular, the mutual misclassifica
ated. Due to the complex stratigraphic context in the tion between sands and clays never occurs. Further
investigated area of Terre del Reno (Fe), each CPT more, the lower PPV values are associated with
profile has been considered representative of Units #1 (silt, silty clay, clayey silt) and #2 (sandy
a circular area having a radius equal to 30 m. This silt, silty sand), located in the upper 10 m of the
distance has been chosen because the cone tip resist subsoil and characterized by a complex geological
ance of the selected site is characterized by history.
712
recognition of the surveys disseminated on the inves
tigated area (Figure 7b).
4 CONCLUSION
This work aims to propose a methodology for

subsoil stratigraphic reconstruction based on the
most diffused geotechnical survey, combining stat
istical tools and artificial intelligence algorithms.
In particular, the proposed methodology identifies
the layer discontinuities from the sectioning statis
tical test of the CPT profiles, the main lithologies
from borehole log stratigraphies, and the auto
mized stratigraphic classification of each statistic
ally homogeneous layer from AI. The procedure
has been implemented and validated in the com
plex geological context of the alluvial plain of
Terre del Reno in Italy, characterized by a large
geodatabase. The automatized methodology is
easily replicable in other contexts for the strati
graphic recognition of the subsoil on a large scale,
providing a valid tool for several applications as
liquefaction risk analysis at an urban scale on
structure and infrastructure (Paolella et al. 2021,
Baris et al. 2021, Modoni et al. 2019) and subsid
ence study (Spacagna et al. 2020, Spacagna &
Modoni, 2018). The data-driven analysis provides
reliable and indispensable results for the construc
Figure 6. Efficiency of the proposed LDA algorithm. tion of a robust and coherent geotechnical model
of the subsoil. The next step concerns the inter
polation of the stratigraphic surfaces to obtain
a complete reconstruction of the subsoil.
REFERENCES
Baris, A., Spacagna, R.L., Paolella, L., Koseki, J. and
Modoni, G. 2021. Liquefaction fragility of sewer pipes
derived from the case study of Urayasu (Japan). Bull.
Earthquake Eng. 19, 3963–3986. https://doi.org/
10.1007/s10518-020-00957-2
Calabrese, L., Martelli, L., Severi, P. 2012. Stratigrafia del
l’area interessata dai fenomeni di liquefazione durante il
terremoto dell’Emilia (maggio 2012). Atti 31° Convegno
GNGTS, Potenza, 20-22 novembre 2012, sess. 2.2,
119–126.
Chilès, JP. & Delfiner, P. 2012. Geostatistics: modeling spa
tial uncertainty. 2nd edn. Wiley, Hoboken, p 726. ISBN
978-0-470-18315-1.
Ching, J., Phoon, K. K., Ho, Y. H. and Weng, M. C. 2020.
Quasi-site-specific Prediction for Deformation Modulus
of Rock Mass. Canadian Geotechnical Journal.
doi:10.1139/cgj-2020-0168.
Ching, J., Wu, S. and Phoon, K. K. 2020. Constructing
Figure 7. a) Input parameters and output for the selected
QuasiSite-specific Multivariate Probability Distribution
LDA algorithm and b) application on the surveys dissemin
Using Hierarchical Bayesian Model. Journal of Engin
ated all over the area.
eering Mechanics, under review computing.
R Foundation for Statistical Computing, Vienna,
Austria.
Once the effectiveness was ascertained, the algo Facciorusso, J. & Uzielli, M. 2004. Stratigraphic profiling
rithm (Figure 7a) was applied to the remaining cone by cluster analysis and fuzzy soil classification from
penetration tests, obtaining the stratigraphic mechanical cone penetration tests. Proc. ISC-2 on
713
Geotechnical and Geophysical Site Characterization, R Core Team. 2021. R: A language and environment for
Porto. Vol. 1, Millpress, 905–912. statistical.
Fioravante, V. et al. 2013. Earthquake geotechnical engin Reale, C., Gavin, K., Librić, L. and Jurić-Kaćunić D. 2018.
eering aspects: the 2012 Emilia Romagna earthquake Automatic classification of fine-grained soils using CPT
(Italy). Seventh international Conference on Case His measurements and Artificial Neural Networks.
tories in Geotechnical Engineering, April 29 th – Advanced Engineering Informatics, Volume 36, Pages
May 4th, 2013. Chicago (US). 207–215, ISSN 1474-0346. https://doi.org/10.1016/j.
Fisher, R. A. 1918. The correlation between relatives on aei.2018.04.003.
the supposition of mendelian inheritance. Trans. R. Soc. Robertson, P. K. 2016. “Cone Penetration Test (CPT)
Edinb. 53: 399–433. Based Soil Behaviour Type (SBT) Classification
Herzagy, Y., Mayne, P., Rouhani, S. 1996. Geostatistical System – An Update.” Canadian Geotechnical Journal
assessment of spatial variability in piezocone tests. 53 (12): 1910–1927
Uncertainty in geologic environment: from theory to Shuku, T., Phoon, K. K. and Yoshida, I. 2020. “Trend Esti
practice (GSP 58), ASCE, New York, 254–268. mation and Layer Boundary Detection in
Lo Presti, D., Meisina, C. and Squeglia, N. 2009. Applica Depth-dependent Soil Data Using Sparse Bayesian
zione delle prove penetrometriche statiche nella ricos Lasso.” Computers and Geotechnics 128: 103845.
truzione del profilo stratigrafico. Rivisita Italia di Spacagna, R.L., Modoni, G., Saroli, M. 2020. An Inte
Geotecnica. grated Model for the Assessment of Subsidence Risk in
MATLAB 2021. version R2021a. Natick, Massachusetts: the Area of Bologna (Italy) Lecture Notes in Civil
The MathWorks Inc. Engineering 2020, 40, pp. 358–368
Modoni G., Spacagna R.L., Paolella L., Salvatore E., Spacagna, R.L., Modoni, G. (2018) GIS-Based Study of
Rasulo A., Martelli L., 2019. Liquefaction risk assess Land Subsidence in the City of Bologna. In:
ment: lesson learned from a case study. Earthquake Ottaviano E., Pelliccio A., Gattulli V. (eds) Mechatron
Geotechnical Engineering for Protection and Develop ics for Cultural Heritage and Civil Engineering. Intelli
ment of Environment and Constructions. CRC Press. gent Systems, Control and Automation: Science and
https://doi.org/10.1201/9780429031274 Engineering, vol 92. Springer, Cham. https://doi.org/
Paolella, L., Spacagna, R.L., Chiaro, G. and Modoni, G. 10.1007/978-3-319-68646-2_10
2021. A simplified vulnerability model for the extensive Spacagna R.L., de Fouquet, C., Russo, G. 2015. Interpret
liquefaction risk assessment of buildings. Bull. Earth ation of CPTU tests with statistical and geostatistical
quake Eng. 19, 3933–3961. https://doi.org/10.1007/ methods. ISGSR2015 Geotechnical Safety and Risk V T.
s10518-020-00911-2 Schweckendiek et al. (Eds.) © 2015 The authors and
Paolella, L., Salvatore, E., Spacagna, R.L., Modoni, G., IOS Press. 13-16 October 2015
Ochmański, M. 2019. Prediction of liquefaction damage Stone, M. 1974. Cross-validatory choice and assessment of
with artificial neural networks. Earthquake Geotechnical statistical predictions. Journal of the Royal Statistical
Engineering for Protection and Development of Envir Society: Series B (Methodological), 36(2),111–133.
onment and Constructions - Proceedings of the 7th https://doi.org/10.1111/j.2517-6161.1974.tb00994.x
International Conference on Earthquake Geotechnical Uzielli, M. 2008. Statistical analysis of geotechnical data.
Engineering, pp. 4309–4316 3rd International Symposium on Geotehcnical and Geo
Phoon, K. K., S. T. Quek, and P. An. 2003. “Identification physical Site Charaterization, Huangs and Mayers
of Statistically Homogeneous Soil Layers Using Modi (Eds.), 173–193, Taipei, Taiwan, 1-4 april 2008.
fied Bartlett Statistics.” Journal of Geotechnical and Wang, Y., Huang, k. and Cao, Z. 2013. Probabilistic identi
Geoenvironmental Engineering, ASCE 129 (7): fication of underground soil stratification using cone
649–659. penetration tests. Can. Geotech. J. 50: 766–776.
Phoon, K.K., Ching, J. and Shuku, T. 2021. Challenges in Wang, Y., Huang, K. and Cao, Z. 2013. Probabilistic Identi
data-driven site characterization, Georisk: Assessment fication of Underground Soil Stratification Using Cone
and Management of Risk for Engineered Systems and Penetration Tests. Canadian Geotechnical Journal 50
Geohazards. DOI: 10.1080/17499518.2021.1896005. (7): 766–776.
714
© 2022 the Author(s), ISBN 978-1-032-31259-0
Capability of seismic CPTu and DMT in assessing propagation velocity of

body waves: A comparative study
S. Stacul, D. Lo Presti & N. Nenci
University of Pisa, Pisa, Italy
F. Fiera & M. Perini

Geo-Energizers Snc, Cascina (Pisa), Italy
D. Marchetti
Studio Prof. Marchetti s.r.l., Rome, Italy
E. Pagani & M. Siviero

Pagani Geotechnical Equipment s.r.l., Calendasco (Piacenza), Italy
ABSTRACT: The paper shows the geotechnical characterization of a soft clay test site by means of laboratory
testing. Such a test site was established in the campus of the Department of Biology of the University of Pisa at
about 1.6 km from the Leaning Tower square. The main (but not exhaustive) scope of the present research
activity was to compare the propagation velocity of body waves as obtained from various seismic techniques.
A preliminary site characterization, based on both laboratory and in situ tests, was carried out. In situ testing
consisted of a preliminary CPTu, three boreholes with four high quality Osterberg-type undisturbed sampling,
cross-hole, surface seismic reflection test named Multichannel Analysis of Reflection Waves (MARW). Labora
tory tests consisted of soil classification (Atterberg Limits and grain size distribution), incremental loading oed
ometer tests (OE), direct shear tests (DS) and resonant column tests (RC). Moreover, Pagani seismic piezocone
and Marchetti seismic dilatometer were used for the assessment of the propagation velocity of body waves.
1 INTRODUCTION The in-situ characterization consisted of

a preliminary CPTu, boreholes (with undisturbed
The capability of seismic piezocones (SCPTu) and sampling), cross-hole (CH) and surface seismic tests
seismic Marchetti dilatometers (SDMT) in determin (i.e. SH re-flection test as Multichannel Analysis of
ing the propagation velocity of body waves is the Reflection Waves - MARW). A direct stratigraphy
final objective of the present research activity. For was inferred from borehole S1.
such a purpose, a test site was selected in Pisa. This The in-situ characterization was carried out by:
site is in a campus of the University of Pisa, namely – Geoservizi snc (CPTu)
the Department of Biology (Figure 1). – Geo-Energizer snc (surface seismic test:
MARW)
– Servizi Geologici S.a.s. Morbin F. &
C. (boreholes S1, S2, S3)
– SolGeo Srl (Cross-hole. Borehole S2 was used
as source, the other two as receivers)
The laboratory test results were carried out on four
high-quality Osterberg-type samples. In particular, the
results of laboratory classification (Atterberg Limits and
grain size distribution), incremental loading oedometer
tests (OE), direct shear tests (DS) and resonant column
tests (RC) are shown. The recommendations of the Ital
ian Geotechnical Society – AGI were mainly followed
in performing laboratory tests (AGI, 1994). Anyway,
these recommendations mainly conform to the ASTM
Figure 1. Test site layout.
standards that were preferred in case of discrepancies.
DOI: 10.1201/9781003308829-105
715
Table 1 shows the identification number of each Only at a depth of about 20 m a fat clay was identi
sample, the sampling depth, as well as types and fied from classification tests.
number of tests that were performed. Some samples
were apparently non-homogeneous; therefore, classi
2.2 Oedometer and direct shear tests
fication and RC tests were carried out by considering
different portion of the samples. The water table was Tables 4, 5 and 6 summarize the parameters as
located at a depth of three meter from ground level obtained from incremental loading oedometer tests.
when the samples were retrieved.
Table 2. Soil classification.
Table 1. Laboratory testing program.
Depth Sand Clay Pass. #200
Depth Grain Size Atterberg DS OE RC

Sample m % % % USCS
Sample m # # # # #
C1 4.5-5.0 62.12 7.38 37.88 Silty sand
C1 4.5-5.0 1 1 1 1 1 C2(DS) 8.5-9.0 3.33 31.31 96.67 Silt
C2 8.5-9.0 4 4 1 1 2 C2(RC) 8.5-9.0 6.52 23.14 93.48 Silt
C3 19.5-20.0 1 1 1 1 1 C2(RC) 8.5-9.0 18.71 15.8 81.29 Silt clay
C4 26.5-27.0 3 3 1 1 2 with sand
C2(OE) 8.5-9.0 2.87 20.98 97.13 Silt
C3 19.5-20.0 1.57 68.49 98.43 Fat clay
This paper mainly deals with in situ seismic meas C4(DS) 26.5-27.0 58.28 14.93 41.72 Silty
urements. Laboratory tests are shown but not dis Clayey sand
cussed or compared with in situ test results, except C4(RC) 26.5-27.0 33.66 15.6 66.34 Sandy lean
data coming from RC tests. clay
C4(OE) 26.5-27.0 15.44 19.13 84.56 Lean clay
with sand
2 LABORATORY CHARACTERIZATION
Table 3. Physical characteristics.
2.1 Physical characteristics
Tables 2 and 3 summarize the physical characteristics Depth γn γs wn wL PI
of the tested samples. In particular, the natural and
3 3
dry unit weight are reported. As for the unit weight, Sample m kN/m kN/m % % %
the mean and the standard deviation are reported.
Moreover, the natural water content, the liquid limit C1 4.5-5.0 18.63 14.29 31.52 N.A. N.A.
and the plasticity index are shown. Percentages of ± 0.28 ± 0.60 ± 1.69
grain size classes and USCS classification (ASTM C2(DS) 8.5-9.0 18.31 13.75 33.15 38 4
D2487-17e1, 2017) are eventually reported. ± 0.11 ± 0.15 ± 0.87
Figure 2 shows the grain size distribution curves. C2(RC) 8.5-9.0 18.24 13.66 31.75 38 3
According to the laboratory test results the presence C2(RC) 8.5-9.0 18.19 13.66 30.85 34 7
of silt mixtures seems to prevail at various depths. C2(OE) 8.5-9.0 19.32 14.5 31.96 35 3
C3 19.5-20.0 16.14 10.46 54.6 75 45
± 0.33 ± 0.60 ± 4.02
C4(DS) 26.5-27.0 19.14 16.08 19.61 24 11
± 0.15 ± 0.02 ± 0.93
C4(RC) 26.5-27.0 19.85 16.55 18.88 29 16
C4(OE) 26.5-27.0 20.43 16.66 21.96 34 18
γn : natural unit weight; γs : dry unit weight; wn: natural water

content; wL: liquid limit; PI: plasticity index.
Compressibility parameters are shown in Tables 4

and 5, while the hydraulic parameters in Table 6.
The results from oedometer tests confirm what
already observed as for the soil classification. The
sample at the depth of 20 m exhibits a much greater
compressibility and a permeability reduced of about
two orders of magnitude. The coefficient of primary
consolidation was computed by using both Taylor
(1948) and Casagrande (1936) methods. Anyway, in
the case of specimens containing more silts the inter
Figure 2. Grain size distribution curves. pretation of the settlement-time curve was clearer
716
Table 4. Oedometer tests results. Also, direct shear tests were carried out on the
four high-quality Osterberg-type samples. For each
Depth OCR e0 sample, three specimens were tested in the direct
shear box. The applied consolidation pressures were
Sample m kPa - - selected to include the best estimate of the in situ
vertical effective stress.
C1 4.5-5.0 68.8 2.0 0.78 4 Differences between sample C3 and the others were
C2 8.5-9.0 102.9 1.31 0.85 4 evident in terms of strength parameters. As a first esti
C3 19.5-20.0 151.84 1.13 1.80 2 mate at peak, sample C3 exhibits a cohesion of about 5
C4 26.5-27.0 218.27 1.20 0.61 11 kPa and an angle of shear resistance of about 20° whilst
other samples have an apparent cohesion of about 2 kPa
: effective geostatic stress; OCR: overconsolidation
and an angle of shear resistance of about 30°. At large
ratio; e0: void index; M: oedometric modulus.
displacements the apparent cohesion is almost zero
while the angle of shear resistance decreases to about
17° for the C3 sample. For the other samples, the angle
Table 5. Compressibility parameters from OE. of shear resistance remains almost constant. Moreover,
except for the C3 sample, the others exhibit a certain
Depth Cr Cc Cs curvature of the strength envelope.
Sample m - - -
2.3 Shear stiffness and damping ratio
C1 4.5-5.0 0.034 0.226 0.024 1.5E-03
C2 8.5-9.0 0.037 0.258 0.045 9.5E-04 RC tests were carried out according to ASTM
C3 19.5-20.0 0.148 0.587 0.175 5.1E-03 D4015-15e1 (2015). Figures 3 and 4 respectively
C4 26.5-27.0 0.030 0.15 0.033 7.3E-04 show the normalized shear modulus and the damping
Cr: recompression index; Cc: compression index; Cs: swell

index; Cαε: coefficient of secondary consolidation.
Table 6. Hydraulic parameters from OE.
Depth
Sample m cm2/s cm/s
C1 4.5-5.0 3.78E-02 8.4E-07

C2 8.5-9.0 9.21E-03 2.2E-07
C3 19.5-20.0 1.14E-04 5.5E-09
C4 26.5-27.0 2.9E-02 2.7E-07
Cv: coefficient of primary consolidation, k: permeability.

Figure 3. Normalized shear modulus vs. shear strain from RC.
with the Taylor method than with the Casagrande
approach. Therefore, only the results obtained by the
Taylor approach are shown. Obviously, the second
ary consolidation coefficient was inferred from the
semi-log Casagrande representation as slope of the
secondary compression branch of the curve.
In the normally consolidated range, the constrained
modulus exhibits the typical linear trend with the
increase of the consolidation pressure. Only for the
C3 sample a decrease of M until the pre
consolidation pressure was observed. The secondary
compression coefficient increases, as expected until
the applied pressure is 1.5 – 2.0 times the pre
consolidation pressure. On the other hand, the perme
ability linearly decreases with the applied pressure in
a log scale, as usually. Differences between the clay
sample (C3) and the other are evident in terms of
both compressibility and hydraulic parameters. Figure 4. Damping ratio vs. shear strain from RC tests.
717
ratio as inferred from RC tests vs. the log of the shear MARW is an innovative seismic technique based
strain. The reference curves, that have been suggested on 1D simplification of classical reflection survey.
by Vucetic and Dobry (1991) for non-plastic soils Advantages of this technique derive in acquiring
and for soils with a PI of 50%, are also shown. a single Common Depth Point (CDP) or Common
The influence of the PI on the RC results is con Shot Gather (CSG). Dataset is subsequently ana
firmed even though a certain discrepancy is seen for lyzed through a regular processing flow to derive
high plasticity soils between the experimental results a vertical seismostratigraphy. This method reaches
and the suggested curves. The small strain shear its greatest effectiveness in lithological domains
modulus typically ranges in between 36 and 40 MPa characterized by slopes less than 15°.
while the last specimen of the C4 sample gave an Figure 9 represents the acquisition geometry of
initial shear modulus as high as 90 MPa. The speci MARW survey. For this test 96 high sensitivity 4.5
mens were iso-tropically consolidated at the best Hz horizontal geophones connected to a 96 channels
estimate of the in situ vertical effective stress. 24 bit Do.Re.Mi SARA Electronic Instruments Seis
mograph were employed. The energy source con
sisted of a 12 Kg sledgehammer transversely hitting
3 IN SITU TESTS (three times to increase S/N ratio) a wooden plate
placed in B1 position.
3.1 CPTu and DMT tests
Figures 5, 6 and 7 respectively show, in a standard
format, the results of CPTu (ASTM D5778-20, 2020)
and DMT tests (ASTM D6635-15, 2015). In particular,
Figure 6 shows the modified SBTn (I(B)) profile
(Robertson, 2016, 2021). Figure 8 shows the strati
graphic profile as inferred from borehole S1. Boreholes
S2 and S3 were both core destruction surveys. As
a preliminary comment, the indirect stratigraphic profiles
from CPTu and DMT exhibit some differences. This
point should be discussed in the light of laboratory test
results, which is beyond the scope of this paper.
Figure 5. CPTu test results: (a) tip resistance; (b) sleeve Figure 6. Modified Soil Behaviour Type (SBTn) form
friction; (c) pore pressure. CPTu.
3.2 Body waves propagation velocities via different A very accurate data processing has been used to
techniques increase reflected events visibility: trace normaliza
Reflection technique is based on recording time elap tion, gain recovery (AGC), band pass filter, predictive
sing between the instant an elastic impulse is gener deconvolution, muting first break, FX deconvolution,
ated from the surface, reflected from a deep NMO correction.
lithological interface and raised back to the surface. Figure 10 shows from left to right: a) processed
In this sense, reflection exploits seismic imped CSG before velocity analysis; b) the semblance
ance contrast between different materials and its (which shows velocity values consistent with reflec
advantage is the ability to penetrate underground tion events); c) time stack trace (repeated four
with accuracy. times); d) depth converted stack trace.
718
Figure 9. Seismic reflection survey geometry (MARW).
Figure 7. DMT test results: (a) readings; (b) material

index; (c) constrained modulus.
Figure 10. Seismic reflection survey results (MARW).
waves) and transverse (S waves) propagation times

between two or more holes along horizontal trajector
ies. Travel-times, along with distances between source
and receiver points, allow defining seismic wave vel
ocity of investigated soils. The CH tests were carried
out using central hole (S2) for positioning of transmit
ting sources and S1 and S3 holes for simultaneous
acquisition by means respective receivers (P and S).
Therefore, it was possible to verify the subsoil homo
geneity of the survey area. A “Sparker” generator was
employed to create compression waves while
a “GEOS” transducer produced vertical polarized
shear waves (instruments, both source and receivers,
used for cross-hole survey was by Solgeo s.r.l.).
Record of signals (Figure 11) was made by means of
two different types of transducers to receive P and
S waves respectively.
Pre-amplified hydrophones were employed to
receive P waves and pre-amplified 28 Hz AVG geo
phones (Amplified Vertical Geophone) for S waves.
Figure 8. Stratigraphic profile from borehole S1.
Acquisitions were carried out with 1 m vertical con
stant intervals by a 2x16Bit MCHA-GP sampler. Sig
nals were acquired with a 100 kHz frequency
Due to limitations related to acquisition geometry, sampling for S-waves and 200 kHz for P-waves
the accuracy of depth and velocity values is progres (record length 100 ms). First arrival times were
sively reduced beyond depths of about 60 m. manually picked through a PC interface. Velocities
Alongside the geophones array two in-hole seis was simply calculated by operating the distance/times
mic tests (Cross-Hole) were carried out. ratio. A preliminary inclinometric survey was carried
Cross-hole geophysical surveys consist of direct out to define hole deviation from vertical direction
measurements, at various depths, of longitudinal (P through simultaneous measurement of inclination and
719
Figure 12. Vs profiles inferred by various techniques.
velocity as inferred from different in situ and labora

tory tests. The Figure shows a very satisfactory agree
ment among the different testing methods. Moreover,
laboratory testing well compared to in situ
measurements.
Figure 11. Cross-Hole records (S2-S1 left; S2-S3 right). The paper gives preliminary information about
a test site that has been selected for establishing
the precision and accuracy of available seismic
orientation. All three holes showed contained inclin methods in measuring the propagation velocity of
ations ranging between 10 and 20 cm. body waves. At this stage we have a preliminary
Seismic piezocone consisted of a standard cone comparison between MARW, CH, SDMT and
(10 cm2 area) measuring tip resistance, sleeve friction, SCPT. This activity will continue, and we intend
dynamic pore pressure and inclination. The piezocone to invite other producers of seismic piezocone to
was pushed by means of a Pagani TG73 – 150 pene participate into this research. As for the S waves
trometer. The seismic module of SCPT was equipped the comparison among various techniques is very
with two 3D accelerometers. The relative distance good. As for the P waves (less important for
between the accelerometers was 0.5 m. Therefore, engineering applications), it is possible to confirm
two waveforms were recorded for each hit by the that below the water table (as already observed
data acquisition system. The source consisted of by various researchers) Vp is almost constant and
a manual hammer (5 kg) hitting a wooden plate. Test equal to the propagation velocity in water. On the
interpretation was carried out by means of the cross- other hand, only CH test was capable of giving
correlation method and the true interval method. an estimate of Vp.
The SDMT is the combination of the Flat Dilat
ometer with an add-on seismic module for measuring
the shear wave velocity (Marchetti et al., 2008) and ACKNOWLEDGEMENTS
optionally also the compression wave velocity. The seis
mic module is an instrumented steel rod placed just We acknowledge the contribution to the investiga
above the DMT blade and equipped with two receivers tions given by Geoservizi snc, Geo-Energizer snc,
spaced 0.5 m. When a shear or compression wave is SolGeo Srl and Servizi Geologici S.a.s. Morbin F. &
generated at surface, it first arrives to the upper C. These people were very professional and making
receiver, then, after a delay, to the lower receiver. The possible the establishment of such a test site.
wave traces of the two receivers are amplified and digi
tized at depth and transmitted to the computer at sur
face. The software processes the signals and evaluates REFERENCES
the arrival delay, providing a real time interpretation of AGI 1994. Raccomandazioni sulle prove geotecniche di
the wave velocity. True interval method is used for laboratorio. Associazione Geotecnica Italiana.
evaluating the propagation velocity of body waves. ASTM D4015-15e1 2015. Standard Test Methods for
Figure 12 compares the body waves propagation Modulus and Damping of Soils by Fixed-Base Resonant
720
Column Devices, ASTM International, West Consho Marchetti, S., Monaco, P., Totani, G., Marchetti, D.
hocken, PA, www.astm.org. 2008. In Situ Tests by Seismic Dilatometer
ASTM D6635-15 2015. Standard Test Method for Perform (SDMT). Proceedings. From Research to Practice
ing the Flat Plate Dilatometer, ASTM International, in Geotechnical Engineering, ASCE Geotech. Spec.
West Conshohocken, PA, www.astm.org. Publ. No. 180 (honoring J.H. Schmertmann):
ASTM D2487-17e1 2017. Standard Practice for Classifica 292–311.
tion of Soils for Engineering Purposes (Unified Soil Robertson, P. K. 2016. Cone penetration test (CPT)-based
Classification System), ASTM International, West Con soil behaviour type (SBT) classification system—an
shohocken, PA, www.astm.org. update. Canadian Geotechnical Journal, 53(12):
ASTM D5778-20 2020. Standard Test Method for Elec 1910–1927.
tronic Friction Cone and Piezocone Penetration Testing Robertson, P. 2021. Personal communication to D. Lo
of Soils, ASTM International, West Conshohocken, PA, Presti.
2020, www.astm.org. Taylor, D.W. 1948. Fundamentals of Soil Mechanics, John
Casagrande. A. 1936. The Determination of the Pre – Con Wiley and Sons.
solidation Load and its Practical Significance, Proc. 1st Vucetic, M. & Dobry, R. 1991. Effect of Soil Plasticity on
ICSMFE, III, D – 34: 60–64. Harvard University Cam Cyclic Response. Journal of Geotechnical Engineering,
bridge, Mass. 117: 89–107.
721
© 2022 the Author(s), ISBN 978-1-032-31259-0
Bayesian estimation of small-strain shear modulus from offshore CPT tests

in the North Sea
B. Stuyts & C. Sastre Jurado
OWI-Lab, Vrije Universiteit Brussel, Brussels, Belgium
UGent, Ghent, Belgium
D. Gomez Bautista & A. Kheffache

OWI-Lab, UGent, Ghent, Belgium
ABSTRACT: The rapid expansion of offshore wind energy requires accurate and cost-effective site character
isation. The dynamic response of offshore wind turbine structures installed on monopile foundations strongly
depends on the stiffness of the subsoil where the small-strain shear modulus governs the behaviour during oper
ational loading. Direct measurement of the small-strain shear modulus with the seismic CPT (S-PCPT) is increas
ingly performed offshore. However, the additional cost associated with this test leads to incomplete data coverage.
In the past, correlations between cone tip resistance and small-strain shear modulus have been developed for
onshore conditions. The increasing availability of offshore site data has allowed a unique dataset of over 2000
small-strain shear modulus measurements from S-PCPT to be created for North Sea soil units. The comparison of
the small-strain shear modulus inferred from the S-PCPT data to the corresponding cone tip resistance allows
recalibration of the existing correlations. This paper proposes a Bayesian approach to update the correlation model
for sands proposed in the literature based on the developed dataset for offshore conditions in the North Sea.
1 INTRODUCTION 1.2 Offshore measurement of small-strain shear

modulus
1.1 Applications of small-strain shear modulus in
Quantifying the small-strain shear modulus (Gmax ) is
geotechnical design
usually done through in-situ measurements although
For offshore wind monopile foundations, the design is bender element testing on reconstituted samples can
mainly driven by the fatigue limit state (FLS) (Maes also be performed. Due to the lack of direct access to
et al. 2016) and is no longer dominated by the ultimate the seabed, onshore techniques such as cross-hole
limit state (ULS) which was traditionally the design measurements and surface wave measurements are
driver in offshore Oil & Gas platform geotechnical rarely performed.
design. An accurate assessment of the fatigue life of Seismic PCPT tests (S-PCPT) are the dominant
the foundation monopile requires the assessment of offshore measurement techniques. Figure 1 shows
non-linear soil response which shows a strong depend a sketch of the measurement system. A seismic
ence on strain level (Zuccarino et al. 2019). source at the mudline emits shear waves into the soil
While dynamic soil properties and strain depend which propagate downward with a velocity Vs gov
ence of the shear modulus have traditionally been erned by the mechanical properties of the soil mass.
researched mostly in the context of earthquake A dual geophone setup is preferred nowadays. The
engineering (Idriss and Boulanger 2008), the need difference in arrival times between the two geo
for accurate foundation response models for mono- phones can be derived by examining the cross-
pile foundations has seen a renewed interest in the correlation between the two signals. The receiver
characterisation of these properties for marine soils offset (Δz) divided by the difference in arrival time
in areas with low seismic activity. (Δt) then provides an estimate of the shear wave vel
Recent design methods for lateral foundation ocity (Vs ). The shear wave velocity can be converted
response of offshore monopiles in stiff clay (Byrne to a small-strain shear modulus when the density of
et al 2020) and sand (Burd et al. 2020) make use of the soil (ρ) is known (Equation 1).
the small-strain shear modulus as a governing soil
parameter. The improved accuracy of the foundation
response models has been demonstrated based on in-
situ monitoring data (Stuyts et al. 2020).
DOI: 10.1201/9781003308829-106
722
In recent years, the P-S suspension logging
technique has been increasingly used offshore
(Masters et al. 2019) but the empirical correlation
proposed here is based exclusively on S-PCPT
measurements.
Figure 2. Geospatial distribution of S-PCPT tests.
where the small-strain shear modulus Gmax , the cone

tip resistance qc and the vertical effective stress σ0vo
are given in kPa. The coefficients α = 1634 and β =
-0.75 are suggested by the authors.
Several other empirical correlations exist in the
Figure 1. Sketch of S-PCPT measurement system. literature (Hegazy and Mayne 2006). Most of these
correlations were obtained using regression methods
(i.e., least squares fitting) that do not directly quan
tify the uncertainties of predictions. To overcome
2 EXISTING GMAX CORRELATIONS this issue, the Bayesian framework can be adopted,
as it provides the probability distribution of the
Because the S-PCPT test is relatively more expen model parameters instead of a single best estimate.
sive than a conventional PCPT test, shear wave In addition, the Bayesian framework allows further
measurements are generally not available at every updating as additional data becomes available during
foundation location. Instead, several authors have future offshore campaigns.
sought to correlate shear wave velocity with conven In this paper, the principle of Bayesian updating
tional PCPT measurements to allow derivation of of the parameters of the correlation between cone
site-specific stiffness parameters. These studies are resistance and small-strain shear modulus will be
based on onshore measurement campaigns where outlined based on the aforementioned model pro
PCPT test data are compared to shear wave velocity posed by Rix & Stokoe (1991).
measurements at locations where both data types are
available.
3 NORTH SEA GMAX DATASET
The mostly widely used correlation for uncemen
ted silica sands was developed by Rix & Stokoe
(1991) based on calibration chamber measurements 3.1 Data sources
and the results of PCPT, S-PCPT and cross-hole A comprehensive dataset of shear wave velocity and
tests. Equation 2 shows the dependence of the small- small-strain shear modulus of offshore North Sea
strain shear modulus on the cone tip resistance and soils was assembled in the context of research on
the vertical effective stress level. monopile response back-analysis. Table 1 shows the
data compiled for four areas in the Belgian and
Dutch sector of the North Sea. The coordinates of
the S-PCPT test locations are shown in Figure 2.
The geology of the Borssele area and the Belgian
offshore area shows a layered profile with a surface
723
where:
- PðHjdataÞ is the posterior, the updated probabil
ity based on having made certain observations;
- PðdatajHÞ is the likelihood of the data;
- PðHÞ is the prior, the probability before seeing
the data which reflects prior knowledge.
As more observational data are introduced, uncer
tainties can be reduced through updating the mean
Figure 3. Overview of available data on clean sands and values and decreasing the variance of our hypothesis.
the correlation proposed by Rix & Stokoe. The valid range
of the correlation is represented by the dashed lines.
4.2 Probabilistic model
When applied to the derivation of small-strain shear
modulus from PCPT measurements, the multiplier α
and exponent β in Equation 2 are the principal
layer of holocene/pleistocene sand overlying a tertiary parameters to be updated by computing posterior
clay formation which dips towards the northeast. In the distributions using Bayesian inference.
Hollandse Kust area, mainly cohesionless sediments By taking the logarithm of Equation 2, a linear
are found, with channel features leading to lateral regression model can be used to model the relation
variability of the deposits. Finally, the area north of the between Gmax and qc (Equation 4).
Wadden Islands has softer soils at the surface, which
are classified as peat based on the normalized Robert
son chart (Robertson and Cabal 2015). A summary of
the available measurement and its classification
following the the normalized Robertson chart is shown
in Table 1.
This work will focus on clean sands (SBTn where the slope is the coefficient β and log10 α the
zone 6) since this is the soil type the correlation was intercept which will be noted by α0.
initially developed for. In Figure 3 the data are plot The dependent variable log10 ðGmax =qc Þ�is denoted
ted according to the normalisation suggested by Rix pffiffiffiffiffiffiffi�
by y and the independent variable log10 qc = σ0vo
& Stokoe (1991). Although the data confirms the by x. Then, the linear regression model can be writ
general trend suggested by the authors the bias ten in a probabilistic manner as in Equation 5
towards conservative estimates is clear and some (Gelman et al. 1995).
data fall outside the valid range of this correlation.
4 BAYESIAN UPDATING
in other words, the dependent variable y is
4.1 Introduction
assumed to follow a normal distribution with a linear
Bayesian theory treats uncertainty as degrees of trend as mean ( μ) and a standard deviation (ε). An
belief and it allows to combine information from error term (ε) is used to model the mismatch between
multiple sources for the purpose of updating prior the model and the actual data.
knowledge given new information. From After applying Bayes’ rule Equation 5 is the likeli
a geotechnical point of view, the Bayesian method hood of the data. Following the Bayesian method, prior
follows the same logic as the observational method distributions need to be assumed for the unknowns
(originally proposed by Terzaghi) and as a result it is parameters, to compute the term PðHÞ from Equation 3.
a powerful basis for inference (Baecher 2017). Standardization was performed on the left- and right-
The core of Bayesian methods is the Bayes’ Rule. hand side of Equation 4. Standardization consists of
It says that a hypothesis (i.e., parameters to be subtracting the mean from the data and dividing by the
inferred) expressed as probabilities can be modified standard deviation. One advantage of standardizing the
(i.e., updated) by observational information accord data is that the same weakly informative priors can
ing to the conditional probability (likelihood) of always be used without having to think about the scale
those observations were a certain hypothesis true or of the data. A common choice is to assume normal dis
not. If a hypothesis is denoted as H, Bayes’ Rule can tributions as priors for the intercept and slope and
be written in a “friendly form” (Equation 3). a HalfCauchy prior for the error term.
724
Table 1. S-PCPT data sources.
No. of measurements per soil type
Project site Locations Measurements 2 3 4 5 6 7
Hollandse Kust Area 85 2016 1 37 114 333 1484 47

Borssele Area 8 60 0 2 8 11 31 8
TNW Area 15 282 44 0 0 11 224 3
Belgian Area 7 85 0 20 42 8 10 5
Total 115 2443 45 59 164 364 1749 63
2 Organic soils-peats
3 Clays: clay to silty clay
4 Silt mixtures: clayey silt to silty clay
5 Sand mixtures: silty sand to sand silty
6 Sands: clean sands to silty sands
7 Gravelly sand to sand
Table 2. Statistics of α and β from MCMC.
Parameter Mean Standard deviation 95% HPD
α 2241 272 [1711, 2761]

β -0.747 0.016 [-0.778, -0.715]
Figure 4. Computed posterior distribution of correlation

model coefficients. The proposed values by Rix & Stokoe
are represented by vertical grey lines.
5 PROBABILISTIC RECALIBRATION OF GMAX Figure 5. Computed posterior distribution of the model

CORRELATION FOR COHESIONLESS SOILS error represented by the coefficient of variation.
The probabilistic models were built using PyMC3, Posterior Density (HPD) interval. The 95% HPD is
a Python package for Bayesian modelling based on the shortest interval containing the true value of the
Markov Chain Monte Carlo algorithms (Salvatier et al. parameter in question with a probability of 0.95.
2016).The result of a Bayesian analysis is a posterior Transformation uncertainty which is basically the
distribution, not a single value but a distribution of data scatter about the transformation model can be
plausible values given the data and the proposed expressed quantitatively by means of the coefficient
model and there are many methods for computing the of variation (Phoon and Kulhawy 1999). This uncer
posterior distribution. Here, the posterior was esti tainty will be noted by CVM and can be visualised by
mated numerically using an algorithm from the sampling from the posterior distributions as shown
Markov Chain Monte Carlo (MCMC) family, known in Figure 5. Transformation uncertainty follows
as No-U-Turn Sampler (NUTS) (Hoffman et al. 2014). a narrow distribution showing a modal value
The posterior distribution of the inferred parameters about 36%.
α and β can be visualized in Figure 4. The results Finally the inferred model is shown in Figure 6. The
show that the exponent of -0.75 suggested by Rix & average model that fits the data was represented by
Stokoe applies well to the data. However, the multi a black solid line. Note that the most probable value is
plier of 1634 suggested based on the calibration cham given by the mode of the posterior. In this case, the
ber dataset needed to be substantially increased for the mean is a good approximation since the posterior of
North Sea dataset. The statistics of these parameters both coefficients showed a Gaussian shape (Figure 4).
are summarized in Table 2. The mean and standard This model leads to an improvement in the model bias
deviation are shown as well as the 95% Highest- of 28%. This means that the distribution of the ratio of
725
The average model is useful in obtaining a best
estimate of the small-strain shear modulus. The
uncertainty on the estimate of Gmax can be assessed
thanks to the coefficient of variation and by assuming
a normal distribution. The provided statistics can be
used as a basis for a reliability analysis for e.g. lateral
pile response. The effect of the uncertainty on Gmax
on the natural frequency estimate could thus be
quantified.
Following the Bayesian updating philosophy, the
reported statistics of the model coefficients can be
used to build priors distribution for further calibra
tion of the model when new project-specific infor
Figure 6. Recalibrated correlation model based on North mation becomes available.
Sea data: average and standard deviation.
ACKNOWLEDGEMENTS
predicted to measured Gmax shifts closer to unity (the
original Rix & Stokoe correlation showed a bias The authors would like to acknowledge the support
of 0.65. of the Belgian Ministry of Economic Affairs through
The dashed lines in Figure 6 represent the average the ETF project WINDSOIL project. The support of
standard deviations of the model prediction based on VLAIO through the De Blauwe Cluster SBO SOILT
the average CVM equals to 0.36. The impact of the stat WIN project is also acknowledged.
istical uncertainty of the correlation model is repre The geotechnical data from the Dutch offshore
sented by the semitransparent grey band to illustrate wind farms (RVO.nl) was used under a Creative
the 95% HPD. Both errors, statistical and transform Commons license.
ation uncertainties are not constant, increasing when
the ratio decreases. Indeed, for shallow
layers, less data is available and noise in the data (e.g.
due to waves propagating down the cone rod interfer REFERENCES
ing with the shear wave propagating through soil)
Baecher, G. B. (2017). Bayesian thinking in geotechnics. In
introduces more scatter in the small-strain shear modu
Geo-Risk 2017, pp. 1–18.
lus measurements. It is therefore possible to conclude Burd, H. J., D. M. Taborda, L. Zdravković, C. N. Abadie,
that the model does not provide
pffiffiffiffiffiffiffi
ffi accurate predictions B. W. Byrne, G. T. Houlsby, K. G. Gavin, D. J. Igoe,
when the ratio qc = σ0 vo is lower that 1000 R. J. Jardine, C. M. Martin, et al. (2020). Pisa design
approximately. model for monopoles for offshore wind turbines: appli
cation to a marine sand. Géotechnique 70(11),
1048–1066.
6 CONCLUSIONS AND RECOMMENDATIONS Byrne, B. W., G. T. Houlsby, H. J. Burd, K. G. Gavin,
D. J. Igoe, R. J. Jardine, C. M. Martin, R. A. McAdam,
This paper presents the results of Bayesian infer D. M. Potts, D. M. Taborda, et al. (2020). Pisa design
model for monopiles for offshore wind turbines: appli
ence to recalibrate the correlation model between cation to a stiff glacial clay till. Géotechnique 70(11),
cone resistance and small-strain shear modulus 1030–1047.
suggested by Rix & Stokoe (1991). This recali Gelman, A., J. B. Carlin, H. S. Stern, & D. B. Rubin (1995).
bration is based on a dataset of 1749 small-strain Bayesian data analysis. Chapman and Hall/CRC.
modulus measurements with the S-PCPT for Hegazy, Y. A. & P. W. Mayne (2006). A global statistical
North Sea sands. correlation between shear wave velocity and cone pene
The results show that the exponent suggested by tration data. In Site and geomaterial characterization,
Rix & Stokoe is largely confirmed by the assembled pp. 243–248.
dataset. However, the multiplier α needs to be sub Hoffman, M. D., A. Gelman, et al. (2014). The no-u-turn
sampler: adaptively setting path lengths in hamiltonian
stantially increased as the original Rix & Stokoe cor monte carlo. J. Mach. Learn. Res. 15(1), 1593–1623.
relation underestimates the small-strain shear Idriss, I. M. & R. W. Boulanger (2008). Soil liquefaction
modulus for nearly all cases. The recalibrated model during earthquakes. Earthquake Engineering Research
provided in Equation 6. Institute.
Maes, K., A. Iliopoulos, W. Weijtjens, C. Devriendt, &
G. Lombaert (2016, August). Dynamic strain estimation
for fatigue assessment of an offshore monopile wind tur
bine using filtering and modal expansion algorithms.
Mechanical Systems and Signal Processing 76-77,
592–611.
with a coefficient of variation of the prediction error Masters, T. A., P. Juszkiewicz, A. Mandolini, &
of 0.36. H. Christian (2019). A critical appraisal of the benefits
726
of and obstacles to gaining quality data with offshore Salvatier, J., T. V.Wiecki, & C. Fonnesbeck (2016). Prob
seismic cpt and ps logging. In Offshore Technology abilistic programming in python using pymc3. PeerJ
Conference. OnePetro. Computer Science 2, e55.
Phoon, K.-K. & F. H. Kulhawy (1999). Evaluation of geo Stuyts, B., W. Weijtjens, C. Devriendt, H. Versteele, &
technical property variability. Canadian Geotechnical C. V. Haute (2020). Monopile lateral response cali
Journal 36(4), 625–639. bration from insitu monitoring data. In 4th Inter
Rix, G. J. & K. H. Stokoe (1991). Correlation of initial tan national Symposium Frontiers in Offshore
gent modulus and cone penetration resistance. In Calibra Geotechnics, pp. 3557. Deep Foundations Institute.
tion chamber testing. New York: Elsevier, pp. 351–362. Zuccarino, L., A. D. Morandi, & O. Luca (2019). Dynamic
Robertson, P. & K. L. Cabal (2015). Guide to Cone Pene properties of offshore sands. In OTC-29590-MS. Off
tration Testing. Technical report. shore Technology Conference.
727
© 2022 the Author(s), ISBN 978-1-032-31259-0
Factors influencing CPTU Nkt for marine clay in Singapore reclaimed land
C. Tanaka, M. Angeles & J.Y. Wong
Arup Singapore Pte. Ltd., Singapore
ABSTRACT: Cone penetration tests with porewater pressure measurements (CPTU) are extensively adopted
in Singapore due to the presence of soft soils from the young deposits of the Kallang Formation. The most
prominent member of this formation is the soft Singapore marine clay, which can be present in thicknesses of up
to 50m. Singapore is a relatively small island, and as such, over the past few decades has embarked on numerous
reclamation projects to increase land to support their growing economy and population. Therefore, significant
data in reclaimed land overlying soft deposits is available, especially in the east of Singapore. This has presented
the opportunity to review the factors that influences Nkt in the marine clay located within this area. Nkt plays
a significant role in the evaluation of the undrained shear strength (cu), which is often derived from site-specific
calibrations. This paper summarizes the CPTU results carried out in reclaimed land to the east of Singapore. It
discusses the calibration of Nkt with other site data, and proposes a range of Nkt to be adopted in this location.
1 INTRODUCTION Direct dumping was utilized at deeper waters of 6m

to 8m, while hydraulic filling was utilized for shal
Reclamation works have been carried out in the east lower depths of 2m to 6m.
of Singapore in phases, starting with the original rec
lamation scheme throughout the 1970’s and 1980’s.
Subsequently, further lands were reclaimed for the
development of airport infrastructure from the late
1980’s till early 2000’s, forming the land as it is seen
today as shown in Figure 1. Ground improvement
works were also conducted to improve the soft
marine clay in most of the reclaimed site. A series of
site investigation (including CPTU and boreholes)
were carried out on site along the course of the rec
lamation and ground improvement works to under
stand the geological conditions and strength of the
ground. This paper presents the interpreted
undrained shear strength (cu) results from CPTU car
ried after all ground improvement works had been
completed, post 2014. It compares the cu values
obtained at two different locations across the site
(Location A and B, Figure 1). It also presents
a calibration exercise of Nkt while drawing correl
ations to the site history.
2 SITE HISTORY AND GEOLOGY

Figure 1. Singapore topography maps in 1970s (top left)
The site locations A and B were reclaimed in differ
and 2010s (top right) and Location of site (bottom).
ent phases as part of reclamation works that took
place from 1992 to 2005 where 2,000 ha of land
were reclaimed by placing 272 million m3 of granu As the reclaimed site was underlain by 50m of
lar fill in seawater with a depth up to 15m (Bo et al. soft marine clay, prefabricated vertical drains (PVD)
2005). This was done in various ways, including coupled with surcharging were proposed to accelerate
direct dumping, hydraulic filling and sand spreading. consolidation of the underlying deposits during
DOI: 10.1201/9781003308829-107
728
Table 1. Summary of site works at Locations A and B. reclamation works from 1992 to 2005 (Choa et al.
2001) and post reclamation works from 2014 to 2019.
Location A Location B A summary of reclamation and known ground
improvement history across the two locations can be
Reclamation Phase 1A: 1992 to Phase 1B: 1993 to
found in Table 1.
Period1 1997 1999
Extensive site investigations have been carried
Reclamation For sites from -5mCD Hydraulic filling by
Method1 to -10mCD: Direct cutter suction out post ground improvement for land preparation
Dumping from bottom dredgers for future development from 2015 to 2020. More
opening barges then than 2000 CPTU and 1000 boreholes were conducted
hydraulic filling by on site to understand the post-improvement condi
cutter suction dredgers tions as shown in Figure 2.
Others: Hydraulic The site investigation show the ground condi
filling by cutter suc
tions across the site are highly variable due to
tion dredgers
the multiple reclamation methods. Generally, the
Ground No known post rec Post reclamation
Improvement lamation ground 1993 to 1999: PVD
site consists of approximately 10m to 15m of
History improvement installed at square thick reclamation fill underlain by 5m to 40m of
grids of 1.5 m/ Kallang Formation, followed by the stiff deposits
1.7 m with surchar of the Old Alluvium. The most prominent
ging of minimally member of the Kallang formation is the soft
18 months1 marine clay, with thickness varying from 20m
Land preparation and 40 m at Locations A and B, respectively as
for future shown in Figure 3.
development Site investigations carried out includes in-situ
2014 to 2018: PVD tests: CPTU and vane shear tests (VST) and labora
installed at square
grids of 1.0 m/
tory tests: Triaxial unconsolidated undrained tests
1.2 m with surchar (UU) and one dimensional Oedometer tests. These
ging of minimally 6 tests are used to determine the undrained shear
months strength of the ground. From the site investigation,
the Marine clay has varying consolidation states as
1
Note: Choa (1995) discussed in Chapter 5.
Figure 2. Site investigation plan.
Figure 3. Geological long section.

729
3 UNDRAINED SHEAR STRENGTH
With the varying range of recommended Nkt
INTERPRETATION OF CPTU
values, calibration with site-specific undrained
shear strength data were utilised for more accurate
The undrained shear strength (cu) of the soil from interpretation of the cone factor. The undrained
CPTU is derived using standard correlation (Lunne shear strength data was obtained from the meas
et al 2002): ured uncorrected field VST, Laboratory Triaxial
UU test, as well as correlation with one-
dimensional oedometer tests pre-consolidation
pressure (0.22 Pc’).
where:
qt = corrected cone resistance based on the meas 4 CONSOLIDATION STATE OF CLAY
ured pore pressure; σv = total overburden stress; and
Nkt = empirical cone factor. The consolidation state of the ground is deter
The corrected cone resistance (qt), is calculated mined by comparing the in-situ and laboratory cu
from the measured cone resistance (qc) through the with the empirical normally consolidated (NC)
formula line with respect to existing ground level. The
NC line is derived through the formula by Mesri
(1975),
where a = net area ratio of 0.8 to 0.82; and u2 = pore

pressure measured just behind the cone.
Various studies have been conducted to determine where σv’ = effective overburden stress with respect
the range of Nkt values as shown in Table 2. It is to existing ground level.
observed that a wide range of values have been In general terms, the cu that falls below the NC
recommended depending on the varying soil condi line is considered as under-consolidated with respect
tions and locations. to existing ground level, while those that lie above
Theoretical studies have also related the Nkt value are considered over-consolidated with respect to
to plasticity index (Ip). Singapore Marine Clay prop existing ground level.
erties at this location have been extensively studied
by Bo et al (2000), and a Nkt correlation is proposed
based on plasticity index of the clay where: 5 OBSERVATIONS OF NKT VALUES
5.1 Comparison of CPTU cudata with VST, UU and

Oedometer test data
The undrained shear strength interpreted from
With a typical Ip of 40% for the marine clay across various in-situ and laboratory tests for Locations
the site as seen in Figure 4, an Nkt value of approxi A and B are summarised in Figures 5 and 6,
mately 14 is recommended. respectively. Interpretations using Nkt of 14
According to Bo et al. (2000), the Nkt will and 20 at the two locations are plotted for
decrease with increasing Ip although studies by Aas comparison.
et al. (1986) suggest that Nkt will increase with For Location A, Nkt value of 14 appears to over
increasing Ip; La Rochelle et al. (1988) did not find estimate the CPTU cu data as most of the cu data
any correlation between Nkt and Ip. from the other tests fall below the CPTU cu data as
shown in Figure 5a. Nkt value of 20 better correlates
with the other site investigation data as shown in
Table 2. Nkt recommendations from literatures. Figure 5b.
For Location B, Nkt value of 20 appears to under
Nkt estimate the CPTU cu data as most of the cu data from
range Type of Soil Reference the other tests lies above the CPTU cu data as shown in
Figure 6a. Nkt value of 14 appears to better correlate
10 – 15 Normally Consolidated De Ruiter (1982) with the other site investigation data as shown in
Clay Figure 6b.
15 – 20 Under-Consolidated Clay De Ruiter (1982)
15 – 21 Lightly Over-Consolidated Dobie (1988)
Singapore Marine Clay 5.2 Comparison of CPT cu data with NC line
9 – 17 Clay with PI = 33% – 45% Lunne et al. As indicated in Table 1, at Location A, there is
(2005)
not known post reclamation ground improvement
730
Figure 4. Atterberg limits at locations A and B.
Figure 5b. Location A - cu plot for Nkt = 20.

Figure 5a. Location A - cu plot for Nkt = 14.
works been carried out. Based on Terzaghi con On the other hand, at Location B, two rounds of
solidation theory, the marine clay is still undergo extensive ground improvement works were carried
ing consolidation from the reclamation filling out. By Terzaghi consolidation theory, the marine
carried out to form the site’s current ground clay is expected to be at least in normally consoli
level. This is in line with interpreted CPT cu with dated state or over-consolidated. This is in line with
higher Nkt where the CPT cu data falls below the interpreted CPT cu with lower Nkt where the CPT cu
NC line. data lies above the NC line.
731
Figure 6a. Location B - cu plot for Nkt = 14. Figure 6b. Location B - cu plot for Nkt = 20.
Based on the Nkt variability presented, a site-

6 CONCLUSION specific calibration is recommended for derivation of
cu, especially when cu is critical for the geotechnical
This paper has presented CPTU data obtained from design (e.g., assessment of long-term settlements or
two different areas (Location A and B), located excavations schemes), and when the site has varying
within the same site, but with different reclamation consolidation history, such as the one presented in
and ground improvement histories. this paper.
The CPTU data was used to determine cu via
standard correlations (Lunne,2002). The paper
has highlighted that the empirical factor Nkt has REFERENCES
significant influence in the determination of cu.
Therefore, this paper has presented an Nkt cali Aas G., Lacasse S., Lunne T., and Hoeg K. 1986. Use of
bration exercise by comparing the interpreted cu in situ tests for foundation design on clay. Proceedings
from CPTU with other geotechnical tests, includ of the ASCE Specialty Conference: 1–30
ing in-situ vane shear tests, laboratory Triaxial Bo, M. W., Chang, M. F., Arulrajah, A., & Choa, V. 2000.
Undrained shear strength of the Singapore marine clay at
unconsolidated undrained tests and one dimen
Changi from in-situ tests. Geotechnical Engineering, 31(2).
sional Oedometer tests. Bo, M. W., Chu, J., & Choa, V. 2005. The Changi east rec
The calibration exercise showed that for lamation project in Singapore. Elsevier Geo-
Location A, which is under-consolidated to Engineering Book Series 3: 247–276. Elsevier.
existing ground level, an Nkt value closer to 20 Campanella, R. G., RG, C., & PK, R. 1982. Pore pressures
provides consistency with the cu obtained from during cone penetration testing.
the in-situ and laboratory tests data as well as Choa, V. 1995. Changi east reclamation project. In Com
with its site history. For location B, an Nkt pression and consolidation of clayey soils: 1005–1017.
value closer to 14 was selected on the same Choa, V., Bo, M. W., & Chu, J. 2001. Soil improvement
works for Changi East reclamation project. Proceedings
basis.
of the Institution of Civil Engineers-Ground Improve
The Nkt values between 14 (Location B) to 20 ment 5(4): 141–153.
(Location A) recommend based on a site-specific Chu, J., Bo, M. W. & Arulrajah, A. 2009. Reclamation of
data calibration is consistent with the general range a slurry pond in Singapore. Proceedings of the Institu
recommended in other available literature and publi tion of Civil Engineers-Geotechnical Engineering 162
cations (Lunne,2002). (1): 13–20.
732
De Ruiter, J. 1982. The static cone penetration test: Lee, S. L., Karunaratne, G. P., Yong, K. Y., Tan, S. A., &
State-of-the-art report. Proc. of the Second European Tan, T. S. 1989. Soft clay properties in a layered clay-
Symposium on Penetration Testing, Amsterdam, sand reclamation.
May 1982 2:389–405 Lunne, T., Randolph, M. F., Chung, S. F.,
Dobie, M. J. D. 1988. A study of cone penetration tests in Andersen, K. H., & Sjursen, M. 2005. Comparison of
the Singapore marine clay. International Symposium on cone and T-bar factors in two onshore and one off
penetration testing; ISOPT-1. 1: 737–744. shore clay sediments. Frontiers in Offshore Geotech
La Rochelle, P., Zebdi, M., Leroueil, S., Tavenas, F., & nics (Proc. ISFOG, Perth): 981–989.
Virely, D. 1988. Piezocone tests in sensitive clays of Lunne, Tom, John JM Powell, and Peter K. Robertson. Cone
eastern Canada. International Symposium on penetra penetration testing in geotechnical practice. CRC Press,
tion testing; ISOPT-1. 1: 831–841. 2002.
733
© 2022 the Author(s), ISBN 978-1-032-31259-0
Parameters affecting the CPT resistance of reconstituted sands

Y. Tian & B.M. Lehane
The University of Western Australia, Perth, WA, Australia
ABSTRACT: Despite the popularity of the Cone Penetration Test, uncertainty remains regarding the
stress level dependence of qc in a sand of constant relative density, the effect of the mineralogy and the
effect of saturation. This paper provides greater clarity on these effects by presenting results from a large
series of drained cone penetration tests performed under controlled conditions in a laboratory pressure
chamber. The experiments involve dry and fully saturated silica and carbonate sands placed at a number
of relative densities (Dr) and tested at different stress levels and overconsolidation ratios. The observa
tions are compared with existing relationships proposed between qc and Dr and highlight the approximate
nature of such relationships.
1 INTRODUCTION
The relative density (Dr) of coarse-grained soils

is usually assessed indirectly from correlations where σ' = σ'v and n = 0.71
with in-situ test parameters due to the difficul
ties in obtaining undisturbed samples of these
materials. The relationship between CPT cone
resistance (qc) and Dr has been assessed from
experiments in laboratory-based pressure cham
bers that involved CPTs in reconstituted sand where σ'= p' and n = 0.55
samples prepared to a range of Dr values and To allow for shallow penetration effects, which
stress states. The relationships generally employ are important at geotechnical centrifuge scale,
the following format: Lehane et al. (2022) used results from over 70 centri
fuge tests in range of normally consolidated silica
sands to propose the following relationship:
where qc1N = stress normalized cone resistance; where σ' = σ'v, n = 0.7, z=depth, dc = diameter of the
C1 and C2 are empirical constants, patm = atmos cone penetrometer and a is assessed from an initial
pheric pressure (100 kPa), n = stress level expo approximate estimate of Dr:
nent (typically between 0.5 and 0.7) and σ' =
effective stress at the level of the cone tip.
The value of σ' adopted is typically the vertical
effective stress (σ'v) in normally consolidated deposits
and the mean effective stress (p') in overconsolidated Kulhawy & Mayne (1990) recognized the import
deposits. The use of p' in overconsoliated deposits ance of ageing, overconsolidation ratio (OCR) and
reflects the dependence of qc on the in-situ horizontal sand compressibility and proposed:
effective stress (σ'h) illustrated by Houlsby & Hitch
mann (1988). Popular relationships in the form of
Equation (1) proposed by Lunne & Christoffersen
(1983) and Baldi et al. (1986) respectively are:
DOI: 10.1201/9781003308829-108
734
where n = 0.5; CF = compressibility factor between 2.2 Sample preparation
0.91 & 1.09 (higher value for more compressible
Sand samples are prepared in calibration cham
sand) and Cage =1.2+0.05log [t/100], where t is the
bers by pluviation. A sand hopper is fitted on
time in years.
two rails which enables it to move both horizon
This paper presents a re-assessment of these cor
tally and vertically. A constant drop height is
relations using a new series of pressure chamber
maintained during the preparation of uniform
tests conducted at the University of Western Austra
sand while the hopper cycles in the horizontal
lia (UWA). These tests were initially performed to
direction. Different relative densities are
test the general applicability of Equation (4) derived
achieved by varying the horizontal travelling
from centrifuge CPT data. The test series was subse
speed, the drop height and the sieve situated at
quently expanded to examine the effect of sand min
the bottom of the hopper.
eralogy (by comparing qc data in a silica and
Two types of sand are used in this study namely
carbonate sand), sand saturation level (dry and fully
(i) UWA sand, which is a fine silica sand that is
saturated) and overconsolidation ratio (hence σ'h).
sourced commercially and has been tested exten
The test results allow general conclusions to be
sively at the University of Western Australia
drawn regarding the relationship between qc and Dr
(UWA) and (ii) Ledge Point carbonate sand, which
for reconstituted sands.
is obtained from a sand dune near Ledge Point,
Western Australia. Properties of the two types of
sand are listed in Table 1. A miniature cone with
2 TEST PROGRAMME
the diameter (d c) of 7mm is employed for the tests
described here. The corresponding dc/D50 ratio is
2.1 Test setup in excess of 20 for both sands, as recommended by
A schematic diagram of the calibration chamber test Bolton et al. (1999) to avoid particle size effects.
is shown in Figure 1. The soil sample is created by
pluviation of the sand into a cylindrical chamber,
which is then loaded vertically to the desired vertical Table 1. Properties of sands used in the tests.
effective stress via a rigid top plate. This plate has
D50
circular openings to allow for penetrometer access to
the sand. The vertical effective stress is applied by Sand (mm) Specific gravity emax emin φ'cs
a hydraulic jack and is monitored by a load cell
while the cone is driven into the sample at a constant UWA
speed by the actuator located on the steel frame. The 0.18 2.67 0.78 0.50 33o
silica sand
maximum imposed vertical effective stress imposed Ledge Point
in the experiments described here is 75kPa. The 0.21 2.76 1.17 0.78 35o
carbonate sand
walls of the chamber are rigid and a flexible Teflon
sheet is bonded to the inner wall to minimize * emax and emin are the void ratio limits and φ'cs is the crit
friction. ical state friction angle of the sand
2.3 Test conducted

Over 80 CPTs were conducted in the two sands
placed in 12 chambers at a variety of relative densities
and consolidated to three vertical effective stress
levels (25, 50 and 75kPa). The effects of the degree
of saturation (Sr) and overconsolidation ratio (OCR)
in silica sand were also examined. A summary of the
scope of testing is provided in Table 2, noting that
many CPTs were repeated to assess variability and
repeatability. Additionally, CPTs were performed to
assess if the proximity of the boundary to the pene
trometer had an influence on the qc values measured.
2.4 Boundary effect

Initial trials were conducted to assess if the
planned distance of a penetrometer installation
location to the chamber boundary (s) influenced
the tip resistance.
The steady state qc values recorded at σ'v=50 kPa in
Figure 1. Schematic of the calibration chamber test.
medium dense and dense UWA silica sand at 8, 12 and
735
Table 2. Summary of the tests conducted. Dr values but the same vertical effective stress of 50
kPa. It is seen that the qc values in dry silica sand
σ'v increase up to a penetration depth of about 5dc but sta
bilize thereafter at values of about 2.5 MPa and 8 MPa
Sand Dr (kPa) Sr OCR qc1N,avg in the very loose and medium dense silica samples
respectively. Evidence from other chamber tests indi
UWA 0.1 25, 50, 75 1 1 28 cated that the slight reduction of qc with depth seen for
Silica 0.63 25, 50, 75 0 1 118 these samples is associated with a modest reduction in
Sand 0.97 25, 50, 75 0 1-4 283 Dr with depth in these particular samples and is not
0.8 50 1 1 200 related to reduced vertical effective stress due to the
Ledge Point 0.06 25, 50, 75 1 1 37 potential presence of side friction. It should be noted
carbonate sand 0.67 25, 50, 75 0 1 163 that the very loose sand was water-pluviated and the
0.95 25, 50, 75 0 1 469
27 cone diameters (dc) from the chamber wall are pre

sented on Figure 2. It is evident that qc values
increased close to the chamber wall at both Dr values
when the s/dc value was less than about 12 and that
there is a slightly greater relative effect of the boundary
influence in the denser sand. Bolton et al. (1999) found
a very similar result in centrifuge CPTs conducted in
a circular strongbox, observing no significant increase
in qc due to the proximity of the boundary for silica
sand with Dr=81% at s/dc ratios in excess of 11. All
CPT data reported in the following correspond to test
results recorded at s/dc>12.
Figure 2. Influence of normalized proximity to rigid cham

ber boundary (s/dc) in medium dense and dense UWA
silica sand.
It is of interest to note that qc values recorded at

s/dc=8 in Ledge Point carbonate sand were identi
cal to those recorded at s/dc=12 and 27. This
reduced boundary influence may arise due to the
higher compressibility of the carbonate particles.
3 TEST RESULTS
3.1 Normally consolidated dry silica sand

Typical results obtained during penetration of a 7 mm
diameter cone in a pressure chamber are shown in Figure 3. qc profiles measured at σ'v=50 kPa at three rela
Figure 3(a). These were measured in three separate nor tive density values for (a) UWA Silica sand (b) Ledge Point
mally consolidated UWA sand samples with different Carbonate sand.
736
Figure 4. It is evident that, while the steady state qc
value increases with the vertical effective stress in the
sample (σ'v), qc is not linearly proportional to σ'v. This
tendency is typical of sands and is reflected by the ‘n’
exponent in the normalized cone resistance, qc1N
(Equation 1b). It is shown on Figure 4(b) that the data
from the three stress levels are unified when qc1N is
plotted using n=0.7. Data from all 36 CPTs conducted
in normally consolidated UWA silica sand were com
piled to find a best fit relationship in the form of Equa
tion (1). The following relationship was found to
predict measured Dr values to within 10%.
where σ' = σ'v and n = 0.7

Equation (6) is compared with Equation (4) on
Figure 5 which was derived independently for the
same sand from an extensive series of centrifuge CPTs
performed over a larger stress range (from 5 kPa to
330kPa). The plotted centrifuge qc values correspond
to those at the steady state when the hyperbolic tan
gent term in Equation (4a) is unity. Very good agree
ment between both data sets is observed, particularly
in loose and medium dense sand. The reason for the
higher steady state qc values recorded in dense sand in
the centrifuge compared to those in the chamber tests
at the same stress levels is under investigation.
Equation (6) is also compared with Equation (2)
and (3) in Figure 6, which have previously been prop
osed for normally consolidated silica sand. Equation
(2) has been adjusted to correspond with a stress exp
onent (n) of 0.7 at the average stress level used in the
chamber test of 50kPa to facilitate this comparison. It
is evident that agreement is very good between all
three equations for Dr <0.7. However, Equation (6)
implies greater qc values in very dense UWA sand,
which were also found the centrifuge CPTs.
3.2 Normally consolidated dry carbonate sand

Typical qc profiles measured in normally consoli
dated Ledge Point carbonate sand are shown in
Figure 4. Tip resistance in medium dense UWA silica sand Figure 3b. A comparison with the the silica sand
(Dr=0.63) under three stress levels: (a) qc in MPa, (b) qc data on Figure 3a indicates higher qc values are
normalized. developed in the carbonate sand at the same density
and stress level; this trend reflects higher friction
reducing trend of qc with depth in this sample indicates angles of the carbonate particles. The qc profiles in
that this process resulted in slightly looser sand at the carbonate samples also tended to show greater
depth. fluctuations or ‘peakiness’, which is attributed to the
It is also seen on Figure 3(a) that qc values in dense samples’ greater compressibility. Regression ana
silica sand increase with depth and do not tend to lyses of the calibration chamber data indicated a best
a steady state until a penetration depth of about 15-20 fit stress level exponent (n) of 0.7, as for the silica
dc. This trend is typical of that shown by other tests sand and the following best fit relationship:
indense silica sand. Normalized depths, z/dc, required
to develop steady state conditions in carbonate sand
are lower than for silica sand, as seen on Figure 3b.
The influence of stress level on the qc value in
a medium dense sample (Dr=0.63) is illustrated on where σ' = σ'v and n = 0.7
737
Figure 7. Comparison of relationships derived for silica
and carbonate sand (Equation 6 and 7) with Equation (5)
proposed by Kulhawy & Mayne (1990).
Figure 5. Comparison of qc relationship with Dr and σ'v 3.3 Overconsolidated dry silica sand
predicted by Equations (4) and (6) for UWA silica sand
(for steady state conditions). Five CPTs were conducted in dense silica sand (Dr
=0.97) at σ'v values of 75 kPa, 50 kPa and 25kPa
after the samples had been pre-loaded to either 100
kPa or 75 kPa. Figure 8a shows the normalized tip
resistance calculated using σ'v and indicates that this-
normalized value increases with OCR. Figure 8b
plots the same data but uses σ'h to normalize the qc
value for stress level, defined here as:
where σ'h was estimated from the following expres

sion of Mayne & Kulhawy (1982) using the friction
angles (ϕ') given in Table 1:
Figure 6. Comparison of normalized qc relationship with

Dr predicted by Equations (2), (3) and (6) for normally It is seen that the data at all OCRs are unified
consolidated silica sand. on Figure 8b, confirming the suitability of
employing σ'h to normalize the cone resistance
proposed by Houlsby & Hitchman (1988). It is
Equation (7) is compared on Figure 7 with Equation notable, however, that qc1N,h continues to
(6) derived from the calibration chamber data in UWA increase with depth and did not appear to reach
silica sand. This comparison illustrates the relative a steady state value.
magnitudes of qc values in the two sand types. Equa Values of relative density predicted by Equa
tion (5) is also compared with Equation (6) and (7). tions (3) and (5), which are intended for use in
A stress level of 50 kPa (which was the mean used in sands at all OCRs, are compared with the meas
the chamber tests) was assumed to enable Equation (5) ured relative density in Table 3. It is seen that
to be plotted on this figure as this equation employs both equations over-estimate Dr and that the
a stress level exponent of 0.5. In addition, recom degree of over-estimation increases with OCR.
mended compressibility factors (CF) of 0.91 and 1.09 Further studies are ongoing at UWA to investi
were adopted for the silica and carbonate sands gate qc trends in overconsolidated sands.
respectively. It is evident from the comparison that
general agreement is observed between Equation (5)
and Equations (6) and (7) both in terms of the trend 3.4 Effect of saturation
with Dr and the relative magnitudes of normalised The qc values measured in dry and fully saturated
cone resistance. (Sr=0 and 1) silica sand at a relative density of
738
Figure 9. Normalized tip resistance in dry and fully satur
ated silica sand (Dr=0.8) under various penetration rate.
Table 3. Performance of Equations (3) and (5) in dense

silica sand.
σ'v
Dr, calc from Dr, calc from
Dr,meas OCR (kPa) Equation(3) Equation(5)
1 75 1.06 1.00
1.33 75 1.14 1.09
0.97 1.5 50 1.10 1.07
2 50 1.16 1.13
4 25 1.19 1.19
* Dr,calc values quoted are the average predicted values

Figure 8. Tip resistance in overconsolidated dense silica
sand: (a) normalized by vertical effective stress; (b) normal
ized by horizontal effective stress. values arise due to dry crushed carbonate par
ticles forming a slightly enlarged cone base area
on the cone for Sr=0 tests. Jamialkowski et al. (2003)
0.8 are compared on Figure 9. This figure also
also found slightly lower qc values in saturated sand,
plots qc profiles measured at cone penetration
which is not evident on Figure 9 for UWA sand.
rates between 0.1mm/s and 1.0mm/s. It is evident
that qc values, given experimental inaccuracies,
are essentially identical in all cases. Additional
4 CONCLUSIONS
CPTs performed in loose carbonate sand prepared
at Sr=0 and Sr=1 also showed no dependence of
qc on Sr. These findings contrast with centrifuge Over 80 CPTs were performed in calibration
data reported by Giretti et al. (2018) who found chambers at UWA to study the effect on qc of
higher qc values in dry carbonate sand compared stress level, mineralogy, overconsolidation ratio
with saturated sand and suggest that the higher and degree of saturation. Relationships between
739
stress normalized end resistance (qc1N) and rela Bolton, M.D., Gui, M.W., Garnier, J., Corte, J.F.,
tive density (Dr) are developed for UWA silica Bagge, G., Laue, J. & Renzi, R., 1999. Centrifuge cone
sand and Ledge Point carbonate sand. These are penetration tests in sand. Géotechnique, 49(4),
shown to be generally consistent with previously pp.543–552.
published correlations indicating that the relation Giretti, D., Been, K., Fioravante, V., & Dickenson, S.,
ship between qc1N and Dr is relative insensitive 2018. CPT calibration and analysis for a carbonate sand.
to a sand type of given mineralogy. The testing Géotechnique, 68(4),345–357.
Houlsby, G.T. & Hitchman, R., 1988. Calibration chamber
programme highlighted the need for further stud tests of a cone penetrometer in sand. Geotechnique, 38
ies into the development length of the CPT in (1),pp.39–44.
dense sands and the cone resistance in overconso Jamiolkowski, M., Lo Presti, D.C.F. & Manassero, M.,
lidated dense sand. 2003. Evaluation of relative density and shear strength
of sands from CPT and DMT. In Soil behavior and soft
ground construction (pp. 201–238).
ACKNOWLEDGEMENT Kulhawy, F.H. & Mayne, P.W., 1990. Manual on estimating
soil properties for foundation design (No. EPRI-EL
The first Author acknowledges the financial support 6800). Electric Power Research Inst., Palo Alto, CA
through an Australian Government Research Train (USA); Cornell Univ., Ithaca, NY (USA). Geotechnical
ing Program (RTP) Scholarship. Engineering Group.
Lehane, B., Zania, V., Chow, S. & Jensen, M., in press.,
2022.“Interpretation of centrifuge CPT data in nor
mally consolidated sands”. (under final review),
REFERENCES Geotechnique.
Lunne, T. & Christoffersen, H.P., 1983, May. Interpretation
Baldi, G., Bellotti, R. & Ghionna, V., 1986. Interpretation of cone penetrometer data for offshore sands. In Off
of CPTs and CPTUs, Part II: Drained Penetration in shore Technology Conference. OnePetro.
Sands. In Proc. of 4th International Geotechnical Sem Mayne, P.W. & Kulhawy, F.H., 1982. Ko-OCR relation
inar on Field Instrumentation and In Situ Measurements, ships in soil. Journal of the Geotechnical Engineering
Singapore. Division, 108(6),pp.851–872.
740
© 2022 the Author(s), ISBN 978-1-032-31259-0
Practical experience with cone penetration in frozen soils

N.G. Volkov & I.S. Sokolov
GEOINGSERVICE (Fugro Group), Moscow, Russia
R.A. Jewell
Fugro GeoConsulting, Brussels, Belgium
ABSTRACT: This paper summarizes recent practical experience with CPT investigations at several perma
frost sites in Russia. Some of the investigations were performed in unconventional conditions such as with ice
cover, in a crawl space, or from a jack-up platform. All CPT measurements used a cone equipped with
a temperature sensor to confirm the subzero temperature of frozen soil. Stress relaxation tests were performed
in ice-rich permafrost soils to investigate long-term soil strength in both compression and shear. Sampling of
frozen soil using direct push techniques was achieved at two sites.
1 INTRODUCTION 2 PRACTICAL EXAMPLES OF SITES WITH

CPT IN FROZEN GROUND
Frozen soil has ice content and exhibits rheo
logical behavior, meaning the stress-strain Nine cases of TCPTU testing in the Russian Arctic
behaviour and mechanical properties change from 2014 to 2021 are listed below. The sites involve
with time. The failure of frozen soil under a range of permafrost conditions (Figure 1):
long-term loading is important for design. The
long-term soil strength (the resistance of a soil 1. 2014 – Labytnangy Civil Infrastructure
to failure in response to a long-term load 2. 2015 – Vorkuta Railroad
application) is a key parameter in the engineer 3. 2016 – Salekhard College
ing of frozen ground (Vyalov, 1986). Long 4. 2017 – Ob Gulf, Arctic LNG2
5. 2018 – Ob River, Salekhard Bridge
term soil strength can be considered in terms
of two main components: the long-term soil 6. 2019 – Novy Urengoy, Civil Infrastructure
strength in compression (σ c), important for pile 7. 2021 – Kruzenshtern Gas Field, Yamal
8. 2021 – Dikson Area, Kara Sea
end bearing (q p), and the long-term soil
9. 2021 – Norilsk river, Talnakh Bridge
strength in shear (σ f), important for pile unit
side friction (f p). The findings from this testing include:
Cone penetration testing (CPT) in frozen
ground provides valuable data on frozen condi • Frozen sand at -6°C was successfully tested with
tions and soil properties. Both temperature and CPT equipment;
porewater pressure sensors are required for test • Up to 62 m of continuous penetration through
ing in permafrost (a TCPTU test). Temperature permafrost was achieved without predrilling;
measurement is similar to pore pressure dissipa • Ground ice is generally not found to be critical
tion testing. Penetration is paused and the vari for cone penetration in permafrost;
ation of temperature with time is measured to • CPT refusal is generally caused by the soil dens
determine the asymptote. A Stress Relaxation ity rather than ice content or soil temperature.
Test (SRT) can be performed at the same time, Four of these cases (1-4) were presented at the
recording the variation of cone resistance (qc) CPT’18 conference (Volkov et al, 2018).
and sleeve friction (fs) with time, for the cone A description for the five sites (5-9) tested between
clamped in place. This provides an estimate for 2018 and 2021 is given below.
the long-term strength in compression (σc) and
shear (σf) (Sokolov, 2020). An SRT test does
not increase the time or cost, and data process 2.1 Ob river, Salekhard bridge
ing can be fully automated for results to be TCPTU tests were performed at the Salekhard
obtained in the field. Bridge site located in the Russian Arctic, some
DOI: 10.1201/9781003308829-109
741
bearing capacity and uplift capacity were estimated,
for locations characterized by both unfrozen and
frozen soil conditions. The results compared favour
ably with subsequent full scale pile tests at the site
for both static load capacity and pile drivability.
2.2 Novy Urengoy, Civil infrastructure

In the summer of 2019, TCPTU tests with direct
push sampling were performed from a crawl space
under a civil apartment complex in Novy Urengoy,
Western Siberia (Figure 1, site #6).
The structure, supported by piles driven to
10 m below ground level, was experiencing gradual
differential settlement due to unknown reasons. It
was assumed that settlement was due to insufficient
pile bearing capacity. However, the steps taken to
increase pile bearing capacity did not stop the settle
ment. CPT testing was then applied to investigate in
more detail the insitu soil conditions between and
below the piles. The depth of cone penetration was
Figure 1. Permafrost sites in Siberia, Russia with CPT 30 m, comfortably exceeding the pile length.
investigations (for numbers see description in the text). Frozen soil was detected at some locations at
a depth 11 m (Figure 2). Distinctive peaks of high
values in cone resistance and low values in sleeve
friction and electric conductivity were interpreted as
160 km upstream from the Ob River estuary ice lenses. Some of these ice lenses are highlighted
(Figure 1, site #5). Three zones at the site of the Ob in Figure 2, as an example.
River bridge are the left bank, the right bank, and the
river bed itself. The site is located in a cold region
characterized by discontinuous permafrost
(Figure 1). The average annual air temperature in the
Salekhard area is -5.7°C. The CPT tests were per
formed during winter January-March 2018 when the
Ob River was covered by ice. One CPT was per
formed to refusal on the right bank and achieved
a depth 62 m (with no predrilling). The frozen state
of the soil (permafrost) was confirmed by measuring
temperature and collecting samples of frozen ground
from the adjacent geotechnical borehole. The above
is considered to be a record depth for continuous
CPT penetration in permafrost.
An interesting observation was made from the
CPTs on the left bank. The river formed a talik at
this location (an unfrozen soil strata below or near to
a river). Due to this talik, part of the soil depth pene
trated was frozen and part of it not. This provided an
opportunity to compare the same soil horizon in both
a frozen and non-frozen state. Although it is gener
ally known that cone resistance in frozen ground is
higher compared with the same unfrozen ground, all
other conditions being equal, these tests provided
a direct quantitative comparison. The cone resistance
in the unfrozen sand varied between 8 and 35 MPa,
in the frozen sand between 12 and 54 MPa. The
refusal depth was similar both in frozen and non- Figure 2. CPT profile in frozen soil at Novy Urengoy.
frozen sand and varied between 25 and 30 m.
The CPT data were interpreted to provide soil
parameters and associated analysis. For pile design, A direct push probe was then used to collect sam
impact Soil Resistance to Driving (SRD) and Blow ples of the frozen soil, using a regular MOSTAP
Count analyses were completed and axial pile sampler with 35 mm inner diameter (Robertson,
742
2014). The collected samples confirmed the presence Based on these findings, engineering measures were
of frozen soil with ice lenses of thickness up to taken to minimize the heat flow from the building to
10 cm (Figure 3). The frozen soil was tested in the the ground, including thermal insulation of the base
laboratory and was found to have a high density that ment. A cooling system comprising two-phase thermo
would not compress significantly due to thawing syphons was installed to a depth 10 m to cut down any
(Table 1). Rather, the thickness of ice lenses were residual heat flow. The geodetic monitoring of the
such that significant differential settlement would be structure since has shown that the differential settle
expected from gradually melting ground ice. ment ceased and the structure has been stable since the
cooling system was put into operation.
2.3 Kruzenshtern gas field

Twelve TCPTU tests were performed in the period
January-February 2021 on the Sharapov Shar gulf of
the Kara Sea (Figure 1, site #7). Ten of the tests
reached the designed penetration depth in the range
44 m to 54 m. Early refusal of two other tests were
at depths 27 m and 35 m.
The depth of water in the gulf varies between 1
and 4 m. It was unknown if the permafrost would be
continuous or discontinuous at the site, and the field
investigation was to characterize the ground condi
tions and presence of permafrost. Only one test
encountered permafrost from 9 m depth with
a temperature between -0.35 and -1.58°C (Figure 4).
Several peaks of high values in cone resistance and
low values in sleeve friction were observed. They were
not as clear as the ones at the Novy Urengoy site. How
ever, they were interpreted as ice lenses and some of
them are indicated on Figure 4.
Figure 3. Frozen soil samples with ice lenses collected by

direct push sampler mounted on CPT press at Novy
Urengoy.
Table 1. Physical and strength properties of the tested

soil.
Value
Parameter Min max average
Depth, m 10.5 15.0 -

Soil temperature (T), °C -0.3 -0.11 -0.20 (8)*
Freezing temperature, °C -0.15 -0.06 -0.11 (8)
Particle density (ρs), g/cm3 2.60 2.69 2.66 (10)
Density (ρ), g/cm3 1.98 2.17 2.07 (10)
Dry density (ρd), g/cm3 1.60 1.89 1.75 (10)
Porosity (n), % 29 40 34 (9)
Void ratio (e) 0.42 0.67 0.52 (9)
Water content (W), % 15 24 18 (10)
Liquid limit (WL), % 20 28 24 (10)
Plastic limit (Wp), % 13 17 15 (10)
Plasticity index (Ip), % 7 11 9 (10)
Liquidity index (IL) 0.24 0.59 0.39 (10)
Figure 4. CPT profile in frozen soil at Sharapov Shar,
Cone resistance (qc), MPa 4.22 13.52 7.31 (10)
Yamal peninsula.
Long-term compression
strength (σc) MPa 0.08 1.89 0.91 (10)
To confirm the presence of permafrost, direct
* (8) – number of tests. push sampling was applied close to the CPT test
743
location. The sampler was a regular RANGER with The ice rich permafrost was confirmed by geotech
an inner diameter 45 mm. The collected samples nical borehole drilling. The volumetric ice content
confirmed the presence of frozen soil (Figure 5). The (Itot) was measured in soil laboratory. The volumetric
cryogenic structure is massive with rare ice layers. ice content is the ratio of the volume of ice in
a sample to the volume of the whole sample,
expressed as a fraction (Everdingen, 2005).
The measured temperature of the ice-rich perma
frost varied between -0.1 and -0.6 °C. The top layer of
permafrost (10 m to 18 m) comprised clay with
organic matter, a stratified cryostructure with high ice
content 0.40<Itot<0.60. When thawed, the water con
tent was Wtot=110% and density ρ=1.36 g/cm3. The
layer below (18m to 30 m) comprised clay with
organic matter, a stratified cryostructure and high ice
content 0.60<Itot<0.90 (Figure 7). When thawed, the
water content was Wtot=263% and density ρ=1.12
g/cm3.
Figure 5. Frozen soil samples with ice lenses collected by

direct push sampler mounted on a CPT press at Sharapov
Shar.
2.4 Dikson area, Kara sea

TCPTU tests were performed in the summer 2021
from a jackup platform nearshore in the area of
Dikson, Kara Sea (Figure 1, site #8). Permafrost was
not detected although in several locations the clay soil
was found to have a temperature in the range 0 to 1°C.
The clay was not frozen because of the salinity of the
pore water. Conventionally, the measurement of
ground temperature nearshore is problematic. To use
conventional thermistors, these must be installed in Figure 6. CPT profile in ice-rich permafrost near Norilsk.
a dry borehole and then monitored with time. In com
parison, the TCPTU test is not complex and has
a relatively low cost and required time for testing. In
addition to the temperature data, a TCPTU test pro
vides valuable data on the variability of temperature
and mechanical properties with depth.
2.5 Norilsk river, Talnakh bridge

Cone penetration tests with temperature measure
ments (TCPTU) were performed on the right bank of
the Norilsk river, located above the Arctic circle,
15 km from Talnakh city (Figure 1, site #9). The site
is located in a cold region characterized by continu
ous permafrost, with an average annual air tempera
ture in the Norilsk area of -9.6°C. However, the
testing revealed a talik at the site formed due to the
thermal impact of the Norilsk river.
The TCPTU tests detected ice rich permafrost at Figure 7. Ice-rich permafrost samples collected near
the site over a depth range 10 m to 30 m (Figure 6). Norilsk.
744
3 STRESS RELAXATION TEST relaxation period, a body behaves as a Hookean
solid. In the alternative case, it behaves like
3.1 Rheological behavior a Newtonian fluid.
The Relaxation period varies significantly for dif
Rheological behavior is inherent for frozen soils. ferent materials. For instance, limestone 1011 sec
The degree of rheological behavior depends on the (thousands of years), glass 101° sec (hundreds of
physical properties of the frozen soil (ice content, years), ice 102 sec (hundreds of seconds), water 10-11
water content, cryogenic structure, etc.). Rheological sec. In the case of water, for instance, when acted on
behavior is characterized by two interrelated param by a force lasting less than 10-11 sec it behaves elastic
eters: creep and long-term strength (Vyalov, 1986). ally. Rocks experience load over periods of geological
Long-term soil strength corresponds to a stress level time and may develop viscous flow, which can be
below which no failure takes place within observed in folding. Ice behaves like an elastic body,
a practically observable period of load application. failing in its brittle form if struck (or loaded) rapidly,
A stress in excess of the long-term strength results in a force applied for less than 102 sec (Vyalov, 1986).
failure after a certain time. In this paper we consider A long-term load causes ice to flow as a viscous
“practically observable period of load application” material, as occurs in glaciers. Similar behavior, i.e.
as 100 years. If the time period is different, for brittle failure under a rapid load application and vis
instance, 50 years, the term is described as the cous flow due to a long-term load, can be observed in
“long-term strength for 50 years”. frozen soils.
There are two approaches to evaluate long-term It is important that the relaxation period does not
strength (Figure 8). The most common is to measure equal the period of after-effect. The process of relax
deformation under an applied constant load to ation occurs much faster than the process of creep,
describe creep, i.e. deformation change with time. i.e. relaxation period significantly shorter than period
The second approach measures the stress caused by of after-effect, which differs by several orders (10n).
a constant applied deformation to describe stress This fact provides a key advantage for the second
relaxation, i.e. stress change with time. The second approach to evaluate the long-term strength which is
approach is considered in this paper. based on stress relaxation measurements versus the
first approach based on creep measurements.
3.2 Stress relaxation test procedure (SRT)

Relaxation test approach for soils was first proposed
by Vyalov in 1986. It was called an “accelerated
method of testing soils for long-term strength, using
a dynamometric apparatus”. The method description
said “if an initial stress is given which is close to the
hypothetically instantaneous strength (as determined
in advance), the finite value of stress will approach
the ultimate long-term strength”. The stress should
be measured in time.
The stress relaxation curve should then be pro
cessed by a fitting method to determine the empirical
coefficients for the following equation:
Figure 8. Correlation between Creep and Relaxation in the

Rheological testings.
where σc = long-term soil strength; tp = measured

time; and β, T are empirical coefficients.
Stress relaxation is the decrease of stress with Once the empirical coefficients are determined the
time in response to constant deformation applied to long-term strength for a given period of time, for
the soil. Relaxation of stress results from instance 100 years, may be calculated.
a redistribution of elastic and plastic deformation. The proposed laboratory method was not practically
All physical bodies possess both elastic and viscous useful at the time due to the low quality of measure
properties. However, the material behavior depends ment equipment and impossibility to determine the
on the correlation between the period of load appli instantaneous strength in advance. Because of these
cation (i.e. the observation period) and the relaxation limitations, a pure relaxation test on frozen soils is not
period. If the observation period is shorter than the common practice.
745
However, the accelerated method approach can be 4 CONCLUSIONS
applied for the conditions when performing in-situ
a CPT temperature dissipation test (or pore pressure Practical application of CPT in frozen ground has
dissipation test). The only requirement is that the dissi been illustrated for nine different sites with various
pation is performed with the rods clamped (to ensure degrees of permafrost. The CPT is able to penetrate
not movement). In such a case, the deformation is frozen soils, including frozen sands, ice-rich soils
kept constant and the stress relaxation can be meas and ice lenses. This application of CPT testing could
ured by the cone (qc) and sleeve friction (fs) sensors. be significantly enlarged for the current geotechnical
The Stress Relaxation Test (SRT) is described in activities in the Arctic.
detail by Sokolov (2020). A key step in the process Cone penetration testing provides a lot of data
ing was to apply the fitting curve method not for the which can be obtained in one push performed within
entire curve, but for the curve starting at some point. one working day. Perhaps most significant is that
This is the point that separates two parts of the a CPT used with a temperature sensor is capable to
curve: the first part corresponds to the relaxation- detect frozen soils by measuring the soil temperature
creep stage while the second part to the relaxation directly. The pore water pressure sensor provides
stage. Once this separating point is determined it is very useful complementary data. For reference, this
possible to back calculate the empirical coefficients test has been called by the acronym TCPTU.
β and T for equation (1) and thereby calculate the In addition, by using a Stress Relaxation Test
long-term strength in compression and shear. (SRT) it is possible to estimate the long-term soil
strength in both compression (σc) and shear (σs). The
long-term strength results on one of the sites were
3.3 Results
confirmed by comparison with a static pile load test.
SRT results from site #5 (Figure 1) were verified by The results on σc and σs derived for other sites also
comparing with the results of a static pile load test show consistency with the recommended values for
(Volkov, 2019, Sokolov, 2020). The common practice pile unit end bearing and pile unit side friction.
calculates the pile unit end bearing, qp, from the cal Ice content in frozen soil plays a major role and
culated equivalent average cone resistance, qca, multi influences strongly key parameters such as long-term
plied by an end bearing coefficient, kc (Robertson, strength. TCPTU testing shows great applicability
2014). This approach did not work well for frozen for ice-rich frozen soils both to detect ice and evalu
soils because ice contributes significantly to the cone ate long-term soil strength in compression and shear.
penetration resistance, but little to the long-term soil Push sampling methods using the CPT equipment
strength. Thus in case of ice rich permafrost, the end can recover samples for laboratory testing.
bearing coefficient, kc, may be very low compared to
common values for non-frozen soils. For instance,
based on the numbers provided in Table 2, kc (which REFERENCES
is correlated to the relation between σc/qc), varies
from 0.017 to 0.086 and is quite low compared to kc Everdingen, R.O. 2005. Multi-Language Glossary of
Permafrost and Related Ground-Ice Terms: In Chinese,
for non-frozen soils 0.2 to 0.5 (Robertson, 2014). English, French, German, Icelandic, Italian, Norwegian,
Polish, Romanian, Russian, Spanish, and Swedish, The
Arctic Institute of North America, 1998 (revised
Table 2. CPT measured and SRT evaluated results. 2005) – 159 pages.
Robertson, P.K., and Cabal, K.L. 2014. Guide to Cone
qc σc fs σs Penetration Testing for Geotechnical Engineering, 6th
Depth, T, Edition, Signal Hill, California: Gregg Drilling & Test
m °C MPa MPa kPa kPa ing, Inc.
Sokolov I. 2020. Determination method for strength prop
17.4 -0.41 8.96 - 153 67 erties of frozen soils by cone penetration testing. PhD
20.6 -0.53 15.34 0.48 115 51 Thesis – Moscow State University, Moscow, 2020 – 149
23.3 -0.49 13.65 0.48 209 23 pages (in Russian).
27.6 -0.21 22.06 0.43 146 15 Volkov, N., Sokolov, I. & Jewell, R. 2018. CPT Testing in
31.7 -0.09 1.28 0.42 75 15 Permafrost. Proceedings 4th International Symposium
32.0 +0.09 1.12 - 48 on Cone Penetration Testing/N. Volkov [and etc.] // –
14.8 -0.27 15.28 0.48 114 29 CPT’18. – Netherlands, Delft – 2018. 1258–1268.
Volkov N.G., Sokolov I.S., 2019. Estimation of pile bearing
17.7 -0.30 7.08 0.60 187
capacity in permafrost based on stress relaxation meas
20.6 -0.21 18.27 0.49 255 16 ured by cone penetration testing. Geotechnics, Vol. XI,
23.5 -0.15 27.95 0.73 231 12 No. 1, pp. 68–78, http://dx.doi.org/10.25296/2221-5514
26.4 -0.11 31.92 0.55 294 7 2019-11-1-68-78.
30.0 -0.10 1.68 - 70 39 Vyalov, S.S. 1986. Rheological Fundamentals of Soil
30.7 +0.13 2.05 - 60 Mechanics, Volume 36, 1st Edition. Elsevier. ISBN:
0444600566. 564.
746
© 2022 the Author(s), ISBN 978-1-032-31259-0
A site-specific relationship between CPT data and fines content for fine
grained soil in the context of liquefaction analyses
C. Vrettos
Technical University of Kaiserslautern, Germany
ABSTRACT: Fines content is a crucial factor affecting the liquefaction resistance of soils within the frame
of the CPT-based cyclic stress ratio approach. In contrast to highly susceptible sands and silty sands, on one
hand, and non-liquefiable clays, on the other hand, the proper identification and classification of silts is still an
unresolved issue. Several empirical equations have been proposed to correlate fines content with common
CPT soil behavior indices. In the frame of a site-specific study comprising pairs of CPT soundings and adja
cent exploration borings, gradation characteristics of retrieved soil samples are combined with the respective
values of common soil behavior indices inferred from the CPT logs to examine available predictive equations
and calibrate a new hyperbolic equation. The investigation confirms the large scatter in the data and the inher
ent difficulty in establishing a robust correlation.
1 INTRODUCTION While in the early years the focus was placed on

FC values up to 35%, evidence from seismic events
Over the last years, CPT has advanced to the pre and research in the last two decades revealed that fine
ferred method to assess liquefaction susceptibility in soils may also show a degradation under high ampli
code-based design using the cyclic stress ratio tude cyclic loading. It is now distinguished between
approach. This holds in particular for countries out two types of ultimate limit states: i) sand-like (cyclic
side the USA and Japan, where the Standard Pene liquefaction), and ii) clay like (cyclic softening)
tration Test (SPT) does not constitute the primary behavior, as elucidated among others in the recent
means of site investigation via soundings. Several review report by the committee of the National Acad
codes of practice include CPT-based methods as an emies of Sciences, Engineering, and Medicine (2016).
equivalent counterpart to SPT-based methods. An In this context, fine-grained soils may be classified
example is the current draft of the new edition of the based on some measure of their consistency in terms
Eurocode EN 1998-5, CEN (2021), that adopted the of the water content and the Atterberg limits.
methodology by Idriss & Boulanger (2008) and its A plasticity index IP = 15% is considered as
CPT-based update by Boulanger & Idriss (2016). a conservative threshold beyond which the occurrence
In this semi-empirical method, as established in of liquefaction can be excluded. Such a classification
the ’80s by Seed & Idriss (1983), soils other than requires collection of several samples, laboratory test
sands were characterized with regard to liquefaction ing and reliable values of the in-situ water content.
susceptibility by the fines content FC measured in the Even then, this information will be available only for
laboratory. FC is defined according to ASTM D 653 specific points along the soil depth profile.
20 as the portion of soil particles finer than 0.075 mm An evaluation based on recorded data from CPT
(sieve no. 200). In EN ISO 17892-4:2016 the limit is soundings linked with a reliable general equation for
set at the 0.063 mm sieve. An adjustment for the FC constitutes an attractive, widely used alternative.
fines content is made by increasing the measured tip Candidates for the independent variable in such
cone resistance by an increment dependent on FC. a relationship are the well-established soil behavior
Alternatives to consider the soil type in the cyclic type indices, as summarized in the next section.
stress ratio method were proposed by Robertson & Agaiby & Mayne (2020) compiled several of these
Wride (1998) via a soil behavior index derived from equations. The scatter in the data is considerable, in
CPT data, and by Moss et al. (2006) based on particular for silts, which constitute the crucial soil
a combination of CPT tip and sleeve resistance by category in liquefaction hazard assessment.
skipping the intermediate step. Despite its deficien The necessity to derive a site-specific predictive
cies, the fines content concept presently dominates equation for FC in the frame of a critical project
practice and code-based design. motivated the present study. The field investigations
DOI: 10.1201/9781003308829-110
747
performed, the main results of the accompanying An alternative for the soil behavior index is the
laboratory testing program, the essential analysis modified index IB introduced by Schneider et al.
steps, and the conclusions drawn are described next. (2012) and subsequently adopted by Robertson
Similar studies with a much larger spatial extent (2016) in his charts to characterize soils:
have been conducted for soils in Canterbury, New
Zealand, among others by Lees at al. (2015).
2 EQUATIONS FOR THE FINES CONTENT

Various equations relating Ic to FC have been sug
The vast majority of the suggested equations relate gested in the last decades. Some of them are con
the fines content FC to the soil behavior index Ic pro sidered herein in detail. One of the first has been
posed by Robertson & Wride (1998) to capture six proposed by Robertson & Wride (1998):
of the designated zones of the SBTn-chart covering
the basic soil types encountered in practice, cf.
Table 1. Its current form as proposed by Robertson
(2009) is:
0 with the friction ratio
where
Boulanger & Idriss (2016) suggested:
where CFC is a fitting parameter with a default value

where Fr = normalized friction ratio; fs = sleeve fric equal to 0 and a standard deviation ± 0.29.
tion; Qtn = normalized cone resistance; qt = tip resist The above relationships have been examined in
ance corrected for the effects of pore water the frame of the extensive regional liquefaction sus
pressures; n = stress normalization exponent; σv0 = ceptibility study on Christchurch soils by Maurer
in-situ vertical stress; σ0v0 = in-situ vertical effective et al. (2019). Assuming a linear relationship in ana
stress; pa = atmospheric pressure. Except for very logy to equation (8), regression statistics yielded for
soft fine grained soil, it may be assumed that qt the mean estimate:
equals the uncorrected cone resistance qc (Robert
son, 2010).
Table 1. Soil Behavior Type based on Ic. with approximately 68% of the samples having FC
within ±16.56% of the mean prediction. For these
Zone Soil Behavior Type Ic region-specific conditions and FC > 10%, equation
(6) by Robertson & Wride (1998) yields much lower
2 Organic Soils – Clays >3.6 values, i.e. it is too conservative. The many decimal
3 Clay – silt clay to clay 2.95-3.6 digits in the constants of equation (9) suggest an
4 Silt mixtures – clayey silt to silt clay 2.60-2.95 accuracy that is not attainable.
5 Sand mixtures – silty sand to sandy silt 2.05-2.60 A large database from different studies has been
6 Sands – clean sand to silty sand 1.31-2.05 evaluated by Agaiby & Mayne (2020). The derived
7 Gravely sand to dense sand <1.31 relationship reads:
A threshold Ic = 2.6 is often used to distinguish

between clay-like and sand-like soils that are suscep An equation has also been obtained in terms of the
tible to liquefaction (Boulanger & Idriss, 2014). modified index IB:
748
validate back-analysis procedures for the CPT, the
information on the sampling depth of the undisturbed
samples tested in the laboratory was not precise
The observed large scatter in the data in all studies enough to allow an exact mapping to the CPT. The
lies in the nature of Ic, which was introduced to char accuracy was improved by incorporating the antici
acterize soils and not to be associated with the sus pated soil behavior type SBT/SBTn in the selection of
ceptibility to liquefaction. A linear or even power the appropriate depth range for the calculation of the
law relationship for FC vs. Ic is too simplistic if one FC vs. Ic relationship. Hence, the identification of fines
considers the complexity of the physics of the prob content is more accurate for the undisturbed samples.
lem, the variability of the soil composition even over Due to the nature of the glacial till sediments,
short distances, the sensitivity of the CPT response grading for some of the samples spans over silt, sand
as reflected in the strong fluctuation of the recorded and gravel with a large uniformity coefficient. In this
data, as well as the uncertainty in capturing the case and for liquefaction hazard assessment, the
effects of soil plasticity and grain size distribution adverse influence of low fines content is counterbal
on the CPT response. anced by the favorable effects of the gravel fraction
that provides ample drainage. The SBTn charts pres
ently do not include this soil type, and an interpret
3 SITE-SPECIFIC CORRELATIONS ation solely according to Ic may be misleading.
Moreover, the fill is also characterized by an
The investigated site is located in Germany and is inhomogeneous composition.
characterized by a variable stratigraphy. The top
layer with a thickness of approximately 5 m consists
of fillings of sand, silt and partially gravel. This is
followed by alternating layers of silt, silty sand and
gravel, as well as sand-silt mixtures. These soils are
underlain by glacial till deposits: debris loam com
posed of silts and silty sands, and marly till (clays).
Below this is sand with different fine-grain and
gravel proportions. The encountered sediments are
highly variable and hence particularly suited for
checking the accuracy of the various predictive
equations relating FC to Ic.
Field investigations comprised exploration bor
ings at several locations and CPT soundings adjacent
to the boreholes. Disturbed and undisturbed samples
have been extracted for further testing.
The laboratory investigations included tests for
particle size distribution, Atterberg limits, consist
ency, compressibility and shear strength.
Soil classification was conducted according to EN
ISO 14688-2 in conjunction with DIN 18196. It is
very similar to the USCS except for some differ
ences in coding and in including an additional class
of intermediate-plasticity for silts and clays.
The primary purpose of the investigation, and
accordingly the selection of the sampling depths,
was foundation design. Detailed liquefaction assess
ment was conducted at a later stage, since the region
is characterized by low to medium seismicity.
A total of 14 CPT records were evaluated. The tests
reached a depth of 20 m. Essential for the accuracy of
the prediction is an appropriate geodetic survey in
order to assign each sample from the boring to the cor Figure 1. Results at location #13: Borehole log, position
rect depth along the CPT log. At each CPT location, and length of extracted soil samples, CPT results, and
verification checks have been performed at positions derived Ic with colors and numbers in brackets indicating
where an unambiguous identification of the soil type the various soil types in the SBT and SBTn charts.
was possible. Spikes in the CPT records were carefully
inspected, and when an abrupt change in soil type was
not justified, an averaging procedure over a few centi Typical results and the accompanying interpret
meters centered at the location of the sample was ation of the field investigations at one of the loca
applied. As the investigation was not planned to tions are displayed in Figure 1. The delineation, the
749
Table 2. Overview of soil properties and derived soil behavior indices.
Depth Cl/Si/Sa/Gr FC Ic IB wL IP
Loc. [m] [%] [%] Mean CV Mean CV [%]
1 U 11.0 - 11.25 30.1/66.9/3.1/0.0 97.0 2.87 1.7 22.5 4.3 47.5 28.3
18.8 - 20.0 0.0/0.7/98.7/0.6 0.7 1.79 2.9 77.0 7.9
2 U 12.5 - 12.75 25.0/71.5/3.5/0.0 96.5 2.87 1.5 22.4 4.7 59.4 35.1
3 3.5 - 4.0 6.8/17.8/66.3/9.1 24.6 2.32 7.2 39.3 30.2 17.7 5.4
U 12.0 - 12.25 17.3/63.0/18.5/1.1 80.3 3.00 1.4 20.6 3.2 36.2 16.9
4 2.6 - 3.0 15.4/76.2/8.4/0.0 91.6 2.77 4.4 23.8 13.7 32.7 13.5
4.3 - 5.1 0.0/1.6/93.4/5.0 1.6 1.42 5.3 118.1 10.8
5 U 14.0 - 14.25 43.5/56.1/0.4/0.0 99.6 2.91 1.9 21.9 6.2 57.4 36.4
6 11.0 - 11.25 26.0/72.9/1.2/0.0 98.9 2.94 1.3 21.2 3.6 48.9 27.6
7 U 3.0 - 3.25 21.2/44.2/30.2/4.3 65.4 2.69 2.1 24.6 13.2 39.8 23.3
U 9.5 - 9.75 18.7/72.7 /8.6/ 0.0 91.4 2.66 6.6 26.4 20.4 36.5 19.6
8 4.8 - 5.0 0.0/2.0/87.2/10.9 2.0 1.85 1.2 70.3 2.5
U 10.0 - 10.25 5.2/40.9/53.9/0.0 46.1 2.32 7.7 39.4 23.4
U 13.0 - 13.25 16.2/62.9/20.9/0.0 79.1 2.92 1.1 21.2 3.6 47.7 26.5
18.0 - 19.0 0.0/4.9/94.8/0.3 4.9 1.82 5.0 77.0 12.8
9 U 10.0 - 10.25 11.0/51.2/37.8/0.0 62.2 2.83 4.0 24.0 9.8 24.8 9.2
10 U 3.0 - 3.25 11.7/62.6/23.6/2.1 74.3 2.82 5.9 24.0 20.7 29.5 14.5
4.5 - 4.8 16.9/48.4/32.8/1.9 65.3 2.97 1.3 18.8 6.4
11 U 5.0 - 5.25 22.6/66.5/10.6/0.3 89.1 3.12 2.0 18.8 5.9 67.5 39.0
12 4.4 - 5.0 13.5/73.7/12.8/0.0 87.2 3.17 5.3 17.0 15.5 62.0 30.1
U 8.0 - 8.25 17.6/74.2/8.1/0.0 91.8 2.93 2.9 20.8 8.9 37.1 19.2
13 4.4 - 4.9 13.3/61.9/24.1/0.8 75.2 3.05 5.0 16.6 16.0 56.0 23.0
5.0 - 6.0 9.8/73.7/16.5/0.0 83.5 2.79 6.8 23.0 19.0 25.7 8.3
U 10.0 - 10.25 5.9/71.7/22.4/0.0 77.6 2.78 3.9 24.2 11.2 28.8 7.6
14 U 8.0 - 8.25 8.6/80.1/11.4/0.0 88.7 2.89 2.9 22.4 8.3 28.9 9.0
coloring, and the material symbols in the borehole fraction varies between 41 and 80%, the clay frac
logs correspond to the field description of the soil. tion between 5 and 43.5%.
The CPT log is given in terms of tip cone resistance In order to establish a relationship between Ic and
qc and friction ratio Rf. The soil type as inferred FC, a single representative value Ic over the nominal
from the non-normalized SBT-chart by Robertson depth range of the distinct disturbed/undisturbed sam
(2010) is indicated by different colors in the column ples extracted from the borings is required. The fluc
for Rf. The last column shows the Ic profile as calcu tuation of Ic over the sampling depth is considerable,
lated from equations (1) to (4), and also the SBTn and the following method is used for an automated
soil types in color. SBT- classification is the standard evaluation. Samples that show for Ic over the respect
output of the software used, and is considered as ive depth range a coefficient of variation larger than
a good approximation of the more accurate, normal a threshold are dismissed. Comparison of the predic
ized SBTn-chart for vertical effective overburden tion with the actual soil gradation measured in the
stresses up to 150 kPa (Robertson, 2010; Papami laboratory yielded herein a threshold of 8.0. In par
chael & Vrettos, 2018). ticular, undisturbed samples extracted from the fill
Data for all samples are given in Table 2 and did not meet this criterion. If more data are available,
include: sounding number (location); sampling the threshold may be set even lower.
depth; indication (U) for an undisturbed sample; The inherent limitations of the attainable accuracy
material composition in terms of clay (Cl), silt (Si), become evident if one looks at the results for the
sand (Sa) and gravel (Gr) content from the labora undisturbed sample taken from the depth of 10.0 to
tory tests; mean value and coefficient of variation 10.25 m at location #13, cf. Figure 1. While the iden
(CV) of the soil behavior index Ic and the modified tification of the soil type is correct, i.e. zone 4 in the
index IB (CV is the ratio of the standard deviation to SBTn chart matches well the laboratory gradation
the mean); liquid limit wL and plasticity index IP. with 71.7% silt, the derived value Ic ranges between
The ground water table was encountered in depths 2.57 and 2.92, which constitutes a significant vari
between 5 and 7 m. For the fine-grained soil, the silt ation if a reliable prediction of FC is sought.
750
The plasticity chart of EN ISO 14688-2 with all
samples of fine-grained soil is presented in Figure 2.
It can be seen that almost all samples classified as
silts according to the grading are actually clays with
respect to their soil mechanical behavior, as they
plot above the A-line in the plasticity chart.
Since both FC and IP are independent of the
actual soil state in terms of consistency, it is of inter
est to plot the FC versus the plasticity index IP, as
shown in Figure 3. For FC > 60% there is no correl
ation between FC and IP, at least for the soils
encountered at the particular site.
Figure 4 Depicts the core of the investigation, i.e.
data points of FC versus Ic for the samples fulfilling
the coefficient of variation criterion. Figure 5 shows
the corresponding data set FC versus IB. Figure 4 Figure 4. Fines content FC vs. soil behavior index Ic: data
includes the prediction by equations (6), (8) and set, and comparison with the various predictive equations.
(10), and Figure 5 that by equation (11).
Figure 5. Fines content FC vs. modified index IB: data set,

Figure 2. Sampled soil classified in the plasticity chart of and comparison with the predictive equations.
EN ISO 14688-2.
behavior) though, equation (6) matches the data

well. As for the IB-based relationship in Figure 5,
equation (11) is a good approximation.
With regard to the Ic-based estimation of FC, it
turns out that none of the three forecast equations
considered is suitable for the particular site. Equation
(6) is applicable for low Ic and FC values but too
conservative for higher FC values. Equation (10), on
the other hand, yields higher values for low FC
sands (slightly non-conservative), and realistic pre
dictions for medium and high FC values. In order to
increase accuracy and avoid the unboundedness of
the power-law approximations, the following hyper
bolic equation that exhibits a more versatile,
Figure 3. Fines content FC vs. plasticity index IP. “S-like” form is proposed:
It can be deduced from Figure 4 that for high

values of Ic, i.e. for clay-like behavior, equation (6)
predicts consistently lower values than actually
measured, while equations (8) and (10) are closer to The engineering approximation of the data set in
the laboratory results. For low Ic values (sand-like Figure 4 yields:
751
a ¼ 120 ; b ¼ 2:62 ; c ¼ 8 (9): 1165–1177.doi:10.1061/(ASCE)1090-0241(2006)
132:9(1165)
CEN - European Committee for Standardization. 2021.
Similarly, the following counterpart is derived for prEN 1998-5:2021(E). Eurocode 8 - Design of struc
the data in Figure 5: tures for earthquake resistance – Part 5: Geotechnical
aspects. foundations. retaining and underground struc
tures, Draft. September 2021.
CEN - European Committee for Standardization. 2018. EN
ISO 14688–2:2018 Geotechnical investigation and test
ing – Identification and classification of soil – Part 2:
Principles for a classification.
with Idriss, I.M. & Boulanger, R.W. 2008. Soil Liquefaction
During Earthquakes. Monograph MNO-12. Oakland:
Earthquake Engineering Research Institute.
Lees, J., van Ballegooy, S. & Wentz, F.J. 2015. Liquefac
Equations (12) and (13) focus on silts, which consti tion susceptibility and fines content correlations of the
tute the most critical soil type with the strongest Christchurch soils. In Proceedings of the 6th Inter
national Conference on Earthquake Geotechnical
variability in the composition. An accurate predic Engineering, Christchurch, Paper no. 491.
tion is necessary in order to avoid i) the risk of non Maurer, B.W., Green, R.A., van Ballegooy, S. &
conservative design due to an overestimate of fines Wotherspoon, L. 2019. Development of region-specific
content, and ii) non-economical design due to a false soil behavior type index correlations for evaluating
alert by too low FC estimates. The identification of liquefaction hazard in Christchurch, New Zealand. Soil
liquefaction susceptible sands and silty sands of low Dynamics and Earthquake Engineering 117: 96–105.
fines content and low Ic, on the one hand, and of doi:10.1016/j.soildyn.2018.04.059
non-liquefiable clayey silts and clays of high Ic, on Moss, R.E.S., Seed, R.B., Kayen, R.E., Stewart, J.P., Der
the other hand, is in general unambiguous. Kiureghian, A. & Cetin, K.O. 2006. CPT-based prob
abilistic and deterministic assessment of in situ seismic
soil liquefaction potential. Journal of Geotechnical and
4 CONCLUSIONS Geoenvironmental Engineering 132(8): 1032–1051.
doi:10.1061/(ASCE)1090-0241(2006)132:8(1032).
CPT soundings are nowadays an integral part of National Academies of Sciences, Engineering, and Medi
cine. 2016. State of the Art and Practice in the Assess
modern, high-quality field investigations in earth- ment of Earthquake-Induced Soil Liquefaction and Its
quake-prone areas. The present study exemplarily Consequences. Washington, DC: The National Acad
shows the application of a state-of-the-art procedure emies Press. doi: 1017226/23474.
to estimate the fines content of fine soils for subse Papamichael. S. & Vrettos. C. 2018. CPT interpretation and
quent use in liquefaction hazard analyses of correlations to SPT for near-shore marine Mediterranean
a particular site with a strong variability in the sub soils. In M.A. Hicks, F. Pisanò & J. Peuchen (eds),
soil composition. For this type of problem, silts still Cone Penetration Testing 2018: 499–504. London: CRC
constitute the type of soil with the most uncertain Press.
behavior, and gradation is surely not the most appro Robertson, P.K. 2016. Cone penetration test (CPT)-based
soil behaviour type (SBT) classification system – an
priate characteristic to capture it. Targeted field update. Canadian Geotechnical Journal 53: 1910–1927.
investigations on silts, complemented by laboratory doi:10.1139/cgj-2016-0044.
cyclic loading tests, are urgently needed to clarify Robertson. P.K. 2010. Soil behaviour type from the CPT:
the related issues. an update. In P.K. Robertson & P.W. Mayne (eds), Proc.
2nd International Symposium on Cone Penetration Test
ing, CPT’10, Huntington Beach: 575-582.
REFERENCES Robertson, P.K. 2009. Interpretation of cone penetration
Agaiby, S.S. & Mayne, P.W. 2020. Indirect estimation of nal 46(11): 1337–1355. doi:10.1139/T09-065.
fines content using the Modified CPT Material Index. In Robertson, P.K. & Wride, C.E. 1998. Evaluating cyclic
J.P. Hambleton, R. Makhnenko & A.S. Budge (eds) liquefaction potential using the cone penetration test.
Geo-Congress 2020: Modeling, Geomaterials, and Site Canadian Geotechnical Journal 35(1): 151–158.
Characterization (GSP 317): 569–582. Reston, VA: doi:10.1139/t98-017.
ASCE. Schneider, J.A., Hotstream, J.N., Mayne, P.W. &
Boulanger, R.W. & Idriss, I.M. 2016. CPT-based liquefac Randolph, M.F. 2012. Comparing the CPTu Q-F and Q
tion triggering procedure. Journal of Geotechnical and Δu2/σvo’ soil classification charts. Géotechnique Letters
Geoenvironmental Engineering 142(2): 04015065. 2(4): 209–215. doi:10.1680/geolett.12.00044.
doi:10.1061/(ASCE)GT.1943-5606.0001388 Seed, Η.B. & Idriss, Ι.M. 1983. Ground Motions and Soil
Bray, J.D. & Sancio, R.B. 2006. Assessment of the lique Liquefaction during Earthquakes. EERI Monograph
faction susceptibility of fine grained soils. Journal of Series MNO-5. Oakland: Earthquake Engineering
Geotechnical and Geoenvironmental Engineering 132 Research Institute.
752
© 2022 the Author(s), ISBN 978-1-032-31259-0
Numerical modelling of cone penetration tests in spatially variable clays
Ze Zhou Wang, Siang Huat Goh & Xiangxiang Zheng

ABSTRACT: In soft clays, a cone factor Nkt is typically adopted to correlate net cone tip resistance qt with
undrained shear strength Su. Many existing numerical studies on the estimation of Nkt are performed using ana
lyses in homogeneous soil medium. However, soil properties exhibit spatial variability under real field condi
tions. In this study, the effects of spatial variability on the interpretation of CPT data were investigated using
a finite element analysis technique, namely the Press-Replace method (PRM). This method, which can capture
installation effects arising from the penetration of a piezocone, was implemented with the soil domain modelled
as anisotropic random fields. Using a pre-determined Nkt, statistics of Su were back-calculated using the qt pro
files obtained from different random field realizations. Comparisons of the back-calculated statistics with the
true statistics used to generate the random fields revealed that the true statistics may not be fully recovered from
the CPT data.
1 INTRODUCTION Arbitrary Lagrangian Eulerian (ALE) (Lu et al.

2004), the Remeshing Interpolation Technique
In soft clays, the cone penetration test (CPT) is typ with Small-strain (RITSS) (Zhang et al. 2020)
ically adopted to characterize the undrained shear and the Material Point Method (MPM) (Tehrani
strength through a cone factor Nkt as follows: et al. 2016), have been proposed.
In the literature, numerical simulations of CPT
considering both the installation effects and spatial
variability have not been widely reported. This
paper fused random fields and a simplified numer
ical procedure, the Press-Replace Method (PRM),
where σv0 is the total overburden stress, qc is the for simulating piezocone penetration in spatially
cone tip resistance corrected for cone net area ratio, variable soils. The PRM is a simplified technique
and qt is the net cone tip resistance. Although the that can be implemented in standard small-strain
value of Nkt is often obtained using laboratory data finite-element programs without the need for add
and site-specific calibration exercises, extensive ana itional subroutines. In addition, anisotropic random
lytical and numerical studies on the estimation of fields were adopted to realistically express the spa
cone factor values have also been reported in the lit tial variability of soil property. With the consider
erature (Yu 2000; Lu et al. 2004). ation of both features, the impact of spatial
Many studies are performed based on analyses variability on the interpretation of CPT data can be
that assume homogeneity in soil properties. How rigorously investigated.
ever, it is well known that the properties of natural
soils exhibit random variability in both spatial distri
butions and intensities under real field conditions 2 METHODOLOGY
(Phoon & Kulhawy 1999). Many studies reported
that the performance of geotechnical systems can be The Press-Replace Method (PRM) is a simplified
significantly affected by the presence of spatial vari technique to simulate problems associated with
ability (Pan et al. 2018; Zhang et al. 2021). continuous penetration of an object into
Furthermore, the large deformation caused by a continuum. The PRM was first used for simulat
the continuous insertion of a piezocone renders ing the load-controlled penetration of a suction
conventional small-strainnumerical techniques anchor in clay (Andersen et al. 2004) and was fur
ineffective, thus adding to the challenges in the ther used to simulate pile, cone and spudcan pene
numerical analysis of the cone penetration test. tration (Engin et al. 2015; Tehrani et al. 2016;
Therefore, special procedures, such as the Wang & Goh 2018).
DOI: 10.1201/9781003308829-111
753
Figure 1. Illustration of the PRM technique for modelling
cone penetration test.
As its name suggests, the PRM consists of two Figure 2. Model setup and relevant modelling details.
modelling phases. Figure 1 illustrates the procedures
for modelling cone penetration using the method.
Essentially, the entire penetration process is separated Table 1. Key parameter values used in the present study.
into many pairs of press-replace phases. Figure 1a
illustrates the model setup at the start of an arbitrary Value
stage i. In the next step (Figure 1b), a displacement
Parameter Homogeneous Heterogeneous
boundary condition is prescribed to simulate the
penetration. The prescribed displacement, however, Cone diameter, D 60 mm 60 mm
corresponds to only a portion of the total penetration Cone tip-apex 60° 60°
depth. As recommended by Engin et al. (2015), the angle, α
prescribed displacement corresponds to 1/10 of the Model width, w 0.4 m 0.4 m
diameter or width of the penetration object. After the Model depth, d 1.0 m 1.0 m
“press” phase, the “replace” action is performed by Unit weight, γsat 18 kN/m3 18 kN/m3
switching the soil body that was displaced by the Earth pressure 1.0 1.0
piezocone in the preceding “press” phase to the cone coefficient, K0
material (Figure 1c), which marks the end of stage i. Poisson’s ratio, ν 0.5 0.5
The model setup shown in Figure 1c then forms the Mean undrained 30 kPa 30 kPa
initial condition for the next pair of press-replace shear strength, Su
phase (Figure 1d). These phases are repeated until Coefficient of vari - 0.3
the desired penetration depth is achieved. ance, CoV
Horizontal scale of - 5.0 m
fluctuation, δh
3 MODEL SETUP Vertical scale of - 0.5 m
fluctuation, δv
Young’s modulus, 90Su; 300Su; 450Su; 300Su
The finite-element software Optum G2 was used in E 600Su; 900Su kPa
this study. In Figure 2, the model consists of a soil
domain of 0.4 m in width and 1 m in depth. An axi
symmetric piezocone of 36 mm diameter (D) and The soil was modelled as a single-medium mater
60° tip angle was modelled. The penetration started ial using the Tresca model. Table 1 summarizes the
from a pre-embedment depth of 0.03 m, and a key parameter values used in the present study. An
0.0036 m step size (Engin et al. 2015) was adopted. Su of 30 kPa was adopted in the analyses that
Interface elements were used to model the inter assumed a homogeneous soil medium, with the
actions between the piezocone and soil. Three types Young’s modulus E linearly correlated to Su. Five
of interfaces were adopted. First, the shaft-soil inter separate analyses using five E/Su correlations were
face was modelled as a fully smooth contact carried out, and the results are compared with pub
(Figure 2c) while fully smooth and fully rough con lished results to validate the PRM method. In the
tacts were considered for the cone-soil interface RFEM analyses, a lognormal distribution was used
(Figure 2b). Interface extensions (Figure 2b) that are to describe the statistical variability of both Su and
0.0036 m, i.e. 0.1D, in length (Engin et al. 2015), E. The correlation E = 300Su, which is representative
and having the same properties of the neighboring of soft clays, was adopted for the RFEM analyses.
soil, were used to minimize stress oscillations Other statistics, such as CoV, δh and δv , adopted in
around sharp corners of the cone. The model was the present study are also representative of natural
discretized using 6490 15-node triangular elements. soft clays (Phoon & Kulhawy 1999).
754
Figure 3. Comparison of cone factor values with Lim et al. Figure 5. Comparison of cone factor values with other
(2018) (I r ¼ 150; K0 ¼ 1). published results (rough cone-soil contact).
Additional analyses, using different values of the

rigidity index, Ir ¼ G=Su (Table 1), were carried out
to validate the PRM against other published results.
Figure 4 shows the influence of rigidity index on
cone factor values for a smooth cone-soil contact.
All studies indicate that the cone factors increase
approximately linearly with the logarithm of Ir . The
discrepancies between different published studies are
likely caused by the use of different numerical simu
lation techniques. Nevertheless, the results obtained
using the PRM agree reasonably well with other
published results, indicating that the PRM is valid
for simulating the large deformation cone penetra
tion process.
Figure 4. Comparison of cone factor values with other pub Similarly, Figure 5 shows the influence of
lished results (smooth cone-soil contact). rigidity index on cone factor values for a rough
cone-soil contact. The results obtained using the
All random fields were generated using the Kar- PRM also agree reasonably well with other pub
hunen-Loève (KL) expansion (Phoon et al. 2002) lished results, which further confirmed the valid
and a single exponential autocorrelation function. It ity of the PRM.
is not easy to identify the “correct” function (Spry
et al. 1988), but the single exponential autocorrel 4.2 Random field finite-element analysis (RFEM)
ation function is more commonly used (Phoon &
Kulhawy 1999). The K-L expansion was performed Figure 6a shows an example of a random field for
using 1000 terms to ensure sufficient accuracy. the undrained shear strength. Variations in Su values
were observed mainly in the vertical direction
because the vertical scale of fluctuation (δv ) is much
4 NUMERICAL MODELLING OF CONE smaller than the horizontal scale of fluctuation (δh ).
PENETRATION Significant fluctuations in the net cone tip resistances
with respect to depth, as shown by the solid black
line, were observed in Figure 6a. The profile is
4.1 Analysis with homogeneous soil properties
highly sensitive to the presence of weak and strong
This section utilizes published results in the litera soils. For example, at depths of between 0.1 m and
ture to validate the use of the PRM method. Figure 3 0.2 m, the cone is penetrating through soils whose
compares the results of the present analyses with Su values are relatively homogeneous, and hence
those reported in Lim et al. (2018) who also adopted the cone tip profile showed signs of convergence.
the PRM technique. All analyses indicate that the However, as the piezocone continues to penetrate
cone factor values reached a plateau at approxi beyond 0.2m, it will encounter the presence of
mately z/D > 6, even though Lim et al. (2018) started stronger soils at greater depths, thus causing the net
the penetration at the depth of z/D = 5. According to cone tip resistance to pick up quickly. In contrast,
Zhang et al. (2020), a deep penetration mechanism the net cone tip resistance from the analysis based
becomes prominent beyond this depth; therefore, a on the homogeneous soil medium quickly con
z/D = 6 was used in the present analyses to differen verged as the deep mechanism was mobilized at
tiate between a shallow and a deep penetration. z/D > 6 (Figure 6b).
755
4D below the cone tip. However, the variation of the
stresses within the stress bulb generated by the
RFEM analysis is not as smooth as that obtained in
the homogeneous soil medium.
Figure 8 shows the comparison of the plastic
yield zones obtained from the analysis in the homo
genous soil medium and the same selected random
realization at a cone penetration of z/D = 10. In the
homogeneous soil medium, a “butterfly” yield
region was obtained that extended horizontally to
5D from the cone tip, which is in good agreement
with the 4.7D and 4.65D reported by Lu et al.
(2004) and Yu (2000). In contrast, the yield region
obtained using the RFEM analysis for this random
field realization was smaller and more irregular
although the mobilized Tresca stresses were larger
in some zones (Figure 7). This is attributed to the
fact that the yield region propagated through
weaker zones within the soil layer in order to min
imize the overall energy dissipation (Zhang et al.
2020).
Figure 6. An example of random field and the net cone tip

resistance profiles (Ir = 150).
Figure 8. Comparison of yield regions (z/D = 10; Ir = 150).
The comparisons shown in Figures 7 & 8 indicate

that the analysis in the homogenous soil medium
that adopted highly idealized and uniform field con
ditions could not capture important aspects of realis
tic soil responses. To better understand the influence
Figure 7. Comparison of Tresca stress contours (z/D = 10; of the soil’s spatial variability on the cone response,
Ir = 150). 100 random RFEM analyses were carried out, and
the results plotted in Figure 9. For each realization,
Figure 7 plots contours of the Tresca stress, fluctuations in the computed cone resistance with
ðσ1 - σ3 Þ=2, where σ1 and σ3 are, respectively, the depth were observed, which are not unlike those
major and minor principal stresses at the instant recorded from field CPT measurements. Significant
when the cone tip is at z/D = 10, for the analysis in difference in the net cone tip resistances computed at
the homogenous soil medium and a selected RFEM penetration depths of z/D = 10 or greater were also
realization. Both analyses mobilized a stress bulb observed across the 100 realizations. While some
that spans from the ground surface to approximately realizations yielded net cone tip resistances that are
756
where the deep failure mechanism has become dom
inant (Figures 3 and 6), were considered for the
back-calculation.
Figure 9. Net cone tip resistance profiles of 100 random

field finite element analyses.
larger than 600 kPa (which is approximately 2 times

the value obtained in the homogeneous soil ana Figure 10. Procedures to back-calculate undrained shear
lysis), other realizations can yield values of as low strength values using CPT data.
as 150 kPa. These significant differences in the com
puted net cone tip resistances resulting from the con
sideration of spatial variability cannot be captured
by the analysis predicated on a homogeneous soil Figure 11 shows the statistical distribution of the
medium. back-calculated Su values. A lognormal distribution,
represented by the solid line, was fitted to the histo
gram. In general, the statistical distribution of the
5 BACK-CALCULATION OF UNDRAINED back-calculated Su values reasonably resembles the
SHEAR STRENGTH benchmark statistics (dashed line). However, as
shown in Figure 12, while the mean back-calculated
As demonstrated in the preceding section, the pres values of Su (from the 100 RFEM analyses) are
ence of spatial variability can alter the failure mech largely equal to the true (input) value, the back-
anism and, therefore, the net cone tip resistance calculated CoV values from the RFEM results
significantly. In this regard, the impact of spatial underestimated the ‘true’ CoV (prescribed in the
variability on the interpretation of CPT data was fur RFEM analyses to generate the random fields, see
ther investigated by back-calculating values of Table 1) by approximately 13%.
undrained shear strength Su from the net cone tip The same analyses were also repeated for the case
resistances obtained from the random field finite involving a rough cone-soil contact. The results
element analyses (RFEM).
Figure 10 illustrates the procedure to back-
calculate Su, which is similar to that adopted in prac
tice for the interpretation of CPT data in practice.
A deterministic PRM analysis was first performed
using a homogeneous soil medium with a constant
value of Su (Table 1) to obtain a “calibrated” cone
factor, Nkt, homogeneous. This step simulates the cali
bration exercise that is typically used in practice to
obtain the site-specific cone factor. In the second
step, 100 RFEM simulations were performed, each
using a heterogeneous soil medium characterized as
a random field with a spatially varying undrained
shear strength distribution (Table 1). The net cone
tip resistances obtained from the RFEM analyses
were then divided by the Nkt, homogeneous value to
arrive at the back-calculated Su values. Only resist Figure 11. Statistical distribution of the back-calculated Su
ances computed for penetration depths of z/D > 6, (smooth cone-soil contact; Ir = 150).
757
significantly alter the failure mechanism and the
net cone tip resistances. Undrained shear strength
values were then back-calculated following
a procedure that resembles the interpretation pro
cedures adopted in practice. The results indicate
that, while the interpreted mean Su value from
the RFEM analyses was close to the prescribed
(input) value, the value of CoV was underesti
mated by approximately 13-16%. Further research
is needed to provide an improved framework to
interpret CPT data.
REFERENCES
Abu-Farsakh, M., Tumay, M. & Voyiadjis, G., 2003.
Numerical parametric study of piezocone penetration
test in clays. International Journal of Geomechanics, 3
(2), pp.170–181.
Andersen, K.H., Andresen, L., Jostad, H.P. & Clukey, E.C.,
2004, January. Effect of skirt-tip geometry on set-up
outside suction anchors in soft clay. In International
Figure 12. Convergence of the mean and CoV of the Conference on Offshore Mechanics and Arctic Engin
back-calculated Su (smooth cone-soil contact; Ir = 150). eering (Vol. 37432, pp. 1035–1044).
Baligh, M.M., 1985. Strain path method. Journal of Geo-
technical Engineering, 111(9), pp.1108–1136.
indicate that, while the mean back-calculated Su Engin, H.K., Brinkgreve, R.B.J. & Van Tol, A.F., 2015.
agree with the true statistic, the CoV of the Su was Simplified numerical modelling of pile penetration–the
underestimated by approximately 16%, which is press-replace technique. International Journal for
Numerical and Analytical Methods in Geomechanics, 39
consistent with the results shown in Figure 12.
(15), pp.1713–1734.
Lu, Q., Randolph, M.F., Hu, Y. & Bugarski, I.C., 2004.
A numerical study of cone penetration in clay. Géotech
6 DISCUSSIONS nique, 54(4), pp.257–267.
Lim, Y.X., Tan, S.A. & Phoon, K.K., 2018. Application of
Figures 11 to 12 suggest that the conventional inter press-replace method to simulate undrained cone
pretation of CPT data that uses a “calibrated” cone penetration. International Journal of Geomechanics, 18
factor may not fully recover the true statistics of the (7), p.04018066.
parameter due to the presence of spatial variability. Orr, T.L., 2000. Selection of characteristic values and par
tial factors in geotechnical designs to Eurocode 7. Com
The back-calculated data can provide a reasonable
puters and Geotechnics, 26(3-4), pp.263–279.
estimate of the averaged properties of in-situ soils. Peck, R.B., 1969. Advantages and limitations of the obser
When only the most probable performance (Peck vational method in applied soil mechanics. Geotechni
1969) is required, the interpreted properties can be que, 19(2), pp.171–187.
reliably utilized. However, when characteristic Phoon, K.K. & Kulhawy, F.H., 1999. Characterization of
values (Orr 2007) are required for design purposes, geotechnical variability. Canadian geotechnical journal,
an overestimation of the characteristic value may 36(4), pp.612–624.
likely occur, resulting in an optimistic design of the Phoon, K.K., Huang, S.P. & Quek, S.T., 2002. Simulation
geotechnical system under study. While Zhang et al. of second-order processes using Karhunen–Loeve
expansion. Computers & structures, 80(12),
(2020) have provided a correction procedure, further
pp.1049–1060.
research into the impact of spatial variability on the Pan, Y., Shi, G., Liu, Y. & Lee, F.H., 2018. Effect of spatial
interpretation of CPT data is still warranted. variability on performance of cement-treated soil slab
during deep excavation. Construction and Building
Materials, 188, pp.505–519.
7 CONCLUSIONS Spry, M.J., Kulhawy, F.H. and Grigoriu, M.D., 1988. Reli
ability-based foundation design for transmission line
The continuous penetration of a piezocone into structures: Volume 1, Geotechnical site characterization
a statistically heterogeneous soil medium was strategy (No. EPRI-EL-5507-Vol. 1). Electric Power
Research Inst., Palo Alto, CA (USA); Cornell Univ.,
successfully modelled using the Press-Replace Ithaca, NY (USA). Geotechnial Engineering Group.
Method (PRM). The performance of the PRM Teh, C.I. & Houlsby, G.T., 1991. An analytical study of the
was rigorously validated using other published cone penetration test in clay. Geotechnique, 41(1),
results. Using the PRM and a set of statistics that pp.17–34.
is representative of natural soft clays, the pres Tehrani, F.S., Nguyen, P., Brinkgreve, R.B. & van Tol, A.
ence of spatial variability was shown to F., 2016. Comparison of Press-Replace Method and
758
Material Point Method for analysis of jacked piles. Yu, H.S., 2000. Cavity expansion methods in geomecha
Computers and Geotechnics, 78, pp.38–53. nics. Springer Science & Business Media.
Vesic, A.S., 1972. Expansion of cavities in infinite soil mass. Yu, H.S., Herrmann, L.R. & Boulanger, R.W., 2000. Ana
Journal of Soil Mechanics & Foundations Div, 98(sm3). lysis of steady cone penetration in clay. Journal of Geo-
Walker, J. & Yu, H.S., 2006. Adaptive finite element ana technical and Geoenvironmental Engineering, 126(7),
lysis of cone penetration in clay. Acta Geotechnica, 1 pp.594–605.
(1), pp.43–57. Zhang, W., Pan, Y. & Bransby, F., 2020. Scale effects
Wang, Z.Z. & Goh, S.H., 2018. Spudcan installation and during cone penetration in spatially variable clays. Géo
post installation behaviour in soft clay: The press- technique, pp.1–13.
replace method. In Numerical Methods in Geotechnical Zhang, J.Z., Huang, H.W., Zhang, D.M., Zhou, M.L.,
Engineering IX (pp. 1503–1510). CRC Press. Tang, C. & Liu, D.J., 2021. Effect of ground surface sur
Yu, H.S. & Whittle, A.J., 1999. Combining strain path ana charge on deformational performance of tunnel in spa
lysis and cavity expansion theory to estimate cone tially variable soil. Computers and Geotechnics, 136,
resistance in clay. Unpublished notes. p.104229.
759
© 2022 the Author(s), ISBN 978-1-032-31259-0
Some aspects of in situ testing of clay-glacial till mixture redeposited

as man-made fills
J. Wierzbicki, K. Stefaniak & S. Wilczyński
Institute of Geology, Adam Mickiewicz University, Poznań , Poland
B. Brzeziński
Labortest Sp. z o.o. Sp. k., Poznań, Poland
ABSTRACT: The paper presents the results of CPTU and FVT tests carried out in soils, which filled the
former brown coal excavation. In their natural state, these lands were pre-consolidated glacial till and clay.
Detachment of soil from the original deposit, transport and low-energy redeposition led to the creation of
multi-meter-thick non-consolidated embankments, which are now used as a construction subsoil. These
embankments consolidate under their own weight, however, unlike typical geological formations, they are
characterized by a decreasing influence of consolidation on the strength properties along with depth.
1 INTRODUCTION their strength and deformation characteristics, as

opposed to similar soils deposited as a result of geo
The contemporary development of urban areas is logical processes. In this case, the use of typical geo
related to, inter alia, the location of industrial facilities technical procedures may cause overestimation of soil
on the outskirts of cities. Simultaneously, the high strength and stiffness. Due to a complex structure of
cost of land properties within cities make it econom the man-made fill the in situ tests, as CPTU, seem to
ically viable to use advanced foundation techniques be the best solution for geotechnical investigations.
in case of weak, but cheap, grounds. Therefore, more Nevertheless, it is worth seeing how the geotechnical
and more often commercial buildings are located in properties of overconsolidated soil can be changed
areas not previously used for construction purposes. because of excavating and low energy redeposition.
One of the most efficient trends in such cases in For this purpose, the results of CPTU and FVT inves
Poland is the use of land reclaimed after a large scale tigations in the reclaimed land were compared with
industrial facilities as brown coal mines. Contempor the results of investigations carried out in the areas of
ary an ecological purpose makes it necessary to natural occurrence of glacial till of the same glaci
resign from the simple, but highly devastating from ation in Poland.
an environmental point of view, use of brown coal as
an energy source. Thus, brown coal mines in Poland
have been closing for a dozen of years and the rate of 2 MAN-MADE FILL TEST SITE
these changes will increase in the future. Some of the
old brown coal outcrops are being changed into green The studies of the embankments of the brown coal
areas, some of them are used for recreation with mine dump were located on the southern outskirts of
a newly built infrastructure, but some of them, espe the central part of the former Niesłusz outcrop, which
cially the ones close to the cities, are going to be is the oldest part of the Konin mining complex
used as lands for clean industry facilities. (Figure 1).
One of the good examples of such a use of The primary geological structure of this area con
reclaimed land are the areas of brown coal dumps in sists of the Quaternary sediments dominated by gla
the vicinity of the city of Konin, in central Poland. cial till. Two types can be distinguished: upper
The soil that has been dumped is a chaotic mixture of moraine, representing the Vistula glaciation, and
fragments of previously overconsolidated glacial till, lower moraine related to the Odra glaciation. In
sands and clays, very often over 20-meter thick, that clays, there are numerous intercalcations of sand-
consolidate naturally over time. The method of depos gravel and fluvioglacial lenses, as well as varvic
iting man-made fill, despite the fact that the dumps clays and erratic boulders (Widera 2001). These
are made of natural soil, leads to clear differences in sediments lie on the Pliocene and Miocene
DOI: 10.1201/9781003308829-112
760
formations, among which brown coal accumulated which, as a result, lead to the extraction of the
in sedimentation basins is a characteristic element. mineral. A characteristic element of opencast
Clays lay directly above the coal, intercalated with mining is the removal of the overburden of vari
dusty sand, sandy silt or fine-grained quartz sand ous thicknesses and compositions, the mining of
(Ratajczak & Hycnar 2017). the mineral and its transport to the destination as
well as the transport and tipping of the removed
overburden (Nowak & Kozłowski 2009). Almost
from the beginning of brown coal mining in the
Konin region, this process has been taking place
in a system in which the excavated material is
transferred to belt conveyors, which are then
transported to a gravity stacker that forms an
embankment - a dump.
The coal deposits in the Niesłusz outcrop were
exploited in the 1950s and the area was reclaimed
with little investment: at the site of the dump (where
the test site is located), works were undertaken to
restore the land for agricultural or forest use and the
largest depressions were developed as water reser
voirs. The lack of success in restoring the land to
agriculture and the close vicinity of a dynamically
developing city meant that since the 1990s, these
areas have been considered as areas for industrial
and commercial development (Gilewska & Otremba
2013) (Figure 2).
Figure 2. Flat surface of the man-made fill at the Niesłusz

test site.
The end of brown coal exploitation falls on

circa 1961 and agricultural reclamation has been
carried out since then, thus it can be assumed
that the test site has been subject to consolidation
for 60 years under its own weight of deposited
Figure 1. Location of the test sites. material.
The current geological structure of the subsur
face zone of the test site is between a few to
Brown coal, both in Poland and around the over a dozen meters of embankments lying on
world, is exploited almost exclusively using the the glacial till (Figure 3).The basic physical prop
opencast method. Exploitation of minerals is erties of the soil are given in Table 1 and the
related to a number of activities necessary to be typical grain size distribution is showed in
performed in the deposit and in its surroundings, Figure 5.
761
sediments of both phases, which was described,
among others, by Wierzbicki et al. (2008). This differ
entiation is visible, for example, in the natural mois
ture content (approx. 16% for the clays of the
Pomeranian phase and approx. 11% for the clays of
the Poznań phase), and especially in the values of
undrained shear strength (maximum 290 kPa for clays
of the Pomeranian phase and maximum 670 kPa for
clays of the Poznań phase) (Wierzbicki 2010).
Figure 3. Typical soil profile at the Niesłusz test site (soil

description according to ISO14688-2, gwl – ground water
level, wn – natural moisture, PI – plasticity index).
The ground water found in the subsoil behaves

like socalled suspended waters, which can be found
at different levels and do not constitute a single
piezometric level.
Table 1. Typical values of physical properties of the gla Figure 4. Typical soil profile of glacial till within the Par
cial till at the Niesłusz test site. sęta Lobe (soil description according to ISO 14688-2,
gwl – ground water level, wn – natural moisture, PI – plasti
Gs γ e Sr city index).
[-] [kN/m3] [-] [-]

2.69 19.42 0.58 0.74
Table 2. Typical values of physical properties of the gla
where: Gs – specific gravity, γ – soil unit weight, e – void cial till within the Parsęta Lobe.
ratio, Sr – degree of saturation.
Gs γ e Sr
[-] [kN/m3] [-] [-]

3 GLACIAL TILL TEST SITE
2.68 20.11 0.46 0.67
Research in naturally deposited glacial soil was car where: Gs – specific gravity, γ – soil unit weight, e – void
ried out in the area of the Parsęta Lob of the Vistula ratio, Sr – degree of saturation.
glaciation, which was formed during the local trans
gression of the already retreating ice sheet
(Figure 1). As a result of this event, two glacial till
deposits were deposited: the older, from the previous
glaciation phase (the Poznań phase), pre
consolidated by the transgressing Parsęta Lob, and
the younger, from the melt out phase, normally con
solidated or slightly preconsolidated (Pomeranian
phase) (Figure 4).
Granulometrically, these sediments are very
similar to each other, they are a mixture of sand
and gravel in about 50-70%, silt in about 20%
and a supplementary part of the till (Figure 5).
These soils were created within approx. 4,000
years, several thousand years ago, so the aging
processes did not play a decisive role in shaping
their geotechnical properties. The basic physical
properties of these soils are given in Table 2. Figure 5. Typical grain size distribution of Parsęta Lobe
However, the different load history led to a clear till in comparison with grain size distribution of some sam
differentiation of the geotechnical properties of the ples from CPTU 3 testing point at Niesłusz test site.
762
4 RESULTS The relatively low and sometimes even negative
u2 values are common in the case of the Vistula gla
The results of the research carried out on glacial ciation and are caused by a relatively low degree of
deposits in their natural habitat were presented in saturation and low clay content combined with the
detail by Wierzbicki (2010). On this basis, it can be overconsolidation effect. In case of this land, Wierz
concluded that the typical CPTU profile of the soils bicki (2010) also carried out a series of studies of
of the Vistula glaciation is shown in Figure 6. geotechnical features, which indicate, among others,
a different degree of pre-consolidation of individual
parts of the profile. On the other hand, the triaxial
CIU studies showed that the values of cone factor
Nkt range from 15 to 23 (in case of sediments with
high OCR and high content of gravel fraction).
As far as redeposited sediments are concerned,
data from three CPTU soundings located approxi
mately 30 m from each other were used. Boreholes
were made in the same places, samples were taken
for laboratory tests and FVT soundings were per
formed. The results of the CPTU tests indicate
a general similarity of the results at individual
research points (Figure 7). However, this similarity
is more visible along of the entire profile rather than
locally at individual depths. The results of the FVT
test, as well as the CPTU test, indicate a strong local
Figure 6. Typical CPTU results at glacial till test site (test differentiation of the geotechnical features of the
no. 19) (on the basis of Wierzbicki 2010). subsoil (Figure 8). Again, however, the general trend
of strength alterations is increasing along with the
depth of deposition (Figure 9). It is worth to notice
that no correction for plasticity was used during the
cu(FVT) calculations, due to low plasticity of the
investigated soil (PI <20).
Figure 8. Results of FVT at the depth of 4,5 m at the Nie

słusz test site (tests no. 2, 3 & 4).
The values of Nkt obtained on the basis of the

comparison with the results of CIU (natural sedi
ments) and FVT (redeposited sediments), allowed to
determine the undrained shear strength cu(su) in
accordance with the Equation 1 (Figure 10).
Figure 7. CPTU results at the Niesłusz test site (a) tests

no. 2, 3 & 4 and (b) average value of qt (av.) with plus and
minus standard deviation values (av. + st.dev. & av. – st. where: cu(su) = undrained shear strength, qn = nett
dev.). cone resistance, Nkt = cone factor
763
excessively influences the normalization, that is, the
trend along with the depth is permanently negative
across the entire profile. This means that the
undrained shear strength of the soil increases slower
and slower with increasing depth. Such an effect may
be characteristic of underconsolidated deposits. It is
particularly worth emphasizing that when the above
dependence is presented on a semi-logarithmic scale,
we observe variability around 1 in natural sediments,
while in sediments redeposited with depth, we still
observe a decreasing trend with depth (Figure 12). It
is probable that this rule can be universal for all non-
compacted man made but for now, it can be proved
only in the case of fills which are 16 m thick. Thus,
while the dependence of the strength properties in
glacial till can be described by a logarithmic function
Figure 9. The undrained shear strength profiles from FVT (which is typical in most geotechnical situations), in
at different testing points and the average profile of this case of the same granulometric but redeposited soil, it
parameter (av.) with plus and minus standard deviation
is not appropriate.
values (av. + st.dev. & av. – st. dev.).
Figure 10. The undrained shear strength of glacial till

(CPTU 19) and man-made fill (CPTU 2, 3, 4) on the basis Figure 11. Undrained shear strength normalized by vertical
of CPTU results (average values (av.) and plus and minus stress for glacial till and mam-made fill profiles.
standard deviation values (av. + st.dev. & av. – st. dev.).
The Nkt value for the ranges between 13 and 17 (depending
on the specific soil layer).
5 DISCUSION
In case of redeposited sediments, undrained shear

strength cu(su), apart from low values for glacial
sediments (Wierzbicki 2010), is also characterized
by a clear trend increasing with depth and thus with
vertical geostatic stress.
It is worth noting that a similar trend was not
observed in naturally deposited land, where the the
values of cu(su) are more related to different depos
ition phases and different pre-consolidation effects
than only to the increase of depth. The different
dependence of the value of cu(su) on the depth of
occurrence in both test sites is very clearly visible
after the normalization of this parameter by effective
vertical stress (Figure 11). Figure 12. Undrained shear strength normalized by vertical
It can be concluded that in case of redeposited stress in semi-log scale, for glacial till and mam-made fill
sediments, the inclusion of the stress component profiles.
764
6 SUMMARY ISO 14688-2: 2017. Geotechnical investigation and testing
- Identification and classification of soil – Part 2: Prin
Glacial tills, which in their natural habitat are pre ciples for a classification.
consolidated soils, lose some of their mechanical prop Nowak J. & Kozłowski Z. 2009. Technology of managing
erties during excavation and transportation in the form lignite deposits. In: Z. Kozłowski [ed.]: Techniczno
of large fragments and redeposition. They are still ekonomiczny ranking zagospodarowania złóż węgla
soils with sufficiently good geotechnical properties to brunatnego w aspekcie założeń polityki energetycznej
Polski. Wyd. Politechniki Wrocławskiej, Wrocław (in
place building structures directly on them. However, Polish).
the loss of strength properties is so great that the soils Radaszewski R. & Wierzbicki J. 2019. Characterization
have the profile characteristics corresponding to under- and engineering properties of AMU Morasko soft clay.
consolidated soils and therefore susceptible to consoli AIMS Geosciences, 5(2), 235–264.
dation subsidence. Thus, in case of land deposited Ratajczak T & Hycnar E. 2017. Associated minerals in lig
within the post outcrops of brown coal mines, the nite deposits. Wydawnictwo IGSMiE PAN (in Polish).
basic problem is not so much reaching the bearing Wierzbicki, J. 2010. Evaluation of subsoil overconsolida
limit state, but the serviceability limit state. Moreover, tion by means of in situ tests at the aspect of its origin.
Scientific dissertations No. 410, University of Life Sci
the research results indicate that the problem of land
ences in Poznań Publishing, Poland, 181–182 (in
unconsolidation increases with depth, which makes Polish).
this situation unique compared to, for example, natural Widera M. 2001. Occurrence and development of accom
soils (Radaszewski and Wierzbicki 2019). panying minerals in lignite open-pits in Wielkopolska.
Summary of papers presented at meetings PTGeol. in
Poznan X, 61–82 (in Polish).
REFERENCES Wierzbicki J., Paluszkiewicz R. & Paluszkiewicz R. 2008.
Shear strength of the deposits with relation to their
Gilewska M. & Oremba K. 2013. Revitalisation of origin, the glacial till of Vistula glaciation case. Land-
post-mining resions in the area of the town of Konin. form Analysis Vol. 9, Poznań, Poland, 390–393 (in
Inżynieria Środowiska no 29, 59–67 (in Polish). Polish).
765
© 2022 the Author(s), ISBN 978-1-032-31259-0
Cone penetration in a thin medium dense sand layer sandwiched by

different clay layers – LDFE analysis
Q. Xie & Y.X. Hu
University of Western Australia, Perth, Australia
M.J. Cassidy
University of Melbourne, Melbourne, Australia
M. Zhou
South China University of Technology, Guangzhou, China
ABSTRACT: The cone penetration test (CPT) has been the most common site investigation practice for
decades. This paper presents the results of large deformation finite element (LDFE) analysis that modeled
a CPT into soil comprising a thin medium dense sand layer sandwiched between two clay layers of different
undrained shear strengths. An extended critical state Mohr-Coulomb model is utilized to capture the stress-
dependent drained behavior of sand, while Tresca failure criterion is adopted for the undrained behavior of
clay. Because the sand layer and the surrounding clay layers work as a system that defines the mobilized soil
stiffness and soil interface deformation, two soil layer profiles are analyzed to study cone penetration resist
ance responses, including soft clay on top and stiff clay at bottom, and vice versa. It is found that a peak
resistance is registered in the thin sand layer, through the top and bottom clay layers show impact to the peak
as well. The impact on the peak resistance from the clay layer ahead (i.e. bottom layer) weights more than the
clay layer behind (i.e. top layer). The evolving soil flow mechanisms are revealed to explain the findings.
1 INTRODUCTION cone diameters both behind and ahead of the cone tip
(Lunne et al., 2002). When the influence zone crosses
1.1 Historical background layer interfaces, the measured cone resistance repre
sents not only the properties of the local soil around
The practice of inserting a rod into soft soil to find
the cone tip, but also the soils in adjacent layers.
embedded stiff layers dates back to the 19th century.
The inability to accurately estimate soil properties
However, the first recognizable form of the modern
of individual layers (particularly thin layers) in soil
cone penetration test (CPT) was developed in Nether
strata has attracted recent attention, with research
lands in 1934 to predict the ultimate bearing capacity
focused on idealized soil profiles of two or three uni
of piles (Barentsen, 1936). The early forms of CPTs
form soil layers in different sequences (e.g. weak soil
comprised a mechanical system of pushing rods to
over strong soil and vice versa, strong layer embedded
measure the rod tip resistance. In the late 1960s, elec
in soft soils and vice versa). Methodologies included
tric penetrometers came into general usage, enabling
elastic analysis (Vreugdenhil et al., 1994, Yue & Yin,
resistance to be continuously and automatically
1999), cavity expansion analysis (Mo et al., 2016),
recorded (Walker & Yu, 2010). Today the CPT is the
chamber testing (Tehrani et al., 2017), centrifuge test
most widely used in-situ testing method to investigate
ing (Mo et al., 2015) and analysis of field data (Youd
soil stratification due to its reliability and repeatability.
& Idriss, 2001, Yost et al., 2019). Moreover, large
deformation finite element (LDFE) analysis has been
1.2 CPT in layered soils applied for layered clays (Ma et al., 2016, Ma et al.,
2017).
The traditional design charts and formulas to identify
For CPT in clay-sand-clay soils, an early analyt
soil properties from CPT data were formulated on
ical solution was obtained by elastic analysis to
uniform soils of a single layer (Silva & Bolton,
investigate the “error” between the resistance in
2004). To directly apply such interpretation method to
a thin embedded stiff layer and the ultimate resist
layered soils can be problematic. This is because the
ance in an infinitely thick layer of the same stiffness
influence zone of a cone can span a distance several
(Vreugdenhil et al., 1994). As suggested by this
DOI: 10.1201/9781003308829-113
766
study, the “error” was a function of the embedded 2.2 Soil models
layer thickness and the stiffness ratio between adja
In this study, an extended Critical State Mohr-
cent layers. The finding was later confirmed by field
Coulomb (CSMC) model is utilized to capture the
data (Youd & Idriss, 2001) and examined numeric
stress-dependent behavior of sand (Li et al., 2013). It
ally (Ahmadi & Robertson, 2005). The latter sug
introduces the critical state concept into the classical
gested the “error” function only applied to thin and
Mohr-Coulomb (MC) model by linking the friction
normally consolidated sand layers embedded in soft
angle and dilation angle with the soil state parameter
normally consolidated clays. Thus, caution needs to
(i.e. the difference between the current void ratio and
be taken to apply these formulas for different cases.
the critical state void ratio at current stress level).
In the above studies, the stiffnesses of the top and
Elastic parameters (i.e. Young’s modulus) are func
bottom clays were assumed identical. Thus, the
tions of the void ratio and mean effective stress. To
existing solutions may not be suitable for the situ
fit the large deformation analysis computation,
ation where the top clay stiffness is different from
CSMC model balances its complexity with simpli
the bottom clay, with a sand layer sandwiched in the
city to avoid numerical difficulties. In contrast to the
middle. This study is set to explore the CPT in these
advanced model for sand, a simple Tresca model is
complex soil profiles.
used for clay. There is only one plastic parameter,
undrained shear strength (su), to determine an
1.3 Objective unchanged yielding surface. Young’s modulus is
assumed as proportional to the shear strength (E =
This study numerically investigates soil responses to
500 × su), which is within the commonly adopted
cone penetration in a medium dense sand layer sand
range (Hossain et al., 2005). Poisson ratio for clay is
wiched by two different clay layers. The two soil
set at 0.49 for undrained analysis while that of sand
layer arrangements for the top and bottom clay
is set at 0.3 for drained analysis.
layers include (i) the top clay being stiffer than the
The CSMC model along with the RITSS method
bottom clay, and (ii) the top clay being softer than
for CPT in layered soil has been extensively verified
the bottom clay. The soil flow mechanisms will be
with centrifuge tests data for two types of sand (Xie,
revealed and correlated to the features observed in
2020). Good agreement has been achieved providing
the cone resistance profiles. Additionally, stress his
confidence in deploying the model in this study. The
tory will be analyzed to further explain the cone
sand parameters utilized in this paper are calibrated
resistance profiles.
for Toyoura sand and are provided in Table 1. emax
and emin are the maximum and minimum void ratios;
ᶲc is the critical state friction angle; A, m and n are
2 METHODOLOGY
three scaling factors.
2.1 RITSS method
Cone penetration is a typical large strain problem Table 1. CSMC model parameters for Toyoura sand.
in finite element analysis. To overcome mesh dis
tortion in numerical analysis, the remeshing and emax emin ϕc/° A m N
interpolation technique with small strain (RITSS)
method is deployed in this study (Hu & Randolph, 0.985 0.611 32 0.5 7 0.75
1998a). RITSS method belongs to the arbitrary
Lagrangian-Eulerian (ALE) FE methods. It starts
with a series of small strain incremental displace 3 NUMERICAL MODELLING
ment analysis with an initial mesh, followed by
updating the co-ordinates of all nodes in the mesh
3.1 Problem definition
by the displacements calculated in the first steps
and then automatically remeshing the whole Figure 1 graphically defines the problem. A cone
domain accordingly. On the newly established penetrates in clay-sand-clay soil to a depth, d, meas
mesh, the field variables will be interpolated from ured from the cone shoulder to the original elevation
the old mesh. A new round of small strain analysis of the soil surface. The sand is medium dense
will be carried based on the new mesh and the pro (ID = 60%). The strengths of clay layers vary in two
cess will be cycled until the desired cone penetra cases: Case A, top clay (sut = 80 kPa) is stiffer than
tion depth is achieved. the bottom clay (sub = 10 kPa) and Case B, top clay
The FE based computer program AFENA (Carter (sut = 10 kPa) is softer than the bottom clay (sub = 80
& Balaam, 1995) is used in the study with the kPa). The reaction force acting on the cone, during
RITSS method implemented. H-adaptive mesh continuous penetration, is obtained through the
refinement cycles are also implemented in the pro LDFE analysis. The cone shaft and cone face are set
gram to optimize the mesh density and minimize the as fully smooth as it is normally manufactured with
discretization errors (Hu & Randolph, 1998b). polished stainless steel.
767
embedded sand layer in both cases: (I) blue zone –
cone resistance increases linearly and sharply with
the cone tip entering the sand layer, starting from the
kink point registered above the C-S interface; (II)
red zone - another linear increasing stage in resist
ance, but with lower gradient, leading to the peak
resistance in the sand layer; (III) orange zone – cone
resistance decreases after the peak, and converges to
the ultimate resistance in the bottom clay layer.
To illustrate the evolution of soil failure mech
anisms, eight stages are denoted on the resistance
profiles in Figure 3. For Case A: Stage A1
when the cone resistance is stabilized at its ultim
Figure 1. Cone penetration in clay-sand-clay soil. ate resistance in the top clay; Stage A2 - when
the resistance decreases after the stabilized resist
ance is reached; Stage A3 - when the kink point
3.2 Mesh setting is formed above the clay-sand interface; Stage A4
A finite element mesh setting where the cone pene - when cone resistance reaches a turning point
trates in the bottom clay layer is depicted in Figure 2. with changed gradient; Stage A5 - when the
The numerical model is simplified as 2D resistance is increasing at a reduced gradient;
axisymmetric, with the soil domain of equivalent 55D Stage A6 - when the peak resistance is reached in
in depth and 55D in radius. The lateral side is set as sand; Stage A7 to A8- when the resistance
a roller condition to eliminate soil horizontal move decreases gradually till the ultimate resistance in
ments and the bottom boundary is set as a hinge con the bottom clay is reached. Correspondingly, eight
dition to eliminate both vertical and horizontal Stages are chosen for Case B in a similar manner.
movements. The analysis uses six-node triangular The normalized penetration depths, d/D, at differ
elements with three Gauss points. The cone shoulder ent stages are listed in Table 2
is initially buried in the soil to avoid numerical diffi The differences observed between the two profiles
culties. A mesh refined zone is added adjacent to the are: (i) above the blue zone, the cone resistance
cone, which moves with the cone, to improve numer shows a slight decrease before reaching the sand
ical stability in the area of high strain concentration. layer in Case A, but not in Case B; (ii) in the blue
zone, the cone resistance increases more sharply in
Case A than in Case B; (iii) in the red zone, the cone
resistance reaches a higher peak in Case B than in
Case A; (iv) in the orange zone, the cone resistance
reduces more sharply in Case B than in Case
A. Discussions on these observed features are pro
vided in the following section.
Figure 2. Mesh setting when cone penetrates in bottom clay.
4.1 Cone resistance profiles Figure 3. CPT resistance profiles in clay-sand-clay soil: (a)
Case A: sut/sub = 8; (b) Case B: sut/sub = 0.125 (C-S refers
Figure 3 shows the cone resistance profiles of Cases to clay-sand interface while S-C refers to sand-clay
A and B. There are three zones identified around the interface).
768
Table 2. Normalized depths at different pene
tration stages.
Normalized depth (d/D)
Stage No. Case A Case B
1 12 12
2 16.5 16.5
3 19.3 19.25
4 20.3 20.05
5 22 23.7
6 22.7 27.3
7 29 29
8 30.2 30.85
4.2 Soil flow mechanism
4.2.1 Case A: A sand layer embedded in stiff over

soft clays
Figures on the left column of Figure 4 depict the
flow mechanism for Case A. In the top clay, a cavity
expansion failure is formed through Stages A1 to A2
until the cone tip touches the sand surface at Stage
A3. The localized failure mode should result in
a constant ultimate net cone resistance as the resist
ance only depends on the local clay stiffness. There
fore, the reduction in resistance at Stage A2 suggests
the cone is impacted by the sand layer ahead of the
cone, as soil displacements bend slightly downwards
comparing with Stage A1. From Stage A3 to A4, the
cone passes through the clay-sand interface as
grouped in the blue zone in Figure 3(a). Because the
magnitude of mobilized stress in sand is much
higher than that in clay, the resistance increases
sharply over these stages, resulting in a kink regis
tered at Stage A3. Soil movement is disconnected by
Figure 4. Soil flow mechanisms in Cases A (left) and

B (right).
769
the layer interface: cone squeezes the top clay side
ways while pushing the sand downwards. From
Stage A4, the cone shoulder detaches from the
sagged interface, which marks the full cone being
buried in the sand layer, and the resistance develop
ment enters a new stage as shown by the red zone in
Figure 3(a). The cavity expansion merges again at
Stage A5, implying the cone senses little impact
from surrounding clays except the overburden pres
sure. Moving to Stage A6, the soil displacements
bend downwards. This implies that the cone is sens
ing of the bottom soft clay, thus the cone resistance
starts to reduce. From Stages A7 to A8, soil move
ments keep bending downward as the cone
approaching the bottom clay. The mobilized sand
keeps reducing until the resistance reaches the stable
value in the bottom clay. At Stage A8, less than 1D
of sand is mobilized and the soil movement is
largely attracted to the bottom clay. This reduced
mobilization in sand causes the resistance to
decrease in the yellow zone in Figure 3(a). The
S-C interface sags extensively by ~1D, compared to Figure 5. Mean stress contours at (a) Stage A1, (b) Stage
0.18D for the C-S interface at Stage A3. A kink is B1, (c) Stage A2, (d) Stage B2 (MS refers the mean stress).
formed far from the original elevation of the inter

face (e.g. dkb > 1D).
4.2.2 Case B: A sand layer embedded in soft over Case B. This may be due to the influence of the top
stiff clays clay layer to the mobilized sand layer. In Case A, the
The soil flow mechanisms for Case B are depicted in top clay is stiffer than that in Case B, which mobil
the right column of Figure 4. Many features izes higher stress in sand layer, hence a stiffer mobil
observed in Case A can also be noted in Case B, ized sand layer. The sand stiffness promotes
including: (1) cavity expansion in the top clay before a sharper increase in the cone resistance.
the cone tip touches sand (Stage B1 to B2); (2) cavity Difference (iii): in the red zone, the cone resist
expansion in sand before the cone senses the bottom ance in Case B increases with a higher gradient and
clay (Stage B5); (3) the downwards bending soil dis lasts longer than those in Case A, and it results in
placement due to the attraction of the bottom clay a higher peak. This may be due to the influence of
(Stage B6 to B8). The similarities in the flow mech the bottom clay. In Case B, the bottom soil is stiffer.
anisms between Cases A and Case B can explain the The stiffer bottom clay reduces the sand layer sag
similar observations in the resistance profiles, while ging and promotes higher peak resistance.
the differences between the cases are discussed Difference (iv): in the orange zone, the cone resist
below. ance reduces more sharply in Case B than in Case
A. This can also be explained by the stress-dependent
4.2.3 Further discussion on the profile differences behavior of sand with the impacts from surrounding
Difference (i): above the blue zone, cone resistance clays. As discussed before, the mobilized stiffness of
reduction in top clay only appears in Case A. This sand depends on its in-situ mean effective stress. Soil
can be explained by the stress-dependent sand stiff stiffness determines the length of sensing distance to
ness with the impacts from the surrounding clays. the new layer ahead. The observation is consistent to
Figure 5 depicts the soil mean stress contours for that by Lunne et al. (2002) where a cone senses
Cases A and B. In CSMC model, sand stiffness (Es) a new layer earlier in a stiff layer than in a soft layer.
is a function of mean effective stress and void ratio. Thus, in Case A, the sand layer is stiffer than the
Extracted for Point E in Figure 5 (measured 1D from bottom clay, hence the cone can sense the bottom
the central line and 1D below the sand surface) from soft clay earlier and the reduction in resistance is
the output files: Es = 20.8 MPa at Stage A2, and 10.5 gradual. On the contrary, in Case B, the bottom clay
MPa at Stage B2. The stiffness of top clay (Ec) is 40 is stiffer, hence the cone senses the bottom clay later
MPa in Case A and 5 MPa in Case B. Therefore, the and the reduction in resistance is more dramatic.
sand layer is mobilized as a soft layer below the cone Therefore, it was observed that the top clay has
at Stage A2, resulting the cone resistance reduction more influence on the cone resistance profile in the
above the blue zone in Case A. This phenomenon blue zone and the zone above. The bottom clay has
remains until the cone penetrates into the sand layer. more influence on the cone resistance profile in the
Difference (ii): in the blue zone, cone resistance red zone and orange zone. As the peak resistance is
profile exhibits sharper increase in Case A than in developed in the red zone, the bottom clay has more
770
influence on the cone peak resistance. The peak sand. Proc. 18th International Conference on Soil Mech
resistance in Case A is about 2/3 of that in Case B, as anics and Geotechnical Engineering (pp. 2359–2362).
the bottom clay is softer in Case A than in Case B. Lunne, T., Powell, J. J. M. & Robertson, P. K. (2002) Cone
penetration testing in geotechnical practice, CRC Press.
Ma, H., Zhou, M., Hu, Y. & Hossain, M. S. (2016) Inter
5 CONCLUSION pretation of Layer Boundaries and Shear Strengths for
Soft-Stiff-Soft Clays Using CPT Data: LDFE Analyses.
Journal of Geotechnical and Geoenvironmental Engin
The paper studies CPT in a thin medium dense sand eering, 142.
layer sandwiched by two different clay layers. Two Ma, H., Zhou, M., Hu, Y. & Hossain, M. S. (2017) Inter
Cases of soil profiles are established, such as stiff pretation of layer boundaries and shear strengths for
over soft clay and vice versa. By deploying the stiff-soft-stiff clays using cone penetration test: LDFE
CSMC model, the stress-dependent behaviors of sand analyses. International Journal of Geomechanics, 17,
are well captured to reveal a systematic soil responses 06017011.
of cone penetration into clay-sand-clay soils. Mo, P.-Q., Marshall, A. M. & Yu, H.-S. (2016) Interpret
The study shows that the top clay has more influ ation of cone penetration test data in layered soils using
cavity expansion analysis. Journal of Geotechnical and
ence on the cone resistance profile in the top layer and
Geoenvironmental Engineering, 143, 04016084.
the blue zone. The bottom clay layer has more influ Mo, P. Q., Marshall, A. M. & Yu, H. S. (2015) Centrifuge
ence on the cone resistance profile in the red zone and modelling of cone penetration tests in layered soils.
the orange zone (refer to Figure 3 for coloring zones). Géotechnique, 65, 468–481.
A softer bottom clay in Case A induces more bending Silva, M. F. & Bolton, M. D. (2004) Centrifuge penetration
of the sand layer. A stiffer bottom clay in Case tests in saturated layered sands. Proceedings of 2nd
B induces higher cone peak resistance in the sand International Conference on Site Characterization, vols.
layer. (Vol. 1, pp. 377–384).
Further studies are needed to investigate more Tehrani, F. S., Arshad, M. I., Prezzi, M. & Salgado, R.
triple-layer profiles to provide guidance in CPT data (2017) Physical modeling of cone penetration in layered
sand. Journal of Geotechnical and Geoenvironmental
interpretations. Engineering, 144, 04017101.
Vreugdenhil, R., Davis, R. & Berrill, J. (1994) Interpret
ation of cone penetration results in multilayered soils.
REFERENCES International Journal for numerical and analytical
methods in geomechanics, 18, 585–599.
Ahmadi, M. M. & Robertson, P. K. (2005) Thin-layer Walker, J. & Yu, H. S. (2010) Analysis of the cone penetra
effects on the CPT qc measurement. Canadian Geotech tion test in layered clay. Geotechnique, 60, 939–948.
nical Journal, 42, 1302–1317. Xie, Q. (2020) Large deformation finite element analysis
Barentsen, P. (1936) Short description of a field testing on cone penetration test in layered sand-clay soils.
method with cone-shaped sounding apparatus. Proceed School of Engineering. Australia, University of Western
ings 1st International Conference on Soil Mechanics Australia.
and Foundation Engineering (pp. 6–10). Yost, K. M., Cox, B. R., Wotherspoon, L.,
Carter, J. P. & Balaam, N. P. (1995) AFENA users’ manual. Boulanger, R. W., Van Ballegooy, S. & Cubrinovski, M.
Centre for Geotechnical Research, Department of Civil (2019) In Situ Investigation of False-Positive Liquefac
Engineering, University of Sydney, Australia. tion Sites in Christchurch, New Zealand: Palinurus
Hossain, M., Hu, Y., Randolph, M. & White, D. (2005) Road Case History. Eighth International Conference on
Limiting cavity depth for spudcan foundations penetrat Case Histories in Geotechnical Engineering (pp.
ing clay. Géotechnique, 55, 679–690. 436–451). American Society of Civil Engineers.
Hu, Y. & Randolph, M. F. (1998a) A practical numerical Youd, T. L. & Idriss, I. M. (2001) Liquefaction resistance
approach for large deformation problems in soil. Inter of soils: summary report from the 1996 NCEER and
national Journal for Numerical and Analytical Methods 1998 NCEER/NSF workshops on evaluation of lique
in Geomechanics, 22, 327–350. faction resistance of soils. Journal of geotechnical and
Hu, Y. & Randolph, M. F. (1998b) H-adaptive FE analysis geoenvironmental engineering, 127, 297–313.
of elasto-plastic non-homogeneous soil with large Yue, Z. Q. & Yin, J. H. (1999) Layered elastic model for
deformation. Computers and Geotechnics, 23, 61–83. analysis of cone penetration testing. International jour
Li, X., Hu, Y. & White, D. J. (2013) A large deformation nal for numerical and analytical methods in geomecha
finite element analysis solution for modelling dense nics, 23, 829–843.
771
© 2022 the Author(s), ISBN 978-1-032-31259-0
Study on SPT N-values and relative density through various soundings in

full-scale chamber test ground
H. Yabe, K. Harada, T. Ito & E. Watanabe
Geotechnical Division, Fudo Tetra Corporation, Chuo-ku, Tokyo, Japan
ABSTRACT: In Japan, the evaluation of ground strength, such as the liquefaction resistance, is often carried
out through the N-value of the standard penetration test (SPT), and it is used in various designs. Other soundings
are also used, including the cone penetration test (CPT) and the Swedish weight sounding (SWS) that is compara
tively simpler when compared with SPT, and conversion formulae to N-value have also been proposed by various
researchers. In this paper, the various soundings described above (SPT, CPT, and SWS) were carried out in
a sandy ground, where the density was controlled and measured by Radio Isotope (RI) method. Then, the correl
ation between the SPT N-value and the relative density Dr was examined. Based on the results, the
N-Dr correlation was examined using the results of various soundings in a full-scale chamber test where the
ground density was controlled. The results confirmed the applicability of the correlation formulae, with good cor
relation confirmed not only between Dr and SPT N-value, but also with the N-value estimated from each
sounding.
1 INTRODUCTION The authors carried out the above-mentioned

soundings in an artificially constructed sandy ground
Japan has long been prone to ground disasters due to to conduct full-scale tests on ground improvement
earthquakes and heavy rain, and recent events indicate methods. The correlation between SPT N-values and
that it is more prone to these hazards than ever. Coun Dr was examined by comparing: (1) the SPT
termeasure works have generally been implemented to N-values with the normalized N-values estimated
prevent ground disasters. Ground surveys are essential from the CPT/SWS penetration resistance values; and
when determining the necessity and the scale of coun (2) the estimated Dr obtained using available conver
termeasure works and confirming the effectiveness sion formulae with the measured Dr during the con
after their implementation. The methods frequently struction of the full-scale test ground. This paper
used for such ground surveys include in-situ investiga reports on comparative studies carried out for two
tions (soundings) through the standard penetration test cases of ground with different target relative densities.
(SPT). The SPT N-values obtained have been used in
the design of various facilities. For example, in the
case of the sand compaction pile method used as 2 OUTLINE OF FULL-SCALE TEST GROUND
a countermeasure against liquefaction, SPT N-value is
used to determine if the implementation of the coun The full-scale test ground is shown in Figure 1. The
termeasure works is necessary and to confirm the test ground covered an area measuring 12.2 m ×
level of improvement after their execution. Also, con 23.4 m and a depth of 7.5 m, enabling the use of
sidering that the design is carried out considering the actual construction machines.
relative density (Dr) estimated from the SPT N-values As shown in the figure, the full-scale test ground
obtained, the appropriateness of SPT N-values and had a deck slab made of concrete and was sur
their evaluation (Dr-conversion) are important. rounded by side walls made of steel sheet piles such
In addition to SPT, other soundings used include the that a saturated ground could be reproduced. The test
cone penetration test (CPT) and the Swedish weight ground was saturated after backfilling with sand. For
sounding test (SWS). Because there are cases where this purpose, vertical pipes were installed at several
CPT and SWS can be implemented relatively easier locations on the sidewalls with a 50 cm thick crushed
than SPT, formulae to obtain the SPT N-value and stone layer at the bottom. Water was injected and dis
Dr from the results of CPT and SWS have been tributed through the pipes and the crushed stone layer
developed. to saturate the full-scale test ground from the bottom.
DOI: 10.1201/9781003308829-114
772
The pit sand produced in Namegata (Japan) was a roller and other equipment to ensure trafficability
used to backfill the test ground. The physical proper for the heavy machines to be used on the test ground.
ties of the pit sand are shown in Table 1 and Figure 2. The density of the full-scale test ground was con
Table 1 also shows the physical properties of Toyoura trolled by measuring the relative density of each layer
sand, a Japanese standard sand, for comparison. at 16 locations through the RI (Radio Isotope) method.
Then, after completing the construction of the full-
scale test ground, various soundings were conducted.
Figure 3 shows the locations where the density meas
urement and various soundings were conducted. In the
figure, the locations of the density measurement are
indicated by numbers, and the sounding locations are
indicated by letters (such as SPT, CPT, and SWS).
Also, in Case 1, the full-scale ground was constructed
with Dr = 60% while in Case 2, the ground has
Figure 1. Photos of the full-scale chamber. Dr = 80%.
Table 1. Physical properties of the backfilling material

and Toyoura sand.
Fc ρs
Material (%) (g/cm3) emax emin emin* Uc
Toyoura 0.0 2.65 0.985 0.611 0.674 1.51

Sand※
Namegata Sand 4.6 2.716 1.121 0.682 0.586 2.06
emin*:Minimum void ratio from compaction test

※Harada et al. (2003)
Figure 3. Relationship between the locations of density

measurement and soundings within the chamber.
3.2 SPT, CPT, and SWS sounding results

Figure 4 shows the measured profiles of the penetra
tion resistance values (SPT N-values, qc-values, and
NSW values) for each sounding. Comparing the meas
ured results for the same type of sounding conducted
more than once (except for CPT in Case 1), a similar
trend can be observed regardless of the measurement
location. In Case 1, the values of the penetration
Figure 2. Grain size distribution curve of backfilling resistance for each sounding do not change much
material.
with depth. However, in Case 2, the penetration
resistance values for each sounding increases with
3 DENSITY CONTROL DURING depth, and the overall values are larger than those of
CONSTRUCTION OF FULL-SCALE TEST Case 1 (which has lower relative density).
GROUND AND SOUNDING TEST RESULTS Figure 5 shows the RI value profiles measured in
the vicinity of the sounding points. Two to four points
in the vicinity of the sounding were selected based on
3.1 Density control method
Figure 3, and the average of the RI values is shown.
The full-scale test ground was constructed for each In Case 1 (target relative density Dr = 60%), the
of the two cases through backfilling with sand mater overall variation of the measurement results was
ial layer by layer, with each 0.3 m thick layer spread large, and there were places where the measurement
and compacted to the required density. More specific results exceeded the control target density. However,
ally, the first layer was spread to level the ground. the overall average of the measurement results, with
The remaining three layers were compacted by dry density ρd = 1.45-1.50 g/cm3 and relative density
773
Figure 5. Results of density measurement corresponding to
Figure 4. Results of soundings for each case. each sounding.
Dr = 55-70%, was within the acceptable range when values obtained in natural or artificial grounds. The
compared with the target values. conversion formulae to obtain Dr are derived from
In Case 2 (target relative density Dr = 80%), similar the relation between the penetration resistance and
variation in the measurement results was observed, relative density in the range of void ratios reported in
with most of the values exceeding the target value. the available studies shown in Figure 6.
The overall average relative density was in the range Figure 7 shows the procedure to estimate Dr from
of 85-100%, which was greater than the target value the CPT/SWS penetration resistance and SPT
of 80%. N-values. What follows below is a discussion on the
Comparing the profiles shown in Figures 4 and 5, consistency between the sounding results in the sand
there was variation in the measured RI values at vari layers of the full-scale test ground constructed with the
ous points around the sounding locations. However, controlled density and those from studies in natural
the overall average RI values generally showed ground.
a similar trend to the sounding results.
4.1 Comparison between SPT N-value and
normalized N-value
4 DISCUSSION ON THE CORRELATION
BETWEEN THE SOUNDING RESULTS Figure 8 shows the comparisons between the SPT
N-value and the normalized N-value calculated from
Table 2 shows the correlations available in the litera the penetration resistance of CPT and SWS using the
ture between Dr and penetration resistance as well as conversion formulae shown in Table 2 for the rele
the relations between SPT N-values and CPT/SWS vant cases. According to the figure, the SPT N-values
774
Table 2. Conversion equation each sounding test.
Standard Penetration Swedish weight Sounding

Test Cone Penetration Test Test
SPT CPT SWS
Penetration resistance N-value, N Penetration resistance, qc (qt>0.2MPa) Normalized N-value, N

ffiffiffiffiffi
(normalized value) N=2WSW+0.06NSW
WSW : Static weights to
Meyerhof (1957) cause static penetra
tion of the rod
NSW : Number of half a turn
of a handle per meter
JGS (2013)
ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Correlating N-value, ————
equation N
Suzuki et al. (2003) Tsukamoto et al. (1999)
Relative
density,
Cubrinovski et al.(1999) Harada et al. (2008) Tsukamoto et al. (1999)
Dr(%)
Figure 6. Relationship between the range of void ratio and gradient (SPT, CPT and SWS).
Figure 7. Correlations among each sounding test.
775
and the normalized N-values have distribution profiles
which are similar to each other. Also, according to
Figure 9, which shows the correlation between the
SPT N-values and the normalized N-values (estimated
from the penetration resistance of CPT and SWS using
the conversion formulae), the SPT N-values correlate
well with the normalized N-values.
4.2 Comparison of RI-measured Dr and

sounding-estimated Dr
Figure 10 shows the comparisons between the
Dr estimated from the relevant soundings (SPT,
CPT, and SWS) and the Dr measured through the RI
method for each respective case. The Dr obtained
through the soundings was calculated from the aver
age values of the respective cases. The RI values
were the average of the measurements at the loca
tions near the sounding locations.
From the figure, the Dr measured through the RI
method is generally larger than the Dr estimated from Figure 9. Comparison between measured and estimated
the results of respective soundings. Also, the Dr based N-value.
on CPT and SWS is close to the Dr measured through
the RI method, but the Dr estimated from SPT largely
differs from the Dr measured by the RI method. Simi
lar tendencies can be found in Figure 11.
The reasons for these observed trends are con
sidered to be threefold. The first reason is the differ
ence between the full-scale test ground and an actual
ground. The test ground was constructed by stacking
sand layers each with thickness of 30 cm and satur
ated immediately before conducting the soundings.
In contrast, the actual ground has complex strength
distribution due to the stress history developed
during a long accumulation period and aging effects.
Thus, it is considered that the Dr measured in the
full-scale test ground with younger age differs from
Figure 8. Measured N-values and those estimated from Figure 10. Measured relative density and estimated values
CPT and SWS. from each sounding.
776
Figure 12. Relationship between cyclic strength and rela
tive density of Namegata sand and of Toyoura sand.
be approximately applied to the correlation between

control values and the Dr converted from the results of
Figure 11. Comparison between measured and estimated
relative density. various soundings. We will continue our comparative
study between the measured density in the test ground
and the density obtained through the correlation formu
the Dr calculated by the conversion formulae devel
lae to further verify the accuracy of the correlation
oped for natural ground.
formulae.
The second reason is the difference in penetration
mechanisms. The SPT, which dynamically applies
loads by dropping a hammer into the ground, tends REFERENCES
to have a wider variation in relative density than the
CPT and SWS, which statically apply loads to the Cubrinovski, M., Ishihara, K. 1999. Empirical correlation
ground. It is also possible that the degree of satur between SPT N-value and relative density for sandy
ation of the test ground affected the Dr. soils. Soils and Foundations, 39(5): 61–71.
The third reason is the difference in the concepts Harada, K., Ishihara, K. Orense, R.P., Mukai, J. 2008. Rela
of the minimum void ratio, emin, used to calculate the tions between penetration resistance and cyclic strength
relative density. There are cases where the minimum to liquefaction as affected by Kc–conditions. Proc, Geo-
technical Earthquake Engineering and Soil Dynamics
void ratios obtained through compaction tests (emin*)
IV, Sacramento CA, Paper 111.
are smaller (denser) than the emin based on its original Harada, K., Yasuda, S., Niwa, T., Shinkawa, N. and
definition (specified in standard). It is considered that Ideno, T. 2003 Evaluation of im-proved ground by com
there is a large difference between the measured and paction containing fines, JSCE Journal of Earthquake
estimated Dr values because the conversion formulae Engineering, Vol. 27, No.35 (in Japanese).
listed in Table 2 are based on the conventional Dr, Japanese Geotechnical Society 2013. Chapter 4: Method
while the relative density from the RI method repre for Swedish weight sounding test. Geotechnical and
sents compacted state. Note that there is a significant Geoenvironmental Investigation Methods, 331.
difference between emin and emin* for the backfilling Kokusho, T., Yoshida, Y., Nishi, K., Esashi, Y. 1983.
A study on evaluation method of seismic stability of
sand, as shown in Table 1.
dense sandy ground: Part 1 - Dynamic strength proper
Figure 12 compares the relationship between the ties of dense sand. Central Research Institute of Electric
cyclic strength and relative density for Namegata Power Industry Report, No. 383025, 1–44.
sand and for Toyoura sand. As shown in the figure, Liao, S.C., Whitman, R.V. 1986. Overburden correction
the cyclic strength of Toyoura sand increases rapidly factors for SPT in sand. J. of Geotech. Engrg., 112(3):
at relative density of 80%, while that of Namegata 373–377.
sand increases slowly. In other words, the actual Meyerhof, G.G. 1957. Discussion on research on determin
minimum void ratio of the backfilling sand is smaller ing the density of sands by spoon penetration testing.
than that from the standard test, and the relative Proc. 4th Int. Conf. on SMFE, 3: 110.
Suzuki, Y., Tokimatsu, K., Sanematsu, T. 2003. Correl
density is considered to be overestimated.
ations between CPT data and soil characteristics
obtained from SPT. Journal of Structural and Construc
tion Engineering, 566: 73–80.
5 SUMMARY Tsukamoto, Y., Ishihara, K., Sawada, S. 2004. Correlation
between penetration resistance of Swedish weight
The various soundings in the full-scale test ground sounding tests and SPT blow counts in sandy soils. Soils
confirmed that the existing correlation formulae could and Foundations, 44(3): 13–24.
777
© 2022 the Author(s), ISBN 978-1-032-31259-0
Correcting measured CPT tip resistance for multiple thin-layer effects

K.M. Yost, J. Cooper, R.A. Green, E.R. Martin & A. Yerro
Virginia Tech, Blacksburg, VA, USA
ABSTRACT: Multiple interbedded fine-grained layers in a sand deposit have a “smoothing” effect on the
measured tip resistance (qc) from the cone penetrometer test (CPT). This can result in an underestimation of
the predicted liquefaction resistance of the sand layers. Herein, the efficacies of two multiple-thin-layer cor
rection procedures are evaluated using published calibration chamber test data. The results highlight limita
tions of the assessed procedures for profiles with layers less than 40 mm thick. A new approach to estimate
the “true” qc (i.e., values that would be measured in a stratum absent of multiple thin-layer effects) from
measured qc is explored. The proposed numerical optimization algorithm searches for “true” soil profiles with
a finite number of layers. We compare two versions of the algorithm that numerically optimize different func
tions, one of which uses a logarithm to refine fine-scale details, but which requires longer calculation times to
yield improved corrected qc profiles.
1 INTRODUCTION Forward procedures, like the ones proposed by

Youd et al. (2001), Ahmadi and Robertson (2005),
Cone penetrometer test (CPT) data (i.e., tip resist de Greef and Lengkeek (2018), and others, apply
ance, qc, and sleeve friction, fs) are typically a series of corrections directly to the measured qc
reported at 1 to 2 cm depth increments. However, (qm) to obtain a “corrected” qc (qc°rr) that is
data measured at a given depth are not representa a best estimate of the “true” qc (i.e., qc that would
tive only of the soil at that discrete depth, but are be measured in the profile absent of multiple thin-
actually averaged or “blurred” values of the “true” layer effects, qt). Inverse procedures (e.g., Bou
values that fall within a zone of influence above langer and DeJong 2018; Cooper et al. 2022) start
and below the cone tip. For example, this zone of by making a guess of qt (qinv) and then apply an
influence can include soils that are as far away as artificial blurring model (representative of the
10 to 30 times the cone diameter (dcone) ahead of blurring effect of the cone in layered soils) to qinv
the cone tip (Ahmadi and Robertson 2005). Thus, to obtain a “simulated” qm (qm,sim). Then, qm,sim is
the presence of multiple interbedded fine-grained compared to the actual qm. If the misfit between
layers in a sand deposit can result in a significant qm,sim and qm is too large, an update is automatic
underestimation of the predicted liquefaction ally applied to provide a new, improved qinv
resistance of the sand layers. This phenomenon is guess. The procedure iterates until the misfit is
referred to as multiple thin-layer effects. acceptable, at which point the last guessed qinv is
Several methods have been proposed to correct considered to be a best estimate of qt.
CPT data for multiple thin-layer effects. These To develop and validate multiple-thin-layer cor
procedures seek to use the measured CPT data in rection procedures (forward or inverse), one must
an interlayered soil profile and estimate the “true” know both qm and qt for a given layered soil pro
CPT data that would be measured in the profile file. Typically, we only know qm. CPT calibration
absent of multiple thin-layer effects (i.e., a true chamber tests or numerical simulations can be
representation of the CPT data at a discrete used to obtain both qm and qt. In the following sec
depth in the profile). In general, approaches to cor tions, the efficacy of existing forward and inverse
rect for multiple thin-layer effects can be split into procedures for correcting multiple thin-layer
two categories: forward procedures and inverse effects is assessed directly using calibration cham
procedures. In both cases, all procedures proposed ber data. Then, an alternate inverse procedure is
thus far focus on correcting qc, not fs (though some proposed and assessed using the same calibration
provide methods to retroactively adjust fs based on chamber data. Finally, a brief discussion of limita
the corrected qc; complexities in correcting fs are tions and areas for future procedure improvement
briefly discussed in Section 4 of this paper). is provided.
DOI: 10.1201/9781003308829-115
778
2 ASSESSING EXISTING PROCEDURE a correction factor KH (a function of: layer thickness,
EFFICACY H; dcone; and normalized ratio between minimum
and maximum qm in a layer, qratio) that increases tip
2.1 Overview of Deltares calibration chamber tests resistance in thin dense layers. In short, the DEL
procedure requires an input of qm and outputs
A series of CPT calibration chamber tests performed at a corrected tip resistance (qc°rr) that is an estimate of
Deltares by de Lange (2018) were used to assess the qt. KH factors developed for the DEL procedure are
efficacy of existing multiple-thin-layer correction pro provided in Figure 1.
cedures. Several soil profiles were considered in these
tests, including layered sand-clay profiles and refer
ence (single layer) sand profiles. The CPTs performed 2.3 Overview of BD18 inverse procedure
in the reference sand profiles were used to estimate qt The Boulanger and DeJong (2018) [BD18] inverse
for the sand layers in the sand-clay models, where the multiple-thin-layer correction procedure proposes
reference sand profiles had similar relative densities that qm is equal to qt convolved with a depth-
(DR) and overburden pressures (σ’v) to the sand layers dependent spatial filter, wc:
in the layered sand-clay models. No reference clay pro
files were constructed, however, qt for the clay could
be estimated from one of the sand-clay models that
had clay layers that were thick enough (200 mm, or
8dcone, thick) for qt to fully develop. Details of how the
tests were performed and information about the soils where � represents a convolution, and qm, qt, and wc
used to create the profiles are excluded for brevity and are all functions of depth (z). The spatial filter wc is
can be found in de Lange (2018). a discretization of a continuous function that
represents the influence of soil above and below
the cone tip on qc at a particular depth. The BD18
procedure uses an iterative splitting optimization
technique to solve the misfit function defined by:
The BD18 procedure includes two steps to smooth

the results to prevent them from becoming unstable:
first, a smoothing step performed after each iteration of
the inversion that computes a moving average of qinv
over a pre-defined smoothing window, and second,
a low-pass spatial filtering step performed once the
optimization procedure has converged to a solution.
After the optimization procedure is performed,
a separate interface correction procedure is applied in
which sharp transitions in qinv (which presumably cor
respond to layer interfaces) are identified. A constant
value of qinv (either the maximum or minimum qinv
identified in the transition zone) is then applied to the
entire transition zone within a layer, effectively setting
Figure 1. Thin-layer correction factor (KH) values derived a single qinv value for each identified soil layer.
from the de Lange (2018) calibration chamber tests (shown A modified version of this procedure, termed
as points) were used to define KH curves for several nor BD18MOD, was also explored in this study. For
malized thin-layer thicknesses (H/dcone) for the Deltares this variant of the procedure, the recommended
procedure (Yost et al. 2021a).
smoothing and filtering steps that ensure conver
gence of the solution were adjusted. Specifically,
the smoothing window was reduced to a maximum
2.2 Overview of Deltares forward procedure
of three qm data points (the default smoothing
Trends observed by de Lange (2018) using the CPT window is max[3, ceiling(0.866dcone/Δz)], where Δz
calibration chamber dataset were used to develop is the depth interval at which the data were col
a forward multiple-thin-layer correction procedure, lected), and the low pass spatial filtering step
the details of which are discussed thoroughly in Yost applied after the inversion was eliminated. We
et al. (2021a). This procedure, termed the Deltares found that this improved the performance of the
[DEL] procedure, consists of two parts: (1) identifi procedure in identifying very thin layers, but also
cation of layer interfaces based on peaks and troughs destabilized the solution, resulting in non-
in qm, and (2) computation and application of convergence for some scenarios.
779
Figure 2. Results from application of multiple-thin-layer correction procedures to de Lange (2018) calibration chamber
data for: (a) Soil Model 4 with 40-mm-thick clay layers; and (b) Soil Model 8 with 20-mm-thick-clay layers (Yost et al.
2021a).
2.4 Assessment of procedures using calibration identifying all of the 20-mm-thick layers in this pro
chamber dataset file or providing a good estimate of qt for those
layers, although the DEL procedure did a slightly
Direct assessments of the efficacies of the DEL for
better job at identifying the layer interfaces com
ward procedure, and the BD18 and BD18MOD
pared to the BD18 and BD18MOD procedures.
inverse procedures were performed using the calibra
Several other useful observations were made
tion chamber data from de Lange (2018). Note that
during this exercise. None of the procedures per
this is a biased comparison of the efficacies of the pro
formed well on the reference sand (single layer) soil
cedures, since the DEL procedure was developed and
models (i.e., all procedures erroneously identified
calibrated using this dataset. Regardless, the DEL and
and attempted to correct for thin, interbedded layers
BD18/BD18MOD procedures were applied to the qm
that were not present). In general, the BD18MOD
for the layered sand-clay soil profiles reported by de
procedure performed better than the BD18 procedure
Lange (2018). The resulting qc°rr (from the DEL pro
on this dataset, however, the modifications to the
cedure), qinv (from the BD18 procedure), and qinvm°d
smoothing steps tended to de-stabilize the solution
(from the BD18MOD procedure) were compared with
(e.g., see phantom peak and trough between 0.4 and
the qt determined from the reference sand profiles.
0.55 m in Figure 2b). Complete results from this
Select results from this exercise are shown in Figure 2.
analysis are provided in Yost et al. (2021a).
As shown in Figure 2a, the DEL procedure was
effective at identifying the interbedded sand layers
and outputting a qc°rr that is a good estimate of qt for 3 PROPOSED ALTERNATIVE INVERSE
those layers. Conversely, the BD18 procedure does PROCEDURE
not identify the interbedded layers, and actually esti
mates a qinv slightly less than the qm (and signifi Because none of the procedures discussed in the pre
cantly less than qt) in the interbedded sand layers. vious section were shown to be especially effective
The qinv values estimated using the BD18MOD pro at resolving multiple thin-layer effects for profiles
cedure in the interbedded sand layers are a better with thin layers less than 40 mm thick, an alternative
estimate of qt, but the layer interfaces are not well procedure is desired. The inverse approach proposed
defined. The performance of these procedures was by Boulanger and DeJong (2018) is attractive
similar for all soil profiles in this dataset that con because it is fully automated and incorporates an
tained layers that were 40 mm (or 1.6dcone) thick. actual description of the physics behind multiple
For the soil profiles in the dataset with interbed thin-layer effects (i.e., through the blurring model
ded layers less than 40 mm thick, the efficacy of all described by the convolution of qt with the spatial
procedures was poor, as exemplified in Figure 2b. filter wc). Building on this approach, we pose the
None of the procedures were successful at inverse problem in a new way by assuming that qt is
780
a piecewise constant function, and forcing guesses
of qt (i.e., qinv) to be a piecewise constant function.
Thus, the procedure searches for a finite number of
layers in a soil profile, each having a thickness and
This is not the only way to pose the optimization
constant qinv. This approach differs from the BD18
problem. For applications where both large-scale
approach, which solves for an independent qinv value
and fine-scale features contribute to the misfit,
at every depth, and then subsequently applies
a logarithmic misfit function can be more appropri
a procedure to impose a constant qinv within each
ate and is thus proposed as an alternative to
identified layer. Reducing the number of degrees of
Equation 3:
freedom in the problem and eliminating the interface
correction step results in a more computationally
efficient procedure.
The inverse problem is posed to minimize the
misfit function that describes the difference between
the actual measured tip resistance profile (qm) and
the simulated measured tip resistance profile (qm,sim),
which is created by applying an artificial blurring This procedure, including both forms of the misfit
filter to the qinv guess. We restrict qinv to be function, is detailed in Cooper et al. (2022) [Cea22]
a piecewise constant function defined by N layers, and is summarized in Figure 3.
each paired with a qinv value. Therefore, each pro In addition to posing the optimization problem, it
posed qinv profile is described by a material property is necessary to select a numerical optimization
vector, m, that has 2N components (i.e., thickness algorithm to iteratively update the qinv guess. Cooper
and qinv for N layers), where N can be adjusted et al. (2022) utilizes a Particle Swarm Optimization
throughout the optimization. For any assumed m, we (PSO) algorithm that identifies minima of the selected
can extract the qinv values represented by the piece misfit function. PSO was selected because it is able to
wise function at every depth of where CPT data test many widely varying guesses of m, overcoming
were measured. The qinv profile resulting the challenges often associated with global versus local
from this reconstruction process is denoted by qinv minima. Consequently, small adjustments to layer
(m). The qinv profile with the minimized misfit is thicknesses or assumed qinv only marginally affect qm –
likely to be a good estimate of the qt profile, but qm,sim. To optimize the PSO algorithm, two additional
numerical optimization algorithms may yield differ computational procedures are proposed. An add-one-in
ent answers depending on the choice of the misfit (AOI) algorithm is utilized to automatically add new
function. Written as an equation, this algorithm layers between existing layers to assess whether the
optimizes: addition of that layer reduces the misfit function of the
Figure 3. Cea22 inverse approach to correct for multiple-thin-layer effects in CPT tip resistance (modified from Cooper
et al. 2022).
781
proposed profile. A leave-one-out (LOO) algorithm is
utilized to remove insignificant layers from the guessed
values of qinv that are not physically realistic and con
tribute to unnecessary additional degrees of freedom.
A key component of this procedure is the applica
tion of an artificial blurring filter to the guessed
value of qinv. The Cea22 procedure adopts the same
framework to describe this blurring as proposed by
BD18 (i.e., Equation 1). However, they propose
a blurring filter that is a scaled and truncated chi-
squared distribution, selected for its asymmetry,
smoothness, and relatively good match with the de
Lange (2018) calibration chamber data:
Figure 4. Comparison between Cea22 inverse procedure

using the standard misfit function and the logarithmic func
tion on data from Soil Model 9 CPT 3 from de Lange
(2018) (modified from Cooper et al. 2022).
R∞ 4 DISCUSSION AND CONCLUSIONS

where -∞ pðzÞdz ¼ 1 and pðzÞ 2 0 for all z. In
practice, this integral is only calculated over a finite Multiple thin-layer effects in CPT data can be
interval. This blurring function was chosen because addressed using forward or inverse procedures. In
it is simple to implement and only requires the use this paper, a new forward method (DEL procedure)
of a matrix convolution function (“conv” in and a new inverse method (Cea22 procedure) are
MATLAB). This blurring function results in qm,sim compared to the existing inverse procedure proposed
values that represent a weighted combinations of by Boulanger and DeJong (2018) [BD18] using a set
the qt values of the surrounding soil layers at a given of calibration chamber data collected by de Lange
depth. Although this method can quickly compute (2018). It is shown that neither the DEL nor the
qm,sim for any qinv guess, it is a simplification of true BD18 procedure are effective in correcting for mul
physics and could be improved upon in future work. tiple thin-layer effects in profiles with soil layers less
This is exemplified by the difference between the than 40 mm thick, with the DEL procedure yielding
actual observed qm in the layered soil profile (shown slightly better results than the BD18 procedure.
in solid black) and the qm derived from applying the Because this type of profile is of particular interest
blurring filter to the known qt (shown in solid blue) for liquefaction assessment, it is desirable to improve
in Figure 4, and is discussed further in Section 4. on these procedures.
The Cea22 procedure was applied to the de Lange The Cea22 procedure poses the inverse problem
(2018) calibration chamber data using the qm derived that was first proposed by Boulanger and DeJong
from applying the blurring filter to qt (i.e., the blue (2018), but in a new way - searching for a finite
solid line in Figure 4), in lieu of using the actual number of soil layers each with a thickness and con
measured qm. It was found that the algorithm with stant qt. Application of this procedure indicates that
the logarithmic misfit function (i.e., Equation 4) was this new formulation is better able to identify thin,
more effective than the standard misfit function (i.e., interbedded layers than the DEL, BD18, or
Equation 3) at refining the very thin layers in the BD18MOD procedures. Furthermore, the restriction
calibration chamber test soil profiles; see Figure 4. on number of degrees of freedom is computationally
The qinv resulting from the standard misfit function efficient. It was shown that the logarithmic misfit
missed several thin layers in qt, while the qinv result function is more effective at identifying very thin
ing from the logarithmic misfit detects each of the layers than the standard misfit function. There are
thin layers in the layered zone. tradeoffs in using the standard versus logarithmic
Both algorithms (standard and logarithmic misfit) misfit function, namely, use of the logarithmic misfit
had qm,sim closely matching qm derived from appli function is significantly more computationally
cation of the assumed blurring filter to the known expensive (i.e., 5 to 10 minutes runtime for CPT
qt, indicating that the increase in computational soundings with several hundred data points, com
rigor provided by the logarithmic misfit algorithm pared to 1 to 2 minutes for the standard misfit func
was required to achieve the detailed match between tion). However, the use of the logarithmic misfit
qinv and qt for this profile. Results were similar for function is anticipated to be worth the extra compu
other profiles in the de Lange (2018) dataset and tational time for highly stratified soil profiles.
are examined in more detail in Cooper et al. Inverse procedures like the BD18 and Cea22 pro
(2022). cedure require an artificial “blurring” that mimics
782
multiple thin-layer effects in CPT data. Thus far, triggering calculations, it is critical that future work
simple blurring models have been adopted. For addresses multiple thin-layer effects on fs.
example, the Cea22 procedure uses a convolution of
a point spread function derived from the Chi squared REFERENCES
probability density function. As shown in Figure 4,
this method does not capture well the true complex Ahmadi, M.M. & Robertson, P.K. 2005. Thin-layer effects
ity of the physics involved with cone penetration on the CPT qc measurement. Canadian Geotechnical
through layered profiles and can be improved upon. Journal, 42(5): 1302–1317.
If the qt is known for a profile, a qm derived from Boulanger, R.W. & DeJong, J.T. 2018. Inverse filtering pro
applying the blurring filter to qt can be used as input cedure to correct cone penetration data for thin-layer
to the multiple-thin-layer correction procedure to and transition effects. Proc. of Cone Penetration Testing
2018, Hicks, Pisano, and Peuchen, eds., CRC Press,
assess the efficacy of the procedure itself, without Delft, The Netherlands: 25–44.
assessing the accuracy of the blurring filter. This is Cooper, J., Martin, E., Yost, K.M., Yerro-Colom, A., &
the approach taken in this paper to assess the Cea22 Green, R.A. 2022. Robust Identification and Character
procedure. Developing a more accurate blurring ization of Thin Soil Layers in Cone Penetration Data by
filter is the focus of ongoing work. Piecewise Layer Optimization. Computers and Geotech
A persistent challenge in developing and validat nics, 141:104404.
ing multiple-thin-layer correction procedures is the de Greef, J. & Lengkeek, H.J. 2018. Transition-and thin
lack of available qm and qt pairs for a given layered layer corrections for CPT based liquefaction analysis.
soil profile. These data can only come from calibra Proc. of Cone Penetration Testing 2018, Hicks, Pisano,
and Peuchen, eds., CRC Press, Delft, The Netherlands:
tion chamber tests (e.g., de Lange 2018) or from
317–322.
numerical simulations. Towards this end, numerical de Lange, D.A. 2018. CPT in Thinly Layered Soils. In: van
simulations of CPT in layered profiles, like those Elk, J., Doornhof, D., eds., Delft, The Netherlands.
from Yost et al. (2021b), can supplement the limited Yost, K.M., Green, R.A., Upadhyaya, S., Maurer, B.W.,
available calibration chamber data and be used to Yerro-Colom, A., Martin, E.R., & Cooper, J. 2021a.
develop and calibrate these methods. Numerical Assessment of the efficacies of correction procedures
simulations should be calibrated and validated with for multiple thin layer effects on Cone Penetration
laboratory calibration chamber data if possible. Tests. Soil Dynamics and Earthquake Engineering,
Finally, all the procedures discussed in this paper 144:106677.
Yost, K. M., Yerro, A., Green, R. A., Martin, E., &
focus on developing methods to correct CPT qc for
Cooper, J. 2021b. MPM Modeling of Cone Penetrom
multiple thin-layer effects. However, CPT fs is per eter Testing for Multiple Thin-Layer Effects in Complex
haps even more subject to multiple thin-layer effects Soil Stratigraphy. Journal of Geotechnical and Geoen
owing to the large size of the sleeve friction sensor vironmental Engineering. (in press)
(typically ~110 to 134 mm in length). As a result, Youd, T. L., Idriss, I. M., Andrus, R. D., Arango, I.,
the friction sleeve will likely be in contact with mul Castro, G., Christian, J. T., Dobry, R., Finn, W. D. L.,
tiple soil layers at once in a highly interlayered pro Harder, L. F., Hynes, M. E., Ishihara, K., Koester, J. P.,
file. Additionally, resolving the fs for a given layer is Liao, S. S. C., Marcuson, W. F., Martin, G. R.,
further hampered by soil from overlying layers being Mitchell, J. K., Moriwaki, Y., Power, M. S.,
Robertson, P. K., Seed, R. B., & Stokoe, K. H. 2001.
dragged down into underlying layers as the cone
Liquefaction Resistance of Soils: Summary Report from
advances, which has been observed both experimen the 1996 NCEER and 1998 NCEER/NSF Workshops on
tally (i.e., de Lange 2018) and numerically (i.e., Yost Evaluation of Liquefaction Resistance of Soils. Journal
et al. 2021b). Since fs is required to compute normal of Geotechnical and Geoenvironmental Engineering,
ized soil behavior type index (Ic) for liquefaction 127(10): 817–833.
783
© 2022 the Author(s), ISBN 978-1-032-31259-0
Spatial interpolation of consolidation property of clays from limited CPTU

dissipation data
Zening Zhao
Wei Duan
College of Civil Engineering, Taiyuan University of Technology, Taiyuan, China
Guojun Cai*
School of Civil Engineering, Anhui Jianzhu University, Hefei, China
Meng Wu
Anand J. Puppala
Songyu Liu
Surya Sarat Chandra Congress

ABSTRACT: The coefficient of consolidation (ch), an important soil property, is often estimated from the
piezocone penetration test (CPTU) data. However, the number of interpreted ch data is usually limited espe
cially in the horizontal direction because the CPTU is usually performed vertically and the dissipation test is
time-consuming. In the present study, a novel method is proposed for the interpolation of ch data in a 2D
vertical cross-section. Firstly, the original ch is interpolated by CPTU dissipation data. Then, the ch values are
interpreted based on the dissipation data at every location including the untested locations using random field
theory. Finally, real case is illustrated to prove the applicability of the proposed method. It is shown that the
proposed method can reasonably interpolate the ch at untested locations. Overall, the new method can spa
tially interpolate the ch from limited CPTU dissipation data and significantly save the test time.
1 INTRODUCTION used to determine ch in the field since it is fast, repeat

able, and economical (Cai et al., 2011; Lunne et al.,
The coefficient of consolidation (ch), an important geo- 1997; Robertson et al., 1992; Zhao et al., 2021b).
technical parameter that reflects the consolidation prop- The basic principle for estimating ch by CPTU is as
erty of clays, is often used to estimate the seepage follows: The excess pore pressure (Δu) can be gener
characteristics and consolidation settlement of clays ated around the cone during the penetration process.
(Burns and Mayne, 2002; Duan et al., 2018; Zhao The penetration process will be halted when
et al., 2021a). It can be interpreted by either laboratory- a predetermined depth is reached, then the dissipation
based oedometer tests or in-situ tests. Due to the draw- of the penetration-induced excess pore pressure around
backs of oedometer tests being time-consuming, high the cone will be recorded at a certain time interval and
cost, and disturbance of soil sample, determining ch by plotted as the excess pore pressure dissipation curve.
in-situ tests has received extensive attention (Krage As a result, the time corresponding to 50% dissipation
et al., 2015). Among various in-situ test methods, the of the excess pore pressure (t50) can be recorded and
piezocone penetration test (CPTU) has been widely used to interpret ch for lightly over-consolidated to
DOI: 10.1201/9781003308829-116
784
normally consolidated clays (Chu et al., 2002; Sully where T* is a modified dimensionless time factor,
et al., 1999). For heavily over-consolidated clays, the which is taken as 0.245 in this paper; t50 is the time
excess pore pressure rises initially, and then gradually corresponding to 50% dissipation of excess pore pres
dissipates until it becomes zero due to the dilatory sure (s); r is the radius of the CPTU cone (17.85mm).
response (Burns and Mayne, 2002; Lunne et al., 1997; The key to interpret ch is to determine the time t50
Wu et al., 2021; Zhao et al., 2021a). To address this corresponding to 50% dissipation of excess pore
problem, Chai et al. (2014) proposed an empirical pressure. For non-monotonic dissipation curves, the
equation to correct the value of t50 determined from time should be corrected. Chai et al. (2014) proposed
non-monotonic dissipation curves using an uncoupled an empirical equation to correct the value of t50
plane strain radial consolidation analysis. Then, the cor determined from non-monotonic dissipation curves
rected t50 was used in existing interpretation models. using an uncoupled plane strain radial consolidation
However, the CPTU dissipation test in super soft analysis. Then, the corrected time was used in Eq.
soils and marina soils with low permeability can be (1). The proposed empirical equation is as follows:
time-consuming. In these cases, a single cone sounding
can take over a day. Moreover, CPTU is only applic
able to one-dimensional (1D) analysis, i.e., along the
depth. In engineering practice, however, two- or three-
dimensional (2D or 3D) geotechnical analysis is often
performed in which 2D or 3D information on subsur
face soil stratification and zonation is needed (Chen where t50m is the corrected time for 50% excess pore
et al., 2016a; Duan et al., 2021; Wang et al., 2020). It pressure dissipation (s); tumax is the time elapsed for
is impractical to perform a large number of CPTU dis reaching the maximum measured excess pore pres
sipation tests over a whole site. Fortunately, random sure (s).
field theory-based spatial interpolation method can pro
vide 2D geotechnical analysis efficiently. For example, 3 RANDOM FIELD THEORY
Chen et al. (2016a) used random field theory to spa
tially assess liquefaction-induced settlements. The visu To perform 2D analysis over the entire area of inter
alization of the spatial map of ch can provide useful est, random field theory is introduced in this section.
information for guiding the design and construction of The main steps are as follows (Chen et al.,
ground or subgrade treatment (Zhao et al., 2021a). 2016b): 1) evaluation of ch at individual testing loca
Therefore, the development of a simple spatial map is tions; 2) statistical and spatial characterization of the
appealing and expected in engineering practice. index; 3) random field realization and Monte Carlo
In the present study, a novel method was proposed (MC) simulation to generate 2D ch maps.
for the interpolation of ch data in a 2D vertical cross-
section using random field theory. Firstly, the ori 3.1 Spatial correlation
ginal CPTU dissipation data was interpolated in the
2D vertical cross-section. Then, the ch values were The semivariogram is used to describe the spatial
interpretated based on the dissipation data at every correlation of the predicted PL value, which can be
location in the 2D vertical cross-section including obtained based on the half variance of two random
the untested locations. Finally, real CPTU dissipation variables separated by a distance h:
data was illustrated to prove the applicability of the
proposed method.
2 CPTU-BASED INTERPRETATION MODEL

where Z(u) is a Gaussian random variable at location
In the past few decades, many CPTU-based theoret u. There are several semivariogram theoretical
ical and empirical interpretation models were pro models, including the linear, spherical, exponential,
posed for the estimation of ch (Baligh and Levadoux, and Gaussian models.
1986; Burns and Mayne, 2002; Cai et al., 2011; Then, the semivariogram is related to the spatial
Krage et al., 2015; Robertson et al., 1992; Sully correlation ρ(h) by:
et al., 1999; Teh and Houlsby, 1991; Torstensson,
1977). One representative model proposed by Teh
and Houlsby (1991) is considered in this study.
This model considers the effect of rigidity index (Ir)
The vector distance h can account for both separ
based on large strain path analysis and finite difference
ation distance and orientation as:
analysis:
785
where hx, hy are the scalar components of the vector China. The geological formation in this area belongs to
h along the field’s principal axes; scalar quantities the floodplain geomorphic unit of the Yangtze River,
ax, ay specify how quickly spatial dependence and the depth of the groundwater level is 1.5m below
decreases along those axes. The ratio of ax/by =1 the ground surface. Figure 1 shows the profiles of the
represents the correlation decreases with distance CPTU site.
equally in all directions.
3.2 Sequential simulation process

To generate random field realizations of the variables
of interest, a conditional sequential Gaussian simula
tion method is implemented, which has been exten
sively used by mining scientists and geostatisticians for
natural resource evaluations and spatial prediction of
geohazards. It is worth noting that a multiscale exten
sion of this conditional sequential Gaussian simulation
method has been developed in recent studies (Wang
et al., 2017). The procedure can be illustrated by:
Figure 1. Layout of CPTU dissipation tests at the Yangtze

River tunnel of Jiangyin, Jingjiang City, Jiangsu Province,
China.
where Zn is the next realization to be simulated; Zp is
the vector or of all known and previously simulated
points. At this test site, a total of 4 CPTU boreholes were
Once Zn is simulated, it becomes a known data performed. The CPTU equipment is shown in
point in the vector of Zp to be conditioned upon by Figure 2. The dissipation test was performed using
all subsequent data locations. This process is a multifunctional, digital, and vehicle-mounted CPTU
repeated to other unknown points until all locations system produced by Vertek-Hogentogler (Randolph,
in the field have been simulated. OH), USA (Cai et al., 2011). The parammeters of the
Random field models incorporate the spatial CPTU probe are as follows: cone apex angle of 60°,
dependence of the measured parameter through the a cone bottom diameter of 35.7 mm, a cross-sectional
covariance matrix. The covariance of values at two area of 10 cm2, a sleeve surface area of 150 cm2, the
separated locations could be expressed as: pore pressure is measured at the u2 position, penetra
tion rate is 20 mm/s. During the penetration process,
the Δu can be measured at regular intervals. Then,
the dissipation of Δu over time was recorded until
the pore pressure reaches the equilibrium hydrostatic
pressure. After that, the penetration was continued
where ρZi ;Zj is the spatial correlation (calculated by further to the next desired depth and the above steps
Eq. (4)) between the random variables Zi and Zj with were repeated. Finally, a total of 14 dissipation tests
standard deviations of σZi and σZj , respectively. were performed in this site.
Once the empirical semivariogram γ(h) is character
ized, it will be plugged into the covariance matrix Eqs.
(4) and (7). Thus, the unknown value Zn at location
n could be drawn using Eq. (6). The generated value is
then assigned to location n and treated as known data.
This process is repeated until all the unsampled loca
tions are assigned with values. Detailed process of
random field modeling may be found in Chen et al.
(2016a, b).
4 ILLUSTRATIVE EXAMPLE
4.1 Interpretation of dissipation tests

The CPTU test site is located in the Yangtze River
tunnel of Jiangyin, Jingjiang City, Jiangsu Province, Figure 2. CPTU test equipment.
786
High-quality soil samples were also taken at dif Table 1. Typical soil properties in the CPTU soundings.
ferent depths, corresponding to the depths of dissipa
tion tests, using a stationary piston sampler. The Sounding γ (kN/m3) su (kPa) Es (MPa) OCR
diameter of the stationary piston sampler is 76 mm.
The soil sample at the end of the tube was extruded CPTU1 17.5 18.34 3.58 1.22
for waxing and sealing at both ends after withdraw CPTU2 17.6 17.85 3.26 1.21
ing the stationary piston sampler from the borehole. CPTU3 17.4 20.16 3.81 1.29
The collected soil samples were temporarily stored CPTU4 17.3 19.13 3.50 1.18
near the site before being transported back to the
laboratory. In this study, unconfined compression
tests and the Ko-consolidated undrained triaxial tests It can be noted from Figure 3 that the dissipation
(CKoU) were performed on undisturbed samples to curves are non-monotonic dissipation curves, that is,
estimate undrained shear strength (su) and shear the excess pore pressure rises initially and then grad
modulus (G) of the soils. Then, the Ir can be calcu ually dissipates until it is zero. The non-monotonic
lated by Ir = G/su. Other important soil parameters dissipation curves are mainly caused by (a) shear-
such as unit weight (γ) and overconsolidation ratio induced dilatancy of over-consolidated clays or dense
(OCR) were also obtained from laboratory tests, as sandy soils around the cone and (b) possible unload
listed in Table 1. ing effects of the soil elements moving from the vicin
It can be noted that the OCR is in the range of ity of the face to the shoulder of the cone, which can
1.18-1.29, indicating the clays at the test site are usu cause a large gradient in pore pressure going from the
ally lightly over-consolidated. cone to the shaft (Zhao et al., 2021a). Therefore, the
The key to interpret ch is to determine the time t50 ch should be interpreted using Eqs. (1) and (2). The
corresponding to 50% dissipation of excess pore results are listed in Table 2. It can be noted that the ch
pressure. The excess pore water dissipation curves varies significantly in the spatial, it is therefore to be
are shown in Figure 3. interpolated spatially using random field theory.
Figure 3. Dissipation curves at soundings (a) CPTU1; (b) CPTU2; (c) CPTU3; (d) CPTU4.
787
Table 2. Interpretation results of dissipation tests.
Sounding No. Depth (m) Ir tumax (s) t50 (s) t50m (s) ch (mm2/s)
CPTU1 1 5.4 160 107 513 73 13.577

2 8.5 90 98 392 58 12.754
3 13.6 115 340 900 98 8.522
4 18.2 95 1060 3560 470 1.618
CPTU2 5 4.9 190 195 860 111 9.685
6 8.4 120 245 775 93 9.199
7 14.1 160 79 1410 402 2.457
8 17.2 90 320 2200 440 1.683
CPTU3 9 6.9 80 53 260 41 16.015
10 8.8 110 55 420 75 9.670
11 13.6 120 53 707 182 4.593
CPTU4 12 8.4 125 120 2520 73 13.577
13 12.6 135 180 640 80 11.378
14 18.5 125 580 2390 331 2.637
Mean 1156 168 8.825
SD 943 151 4.657
COV 0.816 0.894 0.528
4.2 Spatial interpolation

The length of the area is 20 m, whereas the depth of
the area is 20 m. Of the 14 dissipation tests, 12 dissi
pation tests are used to develop the hazard map of
liquefaction-induced lateral spread based on random
field theory, while the other 2 dissipation tests (i.e.,
No. 6 and 11) are used for the validation.
As suggested by Chen et al. (2016a), the lognormal
distribution is used to fit the distribution of ch. Then,
the fitted lognormal distribution is adopted to charac
terize the semivariogram and then generate random
fields in the sequential Gaussian simulation. In the pre
sent study, the typical exponential semivariogram has
been used to fit the empirical semivariogram. The grid
size of the random field in this site is set as 0.1 m. A
parametric analysis of MC simulation number ranging Figure 4. Variations of COV of ch as the number of Monte
from 1,000 to 10,000 has been performed, as shown Carlo simulations increases.
in Figure 4. It is shown that 1,000 MC simulations are
enough as COV of the ch values tends to be stable
with the MCS number goes beyond 1,000.
The spatial map of ch is shown in Figure 5. The ch
can be characterized and visualized. The ch values at
dissipation tests No. 6 and 11 are about 11.586 mm2/s
and 6.325 mm2/s, respectively. It is evident that the ch
values at dissipation tests No. 6 and 11 interpolated
based on random field theory are similar to the valid
ation values, and can be viewed as the smoothed repre
sentation of the actual data. This proves the reliability
of the random field theory in spatial interpolation.
The proposed method can be used to spatially inter
polate any fields in the future. It can generate ch values
at untested sites and reflect the actual values in the
field, which are often not available in the engineering
practice. Therefore, the proposed method can signifi
cantly save the test time especially in super soft soils
and marine soils. Figure 5. Spatial map of mean values of ch in the testing site.
788
5 CONCLUSIONS Chu, J., Bo, M.W., Chang, M.F. & Choa V. 2002. Consoli
dation and permeability properties of singapore marine
In this study, the random field theory was incorpor clay. Journal of Geotechnical and Geoenvironmental
ated in the interpretation model of ch, which can be Engineering 128(9): 724–732.
used for areal interpolation and uncertainty quantifi Duan, W., Cai, G., Liu, S., Puppala, A.J. & Chen, R. 2018.
cation. It can provide the spatial map of ch over In-Situ Evaluation of Undrained Shear Strength from
a field using limited CPTU data. The spatial map can Seismic Piezocone Penetration Tests for Soft Marine
Clay in Jiangsu, China. Transportation Geotechnics
provide useful information on the need of project- 20:100253.
specific geotechnical investigation. Duan, W., Congress, S.S.C., Cai, G., Liu, S., Dong, X.,
The proposed method can interpolate ch values at Chen, R. & Liu, X. 2021. A hybrid GMDH neural net
untested sites and reflect the actual values in the work and logistic regression framework for state
field, which are often not available in the engineer parameter-based liquefaction evaluation. Canadian Geo-
ing practice. Therefore, the proposed method can technical Journal 58(12): 1801–1811.
significantly save the test time especially in super Krage, C.P., DeJong, J.T. & Schnaid, F. 2015. Estimation
soft soils and marine soils. of the coefficient of consolidation from incomplete cone
penetration test dissipation tests. Journal of Geotech
nical and Geoenvironmental Engineering 141(2):
ACKNOWLEDGEMENTS 06014016.
Lunne, T., Robertson, P.K. & Powell, J.J.M. 1997. Cone
The majority of the work presented in this paper was Penetration Testing in Geotechnical Practice. CRC
Press, London.
funded by the National Key R&D Program of China Robertson, P.K., Sully, J.P., Woeller, D.J., Lunne, T.,
(Grant No. 2020YFC1807200), the National Natural Powell, J.J.M. & Gillespie, D.G. 1992. Estimating coef
Science Foundation of China (Grant No. 41877231, ficient of consolidation from piezocone tests. Canadian
No. 42072299, and No. 52108332). The financial Geotechnical Journal 29(4): 539–550.
supports are greatly acknowledged. Sully, J.P., Robertson, P.K., Campanella, R.G. &
Woeller, D.J. 1999. An approach to evaluation of field
CPTU dissipation data in overconsolidated fine-grained
REFERENCES soils. Canadian Geotechnical Journal 36(2): 369–381.
Teh, C.I. & Houlsby, G.T. 1991. An analytical study of the
Baligh, M. M., & Levadoux, J. N. 1986. Consolidation cone penetration test in clay. Géotechnique 41(1):
after undrained piezocone penetration. II: Interpretation. 17–34.
Journal of Geotechnical Engineering, 112(7): 727–745. Torstensson, B.A. 1977. The pore pressure probe. Geote
Burns, S.E. & Mayne, P.W. 2002. Analytical cavity knikkdagen, Norway.
expansion-critical state model for piezocone dissipation Wang, C., Chen, Q., Shen, M., & Juang, C. H. 2017. On
in fine-grained soils. Soils and Foundations 42(2): the spatial variability of CPT-based geotechnical param
131–137. eters for regional liquefaction evaluation. Soil Dynamics
Cai, G., Liu, S. & Puppala, A.J. 2011. Predictions of coeffi and Earthquake Engineering 95: 153–166.
cient of consolidation from CPTU dissipation tests in Wang, Y., Hu, Y., & Zhao, T. 2020. CPT-based subsurface
Quaternary clays. Bulletin of Engineering Geology and soil classification and zonation in a 2d vertical
the Environment 71(2): 337–350. cross-section using Bayesian compressive sampling.
Chai, J.C., Julfikar Hossain, M., Carter, J. & Shen, S.L. Canadian Geotechnical Journal 57(7): 947–958.
2014. Cone penetration-induced pore pressure distribu Wu, M., Cai, G., Liu, L., Jiang, Z., Wang, C. & Sun, Z.
tion and dissipation. Computers and Geotechnics 57: 2021. Quantitative identification of cutoff wall construc
105–113. tion defects using Bayesian approach based on excess
Chen, Q., Wang, C., & Juang, C.H. 2016a. Probabilistic pore water pressure. Acta Geotechnica online.
and spatial assessment of liquefaction-induced settle Zhao, Z., Duan, W. & Cai, G. 2021a. A novel PSO-KELM
ments through multiscale random field models. Engin based soil liquefaction potential evaluation system using
eering Geology 211: 135–149. CPT and Vs measurements. Soil Dynamics and Earth
Chen, Q., Wang, C., & Juang, C.H. 2016b. CPT-based quake Engineering 150: 106930.
evaluation of liquefaction potential accounting for soil Zhao, Z., Congress, S.S.C., Cai, G. & Duan, W. 2021b.
spatial variability at multiple scales. Journal of Geotech Bayesian probabilistic characterization of consolidation
nical and Geoenvironmental Engineering 142(2): behavior of clays using CPTU data. Acta Geotechnica
04015077. 1–18.
789
Session 3: Applications
© 2022 the Author(s), ISBN 978-1-032-31259-0
Evaluation of the geotechnical behavior of mining tailings through CPTU

tests in the soil improvement process for the decharacterization of
upstream heightened dams
J.L. Albino
Universidade Federal de São Carlos – UFSCAR, Brazil
T.A.T. Souza J. & L.S. Machado

Universidade Federal de Viçosa – UFV, Brazil
ABSTRACT: In order to assess the pore pressure conditions, to verify the drained behavior or not of
a reservoir of a tailings dam raised upstream, and to observe the evolution of the resistence of the material,
CPTU surveys were carried out before and after the execution of works for the implementation of vertical
geodrains in the reservoir of this dam. After the comparative tests, it was possible to indicate the safety gains
of the structures resulting from the execution of the works.
Keywords: CPTU test, upstream tailing dam, pore pressure dissipation, geodrains, liquefaction
1 INTRODUCTION and the susceptibility to liquefaction. Tests such as

CPTU, for example, are able to assess the in situ
The decharacterization of dams heightened by conditions of the material, the development of effect
the upstream method is a legal obligation in the ive stress and pore pressure along the depth.
national territory (Brazil). This obligation is Therefore, CPTU tests were carried out prior to the
a reaction the last dam failures in Brazil, in elaboration of the project to de-characterize an
which the failures of the Fundão Dam in 2015 upstream heightened dam work located in the iron
and the B1 Dam in 2019 were mainly caused by quadrangle. Since the behavior of the material was
the liquefaction of the tailings. Mine tailings dis observed as susceptible to liquefaction, interventions
posed hydraulically in dams tend to have were carried out downstream of the dam in order to
a contractile behavior, mainly due to the initial turn the material initially with undrained behavior into
void rate and material saturation. To carry out drained behavior. These interventions were carried out
the decharacterization, the concern of liquefac with the use of vertical geodrains to accelerate the
tion is also present, and it is necessary to control densification process, and mainly, in the dissipation of
the possible triggers. pore pressure.
In this environment, companies must rethink not The vertical geodrains contains a HDPE (High
only their traditional approaches to mining oper Density Polyethylene) core forming small channels.
ations, but also the development of technologies and The core is wrapped in non-woven geotextile, which
studies of tailings deposal. acts as a filter. Thus, the water captured by the drain is
Laboratory evaluations allow testing materials in conducted through the small channels of the HDPE
a controlled manner, under the desired conditions. core to the surface of the land, where it is drained by
However, a limitation of these analyzes is the extrac the upper drainage layer, placed on the surface of the
tion of undisturbed samples, especially those in depth land.
and in regions of low resistance. This is because, in the After the interventions, a second battery of CPTU
sampler crimping process, the sample is disturbed, tests was carried out. Through these investigations, it
making the test results divergent from the field was possible to evidence the dissipation of the pore
condition. pressure in the dam and in the tailings, causing the
On the other hand, measures to assess the in situ material to behave in a drained way. Such findings
tailings behavior can significantly contribute to the were significant for the decharacterization of the
identification of the undrained behavior of the soil dam to be carried out safely.
DOI: 10.1201/9781003308829-117
793
ways of calculating these values are presented below,
equations (1) to (7):
Figure 1. Pore pressure of CPTU 2017 before construction.
1.1 CPTU test

Currently, one of the main forms of sounding for sim
plified identification of soil properties is the CPTU.
The characterization of intermediate sedimentary soils,
especially silty sands, non-plastic silts, silty clays and
fine-grained tailings are often complicated due to the
difficulty of sampling these materials. Thus, CPTU
tests can be used to determine the stratigraphic profiles, Figure 2. CPTU 2017 dissipation test (13.0 meters) before
construction.
in situ conditions of the material and to estimate the
material’s geotechnical parameters. [1, 2]
With the CPTU crimping, automatic information We can define the elements of the formulas writ
is collected on tip resistance (qc), resistance by lat ten above, with qt being the corrected true tip resist
eral friction (fs) and pore pressure at different points ance, σv being the total vertical stress at depth, σʹvo
of the cone (u1, u2 e u3). [3] the vertical effective stress (σʹvo = σvo – u0), fs
Because of the characteristic geometry of the cone, a lateral friction resistance, qc the penetration resist
the record is disturbed by atmospheric pressure and by ance of the tip, α being the ratio of the cone areas
the difference in pore pressure value collected at differ (AN / AT), u2 is the measured pore pressure, u0 is the
ent points (u1, u2 e u3). Thus, it is necessary to correct equilibrium pore pressure.
the values obtained when carrying out the drilling
according to the boundary conditions. Therefore, the
main values that can be generated at first with 1.2 Assessment of soil behavior
the CPTU are the standardized resistance of the tip Qtn, From the parameters obtained above, along with the
the normalized friction ratio Fr, the friction ratio with CPTU drilling depth, several geotechnical character
Rf and the parameter of pore pressure as Bq. [1, 4]. The istics of the soil can be identified.
794
Figure 3. CPTU 2017 Cone tip resistance before
construction. Figure 4. Pressure pore of CPTU 2020 after construction.
With the CPTU test, it is possible to evaluate, For this study, the pore pressure acting on the tail
through correlations, the following soil characteristics: ings will be evaluated in order to identify its reduc
stratigraphy, geotechnical profile, density coefficient tion after the implementation of vertical geodrains in
(Ch e Cv), relative density (Dr), undrained resistance a dam’s reservoir. In addition, the behavior of the
(Su), sand effective friction angle (φʹ), stress history material regarding the dissipation test will also be
(pre-consolidation stress, OCR), permeability coeffi evaluated in order to assess the change from
cient (k), maximum shear modulus (Go), deformability undrained to drained behavior of the tailings.
coefficient (mv), modulus of deformability or Young’s
(E), effective friction angle (φʹ), confined or oedometric
modulus of deformation (M), sensitivity (St), perme 1.3 Dissipation test
ability coefficients (kh and kv) and evaluation of suscep
tibility to liquefaction of a soil. [2, 8] The dissipation test consists of interrupting CPTU
Another very important point is that with the penetration and observing the pore pressure over time.
records of cone tip resistance, lateral friction resist Data is logged and must occur to a minimum dis
ance and pore pressure, the CPTU interpretation is sipation of 50%.
applied for the stratigraphic determination of the Pressures are plotted as a function of the square
soil based on the behavior type graphs of the root of time. The graphical technique suggested by
soil (SBTn). This proposal was presented by Robertson and Campanella (1989) provides a value
Robertson (1990) where plots of two abacuses for t50, which corresponds to the time to consolida
are made (Qt × Fr (%) e Qt × Bq), where the Qt is tion of 50% [9].
the normalized of cone tip resistance (8) [5, 6, 7]: The value of the consolidation coefficient in the
radial or horizontal direction Ch was then calculated
by Houlsby and Teh’s (1988) theory using the fol
lowing equation (9) [10]:
795
Furthermore, with the performance of the dissipa
tion tests, it was observed that at the deepest point of
dissipation (13.0 meters) a permeability value was
found in the house of 2.56 x 10-9 m/s, classifying it
in this way as waterproof material. The dissipation
test will be shown below, Figure 2.
Finally, another point observed is due to the
resistance of the material collected with the execu
tion of the CPTU. For this first moment, low cone
tip resistance was observed up to 6.5 m deep. The
result will be shown below, Figure 3.
With the implementation of vertical geodrains in
order to reduce the pore pressures acting in the crit
ical region of the dam and provide the drained
behavior and increase the local resistance of the
material, new tests were carried out to verify what
was foreseen in the project, the reduction of the pore
pressure and the guarantee of the behavior drained
by the tailings.
After performing the pore pressure tests, it was
initially identified a considerable variation of the
Figure 5. CPTU 2020 dissipation test (15.5 meters) after acting pore pressure up to 6.5 meters in depth. In
the works.
Where T is the time factor given by the theory of

Houlsby and Teh (1988) corresponding to the pore
pressure, r piezocone radius, Ir is the stiffness index,
equal to the shear modulus G divided by the undrained
strength of the clay (Su) and t50 the time corresponding
to 50% consolidation.
Furthermore, the pore pressure dissipation during
a CPTU dissipation test is controlled by the consoli
dation coefficient in the horizontal direction (Ch)
which is influenced by a combination of soil perme
ability (kh) and compressibility (M), like defined by
the following equation (10):
Where M is the restricted modulus and γw is the

unit weight of water, in compatible units.
2 RESULTS
Initially, with the performance of CPTU tests in

2017, before the geodrain implantation works, the
presence of pore pressure was observed at a critical
depth for the structure’s safety assessments. In this
condition, it was noticed that up to a depth of
6.5 m of drilling depth, there were pore pressures
close to 230 kPa, thus indicating that the tailings
deposited in these quotas presented an undrained
behavior. Below we have Figure 1 with the represen
tation of the result of pore pressure observed. Figure 6. CPTU 2017 Cone tip resistance after the works.
796
Figure 8. Comparison between the 2017 and 2020 CPTU
in the critical region of cone tip strength.
Figure 7. Comparison between the 2017 and 2020 CPTU
in the critical region of pore pressure.
3 CONCLUSION
this condition, the behavior of the tailings pre
sented an average pore pressure of 50 kPa and in As can be seen throughout the results presented,
some points a reduction of up to 200 kPa of pres before the works were carried out, there were
sure was observed. This situation can be seen in conditions of high pore pressures in the initial
Figure 4 below. portion of the test. With the execution of
In addition to the gains observed during the the works, there was a considerable reduction in
pore pressure measurement, considerable gains the critical pore pressure condition, improvement
were also observed during the tailing’s dissipation in the tailing’s drainage capacity an increase in
test. As can be seen in Figure 5, below, with the the cone tip resistance of the CPTU in this
dissipation at the deepest point in the CPTU probe, material.
15.5 meters, it was identified that the permeability A comparative graph of the pore pressure reduc
of the material increased to the house of 1.53 x 10 tion in the critical region will be presented below,
7 Figure 7.
, thus improving the drainage capacity of the tail
ings around one hundred times. In this condition, Finally, as already mentioned, for the critical
the tailings started to be classified as low perme layer, up to 6.5 m, a considerable increase in mater
ability soil. ial strength was also observed. A comparative graph
Finally, another point is due to the resistance gain of the increase in cone tip strength in the critical
observed for the material up to 6.5 m, where the region will be presented below, Figure 8.
cone tip resistance initially found values close to 0.0 Therefore, carrying out the CPTU test was of
kPa and after the execution of the works, peaks close great value in proving critical conditions in the field
to 15.0 kPa were observed. This graph of the new and in verifying the effectiveness of the work
CPTU assay is shown in Figure 6, below. performed.
797
REFERENCES [6] Jamiolkowski M.; Lo Presti D.C.F.; Manassero M.
Evaluation of relative density and shear strength of
[1] Lunne, T.; Robertson, P.K.; Powell J.J.M. Cone pene sands from CPT and DMT. Soil Behavior and Soft
tration testing in geotechnical practice. Blackie Aca Ground Construction, 2003, 7(119), p. 201–238.
demic, EF SPON/ Routledge Publishing, 1997. [7] Kim C.; Kim S.; Lee J. Estimating clay undrained
[2] Campos, R. V. A. Análise de investigações geotécnicas shear strength using CPTu results. Proceedings of the
de barragens utilizando o CPTU em uma mina de Institution of Civil Engineers Geotechnical Engineer
ferro. Undergraduate Thesis (Degree in Mining Engin ing. 2009, 162(2), p. 119–127.
eering). Universidade Federal de Ouro Preto. Escola de [8] Robertson P.K. Soil classification using the cone pene
Minas. Núcleo de Geotecnia, Ouro Preto, MG, 2021. tration test. Canadian Geotechnical Journal, 1990, 27
[3] Rainer, J.; Hubert S. Analysis of nderground stratifica (1), p. 151–159.
tion based on CPTu profiles using high-pass spatial [9] Campanella, R. G.; Robertson, P. K.; Davies, M. P.;
filter. Studia Geotechnica et Mechanica, vol. 42, no 4, Sy, A. Use of in-situ tests in pile design. Proceedings
2020, p. 355–65. 12th International Conference on Soil Mechanics and
[4] Schnaid, F.; Odebrecht, E. Ensaios de campo e suas Foundation Engineering, ICSMFE, Rio de Janeiro,
aplicações à engenharia de fundações. 2. ed., Oficina Brazil, 1989, vol. 1, p. 199–203.
de Textos, São Paulo, SP, 2012. [10] Houlsby G. T.; Teh C. I. Analysis of the piezocone
[5] Robertson P.K., Caval K.L. Guide to cone penetration tests in clay. De Ruiter J (ed.), Penetration Testing,
testing for geotechnical engineering. 4th Edition. Proc. 1st Isopt, Orlando, Balkema, Rotterdam, 1988,
Gredd Drilling & Testing, Inc., California, 2010. vol. 2, p. 777–783.
798
© 2022 the Author(s), ISBN 978-1-032-31259-0
VCPT: An in-situ soil investigation method to validate vibratory pile-soil

interaction models
D. Al-Sammarraie, S. Kreiter & T. Mörz
MARUM–Center for Marine Environmental Sciences, University of Bremen, Germany
M.O. Kluger
School of Science/Te Aka Mātuatua, University of Waikato, Hamilton, New Zealand
M. Goodarzi
COWI – Hamburg, Germany
ABSTRACT: Until now the vibratory pile driving method is not widely used because there is no reliable
approach for drivability analysis. One of the main issues with the current drivability analysis methods are the
inability to accurately predict the cyclic soil behavior during vibratory pile driving. In this research, the cyclic
soil behavior was evaluated using a new in-situ soil investigation method, the vibratory cone penetration test
(VCPT). VCPT penetrates the ground, while inducing controlled vertical cyclic strains, and measuring cone
resistance, sleeve friction, and pore water pressure. Nine static CPTs and 15 VCPTs were performed at
a constant frequency with three different amplitudes. The resulted cone resistance-displacement cycles were
compared with the state-of-the-art soil-pile interaction models. The reduction in soil strength against vibratory
loading was found to be amplitude-dependent; and it was observed that a cavity formed between the cone and
soil during the upward movement of the cone for high displacement amplitudes.
1 INTRODUCTION driving on cyclic soil behavior (Jonker, 1987, Holey-

man and Whenham, 2017).
The notion of using vibratory pile driving to install The vibratory cone penetration test (VCPT) is an
the foundations for offshore wind turbines is increas in-situ soil investigation method, which penetrates
ing, because this driving technique produces less the ground while inducing cyclic loads (Sasaki and
noise, causes less damage to the pile, and has Koga, 1982). The first VCPT tool utilized horizontal
a higher installation speed than impact pile driving vibration as it was developed to investigate the lique
technique (Holeyman and Whenham, 2017). The faction potential due to earthquake loading. More
main problem of this technique is the lack of a well- recently, VCPTs induced vertical (down-hole) vibra
established method for drivability analysis. The cur tions (Wise et al., 1999, McGillivray et al., 2000,
rent drivability analysis of vibratory pile driving Mayne, 2000). The vertical cyclic motion of VCPT
derives from soil-pile interaction models modified may be considered to resemble the motion of vibra
from impact pile drive analyses (Jonker, 1987, Wong tory pile driving (Al-Sammarraie et al., 2018, Stähler
et al., 1992, van Baars, 2004, Viking, 2006, Holey- et al., 2018, Viking, 2006). The measured cone resist
man and Whenham, 2017, Lee et al., 2012). The ance and displacement at the tip of the VCPT allow
vibratory pile driving models utilize empirical for evaluating the stress-displacement cycles of load
parameters, having been derived from conventional ing and unloading of the vibrational penetration (Al-
(quasi static) soil investigation methods, such as Sammarraie et al., 2018). From these cycles, the stiff
CPT and SPT, to predict the cyclic soil response ness during each individual loading and unloading
during vibratory pile driving (Wong et al., 1992). cycle may be obtained (Wong et al., 1992, Lee et al.,
The load application of the conventional soil investi 2012). When cyclic penetration exceeds a certain
gation methods differs from that of vibratory pile amplitude threshold, the resistance at the toe
driving. Therefore, the current soil-pile interaction decreases to zero during the upward movement and
models are likely unable to assess (1) degradation the begin of the downward movement of the pene
processes along the pile shaft and pile toe and (2) the trating object. This loss in resistance is called cavita
effect of different amplitudes of vibratory pile tion and is described in terms of a “cavity” even
DOI: 10.1201/9781003308829-118
799
though it is unlikely that a total empty space is by the external displacement sensor. The displacement
formed (Massarsch and Westerberg, 1996, Rodger amplitude of VCPT, measured from the displacement
and Littlejohn, 1980, Viking, 2006). The degree of sensor, is referred to as “VCPT amplitude” hereafter.
cavitation during vibratory pile driving is defined as The actual displacement of the cone was obtained
the length of upward and downward movement of from a three-axis acceleration sensor, which was
the penetrating object without soil contact (Massarsch located in the cone. This acceleration sensor had
and Westerberg, 1996, Massarsch et al., 2017, Vogel- a sampling rate of 500 Hz.
sang et al., 2017, Dierssen, 1994). The cavitation and
the loading and unloading stiffness during vibratory
2.2 Geological and geotechnical setting
penetration have recently been investigated by utiliz
ing the so-called Vibro-Penetration Test VPT, which The study area is located in Cuxhaven, Northern Ger
is a soil investigation test that uses a vertical har many. The soil deposits in the area consist mainly of
monic excitation force to drive a rod with a conical very dense sand. The stratigraphy and the physical
tip into the ground (Cudmani and Manthey, 2019). properties of the sand deposits were determined from
The VPT has a fixed relationship between frequency a drill core in the vicinity of the test field (Figure 2).
and potential maximum force and is furthermore The stratigraphic characterization followed (Sin
influenced by the soil reaction while VCPT allows dowski, 1965, Ehlers et al., 1984, Geo-Engineering,
for independent variation of penetration speed, fre 2014) (Figure 3), grain size analyses were carried out
quency and amplitude. Until now, the VCPT method according to DIN 18123 (2011), and the unit weight
was not applied to study the loading and unloading was provided from (Geo-Engineering, 2014).
stiffness as well as the cavitation that may occur The deposits are of Pleistocene age and consist of
during vibratory pile driving. The influence of the a 4-m-thick Middle Saalian unit composed of fine- to
displacement amplitude on the soil resistance and the medium-grained sands, and an underlying 7-m-thick
loading and unloading stiffness is also still unknown. Older Saalian unit composed of compacted hetero
In this study, the influence of displacement ampli genic stratified sand deposits, with a till layer in
tudes on the degradation of cone resistance was inves between (Figure 3). The 60-cm-thick Drenthe till
tigated. It was explored to what extent the stiffness layer, composed of grains from clay to gravel size.
during loading and unloading cycles and the cavitation The soil behavior types were calculated following
changed with displacement amplitudes. Finally, the Robertson (2009) from CPT S7 (Figure 2). Most of
obtained cyclic soil resistance were utilized to param the sand deposits are soil behavior type 6 (sand like).
etrize the current pile-soil interaction models devel Only the till layer is soil behavior type 3 and 4 (clay
oped by Dierssen (1994), Wong et al. (1992), and like, silt-like). The sand deposits have relative dens
Jonker (1987), the results are quantified and discussed. ities between dense to very dense following the CPT
correlation of Baldi et al. (1986).
2 METHODOLOGY
2.1 VCPT device

The VCPT comprises a 100 kN hydraulic cylinder,
a clamp, a valve unit, a displacement sensor, and
a real time controller (Figure 1).
The hydraulic cylinder was used to push the rods
and the cone into the ground having been controlled
Figure 2. Maps of the layout of test points and the location

of the core. The coordinate system used to create the maps
was UTM zone N32.
2.3 Experimental layout

A cone with an area of 5 cm2 was used to perform
nine static CPTs (SCPTs hereafter) and 15 vibratory
CPTs (VCPTs hereafter) in a systematic grid. In
Figure 1. Vibratory cone penetration test device. Figure 2, the letter “S” refers to SCPTs and “V”
800
refers to VCPT. The spacing between tests was layers at small horizontal distance, all SCPT and
chosen to be 1.4 m, being equivalent to 55 times VCPT datasets were correlated and shifted to
cone diameter. The spacing was chosen to be larger a common depth. The cone resistance dataset of
than the minimum recommended distance between SCPT S7 was selected to be the “reference dataset”
CPTs (Al-Sammarraie et al., 2020, DIN EN ISO and all other SCPT and VCPT datasets were shifted
22476-1, 2012, BS 1377-9, 1990). The SCPTs were to it (Figure 2) (Al-Sammarraie, 2020). After the
performed before the VCPTs. depth shift, the depths of the SCPT and peak VCPT
The VCPTs were performed at a constant fre datasets were resampled to a common vertical depth
quency of f=20 Hz and at three different VCPT amp scale with an increment of 0.5 mm.
litudes of 3, 5, and 7 mm (Figure 2). The CPT
locations were determined by a DGPS system with 2.4.2 Cyclic behavior analysis
a horizontal resolution of ±2 cm. A theodolite with In this step, the loading stiffness, Kb , and unloading
an accuracy of ±0.3 cm was used to ensure stiffness, Ke , and the upward displacement with cavita
a common height reference at the start and the end tion, dcav , were quantified. Since the acceleration and
of each CPT. cone resistance data were affected by noise, The cycles
of cone resistance cycles, vqc;cy , and raw acceleration,
a, were stacked and averaged in order to reduce the
noise level (Al-Sammarraie et al., 2018). Every second
cycle of the correlated and shifted datasets of vibratory
cone resistance and raw acceleration was selected and
processed until the final depth of penetration of 12 m,
while avoiding averaging between different of individ
ual push-strokes. All data points of the selected cycle
were added to the equivalent points of the neighboring
cycles above and below and then divided by the
number of cycles. This approach is similar to
a moving average where the points are overlaid in
phase. For cone resistance cycles and raw acceleration
cycles each point is the average of 41 points. The
moving average windows corresponded to a vertical
length of 4 cm. The stacking and averaging process
resulted in stacked cone resistance cycles, vqc;cy;st , and
stacked acceleration, ast .
Figure 3. Stratigraphy and soil properties of test site (a)

core; (b) grain size distribution (Geo-Engineering, 2014);
(c) bulk unit weight (Geo-Engineering, 2014); (d) soil
be-havior types (SBT); (e) friction ratio; (f) corrected tip
resistance; and (g) relative density derived after Baldi et al.
(1986).
2.4 Evaluation of SCPTs and VCPTs datasets

The data of SCPT and VCPT were evaluated follow
ing four steps: (1) correlation, (2) cyclic behavior
analyses, (3) statistics, and (4) degradation factors.
Detailed explanation of the four steps is included in
Al-Sammarraie (2020).
2.4.1 Correlation
The peak values of each individual cycle of the
entire VCPT datasets were selected. The resulting
VCPT datasets were defined as peak vibratory cone
resistance, vqc;p , peak sleeve friction, vfs;p , and peak
Figure 4. Fit function to determine loading stiffness,
pore water pressure, vup . The vibratory cone resist unloading stiffness and upward displacement with
ance datasets that included the full information about cavitation.
each cycle is hereafter referred to as cone resistance
cycles, vqc;cy .
In order to minimize the effect of local heteroge The stacked acceleration data, ast was integrated
neities and changes in the depth of specific soil twice to obtain the cyclic displacement, dcy . The
801
constants of integration – depth and average vel zero represent high reduction, whereas values close to
ocity – were obtained from the displacement sensor. one represent small reduction in the representative
The cone resistance-displacement cycles were used cone resistances and sleeve frictions due to VCPT.
to determine the loading and unloading stiffnesses as
well as the upward displacement with cavitation
2.5 Pile-soil interaction models
(Figure 4). This was done by using a fit function that
consists of four connected linear sections to fit the
cone resistance-displacement cycle (Figure 4). The 2.5.1 Karlsruhe model
loading stiffness Kb was then calculated from the In the Karlsruhe model, four steps of the cyclic pene
slope of the second section (from point 1 to 2) and the tration are identified to describe the cyclic soil resist
unloading stiffness Ke from the third section (from ance-displacement behavior during vibratory
point 2 to 3). The upward displacement with cavita penetration (Figure 5) (Dierssen, 1994, Cudmani and
tion, d cav , was defined as the length of the fourth sec Manthey, 2019). The straight line between points 1
tion (from point 3 to 4. These steps resulted in and 2 is expressed by:
datasets of loading stiffness, Kb , unloading stiffness,
Ke , and upward displacement with cavitation, d cav .
2.4.3 Statistics
The objective of the statistics step was to calculate Where qT is the resistance force at the toe; Kb is
representative datasets for different tests in the sys the loading stiffness of the soil:
tematic grid (Figure 1). The datasets of the SCPT When the pile moves upward the straight line
and VCPT that had the same VCPT amplitude, were between points 2 and 3 is expressed by:
averaged using Eq 1. The datasets of the loading/
unloading stiffness and the cavitation were also aver
aged with respect to their corresponding VCPT
using Eq 1.
Where qT is the resistance force at the toe; qT;2 is
the resistance at the toe at point 2; Ke is the unload
ing stiffness of the soil.
Where Xr is the representative dataset; Xi is an

individual dataset; n is the total number of CPTs
with the same amplitude; and i running from 1 to n.
2.4.4 Degradation factors

The degradation factors of representative cone resist
ance and sleeve friction was determined from Eqs. 2
and 3, in order to investigate whether or not the
changes in VCPT amplitude affected the cone resist-
ance and sleeve friction.
Figure 5. Description of the four phases of cyclic penetra

tion (Dierssen, 1994).
Where βqc is the degradation factor of representa

tive cone resistance (Jonker, 1987); vqc;p;r is the rep When cyclic amplitude is large enough the resist
resentative peak vibratory cone resistance; and qc;r is ance at the pile toe will reach zero and a cavity is
the representative static cone resistance. formed, represented by the straight line between
points 3 and 4 (Figure 5). At step between points 4
and 1ʹ, the pile moves downward while the stress at
the toe remains zero because of the cavity.
The loading, Kb , and unloading stiffness, Ke , are
empirical parameters, however in this study, they
were obtained from the cone resistance-displacement
Where βfs is the degradation factor of representative cycle and were used as input for the fit (Figure 4).
sleeve friction; vfs;p;r is the representative peak vibra For step between points 4 and 1ʹ, the loading stiff
tory sleeve friction; and fs;r is the representative static ness was restricted between values of 5-10 MPa/m,
sleeve friction. Degradation factors with values near in order to account for the effect of cavity.
802
2.5.2 The model of Wong et al.
VCPT amplitude of 5 mm indicating total loss in
The cyclic penetration in this model is modeled with
shear stress between the sleeve and soil (Figure 7d).
two equations in two parts one for the loading and
The degradation factors of cone resistance
one for the unloading part (Figure 6).
decreased with increasing VCPT amplitude in the
The first part is the loading part, which is sand deposits (Figure 7c). In the silty till layer how
described by (Wong et al., 1992): ever, the degradation factors of cone resistance were
close to one, indicating no effect of VCPT. Theβqc
determined for different amplitudes were compared to
the β-factor obtained from Jonker (1987) (Figure 7c).
The degradation factor proposed by Jonker is 0.25 for
sand and 0.4 for clay (the latter being considered to
have similar soil properties to the silty till layer) and
it is not related to changes of amplitudes.
The degradation in sleeve friction was highly
And the unloading part is described in Eq.7: affected by the VCPT amplitude in the sand deposits
(Figure 7e). The silty till layer exhibited no degrad
ation in sleeve friction during VCPT.
The representative loading and unloading stiffness
values and the upward displacement with cavitation
were plotted along the depth (Figure 8). In general,
both loading and unloading stiffness values
decreased with VCPT amplitude (Figure 8c-d). All
stiffness values were very low in the silty till layer.
The upward displacement with cavitation for ampli
tudes 5 and 7 mm exhibited approximately similar
values (Figures 8e). The upward displacement with
cavitation at a lower amplitude of 3 mm were lower.
Figure 6. Loading and unloading parts and the explanation

of depths of model of Wong (Wong et al., 1992): (a) Load
ing (b) Unloading.
Where, ql and qu are the loading and unloading

resistance at the pile toe respectively; qc is the unit
toe resistance at the previous reversal; Kb and Ke are
the loading and unloading stiffness respectively; Figure 7. Degradation factors of representative cone resist
w, wp , and wc , are the displacements as noted in ance and sleeve friction for different VCPT amplitudes.
Figure 6; q0 is a reference value of toe resistance; Fq
is the fitting factor; m is the constant which controls
the magnitude of the amplitude.
The parameters Kb , Ke , q0 , and nq used in this
model were obtained by fitting the model to the cone
resistance-displacement cycles obtained from VCPT
(Figure 4).
3 RESULTS
The peak vibratory cone resistances of the sand

deposits were nearly always below the static cone
resistance (Figure 7b). The reduction in the peak
vibratory cone resistances to the static cone resist
ance increased with VCPT amplitude. Figure 8. Loading and unloading stiffnesses and the
During VCPT, the sleeve friction was reduced upward displacement with cavitation plotted along the
with increasing amplitude until it reaches 0 MPa at depth of penetration.
803
The Individual cone resistance-displacement cycles soil resistance during vibratory pile driving depends
measured by VCPT at different depths were plotted and on the amplitude of vibration. This conclusion
compared with the corresponding Karlsruhe and Wong agrees with previous studies about assessment of
fits (Figure 9). The results of Karlsruhe model agree vibratory-drivability in sand, in which it was
with the cone resistance-displacement cycles for 5, and assumed that one of the important factors affecting
7 mm amplitudes at depth of 8 and 10 m. For the 3 mm degradation factor for piles, β, is vibration amplitude
amplitude, the Karlsruhe model showed limitations to (Jonker, 1987, van Baars, 2004).
fully capture loading part of the resistance-displacement The degradation in cone resistance during
cycle. On the other hand the Wong model seemed to VCPT may be related to the cavitation (Dierssen,
fully match the cycles at 3 mm amplitude, however 1994). During the upward movement of the cone,
larger difficulties were found when trying to to capture the unloaded soil may lose its strength, collapses
the cyclic behaviour at 5 and 7 mm amplitudes. and starts to flow into the cavity, which might be
filled with liquified material. However, because
there is only a short time interval until the rever
sal to the downward movement, the liquified
material might have still not fully solidified
(Dierssens, 1994). This process may result in
remolding and/or loosening of the soil below the
cone (Massarsch et al., 2017). Therefore,
a reduction in cone resistance occurs when the
cone gets again in contact with the soil during the
downward movement.
The increase in degradation of loading and
unloading stiffness with increasing amplitudes
(Figure 8c-d), agrees with the degradation of the
secant Young’s modulus with increasing strain in
triaxial tests on Ticino Sand (Yamashita et al.,
2000).
The dependency of the observed cyclic behav
iour on the cyclic amplitude, raises questions
regarding the fruitfulness of current soil-pile inter
action models to try to estimate soil cyclic resist
ance from the static cone resistance (Holeyman
and Whenham, 2017a). Nevertheless, Wong et al.
(1992) modelled the cyclic soil behaviour by
including cyclic degradation of the pile-soil inter
face, and the loading and unloading phases during
vibratory driving. However, the model utilizes
empirically implemented parameters for very spe
cific soil-pile systems, with no account for the
effect of the cyclic amplitude (Eqs. 6 and 7)
(Wong et al., 1992, Lee et al., 2012). Further
more, the model does not account for the cavita
tion part of the cone resistance-displacement
cycles which reliably occurs at higher displace
ment amplitudes (Figure 8e). This explains the
inability of the model to capture the cone resist
ance-displacement cycles with amplitudes of 5
and 7 mm (Figure 9). The Karlsruhe model on
the other hand accounts for cavitation part of the
vibratory cycles during pile driving. Therefore, it
could well match the cone resistance-displacement
Figure 9. Stacked cyclic cone resistance – displacement cycles with higher amplitudes of 5 and 7 mm
behavior for the three applied amplitudes and the correspond (Figure 9). Although the measured cone resist
ing Karlsruhe and Wong fits at depths of a: 8 m; b: 10 m.
ance-displacement cycle differed slightly in the
shape with rounder edges compared to Karlsruhe
4 DISCUSSION model; the ability of the Karlsruhe model to cap
ture the cyclic soil response to the different
The increase in degradation of cone resistance and values of the applied cyclic amplitudes (Figure 9),
sleeve friction with VCPT amplitudes (Figure 7), proved the effectiveness of utilizing the cyclic
provided practical evidence that the degradation in soil parameters obtained from VCPT.
804
5 CONCLUSION Technology Conference, 1987. Offshore Technology
Conference.
In this study, VCPTs were performed to investigate Lee, S.-H., Kim, B.-I. & Han, J.-T. 2012. Prediction of
the cyclic soil behavior in-situ. It was found that: penetration rate of sheet pile installed in sand by vibra
tory pile driver. KSCE Journal of Civil Engineering, 16,
1 The degradation in vibratory cone resistance 316–324.
and sleeve friction increased with VCPT ampli Massarsch, K. R., Fellenius, B. H. & Bodare, A. 2017. Fun
tudes. The degradation is suspected to be damentals of the vibratory driving of piles and sheet
related to increase in cavitation. piles. geotechnik, 40, 126–141.
2 The degradation in the loading and unloading Massarsch, K. R. & Westerberg, E. 1996. FREQUENCY
stiffness increased with increasing VCPT VARIABLE VIBRATORS AND THEIR APPLICA
TION TO FOUNDATON ENGINEERING. Korea Geo-
amplitudes.
technical Society, 1996, 25–40.
3 The parameters of the cyclic soil resistance Mayne, P. W. 2000. Evaluating ground liquefaction poten
obtained from VCPT could successfully be tial by piezoviercone. Georgia Institute of Technology.
used in the current pile-soil interaction models. Mcgillivray, A., Casey, T., Mayne, P. W. & Schneider, J. A.
2000. An electro-vibrocone for site-specific evaluation
of soil liquefaction potential. Innovations and Applica
REFERENCES tions in Geotechnical Site Characterization.
Robertson, P., K., 2009. Interpertation of Cone Penetration
Al-Sammarraie, D. 2020. Vibratory cone penetration test to Tests - A Unified Approach. Can. Geotech. J., 46,
investigate cyclic soil behavior in-situ. PhD, University 1337–1355.
of Bremen. Rodger, A. & Littlejohn, G. 1980. A study of vibratory
Al-Sammarraie, D., Kreiter, S., Kluger, M. O. & Mörz, T. driving in granular soils. Geotechnique, 30, 269–293.
2020. Reliability of CPT measurements in sand–influ Sasaki, Y. & Koga, Y. Vibratory cone penetrometer to
ence of spacing. Géotechnique, 1–13. assess the liquefaction potential of the ground. Proceed
Al-Sammarraie, D., Kreiter, S., Stähler, F. T., Goodarzi, M. ings, 1982. 541–555.
& Mörz, T. New Vibratory Cone Penetration Device for Sindowski, K.-H. 1965. Die drenthestadiale Altenwalder
in-Situ Measurement of Cyclic Softening. International Stauchmoräne südlich Cuxhaven. Zeitschrift der
Symposium on Cone Penetration Testing, 2018. Deutschen Gesellschaft für Geowissenschaften (ZDGG),
Baldi, G., Bellotti, R., Ghionna, V., Jamiolkowski, M. & 158–162.
Pasqualini, E. Interpretation of CPT’s and CPTU’s. 2nd Stähler, F., T., Kreiter, S., Goodarzi, M., Al-
Part: Drained Penetration” Proceeding 4th International Sammarraie, D. & Mörz, T. Liquefaction Resistance by
Geotechnical Seminar, 1986 Singapore. 143–15. Static and Vibratory Cone Penetration Tests. In:
BS 1377-9 1990. Methods for test for soils for civil engin Hicks, M., A., Pisanò, F. & Peuchen, J., eds. Proceed
eering purposes. In-situ tests ings of the 4th International Symposium on Cone Pene
Cudmani, R. & Manthey, S. 2019. A novel tration, 2018 Delft.
vibro-penetration test (VPT) for the investigation of Van Baars, S. 2004. Design of sheet pile installation by
cohesionless soils in the field. Soil Dynamics and Earth vibration. Geotechnical & Geological Engineering, 22,
quake Engineering, 126, 105760. 391–400.
Dierssen, G. 1994. Ein bodenmechanisches Modell zur Bes Viking, K. The vibratory pile installation technique. Pro
chreibung des Vibrations-rammens in körningen Böden. ceedings of the International Conference on Vibratory
Doctoral Thesis, University of Karlsruhe. Pile Driving and Deep Soil Compaction, 2006. 65–82.
DIN 18123 2011. Soil, investigation and testing–determin Vogelsang, J., Huber, G. & Triantafyllidis, T. 2017. Experi
ation of grain-size distribution. mental investigation of vibratory pile driving in satur
DIN EN ISO 22476-1 2012. Geotechnical investigation ated sand. Holistic Simulation of Geotechnical
and testing - Field testing Part 1: Electrical cone and Installation Processes. Springer.
piezocone penetration test. Wise, C., Mayne, P. & Schneider, J. Prototype piezovibro
Ehlers, J., Meyer, K.-D. & Stephan, H.-J. 1984. The cone for evaluating soil liquefaction susceptibility.
pre-Weichselian glaciations of north-west Europe. Qua Earthquake geotechnical engineering, 1999. 537–542.
ternary Science Reviews, 3, 1–40. Wong, D., O’neill, M. W. & Vipulanandan, C. 1992. Mod
Geo-Engineering 2014. Geotechnical site investigation elling of vibratory pile driving in sand. International
report VIBRO-project Altenwalde. Journal for numerical and analytical methods in geome
Holeyman, A. & Whenham, V. 2017. Critical review of the chanics, 16, 189–210.
Hypervib1 model to assess pile vibro-drivability. Geo- Yamashita, S., Jamiolkowski, M. & Presti, D. C. L. 2000.
technical and Geological Engineering, 35, 1933–1951. Stiffness nonlinearity of three sands. Journal of Geo-
Jonker, G. Vibratory pile driving hammers for pile installa technical and Geoenvironmental Engineering, 126,
tions and soil improvement projects. Offshore 929–938
805
© 2022 the Author(s), ISBN 978-1-032-31259-0
Monitoring ground improvement by Rammed Aggregate Piers using

a combined CPTU and SDMT approach at a silty sand liquefaction-prone
site in Emilia-Romagna
S. Amoroso
University of Chieti-Pescara, Pescara, Italy
M.F. García Martínez, L. Tonni & G. Gottardi

P. Monaco
K.M. Rollins
Brigham Young University, Provo, Utah, USA
L. Minarelli
D. Marchetti
K.J. Wissmann
Geopier Foundation Company, Davidson, North Carolina, USA
ABSTRACT: Following the 2012 Emilia-Romagna earthquake, widespread liquefaction of silty sands was
observed, providing the opportunity to gain a better understanding of the influence of fines content on lique
faction hazard and mitigation works. This paper presents the results of a thorough geotechnical investigation
performed as part of a full-scale liquefaction experiment involving controlled blast tests in Bondeno, a small
village that suffered liquefaction in 2012. Piezocone (CPTU) and seismic dilatometer (SDMT) tests were per
formed in natural and improved soils after Rammed Aggregate Pier® (RAP) treatment to provide accurate soil
characterization and to evaluate the effectiveness of liquefaction mitigation. CPTU and SDMT results
revealed a good agreement in the geotechnical characterization of the site, detecting homogenous soil proper
ties in both the natural and treated soils and estimating CPTU-DMT coupled parameters in sandy layers (e.g.
overconsolidation ratio, at-rest earth pressure coefficient), that are usually not determinable by the use of
a single type of in situ test. In particular, the combined use of CPTU-DMT data provided verification of the
increase in the lateral stress produced by the RAP installation. Data analyses revealed that the RAPs were an
effective ground improvement technique despite the high percent of fines (≈ 25-35%).
1 INTRODUCTION by extensive research (e.g. Mitchell 1981, Vautherin

et al. 2017, Amoroso et al. 2018). However, their
Several ground improvement solutions are available effectiveness decreases as the fines content and plasti
to mitigate the liquefaction hazard posed by clean city increase (Mitchell 1981). Therefore, other
sands; namely, increasing the soil resistance by densi ground improvement techniques, such as vibratory
fication or reducing the earthquake-induced excess replacement, are often preferred in silty sands or
pore pressures through drainage or reducing the shear sandy silts to protect the soil against liquefaction by
strains through reinforcement. Vibratory compaction increasing soil density, providing drainage for
methods are a common and effective form of densifi excess pore water pressures, and increasing the stiff
cation for cohesionless soils (Castro 1969), as proven ness and shear resistance of the soil (Priebe 1998).
DOI: 10.1201/9781003308829-119
806
Figure 1. (a) Location of the Bondeno Test Site and of 2012 main shocks; (b) geomorphological features from LIDAR map
(modified after Amoroso et al. 2020): greenish color indicates lower elevation above the sea level, while brownish color
refers to higher elevation.
Examples of this type of reinforcement include Stone and after Rammed Aggregate Pier (RAP) treatment of
Columns (SC), Soil Mixed Columns (SMC), and a silty sand site in Bondeno (Italy), a small village
Rammed Aggregate Piers (RAP). This last approach strongly affected by liquefaction following the 2012
appears to be a promising solution in sandy silts and Emilia-Romagna earthquake. The overall details of the
silty sands to increase not only the density, but also research activities can be found in Amoroso et al.
the lateral stress and shear stiffness, which is often (2020) while details regarding the performance of the
neglected and poorly understood (Smith & Wissmann RAP group following a blast test are provided by Roll
2018, Amoroso et al. 2020). ins et al. (2021) and regarding the liquefaction assess
The at-rest earth pressure coefficient (K0) is a key ment and ground improvement are listed in Amoroso
parameter that should be considered with reference et al. (2022).
to liquefaction mitigation works (Schmertmann
1985, Salgado et al. 1997, Harada et al. 2010).
In this respect, in situ tests have an essential role 2 THE BONDENO TEST SITE (BTS)
to play in estimating the horizontal stress in granular
soils before and after treatment. As argued by Mas 2.1 Geological and geomorphological setting
sarsch et al. (2019), using cone penetration test
(CPT) and flat dilatometer test (DMT) results could The Bondeno Test Site (BTS) is located in the south
produce improved estimates of K0. Moreover, Baldi eastern portion of the Quaternary alluvial Po Plain, one
et al. (1986) and later Hossain & Andrus (2016) pro of the largest and most populous plains in Europe. The
posed a combined CPT-DMT K0-interpretation to area was affected in 2012 by an intense seismic activ
take into account both the resistance and stress his ity linked to the tectonic evolution of the fault-fold
tory of the soil, while the use of a CPT-only structures (Figure 1a) that form the front of the Apen
approach would have been overly affected by arch nine chain buried below the plain (e.g. Toscani et al.
ing of stresses around the penetrating sleeve. 2009).
The coupling of CPT and DMT tests with down- The earthquake sequence induced widespread site
hole geophysics (i.e. seismic piezocone SCPTU and effects, including liquefaction manifestations, soil
seismic dilatometer SDMT) provides a more effi fracturing and lateral spreading (Emergeo Working
cient approach to the task of geotechnical site char Group 2013). At BTS the liquefaction hazard is con
acterization, offering clear opportunities for the centrated in a subsurface sand deposit of a Holocene
economical and optimal collection of the data Po meander (Figure 1b). Figure 1b shows the higher
(Mayne et al. 2009). Therefore, direct push technolo elevations (brownish zones) indicating fluvial ridges
gies are more relevant for understanding the changes bounding the lower, relatively flat interfluvial depres
in soil properties following ground improvement sion (greenish zones). The meandering course of the
(e.g. Jendeby 1992, Amoroso et al. 2018). paleochannel is built up within the interfluvial depres
This investigation presents in situ test results from sion and supports the identification of the paleochan
a thorough geotechnical campaign performed before nel axis together with the location of sand boils. The
807
Figure 2. CPTU and SDMT interpreted results in natural (NS) and treated (TS) soils: (a) qt, Ic, DR, φ’ from CPTU; (b) ID,
KD, M, VS from SDMT (modified after Amoroso et al. 2022).
meander base is frequently cut into upper Pleistocene diameter of 0.5 m (area replacement ratio equal to
coarse sand, accumulated during syn-glacial times. 5%). Details on the construction methodology are
The meander unit geometry has been reconstructed reported in Saftner et al. (2018). Each CPTU,
through the analysis of remote sensing data (satellite Medusa DMT (automated dilatometer test,
images and LIDAR) and correlation of underground Marchetti et al. 2019) and SDMT test was per
geotechnical investigations (Amoroso et al. 2020). formed up to a maximum depth of 15 m at the
This meander sand body is partially buried by finer exact center of four RAPs.
grained levee sediment of historic age.
Further details on the site investigations per
2.2 Site investigations formed within the full-scale liquefaction experiment
through controlled blast tests are reported in Amor
To assess the effectiveness of the RAP treatment at oso et al. (2022).
the BTS, in situ tests were performed before and
after pier installations, according to the phases
reported below: 3 SITE CHARACTERIZATION
• Phase I consisted of site investigations per
formed before the treatment (pre-RAP) and 3.1 Natural soil
before the blast (pre-blast). Boreholes with SPTs The stratigraphic arrangement of the subsoil beneath
and disturbed soil sampling, CPTUs, and the test site area was deduced by the combined inter
SDMTs were executed up to a maximum depth pretation of borehole logs, SPTs, CPTUs, Medusa
of 20 m in two relatively small circular areas DMTs and SDMTs carried out before the RAP
(10 m-diameter and 20 m-spacing) associated installation, as reported by Amoroso et al. (2020).
with the blast experiment, one for testing the Apart from a 0.8 m thick topsoil layer (CH, accord
natural soil (Natural Panel, NP) and one for test ing to Unified Soil Classification System, USCS
ing the improved soil (Improved Panel, IP). This ASTM D2487-11 2011), the following well-defined
paper includes only CPTUs and SDMTs per stratigraphic units, also reflecting their sedimentolo
formed in the IP; gical framework, could be identified:
• Phase II included site investigations carried out
approximately one month after the pier installa • a layer of clays and silts (CL), from 0.8 to
tion (post-RAP) and before the blast (pre-blast) approximately 3.3-3.4 m in depth;
within the IP. The treatment consisted of a 4 × 4 • a predominantly silty sand unit, approximately
quadrangular grid (2 m center-to-center spacing) 9 m thick, attributable to Holocene alluvial
of RAP columns, each 9.5 m long and with a final deposits of a Po river paleochannel. Samples
808
Figure 3. CPTU-DMT combined interpreted results in natural (NS) and treated (TS) soils: ID, M/qt, OCR, K0, ψ (modified
after Amoroso et al. 2022).
recovered from this unit can generally be classi appears to be particularly noticeable (qt = 13.10 ±
fied as SM, having a FC typically in the range of 1.76 MPa versus 9.54 ± 1.37 MPa before installation)
25-35%. Thin layers of coarser sediments have from 6 to 8.5 m in depth, but relatively moderate
been occasionally found; from 3.5 to 6 m. Negligible changes in the qt profile
• a thin layer of sandy silt (ML), from 11.8-12.6 to can be observed in the silty sands below the base of
13.0-13.4 m in depth (interfluvial deposits); the piers. Obviously, these changes in qt affect the
• sands-silty sands (SP-SM) of the late Pleistocene computed estimates of the geotechnical parameters
epoch (namely, glacial braided Po River reported in Figure 2, i.e. DR and ’.
deposits), detected below 13.0-13.4 m in depth. The effect of RAP installation is evidently reflected
by the increase in KD (on average 48-53%), and even
Figure 2 provides plots of representative piezo more in the higher M from SDMT (80-87%), at depths
cone and seismic dilatometer profiles carried out in between 4 and 9 m (Figure 2b). The corresponding
the pre-RAP phase and located in the treated panel, average increase in qt is 30-35% (and in DR is approxi
respectively CPTU01 and SDMT01 (black lines). In mately 10% and is limited to a depth between 6 and
particular, the CPTU profiles are shown in terms of 8.5 m). These results point to a significant increase in
the corrected cone resistance (qt), soil behavior type horizontal stress and stiffness resulting from pier instal
index (Ic, Robertson 2009), relative density (DR, lation, in agreement with previous observations (Saftner
Jamilokowski et al. 2001), and friction angle (’0 , et al. 2018). In fact, the horizontal stress strongly influ
Kulhawy & Mayne 1990), while the SDMT results ences both KD and M estimated from DMT using the
are reported in terms of the soil material index (ID), Marchetti (1980) correlation, which incorporates KD.
horizontal stress index (KD), constrained modulus The observed results are in line with previous
(M), and shear wave velocity VS), according to comparisons of pre- vs. post- CPTs and DMTs exe
Marchetti (1980) and Marchetti et al. (2001)
cuted for monitoring ground improvement (e.g. Jen
correlations. deby 1992), since the RAP installation produced an
The comparative analysis of the CPTU and average increase in M from DMT after treatment
SDMT parameters reveals a substantial agreement approximately 2.5 times the corresponding increase
between the measurements collected in the IP, thus in cone penetration resistance qc. The decrease in
indicating negligible horizontal spatial variability in
KD observed in the upper crust may be due in part
the stratigraphic conditions of the test site. to the construction of an overlying working plat
form, but also to RAP installation under low confin
3.2 Improved soil ing stress and to seasonal variations in water
content caused by fluctuation of the GWT from
Figure 2 also provides a comparison between field soil 1.5 m (February 2018) to 0.5 m (March 2018). No
properties before and after RAP installation in the IP, in
improvement was detected in the silty sands below
terms of both CPTU and SDMT profiles (green lines). the toe of piers, unlike RAP case histories in clean
With regard to the piezocone profiles (Figure 2), sands studied in New Zealand (e.g. Vautherin et al.
the increase in the qt values after column construction 2017).
809
The combined interpretation of CPTU and DMT CPTU and SDMT tests revealed good agreement
data provided information on stress history and the in the geotechnical characterization of the site,
state parameter in sand, in both the natural and treated detecting homogenous soil properties in both the nat
soils as shown in Figure 3. Filtering the data for ID ≥ ural and improved panels. Use of both CPTU and
1.8 and Ic ≤ 2.6, in the sand layers the ratio M/qt (with DMT provided better estimates of soil properties in
M estimated from DMT) is shown in Figure 3. The sandy layers (e.g. OCR, K0), that are usually not
average values of M/qt are about 7-10 in natural soil determinable using a single type of in situ tests.
and 13-14 in treated soil. These values are in line with The comparison of the in situ tests performed pre
the available experience from field observations before blast in natural and treated soils highlighted the
and after compaction of sand fills, reported by March effectiveness of the RAP treatment between 4 and
etti et al. (2001) and Marchetti & Monaco (2018), 9 m depth in silty sands. The increase in the DMT
which show an increase in the ratio M from DMT/qc parameters following treatment were more pro
from ≈ 5-10 before compaction to ≈ 12-24 after com nounced relative to those obtained from CPTU data
paction. The finding that compaction increases both (i.e. KD increase ≈ 48-53%, M increase ≈ 80-87%, qt
M from DMT and qc, but M at a faster rate, suggested increase ≈ 30-35%), thus suggesting a higher sensi
the potential use of the ratio M from DMT/qc, as tivity of DMT to the increase of horizontal stress.
a broad indicator of “equivalent” OCR in sands. On the contrary, VS measurements showed a very
The in situ earth pressure coefficient K0 was esti low sensitivity to the ground improvement More
mated using correlations proposed by Baldi et al. over, the combined use of CPTU and DMT tests
(1986), based on both DMT and CPT data, and by Hos showed a significant increase of M/qt and K0 after
sain & Andrus (2016), which require as an additional treatment, supporting the use of the piers to increase
input also OCR (in this case evaluated by Monaco the lateral soil stress and to mitigate liquefaction.
et al. 2014). In the upper silty clay layer OCR and K0
were estimated from DMT (Marchetti 1980).
The OCRs of about 1-2 estimated in the natural soil, ACKNOWLEDGEMENTS
excluding the shallow “crust”, indicate that the deposit
is normally consolidated or slightly overconsolidated, The study was primarily funded by Geopier® Founda
with K0 ≈ 0.5-0.7. As a result of the RAP installation, tion Company (Davidson, North Carolina, United
the “equivalent” OCR increased to about 3-3.5 and K0 States). A special thanks also to Releo s.r.l. (Ferrara,
to about 0.9-1. The values of K0 estimated according to Italy) who provided the installation of the Rammed
Hossain & Andrus (2016) are lower than those esti Aggregate Piers. The in-situ testing campaign was car
mated according to Baldi et al. (1986). The increase of ried out by CIRI Edilizia e Costruzioni, University of
M/qt, OCR and K0 after treatment was more pro Bologna, Italy under the research project TIRISICO
nounced at depths between 7 and 9 m. (“Tecnologie Innovative per la riduzione del rischio
An approximate estimate of the in situ state param sismico delle Costruzioni”, Project no. PG/2015/
eter ψ in sand from DMT was obtained according to 737636, POR-FESR 2014-2020). Financial contribu
Yu (2004), with K0 determined by both Baldi et al. tions to this research activity were provided by INGV
(1986) and Hossain and Andrus (2016) methods. FIRB Abruzzo project (“Indagini ad alta risoluzione
Figure 3 shows that the input K0 has a large influence per la stima della pericolosità e del rischio sismico
on the calculated values of ψ, with an apparent contra nelle aree colpite dal terremoto del 6 aprile 2009”), by
diction versus the expected trend. In fact, the higher K0 INGV-Abruzzo Region project (“Indagini di geologia,
(i.e. higher OCR) estimated according to Baldi et al. sismologia e geodesia per la mitigazione del rischio
(1986) should involve lower negative values of ψ com sismico”, L.R. n. 37/2016), and by Alma Mater Stu
pared to those obtained using K0 from Hossain & diorum – Università di Bologna within AlmaIdea
Andrus (2016), while the opposite is observed in research project (2017, Scient. Resp. Laura Tonni).
Figure 3. On the other hand, the reduction of ψ after Special thanks to Brigham Young University for con
treatment found using both K0 methods is consistent tributing to the realization of the blast test experiment
with the corresponding increase of OCR and K0 before in terms of personnel and technical equipment; to Prof.
and after treatment. Marco Stefani (University of Ferrara, Italy) for kindly
A more complete overview on the ground improve sharing scientific information of the studied area; to
ment effectiveness at BTS using in situ tests is reported Michele Perboni who kindly guested the experimental
in Amoroso et al. (2022). activities; to the Bondeno Municipality and to the
Emilia-Romagna Region (Luca Martelli), who pro
vided all the necessary support to realize the research
4 CONCLUSIONS in collaboration with the other local authorities.
At BTS a comprehensive comparative study based

on CPTU and SDMT testing was carried out in REFERENCES
a liquefaction-prone silty sand site improved by Amoroso, S., Rollins, K.M., Andersen, P., Gottardi, G.,
a group of Rammed Aggregate Piers and subjected Tonni, L., García Martínez, M.F., Wissmann, K.J.,
to controlled blasting.
810
Minarelli, L., Comina, C., Fontana, D., De Martini, P.M., SDMT in Practice. Geotechnical Testing Journal 41 (5):
Monaco, P., Pesci, A., Sapia, V., Vassallo, M., 837–850. https://doi.org/10.1520/GTJ20170386.
Anzidei, M., Carpena, A., Cinti, F., Civico, R., Coco, I., Marchetti, S., Monaco, P., Totani, G., & Calabrese, M.
Conforti, D., Doumaz, F., Giannattasio, F., Di 2001. The Flat Dilatometer Test (DMT) in Soil Investi
Giulio, G., Foti, S., Loddo, F., Lugli, S., Manuel, M.R., gations – A Report by the ISSMGE Committee TC16.
Marchetti, D., Mariotti, M., Materni, V., Metcalfe, B., Proc. 2nd Int. Conf. on the Flat Dilatometer, 7–48.
Milana, G., Pantosti, D., Pesce, A., Salocchi, A.C., Massarsch. K.R., Wersäll, C., & Fellenius, B.H. 2019.
Smedile, A., Stefani, M., Tarabusi, G., & Teza, G. 2020. Horizontal stress increase induced by deep vibratory
Blast-induced liquefaction in silty sands for full-scale compaction. Proceedings of Institution of Civil Engin
testing of ground improvement methods: insights from eers-Geotechnical Engineering 173(3): 228–253.
a multidisciplinary study. Engineering Geolology 265: Mayne, P.W., Coop, M.R., Springman, S.M., Huang, A.B.,
105437. https://doi.org/10.1016/j.enggeo.2019.105437. & Zornberg, J.G. 2009. Geomaterial behavior and
Amoroso, S., Garcia Martinez, M.F., Monaco, P., testing. Proc. 17th Int. Conf. on Soil Mechanics and
Tonni, L., Gottardi, G., Rollins, K.M., Minarelli, L., Geotechnical Engineering 4: 2777–2872.
Marchetti, D., & Wissmann, K.J. 2022. Comparative Mitchell, J.K. 1981. Soil improvement: state-of-the-art.
study of CPTU and SDMT in liquefaction prone silty Proc. 10th Int. Conf. on Soil Mechanics and Foundation
sands with ground improvement. Journal of Geotech Engineering 4: 509–565.
nical and Geoenvironmental Engineering, accepted on Monaco, P., Amoroso, S., Marchetti, S., Marchetti, D.,
January 25th 2022, https://doi.org/10.1061/(ASCE) Totani, G., Cola, S., & Simonini, P. 2014. Overconsoli
GT.1943-5606.0002801 dation and stiffness of Venice lagoon sands and silts
Amoroso, S., Rollins, K.M., Monaco, P., Holtrigter, M., & from SDMT and CPTU. Journal of Geotechnical and
Thorp, A. 2018. Monitoring ground improvement using Geoenvironmental Engineering 140 (1): 215–227.
the seismic dilatometer in Christchurch, New Zealand. https://doi.org/10.1061/(ASCE)GT.1943-5606.0000965.
Geotechnical Testing Journal 41 (5): 946–966. https:// Priebe, H.J. 1998. Vibro replacement to prevent earthquake
doi.org/10.1520/GTJ20170376. induced liquefaction. Ground Engineering 31(9): 30–33.
ASTM D2487-11. 2011. Standard practice for classification UK: Emap Construct.
of soils for engineering purposes (Unified Soil Classifi Robertson, P.K. 2009. Interpretation of cone penetration
cation System). West Conshohocken, PA: ASTM tests – a unified approach. Canadian Geotechnical Jour
International. nal 46 (11): 1337–1355.
Baldi, G., Bellotti, R., Ghionna, V., Jamiolkowski, M., Rollins, K.M., Amoroso, S., Andersen, P., Tonni, L., &
Marchetti, S., & Pasqualini, E. 1986. Flat Dilatometer Wissmann, K.J. 2021. Liquefaction mitigation of silty
Tests in Calibration Chambers. Proc., Specialty Conf. on sands using rammed aggregate piers based on
Use of In Situ Tests in Geotechnical Engineering GSP 6, blast-induced liquefaction testing. Journal of Geotech
431–446. Reston, VA: ASCE. nical and Geoenvironmental Engineering 147 (9)
Castro, G. 1969. “Liquefaction of sands.” Ph.D. Disserta https://doi.org/10.1061/(ASCE)GT.1943-5606.0002563
tion, Harvard University. Saftner, D.A., Zheng, J., Green, R.A., Hryciw, R., &
Emergeo Working Group. 2013. “Liquefaction phenomena Wissmann, K.J. 2018. Rammed aggregate pier installation
associated with the Emilia earthquake sequence of effect on soil properties. Proceedings of Institution of Civil
May-June 2012 (Northern Italy).” Nat. Hazards Earth Engineers-Geotechnical Engineering 171 (2): 63–73.
Syst. Sci., 13 (4), 935–947. Salgado, R., Boulanger, R.W., & Mitchell, J.K. 1997. Lat
Harada, K., Orense, R.P., Ishihara, K., & Mukai, J. 2010. “Lat eral stress effect on CPT liquefaction resistance
eral stress effects on liquefaction resistance correlations.” correlations. Journal of Geotechnical and Geoenviron
Bull. New Zeal. Soc. Earthq. Eng., 43(1),13–23. mental Engineering 123(8),: 726–735.
Hossain, M.A., and Andrus, R.D. 2016. “At-rest lateral Schmertmann, J.H. 1985. Measure and use of the in situ lat
stress coefficient in sands from common field methods.” eral stress. The Practice of Foundation Engineering,
J. Geotech. Geoenviron. Eng. 142(12): 06016016. A Volume Honoring Jorj O. Osterberg, 189–213.
https://doi.org/10.1061/(ASCE)GT.1943-5606.0001560. Smith, M.E., and Wissmann, K.J. 2018. Ground improve
Jamiolkowski, M., Lo Presti, D.C.F., & Manassero, M. 2001. ment reinforcement mechanisms determined for the Mw
“Evaluation of relative density and shear strength of sands 7.8 Muisne, Ecuador, earthquake. Proc. 5th Geotech
from cone penetration test and flat dilatometer test.” In nical Earthquake Engineering and Soil Dynamics Con
Proc. Symp. on Soil Behaviour and Soft ground Construc ference: Liquefaction Triggering, Consequences, and
tion GSP 119, 201–238. Reston, Virginia: ASCE. Mitigation, 286–294. Washington, DC: ASCE.
Jendeby, L. 1992. Deep Compaction by Vibrowing. I Proc. Toscani, G., Burrato, P., Di Bucci, D., Seno, S., &
Nordic Geotechnical Meeting 1: 19–24. Lyngby (Den Valensise, G. 2009. Plio-Quaternary tectonic evolution of
mark): Danish Geotechnical Society. the northern Apennines thrust fronts (Bologna-Ferrara sec
Kulhawy, F.H., & Mayne, P.H. 1990. Manual on estimating tion, Italy): seismotectonic implications. Italian Journal of
soil properties for foundation design. Report No. EL Geosciences 128: 605–613.
6800. Palo Alto, CA: Electric Power Research Institute Vautherin, E., Lambert, C., Barry-Macaulay, D., &
(EPRI). Smith, M. 2017. Performance of rammed aggregate piers
Marchetti, D., Monaco, P., Amoroso, S., & Minarelli, L., as a soil densification method in sandy and silty soils:
2019. In situ tests by Medusa DMT. Proc. XVII Eur. experience from the Christchurch rebuild. Proc. 3rd Int.
Conf. on Soil Mechanics and Geotechnical Engineering, Conf. on Performance-based Design in Earthquake Geo-
https://doi.org/10.32075/17ECSMGE-2019-0657. technical Engineering. London, UK: ISSMGE.
Marchetti, S. 1980. In situ tests by flat dilatometer. Journal Yu, H.S. 2004. James K. Mitchell Lecture – In situ soil test
of Geotechnical Engineering Division 106 (3): 299–321. ing: from mechanics to interpretation. Proc. 2nd Int.
Marchetti, S., & Monaco, P. 2018. Recent improvements in Conf. on Site Characterization 1: 3–38. London, UK:
the use, interpretation, and applications of DMT and Taylor & Francis Group.
811
© 2022 the Author(s), ISBN 978-1-032-31259-0
A CPT-based method for monotonic loading of large diameter monopiles

in sand
S. Bascunan
Ramboll, Hamburg, Germany
K. Kaltekis
B. van Dijk
Arcadis, Amersfoort, The Netherlands
K.G. Gavin
ABSTRACT: A joint academia-industry project, the Pile Soil Analysis (PISA) project, resulted in an empir
ical method for assessing the monotonic lateral loading response of large diameter monopiles. The method
predicts four soil reactions, namely the distributed load and the distributed moment along the pile shaft, the
pile base shear and the pile base moment. The method considers pile load test data and 3D numerical model-
ling. A 1D framework allows prediction of the four soil reactions. In this paper, a CPT-based approach is
proposed to derive the four soil reaction components for use in a 1D model for conceptual design of mono-
piles in sand subject to monotonic lateral loading. The approach relies on results from 3D finite element ana
lyses that were performed considering soil conditions for a sand site used in the PISA project (Dunkirk site).
The results are compared to pile load test data from the PISA project, showing good agreement, particularly
for load levels related to the serviceability limit state.
1 INTRODUCTION A joint academia-industry project, the Pile Soil

Analysis (PISA) project, resulted in an empirical
Monopiles are commonly used as foundations for off method for assessing the monotonic lateral loading
shore wind turbine generators (WTGs). The current response of large diameter monopiles. The method is
trend in the ever-growing offshore wind energy sector based on conventional models for caisson design, pre
is for WTGs to becoming bigger which evidently dicting four soil reactions, namely the distributed load
leads to requirements for monopiles with large diam and the distributed moment along the pile shaft, the
eters up to 10 m to support the superstructure. It is pile base shear and the pile base moment (Figure 1).
expected that the ratio of embedded length to diam The PISA schematisation excludes torsional foundation
eter, L/D (or slenderness ratio) of monopile founda loading (Burd et al., 2020). The empirical method con
tions for the 10 MW+ next-generation wind turbines siders pile load test (PLT) data and 3D numerical mod
could be in the range between 2 and 6 (Panagoulias elling. A 1D framework allows prediction of the four
et al., 2018). Such structures are categorised as inter soil reactions and requires, for sands, profiling of three
mediate foundations according to ISO (2016). soil parameters, namely the relative density, the verti
An industry standard approach for assessing mono- cal effective stress and the shear modulus at small
pile lateral response was a p-y method for long slender strain.
piles, adjusted to large diameter monopiles. The In this paper, a CPT-based approach is proposed to
p-y method is based on the Winkler assumption accord derive the four soil reaction components for use in
ing to which the soil surrounding the pile is modelled a 1D model for conceptual design of monopiles in
as a set of uncoupled, non-linear, elastoplastic springs sand subject to monotonic lateral loading. The
which define the lateral pressure (p) applied to the pile approach relies on results from 3D finite element (FE)
at a given depth, as a function of the lateral displace analyses that were performed considering soil condi
ment (y). The method, however, does not capture the tions for a sand site used in the PISA project (Dunkirk
physics of the monopile behaviour accurately. site). The results are compared to PLT data from the
DOI: 10.1201/9781003308829-120
812
PISA project, showing good agreement, particularly soil reaction curves were extracted via MoDeTo at
for load levels related to the serviceability limit different load steps and pile depths.
state (SLS).
3.2 Soil model
The Dunkirk test site was characterised using
a range of in situ tests and advanced laboratory test
ing (Zdravković et al., 2020). Several CPTs were
performed next to the test pile locations and other
key locations. Figure 2 presents the average cone
resistance at the site. The general soil stratigraphy is
shown in Table 2. The water table is found approxi
mately at 5.4 m below ground level.
The Hardening Soil small strain model (HSsmall)
was used as soil constitutive model. The model was
calibrated against available soil data from the Dun
kirk site, including CPTs, seismic CPTs and labora
tory tests such as triaxial tests with bender element
measurements. The calibration process included
study of several CPT-based and empirical parameter
formulations from the literature (e.g. Robertson and
Cabal, 2015; Brinkgreve et al., 2010), investigation
Figure 1. (a) Schematised soil reaction components acting of parameter interdependency and performance of
on a laterally loaded monopile; (b) 1D design model. (after single element test predictions.
Burd et al., 2020). The focus of the CPT-based approach was accur
ate representation of the SLS, according to which the
2 DATABASE allowance for the total permanent tower axis tilt rota
tion is 0.5° (DNVGL, 2016). By analysing the data
Several piles driven into dense sand at the Dunkirk obtained from the PISA project, this limit is reached
site were tested during the PISA project in order to at approximately 30% to 50% of the maximum hori
investigate the effect of different design aspects such zontal load applied to the monopiles during pile load
as pile geometry, load ratio, unloading/reloading testing; hence only that portion of the horizontal
behaviour and creep. In this paper, the results from load-deformation curve was considered for the
three PLTs on medium diameter piles, D = 762 mm HSsmall calibration process.
(i.e. DM3, DM7 and DM4; see Table 1) were com Table 3 shows an overview of the soil parameter
pared to results from 3D FE analyses. This allowed, values for the calibrated HSsmall soil model.
using the FE-derived resistance components, devel
opment of a CPT-based method.
Table 1. Geometry of PISA piles considered in this study

(Taborda et al., 2020).
Diameter Length Slenderness Wall Thick

Pile (m) (m) ratio (-) ness (mm)
DM3 0.762 6.1 8.0 25

DM4 0.762 4.0 5.3 14
DM7 0.762 2.3 3.0 10
3 FINITE ELEMENT MODEL
3.1 General
The commercial software packages Plaxis 3D and
Plaxis Monopile Design Tool, MoDeTo (Plaxis BV,
2018), were used to perform the FE analyses and
extract the soil reaction curves. Through the latter,
the monopile was modelled and then the FE analysis Figure 2. Cone resistance profile at the Dunkirk site
was performed in Plaxis 3D. Finally, each of the four (Zdravković et al., 2020).
813
Figure 3 illustrates the comparative results
between the measured horizontal load-displacement
responses from the PLTs and the predicted responses
from the performed 3D FE analyses. A fairly good
match is observed at the initial part of the curves,
rendering the prediction of the stiffness response,
which was of primary interest, satisfactory.
Additional (fictional) piles were considered in
order to expand the database and check the influence
of pile geometry on each of the four soil reaction
components. Table 4 shows an overview of the add
itional piles considered for the sensitivity analyses.
Figure 3. Comparison of ground level load-displacement
4 SOIL REACTION CURVES for three piles tested during the PISA project (see Table 1
for details). Solid lines represent the results of the pile load
4.1 Distributed lateral load (p-y) tests (after Taborda et al., 2020), dashed lines represent the
results of the 3D FE calculations.
The relationship between p and y along the pile shaft

has been widely studied. In recent years several for
mulations for p-y curves have been developed by
taking into consideration the cone penetration test
and considering the link between cone resistance Table 4. Geometry of additional (fictional) piles con
sidered in the study.
(qc) and in situ horizontal effective stress of the soil
(Houlsby and Hitchman, 1988). An overview of Diameter Length Slenderness Wall Thick-
some of those formulations together with their cor Pile (m) (m) ratio (-) ness (mm)
responding authors is shown below:
DM3A 1.0 6.1 6.1 25
Table 2. Soil stratigraphy at the Dunkirk site (Zdravković DM3B 1.2 6.1 5.1 25
et al., 2020). DM3D 2.0 6.1 3.1 25
DM7B 0.762 3.0 3.9 10
Depth DM7D 0.762 4.7 6.1 10
(m) Material Description DM7E 0.762 3.8 5.0 10
PL1 0.762 15.0 19.7 20
0 - 3 Hydraulic Sand dredged from offshore Flan- PL2 0.5 15.0 30.0 25
fill drian deposits

PL3 0.6 21.0 35.0 30
3 - 30 Flandrian Marine sand deposited during three

sand local transgressions
> 30 Ypresienne Eocene marine clay located beneath
clay the southern North Sea
Table 3. Summary of soil parameters for HSsmall model.
Depth [m] 0-3 3-5.4 5.4-9 9-12.2 12.2-15
γ’ [kN/m3] 19.1 20.8 11.0 11.8 9.8
E50,ref [MPa] 250 223 174 202 87
(=Eoed,ref)
Eur,ref [MPa] 751 668 523 605 260
’0 [deg] 46 45 43 42 37
ψ[deg] 15 9 9 9 9
γ0.7 [-] 1e-4 1.3e-4 1.3e-4 1.3e-4 1.3e-4
G0,ref [MPa] 321 285 223 259 111
Rf [-] 0.88 0.91 0.91 0.91 0.91
K0 [-] 0.5 1.0 0.8 0.7 0.7
814
where Equation 1 is by Novello (1999), Equation 2

is by Dyson & Randolph (2001), Equation 3 is by Li
et al. (2014), Equation 4 is by Suryasentana &
Lehane (2016), D = pile diameter, γ0 = effective unit
weight of soil, z = depth, Gmax = small strain shear
modulus, pu = ultimate lateral soil resistance (for
more details refer to Suryasentana & Lehane, 2016)
and f(y) = exponential function that depends on lat
eral displacement (for more details refer to Surya
sentana & Lehane, 2016).
Equations 1 to 4 were used to derive p-y curves
which were then inserted in a 1D Timoshenko beam
model for modelling of the pile-soil lateral behaviour.
Long slender (fictional) piles (L/D ≥ ~20) were con
Figure 4. Distributed moment ratio as function of the slen
sidered so that the influence of the other three soil
derness ratio L/D.
reaction components (distributed moment, base shear
and base moment) to the overall response is negli
gible (see Table 4; piles PL1, PL2 and PL3). The
results obtained from the 1D model were thereafter
compared with results from 3D FE analyses and it of the rigid pile, the rotation can be obtained from
was found that Equation 2 (Dyson and Randolph, the horizontal displacements. Figure 5 shows the dis
2001) was providing the better match and was thus tributed moment for various slices along the shaft of
selected to define the p-y component for this study. pile DM4, obtained both from the 3D FEA and the
proposed CPT-based formulation (Equation 6).
4.2 Distributed moment (m-ψ)
The distributed moment (m) is caused by the vertical
shear stresses along the pile shaft due to pile rotation
ðψÞ . It is considered that m is linked to p, which is
acting as a normal force along the shaft, through
consideration of the pile-soil interface friction angle
ðδÞ and the pile geometry (L and D). A fitting param
eter, Fmψ, was adopted in order to investigate the
relationship between the aforementioned parameters
for the range of pile geometries considered.
where δ = pile-soil interface friction angle taken Figure 5. Pile DM4 distributed moment along each slice.
Solid lines correspond to the results from 3D FE models
as 2=3j0 .
and dashed lines correspond to the results from the pro
By considering the maximum value of the distrib posed CPT-based formulation.
uted moment at every slice along the pile shaft
obtained from the 3D analysis, mmax, the influence of
L/D on the ratio mmax/Fmψ was investigated
(Figure 4) and a formulation for determination 4.3 Horizontal base force (HB)
of m is proposed (Equation 6). The relatively low R2
value is attributed to the small dataset and the fact Due to the applied force at the pile head, the base of
that the proposed linear trend might be less suitable the pile tends to move in the opposite direction, gen
as L/D increases. erating a horizontal base force (HB). HB was linked
to the base displacement, vb, via a fitting parameter,
FHB, which is a function of the qc at the pile base
and the pile geometry (Equation 7). Figure 6 shows
the relationship between FHB and the ultimate hori
zontal base force, HB,ult, for all piles in the con
sidered database.
The distributed load and distributed moment are soil
reactions along the pile shaft, thus the pile was div
ided into slices and both soil reactions were com
puted per slice. By considering geometric continuity
815
from the Plaxis 3D models of the database resulted
in the following bi-linear relationship:
Figure 8 shows the pile base moment reactions

obtained from the 3D FE models and the proposed
CPT-based formulation (Equation 9) for a selection
Figure 6. Fitting parameter FHB versus the ultimate hori of piles from the database.
zontal base force, HB,ult.
Curve fitting with results from the Plaxis 3D models

of the database resulted in the following bi-linear
relationship:
Figure 8. Pile base moment reactions. Solid lines corres

pond to the results from 3D FE models and dashed lines
correspond to the results from the proposed CPT-based
formulation.
5 PILE LATERAL RESPONSE
The four soil reaction components, as computed

with the use of the proposed equations, were entered
in a 1D Timoshenko beam model for modelling of
the general monopile response under lateral loading.
Results for the piles of Table 1 are shown in Fig
ures 9 and 10. The predictions of the CPT-based
method show in general good agreement with the
Figure 7. Pile base horizontal reactions. Solid lines corres PLTs and the 3D FE analyses for the initial part of
pond to the results from 3D FE models and dashed lines the load-displacement curve, i.e. until approximately
correspond to the results from the proposed CPT-based
formulation.
half the ultimate lateral load (Figure 9). The initial
stiffness response of the monopiles is, therefore,
fairly captured. Figure 10 depicts the pile deflections
Figure 7 shows the pile base horizontal reactions below ground level at different loads, all with mag
obtained from the 3D FEA in comparison to the reac nitude lower than half of the ultimate lateral load.
tions from the proposed CPT-based formulation Again, a satisfactory agreement between the results
(Equation 8) for a selection of piles from the database. of the CPT-based method, the 3D FE analyses and
the PLTs is observed.
Figure 9 shows that after a certain level of
4.4 Base moment (MB) ground level displacement, the response obtained
The base moment is caused by rotation of the pile from the proposed CPT-based method is stiffer
toe. Similarly to the base horizontal force, the base than the response obtained from the 3D FE ana
moment relationship contains a first linear portion lyses and the PLTs. Therefore, a cut-off point
followed by a plateau. Curve fitting with results needs to be defined beyond which the proposed
816
response in which the displacements at ground
level are not larger than 2% to 3% of the pile outer
diameter. This level of deformation generally cor
responds with the serviceability limit state of
monopiles used in the offshore wind industry.
6 DISCUSSION AND CONCLUSIONS
The paper presents a CPT-based method for predict

ing the response of laterally loaded monopiles in
a sand setting. In order to verify application of the
method to full-scale monopiles and until further
Figure 9. Comparison of ground level load-displacement experience is gained with the use of this method,
for three piles tested during the PISA project. Solid lines a FE analysis, considering typical soil conditions of
represent the results of the pile load tests; thinly dashed the investigated site and expected pile geometry, is
lines represent the results of the 3D FE calculations; thickly recommended as a minimum.
dashed lines represent the results of the CPT-based method. The method allows for performing monotonic
conceptual design calculations for monopile founda
tions supporting WTGs in a time-efficient manner,
requiring only CPT data. Total computing time can
be reduced by up to 90 % with respect to performing
3D FE analyses.
The proposed soil reaction formulations were cali
brated against soil data from the PISA sand site in
Dunkirk and consider a specific limit state, i.e. the
SLS. Therefore, applicability of the method to
marine sites with significantly different soil condi
tions (e.g. in terms of strength, stiffness, sand type)
than the ones at Dunkirk and/ or for different limit
states should be carefully checked. In these occa
sions, FE analyses are required prior to implementa
tion of the approach shown in this paper to develop
a site-specific CPT-based method. Alternatively, the
PISA ‘numerical-based method’ can be employed
(Byrne et al., 2017).
The curve fitting process considered the individual
soil reactions from the 3D FE analyses and not the
actual PLTs, since modelling of each individual soil
reaction component based on measured PLT data has
been shown to be problematic (Foursoff, 2018).
The proposed CPT-based method provides
a representation of the global monopile response
under monotonic lateral loading, although the individ
ual soil reactions at a local level can differ consider
ably between the FE analyses and the CPT-based
Figure 10. Comparison of deflection at and below ground formulations. The latter can be attributed to factors
level for three piles tested during the PISA project. Solid
such as imperfect curve fitting and inherent limitations
lines represent the results of the pile load tests; thinly
dashed lines represent the results of the 3D FE calculations; of the 1D model which cannot accurately represent all
thickly dashed lines represent the results of the CPT-based mechanisms of soil-pile interaction at a local level.
method. Values within brackets denote the applied load in The CPT-based method should be employed in its
kN. entirety, i.e., individual soil reaction components
should not be excluded from the analysis or used
method is less accurate. This point was defined by independently.
analysing, for all piles of the database, the differ
ence in stiffness magnitude between the 3D FE
REFERENCES
analyses and the CPT-based method, at ground
level and at all load levels. Consequently, it is Brinkgeve, R. Engin, E. & Engin, H. 2010. Validation of
recommended that the proposed CPT-based empirical formulas to derive model parameters for
method is used for predictions of monopile lateral sands. In Proceedings of the 7th European Conf. on
817
Numerical Methods in Geotechnical Eng., Trondheim – Li, W., Igoe, D. & Gavin, K. 2014. Evaluation of CPT-
Leiden, 137–142. based p–y models for laterally loaded piles in siliceous
Burd, H.J., Taborda, D.M.G., Zdravković, L., Abadie, C. sand. Géotechnique Letters 4: 110–117.
N., Byrne, B.W., Houlsby, G.T., … Potts, D.M. 2020. Novello, E. 1999. From static to cyclic p–y data in calcar
PISA design model for monopiles for offshore wind eous sediments. 2nd Intl. Conf. on Eng. for Calc.
turbines: application to a marine sand. Géotechnique Sedim., Perth, 17–27.
70(11): 1048–1066. Panagoulias, S., Brinkgeve, R.B.J., Minga, E., Burd, H.J. &
Byrne, B.W., McAdam, R., Burd, H.J., Houlsby, G.T., McAdam, R.A. 2018. Application of the PISA frame
Martin, C.M., Beuckelaers, W.J.A.P., … Plummer, M.A. work to the design of offshore wind turbine monopile
L. 2017. PISA: new design methods for offshore wind foundations. In Proc. of the WindEurope Conf. 2018,
turbine monopiles. In Proc. of the 8th Intl. conf. on off Hamburg, Germany.
shore site investig. and geot., smarter solutions for Plaxis BV. 2018. PLAXIS MoDeTo Manual, The
future offshore developments, Vol 1: 142–161, London, Netherlands.
UK: Soc. Underwater Tech. Robertson, P.K., & Cabal, K.L. 2010. Estimating soil unit
DNVGL. 2018. DNVGL-ST-0126 – Support structure for weight from CPT. In Proc. of the 2nd Intl. Symposium
wind turbines. Oslo, Norway: DNV GL. on Cone Penetration Testing, Huntington Beach,
Dyson, G.J. & Randolph, M.F., 2001. Monotonic lateral California.
loading of piles in calcareous sand. Journal of Geotechi Suryasentana, S.K. & Lehane, B.M. 2016. Updated CPT-
cal and Geoenvirontal. Engineering, 127(4): 346–352. based p–y formulation for laterally loaded piles in cohe
Foursoff, W. 2018. Investigation into a new CPT-based sionless soil under static loading. Géotechnique 66(6):
design method for large diameter monopiles in sand. 445–453.
MSc thesis, Technical University of Delft, The Taborda, D.M.G., Zdravković, L., Potts, D.M., Burd, H.J.,
Netherlands. Byrne, B.W., Gavin, K.G., … McAdam, R.A. 2020.
Houlsby, G.T. & Hitchman, R. 1988. Calibration chamber Finite-element modelling of laterally loaded piles in
tests of a cone penetrometer in sand. Géotechnique 38 a dense marine sand at Dunkirk. Géotechnique 70-11,
(1): 39–44. 1014–1029
International Organization for Standardization, 2016. ISO Zdravković, L., Jardine, R.J., Taborda, D.M.G., Abadia, D.,
19901–4:2016 Petroleum and Natural Gas Industries – Spe Burd, H.J, Byrne, B.W., … Ushev, E. 2020. Ground
cific Requirements for Offshore Struc. – Part 4: Geotech characterisation for PISA pile testing and analysis. Géo
nical and Foundation Design Considerations. Geneva: ISO. technique 70(11): 945–960.
818
© 2022 the Author(s), ISBN 978-1-032-31259-0
Automatic interpretation and statistical evaluation of soil conditions for

preliminary design of offshore foundations using the cone penetration test
L. Berenguer Todo Bom & M. Kanitz
Ramboll, Hamburg, Germany
ABSTRACT: The increasing need for renewable energy has led to a substantial growth of the offshore wind
industry. To meet the need of the demanding timelines of the industry, while maintaining a thorough and
robust design of the offshore foundations, automation is key in supporting rapid and accurate geotechnical
screening of any investigated site. The preliminary design of foundations for offshore wind turbines is
a crucial step to determine the feasibility of the planned structures to build an offshore wind park at the site.
Therefore, an initial estimate of the ground conditions, in terms of both soil stratigraphy as well as soil
strength and stiffness properties are necessary. This contribution presents a MATLAB-based tool which per
forms an automatic interpretation and statistical evaluation of soil conditions based on solely cone penetration
testing (CPT) data. The tool derives a preliminary ground model by assembly of similar soil type and strength
and stiffness parameters are determined through available CPT correlations. A depth-dependent statistical
evaluation of the strength and stiffness parameters for each soil layer in the respective soil profile is computed.
The output profiles can be adopted directly for the preliminary design of offshore foundations such as jacket
pin-piles, suction bucket or monopiles using the PISA rule-based method.
1 GENERAL 2 BACKGROUND
1.1 Introduction 2.1 Offshore foundation design practice

The increasing demand in the contribution of renew The most common foundation types in the offshore
able wind power to the national energy grids of most wind industry are the monopile and jacket structures.
countries in the world continues to accelerate at The monopile consists of a single tubular steel element
a rapid pace. Due to the increasing size of these supporting the wind turbine generator (WTG).
structures to improve efficiency and increase energy A jacket structure consists of a welded tubular space
production, the offshore environment with steadier frame with vertical or battered legs and bracing
wind speeds than on land has become very attractive system, which is then in turn commonly fixed to the
for these types of investments. seabed by pile elements, among other foundation
The design cycles of offshore wind turbine foun options.
dation typically consist of 3 phases. An initial phase The preliminary design of these foundations typ
termed ‘concept’, followed by ‘Front-end engineering ically resorts to the methods defined in the API RP
design’ (FEED) and ending in the ‘detailed’ stage. 2GEO (2014) and DNVGL-RP-C212 (2017) stand
The geotechnical investigation campaigns are typic ards. More recent developments, namely the CPT-
ally undertaken during the concept and FEED stages based methods for the calculation of axial capacity,
of development, where cone penetration tests (CPT) Jardine et al. (2015) and Lehane et al. (2005), and
are undertaken at representative and all planned foun the PISA working group publications, Byrne et al.
dation locations, respectively. Whilst non-binding, (2020), are currently adopted already during early
important decisions concerning the foundation type stages of design for jacket piles and monopiles,
and sometimes geometrical and installation method respectively.
ologies are made early in the design process. Improv The definition of a ground model and derivation
ing the methodologies and processes involved in the of geotechnical properties is a necessary initial step
preliminary design of the foundations, resorting to for these design approaches. The following sections
the significant information provided by the CPT describe in detail an automatic procedure to under
in situ testing is the main goal of this contribution. take it supported on the CPT data.
DOI: 10.1201/9781003308829-121
819
Figure 1. Soil classification based on Ic (left), based on the Robertson textural chart (middle) and Robertson chart (right).
2.2 CPT classification charts and ground model 2.3 Correlations for strength and stiffness soil
definition parameters
CPT measurements can be used to determine the soil For coarse-grained soils (sands), the dry density and
stratigraphy and the soil type. For this purpose, clas the friction angle are evaluated based on CPT correl
sification charts can be used that link the CPT meas ations. For sands, the CPT correlations according to
urements to the soil type. Robertson (1990) Robertson & Campanella (1983), Equation 3, as well
proposed two charts based on either the normalized as Meyerhof & Hanna (1979), Equation 4, are well-
cone penetration resistance with variable stress expo established to derive the internal friction angle j0 :
nent Qtn and the pore pressure ratio Bq or based on
Qtn and the normalized friction ratio Fr . These charts
are subdivided into 9 regions ranging from sensitive
fine grained over sand to very stiff fine-grained soil.
The normalized Soil Behavior Type Index, Ic, based
on Robertson and Wride (1998) is a parameter for
mechanical behavior classification of the soil and with qt = corrected cone resistance [MPa] and =
defined in Equation 1. vertical effective stress [MPa].
For fine-grained soils, the undrained shear strength
Cu can be evaluated based on CPT measurements by
the cone factor Nkt through:
It defines the soil as either sand (for 1.31 < Ic <

2.05), sand/silt mixtures (for 2.05 < Ic < 2.6), clay/
silt mixtures (for 2.6 < Ic < 2.95) or clay (2.95 < Ic The shear wave velocity in sands can be derived
< 3.6). Robertson (2010) additionally proposed with the CPT correlation after Baldi et al. (1989)
a non-normalized Soil Behavior Type Index ISBT with the cone resistance qc [MPa] and [MPa]:
that is based on basic CPT measurement values:
For the CPT correlation of the small strain shear

modulus, G0, in sands, the correlation after Rix and
Stokoe (1991) can be used via qc [kPa] and [kPa]:
820
Figure 2. Identified soil types based on Ic (right), preliminary ground model (middle), simplified ground model (right).
soil behavior type index Ic, Robertson & Wride (1998).

Figure 1 shows the derived normalized soil behavior
type index Ic and ISBT (left), the net cone resistance
For clays, the shear wave velocity is derived with over depth including a description of the soil type
the CPT correlation after Mayne and Rix (1995) based on the textural chart (middle) and the textural
through Equation 8 with qc in [kPa]. chart type plots for a CPT measurement (right), for
a given turbine location.
Since CPT measurements can vary substantially
due to repushes or small rocks at the cone tip, the
soil behavior type index is filtered with a moving
Finally, the small strain shear modulus in sands median function to smooth these variations. Through
is based on the CPT correlation proposed by this investigation and presentation, the soil types pre
Mayne and Rix (1993) with qc in [kPa]: sent at the site are identified and an initial assess
ment can be made regarding the design of the
foundations. In Figure 1, the location is character
ized by clay layers with high thickness and embed
ded sand layers of medium to high thickness. Some
3 CPT INTERPRETATION minor stratification of the soil can also be observed
in a depth of 20 m to 25 m. The textural chart add
For a preliminary assessment of the design of the itionally indicates that the location is mostly com
offshore wind turbine foundations, the CPT meas prised of fine soils since most data points plots into
urements at the turbine locations are analyzed. To section 3 (clay to silty clay) and 4 (clayey silt to silty
derive the soil information, including strength and clay). The comparison between Ic and the textural
stiffness values, the cone tip resistance qc, the sleeve chart shows a good agreement in terms of soil classi
friction fs and the excess pore water pressure u2 are fication. For preliminary ground models a low level
required in depth. The following sections explain the of soil variability and stratification is preferred prior
necessary steps to derive a soil profile including the to additional soil information being available. The
required soil parameters for the foundation design aim of the ground model is to identify soil units that
from the CPT measurements. can be characterized for the entire offshore wind
farm site. Thin layers tend to do not influence the
3.1 Preliminary ground model foundation design. To improve the efficiency of the
analysis of the foundation design, a ground model is
As a first step, the soil type is determined based on automatically created, neglecting thin soil layers to
either the soil textural chart, Robertson (1990) or the derive a general ground model.
821
Figure 3. Derived ground model (right) with relevant soil parameters for design including lower bound, best estimate and
upper bound values.
Figure 2 presents the output of this automatic pro value obtained between Equation (3) and Equation (4)
cedure which details the ground model derived. On is adopted. The relative density is determined based
the left side of the figure, the Soil Behavior Type on the correlation of Jamiolkowski et al. (2003).
indexes over depth are shown as in Figure 2. In the Figure 3 presents the correlated strength parameters
middle, the detailed ground model is shown which is over depth for the investigated soil profile, all of
derived directly from Ic, still including thin layers. which are obtained automatically following the
By defining a threshold thickness of the soil layers ground model definition.
to be acceptable in the ground model, the simplified
ground model is derived (Figure 2 right). The soil
3.3 Statistical methodology for parameter
layering procedure is based on a moving median
derivation
approach of around 50 cm in both directions. In this
analysis, the limit thickness of a soil layer is set to Since soil parameters can vary significantly,
0.55 m. a statistical assessment of their distribution in
depth is preferred. The local scatter present in
CPT measurements should not be taken directly
3.2 Determination of relevant soil parameters
as input values for parameter derivation but
Following the ground model definition, the derivation rather statistically quantified. Figure 3 shows the
of the soil parameters relevant to the foundation soil parameters over depth for the investigated
design is undertaken. The CPT measurements are ana CPT profile. For the design of offshore wind tur
lyzed to derive the strength and stiffness parameter of bine foundations, the statistical derivation of
the determined soil units. In the preliminary stage of a lower bound (LB), upper bound (UB) and best
the design process, typically, laboratory tests results estimate (BE) of the soil parameters, defined as
of the soil are not yet available. Well-established cor the 25th, 75th and the minimum between the
relations for the relevant soil parameters are hence mean and the median, respectively, is required.
employed to derive a first estimate of the soil proper Therefore, a depth-dependent statistical evaluation
ties from the in-situ CPT measurements. For all soil considering a linear variation of the soil param
types, the submerged unit weight γ’ and the small eter within each layer is performed for every soil
strain shear modulus G0 are derived. The soil type unit in the preliminary ground model. These per
determines which correlation (see Sec. 2.3) is used to centiles are typical assumptions in offshore foun
derive G0. For clay and silty clay soils, the undrained dation designs for a normal distribution of the
shear strength cu is determined while for sand and parameters. The statistical evaluation of the mean
silty sands, the internal friction angle φ0 and the rela determines if a soil parameter is constant or vary
tive density Dr are calculated. For φ0, the minimum ing in depth as can be seen in Figure 3.
822
4 DESIGN APPROACHES
Different approaches can be used for the design of

offshore wind turbine foundations based on design
phase, client or regional specifications and the founda
tion type. For monopiles, the lateral bearing capacity
is design-driving while for jacket-pile foundations, the
axial bearing capacity is more relevant.
Figure 4. Pile geometrical definitions, as according to
4.1 Monopile design approaches Byrne et al. (2020).
For monopiles, the most used approaches are based
on the API RP 2GEO (2014) or the PISA rule-based
methodology proposed by Byrne et al. (2020) and
Burd et al. (2020).
4.1.1 Enhanced API approach

The soil interaction curves described in the API RP 4.1.2 PISA rule-based methodology
2GEO (2014) and DNVGL-RP-C212 (2017) stand In the PISA methodology, the lateral resistance of the
ards do not include the local distributed moment monopile is determined considering the lateral soil
component. This component was re-introduced by reaction (P-y), the distributed moment along the pile
the PISA working group, Byrne et al. (2020), due to shaft (m- ψ), the base shear reaction (S-y) and the base
its relevance in the lateral component of resistance moment (MB-ψ). The soil reaction curves are deter
and therefore, in the design of large diameter pile mined by a normalized analytical function that is
foundations. The incorporation of the soil resistance described by the ultimate resistance, ultimate displace
component is usually achieved by either performing ment, curvature of the reaction function and initial
Finite Element Analyses (FEA) or adopting the stiffness. Details of the analytical function, and correl
PISA rule-based formulations. For the preliminary ations from soil parameters described in the previous
design of pile foundations, however, FEA are not sections to obtain the normalized parameters, can be
usually performed and the database supporting the found in Burd et al. (2020) and Byrne et al. (2020).
correlations proposed by the PISA rule-based meth These four parameters are either taken from the rule-
odology is empirically verified only for three based sets provided in Burd et al. (2020) and Byrne
onshore sites in northern Europe. To bridge this gap et al. (2020) or, more adequately, calibrated based on
in the adequacy of the correlations to other regions advanced 3D FE models.
of the world, and avoid conservativism in prelimin
ary designs, a simple calculation of this component 4.2 Jacket piles design approach
can be adopted. Considering a simple geometrical The design of the jacket piles follows the guidelines
assumption, the distributed moment component can presented in API RP 2GEO (2014). For the axial
be determined from the T-z curves defined in bearing capacity, the shaft friction resistance (T-z)
DNVGL-RP-C212 (2017). The distributed moment and the tip resistance (Q-w) are determined. The
component is defined in Byrne et al. (2020) as T-z curves describe the relationship between mobil
shown in Equation 10. ized soil-pile shear transfer and local pile deflection
at any depth while the mobilized end bearing resist
ance and axial tip deflection is described by the
Q-w curves. The soil parameters defined using the
automated procedure described allow for an
advanced preliminary design of this foundation type.
The variable tz represents the vertical shear stresses
on the pile shaft, D the pile diameter and the angle
is defined in the Figure 4. Assuming uniformly dis 5 EXAMPLE MONOPILE MODEL
tributed vertical shear stresses on the pile shaft the
distributed moment’s component can be easily calcu The derived ground model and geotechnical parameters
lated. The rotation variable ψ can be simply deter can be used in the different design approaches to obtain
mined by considering that the shear strain in each pile a monopile design. In Figure 5, the results obtained
element to be negligible which results in θ=ψ, since from laterally loading a monopile with an outer diam
the ratio of shear stress to the pile’s shear modulus, eter of 8.4m, wall thickness of 80mm and an embed
pile cross section and shear factor is assumed infini ment length of 32m in the analyzed soil profile, are
tesimal. Hence the rotation variable ψ can be simply shown. A lever arm of 38m of the horizontal load was
defined from the z component of the T-z curves as considered based on the wind and wave loading
shown in Equation 11. conditions.
823
Figure 5. Global monopile lateral stiffness curves - FEA, PISA rule-based, API and enhanced API.
As additional background information, the Byrne, B.W., Houlsby, G.T., Burd, H.J., Gavin, K.G.,
derived load-displacement curve from advanced Igoe, D.J.P., Jardine, R.J., Martin, C.M., McAdam, R.
Finite Element Analyses (FEA), undertaken in the A., Potts, D.M., Tarborda, D.M.G. & Zdravković, L.,
detailed design stage, is also plotted. 2020. PISA Design Model for Monopiles for Offshore
The results shown corroborate the conservativism Wind Turbines: Application to a Stiff Glacial Clay Till.
associated with the API methodology and the signifi Géotechnique, 70(11): 1030–1047.
DNVGL-RP-C212, Offshore soil mechanics and geotech
cant improvement seen in both the PISA rule-based
nical engineering, DNVGL, August 2017.
approach and the enhanced API. Moreover, the sim Jamiolkowski, M., Presti, D.C., Manassero, M., 2003.
plified calculation of the distributed moment curves Evaluation of Relative Density and Shear Strength of
option falls within the expected range of lateral Sands from CPT and DMT. J.T. Germaine, T.
monopile stiffness behavior observed in the more C. Sheahan, R.V. Whitman (Eds.), Soil Behavior and
established approaches. Soft Ground Construction, ASCE Geotechnical Special
Publication, American Society of Civil Engineers
(ASCE), 119: 201–238.
6 CONCLUSIONS Jardine, R., Chow, F., Overy, R. and Standing, J., 2005.
ICP Design Methods for Driven Piles in Sands and
Clays. Imperial College, London
An automatic procedure to derive geotechnical Lehane, B.M., Schneider, J.A., Xu, X., 2005. The UWA-05
ground models and parameters based on the CPT method for prediction of axial capacity of driven piles in
data has been described. The employed soil classifi sand. Frontiers in Offshore Geotechnics
cation charts and parameter correlations are state of Mayne, P. & Rix, G., 1993. Gmax-qc Relationship for
the art and the most adequate per region of the world Clays. Geotechnical Testing Journal. 16(1): 54–60.
and soil type can be adopted. Finally, a simplified Mayne, P. & Rix, G., 1995. Correlations between Shear
approach to derive the distributed moment lateral Wave Velocity and Cone Tip Resistance in Natural
contribution from the axial resistance curves is Clays. Soils and Foundation, 35(2): 107–110.
included. An exemplary monopile design is shown Meyerhof, G.G. & Hanna, A.M., 1974. Ultimate bear
ing capacity of foundation on sand layer overlying
and the lateral load displacement curves plotted for clay. Canadian Geotechnical Journal, 11(2):
different design methodologies. 223–229.
Rix, G. & Stokoe, K., 1991. Correlation of initial tangent
modulus und cone penetration resistance. In: 1st Inter
REFERENCES national Symposium on Calibration Chamber Testing
(ISOCCT1), Potsdam, NY.
API RP 2GEO/ISO 19901-4, 2014. Geotechnical and Robertson, P.K., 2010: Soil behaviour type from the CPT:
Foundation Design Considerations, American Petrol an update. International Symposium on Cone Penetra
eum Institute, 2nd edition. tion Testing, CPT’10, Huntington Beach, CA, USA
Baldi, G., Belotti, R., Ghionna, V. and Jamiolkowski, L.P. Robertson, P.K. 1990. Soil classification using the cone pene
D., 1989. Modulus of Sands from CPT’s and DMT’s. In: tration test. Journal of Geotechnical and Geoenvironmen
Proceedings of XII ICSMFE, Rio de Janiero. tal Engineering, 27(1): 151–158, doi: 10.1139/t90-014.
Burd, H.J., Taborda, D.M.G., Zdravković, L., Abadie, C. Robertson, P.K. & Campanella, R.G., 1983. Interpretation
N., Byrne, B.W., Houlsby, G.T., Gavin, K.G., Igoe, D.J. of cone penetration tests, Part I: Sand. Canadian Geo-
P., Jardine, R.J., Martin, C.M., McAdam, R.A., technical Journal, 20(4): 718–733.
Pedro, A.M.G. & Potts, D.M., 2020. PISA design model Robertson, P.K. & Wride, C.E., 1998. Evaluating cyclic
for monopiles for offshore wind turbines: application to liquefaction potential using the cone penetration test.
a marine sand. Géotechnique, 70(11): 1048–1066. Canadian Geotechnical Journal, 35(3): 442–459.
824
© 2022 the Author(s), ISBN 978-1-032-31259-0
Effect of dynamic pile driving parameters on vibratory penetration

Anchal Bhaskar, Stefan Kreiter & Dina Al-Sammarraie
Tobias Mörz
GeoEngineering.org
ABSTRACT: To reduce the harmful noise for marine life, generated during impact driving of offshore founda
tions, vibratory driving was presented as an alternative. Vibratory driving produces less noise and allows faster
installation in sandy soil. This technique is cost-efficient for offshore pile installations and causes less pile fatigue
than impact driving. One of the main issues with vibratory driving is the lack of a reliable driveability analysis. This
is due to uncertainties in understanding of cyclic soil response during vibratory driving and the effect of different
vibratory driving parameters on the soil response. In this study Vibratory Cone Penetration Test (VCPT) is used to
assess the cyclic soil response during vibratory penetration. VCPT penetrates the soil while inducing cyclic loads
and measures cone resistance. VCPTs were conducted in a calibration chamber to investigate the reduction of the
cyclic cone resistance due to variation in frequency for the same penetration path. The VCPT results are discussed in
terms of acceleration, energy, reduction ratio, which are considered as primary parameters in the different driveability
analyses methods. The effect of cavitation during upward movement of the cone is investigated and discussed.
1 INTRODUCTION value introduced by Jonker (1987) and the acceler

ation-based Vibdrive model using a quasi static
With the increasing demand for renewable energy, approach (Holeyman 1993). In the former, the degrad
there is an increased need to deploy offshore wind ation factors were obtained from observations of full-
turbines around the world. Most wind turbines are scale vibratory pile installation in different soil types.
constructed on pile foundations which are commonly In the latter, the effect of dynamic pile driving param
installed by impact driving (Fischer et al. 2013). eters on the degradation of cyclic soil resistance is
However, such construction activities adversely considered implicitly by providing an empirical lique
impact marine life due to underwater high-pressure faction factor. Until now there is no clear understand
noise generated during impact driving (Dahl et al. ing of the effect of different dynamic pile driving
2015). Alternative installation options such as vibra parameters on the degradation of cyclic soil resistance.
tory pile driving, widely used in onshore engineering, Various studies have identified different key
have been gaining attention in offshore wind farm parameters influencing the degradation of soil resist
projects. This is because vibratory pile driving gener ance due to vibratory driving. Acceleration, fre
ates less noise in the water, causes less structural quency, displacement amplitude, energy and soil
fatigue to the piles, enables fast installation in sandy type dependent β value have been proposed as key
soil, and allows easy pile handling (Tsouvalas 2020). parameters (Barkan 1962); (Rodger and Littlejohn
It suffers from drawbacks such as, inefficiency in fine 1980); (Jonker 1987); (O’Neill and Vipulanandan
grained soil (Massarsch et al. 2017), and the lack reli 1989); (O’Neill et al. 1990); (Wang 1994); (Viking
able driveability model. It has also been proved that 2002); (Cudmani and Manthey 2019). For one vibra
the bearing capacity of vibrated piles is less than that tory driver-pile combination, all these parameters are
of impact driven ones (Holeyman and Whenham interlinked or fixed. There is no key parameter
2017); (O’Neill et al. 1990). The main issue with the agreed upon by the past studies that dominates the
current driveability models is the lack of understand degradation of soil resistance. Until now no system
ing of the effect of dynamic pile driving parameters atic study has been made to isolate the dynamic and
on the degradation of cyclic soil resistance. quasi static effects of vibratory pile driving.
Different methods have been developed to predict There are different suggestions explaining the
the degradation of the cyclic soil resistance by vibra degradation of cyclic soil resistance during vibratory
tory pile driving. The most prevalent ones are the β pile driving. In the work of Dierssen (1994) a total
DOI: 10.1201/9781003308829-122
825
loss of stress between the soil and the penetrating
object was observed during the upward movement of
the pile. This zone of total loss of stress is called
cavity and the process is called cavitation. The soil
is presumed to flow in this cavity and to be disturbed
during this process. This disturbation is postulated to
be responsible for the degradation of the cyclic soil
resistance. Acceleration is also an important param
eter that affect the degradation of cyclic soil resist
ance. This was proven in study by Barkan (1962)
where steel balls placed on a unconfined dry sand
bed started to penetrate when the acceleration of the
whole bed exceeded a value near to the acceleration
of gravity, this process is called fluidization (Barkan
1962). Furthermore, if the excess pore water pres
sure generated by cyclic loading reaches the value of
the total stress, liquefaction and thus total loss of
shear strength occurs (Kramer 1996). Therefore, Figure 2. Illustration of VCPT penetration modes of 3 mm
liquefaction is another possible cause for the degrad amplitude with varying frequency and average penetration
ation of cyclic soil resistance during vibratory velocity; depth vs. average depth plane. Orange stars repre
penetration. sent the penetration after 4 cycles. Compare to Figure 1.
The effect of various vibratory pile driving param
eters can be investigated with VCPTs. The penetration
of the cone is controlled by amplitude, frequency and When the vibration amplitude of different pene
average penetration velocity and these parameters can tration modes are the same and if the frequency and
be varied independently. These parameters are also dir average penetration velocity for each of these pene
ectly linked to the acceleration. The soil behavior itself tration modes vary with the same factor, then they
is linked to cavitation, cyclic stress amplitude, confin have the same penetration path in the depth vs. aver
ing stress, number of cycles and drainage. In this study age depth plane. This is best seen in Figure 1 and 2,
the influence of dynamic parameters was isolated from where the four stars marked in the time domain
quasi static influence by performing tests with different (Figure 1) fall in the same spot in the average depth
frequencies and proportionally changed average pene domain (Figure 2). In fact all the points of the three
tration velocities, leading to identical penetration paths penetration modes fall directly on each other in the
in the depth vs average depth plane (Figure 1, 2). average depth domain. If the soil reacts only quasi
A change in frequency changes the dynamic param statically then, the resistance and other soil behavior
eters like acceleration and time for drainage, while would be the same for all three penetration modes.
leaving quasi static parameters like deformation per The differences among the three penetration modes
cycle unchanged (Figure 1). are then the dynamic effects. This study systematic
ally quantifies the influence of currently proposed
key parameters on vibratory penetration.
Mini Vibratory Cone Penetration Tests (VCPT)
were carried out in the Marum Calibration Chamber
(MARCC). The VCPT, has a cone which penetrates
the sample while applying vertical cyclic loading and
measures the cone resistance. It was used for in-situ
soil investigation to obtain dynamic soil parameters
(Al-Sammarraie 2020). However, laboratory investi
gations eliminate the effect of inherent soil variability
encountered during field investigations. Hence, this
investigation is carried out in the MARCC to investi
gate the fundamental soil behavior.
2 METHODOLOGY
2.1 Properties of Cuxhaven Sand

Figure 1. Illustration of VCPT penetration modes of 3 mm The Cuxhaven Sand for the experiment comes from
amplitude with varying frequency and average penetration a sand pit in Altenwalde, Cuxhaven, Northern Ger
velocity, depth vs. time. Orange stars represent the penetra many. Cuxhaven Sand has similar properties as
tion after 4 cycles. Compare to Figure 2. North Sea sand with cone resistances greater than 60
826
MPa in situ (Naumann et al. 2013). It is (BC5) (Huang and Hsu 2005). Expansion is induced
a Pleistocence glaciofluvial sand deposited in the during CPT penetration and the stress increase with
Saalian glacial (Stähler 2020). It is uniform, well BC5 is equivalent to the reaction of a infinite half-
sorted, fine to medium quartz sand with less than 5% space of the same material. The soil stiffness for this
of fines (Fleischer et al. 2016). The properties of the increase is based on the polynomial function obtained
sand is given in Table 1. in lateral compression test performed prior to penetra
tion of the cone (Kluger et al. 2021). The soil is
finally penetrated with a small cone (Table 2).
Table 1. Parameters of Cuxhaven Sand.
Property of the sand Value Table 2. Specification of the cone in VCPT.

Uniformity coefficient (Cu ) 3.09 Cone diameter Sampling Rate Sample to cone ratio
Median diameter (D50 Þ 0.25 mm
Minimum void ratio (emin ) 0.48 12 mm 1 kHz 25
Maximum void ratio emax 0.82
Specific gravity (Gs ) 2.644
Quartz content 495%
2.2.1 Sample preparation
The sample is prepared by air pluviation with
a hand-held device similar to the one described in
the work of Baldi et al. (1982). The sand is pluviated
at a constant falling height of 0.60 m in a sample
mould. The flow rate of the pluviation is adjusted by
varying the aperture size of the hand-held device.
The constant drop height with the chosen flow rate
led to a dense sample of Cuxhaven Sand. After plu
viation of the sample, a vacuum of -50 kPa is applied
through the pore pressure tubes to keep the sample
stable before removing the surrounding mould.
2.2.2 Saturation
The sample is saturated by increasing the vacuum at
the rate of 10 kPa/steps until nearly full vacuum of
ca. -100 kPa is reached. The effective stress is held
constant at ca. 50 kPa by changing cell pressure and
pore pressure simultaneously. Deaired deionized water
is added to the sample from the pore pressure tubing
at the bottom of the sample at a constant flow rate of
500 cm3/minute until the sample is saturated. Then
the pore pressure and the cell pressure are increased
Figure 3. MARUM Calibration Chamber (MARCC). again in 10 kPa steps holding the effective stress of
ca. 50 kPa constant. Later, after the application of 300
kPa back pressure a saturation test is performed and
2.2 Calibration chamber all tests had degree of saturation of 95% i.e., Skemp
In MARCC, the sample is fitted inside the chamber ton’s B value of 0.95 was achieved (D7181 2020).
with a sinter plate on top and a water-filled cushion at
the bottom (Fleischer et al. 2016). The soil sample is 2.3 Lateral compression test
laterally contained in a latex membrane with
a pressure-controlled cell water around it. The pore The lateral compression test determines the unique
pressure is connected to the sample through small stress - strain relationship for the application of BC5
tubes located below a geotextile at the bottom of the during the penetration. In this test, the cell water
sample and the sinter plate on top. The axial strain is pressure is increased until lateral strain εh of 0.08%
computed by the volumetric changes in the water is reached, while keeping the axial strain fixed and
filled cushion. The axial, cell and pore pressures are maintaining a constant pore pressure. The resulted
controlled using three syringe pumps. The lateral stress-strain curve is fitted by a unique fifth order
strain is obtained from three lateral circumferential polynomial function to compute the compensating
displacement sensors installed around the sample at pressure for the strain in BC5 (Huang and Hsu
height of 0.15 m, 0.25 m, 0.35 m from the top 2005). At the end of test the lateral stress is reversed
(Figure 3). The lateral stress caused by an infinite soil back to initial stress. The influence of lateral com
medium that would be around in the field is simulated pression test on the cone resistance was found negli
in the chamber by applying Boundary Condition 5 gible (Stähler 2020).
827
2.3.1 Consolidation independently measured points with the same phase
The consolidation stress is is chosen to be equivalent for each point of two full sinusoidal cycles. The over
to at depth of 10 m; therefore, the sample was aniso layed double cycles were then averaged yielding the
tropically loaded to an effective vertical stress σ0 v of mean cyclic cone resistance qc;cyc;mean and the mean
100 kPa, horizontal effective stress σ0 h of 45 kPa. displacement d (Figure 4).
This corresponds to a lateral earth pressure of 0.45
(Fleischer et al. 2016). The sample was allowed to
2.5 Pile-soil interaction models
consolidate for a minimum of 90 minutes.
In order to compare the effect of dynamic parameters
on the cyclic cone resistance qc;cyc , following predic
Table 3. VCPT with varying dynamic pile driving param tion models are evaluated.
eters investigated in MARCC.
Frequency Velocity Amplitude

Experiment Name (Hz) (mm/s) (mm)
Push 1 V1 5 10 3
V2 2.5 5 3
V3 1 2 3
Push 2 V4 10 20 3
V5 0.5 1 3
V6 0.25 0.5 3
V7 0.125 0.25 3
Push 3 S1 - 20 -
2.3.2 Vibro Penetration

The tests are performed at a constant amplitude of
3 mm with different frequencies and proportional
varied average penetration velocities (Table 3). The
total penetration depth is 0.42 m and a steady state
penetration is expected between 0.18 m and
0.32 m from the top of the sample (Kluger et al. Figure 4. Stacking of 25 double cycles of cone resistance
2021). In two experiments, two VCPT Pushes were and displacement from V1 (5 Hz). The shaded colour
conducted. Each Push had 3-4 tests in the same along the mean curves represents 25 overlayed double
penetration. Tests V1 to V4 were performed in Push cycles.
1 in the the first experiment, and tests V5 to V7 were
performed in Push 2 in the second experiment
(Table 3). At frequency greater than 10 Hz, the 2.5.1 Jonker’s model
applied displacement amplitude of 3 mm was not Jonker estimated the cyclic cone resistance qc;cyc
completely achieved. Initially the cone penetrated from static qc by using the β value. This is an empir
for 0.18 m in the sample to reach steady state condi ical value estimated from the past project experi
tions with the same penetration mode. Then the fre ences of different soil types (Jonker 1987). For
quency and average penetration velocity were comparison with the experimental results of the cur
changed approximately after every additional rent study, a β value of 0.15 is used from Jonker’s
50 mm. It was observed that after each driving par recommendation as a “representative value” of
ameter change, the steady state cyclic soil response a clean sand.
was achieved in the first cycle. For comparison,
a static CPT with BC5 is also conducted in a third
experiment.
where, qc;j is the cyclic resistance based on β

2.4 Data analysis value [kPa], β is the empirical degradation value for
The maximum cone resistance of each cycle is the cyclic cone resistance, static qc is the static cone
maximum cyclic cone resistance qc;cyc;max . For every resistance [kPa]
test, a dataset of cone resistance and displacement
which consist of 50 to 80 steady state cycles are 2.5.2 Energy based model
chosen. Two consecutive cycles were then overlayed The total energy dissipated for each penetration
25 - 40 times with the following two consecutive cycle was computed from the force - displacement
cycles (Figure 4). This resulted in 25 - 40 curves (Cudmani and Manthey 2019) (Figure 5).
828
comparing the static qc with maximum cyclic cone
resistance qc;cyc;max , there is 25 - 50% reduction in
the qc;cyc;max for all the vibratory CPTs (Figure 6).
The Jonker’s model underestimates the cyclic cone
resistance value qc;j (Figure 6).
At the steady state, the qc;cyc;max of the mean
cyclic cone resistance value from different vibrational
modes exhibited minimal changes with changing fre
quency (Figure 7). The energy dissipated during the
penetration of each cycle was roughly proportional to
the qc;cyc;max (Figure 7). However, the energy for tests
with penetration modes of 5 and 10 Hz were low
even though the qc;cyc;max is in comparison high.
Figure 5. Energy for each cyclic cone resistance - displace

ment cycle of VCPT. The shaded part represents energy dis
sipated for each cycle in Karlsruhe model. Modified after
Cudmani and Manthey (2019).
Figure 7. Maximum cyclic cone resistance at various driv

ing frequency, Energy calculated using (Cudmani and Man-
they 2019) for various driving frequency.
The reduction of cone resistance in comparison to

the static penetration is also seen when looking at the
full details of the penetration cycles (Figure 8). The
mean cone resistance - displacement cycle is similar
to Karlsruhe model, but more roundish and not
exactly reaching zero (Dierssen 1994) (Figure 5, 8).
During downward motion, the cone resistance rises
linearly until it reaches qc;cyc;max . In the current study,
it is observed that the magnitude of cyclic cone
Figure 6. Static qc vs penetration depth, qc;cyc;max vs pene resistance degradation remains almost unchanged
tration depth for test V1 - V7 (represented by penetrating
frequency).
even at much lower frequencies and way longer time
intervals of upward motion of cone and cavitation.
From the cone resistance - displacement cycles, the
3 RESULTS upward displacement with cavitation could be
observed for all the performed VCPTs. The upward
The maximum cyclic cone resistance qc;cyc;max pro displacement with cavitation ranged from 3.9 to
files for each Push seem to be continuous even when 4.2 mm for all VCPTs. Both in the lowest frequency
the vibro penetration mode is changed (Figure 6). of 0.125 Hz, where 8 s are needed for the cone to
On the other hand, the vibratory penetrations as well reverse to downward motion and in the highest fre
as the static penetration are not in perfectly constant quency of 10 Hz where only 0.1 s is spent for the start
steady state conditions but undulate in a resistance of downward motion, it was observed that the upward
range between 2 - 4 MPa. A gradual reduction in displacement with cavitation were nearly identical
qc;cyc;max are observed along the bottom boundary (Figure 8). When comparing the first and second
predominantly in Push 2 (Figure 6). When cycles of cone resistance - displacement curve of each
829
in Cuxhaven Sand, since the pore pressure would have
ample time to drain at lower frequencies.
The gradual reduction in the qc;cyc;max for the
VCPTs performed in Push 2, could be an effect of soft
bottom boundary of the chamber. The influence of
bottom boundary in MARCC was observed in the
static CPTs (Kluger et al. 2021). However, since this
gradual reduction in qc;cyc;max was not observed in the
Push 1, the explanation of bottom boundary could be
disregarded (Figure 6). The more plausible explanation
for this reduction is the heterogeneity in the sand fabric
and relative density that occur due to imperfect execu
tion of the air pluviation method (Lagioia et al. 2006).
The underestimation of qc;j represented in
Figure 6 calculated from β value is almost certainly
caused by the oversimplification of the multi-variate
degradation of cyclic soil resistance due to dynamic
loading (Viking 2002). The proposed β values are
Figure 8. Mean cone resistance - displacement cycles for based on data from real vibratory driver-pile com
different vibrational modes of VCPT. binations with amplitudes probably in the range of
6 mm, which is larger than the amplitude of 3 mm
used in this study.
VCPT, it is clear that the cone penetrated approxi
The proportionality between energy and qc;cyc;max
mately 2 mm in virgin soil in the second cycle. How
is most likely because the general shape of the cone
ever, the value of qc;cyc;max of this cycle was still well
resistance - displacement cycle is very similar for all
below the static cone resistance (Figure 8).
the VCPTs in this study. Therefore the area below the
curve is proportional to the maximum value of cone
resistance. If the upward displacement with cavitation
4 DISCUSSION would differ a lot, this proportionality would break
down. The slightly lower energy dissipation at the
The lower value of qc;cyc;max compared to static qc high driving frequency, i.e., 5 Hz and 10 Hz is prob
might be attributed to cavitation as proposed by ably due to the inability of the system to achieve com
Dierssen (1994) (Figure 6). But Dierssen (1994) plete displacement amplitude of 3 mm (Figure 7). The
attributed the loss of contact from the soil mainly to dissipated energy might be an important parameter to
the short time span for the cone to go up and to predict the degradation of the cone resistance since
reverse for downward motion, which might indicate both are constant and ca. proportional in this study.
that the cavitation is a dynamic process. In this study However,like for the cavitation, since both values are
however, no difference was found in the upward dis constant more data with different boundary conditions
placement with cavitation for 0.1 s and 8 s reversal is needed to prove it. It can be concluded that the
time of the cone, therefore there seem to be no upward displacement with cavitation does not depend
dynamic effect in the cavitation for the tested pene on frequency and average penetration velocity and
tration modes and material. The similar values of that the energy is constant for same penetration path
upward displacement with cavitation might lead to regardless of the frequency, for tested penetration
approximately similar degree of remolding and loos modes, soil type and state.
ening of the soil for all VCPTs which might be The dynamic parameters have surprisingly little
reflected in the similar reduction in the cyclic cone effect on the degradation of the cyclic soil resistance
resistance. A relation between length of upward dis in this study (Figure 7), however there seem to
placement with cavitation and reduction in the cyclic be a complex quasi static behavior of the soil
cone resistance could not be established since this for frequencies between 0.125 and 10 Hz and corres
length was almost constant and experiments with ponding maximum accelerations between 0.00185
a lot more different penetration modes would be and 11.8 m/s2 With higher frequencies the dynamic
necessary to establish such a relation. effects will almost certainly appear, but its signifi
A classical reduction in cyclic cone resistance due cance for the vibratory driving of monopiles with
to liquefaction was not expected from the beginning, driving frequency of 20 Hz is unclear and would need
because a recent field study suggested that there was futher experiments, for which it is planned to upgrade
no development of pore water pressure when perform the MARCC system. The observed quasi static effects
ing vibratory CPTu (Al-Sammarraie 2020); (Wong are probably very important for vibratory driving of
et al. 1992); (O’Neill and Vipulanandan 1989). The big monopiles and classic liquefaction and acceler
non frequency i.e., time dependent reduction of ation driven fluidization may not be the only process
qc;cyc;max supports the notion that liquefaction is not allowing for penetration with little resistance during
the main cause of cyclic cone resistance degradation vibratory driving.
830
5 CONCLUSION driven and vibratory driven steel piles in the german
north sea. In Proceedings Conference on Maritime
A static CPT and several VCPTs were performed in Energy (COME), Hamburg, pp. 21–22.
MARCC in Cuxhaven Sand to investigate the influ Fleischer, M., S. Kreiter, T. Mörz, & M. Huhndorf (2016).
ence of dynamic pile driving parameters on cone A small volume calibration chamber for cone penetra
resistance. The degradation in the cyclic cone resist tion testing (cpt) on submarine soils. In Submarine Mass
ance is most likely related to the soil disturbance Movements and their Consequences, pp. 181–189.
Springer.
which might be caused during the loss of contact Holeyman, A. (1993). An analytical model-based computer
during the upward motion of the cone tip, a process program to evalaute the penetration speed of vibratory
named cavitation. The dynamic pile driving param driven sheet piles. Geotechnique 43(18), 65–78.
eters, acceleration and drainage time have little to no Holeyman, A. & V. Whenham (2017). Critical review of
influence on degradation of the cone resistance for the hypervib1 model to assess pile vibro-drivability.
tested penetration modes. Jonker’s β value underesti Geotechnical and Geological Engineering 35(5),
mates the cyclic cone resistance for 3 mm amplitude. 1933–1951.
The energy dissipation provides a proportional esti Huang, A.-B. & H.-H. Hsu (2005). Cone penetration tests
under simulated field conditions. Geotechnique 55(5),
mate of cyclic cone resistance but since both param
345–354.
eters were constant no conclusive relation could be Jonker, G. (1987). Vibratory pile driving hammers for pile
determined. The penetration path in the depth vs installations and soil improvement projects. In Offshore
average depth plane should be used more to discern Technology Conference. OnePetro.
the processes leading to the degradation in the cyclic Kluger, M. O., S. Kreiter, F. T. Stähler, M. Goodarzi,
soil resistance, to finally come to reliable driveability T. Stanski, & T. Mörz (2021). Cone penetration tests in
predictions. dry and saturated ticino sand. Bulletin of Engineering
Geology and the Environment 80(5), 4079–4088.
Kramer, S. L. (1996). Geotechnical earthquake engineer
ACKNOWLEDGEMENT ing. Pearson Education India.
Lagioia, R., A. Sanzeni, & F. Colleselli (2006). Air, water
The authors acknowledge the support of the project and vacuum pluviation of sand specimens for the triaxial
apparatus. Soils and foundations 46(1), 61–67.
“VCPTu2PDA”, FKZ: 03EE3025A by the Federal Massarsch, K. R., B. H. Fellenius, & A. Bodare (2017).
Ministry for Economic Affairs and Energy (BMWi). Fundamentals of the vibratory driving of piles and sheet
We thank, Wolfgang Schunn, Lukas Urbainczyk for piles. geotechnik 40(2), 126–141.
invaluable technical support with upgradation of Naumann, M., C. Schnabel, J. Fritz, & D. Djuren (2013).
MARCC and VCPT cone, and also Hammed Ade Erstellung von baugrundabschnitten in der deutschen
niyi, Atakan Acar for assisting in lab work. nordsee. Geopotential Deutsche Nordsee Modul B 9,
1–22.
O’Neill, M. W., C. Vipulanadan, & D. O. Wong (1990).
REFERENCES Evaluation of bearing capacity of vibro-driven piles
from laboratory experiments. Transportation Research
Al-Sammarraie, D. (2020). Vibratory cone penetration test Record (1277).
to investigate cyclic soil behavior in-situ. Ph. D. thesis. O’Neill, M. W. & C. Vipulanandan (1989). Laboratory
Baldi, G. et al. (1982). Design parameters for sands from evaluation of piles installed with vibratory drivers.
cpt. Number 316.
Barkan, D. D. (1962). Dynamics of bases and foundations. Rodger, A. & G. Littlejohn (1980). A study of vibratory
McGraw-Hill Companies. driving in granular soils. Geotechnique 30(3), 269–293.
Cudmani, R. & S. Manthey (2019). A novel vibro Stähler, F. T. (2020). Cone Penetration Tests in a Small
penetration test (vpt) for the investigation of cohesion- Volume Calibration Chamber: Effects Related to Sand
less soils in the field. Soil Dynamics and Earthquake Type, Saturation State, Cyclic Pre-loading, Vibratory
Engineering 126, 105760. Penetration Mode, and Boundary Condition: Kumula
D7181, A. (2020). Standard Test Method for Consolidated tive Dissertationsarbeit. Ph. D. thesis, Universität
Drained Triaxial Compression Test for Soils. ASTM Bremen.
International, West Conshohocken, PA,. Tsouvalas, A. (2020). Underwater noise emission due to
Dahl, P. H., C. A. de Jong, & A. N. Popper (2015). The offshore pile installation: A review. Energies 13(12),
underwater sound field from impact pile driving and its 3037.
potential effects on marine life. Acoustics Today 11(2), Viking, K. (2002). Vibro-driveability-a field study of vibra
18–25. tory driven sheet piles in non-cohesive soils. Ph.
Dierssen, G. (1994). Ein bodenmechanisches Modell zur D. thesis, Byggvetenskap.
Beschreibung des Vibrationsrammens in körnigen Wang, H. (1994). Experimental study and finite element
Böden. Ph. D. thesis. Karlsruhe 1994. (Veröffentlichun analysis of drivability and static behavior of various
gen des Institutes für Bodenmechanik und Felsmechanik piles installed by vibratory driving. Ph. D. thesis, Uni
der Universität Fridericiana in Karlsruhe. 133.) Fak. versity of Houston.
f. Bauingenieur- und Vermessungswesen, Diss. Wong, D., M. W. O’Neill, & C. Vipulanandan (1992).
v. 9.7.1993. Modelling of vibratory pile driving in sand. Inter
Fischer, J., H. Sychla, J. Bakker, L. de Neef, & national Journal for numerical and analytical methods
J. Stahlmann (2013). A comparison between impact in geomechanics 16(3), 189–210.
831
© 2022 the Author(s), ISBN 978-1-032-31259-0
Application of a new qc averaging approach for end bearing of driven piles

in sand
E.J. Bittar, Y. Tian & B.M. Lehane
The University of Western Australia, Perth, Australia
ABSTRACT: CPT-based methods to estimate axial pile capacity generally relate the pile base resistance
with a qc value averaged in the vicinity of the pile tip (qc,avg) and hence implicitly acknowledge the greater
zone of influence of a full scale pile compared to a cone. To account for layered deposits (e.g. loose over
dense sand), most common designs methods average the cone tip resistance over 1.5 diameters (D) above and
below the cone tip or use the Dutch methodology where the averaging is conducted over a zone of 0.7D to
4D below the pile tip and 6D to 8D above the pile tip. This paper examines an alternative averaging approach
based on an algorithm which allows estimation of the steady state end bearing resistance of penetrometer with
the same diameter as a pile (qp). End bearing stresses determined at a base displacement of 10% of the pile
diameter (qb0.1) in a database of instrumented static load tests on driven piles are compared with the corres
ponding qp values. It is shown that the qb0.1/qp ratio varies with the effective area ratio of driven piles and is
independent of the pile diameter. The best-fit equation of the database of end bearing measurements provide
a rational and improved means of determining end bearing of driven piles.
1 INTRODUCTION The need for use of an appropriate average qc value

(qc,avg) for estimation of pile end bearing is acknow
1.1 CPT and pile end bearing resistance ledged in current CPT methods. Most of the CPT-
based methods use the LCPC method (Bustamante &
Cone Penetration Test (CPT) data are widely used
Gianeselli, 1982) where qc,avg is the average qc value
directly in the design of deep foundations. The simi
in the zone 1.5D above and below the pile tip (qc,1.5D).
larities between a cone and a closed-ended displace
However, there is a concern that the zone of influence
ment pile in terms of their geometry and installation
in the LCPC method (±1.5D from the pile tip) is too
mode provide the basis for direct correlations.
small. Lehane (2019), for example, describes a case
The steady-state penetration resistance for
history where this averaging approach over-predicted
a 36 mm diameter cone (qc) and that of a field scale
the capacity of a closed-ended pile driven a distance of
displacement pile (qb) can be expected to be theoret
2.5D into a sand stratum underlying a soft estuarine
ically equivalent in a homogeneous soil. However,
clay.
natural soils comprise layers with different proper
This paper examines an alternative qc averaging
ties and thicknesses leading to a scale effect. This
approach based on an algorithm, devised by Boulan
scale effect arises because of the greater sensing and
ger & DeJong (2018), which allows estimation of
development distances of the larger diameter pile.
the steady state end bearing resistance of
The steady-state end resistance profile of a pile (qp)
a penetrometer with the same diameter as a pile (qp).
is therefore a ‘smoothened’ or filtered variant of the
Laboratory and field experiments are described that
actual CPT qc profile (Bittar et al., 2020a).
assess the applicability of this approach. qb0.1 values
CPT design methods for driven piles generally
measured in a database of instrumented static load
assume that the base resistance at a pile base settle
tests on driven piles are then compared with the cor
ment of 10% of the diameter (qb0.1) is directly propor
responding qp values with a view to determining
tional to the cone tip resistance qc. The constant of
a more rational means of estimating the end bearing
proportionality (α or kc) is less than one due to effects
of driven piles that allows for scale effects.
of partial mobilization (i.e. a settlement of 10% of the
diameter, D, is not sufficient to attain steady state
1.2 Alternative averaging technique
conditions) and due to effects of partial embedment
(as the pile has not penetrated far enough to generate Boulanger & DeJong (2018) proposed a procedure to
a steady state resistance). account for multiple thin-layer effects by considering
DOI: 10.1201/9781003308829-123
832
cone penetration as an inverse problem, assuming that
the measured qc is equal to the “true” qc (qt) convolved
with a depth-dependent spatial filter (Yost et al., 2021).
The filter is depicted in Figure 1 (where z’ is the dis
tance from the cone tip normalized by the penetrom
eter/pile diameter) and encapsulates experimental
observations via filter parameters to represent the
dependency on layers in the vicinity of the cone tip on
(i) the z’ value, (ii) the relative strengths (or qt values)
of adjacent layers and (iii) whether the cone is above
or below a given layer (i.e. negative or positive z’
value).
Bittar et al., (2020a) present experimental obser
vations showing that the procedure can provide
a reasonable estimate of the steady state end resist
ance of a full-scale pile (qp) value. Further compari
sons with experimental observations in the
laboratory and field are presented here before exam
ining the application of the approach for estimating
the end bearing resistance of driven piles in sand.
Figure 1. Normalized cone penetration filter versus nor

malized depth from the cone tip with lines for qtz’/ qtz’=0 =
0.01, 0.1, 1, 10, and 100 (Boulanger & DeJong, 2018).
2 EXPERIMENTAL ASSESSMENT OF
BOULANGER & DEJONG (2018)
2.1 Laboratory penetrometer testing

A number of penetration tests were conducted in
a cylindrical 400 mm diameter, 400mm high steel
pressure chamber. Full details of the test set-up and
chamber are provided in Tian & Lehane (2022).
An illustration of the effects of development
Figure 2. (a) CPT resistance in uniform and layered pro
length and layering on the penetration resistance
files, (b) measured and predicted CPT resistance in sample
measured by a 7mm diameter cone is shown on with dense sand underlying medium dense sand. Note cone
Figure 2. The penetrometer resistances in the diameter =7mm and vertical effective stress=50 kPa.
medium dense and dense sands (with respective rela
tive densities, Dr, of 72% and 97%) attain a steady
state after a penetration of about 6D in sand with Dr
=72% and about 15D in that with Dr=97%; the ten (2005), Tehrani et al., (2017) and others. Figure 2a
dency for greater development distances in denser also plots the penetration resistance in a layered pro
sand has been observed by Ahmadi & Robertson, file with medium dense sand overlying dense sand. It
833
is seen that the penetrometer senses the dense sand instrumented jacked piles reported by Lim & Lehane
at a distance of about 1D above the interface (2014).
between the two layers and then The end resistance corresponding to the 35.7mm
requires a penetration of about 8D before reaching cone was converted to a true penetration resistance
the steady state resistance of the dense layer. (qt) using the Boulanger & DeJong (2018) default
The calculated penetration resistance, assuming parameters (noting qt is equivalent to the end resist
steady state qc values of 10 MPa and 18 MPa in the ance measured with an infinitesimally small cone).
medium dense and dense sand respectively is com This qt profile was then used as input to calculate the
pared with the measured response on Figure 2b. These penetration resistance of the 60mm diameter pile.
calculations were performed using the default param The calculated and measured profiles of end
eters proposed by Boulanger & DeJong (2018) and resistance are plotted on Figure 3. This comparison
lead to reasonable, although certainly not perfect, shows that the Boulanger & DeJong (2018) calcula
predictions. tion for the 60mm diameter pile is a reasonable
approximation to the measured profile, albeit over
estimating resistance by up to 10% at around 0.8m
2.2 Field scale penetrometer testing
depth. However, as seen on Figure 3, the calculated
Penetrometer tests were carried out at the University resistance is a substantial improvement on the stand
of Western Australia (UWA) Shenton Park Field Sta ard approach of averaging qc values within 1.5D of
tion. This site, which comprises a 6m deep deposit the penetrometer tip.
of Aeolian sand, has been used for a range of studies The comparisons made on Figures 2 and 3 pro
with details reported in Bittar et al. (2020b), Lehane vide evidence in general support of the Boulanger &
et al. (2004), and elsewhere. The penetrometer tests DeJong (2018) algorithm for prediction of penetrom
plotted on Figure 3 at this site were conducted using eter resistance in layered sands. The value of qp
a standard 35.7mm diameter cone and a 65mm diam determined using this approach is equivalent to the
eter closed-ended pile. Both devices were pushed at bearing resistance of a pile with the same diameter
the standard rate of 20mm/s and correction for shaft as a penetrometer and is considered a rational means
friction on the pile to allow derivation of end bearing of determining an average qc value in the vicinity of
from a head load cell was determined with a high a pile tip (qc,avg).
level of confidence using results from fully
3 DATABASE ANALYSIS
The feasibility of employing qp as a measure of the

average qc value near a pile tip (qc,avg) for estimating
qb0.1 is examined here using databases of end bear
ing capacity measurements (qb0.1) for closed and
open ended driven piles. These databases, which are
summarised in Tables 1 and 2, are an extension
of those presented by Xu et al. (2008) and were used
in the development of the ‘Unified’ CPT-based
method for axial pile capacity calculation of driven
piles in sand (Lehane et al., 2020).
3.1 Closed-ended driven piles

The Boulanger and DeJong (2018) algorithm was
employed to determine qp for each database pile
using the published default values of the algorithm
(as employed for Figure 2 and 3). These qp values
are assumed equivalent to qc,avg values where they
are compared on Figure 3 with qb0.1 values measured
for the closed-ended pile database. A proportional
relationship is observed for closed-ended piles,
where:
Figure 3. Measured and predicted resistance at the site:

cone tip resistance, qc,1.5D, Boulanger & DeJong method The coefficient of variation for ratios of measured
and 65mm pile tip resistance. qb0.1 values to those calculated using Equation (1) is
834
Table 1. Database of qb0.1 measurements for closed-ended piles.
D L qb0.1 qc,tip qp
Num. Site name m m MPa MPa MPa qb0.1/ qp qb0.1/qc,1.5D
1 Akasaka 0.2 11 15.18 26.3 29.66 147.4 0.64

2 Drammen 0.28 8 1.14 2.86 2.72 89.5 0.42
3 Drammen 0.28 16 1.79 5 5.2 177.2 0.36
4 Hoogzand 0.36 6.8 13.96 41.16 40.91 97.5 0.38
5 Hsin Ta 0.61 34.3 3.11 7.47 5.77 310.5 0.48
6 Hunter’s P 0.27 9.2 4.97 8.6 8.28 99.6 0.64
7 Kallo 0.91 9.7 8.96 27.9 27.3 108.9 0.46
8 Kallo 0.54 9.7 10.69 28.1 29.3 109.1 0.48
9 Kallo 0.62 9.8 9.73 29.2 30.1 110.2 0.43
10 Kallo 0.82 9.8 9.22 28.9 29 110 0.44
11 Kallo 0.41 9.3 10.74 25.9 25 105.3 0.5
12 Kallo 0.61 9.4 8.55 26.5 24.9 105.7 0.43
13 Pigeon R 0.36 6.9 10.96 20.25 19.94 89.6 0.57
14 Wadinxveen 0.4 9.8 6.28 8.42 8.64 73.77 0.74
15 Rio 0.7 26.5 5.46 15.07 13.63 232.05 0.46
16 Bennett 0.61 45 7.53 14.63 12.16 409.65 0.7
17 Hampton 0.69 16.8 1.85 7.17 6.99 161.16 0.35
18 Fittja Sraits 0.27 12.8 2.36 6.19 5.62 134.85 0.43
19 Sermide 0.51 35.9 10.17 16.4 16.3 315 0.61
Average 0.50 0.45
ST. dev 0.11 0.13
COV 0.22 0.29
Table 2. Database of qb0.1 measurements for open-ended piles.
D L qb0.1 qc,tip q± 1.5D qp
Num. Site name m m IFR Are MPa MPa MPa MPa qb0.1/ qp qb0.1/ qc,1.5D
1 Dunkirk 0.32 11.3 0.45 0.65 7 26.7 24 20.7 0.34 0.29

2 Dunkirk 0.32 11.3 0.48 0.59 6.1 26.7 24 20.7 0.3 0.25
3 Euripides 0.76 30.5 0.99 0.19 12.3 61.5 60.8 60.2 0.21 0.2
4 Euripides 0.76 38.7 0.9 0.26 9.9 50.8 50.8 50.8 0.2 0.2
5 Euripides 0.76 47 0.89 0.27 15.3 65.9 66.4 63 0.24 0.23
6 Euripides 0.76 46.7 0.82 0.33 16 63.3 63.1 63 0.25 0.25
7 Euripides 0.76 8.5 0.99 0.19 9.6 62.2 60.9 58.6 0.16 0.16
8 Euripides 0.76 16.7 0.9 0.26 9.6 51.4 51.2 51.3 0.19 0.19
9 Euripides 0.76 25 0.89 0.27 15.5 63.6 66.9 61.9 0.25 0.23
10 Hoogzand 0.36 7 0.66 0.45 10.9 37.7 42.3 32.9 0.33 0.26
11 Hoogzand 0.36 5.3 0.77 0.39 11.2 45.5 45.5 35.4 0.32 0.25
12 Pigeon 0.36 7 0.8 0.46 5.9 19.5 19.7 20.5 0.29 0.3
13 Rastanajib 0.76 25 1.13 0.09 12.3 77.1 86.8 85.9 0.14 0.14
14 Shanghai 0.91 79 0.8 0.27 5.9 23.3 23.3 23.2 0.25 0.25
15 Shanghai 0.91 79.1 0.85 0.22 5.1 23.3 23.3 23.2 0.22 0.22
16 Lafayette 0.66 31 0.7 0.5 7.4 25 26.5 25 0.29 0.28
17 Tokyo P Bay 1.5 73.5 1 0.07 8.9 80.7 84.4 72 0.12 0.11
18 Tokyo P Bay 1.5 86 1 0.07 6.4 42.1 47.9 41 0.16 0.13
19 Tokyo 2 30.6 1.08 0.01 2 30.4 23.9 21.7 0.09 0.08
835
Figure 4. End bearing values qb0.1 vs qc,avg for Figure 5. Qb0.1/qp variation with effective area ratio for
closed-ended piles with qc,avg equal to qc,1.5D and qp. open-ended piles.
0.22. Figure 3 also compares measured qb0.1 values Deq. A consistent linear increase in qb0.1/qp with Are
with average qc values taken equal to qc,1.5D Greater is apparent on Figure 5 with an extrapolated qb0.1/qp
scatter in the proportional relationship is seen and ratio at Are = 1 corresponding to the best estimate for
there is a clear tendency for the qb0.1/qc,1.5D ratio to closed-ended piles on Figure 4. The best-fit linear fit
reduce with increasing pile diameter. to the database is:
3.2 Open-ended driven piles

The mode of penetration of a pipe pile can be fully
plugged, partially plugged or fully coring. The
degree of sand displacement near the base can be
described by the final filling ratio (FFR), which is This best-fit line is seen to represent all the
taken to be the average incremental filling ratio data with good accuracy except for the case with
(IFR) over the final 3 diameters of installation. The the very low Are value of 0.01 which is believed to
effective area ratio of a given open-ended pile (i.e. be affected by the proximity of a low strength clay
the ratio of the displacement induced to that of near the pile base (Randolph, 2003). Further ana
a fully plugged pile) is then given as (where Di is the lyses showed that Equation (5) is much better per
internal pile diameter): forming that an equivalent relationship using
qc1.5D (see data points on Figure 5) and also has no
bias with respect to pile diameter and relative
density of the sand.
When FFR data were not reported for the database 4 CONCLUSIONS
piles, the value of the FFR was estimated using the
following equation proposed by Lehane et al. This paper demonstrates that the Boulanger &
(2020), where dcpt=35.7mm DeJong (2018) algorithm provides a reasonable
means of predicting pile/penetrometer end resist
ance in layered stratigraphy. This algorithm is
then used to deduce steady state end bearing
resistances for penetrometers with the same
diameter or equivalent diameter of piles. This
The equivalent pile diameter causing the same level steady state resistance (qp) is shown to be dir
of displacement is then obtained as: ectly proportional to the measured end bearing
of a database of driven piles at a displacements
of 10% of their diameter (qb0.1). A simple equa
tion which includes an effective area ratio term
as well as qp is proposed (Equation 5) and is
Values of qp=qc,avg were calculated for each considered to provide a more rational and reli
open-ended pile in the database using the Boulanger able means of assessing pile base resistance in
and DeJong (2018) algorithm and a pile diameter of sand.
836
ACKNOWLEDGMENTS Lehane, B.M., 2019. EH Davis Memorial Lecture (2017)
CPT-based design of foundations. Australian Geome
The first and second author acknowledge the support chanics Journal, 54(4), pp.23–45.
of the Australian Postgraduate Award scheme at The Lehane, B.M., Liu, Z., Bittar, E., Nadim, F., Lacasse, S.,
University of Western Australia. Jardine, R., Carotenuto, P., Rattley, M., Jeanjean, P.,
Gavin, K., Gilbert, R., Bergan-haavik, J., & Morgan, N.,
2020. A new CPT-based axial pile capacity design
method for driven piles in sand. 4th International Sym
REFERENCES posium on Frontiers in Offshore Geotechnics, ISFOG-4.
Lehane, B.M., Ismail, M.A., & Fahey, M., 2004. Seasonal
Ahmadi, M. M., & Robertson, P. K. (2005). Thin-layer dependence of in situ test parameters in sand above the
effects on the CPT qc measurement. Canadian Geotech water table. Geotechnique, 54(3), 215–218.
nical Journal, 42(5), 1302–1317. Lim J.K. and Lehane B.M., 2015. Time effects on the shaft
Boulanger, R.W., & DeJong, J.T., 2018. Inverse filtering capacity of jacked piles in sand. Canadian Geotechnical
procedure to correct cone penetration data for thin-layer Journal, 52(11),1637–1648.
and transition effects. Cone Penetration Testing 2018: Randolph, M.F., 2003. Science and empiricism in pile
Proceedings of the 4th International Symposium on foundation design. Géotechnique, 53, 847–875.
Cone Penetration Testing (CPT’18), 21-22 June, 2018, Tehrani, F. S., Arshad, M. I., Prezzi, M., & Salgado, R.
Delft, The Netherlands, 25. (2017). Physical Modeling of Cone Penetration in
Bittar, E., Lehane, B.M., Boulanger, R.W., & Dejong, J.T., Layered Sand. Journal of Geotechnical and Geoenvir
(2020a). CPT Filter to Estimate the End Bearing of onmental Engineering, 144(1), 04017101.
Closed-Ended Driven Piles in Layered Sands. In Pro Tian, Y., & Lehane, B.M. 2022. Parameters affecting the
ceedings of the 4th International Symposium on Fron CPT resistance of reconstituted sands. Proc. 5th Int.
tiers in Offshore Geotechnics (pp. 520–528). Deep Symp. Penetration Testing, CPT22, Bologne, June 2022
Foundations Institute. Xu, X., Schneider, J.A., & Lehane, B.M., 2008. Cone pene
Bittar, E., Lehane, B., Watson, P., & Deeks, A. (2020b). tration test (CPT) methods for end-bearing assessment
Effect of cyclic history on the ageing of shaft friction of of open- and closed-ended driven piles in siliceous sand.
driven piles in sand. In 4th International Symposium on Canadian Geotechnical Journal, 45(8), 1130–1141.
Frontiers in Offshore Geotechnics. American Society of Yost, K. M., Green, R.A., Upadhyaya, S., Maurer, B. W.,
Civil Engineers. Yerro-colom, A., Martin, E. R., & Cooper, J., 2021.
Bustamante, M., & Gianeselli, L.,1982. Pile bearing cap Assessment of the efficacies of correction procedures for
acity prediction by means of static penetrometer CPT. In multiple thin layer effects on Cone Penetration Tests. Soil
Proceedings of the 2-nd European symposium on pene Dynamics and Earthquake Engineering, 144(March),
tration testing (pp. 493–500). 106677.
837
© 2022 the Author(s), ISBN 978-1-032-31259-0
A review of a CPT based axial capacity prediction of screw piles in sand

E.J. Bittar, B.M. Lehane & S. Mahdavi
The University of Western Australia, Australia
A.P. Blake, D.J. Richards & D.J. White

The University of Southampton, UK
ABSTRACT: Screw piles are widely used in onshore engineering applications and have recently been con
sidered as an alternative foundation solution for offshore wind turbines (OWTs) supported on jacket structures.
The high loads required to support such structures demand a considerable up-scaling of the screw pile geom
etry typically used onshore. Driven piles followed a similar upscaling process for their use in offshore struc
tures and the CPT tip resistance (qc) value is now commonly used directly in design methods for both onshore
and offshore driven piles. This paper evaluates the performance of a new CPT-based design method (UWA-SP
-21) to predict the axial capacity of screw piles in tension and compression. To achieve this, a database of
single helix load tested screw piles sand is collated and used for assessment of the method. The development
of this CPT design approach reflects the process that was following for driven piles. UWA-SP21 has been
developed through studies of onshore piles. The findings from this study can in the future be extrapolated
appropriately to the larger pile sizes required offshore, in the same way that driven pile design has evolved.
1 INTRODUCTION capacity is dominated by a localized ‘deep’ failure

mechanism of the helix.
Screw (or helical) piles have been increasingly used The axial capacity of screw piles is often esti
to support a variety of structures such as pipelines, mated using conventional bearing capacity theory or
transmission towers, bridges and commercial build ‘semi-empirical methods’ derived from anchor test
ings. Screw piles are steel tubes with one or more hel results from the laboratory or the field (Mitsch &
ical elements fabricated on the shaft that are screwed Clemence, 1985). These so-called ‘theoretical
into the ground through the application of torque and methods’ are limited as they require estimation of
thrust (vertical force) (Richards et al. 2019). This pile appropriate soil strength parameters and the capacity
configuration increases the axial base resistance, estimates are very sensitive to the adopted strength
allows a rapid, quiet and low-vibration installation, parameters. In situ test data are rarely used directly in
and enables reusability. These advantages have the assessment of axial capacity (Fateh et al. 2017).
increased interest in their use as an alternative founda In practice, an estimate of the screw pile capacity
tion solution for offshore wind turbines (OWTs) sup is obtained during construction via an empirical rela
ported on jacket structures. However, given the much tionship with the torque measured at the end of instal
larger scale required to meet the axial load require lation. Tsuha & Aoki (2010), Spagnoli et al. (2020),
ments of OWTs compared to current onshore applica and others, assume that the torsional resistance meas
tions, concerns have been raised regarding the large ured during installation is related directly to a pile’s
installation torque necessary for their installation axial capacity. This is because the normal forces gen
(Sharif et al., 2020; Bittar et al., 2021). erated on the helix that lead to its frictional and hence
The total axial resistance of a single helix pile, torque resistance are a measure of the axial resistance
Qult, is usually considered as a combination of the associated with a localized failure. The torsional
helix bearing resistance (Qh) and shaft resistance resistance is also needed as a check to assess if the
developed along the pile shaft (Qs). In tension, two helix has sufficient structural strength under the appli
failure mechanisms are commonly assessed depend cation of axial load (Spagnoli & Gavin 2015).
ing on the helix embedded depth (H) or embedment The prediction of the installation torque for typ
ratio (H/Dh), where H is the helix embedded depth ical onshore piles is often based on one of three
and Dh the helix diameter. These failure mechanisms methods (Davidson et al. 2020), namely (i) correl
are known as the ‘shallow’ failure (cylindrical, con ation of field measured torque with anticipated or
ical, circular) and the ‘deep’ failure. The compression measured pile capacity (Hoyt & Clemence, 1989;
DOI: 10.1201/9781003308829-124
838
Perko, 2009); (ii) modification of empirical pile cap The method assumes that a deep localized failure
acity design methods (Ghaly & Hanna 1991; Tsuha mechanism applies (i.e. Figure 1b), which is
& Aoki, 2010; Sakr, 2015) and (iii) direct correlation expected for H/Dh > 3, and draws on parallel
with the cone penetration test (CPT) end resistance methods for the similar situation of non-
(Gavin et al. 2013; Spagnoli et al. 2016; Al- displacement piles. Recommended empirical factors
Baghdadi et al. 2017; Davidson et al. 2018a). for αhi are 0.2 for compression and 0.15 for tension.
Although CPT-based capacity design methods for These factors are closely comparable to factors
screw piles have received little attention until recently, observed for footings and at the base of bored piles
such methods are now commonly used in the design of at a displacement of 10% of the foundation diameter
driven offshore piles in sand. The 22nd edition of the (e.g. Lehane 2012) and emerge, as discussed by
American Petroleum Institute (API) recommendations Lehane (2019), because of the local hemispherical
presents driven pile design methods, including the con expansion type mechanism of footings at this level
ventional (API) approach and four CPT-based methods. of displacement and the proportional relationship
Lehane et al. (2005) performed an assessment of the between qc and cavity expansion limit pressure. For
reliability of the CPT methods and noted considerable evaluation of shaft capacity (Qs), Bittar et al. (2022)
improvement in their predictive performance compared proposed a βc value of 230, which is close to the βc
with the traditional API main text approach. The com value proposed for non-displacement piles by Doan
mentary of the API design guidelines states that CPT & Lehane (2021). The contribution of Qs to Qult is
methods are “fundamentally better and show statistic relatively small.
ally closer predictions of pile load test results” than the It is of interest to compare equation (1) with other
API main text approach and are preferred to the main formulations. For example, Gavin, et al., (2013) pro
text approach (Igoe et al. 2014). posed a similar relationship but employ an αhi factor
Given the experience with offshore driven piles and of 0.065 and assume that the unit shaft friction of the
the potential use of screw piles offshore, it is surpris shaft helix is 60% of the CPT friction sleeve value.
ing that studies on the use of CPT qc to design screw This method was adopted by Spagnoli (2017) to
piles were rare until the recent studies listed above. evaluate torque predictions and it has to be noted
Bittar et al. (2022) recently presented the development that it was developed for uplift capacity only.
of one such CPT approach (UWA-SP-21) for predic Perko (2009) and Tappeden (2004) amongst
tion of screw pile axial capacity and the torque others use of the LCPC-1982 (LCPC) method pro
required for installation. This paper examines the per posed by Bustamante and Gianeselli (1982) for the
formance of UWA-SP-21 against a database of full calculation of screw pile capacity. It should be
scale field tests involving various single-helix screw noted, however, that this method was established for
pile geometries in a range of different sands. predicting the axial capacity of conventional piles,
and the applicability of the method to screw piles is
questionable.
2 CPT METHOD FORMULATIONS FOR The predictive ability of the Gavin et al. (2013)
AXIAL CAPACITY and LCPC-1982 methods is examined along with that
of UWA-SP-21 (i.e. Equation 1) in the following.
Informed by tests on instrumented screw piles, Bittar
et al. (2022) proposed a direct application using
empirical factors (βc and αh) applied to the CPT end 3 INSTALLATION TORQUE PREDICTION
resistance (qc) for the estimation of axial capacity at
a displacement of 10% of the pile diameter (Qult). The determination of Qult for helical piles in industry
The proposed formulation for ultimate capacity is: is typically based on an empirical relationship with
torque via a dimensional torque factor, Kt (Li &
Deng 2019). Hoyt & Clemence (1989) assume
a direct proportional relationship between Qult
and T as:
where Qhi is the axial capacity of a single helix (i) However, the dimensionality of the Kt factor means
separated by more than three helix diameters (Dh) that the value is likely to change as piles are
from an adjacent helix, Qs is the axial capacity of the upscaled from onshore to offshore applications. The
pile shaft, ds is the diameter of the pile shaft, Ls is same authors propose a formulation for Kt that
the length of the pile shaft, qci is the cone resistance depends on the pile shaft diameter, ds. Perko (2009)
at the level of the helix, qc,avg is average cone resist proposed the following equation which relates Kt
ance along the pile shaft. with the shaft diameter of screw piles (ds):
839
(which was typically 15 to 20% higher than the
measured load at a displacement of 0.05Dh.
It is noteworthy that existing guidance on the instal
lation of screw piles suggests that installation should
Perko (2009) collected data of Qult and T for single be performed in a pitch-matched manner (Perko 2009)
helix screw piles data and correlated these by a linear i.e. one full rotation to advance the helix by a depth
regression (Figure 1) – which is a minor simplification equal to the helix pitch (p). No specific information
of Equation 3, altering the power applied to ds to unity. regarding the advancement ratio of the database piles
The data were from tests performed on piles with helix is available although it is expected that the ratio is
diameters up to 350 mm (for which Kt ≥ 6.54, Eqn 3) somewhat less than unity as operators often allow the
and lengths up to 8 m (Spagnoli 2017). However, the auger to pull itself into the ground without the assist
torque measurements in most of this database were less ance of additional axial thrust. Characteristics of the
than 30 kNm and the Qult vs T relationship is highly screw pile database are provided in Table 1.
scattered in this region. As a consequence, Spagnoli
(2017) suggests no unique Qt - T correlation exist.
5 METHOD PREDICTIVE PERFORMANCE
The ratio of capacities calculated using UWA-SP-21

to measured capacities (Qm/Qc) was evaluated and
the predictive performance was expressed in terms
of the mean (µ) and coefficient of variation (CoV) of
the Qm/Qc ratios. For comparison purposes, the same
approach was adopted to evaluate the CPT-method
suggested by Gavin et al. (2013) and the LCPC
approach mentioned previously. Measured and cal
culated capacities and the associated statistics are
provided in Tables 2 and 3. It is observed that:
• The UWA-SP-21 method (Bittar et al. 2022) pro
vides the lowest CoV and an average value close
to unity both in compression and in tension.
• The LCPC method significantly over-predicts the
capacity of screw piles in compression (μ=0.42)
and tension (μ=0.38). This arises as the multiple
Figure 1. Torque vs. uplift capacity for previously-pub
on qc in the LCPC method varies 0.3 and 0.5
lished case studies of single-helix piles (modified from
Perko 2009). while the corresponding multiple UWA-SP-21 is
between 0.15 and 0.2.
• The high average value of the Gavin et al. (2013)
4 DATABASE OF LOAD TESTS ON SCREW approach (μ=1.37) indicates that the method
PILES IN SAND underpredicts the uplift capacity of screw piles.
The CoV of Qm/Qc for this method is also signifi
A database of static load tests on screw piles was
cantly higher than that of UWA-SP-21.
compiled by the authors. This database comprises 18
pile tests with 7 tests in tension and 11 in compres UWA-SP-21 is evidently the best performing
sion; the shafts of all piles were open ended. Many of method with an excellent CoV value for Qm/Qc of
the cases were derived from tests performed by Screw only about 10%. The method does, however, over-
Pile Australia Pty Ltd while others were sourced from predict the capacity of the database piles on average
the literature. All the piles included in the database by 8% in tension and 10% in compression.
contain CPT results in the vicinity of the test piles.
All CPT data were digitized at depth intervals
of 0.1 m. Pile capacity was defined as the load at 6 TORQUE PREDICTION ASSESSMENT
a pile head displacement of 10% of the helix
diameter (0.1Dh). The buoyant weight of piles Table 4 lists the torques measured at the end of instal
tested in tension was deducted from the max lation of the database piles (Tm). These were meas
imum head load to determine the tension cap ured using a pressure gauge on a hydraulic system.
acity. The documented cases provided by Screw The accuracy of this gauge is dependent on the rota
Pile Australia included piles that were typically tional speed applied and the hydraulic differential
loaded to a displacement of about 5% of Dh. The pressure across the gear motor. An accuracy of ±20%
measured load displacement curves were therefore is expected using this system. It is important for oper
extrapolated using Chin’s method (Chin 1970) to ators to know in advance what magnitude of torque is
determine the ultimate capacity, defined at 0.1Dh required to give confidence that the installed pile will
840
Table 1. Database of load tests on screw piles.
N Pile Site Loading direction Ls (m) ds (m) Dh (m) p (m) Reference
1 S2P3 Alberta, Canada C 4.57 0.114 0.406 0.076 Li & Deng (2019)
4 P1 Blessington C 2.61 0.11 0.4 0.1 Gavin et al. (2014)
5 SC1 Port Hedland C 6 0.219 0.6 0.1 Bittar et al. (2022)
6 SC1 Orrong Rd, Perth C 6 0.219 0.6 0.1 Bittar et al. (2022)
7 SC2 Orrong Rd, Perth C 8 0.219 0.6 0.1 Bittar et al. (2022)
8 SC1 Karrakatta, Perth C 4 0.168 0.475 0.1 Bittar et al. (2022)
9 SC2 Karrakatta, Perth C 4 0.168 0.475 0.1 Bittar et al. (2022)
10 SC1 Henry st, Perth C 5.5 0.168 0.6 0.1 Bittar et al. (2022)
11 SC1 Ellenbrook, Perth C 6 0.219 0.6 0.1 Bittar et al. (2022)
12 S2P3 Alberta, Canada T 4.57 0.114 0.406 0.1 Li & Deng (2019)
15 T9 Alberta, Canada T 4.9 0.273 0.762 NA Tappeden (2007)
16 P1 Blessington T 2.61 0.11 0.4 0.1 Gavin et al. (2014)
17 ST1 Perth airport T 6 0.323 0.6 0.1 Bittar et al. (2022)
18 ST1 Ellenbrook, Perth T 6 0.219 0.6 0.1 Bittar et al. (2022)
T=Tension; C=Compression
Table 2. CPT methods assessment for screw piles loaded in compression.
N Qc LCPC Qc Bittar et al. (2022) Qm Qm/Qc LCPC Qm/Qc Bittar et al. (2022)
1 365.4 162.1 135.0 0.37 0.83

2 303.0 131.3 114.0 0.38 0.87
3 360.8 160.9 143.0 0.40 0.89
4 858.3 415.0 420.0 0.49 1.01
5 3983.5 2227.8 2123.0 0.53 0.95
6 1753.3 856.2 617.0 0.35 0.72
7 2369.2 1216.0 1027.0 0.43 0.84
8 579.0 252.0 265.0 0.46 1.05
9 831.9 359.2 345.0 0.41 0.96
10 1461.4 609.1 525.0 0.36 0.86
11 3086.1 1594.8 1399.0 0.45 0.88
Avg 0.42 0.90
CoV 0.13 0.10
Table 3. CPT methods assessment for screw piles loaded in tension.
Qt QcGavin et al. Qc Bittar et al. Qm/Qc Qm/Qc Gavin et al. Qm/Qc Bittar et al.
N LCPC (2013) (2022) Qm LCPC (2013) (2022)
12 354.0 100.8 182.8 189.0 0.53 1.87 1.03

13 296.4 45.7 104.3 88.0 0.30 1.92 0.84
14 350.9 77.3 127.1 110.0 0.31 1.42 0.87
15 4154.4 2731.6 2365.4 2025.0 0.49 0.74 0.86
16 842.9 161.7 303.6 240.0 0.28 1.48 0.79
17 1484.9 770.0 601.9 554.0 0.37 0.72 0.92
18 3036.6 794.3 1219.7 1145.0 0.38 1.44 0.94
Avg 0.38 1.37 0.89
CoV 0.23 0.33 0.08
841
develop a certain axial capacity. It is therefore of
nterest to assess if there is a relationship between
end of installation torque and axial capacity (as
described in equations 2 and 3) and, furthermore,
if the UWA-SP-21 method can allow a reasonable
estimation of final installation torque using Equa
tion (3).
The relevant data for this assessment are provided
in Table 4 where the final installation torques are cal
culated using Equation (3) either (i) from the meas
ured load test capacity, Qult,m or (ii) from the
ultimate capacity calculated using UWA-SP-21,
Qult,c.
The average ratio of measured to calculated
torque was 1.06 when applying Equation (3) to the
measured tension capacity and was 1.03 when apply
ing the same equation to the tension capacity calcu Figure 2. Measured torque vs. uplift capacity for single-
lated using UWA-SP-21. The coefficient of variation helix piles in the presented database.
of measured to calculated torque using both
approaches was about 22%, which supports the val
idity of a direct relationship between capacity and 7 CONCLUSIONS
torque as proposed by Perko (Equation 3) - at least
for the range of screw pile geometries covered by A database comprising 18 static load tests on single
the database. helix screw piles is presented. These piles had helix
Figure 2 compares the measured uplift capaci diameters ranging from 0.3m to 0.6m and lengths
ties with the final measured installation torque between 2.5m and 8m. It is demonstrated that the
from the piles in the database (Table 1 and 4). UWA-SP-21 CPT method (Bittar et al. 2022) provided
The regression analysis shows great consistency pre-dictions of the axial compression and tension cap
when compared with the results found by Perko acity of these piles that were generally within about
(2009) in Figure 1. 12% of the measured capacities.
Table 4. Measured and calculated final installation torque for each pile in the database.
N Pile Loading direction Tc from Qult,m Tc from Qult,c Tm Tm/Tc from Qult,m Tm/Tc from Qult,c
1 S2P3 C 7.4 10.0 7.0 0.95 0.70

2 S2P1 C 4.1 3.8 4.0 0.97 1.06
3 S2P2 C 6.2 5.5 5.0 0.81 0.91
4 P1 C 22.1 16.0 18.7 0.84 1.17
5 SC1 C 210.8 183.3 250.0 1.19 1.36
6 SC1 C 61.3 68.9 95.0 1.55 1.38
7 SC2 C 102.0 99.7 115.0 1.13 1.15
8 SC1 C 20.6 15.9 25.0 1.21 1.57
9 SC2 C 26.8 22.6 25.0 0.93 1.10
10 SC1 C 40.9 38.6 45.0 1.10 1.17
11 SC1 C 138.9 129.7 115.0 0.83 0.89
12 S2P3 T 10.3 10.0 9.0 0.87 0.90
13 S2P1 T 3.2 3.8 3.1 0.97 0.82
14 S2P1 T 4.8 5.5 4.2 0.88 0.76
15 T9 T 246.3 287.7 257.6 1.05 0.90
16 P1 T 12.6 16.0 18.7 1.48 1.17
17 ST1 T 78.7 92.0 103.0 1.31 1.12
18 ST1 T 113.7 129.7 115.0 1.01 0.89
Avg 1.06 1.03
CoV 0.21 0.22
Tm = measured torque; Tc = calculated torque using Equation 3, Perko (2009)
Tc from Qt, m = torque calculated from the measured Qult,m
Tc from Qult,c = torque calculated from Qult,c, the capacity calculated via Bittar et. al. (2022), see Equations 1-2.
842
This promising predictive performance is a sub screw piles for offshore jacket supported wind energy
stantial improvement on existing approaches and structures. Géotechnique. (September): 1–50
arises from the simple relationship between spread Doan, L. V., & Lehane, B. M. (2021). CPT-Based Design
footing bearing stress and CPT qc at a displacement Method for Axial Capacities of Drilled Shafts and
ratio of 10%. A combination of this method and the Auger Cast-in-Place Piles. Journal of Geotechnical and
relationship between final installation torque and ten Geoenvironmental Engineering, 147(8), 04021077.
sion capacity proposed by Perko (2009) enables Fateh, A. M. A., Eslami, A., & Fahimifar, A. (2017). Direct
CPT and CPTu methods for determining bearing cap
assessment of the torque required for installation of acity of helical piles. Marine Georesources & Geotech.,
single helix piles to an accuracy of about 25%. 35(2), 193–207.
This method has the potential for scaling up to Gavin, K., Doherty, P., & Tolooiyan, A. (2014). Field
the dimensions of offshore piles, following the investigation of the axial resistance of helical piles in
same research direction that has led to CPT-based dense sand. Canadian Geotechnical Journal, 51(11),
methods now being favoured for predicting the cap 1343–1354.
acity of driven tubular piles. This research will sup Hoyt, R. M., & Clemence, S. P. (1989). Uplift capacity of
port the adoption of screw piles as an alternative helical anchors in soil. In Proceedings of the 12th Inter
foundation solution for offshore wind turbine jacket national Conference on Soil Mechanics and Foundation
Engineering, Rio de Janeiro, Brazil, Vol. 2, pp. 1019–
structures. 1022
Igoe, D., Kirwan, L., & Gavin, K. G. (2014). Ageing
effects and CPT based design methods for driven piles
ACKNOWLEDGMENTS in sands. Int. Symposium on Cone Penetration Testing,
1073–1082.
The first author acknowledges the support of the Lehane B.M. (2012). Foundation capacity from the CPT.
Australian Postgraduate Award at The University of Keynote Lecture, Proc. 4th Int. Conf. on Geotechnical
Western Australia. and Geophysical Site Characterisation, 1, ISC4, Recife,
Brazil, 63–82.
Lehane, B.M., 2019. EH Davis Memorial Lecture (2017)
REFERENCES CPT-based design of foundations. Australian Geome
chanics Journal, 54(4), pp.23–45.
Al-Baghdadi, T.A., Davidson, C., Brown, M.J., Lehane, B.M., Schneider, J.A., and Xu, X. 2005. CPT
Knappett, J.A., Brennan, A., Augarde, C., Coombs, W., based design ofdriven piles in sand for offshore struc
and Wang, L. 2017. CPT based design procedure for tures. The University of Western Australia.
installation torque prediction for screw piles installed in Li, W., & Deng, L. (2019). Axial load tests and numerical
sand. Proc. 8th Int. Conf. on Offshore Site Investigation modeling of single-helix piles in cohesive and cohesion-
& Geotechnics. pp. 346–353. less soils. Acta Geotechnica, 14(2), 461–475.
Bittar E., Lehane B., Richards, D., Blake A., White D., Perko, H. A. (2009). Helical piles: a practical guide to
Davidson C. and Brown, M. (2021). Field investigation design and installation. John Wiley & Sons.
to evaluate the uplift capacity and installation perform Richards, D., Blake, A., White, D., J, B. E., & B.M,
ance of screw piles in sand. Proc. 20th Int. Conference L. (2019). Field tests assessing the installation perform
on Soil Mechanics and Geotechnical Engineering, ance of screw pile geometries optimised for offshore
Sydney 2021, wind applications. Proc. 1st International Screw Pile
Bittar E., Lehane B., Richards, D., Blake A., White D., Symposium on Screw Piles for Energy Applications
Davidson C. and Brown, M. (2022). CPT-based design (pp. 47–54).
method for axial capacities of screw piles in sand (in Sakr, M. (2015). Relationship between installation torque
press). and axial capacities of helical piles in cohesionless soils.
Bustamante, M., and L. Gianeselli. 1982. “Pile bearing cap Canadian Geotechnical Journal, 52(6), 747–759.
acity prediction by means of static penetrometer CPT.” Sharif, Y. U., Brown, M., Ciantia, M. O., Cerfontaine, B.,
Proc., 2nd European Symp. on Penetration Testing Davidson, C., Knappett, J., … Ball, J. D. (2020). Using
(ESOPT II), 493–500. Rotterdam, Netherlands: A.A. DEM to create a CPT based method to estimate the
Balkema. installation requirements of rotary installed piles in
Das, B. M. & Shukla, S. K. (2013). Earth anchors. Planta sand. Canadian Geotechnical Journal, cgj–2020–0017.
tion, FL, USA: J. Ross Publishing. Spagnoli, G. 2017. A CPT-based model to predict the
Davidson, C., Al-Baghdadi, T., Brown, M., Brennan, A., installation torque of helical piles in sand. Marine Geor
Knappett, J., Augarde, C., Coombs, W., Wang, L., esources and Geotechnology, 35(4): 578–585. Taylor &
Richards, D., Blake, A., and Ball, J. (2018). A modified Francis.
CPT based installation torque prediction for large screw Tappenden, K. M., & Sego, D. C. (2007, October). Predict
piles in sand. Proc. 4th International Symposium on ing the axial capacity of screw piles installed in Canad
Cone Penetration Testing, CPT 2018,: 255–261. ian soils. OttawaGeo2007 Conference (pp. 1608–1615).
Davidson, C., Brown, M.J., Cerfontaine, B., Al-Baghdadi, Tsuha, C. de H.C., and Aoki, N. 2010. Relationship
T., Knappett, J., Brennan, A., Augarde, C., Coombs, W., between installation torque and uplift capacity of deep
Wang, L., Blake, A., Richards, D., and Ball, J.D. (2020). helical piles in sand. Canadian Geotechnical Journal, 47
Physical modelling to demonstrate the feasibility of (6): 635–647.
843
© 2022 the Author(s), ISBN 978-1-032-31259-0
CPT-based liquefaction ejecta evaluation procedure

Jonathan D. Bray & Daniel Hutabarat
University of California, Berkeley
ABSTRACT: A CPT-based procedure to estimate the severity of liquefaction ejecta is presented. It employs
a liquefaction ejecta demand parameter (LD) that captures the amount of upward seepage pressure that can
produce artesian flow due to elevated excess hydraulic head and a crust layer resistance parameter (CR) that
captures the strength and thickness of the nonliquefiable crust layer. LD tends to increase systematically as
ejecta severity increases at the thick, clean sand sites, and low LD values are estimated at stratified soil sites
that did not produce ejecta. CR captures the differing performances of sites with and without a competent
crust layer overlying a thick liquefiable layer with a high LD value. The proposed LD – CR liquefaction ejecta
severity chart separates cases with severe or extreme ejecta, which have high LD and low CR values, from
cases with minor or no ejecta, which have low LD and high CR values.
1 INTRODUCTION equivalent shear strength which suppress manifest

ations of liquefaction. These parameters are used to
Ejecta-induced ground failure contributed nearly 90% estimate the amount of ejecta likely to occur.
of the liquefaction land damage in the residential area
of Christchurch, New Zealand during the 2010-2011
Canterbury earthquake sequence (van Ballegooy et al. 2 LIQUEFACTION EJECTA SEVERITY
2014). The severity of land damage increased as more
liquefied sediment was ejected. Ground-failure indices, LPI estimates the severity of liquefaction ground
such as Liquefaction Potential Index (LPI, Iwasaki damage using a linear depth-weighted integration of
et al. 1978) and Liquefaction Severity Number (LSN, the factor of safety against liquefaction (FSL) to
van Ballegooy et al. 2014), often overestimated the a depth of 20 m as
severity of sediment ejecta produced at level-ground
stratified silty soil sites in Christchurch (e.g., Maurer
et al. 2014).
Hutabarat and Bray (2021a,b) performed nonlinear
dynamic effective stress analysis (ESA) to find that
the location of liquefaction triggering, soil layer strati where F ðFSL Þ = 1 – FSL for FSL < 1.0; F ðFSL Þ = 0
fication, and vertical hydraulic conductivity (kv) profile for FSL21.0, and z is the depth below ground surface
govern the post-shaking ejecta mechanism and system (m). LPI ranges from 0 to 100, where a higher value
response of sites that contain liquefiable layers that indicates more severe liquefaction-induced ground
either produce sediment ejecta or not. The insights failure. LPI’s performance has been evaluated exten
obtained from performing ESA are noteworthy; how sively (e.g., Maurer et al. 2014). LSN estimates the
ever, it may be infeasible to perform ESA for projects severity of liquefaction ground damage as
when subsurface data are limited. Cone penetration
test (CPT)-based simplified procedures for evaluating
liquefaction triggering have proved reliable for evalu
ating liquefaction triggering. A CPT-based procedure
for estimating the severity of ejecta-induced ground
failure for level-ground conditions is presented in this where z = depth > 0 and εv = post-liquefaction volu
paper. The procedure quantifies two governing factors: metric strain estimated using Zhang et al. (2002).
liquefaction ejecta demand (LD) and crust layer resist LSN utilizes a power-law depth-weighting function
ance (CR). The LD parameter considers the excess to emphasize the relative importance of shallow
hydraulic head (hexc) and artesian water pressure that liquefied layers to cause severe ground failure. Add
can develop at a site, and the CR parameter considers itionally, Towhata et al. (2016) developed
the thickness of the nonliquefiable crust layer and its a liquefaction severity chart by comparing the LPI
DOI: 10.1201/9781003308829-125
844
values and thickness of the crust (H1) values for Tonkin+Taylor investigated liquefiable sites in
cases with and without liquefaction-induced damage. Christchurch with underestimations or overestim
Like Ishihara (1985), Towhata et al. (2016) compare ations (e.g., Cubrinovski et al. 2019). Of these sites,
liquefaction demand and crust resistance, but instead 44 well-investigated free-field level-ground sites
of using the thickness of the liquefiable layer (H2), were selected to represent various soil profiles,
they use LPI to represent liquefaction demand. ground shaking intensities, and observed ejecta
The poor performance of these liquefaction indi amounts. Their performance during the four Canter
ces in stratified silty soil sites in Christchurch is bury earthquakes: 2010 Mw7.1 Darfield, 2011 Mw
affected primarily by the limitations of simplified 6.2 February, 2011 Mw6.0 June, and 2011 Mw
liquefaction triggering procedures in capturing their 5.9 December events were considered. The 44 sites
response (e.g., Beyzaei et al. 2018 and Cubrinovski were classified as either thick, clean sand sites or
et al. 2019). Sediment ejecta is a post-shaking partially-to-highly stratified silty soil sites to evalu
hydraulic phenomenon resulting from the migration ate the effect of this attribute. Details of the sites are
and redistribution of excess-pore-water-pressure (ue) described in Hutabarat and Bray (2021b).
generated during earthquake shaking. The dissipation Figure 1 shows representative thick sand sites and
process of residual ue can trigger high-gradient stratified silty soil sites. The Robertson (2016) modi
upward seepage that induces hydraulic fracturing in fied soil behavior type index (IB) zones are used to
the upper crust layer. With a sufficiently high classify the sand-like (IB > 32), intermediate (22 < IB
hydraulic gradient, artesian flow can be produced < 32), and clay-like (IB < 22) soil responses of the
that exploits cracks in the crust layer to eject lique layers. Thick sand sites (Figure 1a-b) contain at least
fied sediment onto the ground surface. Ejecta produc 4.5 m thick of continuous sand-like and high-kv soils
tion is governed largely by the post-shaking upward within the depth of the groundwater level (GWL) to
seepage developed in the soil profile (Hutabarat and 15 m. Stratified soil sites (Figure 1c-d) contain par-
Bray 2021a,b). The amount of excess hydraulic head tially-to-highly stratified deposits of sand, intermedi
(hexc = ue / γw, where γw is the unit weight of water), ate, and clay soil layers where there is no continuous
soil’s hydraulic conductivity (kv), and the degree of sand layer thicker than 4.5 m within the GWL to
soil layer stratification are key factors that largely 15 m depth. The parameter zAB highlighted in the fig
determine the hydraulic gradient (i) and volume of ures is the thickness of the sand-like soil layer that
upward seepage in the liquefiable layer. may contribute to ejecta production if liquefaction is
ESA results indicate the hexc developed at shallow triggered measured from the top depth of zA to the
depths in sites without sediment ejecta remains low bottom depth of zB. The parameter zAB is measured
during and after shaking. Low-kv layers within from directly below the nonliquefiable crust layer to
a highly stratified deposit impede upward seepage so the bottom of the first continuous sand-like soil layer.
the generated hexc is insufficient to produce artesian CPT-based liquefaction triggering back-analyses
flow. Although liquefaction is triggered in isolated of all cases were performed using the Boulanger and
layers of loose sand at deeper depths, upward seep Idriss (2016) “BI-16” procedure. The analyses con
age is impeded by an overlying low-kv layer. Lique sidering LPI and LSN used the probability of lique
faction triggering in a deeper layer can also reduce faction (PL) of 15% to be consistent with their
the seismic demand so that liquefaction is not trig criteria. The liquefaction triggering calculations
gered at shallow depths. Conversely, Hutabarat and were performed using the median peak ground accel
Bray (2021a,b) showed that ejecta amount largely eration (PGA) values and a fines-content correction
depends on the thickness of the liquefiable high-kv (CFC) factor of 0.13 for Christchurch soil (Hutabarat
soil layer and homogeneity of the kv-profile at sites and Bray 2022). The GWL for each site was esti
with significant ejecta. The simulations indicated the mated based on nearby well records and Vp measure
high-hydraulic gradient upward seepage that flows ments. Only the FSL of soil with the Robertson
within a thick, continuous, high-kv deposit can trig (2009) SBT Index (Ic) < 2.60 is computed to calcu
ger post-shaking secondary liquefaction within the late LPI and LSN for each case using Eqs. 1 and 2.
sand layer directly beneath the crust. A thick, con The distributions of LPI and LSN for each case
tinuous, and high-kv sand site requires more time to within the two site groups categorized by ejecta sever
return to its initial hydrostatic state; thus, it can pro ity are presented in Figure 2 with box-and-whisker
duce more ejecta. Partially stratified sites with thin plots (Tukey 1977) to evaluate their performance in
and less permeable sand layers produce less upward estimating ejecta amounts at thick sand sites and par-
seepage which in turn produces no or minor-to tially-to-highly stratified soil sites. The box height
moderate ejecta (Hutabarat and Bray 2021b). represents the range of the first (Q1) to third (Q3)
quartile of the distribution and the horizontal orange
line in the box is the median value. The whiskers
3 CHRISTCHURCH LIQUEFACTION CASE extend 1.5 times the interquartile range (IQR = Q3
HISTORIES -Q1) and data points beyond the whiskers are out-
liners. LPI and LSN show a flat trend of increasing
Researchers from the Univ. of Canterbury, Univ. of ejecta amounts with increasing median values for
California at Berkeley, Univ. of Texas at Austin, and thick sand sites (Figure 2a). There is an overlap in the
845
ejecta severity. Ejecta results from high-gradient
upward seepage developed in hydraulically continuous
liquefied soil layers beneath the crust. For highly
stratified sites, the liquefiable soil layer directly
beneath the crust does not develop sufficiently high-
gradient upward seepage, either because liquefaction
is not triggered from reduced seismic demand due to
liquefaction at depth or an underlying low-kv soil layer
impedes the upward water flow during the advection
process. The intensity of the seepage is higher if
a thicker liquefied layer develops hexc higher than the
critical head to produce artesian flow and it is com
posed of a high-kv soil (Hutabarat and Bray 2021b).
Figure 1. Examples of thick sand (a-b) and stratified soil

(c-d) profiles of qc1Ncs, IB, and observed ejecta severity
(given below site name in the sequence of the DAR, CHC,
JUN, and DEC earthquakes where N = None; Mi = Minor;
Mo = Moderate; S = Severe; E = Extreme).
Figure 3. (a) Sediment ejecta mechanisms in a typical

thick sand site and (b) Artesian Flow Potential concept.
During shaking, the elevated ue increases the total

Figure 2. Box-and-whisker plots of liquefaction ejecta hydraulic head. Tokimatsu and Yoshimi (1983) pro
demand parameters grouped based on observed ejecta posed this relationship to estimate the pore pressure
severity for: (a) 96 case histories with thick sand deposits ratio (ru) as the function of FSL:
and (b) 80 case histories with stratified deposits.
boxes representing the middle 50% of their distribu

tions. There are numerous No ejecta cases above the
top whisker with values higher than the Minor and
Moderate top whisker. Also, LPI and LSN are not
reliable for highly stratified soil sites as shown in
Figure 2b. Several stratified sites have high index where they proposed to use α = 1.0. The β parameter
values but did not produce ejecta. Similar results are is set to -0.2, which gives 1/(α β) = -5.0. The inverse
obtained if the input parameters are varied. The per sine term must be calculated in radians, and ru = 1.0
formance of the new liquefaction demand parameter for FSL = 1.0. The FSL in Eq. 3 is computed using the
(LD) shown in Figure 2 is discussed next. BI-16 liquefaction triggering procedure. Once ru is
computed, the hexc at each depth is estimated using
4 LIQUEFACTION DEMAND PARAMETER
Figure 3 illustrates the conceptual framework

employed in the CPT-based procedure to estimate
846
where σ’vo = initial effective vertical stress and γw = that is at least 250-mm thick; zB = depth from the
unit weight of water (9.81 kN/m3). ground surface to the top of the shallowest soil layer
Liquefaction ejecta demand from upward seepage between the depths of zA and 15 m with Ic ≥ 2.6 that is
is estimated by the head difference between at least 250-mm thick. The depth zB will be
a specified depth and the ground surface. A thicker 15 m if there is no soil layer (at least 250-mm thick)
H1 creates a longer travel path for water to flow, with Ic ≥ 2.6 below zA. LD is calculated in these steps:
which reduces the seepage gradient and thus increases
1. Estimate zA, and zB based on the Ic profile.
the resistance parameter. Additionally, a stronger crust
2. Compute FSL at each depth.
provides more resistance against ejecta being pro
duced than a weak crust. 3. Compute ru at each depth interval using Eq. 3.
To produce ejecta, the hexc in liquefied layer must 4. Compute hexc at each depth using Eq. 4.
exceed a critical value hA which is the required hexc at 5. Compute normalized-kv weighting factor (kv/kcs)
a depth z to produce artesian flow above the ground at each depth using Eq. 5 to calculate kv.
surface to exploit cracks in the crust layer and produce 6. Compute LD using Eq. 6 where hA is defined by
a flow that transports the liquefied soil to the ground the 1V:1H line shown in Figure 3b.
surface. The Artesian Flow Potential (AFP) concept of Figure 2 shows the distribution of the computed LD
Hutabarat and Bray (2021a) assumes the hA value value for each case history with its observed severity
required to cause artesian flow at a depth z is equal to of ejecta. The results indicate:
z which is represented by the hA line with slope 1H:1V
down from the ground surface. AFP equals the red • LD has a lower variation than LSN and LPI as
area in Figure 3b between the profile of hexc and the hA reflected by its shorter box-and-whisker box for
line when hexc > hA. More ejecta are produced at sites the No ejecta cases shown in Figure 2a for the
with higher AFP values that remain high after shaking. thick sand sites. Also, 87% of these cases have
A soil layer’s kv directly influences the upward LD < 5.
flow of water in a soil column that can induce post- • There is some overlap of the LD boxes of the
shaking secondary liquefaction at shallow depths. Minor and Moderate cases shown in Figure 2a, but
The water flowing upward from a deep liquefied the median values of LD increase systematically as
layer can increase the hexc in a shallow layer if the the severity of ejecta increases. Some Minor-to-
intermediate soil layers have high kv values. Con Moderate cases have LD < 5, which overlaps with
versely, a low-kv intermediate depth soil layer with the None cases, which is likely due to the bias in
sufficient thickness can restrict the upward flow of the liquefaction triggering calculation.
water from deep liquefiable layers. To capture this • The Minor and Moderate LD boxes do not overlap
effect a normalized-kv weighting factor is employted the Severe LD box, and the Severe LD box does
where a low permeability layer lowers ejecta poten not overlap the Extreme LD box. More than 75%
tial and a high permeability layer increases ejecta of the Severe ejecta cases have LD > 15.
potential. The Robertson and Cabal (2015) CPT cor • LD resolves the overestimation problem of the
relation is adopted to estimate kv using CPT data: other liquefaction ground damage indices for the
highly stratified sites (Figure 2b). The range of the
None cases is smaller than the Minor and Moder
kv ðm=sÞ ¼ 10ð0:952-3:04Ic Þ ; when 1:05Ic 53:27 ð5Þ ate cases. The use of zAB which considers the
impedance of upward seepage by a low-kv profile
A normalized kv factor is used as the weighting func is primarily responsible for this improvement.
tion where the kv value of clean sand calculated for
Ic = 1.8 is selected as the baseline hydraulic conduct
ivity. Nearly all Moderate, Severe, and Extreme 5 CRUST RESISTANCE PARAMETER
ejecta cases investigated in this study contain a thick
deposit of clean sand with Ic ≤ 1.8. Soil layers where The crust layer resistance parameter (CR) captures
hexc exceeds hA with low Ic values and thus high kv the strength and thickness of the nonliquefiable crust
values are multiplied by factors greater than one, and layer. It captures the differing performances of sites
soil layers with high Ic values and thus low kv values with and without a competent crust layer overlying
are multiplied by factors less than one. a thick liquefiable layer with a high LD value. The
The liquefaction ejecta demand parameter (LD) is crust resistance parameter (CR) is estimated as
estimated as
where zA = depth from the ground surface to the top of

the shallowest soil layer below the GWL with Ic < 2.6
847
where H1 (m) = thickness of the layer above zA • The Severe ejecta cases generally have lower LD
until the ground surface; su (kN/m2) = shear strength values and higher CR values than the Extreme
of the crust layers estimated using the CPT data; Ko cases. CR helps distinguish the severity of ejecta
= coefficient of lateral pressure, which is assumed to by shifting a data point left or right.
be 0.5; ϕcs is the critical state friction angle which is • CR with LD helps distinguish Moderate ejecta
assumed to be 33 degrees; and Nkt = 15 in the tip cases from Severe ejecta cases at thick sand sites;
resistance (qt) correlation used for clay. A soil with whereas it is difficult to distinguish Minor ejecta
IB > 22 will have a lower su-eq than a crust layer cases from Moderate and None cases.
composed of clay defined by IB < 22, because the • The No ejecta cases generally have LD < 2.5 so
vertical effective stress is low at shallow depth. Thus, CR does not aid significantly in distinguishing
su represents the strength (and integrity) of the crust. these sites from cases with ejecta.
The LD – CR charts (Figure 4) generally estimates
6 EJECTA SEVERITY EVALUATION ejecta severity well for the 176 case histories investi
PROCEDURE gated by Hutabarat and Bray (2022). The chart
resolves the overestimation problem of other indices
The Christchurch case history data are plotted on the at highly stratified soil sites (Figure 4b). The improve
LD-CR charts shown in Figure 4 with zones for classi ment is largely because LD captures the important role
fying ejecta severity potential. The horizontal bound of low-kv layers in impeding upward seepage, which
ary line for each zone is determined qualitatively prevents ejecta production at stratified soil sites.
using the Q1-Q3 ranges of the box-and-whisker plot
shown in Figure 2. The boundaries of the ejecta zones Table 1. Severity criteria used in this study.
were adjusted for larger CR values to bring in its
effect. The classification boundaries were selected Area within Best-estimate
qualitatively to separate data to minimize bias and 20 m covered ejecta-induced
misclassifications. The bilinear boundary lines shown Category by ejecta (%) settlement (mm)*
in Figure 4 are defined by three [CR, LD] data points
that separate adjacent zones as: None 0 0
Minor <5 < 50
• None-to-Minor: [0, 2.5] [100, 2.5] & [250, 25] Moderate 5 - 20 50 - 100
• Minor-to-Moderate: [0, 6] [90, 6] & [250, 70] Severe 20 - 50 100 - 300
• Moderate-to-Severe: [0, 15] [85, 15] & Extreme > 50 > 300
[250, 150]
• Severe-to-Extreme: [0,85] [75, 85] & [200, 250] * Based on settlement measured at sites with localized
ejecta.
Examination of the data plotted on the Liquefac
tion Severity Charts shown in Figure 4 indicates:
• The 2 Extreme ejecta cases with LD > 85 and CR
< 75 are located at the top left part of the chart in
the Extreme zone.
Figure 5 . Liquefaction-induced ejecta severity chart.
7 CONCLUSIONS
Figure 4. Ejecta severity using Liquefaction Demand (LD)
and Crust Resistance (CR) parameters at: (a) thick sand The LD – CR chart shown in Figure 5 provides
sites and (b) stratified soil sites. The inserts show data for a straightforward means to assess the severity of
LD < 6 kN/m. ejecta at a level-ground, free-field site. The proposed
848
chart captures two key attributes of the problem: National Science Foundation (NSF) with Grant
liquefaction ejecta demand through LD and crust CMMI-1561932. All opinions, findings, and conclu
resistance through CR. Additionally, an estimate of sions expressed in this paper are those of the authors
ejecta-induced 1D free-field ground settlement is and do not necessarily reflect their views.
provided using the values of Table 1 based on the
liquefaction severity category zone in the LD – CR
chart shown in Figure 5. Although the LD – CR chart REFERENCES
proved useful in evaluating ejecta potential, the
chart’s accuracy is hindered by the limitations of the Beyzaei, C. Z. et al. (2018). Depositional environment
effects on observed liquefaction performance in silt
simplified liquefaction triggering procedure used to swamps during the Canterbury earthquake sequence.
perform the liquefaction assessment. SDEE, Elsevier, 107, 898303–321.
As discussed previously, the LD – CR chart may pro Boulanger, R. W. and Idriss, I. (2016). CPT-based liquefac
duce misestimations for some cases. Often dynamic tion triggering procedure. JGGE, ASCE, 142
ESA can resolve these cases of overestimation or (2),04015065.
underestimation. Overall, the proposed chart with Cubrinovski, M., Rhodes, A., Ntritsos, N., and Van
a parameter that captures liquefaction ejecta demand Ballegooy, S. (2019). System response of liquefiable
(LD) and a parameter that captures crust resistance deposits. SDEE, Elsevier, 124, 212–229.
(CR) provides reasonable estimates of ejecta severity. Hutabarat, D. and Bray, J.D. (2021a). Effective stress ana
lysis of liquefiable sites to estimate the severity of sedi
LD captures the influence of depth (represented by the
ment ejecta. JGGE, ASCE, 147 (5):04021024.
hA line) and the hydraulic conductivity contrast (repre Hutabarat, D. and Bray, J.D. (2021b). Seismic response
sented by the kv/kcs weighting factor), and CR captures characteristics of liquefiable sites with and without sedi
the resistance of the crust to discern when the crust ment ejecta manifestation. JGGE, ASCE, 147
might prevent ejecta formation. (6):04021040.
Estimating kv is challenging, and hence, estimat Hutabarat, D. and Bray, J.D. (2022). Estimating the Sever
ing kv/kcs in Eq. 6 is challenging. To achieve results ity of Liquefaction Ejecta using the Cone Penetration
consistent with this study, the Robertson & Cabal Test JGGE, DOI: 10.1061/(ASCE)GT.1943
(2015) CPT correlation should be used to estimate 5606.0002744.
Ishihara, K. (1985). Stability of natural deposits during
kv. The kv/kcs weighting factor captures the post-
earthquakes. Proc., 11th Int. Conf. on Soil Mechanics
shaking upward seepage mechanism where a high-kv and Foundation Eng., ISSMGE, San Francisco, CA,
deposit tends to produce a higher rate of upward 321–376.
seepage that can more readily erode the crust layer. Iwasaki, T., Tatsuoka, F., Tokida, K., and Yasuda, S.
The normalized-kv weighting factor incorporated in (1978). A practical method for assessing soil liquefac
LD enables it to capture important features of lique tion potential based on case studies at various sites in
faction induced upward seepage which are not cap Japan. 2nd Int. Eq. Micro. Conf, San Francisco, CA,
tured by other ground failure indices. 885–896.
The LD – CR chart attempts to assess reliably the Maurer, B. W., Green, R. A., Cubrinovski, M., and
Bradley, B. A. (2014). “Evaluation of the liquefaction
highly complex phenomenon of ejecta production
potential index for assessing liquefaction hazard in
in a practical manner. Investigation of additional Christchurch, NZ.” JGGE, ASCE, 140 (7).
liquefaction field case histories that produced dif Robertson, P. (2009). Interpretation of cone penetration
ferent amounts of ejecta is warranted to refine the tests—a unified approach. Can. Geotech. J., 46
proposed chart. Limitations include the proposed (11),1337–1355.
procedure cannot capture the system site response Robertson, P. (2016). CPT-based SBT classification system
during earthquake shaking. Additionally, the FSL – —an update. Can. Geotech. J., 53(12),1910–1927.
ru relationship used in the calculation is for fine- Robertson, P. K. and Cabal, K. L. (2015). 6th Ed Guide to
clean sand, but it is used to estimate ru for silty CPT for geotechnical engineering. Gregg Drilling Test
ing, Inc.
soils. Lastly, the upward seepage pressure is com
Tokimatsu, K. and Yoshimi., Y. (1983). Empirical correl
puted based on the unit weight of water, though the ation of soil liquefaction based on SPT N-value and
fluidized sediment will be heavier than water; how fines content. Soils and Foundations, 23(4),56–73.
ever, it is difficult to estimate its value accurately, Tonkin and Taylor (2013). Liquefaction vulnerability study.
and the unit weight of water is used to keep the Tonkin and Taylor, Ltd. prepared for NZ EQ Comm.
method straightforward. Estimating zB is important Towhata, I. et al. (2016). Qualification of residential land
as a thicker zAB may result in an unrealistically from the viewpoint of liquefaction vulnerability. SDEE,
higher LD. 91, 260–271.
van Ballegooy, S. et al. (2014). Assessment of
liquefaction-induced land damage for residential
ACKNOWLEDGEMENTS Christchurch. Earthquake Spectra J., 30(1),31–55.
Zhang, G., Robertson, P., and Brachman, R. W. (2002).
This study was funded by the Pacific Earthquake Estimating liquefaction-induced ground settlements
Engineering Research (PEER) Center through the from CPT for level ground. Can. Geotech. J., 39
Transportation Systems Research Program and the (5),1168–1180.
849
© 2022 the Author(s), ISBN 978-1-032-31259-0
Settlement estimations for buildings founded on saturated silty sands from

CPT and DMT results
Maxwell Cáceres, Javier Fumeron & Felipe A. Villalobos
Department of Civil Engineering, Universidad Católica de la Santísima Concepción, Chile
Ricardo Moffat
LMMG Geotechnics & Faculty of Engineering and Sciences, Universidad Adolfo Ibáñez, Chile
ABSTRACT: The estimation of building settlements founded in saturated sandy and silty soils is a relevant
part of foundation design. Settlement calculation methodologies are still based on SPT and plate-load tests.
SPT results can have significant dispersion and soil stiffness estimations are obtained from correlations with
out a physical meaning and valid only for particular soils and geology conditions. Moreover, plate-load tests
are normally limited to shallow depths. CPT and DMT can obtain reliable results in engineering units. These
tests are operator independent and the equipment is truly standard worldwide. Results obtained using CPT and
DMT equipment for an urban renovation project with buildings founded on saturated sands and silts in Con
cepción city in Chile, are presented. These results allow estimations of static and liquefaction-induced settle
ments obtained from calculation methods developed for CPT and DMT.
1 INTRODUCTION without major operator influence as with SPT (Lunne

et al., 1997). A CPT based method to estimate static
Bearing capacity and settlements are key components settlements of shallow foundations in sand requires
in the analysis and design of foundations. Bearing the soil modulus of deformation Es, which is nor
capacity of shallow foundations is usually calculated mally obtained from the cone tip resistance.
adopting the Terzaghi procedure which has been The flat dilatometer test DMT is another in situ
extended for different conditions (geometry, loading). test which has had a particular success in the estima
However, in silty sands the design becomes controlled tion of settlements of shallow foundations (Schmert
in most of the cases by the allowable settlement and mann, 1986; Schnaid, 2009; Marchetti, 2015). The
not by bearing capacity. Therefore, estimations of settlement estimation is mainly based on the deter
settlements are crucial in almost every project. Settle mination of the 1D dilatometer modulus MDMT.
ment estimation involves usually the determination of Schmertmann (1986) has shown 16 case histories for
the soil stiffness by means of an operational deform different structures on sand, silt, clay, peat and mix
ation modulus related to the design load range, which tures of these soils where the use of MDMT led to an
can be obtained from laboratory tests such as triaxial average and standard deviation values of the esti
tests. Alternatively, in situ tests can be carried out mated/measured settlement ratio se/sm of 1.18 and
instead or to complement the information from 0.36 respectively, with se/sm varying from 0.71 to
laboratory. In situ testing can have benefits in terms 2.23 for measured settlements between 3 mm and
of measuring directly in the soil without the hassles 2.85 m. If only sand and silt and their mixtures were
of extracting, transporting, storing and the preparation considered (9 cases), se/sm average and standard
of samples. A traditional in situ test carried out in deviation are 1.10 (0.73 – 1.34) and 0.21 for settle
projects is the standard penetration test SPT. How ments between 3 and 58 mm. Moreover, other
ever, there are concerns about its reliability and result researchers have collected more cases with favour
interpretation for soil stiffness due to poor reproduci able performance of the DMT (Hayes, 1986; Monaco
bility and lack of continuity (Robertson, 2012). In et al., 2006; Failmezger et al., 2015). For that reason,
situ load plate test can provide a scaled footing load- settlement studies based on CPT have often been
displacement response. However, its applicability is benchmarked against results from the DMT (Kag
limited to surface or shallow depths. gawa et al., 1996; Lehane and Fahey, 2004).
Cone penetration test CPT is versatile and reliable This work focuses on the Aurora de Chile urban
since results can be faster and continuously obtained recovery project located at the Bío Bío north
DOI: 10.1201/9781003308829-126
850
riverbank in the city of Concepción, Chile, which where qt is the corrected tip resistance, σv0 is the
started in 2016. The project considered the construc in situ vertical stress and αM is a modulus factor:
tion of 8 four-storey buildings of with 128 flats and If
78 two-storey houses. As part of the geotechnical
site investigation, one DMT and two CPT tests were
carried out. This offers the opportunity for the com
parison of settlement estimations based on results
from CPT and DMT. The soils found in the project
area correspond to fluvial deposits mainly of sands
and silts and a mixture of them. Static settlements If
are determined from CPT and DMT results consider
ing 1D deformation modulus for each soil layer and
different vertical load increments. Liquefaction-
induced settlement analysis based on the Ishihara
(1996) chart and the liquefaction potential are per
formed. The latter follows the modified Seed and where Ic is a soil behaviour type SBT index that can
Idriss (1971) simplified method for CPT (Youd and represent the SBT zones in the Qtn-Fr chart, Qtn is
Idriss, 2001) and DMT (Monaco et al. 2005; March the normalized tip resistance and Fr is the friction
etti et al. 2013). ratio. Ic represents the radius of a concentric circle:
2 STATIC SETTLEMENT ESTIMATION
Static settlement can be calculated based on a 1D

compression modulus also referred to as oedometric
or constrained modulus M which is commonly
obtained from consolidation tests. This modulus Þ
M corresponds to the slope of a straight line between
two points in the 1D effective vertical stress - verti where n is an exponent which varies according to the
cal strain curve, which is therefore valid for that soil type (Robertson, 2009a). αM increases from soft
stress increment applied by a structure. soils to dense granular soils with a division around
the middle for Ic = 2.2. Expression (9) is used to cal
2.1 CPT-based settlement prediction culate static settlements, where M is obtained with
(1) to (6).
The CPT test consists in the continuous penetration
through the soil of a standard steel bar with a conical
tip at a constant rate of 20 mm/s. An electronically 2.2 DMT-based settlement prediction
instrumented cone with load cells allows the meas The flat dilatometer test DMT provides subsurface
urement of the cone tip resistance qc and sleeve fric information through two horizontal pressure read
tion resistance fs. The use of a piezocone also allows ings in a circular membrane. The DMT is recognized
the continuous measurement of the pore water pres for being a suitable test to acquire information
sure u. The CPT equipment and test procedures are related to the stiffness of the soil. Moreover, it is
standardized (ASTM D5778, 2020; EN ISO 22476 also sensitive to the soil stress history, therefore, it
1, 2012). The CPT results can provide a detailed can provide reasonable estimations of parameters
record for the evaluation of the ground stratigraphy such as the coefficient of lateral earth pressure K0,
and geotechnical properties (Lunne et al., 1997; over-consolidation ratio OCR and 1D constrained
Robertson, 2009a). Settlement estimations for foot modulus M (Marchetti et al., 2001; Schnaid, 2009;
ings in sand can be carried out by methods that use Marchetti, 2015). M can be estimated according to
the drained elastic modulus E. 1D constrained modu the following expression:
lus M are generally used for the estimation of long
term consolidation settlements. However, for stresses
below the preconsolidation stress, it can be assumed
that M is approximately constant and possible to cor
relate with the net cone resistance ðqt - σv0 Þ by
means of DMT-CPT relationships represented in the where ED is the dilatometer modulus, determined by
Qtn-Fr chart (Robertson, 2009a, 2009b). the elasticity theory for the 60 mm diameter mem
brane displacing 1.1 mm.
M ¼ αM ðqt - σv0 Þ ð1Þ
851
where p0 and p1 are the corrected lift-off and full
expansion pressures, respectively. RM is a correction
factor applied to ED, which is a function of the hori
zontal stress index KD and the material index ID;
therefore, it is calculated according to the soil type.
Static settlements are calculated using the following
1D relationship:
where Δσv is the stress increment applied by build

ings or embankments in the middle of a layer of
thickness Δz, which can be adjusted using the theory
of Boussinesq for deep layers.
3 IN SITU TESTING RESULTS
Figure 1 shows the qt and fs variation with depth as

well as the soil behaviour type SBT interpretation for
the CPT profiles. The CPT1 (Figure 1a) and CPT2
(Figure 1b) tests reached 22 and 17.5 m, respectively.
The groundwater was detected at approximately
5.5 m depth from the ground level. In general, sand
layers from fluvial deposits are detected with max
imum thicknesses of 10 m mixed with silt and clay
lenses with thicknesses that vary between 0.8 to 1.5 m.
The DMT test reached a depth of 10 m and the
variation of ID and KD with depth are shown in
Figure 2. The stratigraphy is comparable to the
CPT1 profile, both coinciding in the presence of
a soft layer of clay approximately 1.5 m thick at
4 m depth. It is worth noting that when KD = 2 soils
are normally consolidated.
In Figure 3 some of the soil parameters that are
calculated based on the measurements from both test Figure 1. CPT profiles of qt and fs: a) CPT1 and b) CPT-2.
equipment are compared. Similarities can be seen
between DMT and CPT1 for the profiles of unit
weight γ and friction angle 0 , where 0 is around
35° in the first 4 m and 40° below 6 m. In Figure 4,
the 1D compression modulus M shows low values in
the first 3 m, and in the clay layer even lower values
between 2 and 6 MPa according to the DMT and
between 2 and 12 MPa according to CPT1.
Below 6 m, M increases with much larger values
with better agreement between CPT and DMT for
values in the order of 180 and 200 MPa. Figure 4 also
shows the MDMT/MCPT1 ratio, where it can be observed
that CPT1 tend to overestimate M respect to DMT.
4 STATIC SETTLEMENT CALCULATIONS
Settlements are calculated with the method previ

ously explained in 2.1 and 2.2 considering
a 2 m wide square footing and founded 2 m below
ground level. An increasing sequence of vertical Figure 2. ID and KD variation with depth from DMT
load is considered as the stress increment Δσv . results.
852
Figure 5. Shallow foundation static settlement estimated
from DMT and CPT test results.
5 LIQUEFACTION ANALYSIS
5.1 Liquefaction occurrence

Figure 3. Comparison of the variation with depth of the
Chile is a highly seismic country and in particular
unit weight γ and angle of friction 0 from DMT and CPT
data. Concepción is a region where several mega-thrust
earthquakes (Mw ≥ 8.0) have occurred (in year 1570,
1657, 1730, 1751, 1835, 1939, 1960 and 2010).
Moreover, the Aurora de Chile project site is next to
Figure 5 shows the estimated settlements for each Bío Bío River, the largest of Chile in terms of width
CPT and DMT test. It can be observed that DMT and water flow volume. For that reason, it is import
settlement estimations tend to be higher than those ant to assess the liquefaction potential of the satur
estimated with CPT. settlement estimations tend to ated sandy ground and the associated settlements.
be higher than those estimated with CPT. The procedure to assess liquefaction potential is
mainly based on the simplified method by Seed and
Idriss (1971) developed initially for SPT, from
which the cyclic stress ratio CSR imposed by the
earthquake can be estimated.
where rd is a stress reduction coefficient which account

for the reduction of amax with depth in absence of
liquefaction. The expression used of rd is too big to
include it here (see (3) in Youd and Idriss, 2001).
For CPT data, a modified, although similar empir
ical procedure presented by Youd and Idriss (2001)
has been adopted, where qt is normalized and trans
formed to:
where Kc is a grain characteristic factor that is

a function of Ic and in this form a clean sand value
qt1Ncs is obtained, from which the cyclic resistance
Figure 4. Variation with depth of the constrained modulus ratio CRR can be determined (Robertson and Wride,
M determined from DMT and CPT and the M ratio. 1998).
853
liquefaction is highly likely to occur there. It is worth
mentioning that liquefaction has been reported along
the Bío Bío river promenade during the 2010 earth
quake (Verdugo et al., 2010). Liquefaction is not
expected in the clay layer as shown for CPT2 and
DMT. However, FSL < 1 for CPT1 due to the very
low values of the clay tip resistance. It is important to
bear in mind that liquefaction does not occur in clay.
Below the clay layer liquefaction may still take
place down to 7.5 m according to the CPT results.
Subsequently, a liquefaction factor of safety FSL

can be obtained as:
where MSF is a magnitude scale factor to account

for earthquakes different from Mw = 7.5.
which results in MSF = 0.66 for Mw = 8.8.

A correction factor for confining stresses higher than
100 kPa is determined as:
where f is a factor related to the site conditions such

as relative density DR, ageing and overconsolidation
ratio OCR. In these analyses f has been assumed
either 0.65 for DR > 60% or 0.75 for DR ≤ 60%. Figure 6. Liquefaction factor of safety obtained from CPT
For DMT, FSL is also calculated using (14) with and DMT tests.
the same components as for CPT, except CRR7.5
which is determined with the following expression
(Monaco et al., 2005):
Between 7.5 and 11 m FSL > 1 for CPT results,
although for DMT between 8.5 and 9 m, FSL = 1.0
CRR7:5 ¼ 0:0107KD 3 - 0:0741KD 2 þ 0:2169KD - 0:1306 1.2, which is somewhat close to CPT2 results. This
ð17Þ zone of no liquefaction may stop propagation of lique
faction occurring below. Indeed, below 11 m FSL <
1.0 and according to CPT1 a layer of 6 m thick may
The liquefaction analyses assume Mw = 8.8 earth liquefy.
quake and a maximum acceleration amax = 0.4g,
which are actually the values recorded in the centre
of Concepción during the 2010 earthquake and nor 5.2 Liquefaction-induced settlements
mally adopted in practice. The estimation of free-field settlements caused by
Figure 6 shows the variation with depth of the earthquake-induced liquefaction is carried out usu
liquefaction factor of safety FSL determined from the ally based on the soil incremental volumetric
CPT and DMT results. The groundwater level is strain εvol Dissipation of excess pore pressure
assumed to be at 2 m as an unfavourable condition densifies the soil which may result in important
during winter when the river water level is very high. volume changes leading to settlements. A chart
A vertical line for FSL = 1 separates where liquefac proposed to estimate εvol as a function of FSL and
tion is likely or not to occur. It can be observed that N1 from the SPT or alternatively qc or DR can be
between 2 m and 3.5 or 4 m (where starts the clay used (Ishihara and Yoshimine, 1992; Ishihara,
layer) FSL < 1 for CPT and DMT results, hence, 1996). From CPT and DMT data this chart was
854
directly used for εvol = f(FSL, qt) and εvol = f(FSL, It has been found that free-field liquefaction-
DR), respectively. The DR expressions used are induced settlement can become different from that
a best fit to the DR-KD plot by Reyna and Cha occurring underneath buildings founded on shallow
meau (1991) for normally consolidated sands and foundations (Bertalot, 2011). The overburden pres
the DR-KD-OCR plot by Lee et al. (2011) for over- sure imposed by a building applies large confining
consolidated sands: stresses in the soil which tend to modify settlements.
Overburden increases settlements because of higher
loads, but only until a certain value of bearing pres
DR ¼ 50:66ln KD þ 7:95 sure owing to the reduction of excess pore pressure
ð18Þ
DR ¼ a ln KD þ b ratio with confinement (Bertalot et al., 2013).
A chart has been developed to estimate building
settlements sL based on data that include the Con
Figure 7a shows the variation with depth of the cepción 2010 earthquake (Bertalot et al., 2013). sL is
increments of εvol for each layer analysed from CPT determined based on the thickness of liquefied soil
DL, building width B and building bearing pressure
q. For B = 15 m, q = 50 kPa, sL/DL = 0.1 and consid
ering DL = 2, 5 and 7 m results in sL = 200, 500 and
700 mm, which are larger than those from free-field.
6 CONCLUSIONS
CPT and DMT tests were carried out in fluvial silty

sands in Concepción Chile. A general good agree
ment was found among the results of soil profile
interpretation, constrained modulus and resistance
parameters. Static settlements estimated through the
1D MDMT were approximately twice larger than
those estimated with CPT data. Free-field liquefac
tion-induced settlements resulted in much larger than
that from static analyses, from 100 mm (DMT) up to
150 mm (CPT) for the first 10 m. Including the con
fining stress imposed by a building increases even
more the liquefaction-induced settlements.
REFERENCES
ASTM D5778. 2020. Standard test method for electronic
friction cone and piezocone penetration testing of soils.
Figure 7. For CPT and DMT: a) incremental volumetric ASTM International, West Conshohocken, PA, USA
strain with depth and b) accumulated vertical displacement Bertalot, D. 2011. An overview on field and experimental
with depth. evidences concerning seismic liquefaction induced
settlement of buildings with shallow foundations. Obras
y Proyectos 10, 36–45
and DMT data. It can be observed that in the first Bertalot, D., Brennan, A.J. and Villalobos, F.A. 2013. Influ
2 m and below 15 m there are high values of Evol . ence of bearing pressure on liquefaction-induced settle
The three curves are relatively close to each other ment of shallow foundations. Géotechnique 63(5):
denoting similar variation trends of Evol . Figure 7b 391–399
shows the accumulated integration of Evol with depth EN ISO 22476-1. 2012. Geotechnical investigation and
which is the accumulated vertical displacement at testing. Field testing - Part 1: Electrical cone and piezo
each depth. It is clear to observe that DMT data lead cone penetration test. European Committee for Stand
to smaller settlements sL compared with CPT data. sL ardization CEN, Brussels, Belgium
increase rate is higher in the first 4 m and then stabil Failmezger, R., Till, P. Frizzel, J. & Kight, S. 2015.
Redesign of shallow foundations using dilatometer tests
ises and increases again around 11 m and 15 m for —more case studies after DMT’06 conference. 3rd Int.
CPT2 and CPT1, respectively. For the first 10 m, sL Conf. on the Flat Dilatometer DMT’15. Roma, Italy.
is estimated to be around 100 mm for DMT, whereas Hayes, J.A. 1986. Comparison of flat dilatometer in-situ
for CPT1 sL ≈ 150 mm and sL ≈ 130 mm for CPT2. test results with observed settlement of structures and
If the liquefied soil below 10 m manages to dissipate earthwork. 39th Geotechnical Conference. Ontario,
the excess pore pressure, sL values may increase Canada, 311–316.
even more to values of 300 mm at 17.5 m (CPT2) Ishihara, K. (1996). Soil behaviour in earthquake geotech
and up to 550 mm at 22 m (CPT1). nics. Oxford University Press, UK
855
Ishihara, K. & Yoshimine, M. 1992. Evaluation of settle International Conference on Recent Advances in Geo-
ments in sand deposits following liquefaction during technical Earthquake Engineering and Soil Dynamics,
earthquakes. Soils and Foundations 32(1),173–188 St. Louis, Missouri, USA, 385–392
Kaggwa, W.S., Jha, R.K. & Jaksa, M.B. 1996. Use of dilat Robertson, P.K. 2009a. Interpretation of cone penetration
ometer and cone penetration tests to estimate settlements tests – a unified approach. Canadian Geotechnical Jour
of footings on calcareous sand. 7th Australia New Zealand nal 46(11): 1337–1355
Conf. on Geomechanics. Adelaide, Australia, 909–914 Robertson, P.K. 2009b. CPT – DMT correlations. Journal
Lee, M.J., Choi, S.K., Kim, M.T. and Lee, W. 2011. Effect of Geotechnical and Geonvironmental Engineering 135
of stress history on CPT and DMT results in sand. (11): 1762–1771
Engineering Geology 117, 259–265 Robertson, P.K. 2012. Interpretation of in-situ tests - some
Lehane, B. & Fahey, M. 2004. Using SCPT and DMT data insights. In Coutinho & Mayne (eds.), 4th International
for settlement prediction in sand. 2nd International Con Conference on Geotechnical and Geophysical Site
ference on Geotechnical and Geophysical Site Charac Characterization, ISC4, Porto de Galinhas, Brazil.
terization ISC2, Millpress, Rotterdam, 1673–1679. Taylor & Francis, vol.1, 3–24
Lunne, T., Robertson, P. K. & Powell, J. 1997. Cone Pene Robertson, P.K. & Wride, C.E. 1998. Evaluating cyclic
tration Testing in Geotechnical Practice. CRC Press liquefaction potential using the cone penetration test.
Marchetti, S. 2015. Some 2015 updates to the TC16 DMT Canadian Geotechnical Journal 35(3),442–459
report 2001. 3rd International Conference on the Flat Schmertmann, J.H. 1986. Dilatometer to compute founda
Dilatometer DMT2015, Rome, Italy, 43–65 tion settlement. In Situ ‘86, ASCE Spec. Conf. on Use of
Marchetti, S., Monaco, P., Totani, G. & Calabrese, M. in situ Tests in Geotechn. Engineering. Virginia Tech,
2001. The DMT in soil investigations. ISSMGE TC 16 Blacksburg, USA, 303–321
report. Int. Conf. on In Situ Measurement of Soil Proper Schnaid, F. 2009. In situ testing in geomechanics: The
ties and Case Histories. Bandung, Indonesia, 95–132 main tests. Taylor & Francis, Abingdon:
Marchetti, S., Marchetti, D. & Villalobos, F. 2013. The Seed, H.B. & Idriss, I.M. 1971. Simplified procedure for
seismic dilatometer SDMT for in situ soil testing. Obras evaluating soil liquefaction potential. Journal of the Soil
y Proyectos 13, 20–29 (in Spanish) Mechanics and Foundations Division 97(9),1249–1273
Monaco, P., Totani G. & Calabrese M. 2006. DMT pre Verdugo, R., Villalobos, F., Yasuda, S., Konagai, K.,
dicted vs observed settlement: a review of the available Sugano, T., Okamura, M., Tobita, T. & Torres, A. 2010.
experience. DMT 2006, Washington DC, 244–252 Description and analysis of geotechnical aspects associ
Monaco, P., Marchetti, S., Totani, G. & Calabrese, M. ated to the large 2010 Chile earthquake. Obras
2005. Sand liquefiability assessment by flat dilatometer y Proyectos 8, 25–36
test (DMT). 16th International Conference on Soil Youd, T.L. & Idriss, I.M. 2001. Liquefaction resistance of
Mechanics and Geotechnical Engineering, IOS Press, soils: summary report from the 1996 NCEER and 1998
vol. 4, 2693–2698 NCEER/NSF workshops on evaluation of liquefaction
Reyna, F. & Chameau, J.L. 1991. Dilatometer based lique resistance of soils. Journal of Geotechnical and Geoen
faction potential of sites in the Imperial Valley. 2nd vironmental Engineering 127(4),297–313
856
© 2022 the Author(s), ISBN 978-1-032-31259-0
Numerical modeling of static load test in drilled shaft using CPTu results
M.A. Camacho
Independent Engineer, Cochabamba, Bolivia
C.B. Camacho
Universidad Mayor de San Simón, Cochabamba, Bolivia
V.H. Miranda
Independent Engineer, Cochabamba, Bolivia
ABSTRACT: The projects with deep foundations use information from field tests (e.g. CPTu and static load
test) to predict an ultimate capacity. The static load test verifies foundation design parameters and installation
method. However, this test establishes the real behavior of the drilled shaft under specific conditions and the
possibility of experimenting in different conditions becomes extremely expensive and unfeasible. Therefore,
the numerical modeling of the static load test becomes important because the initial soil profile and dimen
sions of the drilled shaft could be changed. This article presents the comparison of the load–displacement
curves by numerical modeling in fine soils with that load–displacement curve through a static load test in
a drilled shaft built in Cochabamba, Bolivia.
1 INTRODUCTION soil unit weight, the friction angle, and the undrained
shear strength. This data is needed for the model.
The objective of a load test is to verify that the pile
prototype in real conditions of the underground has
2.1 Water table
enough bearing capacity. However, this test estab
lishes the real behavior of the pile in singular condi The CPTu did not find any water table (the pore pres
tions, and the possibility of testing in different settings sure transducer has been damaged at the beginning of
make them extremely expensive and infeasible. the test), however; it was evidenced that it was 2
The necessity of developing mathematical models meters below the surface, because an excavation has
to predict the behavior of a pile under different condi been carried out to that depth.
tions is evident. The implementation of a numerically
model of load test allows to explore different settings
2.2 Soil classification
demanding a little effort during its execution. These
models require detailed study of the field conditions Figure 1 shows the evaluation of the CPTu test results
in a load test to be modeled. using the ISBT behavior index proposed by Robertson
The project “Nudo Viario Beijing” in the city (2010). The characterization of the soil based on the
of Cochabamba (Bolivia) is a project of great fines content proposed by F. Yi (2014) allowed an
magnitude that its infrastructure required the identical characterization to that obtained utilizing the
execution of load tests (e.g., CPTu and static soil type behavior indexes (ISBT), moreover, being
load tests). The present document analyzes the known the percentage of fines content present in the
numerical model for the load test of a pile layer, it allowed a precise definition of the type of soil.
through the interpretation of results obtained There are 15 strata in the underground profile, 8
from the field test using FB-Multipier v.5.5 of them of clay and 7 of silt.
software.
2.3 Unit weight
2 GEOTECHNICAL MODEL Within the evaluation process of the unit weight from
the CPTu results, it was applied the equations proposed
The geotechnical model based on the CPTu results by Mayne (2012); Robertson & Cabal (2010). Figure 2
analyzes the water table, the soil classification, the indicates the estimation of the unit weight values.
DOI: 10.1201/9781003308829-127
857
The principal geotechnical parameters are indicated.
The unit weight and undrained shear strength have
been defined in low, medium, and high values for
each clay strata. For silt strata, it was defined the
friction angle with constant values between 30°
and 36°.
2.6 Selection of geotechnical parameters

The unit weight and undrained shear strength
data in a soil stratum vary in depth, which value
is representative? The comparison between these
parameters has been made (low, medium, or high
values)
Figure 1. ISBT behavior index proposed by Robertson

(2010).
2.4 Undrained shear strength

Figure 3 indicates the variation of the undrained
shear strength results depending on the depth. The
equation proposed by Mayne & Kemper (1988) to
evaluate the undrained shear strength is based on the
Nkt factor:
where qt is the resistance to the corrected tip σvo is

initial vertical total stress Nkt is a factor that varies
between 10 to 20 (Ameratunga et al., 2016). The
results of Nkt = 12.76 have been used.
2.5 Geotechnical model based on the results of the

CPTu test
In summary, Figure 4 shows the geotechnical model Figure 2. Unit weight estimation from CPTu results.
based on the CPTu results for the numerical model. (Mejia, 2016).
858
3 LOAD TEST 4.1 Numerical model in FB–Multipier software
In order to get the geotechnical model of the under
The static compression load test was performed on
ground, it has been carried out the numerical model
the top of the drilled shaft based on ASTM D1143–
of a compression load test of a drilled shaft in the
07. Figure 5 shows the results obtained reaching up
FB – Multipier software. The principal parameters
to 4347 kN with a registered maximum displacement
used are the unit weight and the undrained shear
of 17,89 mm (1,50% strain based on the drilled shaft
strength since these values influence the t–z and q–z
diameter).
curves which are based on the research made by
Reese & O’Neill (1988) and Wang & Reese (1993)
for drilled shafts.
4 SOIL–STRUCTURE INTERACTION IN FB–
MULTIPIER SOFTWARE
Due to the fact that the models for the soil and piles
are not lineal, FB–Multipier carries out an iterative
solution process through a secant method for solving
the non-linear equations. The software uses
a Newton–Raphson iteration scheme where in each
iteration the stiffnesses of the soil and piles are used
to form the stiffness matrix and calculate the deform
ations. These deformations were used then to find
the internal loads in each discretization of the piles.
Figure 4. Geotechnical model based on the CPTu test

results.
4.1.1 Global data

In the FB–Multipier v.5.5 software, the numerical
Figure 3. Undrained shear strength estimation (Mejia, 2016). models based on the conceptual geotechnical model
859
have been carried out in a drilled shaft simulating The results of the sensitivity study of the geotech
the compression load test for stages. nical parameters that influence the t-z and q-z curves
are presented in Figure 6.
4.1.2 Pile data The general result of the geotechnical parameters
The drilled shaft considered in this study is of diam shows a proportional direct relationship between the
eter 1,20 m, length 22 m, and made of concrete with parameters and the peak load. The higher the
a resistance of 41,37 MPa. The drilled shaft has been undrained shear strength and unit weight, the higher
built with 24 steel bars of 25 mm. All of these the peak load. Three groups of curves have stood out
parameters have been included in the FB-Multipier due to undrained shear strength, being the geotechnical
software. parameter that defines a difference in peak loads in
a range of 3000 kN. The unit weight, on the other
4.1.3 Soil parameters hand, varies the peak loads in a range of 300 kN. The
The soil´s parameters were defined according curves with the highest correlation to the static load
to the geotechnical model. It is indicated the test curve are the CPT02, CPT05, and CPT08. Accord
phreatic level and each stratum was defined with ing to Table 1, all the strata have a medium undrained
the unit weight and undrained shear strength shear strength and each curve represents a low
data. (CPT02), medium (CPT05), and high (CPT08) unit
weight in all strata.
Table 1. Curves’ names.
Undrained shear strength

Unit weight Low Medium High
Low CPT07 CPT08 CPT09

Medium CPT04 CPT05 CPT06
High CPT01 CPT02 CPT03
5 RESULT ANALYSIS
The analysis has been done from a geotechnical

model with low, medium, and high values of unit
weight and undrained shear strength. A model has
Figure 5. Load test results (Camacho & Gomez, 2016). been developed in the FB-Multipier software obtain
ing 9 load-displacement curves that have been com
pared with the executed static load test curve. From
this group of 9 curves, it has been chosen 1 curve
4.1.4 Deformations in the model which best represents the load test in a numerical
According to the results of the static load test, it has model.
been input the displacements in the head of the drilled
shaft.
5.1 Conceptual geotechnical model analysis
Two stages of displacements were input. First,
displacements between 0,20 mm to 6,95 mm with The CPTu test indicates a geotechnical profile as
a step of 0,15 mm, and secondly, displacements intercalation of fine-grained saturated soils, predom
between 6,95 mm to 20,15 mm with a step of inantly lean clays with sand (8 strata) with intercal
0,30 mm. Ninety load states were obtained with ations of sandy silt strata (7 strata).
these displacements. The low, medium, and high values of the unit
weight have been obtained with the correlation of
the CPTu proposed by Robertson & Cabal (2010).
4.2 Results of the numerical model
In Table 1, the numerical model has been labeled
5.2 Load–displacement curves
with the names described for each case. For
example, for a low undrained shear strength and From the FB–Multipier software it has been obtained
a high unit weight, it has been called CPT01 curve. load–displacement curves. The static load test curve
860
Figure 6. Load test and numerical models results based on the CPTu test.
Figure 7. Load – Displacement curves from CPTu test.
plot is a group of points joined by lines. It has been 6 CONCLUSIONS

observed a relationship and curvy trend between
a load of 635,17 kN to 4.045,71 kN, whereas the The geotechnical model extends from the surface to
lastest load and displacement data indicates an a depth of 25,50 m. where the lean clay strata with
abrupt drop in drilled shaft capacity. sand (75%) interspersed with sandy silt strata (25%)
Therefore, it has been chosen to correlate the resulting in a slightly plastic over consolidated pro
curves obtained from the numerical models with the file with a low expansion and variable consistency.
points of the load test between 635,17 kN to The water table is ubicated 2 meters below the sur
4.045,71 kN. The correlation indicator is r2 calcu face (Figure 4).
lated with the displacement obtained by the load. In the static load test, the load–displacement
curve has been developed until reaching the
maximum load of 4.346,64 kN and the max
5.3 Modeling
imum deformation of 17,89 mm was obtained in
Figure 7 indicates 3 load–displacement curves (FB– stage 12 of peak load. In the unload stage,
Multipier) that correlate with the load test. The cor 3,46 mm has been recovered until the end of the
relation indicator r2 varies between 0,86 to 0,95. test.
861
The load test has been modeled in different scen The load–displacement curve with the correla
arios from the interpretation of field test results (e.g., tor r2 close to 1 has been graphically observed
CPTu and static load test). It has been observed that determining that the numerical model CPT08 is
the strength parameters that influence the t-z and appropriate to be used.
q-z curves are the unit weight and the undrained cohe
sion. Thus, the sensitivity analysis of these two com
bined parameters was carried out resulting in 9 REFERENCES
resistance states.
The load–displacement curves were obtained with Camacho, M.A. & Gomez J. 2016. Determinación
y comparación de la capacidad portante del pilote
the FB-Multipier software that performs the soil–pile
vaciado in situ (P2-3), a partir de ensayos de
interaction through the t-z and q-z curves. An axial dis campo y una prueba de carga estática en el nudo
placement has been induced in the drilled shaft head viario Beijing, Cochabamba, Bolivia. Cochabamba:
(simulating the static load test) to obtain the resultant Universidad Mayor de San Simón.
force at the same point. Ninety cases of load states Mayne, P.W. 2012. Regional report for North America.
have been determined and then a load–displacement Atlanta.
curve has been plotted. Mejia, J. 2016. Caracterización de suelos estratificados
The load–displacement curves were analyzed for mediante el ensayo de penetración de piezocono
each geotechnical model. In addition, it has been CPTu. Cochabamba: Universidad Mayor de San
Simón.
chosen to correlate the curves obtained from the
Robertson, P.K. & Cabal K.L. 2016. Cone penetration test
numerical models with the points of the load test ing for geotechnical engineering. California: Gregg.
between 635,17 kN to 4.045,71 kN. The correlation Reese, L.C. & O’NEILL, M.W. 1988. Drilled shafts: Con
indicator is r2, calculated with the displacement struction and Design. USA: FHWA Publication.
obtained by the load. Wang, S.T. & Reese, L.C. 1993. COMP624P – Laterally
The correlation indicator r2 has varied between loaded pile analysis for the microcomputer, ver. 2.0.
0,86 and 0,95 based on the CPTu field test. Springfield: FHWA Publication.
862
© 2022 the Author(s), ISBN 978-1-032-31259-0
Characterization of geotechnical spatial variability in river embankments

from spatially adjacent SCPT
F. Ceccato
DICEA-University of Padua, Italy
M. Uzielli
DICEA – University of Florence, Italy
P. Simonini
DICEA-University of Padua, Italy
ABSTRACT: characterization of the spatial variability of geotechnical properties of river embankment soils is
important for the enhanced modelling and assessment of embankment stability. In practice, available data is
usually limited. This study presents the results of a quantitative statistical analysis of the spatial variability of
cone tip resistance from the results of 16 closely spaced SCPT carried out near a recent breach in a levee on the
Panaro river in northern Italy. Two geotechnical homogeneous soil units are preliminary identified in the levee
to ensure the meaningfulness of the analysis in terms of soil type. For each unit, the horizontal and vertical
spatial variability and spatial correlation structures of cone tip resistance are investigated by a two-step proced
ure involving the calculation of empirical semivariograms and the subsequent fitting of semivariogram models.
Horizontal and vertical scales of fluctuation are estimated based on fitted semivariogram model parameters.
1 INTRODUCTION SCPT. CPT measurements have extensively proved to

be well-suited for assessing inherent soil variability
Levee and dike collapses cause considerable financial because a large volume of near-continuous data can be
and social losses in many countries, especially in collected in a cost-effective way, the test has good
highly developed areas. Therefore, risk assessment of repeatability, the equipment is highly standardized,
river embankment stability is receiving increasing and the procedure is well defined and almost inde
attention worldwide. The quantitative estimation of the pendent of operator skill. The investigations revealed
vulnerability of embankments, which is necessary for that the levee embankment is made of sand-silt mix
the risk assessment process, can be pursued in the tures in different proportions. While friction angle
form of fragility curves from the outputs of slope sta varies within a limited range (30-33 degrees), cohe
bility analyses. A key parameter in slope stability ana sion, which depends primarily on compaction level,
lyses is soil strength, which varies spatially within soil soil suction and cementation, varies significantly even
volume due to factors such as compositional hetero within a limited area.
geneity, level of compaction, degree of saturation, This paper describes the procedures and results
cementations, and presence of ancient breech-repairing of a quantitative investigation into the vertical and
materials. horizontal spatial variability of cone tip resistance
A levee failure occurred along the Panaro River performed on the unsaturated silty sand of the river
in Northern Italy on 6th December 2020. The fail levee embankment. Section 2 describes the results
ure was caused by a combination of concurrent of the geotechnical site investigation campaign
causes including soil heterogeneity, relict animal along with a preliminary descriptive statistical ana
burrows, presence of ancient brick elements in the lysis. CPT results are pre-processed to identify
embankment body, as well as of rhizomes of homogeneous soil units (Sec. 3) in which spatial
Arundo Donax. soil variability is modelled through the application
To corroborate the interpretation of such collapse, of geostatistical techniques (Sec. 4). Results of the
an extensive geotechnical and geophysical campaign geostatistical modelling process are assessed and
was conducted, including a series of 16 closely spaced discussed in Sec. 5
DOI: 10.1201/9781003308829-128
863
2 DESCRIPTION OF GEOTECHNICAL descriptive second-moment sample statistics were cal
CAMPAIGN culated depth-wise for qc, fs, and Vs; more specific
ally: mean (μ), standard deviation (σ), and coefficient
The geotechnical campaign conducted in the area of of variation (COV=σ⁄μ), given by the ratio of the
the breach consisted in 12 CPTU, 2 SCPTU, 2 DMT, standard deviation to the mean. Figure 1c-e plots the
2 boreholes, 7 ERT. A total of 11 undisturbed sam depth-wise mean and COV of qc, fs, and Vs. The grey
ples were collected for laboratory testing. To investi shaded area represents the values within one sample
gate in greater depth the spatial variability of soil standard deviation from the sample mean.
strength, a short stretch of levee embankment, The depth-wise COV of tip resistance and sleeve
located approximately 150m west of the breach, was friction varies between 0.15 and 0.6, with an average
investigated through 16 seismic CPT soundings at of 0.37 for qc and 0.49 for fs. This indicates that the
a constant horizontal spacing of 2.5 m. horizontal variability of the deposit is relatively high,
Cone tip resistance qc and sleeve friction fs were according to the “rule of thumb” provided by Harr
measured at vertical intervals of 1cm and the shear (1987), by which coefficients of variation below 10%
wave velocity Vs was measured at vertical intervals are considered to be “low”, between 15% and 30%
of 25cm with the true interval method. These tests “moderate”, and greater than 30%, “high”. The COV
reached a maximum depth of 5.5m. Figure 1a-b of shear wave velocity varies between 0.11 and 0.78
plots the complete set of results of the SCPT cam with an average of 0.22; indicating that the overall
paign. A smoothing procedure based on a moving level of horizontal variability of this parameter is
average with a 40cm-wide window is applied to moderate.
each SCPT vertical to remove small-scale noise. From a geotechnical perspective, it is important to
Although the tests are very closely spaced and the parameterize not only the degree of horizontal and
boreholes show that the material is classifiable con vertical scatter in data measurements, but also the spa
sistently as a mixture of sand and silt, simple visual tial correlation structure, i.e., whether the spatial vari
inspection of Figure 1a-b reveals significant horizon ation of mechanical behavior as described by SCPT
tal inter-sounding variations in the measured proper occurs abruptly or with continuity. The above descrip
ties at the same depth, and intra-sounding, depth-wise tive statistical analysis is not suited to fully describe
variability in the vertical direction. While vertical variability as it cannot provide information regarding
variability can be expected due to in-situ stress effects spatial variability patterns. Moreover, the descriptive
and to stratigraphic layering, horizontal variability is analysis does not account for the possible presence of
more significant than could be foreseen given the different stratigraphic units which could display dis
close spacing of the soundings and the limited exten tinct geotechnical properties. In the following we
sion of the area. To quantify this variability, focus on the variability of the tip resistance.
Figure 1. (a) qc and Vs measured in the 16 SCPT; (b) fs measured in the 16 SCPT; (c-e) average values and COV in hori
zontal direction of qc, fs and Vs respectively.
864
3 IDENTIFICATION OF HOMOGENEOUS between 1.8m and 2.3m as well as between 5.0 and
SOIL UNITS 5.3m, which correspond to the transition zone
between different HSU (Figure 2b). Qtn is used for
Sample statistics and spatial variability parameters the identification of HSU because it is the parameter
aimed at characterizing specific soil types are only that best captures soil compaction effects. Indeed, it
meaningful if conducted on soil volumes which are is clear from the borehole that the embankment is
sufficiently homogeneous for geotechnical purposes. built with silty sand but compacted at different dens
A moving-window procedure proposed by Uzielli ities. The threshold value for COV is set to 0.15.
et al. (2008) is employed to identify Homogeneous With this procedure it is possible to identify two
Soil Units (HSUs) statistically. The normalized tip HSUs: Unit A (between the depths of 0.30m and
resistance Qtn , normalized friction ratio Fr and soil 1.80m) and Unit B (between 2.35m and 5.00m).
behaviour type index Ic are computed from field
measurements according to Robertson (2009).
where
The vertical effective stress is estimated consider

ing soil suction (s) by applying the Bishop effective
stress principle
Sr is the degree of saturation computed from soil

suction assuming the Van Genuchten soil water
retention model (Eq. 6) with typical parameters for
these soils derived from the laboratory tests, i.e., p0
=5kPa, λ=0.3, Ssat ¼ 1, Sres ¼ 0:16. Figure 2. Identification of HSU.
1 -λ
4 MODELLING OF SPATIAL VARIABILITY
The spatial correlation structure of tip resistance is

A linear suction distribution is assumed above investigated using a geostatistical approach by means
water level, which is 10m-deep in this site. of semivariograms. Given the well-known anisotropy
For each investigated depth, average values of in geotechnical properties stemming from in-situ
Qtn , Fr and Icn are computed over the horizontal dir stress effects and other site-specific factors, horizontal
ection and used to identify the homogeneous soil and vertical variability are addressed separately. The
units (HSU), see Figure 2a. The procedure is based adoption of specific geostatistical techniques and
on the evaluation of the coefficient of variation models relies heavily on the hypothesis of data statio
(COV) of the data within a 40cm-wide moving narity, which denotes the invariance of a data set’s
window. Each position of the moving window statistics to spatial location. Stationarity can be
defines two semi-windows of equal height above and achieved through a number of data transformation
below a centre point. The COV shows a peak at the techniques. Here, data decomposition is implemented,
interface between different homogeneous soil by which the “total” spatial variability of a spatially
units. Qtn , Fr and Icn show an increase of COV ordered measured geotechnical property [q(z1…zn)]
865
in a sufficiently physically homogeneous soil unit is depth interval and the average tip resistance in of
broken down into a trend function [t(z1…zn)] and each sub-unit are considered as reference values for
a set of residuals about the trend [r(z1…zn)]. In the geostatistical modelling purposes. A linear trend in
one-dimensional case, for instance, taking depth (z) horizontal direction is determined for each reference
as the single spatial coordinate, decomposition is depth (qc;trend ðxÞ ¼ a1 x þ a0 ) and the residuals are
expressed by the following additive relation calculated through data decomposition. Application
of the Mann-Kendall test assessed the stationarity of
the residuals of linear detrending for all sub-units.
The empirical semivariograms of the residuals for
all sub-units are plotted in Figure 3. These are fitted
Stationarity of the residuals is verified with the with the Gaussian (GAU), Spherical (SPH) and
Mann-Kendall test (Kendall, 1938, 1955). This non Exponential (EXP) semivariogram models summar
parametric statistical test involves the calculation of ized in Table 1. Though semivariogram model fitting
the test statistic τ. Low values of τ indicate a low sig is performed automatically, best-fit models were sub
nificance of spatial correlation (and, thus, a more sequently scrutinized critically to assess their adequ
probable stationarity of data), while τ values close to ateness. The GAU and SPH model are those
+1 or -1 indicate positive or negative correlation providing the best fits overall. An example of the
respectively. best-fit model for the depth of 3.53m is shown in
The spatial correlation structure of residuals is Figure 4.
investigated through a sequential process involv Figure 5 plots the scale of fluctuation for each ref
ing: (1) the calculation of empirical semivario erence depth for the selected semivariogram models.
grams; (2) the fitting of semivariogram models; GAU and SPH model provide similar values of δ,
and (3) the estimation of the scale of fluctuation while higher values are obtained with the EXP
from semivariogram model parameters. The scale model. The horizontal scale of fluctuation (Table 2)
of fluctuation (δ) describes the distance over ranges between 3.7 and 21.1m. Average values of
which the parameters of a soil are significantly 6.8m, 7.4m and 11.5m are obtained for the GAU,
correlated. A low scale of fluctuation attests to SHP and EXP models, respectively. These average
less gradual spatial variability. The scale of fluc scale of fluctuations are lower than reported in other
tuation can be calculated from the values of the studies (e.g. (Cami et al., 2020)), but it must be con
characteristic parameters of the semivariogram sidered that they highly depend not only on the data
models (a: range; c0 : nugget; c: sill) which are base but also on the reduced horizontal spacing
fitted to empirical semivariograms. Table 1 sum between soundings, which allows full exploitation of
marizes the semivariogram models used in this the typically existing nested correlation structure as
study and the model-specific functions used to discussed in Cami et al. (2020).
calculate the scale of fluctuation (Elkateb et al.
2003). In the model equations given in Table 1,
h is the lag distance, i.e., the distance between
observations.
Table 1. Semivariogram models and analytical expres

sions for the scale of fluctuation (Onyejekwe et al. 2016).
Model Equation δ
� � �� p
Gaussian (GAU) c 1 - exp - ha þ c0 πa
( f n
h3
Spherical (SPH) 2a - 2a3 þ c0 for h � a
c 3h
0:75a
c þ c0 for h4a
f f 2 nn
Exponential (EXP) c 1 - exp - ha2 þ c0 2a
4.1 Horizontal spatial variability

Horizontal spatial variability is investigated by sli
cing each homogeneous soil unit into 30cm-thick
sub-units and conducting two-dimensional geostatis
tical modelling on each sub-unit. A total of 15 sub
units were obtained. The central value of the vertical Figure 3. Empirical semivariogram in horizontal direction.
866
Table 2. Horizontal scale of fluctuation calculated from
semivariogram model parameters.
z[m] GAU SPH EXP
0.4 5.1 4.6 4.3

0.7 6.2 5.1 6.8
1.0 3.5 3.8 6.4
1.3 3.3 3.5 5.4
1.6 4.6 5.3 9.1
2.5 4.2 4.7 7.9
2.8 3.7 4.1 6.8
3.0 4.0 4.6 7.5
3.3 3.8 4.3 6.9
3.5 5.0 5.5 9.6
3.8 12.8 14.8 21.1
Figure 4. Best fitting semivariogram model for the refer
ence depth of 3.53m. 4.1 13.0 12.3 17.6
4.3 7.8 9.1 14.0
4.6 8.6 8.7 14.1
4.8 4.7 5.9 9.4
Mean 6.0 6.4 9.8

Min 3.7 4.1 6.8
Max 13.0 14.8 21.1
Figure 5. Horizontal scales of fluctuation by reference

depth obtained from semivariogram model fitting.
4.2 Vertical spatial variability

Vertical spatial variability is investigated by soil
unit and sounding. A cubic trend model
(qc;trend ðzÞ ¼ a3 z3 þ a2 z2 þ a1 z þ a0 ) is applied to
each instance and the residuals are calculated. The
Mann-Kendall test is performed to assess stationar
ity. The empirical semivariograms for all soil units
and SCPT soundings are shown in Figure 6. These
are fitted with the semivariogram models summar
ized in Table 1, yielding the vertical scales of fluctu
ation given in Table 3 and plotted in Figure 7.
The EXP model predicts the largest values of
scale of fluctuation, followed by the SHP and GAU
models. In Unit A, δ varies between 0.1m and 0.6m,
thus generally lower than in Unit B, where calcu
lated values range between 0.2m and 0.9m. These
values are consistent with the results of previous
Figure 6. Empirical semivariogram in vertical direction.
studies, e.g. (Cami et al., 2020).
867
geostatistical modelling of the results of 16 closely
spaced SCPT soundings. Results of the
descriptive second-moment statistical analysis attest
to a high degree of scatter in measured data, presum
ably due to the effect of partial soil saturation and
other site-specific phenomena.
To supplement the outputs of the statistical ana
lysis and to overcome its limitations with respect to
the quantitative characterization of the spatial correl
ation structure of mechanical resistance to cone
penetration, geostatistical modelling of the spatial
correlation structure was conducted both in the hori
zontal and vertical directions for two depth intervals
referring to highly homogenous soil units.
The horizontal correlation structure, parameterized by
the horizontal scale of fluctuation, proved to be stronger
(i.e., with cone resistance varying significantly over
smaller horizontal distances) than typically assessed in
existing literature. This result could be reconducted to
Figure 7. Vertical scales of fluctuation by semivariogram both the high quality of the dataset (the small horizontal
model and soil unit. spacing between consecutive soundings allows the
appreciation of nested correlation structures) and to the
specific site effects which result in the surprisingly high
Table 3. Vertical scale of fluctuation from variogram
models. degree of inter-sounding variability. Results of vertical
spatial correlation modelling are fully in line with previ
ous studies, thus attesting to the general correctness in
Unit A Unit B
the modelling approach.
The results obtained in the study confirm the par
x[m] GAU SPH EXP GAU SPH EXP ticular site conditions which lead to the significant
horizontal inter-sounding variability observed through
0.0 0.3 0.4 0.5 0.4 0.4 0.6
the geotechnical testing campaign. While the com
2.5 0.3 0.4 0.5 0.5 0.6 0.8 plexity of the physical phenomena which lead to such
5.0 0.2 0.3 0.4 0.4 0.4 0.6 variability require further and more extensive investi
7.5 0.3 0.4 0.5 0.5 0.6 0.9 gation, the quantitative assessment of the spatial
correlation structure and its anisotropy attest to the
10.0 0.3 0.4 0.5 0.2 0.2 0.3
importance of statistical and geostatistical analyses
12.5 0.3 0.3 0.4 0.5 0.6 0.9 for geotechnical modelling purposes. The availability
15.0 0.2 0.2 0.3 0.5 0.7 0.9 of quantitative spatial variability parameters allows
17.5 0.3 0.4 0.5 0.4 0.5 0.7 the enhanced modelling of the geotechnical systems
20.0 0.3 0.4 0.5 0.4 0.5 0.8 by providing realistic inputs to, for instance, limit
equilibrium and numerical analyses.
22.5 0.3 0.4 0.6 0.5 0.6 0.9
25.0 0.2 0.3 0.4 0.3 0.3 0.5
27.5 0.2 0.2 0.3 0.5 0.6 0.8 ACKNOWLEDGMENTS
30.0 0.2 0.2 0.3 0.5 0.6 0.8
The authors would like to thank AIPO that financed
32.5 0.3 0.4 0.5 0.4 0.4 0.8 the geotechnical site investigations.
35.0 0.1 0.2 0.2 0.3 0.3 0.5 The authors would like to acknowledge the finan
37.5 0.1 0.1 0.2 0.5 0.5 0.7 cial support from MIUR (Redreef - PRIN 2017 Call,
prot. 2017YPMBWJ)
Mean 0.2 0.3 0.4 0.4 0.5 0.7
Min 0.1 0.1 0.2 0.2 0.2 0.3
Max 0.3 0.4 0.6 0.5 0.7 0.9 REFERENCES
Cami, B. et al. (2020) ‘Scale of Fluctuation for Spatially
Varying Soils: Estimation Methods and Values’, ASCE
5 DISCUSSION AND CONCLUSIONS ASME Journal of Risk and Uncertainty in Engineering
Systems, Part A: Civil Engineering, 6(4), p. 03120002.
The vertical and horizontal spatial variability in cone doi: 10.1061/ajrua6.0001083.
tip resistance of the unsaturated silty sand forming Elkateb, T., Chalaturnyk, R. and Robertson, P. K. (2003) ‘An
a levee is investigated through statistical and overview of soil heterogeneity: Quantification and
868
implications on geotechnical field problems’, Canadian Geology. Elsevier B.V., 214, pp. 43–49. doi:
Geotechnical Journal, 40(1), pp. 1–15. doi: 10.1139/t02 10.1016/j.enggeo.2016.09.014.
090. Robertson, P. K. (2009) ‘Interpretation of cone penetration
Harr, E. M. (1987) Reliability-based design in civil engin tests — a unified approach’, Canadian Geotechnical Jour
eering. Department of Civil Engineering, Schoool of nal, 46(11), pp. 1337–1355. doi: 10.1139/T09-065.
Engineering, Noth Carolina State University. Uzielli, M. (2004) Variability of stress-normalized CPT
Kendall, M. G. (1938) ‘A new measure of rank parameters and application to seismic liquefaction initi
correlation’, Biometrika, 30, pp. 81–93. ation analysis.
Kendall, M. G. (1955) Rank correlation methods. Uzielli, M., Simonini, P. and Cola, S. (2008) ‘Statistical
New York: Hafner Publishing Co. identification of homogeneous soil layers in Venice
Onyejekwe, S., Kang, X. and Ge, L. (2016) lagoon soils’, in Proceedings of the 3rd International
‘Evaluation of the scale of fluctuation of geo Conference on Site Characterization ISC’3. Taipei:
technical parameters by autocorrelation function Taylor & Francis. Available at: https://flore.unifi.it/
and semivariogram function’, Engineering handle/2158/1181074.
869
© 2022 the Author(s), ISBN 978-1-032-31259-0
Coupling site wide CPT profiles and genetic algorithms for whole-site
offshore windfarm layout optimization
J.A. Charles & S.M. Gourvenec
University of Southampton, Southampton, UK
M.E. Vardy
SAND Geophysics, Southampton, UK
ABSTRACT: Offshore windfarm layout is driven by optimizing wind potential at the selected site subject to
minimum spacing between turbines. Within these constraints, this paper explores the optimization of turbine
layout within a selected site based on geotechnical site conditions. Utilizing an irregular but site-wide grid of
CPT profiles, a simple pile design method and a cubic interpolation of pile dimensions, heatmaps of pile
length and diameter at every possible coordinate within a site can be generated, providing the opportunity for
optimization of layout. This paper demonstrates the ability to optimize the location of a prescribed number of
wind turbine foundations, in conjunction with other constraints such as minimum spacing, to minimize total
steel usage (a simple proxy for foundation cost and embodied carbon) in the foundation system using a genetic
algorithm (GA)-based approach. Existing work on windfarm layout optimization uses such techniques to maxi
mize wind energy generation but negligible work exists on extending the methodologies to minimize founda
tion costs. This paper demonstrates the viability of using a set of CPT profiles with a GA-based approach for
a geotechnically-informed windfarm layout and explores the effects of various meta parameters using publicly
available datasets. The work demonstrated in this paper is directly relevant to ongoing advances in geophysics
and machine learning that would allow for the generation of a synthetic CPT profile at any point on the site,
eliminating the drawbacks of interpolation between actual CPT profiles for design parameters.
1 INTRODUCTION AND LITERATURE With continuous site investigation data, a monopile

REVIEW foundation can be designed at any location on a site,
from which it is possible to use optimization tech
The construction of offshore windfarms is a key part niques to minimize the overall foundation cost - in
of the UK government’s renewable energy strategy terms of volume of steel used - for the windfarm for
(BEIS 2019, Supergen ORE, 2021), and internation a specified number and spacing of turbines. Windfarm
ally, the Global Wind Energy Council (GWEC) fore layout optimization is typically carried out to maxi
casts that by 2030, more than 205GW of new offshore mize the wind energy generated based on aero
wind capacity will be added (Global Wind Energy dynamic considerations (Pérez et al., 2013; Samorani,
Council, 2020). 2010). This paper seeks to demonstrate that founda
Securing offshore wind turbines to the seabed – tion efficiency can also be optimized for, and hence
whether as fixed or floating devices – requires a step include geotechnical parameters as additional vari
change in site investigation and geotechnical design ables in typical wind turbine layout optimization.
approaches. Due to the size of offshore windfarms, Combining or comparing the significance or trade-
with Hornsea 1, currently the world’s largest, covering offs of these different considerations is beyond the
an area of 407km2 and consisting of 174 turbines scope of this paper, which seeks to explore and dem
(Ørsted, 2020a), ground investigation datapoints such onstrate the opportunity to include geotechnical
as borehole logs and CPT profiles are relatively sparse assessment in wind farm layout optimization.
per structure compared to e.g. a building foundation, Designing a monopile at a location with a CPT pro
or even an offshore hydrocarbon platform. However, file is straightforward. Many techniques with varying
using interpolation between CPTs, or synthetic CPT complexity exist in literature and in commercial soft
profiles derived from continuous seismic data, continu ware packages (Bhattacharya, 2019, Kay et al. 2021).
ous ground models of a site can be derived (Vardy Designing a monopile at locations in between available
et al., 2017). CPT profiles may apply similar methods but is
DOI: 10.1201/9781003308829-129
870
associated with greater uncertainty. Since this paper 2 GEOTECHNICAL DESIGN OF MONOPILES
aims to demonstrate a proof of concept for
a geotechnically-informed windfarm layout optimiza There are many methods by which a monopile can
tion procedure, a simple cubic interpolation approach be designed for a given CPT profile with varying
has been adopted to derive pile dimensions between levels of complexity. Although more complex
locations of actual CPT profiles based on the calcu methods such as PISA (Byrne et al., 2015) or the
lated dimensions at each CPT location. methodologies implemented in the web app based
The site examined as a case study, Burbo Bank LAP software (Doherty, 2017, 2020) are available
Extension (Ørsted, 2020b), located off the coast of Liv which consider factors such as serviceability, for this
erpool, UK, in the Irish Sea and has an area of the order proof of concept only capacity will be considered.
of 40km2 and around 42 CPT profiles in an irregular The primary loading on a wind turbine foundation
grid as shown in Figure 1. Coordinates are given in is horizontal loading, both hydrodynamic (wave, cur
northings and eastings which are measured in kilo rent) and aerodynamic (wind), taken together as 16MN
meters. For Burbo Bank Extension, the maximum north- for this study which is a typical value cited in literature
south distance is approximately 4.2km and the max (Lesny & Wiemann, 2005). A typical safety factor of 3
imum east-west distance is approximately 11.8km. The has been applied to this load for a total 48MN. Note
final constructed windfarm extension consisted of 32 that factoring has been applied only to the applied load
8MW turbines that together can power around 230,000 and not to soil or material properties. The failure
homes (Ørsted, 2020b). Although not utilized in this mechanism examined is rotational failure. To ensure
paper, the Burbo Bank Extension site investigation (and the foundation is sufficient to resist this load, lateral
many others) feature large amounts of geophysical data. resistance can be calculated at each point along the
A geophysical survey involves sending seismic waves length of the pile based on the values obtained in the
into the seafloor and determining soil properties based CPT profile. Taking moments about a representative
on reflected waves. Whereas CPT and borehole data is lever arm for both wind and wave loading, 35 m in
discrete, geophysical data is continuous, but does not dir this case, the depth to the rotation point can be found.
ectly measure physical parameters that have to date been The local lateral resistance values can then be multi
robustly linked to engineering properties needed for geo plied by the length they apply to and summed, with
technical design. However, recent advances allow for those below the point of rotation being negative.
synthetic CPT data to be generated from seismic data Several equations exist to calculate discretized
via machine learning at any point on a site with geo resistance along the pile. Equation 1 shows the pie
physical data of sufficient quality (Sauvin et al., 2019). cewise resistance in clay (Truong & Lehane, 2014)
This process, in future work, would allow for a drop-in and Equation 2 is the equivalent for sand (Suryasen
replacement to the simple interpolation process used in tana & Lehane, 2014). Note that there are various
this proof of concept. other possibilities for sands and clays as well as
This paper first details a simple CPT based pile equations for other types of soil and rock (Randolph
design method and provides a brief summary of alter- & Gourvenec, 2017).
native/more detailed approaches. The process by
which pile length and diameter are interpolated across
the site is provided, along with discussions on the
emerging techniques that would allow for improve
ments. Finally, a demonstration of how a genetic algo
rithm (GA) can be used to automatically select the
locations for a specified number of monopiles to min
imize a given condition, in this case steel volume, is
presented. The general method presented here can be
applied to any potential site or foundation type, and
involve other constraints such as bathymetry or areas
of ecological significance or anthropogenic activity.
In Equations 1 and 2, pu refers to local lateral resist
ance per unit length, D is the diameter of the mono-
pile, Ir is rigidity index (assumed to be 125 for this
study), d is depth below the seafloor to the local point
being calculated, σ0v is effective vertical stress calcu
lated at the local point, qc is cone resistance where
qc ¼ Qc =Ac , i.e. the force required to push the cone
into the ground over cone area, and qnet is the net
cone resistance where qnet ¼ qt - σv , i.e. corrected
cone resistance minus total vertical stress. Corrected
cone resistance qt ¼ qc þ u2 ð1 - aÞ where u2 is pore
Figure 1. Locations of CPT tests at the Burbo Bank Extension. water pressure and a is a constant (Robertson, 1990).
871
As these equations are defined in terms of pile 3 WHOLE SITE INTERPOLATION
diameter (along with the pile length determining how
many discretization points are to be summed together) Although it is ideal to design a monopile that is
the easiest way to size a pile is a brute force approach. co-located with a CPT profile, often it is neces
Table 1 shows the range of lengths and diameters sary to design a pile that is located between
used in this study which are based on the selection actual CPT profile locations. To achieve this,
used in Sauvin et al. (2019), with the addition of a monopile was designed at each CPT location
lower 5m and 6m diameters. It should be noted that across the site, with pile length and diameter then
due to the simplicity of the design criteria chosen that interpolated for locations in between the discrete
the possible pile dimensions in Table 1 are illustrative irregular grid.
and not intended to represent a robust detailed design It should be noted that this methodology has sev
methodology for the loading conditions, and do not eral limitations, including overlooking the actual
take into account other considerations such as install- variation in soil conditions in between CPT loca
ability or other failure mechanisms such as bending. tions. Nonetheless, for the purpose of this proof of
concept, using interpolation allows the construction
of heatmaps of monopile length, diameter, and
Table 1. Typical pile dimensions as used in this study.
volume to be constructed over the whole site, which
Property
is a necessary step in freely selecting the location of
a set of monopiles.
Diameter (m) 5 6 7 7.7 8.5
There are emerging techniques that would
Length (m) 3D 3.5D 4D 4.5D 5D
potentially eliminate the issues of this interpolation
step. On a site with continuous geophysical survey
data and discrete CPT data, it is possible to com
Python code was written to automatically identify bine seismic inversion methods with machine
soil type based on the standard Robertson charts learning algorithms predict a synthetic CPT profile
(Robertson, 1990, 2009). For simplicity, datapoints at any location on the site (Vardy et al., 2018;
along the CPT profiles were classified as either sand Sauvin et al., 2019). This technique would poten
or clay allowing for automated selection of the rele tially represent a significant improvement to site
vant lateral resistance equation. Of the nine possible wide characterization and work is actively ongoing
Robertson chart classifications, sensitive fine grained, in this area, but has not seen widespread applica
organic, clay, silt-mixtures, and very stiff fine-grained tion, and therefore the suitability of the method
were labelled as clay, and sand-mixtures, sand, grav for different site conditions is unclear.
elly sand to sand, and very stiff sand to clayey sand It is the hope of the authors that such techniques
were labelled as sand. Figure 2 shows a plot of would represent a drop-in replacement for the inter
a Burbo Bank Extension CPT profile colored based polation step presented in this paper. However, for
on soil classification. The axis in the figure are nor now, cubic interpolation serves as an acceptable
malized cone resistance Qt ¼ ðqt - σv Þ=σ0v and fric placeholder to allow for the demonstration of foun
tion ratio Rf ¼ ðfs =qt Þ x 100% where sleeve friction dation-based site layout optimization.
fs ¼ Fs =As with Fs and As being the recorded fric Figure 3 shows the heatmaps generated for pile
tional force and friction sleeve area respectively diameter, length, and volume across the site. It
(Robertson, 1990). should be noted that due to the interpolation process
it is possible that some regions on the heatmap drop
below the discrete pile dimensions provided in
Table 1.
Calculation of pile volume requires an additional
wall thickness term. The American Petroleum Insti
tute (API, 2005) recommend a minimum thickness t
based on Equation 3. A thickness of 0.1m was found
to satisfy this requirement for all pile diameters con
sidered in this study.
4 LAYOUT OPTIMISATION USING

A GENETIC ALGORITHM
With heatmaps produced for required pile diameter

Figure 2. Example CPT profile from Burbo Bank Exten and length across the whole site, it is possible to check
sion categorized into sand and clay. the pile dimensions and hence steel volume necessary
872
allow for a subset of specimens to advance as
they are; a second subset are mutated by ran
domly adjusting a random selection of the associ
ated coordinates; and a third subset is adjusted by
randomly mutating the coordinates of the highest
volume piles.
It is noted that advancing the best specimens with
out changes guarantees that the fittest solution at
a given time will never become worse with add
itional generations. Further, for this use case the GA
technique of “crossover” in which two specimens
are merged has been omitted as merging half of the
list of coordinates from two separate specimens
without violating the distance constraint proved to
not be possible without additional logic that effect
ively nullified any benefits of the crossover process.
There are many choices for the aforementioned
meta-parameters, for example, the number of gener
ations, the number of specimens generated, the
number of specimens in each category, the rankings
Figure 3. Heatmaps of pile diameter, length, and volume assigned to each category, whether or not specimens
for the Burbo Bank Extension site. can be included in more than one category etc. Com
mentary and plots are provided on some meta-
at any arbitrary coordinate. It is therefore possible to parameters in the following section, but an exhaust
sum the steel usage for a set of n arbitrarily located ive search is beyond the scope of this paper.
piles, and to compare the sums of steel usage for two
different sets of piles. Using some means of optimiza
tion, in this paper, a genetic algorithm (GA), it is pos 5 RESULTS AND DISCUSSION
sible to select a set of n coordinates that minimize the
required steel usage in foundations across the site. To demonstrate the methodology, the GA was run for
A GA is an evolutionary algorithm (Poli et al., multiple combinations of parameters. Initially, the
2008) that mimics natural selection (Darwin, 1859). method was used to find the optimal layout of 32
Populations of solutions (turbine layouts in this case) piles, the number constructed in the actual Burbo
are ranked by fitness (steel usage in this case) with the Bank Extension site development, with 100 speci
best continuing to future generations and the less good mens (i.e. lists of coordinates) used per generation,
not doing so. After many generations, the solutions with 100 generations.
tend to improve. It should be noted that the real site layout features
Although there are numerous existing libraries 32 piles packed as closely as possible along linear
implementing GA functionality (e.g. Fortin et al., transects to minimize wind turbulence and maximize
2012; Gad, 2021), due to the relative simplicity of wind yield. An industry typical spacing would be 8
the problem, and simply being able to look up the blade diameters (DNV GL Energy, 2019). Enforcing
steel usage for a given coordinate, the GA function this restriction would allow no leeway to demon
ality for this study was implemented from scratch in strate the ability of a GA to select locations to min
Python. The steps are outlined below: imize total foundation costs and as such a lower
minimum spacing of 800m was enforced, representa
1. Generation of a specimen. Code was created to tive of wind turbines of diameter 100m or spacing of
randomly select a coordinate within the site, the actual ~160m diameter wind turbines at 5 blade
check that specified constraints are not violated diameters.
(i.e. being too close to an existing coordinate) and Using smaller turbines than is currently standard
either add it to the list, or discard it, moving on to or adopting smaller spacing would negatively affect
the next randomly selected coordinate until the energy generated, and therefore be an unattractive
desired number of coordinate positions is reached. prospect for windfarm operators, but implementa
2. Implementation of cost function. Code was cre tions of the presented methodology allow for fine
ated to take a specimen as an input and calculate adjustments to more traditional layouts that minim
or look up the required steel volume for each ize foundation and potentially installation cost that
coordinate, and then sum all values, enabling may be beneficially offset against reduced wind
ranking of a population of specimens. power generation.
3. Creation of the next generation. Code was created Figure 4 shows the layouts generated by the GA
to advance the population of specimens to the after 3, 20, and 60 generations. As the genetic algo
next generation. User specified meta-parameters rithm process involves randomness it is not
873
deterministic and rerunning the process will likely provides a more optimal result, i.e. less steel used for
return similar but slightly different layouts. After the same number of monopiles. The implementation in
just 3 generations nearly every turbine is placed in fact ensures that a generation is at least as good as the
a lower volume area but there are still improvements one before. Most improvement happens relatively
that can be seen by eye between the 3 and 20 gener soon, within the first 10 or 20 generations, and although
ation layouts. The differences between 20 and 60 further improvement can be seen there are significant
generations are difficult to spot in the layout maps, diminishing returns after generation 50 or so.
but small improvements can be found when examin The relationship between number of specimens
ing the total site pile volume. Table 2 shows total and total volume has a clear correlation but the
pile volume for each of the layouts in Figure 4. effect is relatively modest in comparison with the
number of generations. It is possible that due to the
impracticality of including the GA crossover tech
nique in the current problem formulation that the
benefits of a larger population are less pronounced
than they would be for a different problem. In this
case, 20 specimens appear to be enough.
Figure 5. Effect of selected meta parameters on total pile

volume.
Figure 4. Layouts of 32 turbines after 3, 20, and 60 generations. It is important to note that there is no guarantee that
a GA can find the global optimum solution, even if the
improvements between generations appear to plateau.
Table 2. Total pile volume per generation for the 32-pile The global best solution would be the ideal solution in
layouts shown in Figure 4. which there can be no improvements. GAs however can
potentially tend towards a local optimum solution.
Difference from
Total mean volume at
A local optimum solution can be thought of as an evolu
volume Mean volume CPT locations tionary “dead end” in which a solution has been found
Generation (m3) per pile (m3) (36.96 m3) that is good but is sufficiently different from the global
optimum that no random mutations exist that will move
3 1100.8 34.4 -6.9% from one to the other. Finding a local rather than global
20 1045.1 32.7 -11.5% minima for site wide foundation cost would be more
60 1039.7 32.5 -12.1% expensive but would not have safety implications.
Figure 5 shows the effects of various meta- 6 CONCLUSIONS AND FUTURE WORK
parameters on the layout and total volume of 32 mono-
piles. Considering the relationship between number of Using a GA to help identify offshore windfarm site lay
generations and volume of steel used, 10 trials were outs that minimize foundation steel volume (and implied
carried out with the average volume plotted per gener cost, both direct, and embodied carbon) has been illus
ation number, and for the plot of specimens, 5 trials trated. The illustrated example includes several idealiza
were carried out for each value. As stated previously, tions and placeholder techniques have been used that
there is an infinite possibility of potential setups for need to be revisited to refine the approach. These include
parameters for a GA and the two plots presented here (1) Incorporating the emerging technique of generating
do not represent an exhaustive analysis but were synthetic CPT data from geophysical surveys via
selected due to their relative simplicity and their effects machine learning; (2) Additional constraints based on
on computational time. It is evident that increasing the probabilistic analysis of the CPTs, i.e. areas in which
number of generations that the GA operates for, there is high uncertainty in synthetic CPT data may be
874
better avoided; and (3) More complex higher order pile algorithms made easy. Journal of Machine Learning
design approaches. Research, 13, 2171–2175.
Finally, it is noted that an approach such as put Gad, A. F. 2021. PyGAD: An Intuitive Genetic Algorithm
forward in this paper is intended to complement, not Python Library. http://arxiv.org/abs/2106.06158
replace other constraints and conditions for wind- Global Wind Energy Council. 2020. Global Offshore
farm layout optimization, such as aerodynamic Wind: Annual Market Report 2020. Global Offshore
effects for optimal energy generation. Wind Report 2020, February, 130.
Kay, S., Gourvenec, S., Palix, E. & Alderlieste, E. 2021
Intermediate Offshore Foundations. CRC/Taylor &
Francis ISBN: ISBN: 978-1-138-35353-4 (hbk) ISBN:
ACKNOWLEDGEMENTS 978-0-429-42384-0 (ebk)
Lesny, K., & Wiemann, J. 2005. Design aspects of mono-
This work forms part of the activities of the Centre piles in German offshore wind farms. In Proceedings of
of Excellence for Intelligent & Resilient Ocean the International Symposium on Frontiers in Offshore
Engineering (https://www.southampton.ac.uk/iroe) Geotechnics (pp. 383–389). AA Balkema Publishing.
and the first and second authors are supported by Ørsted. 2020a. Hornsea One: About the project. https://horn
Royal Academy of Engineering under the Chairs in seaprojectone.co.uk/about-the-project#project-timeline
Emerging Technologies scheme. 2020
Part of the development of the GA approach pre Ørsted. 2020b. Burbo Bank Extension Offshore Windfarm.
https://orstedcdn.azureedge.net/-/media/www/docs/corp/
sented here was carried out by University of South uk/updated-project-summaries-06-19/sept-2020/
ampton undergraduate Etienne Martin during 200819_ps_burbo-bank-extension_v2_web-aw.ashx
a summer internship organized through the Depart Pérez, B., Mínguez, R., & Guanche, R. 2013. Offshore
ment of Civil, Maritime and Environmental Engin wind farm layout optimization using mathematical pro
eering and funded by the Centre of Excellence for gramming techniques. Renewable Energy, 53, 389–399.
Intelligent & Resilient Ocean Engineering. https://doi.org/10.1016/j.renene.2012.12.007
The Burbo Bank extension dataset used was made Poli, R., Langdon, W. B., & McPhee, N. F. 2008. A Field
publicly available by the Crown Estate via the Guide to Genetic Programing. In Wyvern (Issue March).
Marine Data Exchange website (https://www.marine http://www.essex.ac.uk/wyvern/2008-04/Wyvern
April08 7126.pdf
dataexchange.co.uk/). Randolph, M., & Gourvenec, S. 2017. Offshore Geotech
nical Engineering. DOI: 10.1201/9781315272474
Robertson, P. K. 1990. Soil classification using the cone
REFERENCES penetration test. Canadian Geotechnical Journal, 27(1),
151–158. https://doi.org/10.1139/t90-014
American Petroleum Institute. 2005. Recommended practice Robertson, P. K. 2009. Interpretation of cone penetration
for planning, designing and constructing fixed offshore tests - A unified approach. Canadian Geotechnical Jour
platforms—Working stress design. API RP 2A-WSD. nal, 46(11), 1337–1355. https://doi.org/10.1139/T09-065
Bhattacharya, S. 2019. Design of Foundations for Offshore Samorani, M. 2010. The Wind Farm Layout Optimization
Wind Turbines. John Wiley & Sons. Problem. Power, 1–18.
BEIS, Business Energy and Industrial Strategy, Department Sauvin, G., Vanneste, M., Vardy, M. E., Klinkvort, R. T., &
of. 2020. Policy Paper, Offshore wind: Sector Deal. Carl Fredrik, F. 2019. Machine Learning and Quantita
https://www.gov.uk/government/publications/offshore tive Ground Models for Improving Offshore Wind Site
wind-sector-deal Characterization. Offshore Technology Conference,
Byrne, B. W., Mcadam, R., Burd, H. J., Houlsby, G. T., 110(9), 1689–1699. https://doi.org/10.4043/29351-MS
Martin, C. M., Zdravković, L., Taborda, D. M. G., Supergen ORE. 2021. Delivering Net Zero: the role of Off
Potts, D. M., Jardine, R. J., Sideri, M., Schroeder, F. C., shore Renewable Energy A COP26 Briefing note pre
Gavin, K., Doherty, P., Igoe, D., Wood, A. M., pared by the Supergen Offshore Renewable Energy Hub.
Kallehave, D., & Gretlund, J. S. 2015. New design methods https://www.supergen-ore.net/uploads/Supergen-ORE
for large diameter piles under lateral loading for offshore Hub-Delivering-Net-Zero-the-role-of-Offshore-Renew
wind applications. Frontiers in Offshore Geotechnics III, able-Energy-A-COP26-Brief.pdf
705–710. Suryasentana, S. K., & Lehane, B. M. 2014. Numerical deriv
Darwin, C. 1859. On the origin of species. John Murray. ation of CPT-based p-y curves for piles in sand. Geotechni
DNV GL Energy. 2019. Potential to improve Load Factor que, 64(3), 186–194. https://doi.org/10.1680/geot.13.P.026
of offshore wind farms in the UK to 2035. https://assets. Truong, P., & Lehane, B. 2014. Numerically derived
publishing.service.gov.uk/government/uploads/system/ CPT-based py curves for a soft clay modeled as an elas
uploads/attachment_data/file/839515/L2C156060 tic perfectly plastic material. 3rd International Sympo
UKBR-R-05-D_-_potential_to_improve_Load_Factor sium on Cone Penetration Testing, 975–982.
s_of_UK_offshore_wind_to_2035.pdf Vardy, M. E., Vanneste, M., Henstock, T. J., Clare, M. A.,
Doherty, J. P. 2017. A web based application for the lateral Forsberg, C. F., & Provenzano, G. 2017. State-of-the-art
analysis of pile (LAP) foundations. Proceedings of the remote characterization of shallow marine sediments: the
ASME 2017 36th International Conference on Ocean, road to a fully integrated solution. Near Surface Geophysics,
Offshore and Arctic Engineering. 15(4), 387–402. https://doi.org/10.3997/1873-0604.2017024
Doherty, J. P. 2020. Lateral Analysis of Piles User Manual. Vardy, M.E., Clare, M.A., Vanneste, M., Forsberg, C.F., &
https://www.geocalcs.com/lap Dix, J.K. 2018. Seismic Inversion for Site Characterisa
Fortin, F. A., De Rainville, F. M., Gardner, M. A., tion: When, Where and Why Should We Use It? OTC
Parizeau, M., & Gagńe, C. 2012. DEAP: Evolutionary 28730-MS.
875
© 2022 the Author(s), ISBN 978-1-032-31259-0
CPT-based model calibration for effective stress analysis of layered

soil deposits
A. Chiaradonna
N. Ntritsos & M. Cubrinovski

University of Canterbury, Christchurch, New Zealand
ABSTRACT: Recent research recognized the advantages of effective stress dynamic analysis in estimating
the seismic response of layered soil profiles. One of the key challenges in performing effective stress analysis
is the calibration of constitutive models able to simulate complex soil behavior under seismic loading. To over
come this problem, calibration procedures have been developed to guide the definition of model parameters on
the results of in-situ tests. In this study, a reference layered soil profile from Christchurch in New Zealand,
which experienced liquefaction during the 2010-2011 Canterbury earthquake sequence, was used to verify the
goodness of the CPT-based calibration procedure of a simplified pore water pressure model. Comparisons with
the seismic response predicted using an advanced state concept-based constitutive model were also reported.
Despite the simplifications adopted in the calibration process, the analysis results replicate the occurrence of
liquefaction and encourage CPT-based model parameter definition for routine effective stress dynamic analyses.
1 INTRODUCTION generalized to include the results of field tests com

monly used in engineering practice (Chiaradonna et al.
Recent research recognized the importance of effective 2020).
stress dynamic analysis in estimating the seismic In this study, the above-mentioned simplified stress-
response of layered soil profiles with interbedded based pore water pressure model is adopted with the
liquefiable and non-liquefiable soils (Cubrinovski et al. two-fold goal to (1) verify the goodness of the CPT-
2019). Two distinct approaches can be adopted to per based calibration procedure in predicting the liquefac
form an effective stress analysis: (1) a ‘loosely tion of a layered profile, (2) compare the model per
coupled’ approach that predicts seismic-induced pore formances with those provided by an advanced state
pressure build-up by adopting simplified relationships concept-based constitutive model assumed as
used in combination with constitutive models that a reference, which was calibrated on the same CPT.
address total stress (e.g., pore pressure prediction To pursue these aims, a representative layered soil
based on accumulated strains or stresses) and (2) profile from Christchurch in New Zealand, which
a ‘fully coupled’ approach that uses a plasticity-based experienced liquefaction multiple times during the
effective stress constitutive model to predict both the 2010-2011 Canterbury earthquake sequence, was
stress-strain and the pore pressure response of the soil considered in the numerical simulations. The adopted
(Tropeano et al. 2019). studied area was largely investigated in Cubrinovski
One of the key challenges in performing effective et al. (2019), where the geotechnical model and
stress analysis is the calibration of constitutive models effective-stress analyses by using the advanced state
able to simulate the dynamic soil behavior under seis concept-based constitutive model are detailed.
mic loading. To overcome this problem, calibration One-dimensional (1D) dynamic analyses were pre
procedures have been developed to guide the definition liminarily carried out by assuming visco-elastic
of the parameters of advanced constitutive models linear behavior of the soils, to check that the same
based on data from in-situ tests, such as Cone Penetra boundary conditions were applied in both approaches
tion Test - CPT (Ntritsos & Cubrinovski 2020). Fol (advanced and simplified). Comparisons with equiva
lowing this philosophy, the calibration of a simplified lent linear and nonlinear codes were also provided.
stress-based pore water pressure model, originally Then, a 1D effective stress analysis was performed
based only on cyclic laboratory test data, was with the simplified pore water pressure model,
DOI: 10.1201/9781003308829-130
876
according to a loosely coupled approach (Chiara the number of cycles, CSRr is the ordinate of the
donna et al. 2019). The parameters of the stress-based curve for NL = 15 (usually adopted as a reference
pore water pressure model were mainly defined on number of cycles). For a regular shear stress his
the cone tip resistance of each soil layer following the tory, κ is proportional to the number of cycles, N;
CPT-based procedure proposed by Chiaradonna et al. it is, therefore, possible to express the pore pres
(2020). The same geotechnical model was also simu sure ratio, ru (ratio between the excess pore pres
lated through the advanced state concept-based con sure and the initial effective confining pressure), as
stitutive model, whose parameters were calibrated on a function of the damage parameter, through the
the same CPT data by Cubrinovski et al. (2019). relationship proposed by the authors (Chiaradonna
Results of the 1D effective stress analyses provided et al. 2018):
insights on the capability of the loosely coupled
approach to correctly predict the attainment of lique
faction in sites with complex stratigraphic conditions,
also compared to the performance of a fully coupled
approach assumed as reference.
where a, b and d are parameters that control the
2 SIMPLIFIED PORE PRESSURE MODEL AND shape of the curve.
CPT-BASED CALIBRATION According to the CPT-based calibration proposed
by Chiaradonna et al. (2020), The parameters of the
A simplified stress-based pore water pressure model, curve CRR-NL (Eq. 1) can be computed as a function
hereafter called ‘PWP model’ (Chiaradonna et al. of the effective stress state and the normalized and cor
2018; 2019) permits the comparison of the irregular rected cone tip resistance, qc1Ncs, of the CPT; while the
seismic loading with the soil liquefaction resistance, parameters of the curve ru – N/NL (Eq. 2) can be
through an accumulation stress-based variable κ, defined as a function of the fine content and relative
called ‘damage parameter’. It is an incremental func density; this latter being usually estimated from CPT
tion of the applied load that considers the cyclic data (Robertson & Cabal 2015). In so doing, the pore
strength of the soil. This latter is expressed in terms water pressure model parameters can be easily cali
of cyclic resistance curve, analytically described by brated directly on the results of CPTs, through the
the equation: charts in Figure 1.
The PWP model was implemented in the non
linear code SCOSSA which models the soil profile as
a system of consistent lumped masses, connected by
viscous dampers and springs with hysteretic behavior.
The non-linear shear stress-strain relationship is
described by the MKZ model and the modified
where CSR is the shear stress amplitude normal Masing rules. More details about the numerical imple
ized by the initial effective confining pressure; N is mentation can be found in Tropeano et al. (2019).
Figure 1. Parameters of the PWP model: (a) CSRr, (b) α and (c) CSRt of Eq. (1) as a function of qc1Ncs; (d) a, (e) b and (f)
d of the Eq. (2) as a function of the fines content, FC, and the relative density, Dr.
877
3 ADVANCED STATE CONCEPT-BASED SOIL (2019), and are reported in Table 1. The small-strain
CONSTITUTIVE MODEL shear modulus of each layer was inferred from the
shear wave velocity profile shown in Figure 3.
An advanced soil constitutive model, called “Stress- The groundwater table depth identifies the thickness
Density” model (S-D Model) is formulated by of the liquefiable layer between 1.8 and 10 m depth.
Cubrinovski & Ishihara (1998a,b). This liquefiable deposit was divided into four dif
The S-D Model is a state-concept based model ferent materials (from “mat2” to “mat5”) as a function
that accounts for the combined effects of density and of the normalized and corrected cone tip resistance,
confining stress on sand behavior through the state- qc1Ncs, as reported by Cubrinovski et al. (2019).
concept framework. The benefit of this, in addition
to the consistent modeling of stress-density effects
on sand behavior, is that the model is a true material
model with a single set of parameters representing
a given soil across all relevant density-stress states.
Soil properties required for the S-D Model fall into
four categories: critical state line (used to define the
state of the soil relative to the reference state, the crit
ical state); plastic stress-strain parameters (defining
the shear stress-plastic shear strain relationship);
stress-dilatancy parameters (providing the link
between plastic shear strain and plastic volumetric
strain increments); and elastic parameters (determin
ing an incremental stress-strain relationship for elastic
behavior).
The S-D Model is implemented in the finite elem
ent code DIANA-J (Corporation Taisei 1997), allow
ing to perform effective stress dynamic analysis
according to a fully coupled approach.
4 CASE STUDY
The considered case study is the “YY-site” excited by

the 22 February 2011 Canterbury earthquake. Figure 2
reports the outcrop motion of the deconvolved fault-
normal component of the 22 February 2011 earth
quake record at the Canterbury Aero Club (CACS)
strong motion station, considered in the simulations.
Figure 3. Layering and VS profile.
Table 1. Soil properties.
VS D0 k
Soil Gs n (m/s) (%) (m/s)

Figure 2. Reference input motion.
mat1 2.65 0.412 93 1.406 4*10-6
mat2 2.65 0.462 91 1.073 4*10-6
The soil column consists of a 20 m deposit, where mat3 2.65 0.460 101 0.969 1*10-4
eight different soil layers, named from “mat1” to mat4 2.65 0.444 116 0.882 2*10-4
“mat8”, are identified (Figure 3). The soil properties mat5 2.65 0.412 130 0.834 3*10-4
(specific gravity, Gs, porosity, n, shear wave vel mat6 2.65 0.500 202 0.797 1*10-4
ocity, VS, small-strain damping, D0, permeability mat7 2.65 0.500 242 0.750 1*10-4
coefficient, k), of each material are the results of the mat8 2.65 0.500 266 0.712 1*10-4
soil investigations carried out by Cubrinovski et al.
878
5 PRELIMINARY VISCO-ELASTIC LINEAR strain modulus is introduced in the different codes as
ANALYSES an elastic modulus, the small-strain damping D0 is
modeled in different ways in the adopted codes.
Preliminarily visco-elastic linear analyses of the con STRATA adopted a frequency-independent for
sidered case study were performed by both loosely mulation, where the damping ratio is assumed to be
and fully coupled approaches, implemented in the constant with the shear strain level. The nonlinear
SCOSSA and DIANA-J codes respectively. These codes SCOSSA and Deepsoil used a full Rayleigh
preliminarily dynamic analyses aimed to check that damping formulation with two frequencies control.
the boundary conditions and small-strain soil behav Finally, a full Rayleigh damping formulation with
ior were reasonably reproduced in the same way in a single frequency control is used in DIANA-J. To
both approaches. To furtherly identify possible match as close as possible the damping curves used
sources of discrepancies, the same analysis was car in the different analyses, the parameters of the sin
ried out with two additional computer codes widely glefrequency control damping ratio curve in
adopted in the professional practice: the equivalent DIANA-J were defined as equivalent parameters,
linear code in the frequency domain STRATA i.e., yielding the same damping-frequency function
(Kottke et al. 2003), and the nonlinear code in the as the double frequency method used by SCOSSA.
time domain Deepsoil (Hashash et al. 2016). For this The results of the visco-elastic analyses in terms
latter, two conditions of the bedrock were considered, of profiles of maximum acceleration, shear strain
a perfectly rigid half-space, and a deformable one and shear stress are reported in Figure 4.
with a shear wave velocity of 800 m/s, a unit weight It can be observed that all the analyses showed
equal to 22 kN/m3 and a viscous damping ratio equal similar responses, with amplification of the acceler
to 0.5%. The same options are available in the com ation at the surface and higher shear strains attained
puter code SCOSSA. A deformable bedrock was also in the liquefiable layers. Closer results can be noted
adopted in the analysis performed with STRATA. in the analyses where the deformability of the seis
In the DIANA-J code, the soil column is excited at mic bedrock is considered, while higher maximum
the base by a horizontal force-time history which is acceleration and shear strains are ascribed to the ana
proportional to the known velocity time history of the lyses where a rigid bedrock is modeled. This effect is
input ground motion (Ntritos & Cubrinovski 2020). due to the lack of radiation damping through the bed
The small-strain shear modulus inferred from the rock, which usually causes the free surface motion
VS and damping ratio, D0, adopted as a target for the amplitudes of the case of deformable bedrock to be
analyses are reported in Table 1. While the small- smaller than those in the case of rigid bedrock.
Figure 4. Vertical profiles of maximum (a) acceleration, (b) shear strain and (c) shear stress attained during the
visco-elastic linear analyses.
879
6 NON-LINEAR EFFECTIVE STRESS Table 2. Parameters of the PWP model.
ANALYSES
CRR - NL ru – N/NL
6.1 Loosely coupled approach
Soil qc1Ncs CSRr α CSRt a b d
In the loosely coupled approach adopted in
SCOSSA, the nonlinear and dissipative soil behavior mat2 80 4.486 0.013 0.127 0.771 0.723 17.02
were simulated through the definition of the normal mat3 90 4.289 0.012 0.134 0.772 0.692 16.46
ized shear modulus, G/G0, and damping ratio, D, mat4 115 3.755 0.010 0.164 0.774 0.630 15.29
with the shear strain, γ. The curves proposed by Dar mat5 140 3.224 0.008 0.232 0.775 0.582 14.37
endeli (2001) as a function of the effective confining
pressure of each layer were adopted and analytically
fitted by the MKZ model (Figure 5).
6.2 Fully coupled approach

For the considered soil column, the S-D Model
parameters were determined by combining the
use of empirical relationships and generic data
for sandy soils. For non-liquefiable soils, the
target cyclic stress-strain relationship is defined
using strain-dependent stiffness degradation and
damping ratio curves, commonly employed in site
response analyses. For the liquefiable soils, the
model parameters were constrained to simulate
the target liquefaction resistance curves for the
Figure 5. Normalized shear modulus and damping ratio vs
shear strain.
four soil layers determined based on the simpli
fied liquefaction triggering procedure of Boulan
ger and Idriss (2014), as described in
The parameters of the PWP model to be assigned Cubrinovski et al. (2019). The model parameters
to the four liquefiable soils were defined through the and calibrated values of the considered case are
charts in Figure 1, as a function of the geostatic reported in Cubrinovski et al. (2019), and not
effective vertical stress at the mean depth of each reported here for sake of brevity.
layer, the qc1Ncs and fines content equal to 0. The
defined curves are plotted in Figure 6 and the related
model parameters are reported in Table 2. 6.3 Results of the effective stress analyses
The results of the effective stress analyses in terms
of profiles of maximum acceleration, shear strain
and shear stress are reported in Figure 7. The profiles
of maximum acceleration show a certain variability,
with higher values of SCOSSA compared to
DIANA-J in the liquefiable deposit between approxi
mately 5 and 10 m (Figure 7a). In the remaining
layers, the prediction of DIANA-J overcomes that of
SCOSSA, especially at the ground level, with
a predicted value of 0.38 g. The integral of the accel
eration profiles in both analyses leads to a similar
shear stress profile, with a slightly higher prediction
provided by the analysis performed with the
DIANA-J code (Figure 7b).
In the loosely coupled approach, the liquefac
tion condition is attained in the two shallow crit
ical layers, mat2 and mat3, between 2 and
5 m (Figure 7c), where the lowest value of the
cone tip resistance induces the lowest cyclic
resistance curves (Figure 6a). It should be
pointed out that at liquefaction ru= 0.9 to guaran
tee numerical stability (Tropeano et al. 2019). In
the fully coupled approach, the liquefaction is
Figure 6. (a) Cyclic resistance and (b) excess pore water triggered in the mat2, while ru<0.7 is observed in
pressure ratio curves. the other liquefiable soils. The main differences
880
Figure 7. Vertical profiles of maximum (a) acceleration, (b) shear stress and (c) excess porewater pressure ratio attained
during the nonlinear effective stress analyses.
between the two approaches are also due to the ACKNOWLEDGEMENTS

way the dissipation and redistribution of excess
pore water pressure along the soil column are The first author was supported by the Italian Minis
implemented. DIANA-J can simulate the water try of Research through the ‘Attraction and Inter
flow in the shallowest and unsaturated layer national Mobility’ project.
(mat 1), while this latter remains perfectly dry in
the SCOSSA analysis.
Additionally, the simplified PWP model generated REFERENCES
higher pore water pressure than the S-D model
between 5 and 10 m, which is anyway in accordance Boulanger, R.W. & Idriss, I.M., 2014. CPT and SPT lique
with a quite conservative approach expected in the faction triggering procedures. Report No UCD/GCM-14/
simplified soil modeling. 01, University of California at Davis, California, USA.
Chiaradonna, A. Flora, A. d’Onofrio, A. & Bilotta, E.
2020. A pore water pressure model calibration based on
in-situ test results. Soils Found. 60(2): 327–341
7 CONCLUSIONS Chiaradonna, A., Tropeano, G., d’Onofrio, A., &
Silvestri, F. 2018. Development of a simplified model
This paper investigated the performance of the CPT- for pore water pressure build-up induced by cyclic
based calibration procedure of a simplified PWP loading. Bulletin of Earthquake Engineering. 16(9):
model in predicting the liquefaction of a layered 3627–3652
liquefiable soil column. A case study from the lique Chiaradonna, A., Tropeano, G., d’Onofrio, A. &
faction observed in Christchurch after the 2011 Silvestri, F. 2019. Prediction of non-linear soil behav
earthquake is considered, and also a comparison iour in saturated sand: a loosely coupled approach for
1D effective stress analysis. Proc. 7ICEGE, Rome,
with an advanced constitutive model is provided.
June 2019.
Some limitations of the simplified approach are Corporation Taisei 1997. DIANA-J3: finite element pro
mainly associated with the overestimation of the gram for effective stress analysis of two-phase soil
generated pore pressure induced by the seismic medium. Internal report, Software science. 1997 (in
event. However, despite the simplifications adopted Japanese).
in the CPT-based calibration process, the effective Cubrinovski, M. & Ishihara, K. 1998. Modelling of sand
stress analyses performed replicated the occurrence behaviour based on state concept. Soils Found 38
of liquefaction observed at the site and furtherly (3):115–27.
encourage CPT-based model parameter definition for Cubrinovski, M. & Ishihara, K. 1998. State concept and
modified elastoplasticity for sand modelling. Soils
routine effective stress dynamic analyses.
Found 38(4):213–25.
881
Cubrinovski, M., Rhodes, A., Ntritsos, N. & Van Ballegooy, S. Tropeano, G., Chiaradonna, A., d’Onofrio, A., and
2019. System response of liquefiable deposits, Soil Dynam Silvestri, F., 2019. Numerical model for
ics and Earthquake Engineering, 124, 212–229. non-linear coupled analysis on seismic response
Kottke, A.R., Wang, X. & Rathje, E.M. 2003. Technical of liquefiable soils. Computers and Geotechnics
Manual for Strata. Geotechnical Engineering Center, 105: 211–227.
University of Texas, Austin (USA). Hashash, Y.M.A., Musgrove, M.I., Harmon, J.A.,
Hashash, Y. M. A., Musgrove, M. I., Harmon, J. A., Groholski, D.R., Phillips, C.A., Park, D., 2016.
Groholski, D. R., Phillips, C. A., & Park, D. 2016. DEEPSOIL 6.1, User Manual. Urbana, IL, Board
DEEPSOIL 6.1, user manual. Urbana, IL, Board of Trust of Trustees of University of Illinois at Urbana-
ees of University of Illinois at Urbana-Champaign. Champaign.
Ntritsos, N. & Cubrinovski, M. 2020. A CPT-based effective Robertson P.K. & Cabal K.L. 2015. Guide to Cone Penetra
stress analysis procedure for liquefaction assessment. Soil tion Testing for Geotechnical Engineering. Gregg drill
Dynamics and Earthquake Engineering,131,106063. ing. 6th edition.
882
© 2022 the Author(s), ISBN 978-1-032-31259-0
Stress increase induced by impact precast pile driving

V. Colella
Geofondazioni Ingegneria e Lavori s.r.l., Martellago, Italy
G. Cortellazzo
University of Padua, Padua, Italy
A. Dei Svaldi
Desam ingegneria e ambiente s.r.l., Italy
S. Amoroso & L. Minarelli

University of Chieti-Pescara, Istituto Nazionale di Geofisica e Vulcanologia, Italy
K.M. Rollins
Brigham Young University, Utah, USA
ABSTRACT: The paper describes the experimental activities carried out in a test site, set up to evaluate the
increase of soil stiffness and horizontal effective stress, mainly in saturated low-medium density sandy layers
and in silty sand, after the driving of tapered precast piles.
The experimentation consisted in driving some prefabricated tapered piles with different energies and spa
cing between them in an area where some soil layers had a high liquefaction potential. To evaluate the pile
driving effects on the stress state around them, preliminary CPTu and DMT tests were carried out and
repeated after the driving activity.
In particular, the change of the CPTu sleeve resistances was compared with that of the DMT KD data, to
evaluate the increase in horizontal stress using different methods, a phenomenon influencing the soil suscepti
bility to liquefaction and the pile bearing capacity.
1 INTRODUCTION consists in displacement of soil and subsequent place

ment of fluid cement grout within the dislocated
To reduce the potential liquefaction-induced volume. For RAPs, the improvement mechanisms
settlement in cohesionless soils, many ground include increased lateral pressure and increased shear
improvement techniques are adopted, among stiffness.
them vibrocompaction, rammed aggregate piers These processes can result in a measurable densi
(RAP), stone columns, drilled displacement fication and in an increase in lateral stress. There
piles, driven displacement piles, deep dynamic fore, these methods are used to mitigate cyclic
compaction, and blast densification (Han 2015). liquefaction in sandy soils.
In particular numerous studies have been per To recognize the deriving soil improvement, the
formed to evaluate the ability of deep vibratory most widely in situ tests used are the cone penetration
compaction (Van Impe et al., 1994, Massarch & test (CPT), CPT with pore water pressure measurement
Fellenius 2002, Massarch et al. 2020), of drilled (CPTu), the seismic cone penetration test, the March
displacement piles (Siegel et al. 2007, Siegel etti flat dilatometer test (DMT) or the seismic dilat
et al. 2008) and RAP (Rollins et al. 2021) to ometer test (Mayne et al. 2009).
mitigate the risk of liquefaction in sandy soils, Indeed, the CPT and the DMT measured data
while few refer to the use of driven precast enable to detect changes in strength, stiffness, and
tapered piles. horizontal stress in the soil.
With deep vibratory compaction cyclic stresses are The installation process of driven precast piles
generated in the ground, resulting in a denser particle determines, seemingly, similar phenomena, but
arrangement and changes in effective stresses. The there is a lack of knowledge about their
installation process of drilled displacement piles quantification.
DOI: 10.1201/9781003308829-131
883
Figure 1. Configurations of pile groups and relative CPTu/DMT locations.
Therefore, the paper describes the experimental 2 TEST SITE CHARACTERIZATION

activities carried out in a test site, set up to evaluate
the possible existence of these phenomena after the 2.1 Soil characteristics
driving of precast piles.
The experimentation consisted in driving The experimental site was chosen in the area where the
some prefabricated tapered piles with different most significant and widespread liquefaction phenom
energies and spacing between them in an area ena had occurred during the 2012 Emilia seismic
where some soil layers had a high liquefaction sequence. In particular the choice fell on an area near
potential. To evaluate the pile driving effects on the Mirabello village (Ferrara, Italy), previously inves
the stress state around them, preliminary CPTu tigated in detail since a blast test was carried out in
and DMT tests were carried out and repeated at a neighboring area, in 2016 (Amoroso et al. 2017).
various distances from the pile center, after the During these previous studies, geological, geotechnical
driving activity. and geophysical characterization was carried out in
One of the most interesting objectives of the proximity to the observed liquefaction evidence. The
research is to evaluate the change of the tip (qt) geotechnical in situ investigation included several
and sleeve resistances (fs) of the CPTu tests, standard penetration tests, piezocone tests, seismic
comparing them with the data acquired using the dilatometer tests and deep boreholes, while geophysical
DMT tests, principally the horizontal stress tests included down-hole tests (DH1), MASWs (Multi
index KD. Among the various phenomena influ channel Analysis of Surface Waves), P-wave and
encing the soil susceptibility to liquefaction and S-wave topographies, electrical resistivity topographies
the pile bearing capacity, the increase of hori (Amoroso et al. 2017). During the new research further
zontal stresses, resulting in a preloading effect, CPTu and DMT tests were carried out in the specific
and its accurate determination could allow test area (CPTu 3-0, 3-1 and 3-2 and DMT 1, 2 and
economic benefits in the design of deep 3 – Figure 1) to confirm the main soil characteristic
foundations. recognized previously.
884
On the basis of the abovementioned investiga
tions, a geotechnical model was determined and the
layers with the higher probability of liquefaction
were identified. In the following, the main units are
listed with their Unified Soil Classification System
(USCS) descriptors according to ASTM D2487-11
(2011):
- Topsoil from 0 to 1 m bgl (CH) – Layer 1;
- Silty clay from 1 to 4 m bgl (CH) – Layer 2;
- Clayey silt with sand from 4 to 6 m bgl (CL-CH) –
Layer 3;
- Silty sand and sandy silt (fluvial Apenninic
deposits) from 6 to 8 m bgl (ML-SM) – Layer 4;
- Silty sand (paleochannel of the Po River) from 8 to
17 m bgl (SM) – Layer 5;
- Silty sand (Syn-Glacial braided Po River deposits)
from 17 to 20 m bgl (SM) – Layer 6.
2.2 Pile test characteristics

In the area four pile groups, with four piles each, and
one with six piles were set up (Figure 1).
The piles were tapered precast piles 16.0 m long,
with a tip/head diameter of 260/500 mm and a taper
of 15 mm/m.
In pile groups N.2, N.3 and N.4 piles have the
same center distance, but were driven with differ Figure 2. Corrected cone tip resistance qt before and after
ent energies, measured using a Pile Driving Ana pile driving and R(qt) value - Pile group N.3 [E=43 kNm,
lyser (PDA). i=1.4 m].
Chosen the installation energy of the piles ensur
ing the best performance as regards the improvement
of the soil characteristics, two other pile groups were both in the center position of the group and along
set up with a larger center distance than the previous one or two sides of the group of piles.
ones. In Table 1 are shown the data regarding center The influence of tapered precast pile installation
distance and installation energy. may be represented by the ratio post-installation qt
to the pre-installation qt, the Improvement Ratio
[R(qt)], defined by the following expression:
Table 1. Installation energy and center distance of piles in
the pile groups.
Energy (E) Center distance (i)
Pile group kNm m It was computed at each measurement depth and

1 43.0 2.00 shown in the same figures.
2 25.0 1.40 The qtpost installation data differ considerably
3 43.0 1.40 depending on both the pile driving energy used and
4 43.0 1.85 the distance between the piles.
5 20.0 1.40 Considering the group of piles N.3, there is
a high increase in qt along the entire profile of the
pile, after the installation, in the layers 4 and 5
(Figure 2) with an average increase of about 2 in
soils with initial qt greater than 6 MPa and even
3 RESULTS greater with values lower than 6 MPa. In the latter
case, some of the very high values could depend on
3.1 Analysis of qt data the comparison of different type soils (clay in the
pre-installation CPTu and sand in the post-
Figures 2, 3 and 4 show the profiles of the corrected installation CPTu).
cone tip resistance (qt) acquired before and after the In the case of pile group N.2, the piles have the
pile driving activity for the different pile groups. The same center distance as group N.3, but they have been
CPTu tests after the pile driving were performed driven with lower energy (Table 1). R(qt) is
885
distance of approximately 1.85 m, R has a value sub
stantially equal to one, showing a lack of improve
ment, albeit a probable stratigraphic variability is
present at various depths. A similar situation is also
observed in the case of pile group N.1.
3.2 Analysis of KD and fs data

A similar comparison was carried out taking into
account the horizontal stress index KD of DMT tests
carried out before and after the installation of the
piles. The DMTs after installation were performed
along one side of the pile groups.
KD of the dilatometric test is strictly correlated
with the horizontal earth stress coefficient and there
fore with the horizontal stresses.
As in the previous cases, the influence of tapered
precast pile installation may be represented by the
ratio post-installation KD to the pre-installation KD,
the Improvement Ratio [R(KD)], defined by the fol
lowing expression:
Figure 3. Corrected cone tip resistance qt before and after pile

driving and R(qt) value - Pile group N.2 [E=25 kNm, i=1.4 m].
Also in this case for pile group N.3 (Figure 5) there

is a marked increase in post-installation KD with
R(KD) values between 3 and 5, while for pile group
N.4 the improvement is absent, except in the loose
silty sand and sandy silt layer between 6.0 and
8.0 m, having a low cone resistance (Figure 6).
Figure 4. Corrected cone tip resistance qt before and after

pile driving and R(qt) value -Pile group N.4 [E=43 kNm,
i=1.85 m].
significantly greater than 1 (Figure 3) only in the first

meters of the silty sand layer 5, probably due to the
effect of the lateral spreading of the soil, and is equal
to 1 at deeper levels.
Finally, in the case of pile group N.4 (Figure 4),
in which the piles have been driven with the same Figure 5. KD before and after pile driving and R(KD)
energy used for those of stand 3, but with a center value - Pile group N.3 [E=43 kNm, i=1.4 m].
886
Both ratios show the same phenomena, improve
ment of soil characteristics in correspondence
of pile group N.3 (Figure 7) and modest if not
zero improvement in pile group N.4 (Figure 8).
Therefore, both, again, highlight the decrease
of the improvement as the distance between
the piles increases. The R(KD) ratio also is more
sensitive than R(F) to the effects of the pile
driving.
3.3 Cumulative settlement prediction

Adopting the Zhang et al. (2002) method to estimate
liquefaction-induced ground settlements using CPT
data, the effectiveness of the driving of piles on the
decrease of the potential liquefaction effects was
determined.
The Zhang approach combining the CPT estimate
liquefaction resistance with laboratory test results on
clean sand evaluates the liquefaction-induced volu
metric strains for sandy and silty soils.
Figure 6. KD before and after pile driving and R(KD) value
- Pile group N.4 [E=43 kNm, i=1.85 m]. Zhang et al. (2002) assume that little or no
lateral displacement occurs after the earthquake,
such that the volumetric strain will be equal
or close to the vertical strain. Integrating
with depth the vertical strain in each soil layer,
the potential liquefaction-induced ground settle
ment due to an earthquake is estimated by
Equation 4,
Figure 7. R(KD) and R(F) - Pile group N.3 [E=43 kNm,

i=1.40 m].
Robertson (2016) highlighted that the normalized

CPT sleeve resistance (F = fs/σ’vo) can be used to
estimate the DMT KD; therefore, the trend along the
depth between R (KD) and R (F) of various pile
groups was compared, where R (F) is the following Figure 8. R(KD) and R(F) - Pile group N.4 [E=43 kNm,
expression i=1.85 m].
887
Zhang et al. (2002) calculation method, the improve
ment of the soil behavior achievable through the
installation of driving piles was clearly highlighted.
REFERENCES
Amoroso, S., Milana, G., Rollins, K.M., et al. 2018. The
first Italian blast-induced liquefaction test (Mirabello,
Emilia-Romagna, Italy): description of the experiment
and preliminary results. Annals of Geophysics 60(5);
S0556.
ASTM D2487-11, 2011. Standard Practice for Classifica
tion of Soils for Engineering Purposes (Unified Soil
Classification System).
Fontana, D., Amoroso, S., Minarelli, L. & Stefani, M.
2019. Sand liquefaction induced by a blast test: new
insights on source layer and grain size segregation
mechanisms (late Quaternary, Emilia, Italy). Journal of
Sedimentary Research, v. 89, 13–27.
Han, J. 2015. Principles and practices of ground improve
ment. Hoboken, NJ: Wiley.
Massarsch, K.R. & Fellenius, B.H. 2002. Vibratory com
paction of coarse-grained soils. Canadian Geotechnical
Journal 39(3): 695–709.
Massarsch, K.R., Wersäll, C. & Fellenius, B.H. 2020. Hori
Figure 9. Post liquefaction settlement using Zhang et al. zontal stress increase induced by deep vibratory
method (2002) - Pile group N.3 [E=43 kNm, i=1.4 m]. compaction. Proceedings of the Institution of Civil
Engineers – Geotechnical Engineering 173(3):
228–253.
Mayne, P.W., Coop M.R., Springman S., Huang A.B. &
Zornberg, J. 2009. Geo material behavior and testing. In
Hamza M, Shahien M & El-Mossallamy Y. (eds), Pro
where w is the calculated liquefaction-induced ceedings of the 17th International Conference on Soil
ground settlement at the CPT location; εvi is the Mechanics and Geotechnical Engineering (ICSMGE).
postliquefaction volumetric strain for the soil sub Millpress/IOS Press, Rotterdam, the Netherlands, vol. 4,
layer i; Δzi is the thickness of the sublayer i; and j is pp. 2777–2872.
Robertson P.K. 2016. Estimating Ko in sandy soils using
the number of soil sublayers. the CPT. In Lehane, Acosta-Martínez & Kelly (Eds).
The analysis of data related to the pile group N.3 Geotechnical and Geophysical Site Characterisation 5 –
(Figure 9) shows a great decrease of potential post ISC’5, Gold Coast, Queensland, Australia, September 5 –
liquefaction settlement and consequently that the 9,2016.
improvement is large. Therefore, with groups of more Rollins, K.M., Amoroso, S.; Paul Andersen, P., Tonni, L. &
piles than in this case, significant improvements can be Wissmann, K. 2021. Liquefaction mitigation of silty
achieved from the point of view of a potential soil sands using rammed aggregate piers based on
liquefaction settlement lower than the virgin soil. blast-induced liquefaction testing. J. Geotech. Geoen
viron. Eng., 2021 , 147 (9):04021085.
Siegel, T.C., NeSmith, W.M., NeSmith, W.M. & Cargill, P.
4 CONCLUSIONS E. 2007. Ground improvement resulting from installa
tion of drilled displacement piles. 32nd DFI Annual
The CPTu and DMT in situ tests were used to evalu Conference, Colorado Springs, CO.
Siegel, T.C., NeSmith, W.M. & NeSmith, W.M. 2008.
ate the improvement of the soil characteristics due to Increase in Cyclic Liquefaction Resistance of Sandy
the driving of precast tapered piles. Soil Due to Installation of Drilled Displacement Piles.
The data acquired using the CPT and DMT tests Earthquake Engineering and Soil Dynamics IV Confer
show that the installation of prefabricated tapered piles ence, Sacramento CA.
determines an increase of the soil resistance, pointed out Van Impe, V.F., De Cock, F., Massarsch, K.R. & Mengé, P.
by the increase in qt, as well as a change in effective lat 1994. Recent experiences and developments of the res
eral stress highlighted by both the increase of F and KD. onant vibrocompaction technique. Proceedings of Inter
The phenomenon is clearly remarkable up to national Conference on Soil Mechanics and Foundation
a distance of about 4-5 pile diameters; beyond this dis Engineering, New Delhi, India. Balkema, Rotterdam,
the Netherlands, vol. 3, pp. 1151–1156.
tance the phenomenon tends to run out. Zhang, G., Robertson, P.K. & Brachman, R.W.I. 2002.
In the present case, the driving energy played a non- Estimating liquefaction-induced ground settlements
negligible role in the amount of the improvement. from CPT for level ground. Canadian Geotechnical
Referring to the potential post-liquefaction settle Journal. 39(5): 1168–1180. http://dx.doi.org/10.1139/
ment due to an earthquake and determined with the t02-047
888
© 2022 the Author(s), ISBN 978-1-032-31259-0
CPT based liquefaction potential of flood defences in The Netherlands

T. de Gast & K.G. Gavin
Section of Geo-Engineering, Faculty of Civil Engineering and Geosciences, Delft University of Technology, Delft,
The Netherlands
P.D. Notenboom & R. Abraimi

Afdeling Kennis & Advies, Waterschap Hollandse Delta, Ridderkerk, The Netherlands
C. Reale
Department of Architecture & Civil Engineering, Centre for Infrastructure, Geotechnical and Water Engineering
Research (IGWE), University of Bath, Bath, UK
ABSTRACT: The paper describes a study on the liquefaction potential of flood defences along the rivers run
ning through the delta area of the Netherlands. The study concentrates on an area south of Rotterdam. The
dykes used as primary flood defences protect an urban, rural, and industrial area of 102,400 ha. In this paper the
data from more than 4200 Cone Penetration Tests, CPT traces are used to assess more than 200 km of dykes.
The pore pressure, u2 data is analysed, then used to separate the material response into contractive and dilative
zones. Using the separation of liquefaction susceptible soils, and geometry of the riverbed a regional hazard map
is generated. The choices for the data visualisation and their effect on the generated map are discussed and pre
sented. The final liquefaction susceptibility map is used by the water governing authority Waterschap Hollandse
Delta as a decision-making tool to improve the efficacy of liquefaction hazard assessment such as the location
and return period of bathymetry measurements, and the scale of site- and laboratory investigation.
1 INTRODUCTION and in turn reported to the national government. As

part of their most recent assessment, WSHD conducted
In the Dutch delta, the majority of rural, industrial, and a large site-investigation campaign consisting of more
residential land are founded below mean sea level. than 5,000 (or exactly 5,137) Cone Penetration Tests
These areas are reliant on flood defences for everyday (CPT), 800 boreholes and substantial laboratory tests
protection. In 1953 a major flood following a heavy on the soft deltaic soils. This study utilises this dataset
storm led to some of these dykes failing, which in turn to check for the potential of sand liquefaction beneath
led to large scale flood inundation and loss of life. the Dykes.
Since 1953 extensive flood defences along the delta Static liquefaction can occur when loose satur
have reduced the risk of coastline flooding signifi ated cohesion less soils are loaded rapidly, for
cantly. Dykes in the Netherlands are separated into two example due to slope over-steepening. Erosion of
general types i) primary dykes which protect along the rivers in the governance area of WSHD affects the
coast and rivers that have large fluctuations in water slopes of the river channel. This erosion can lead to
level during normal operation; and ii) regional dykes, underwater slope instability potentially leading to
which surround polder-systems where water levels are a static liquefaction type failure. The current
artificially maintained with little to no fluctuation. The assessment criteria (Rijkswaterstaat 2019) proposes
responsibility to protect against inundation in the Neth a stepwise approach, where high hazard locations
erlands has been given to separate water governing are identified based on their geometry, informed by
authorities. The water governing authority Waterschap both the bathymetry of the river basin and the phys
Hollandse Delta (WSHD) are responsible for the ical geometry of the dyke. High hazard locations
safety of the primary and regional dykes around the are then subjected to more detailed analysis. This
Islands in the province of South Holland. There are study proposes to add to the geometrical data used
approximately 200 km of primary dykes preventing to pre-screen high hazard locations by using the
the local area from flooding. These dykes need to be material behaviour chart (Robertson 2016) to iden
assessed frequently, considering several different fail tify underlying soil susceptible to liquefaction in
ure mechanisms and their safety needs to be evaluated the WSHD area.
DOI: 10.1201/9781003308829-132
889
2 GEOLOGY OF AREA along the primary dykes in the governing area of
WSHD. The potential for liquefaction is determined
The area of investigation lies in the Rhine-Meuse by the layer thickness of the contractive sand or the
delta which discharges to the North Sea. The river relative density of the material when it is less than
basin consists of silty clay and organic (clay) 66%. If the relative density decreases or the thick
deposited on top of a Pleistocene sand layer typic ness of the contractive sand layer increases the rela
ally located at depths from Dutch Ordnance Datum tive risk of liquefaction in the area increases.
Level (NAP), NAP -15 m to NAP -20 m depth
(≈15 to 20m below ground level in this low-lying
coastal region). The clay and organic soil layers are 4 CURRENT NATIONAL GUIDELINES
intersected by several meandering streams forming
channel belts, these are mapped by (Cohen, et al. The current Dutch Water Act (DWA, BWBR0025458)
2012). Currently the rivers are ‘locked’ in place, which came into effect on January 29th 2009, gives
meaning that old river channels locations have rules and design recommendations for the use and
been fixed in location by engineering works. Dykes maintenance of Dutch water systems. The water gov
have been built to prevent flooding which can erning authorities in the Netherlands are the primary
change the natural flow path of a river whilst add executive organisations tasked with ensuring the water
itionally the riverbed is frequently dredged to allow system and its defences comply with the DWA. Part
for the safe operation of inland ships serving the of the DWA includes design rules for assessing direct
Port of Rotterdam and maintain the discharge cap and indirect failure mechanisms of dykes. With lique
acity of the river. faction recognised as one of the main indirect failure
mechanisms, meaning that should an event happen, it
will most likely not be the direct cause of a dyke fail
3 METHODOLOGY ure but may contribute to one.
The recommended approach contains three assess
Loose silts and fine sands will tend to contract when ments, each more detailed than the one before. The
loaded. During regular static loading where loads are test are: a simple test, a detailed test, and a custom
applied slowly (e.g. during construction of a building test. The methods of the simple and detailed test are
or dyke) pore pressures dissipate, and settlement prescribed in the technical guidelines whilst the
occurs. However, if the rate of loading (or unloading) custom test allows one to utilise the latest scientific
is high, e.g. during erosion, excess pore pressures insights to make the assessment.
develop and may lead to liquefaction and sudden large The simple test is a geometric test with conserva
failure e.g. as shown in Figure 1. In dense sands this is tive assumptions. The detailed test considers soil
not an issue as the particles want to dilate under load properties, geometry, water levels and load types.
ing, increasing the voids between particles and conse The test itself determines whether there is enough
quently the volume for water to occupy. One of the dike forefront left after a liquefaction event to pre
indicators of contractive/dilative behaviour is a sands vent failure (inundation). The DWA is revaluated
relative density. regularly to incorporate the latest insights in water
In an extensive site-investigation campaign safety. The latest official code came into effect in
WSHD has performed, 2 CPTs per 100 m length on 2017 (WBI 2017). While the next instalment is
both the top of the dyke and the inner slope (the expected in 2023.
slope facing the water being the outer slope). This Currently there is a large difference in the com
dataset is used to assess the liquefaction potential plexity and as a result the level of detail required in
parameter determination between the simple test and
the detailed test. Incorporating information from the
material behaviour chart into the simple test would
allow the WSHD to identify areas more susceptible
to liquefaction. This would facilitate them to focus
their subsequent detailed assessments on areas with
a high liquefaction hazard.
5 DETERMINING CONTRACTIVE ZONES
To identify whether a soil is contractive and subse

quently susceptible to liquefaction, the following
CPT based procedure developed by (Robertson
1990, Robertson 2016) has been followed. First
Figure 1. Example liquefaction occurrence, 1968-10-04,
measured qc values are normalised into Qt using the
Oud-Kempenshofstedepolder (Tholen), by Kotvis, (1986).
following relations see equations 1 to 3:
890
where pa being the atmospheric reference pressure
and n is the stress exponent defined by equation 6.
After obtaining Qtn and Fr and using the boundaries
based on Ib and CD suggested by (Robertson 2016)
the soil can be classified by the soil type behaviour.
The relative density, Dr of young, uncemented silica
Where Qt is the normalized cone resistance, qt is sands (Kulhawy and Mayne 1990) can be obtained
the cone resistance corrected for water effects, σv0 is using:
the current in-situ total vertical stress, σ0 v0 is the cur
rent in-situ effective vertical stress, Fr is the normal
ized friction ratio, fs is the measured sleeve
resistance, qc is the measured cone resistance, u2 is
the shoulder penetration pore pressure (behind cone
tip) and a is the cone area ratio. When normalised
CPT parameters have been obtained the soil behav Figure 2 gives two examples of analysed CPTs
iour indices Ic and IB as well as the contractive dila using the soil behaviour chart as suggested by
tive boundary (CD) can be obtained using the Robertson (2016). The following zones are named,
following relations, see Equations 4 to 8. CCS (clay-like, contractive, sensitive) CC (clay-like,
contractive), CD (clay-like, dilative), TC (transi
tional, contractive), TD (transitional, dilative), SC
(sand-like, contractive) and SD (sand-like, dilative).
Figure 2a shows a CPT having a large number (606)
of contractive points, and in Figure 2b fewer (296)
contractive points are found.
6 IDENTIFICATION AND VISUALISATION OF

LIQUEFACTION HAZARD
To identify the liquefaction potential the CPTs are

automatically processed. For the calculation of
Figure 2. Example soil behaviour chart after Robertson (2016) for two CPTs a) containing a large number (606) of con
tractive soil points and b) containing fewer (296) contractive points.
891
Figure 3. Governing area of WSHD overlaid with the old and current channel belts map, SC-SD-TC-TD 66% window is
200cm, CPT point size scaled to number of liquefaction susceptible points.
Figure 4. Governing area of WSHD overlaid with the old and current channel belts map, SC-TC 100% window is 200cm,
CPT point size scaled to number of liquefaction susceptible points.
effective stress, the phreatic level is assumed to be analysed using a moving window approach, once
fixed at the reference level of NAP +0.0 m with points in the window are identified as being suscep
hydrostatic conditions assumed with depth. The tible to liquefaction, they are aggregated and the
liquefaction potential is analysed over a fixed depth total number of susceptible points at a CPT location
of NAP +0.0 m to NAP -12.0 m. Each CPT is is used as an indicator of liquefaction hazard.
892
For all the CPTs analysed, two different criteria of large dyke failures to occur these areas will be more
liquefaction susceptibility have been used i) con closly monitored by WSHD.
tractive sands and contractive transitional soils and The analyses included some simplifications that can
ii) sands and transitional soils with a relative density be examined in future studies. Rather than assuming
below 66%. A moving window is used to tally the hydrostatic pore pressures, an accurate assessment of
number of liquefaction susceptible points, the lique in-situ pore water pressures should be conducted. The
faction susceptible points are counted if all points impact of dyke geometry and the resulting effective
meet the requirement criteria. After processing, the stress conditions should be included. Combining the
highest liquefaction potential is associated to the liquefaction analysis with the geometry and bathym
highest number of concurrent liquefiable points. etry of the riverbed is an area of interest for the lique
Figure 3 shows the results of a batch analysis look faction susceptibility analysis. It is important to
ing within all CPTs for zones that contain a relative confirm the applicability of the relative density correl
density below 66% for transitional soils and contract ation and the trigger level of Dr below 66% to the
ive sands, it shows that most of the locations tested in Pleistocene sand layers considered. And finally,
the WSHD area contain significant volumes of a major question remains, on what is the optimum
loosely packed sand that may be susceptible to static moving window size for highlighting when contractive
liquefaction. Figure 4 presents the results for 2.0 m or soil layers can cause large liquefaction induced failure.
larger continuous contractive sands and transitional
soil layers. Figure 4 still highlights a number of lique
faction prone areas in the governing area of WSHD, REFERENCES
however much fewer are identified than when the
Cohen, K., E. Stouthamer, H. Pierik and A. Geurts (2012).
relative density approach is used. “Digitaal Basisbestand Paleogeografie van de Rijn-Maas
Delta.” Dept. Fysische Geografie. Universiteit Utrecht.
Digitale Dataset.
7 DISCUSSION AND CONCLUSIONS Kotvis, C (1986). Photo of liquefaction occurrence,
1968-10-04, at Oud-Kempenshofstedepolder (Tholen),
This paper presents two analyses performed using Beeldbank Zeeland, recordnr. 6895
a large dataset of CPTs in the governing area of Kulhawy, F. H. and P. W. Mayne (1990). Manual on esti
Waterschap Hollandse Delta (WSHD). Each CPT mating soil properties for foundation design. Electric
Power Research Inst., Palo Alto, CA (USA); Cornell
profile was analysed for the liquefaction susceptibil
Univ., Ithaca, NY (USA).
ity based on relative density and contractive behav Rijkswaterstaat, (2019). Schematiseringshandleiding
iour. Based on the relative density analysis, many zettingsvloeiing. Ministerie van Infrastructuur en
dykes were shown to be built on soil profiles con Waterstaat
taining significant depths of loose sand. This loose Robertson, P. (1990). Soil classification using the cone
material has the potential to liquefy but does show penetration test. Canadian Geotechnical Journal. 27(1),
consistent contractive behaviour over 2.0 m depths 151–158.
both according to the CPT analysis. The analysis Robertson, P. K. (2016). Cone penetration test (CPT)-based
based on the soil behaviour type highlights several soil behaviour type (SBT) classification system — an
update. Canadian Geotechnical Journal 53(12),
locations where continuous depths of soil that would
1910–1927.
exhibit contractive behaviour are observed. As this WBI (2017). Beoordelingsinstrumentarium (WBI2017).
has the potential to trigger liquefaction and cause Ministerie van Infrastructuur en Waterstaat
893
© 2022 the Author(s), ISBN 978-1-032-31259-0
Application of the CPT for liquefaction assessment of gravelly reclamations

at the port of Wellington
R. Dhakal & M. Cubrinovski
University of Canterbury, Christchurch, New Zealand
J.D. Bray
University of California, Berkeley, USA
ABSTRACT: Widespread liquefaction occurred in the reclamations at CentrePort, Wellington (New Zealand)
during the 2016 Kaikōura earthquake (Mw7.8) which produced a relatively moderate seismic demand (peak
horizontal ground acceleration of 0.2-0.3g at the site). Most of the damage occurred in the end-dumped fills
(i.e., the Thorndon reclamation) which are comprised of 60-80% gravels and 20-40% sand-silt mixtures. This
study presents results from CPT-based liquefaction assessment on a representative profile from the Thorndon
reclamation using both a semi-empirical simplified method and advanced dynamic effective stress analysis.
The simplified assessment results are generally consistent with observations since the CPT is able to capture
the response of the finer sand-silt fraction in the matrix, which likely controlled the liquefaction behavior
during the Kaikōura earthquake. The preliminary effective stress analysis provides insights on the timing and
onset of liquefaction, and processes involving dynamic interactions within the deposit, thus illustrating import
ant response characteristics which are beyond the capability of simplified methods. This paper illustrates the
use of the CPT in performing conventional simplified and dynamic liquefaction analysis when applied to non-
conventional or problematic soils for the assessment with reference to a well-documented case study.
1 INTRODUCTION 2 CENTREPORT RECLAMATIONS
The cone penetration test (CPT) is commonly used for 2.1 Reclamation history
geotechnical characterization and liquefaction assess
ment (e.g., Boulanger and Idriss, 2014; Robertson and Wellington is located at the southern tip of the
Wride, 1998). However, several issues exist in the use North Island of New Zealand. The city was devel
of the CPT for assessment of soil liquefaction. For oped over the past 170 years, where the original
example, there are challenges in the application of sim coastline sat approximately 200 m to 500 m inland
plified semi-empirical CPT-based methods, developed from the present coastline delineating a belt of
primarily using case histories of naturally deposited reclaimed land. An aerial view of the port of Wel
sands and sands with fines, to nonstandard soils such as lington (CentrePort) is shown in Figure 1, which
reclaimed gravels (Cubrinovski et al., 2019). In add was developed over several periods of major rec
ition, there is a growing need for advanced numerical lamation. In the most recent reclamation works
techniques such as seismic effective stress analysis, (from 1965 to 1976), the area south of the buried
which can provide insights on liquefaction behavior seawall depicted in Figure 1 (i.e., the Thorndon rec
beyond the scope of the simplified procedures. When lamation) was constructed by end-tipping of grav
applying an effective stress analysis, the constitutive elly soils sourced from nearby quarries
model can be calibrated using either laboratory data on (Cubrinovski et al., 2017). Gravelly soils were
high-quality samples or empirical relationships (e.g., dumped into the sea to construct fills of 10 m to
Ntritsos and Cubrinovski, 2020). This paper illustrates 22 m thickness. The fill below 2-3 m depth, align
the use of the CPT for simplified and advanced lique ing with the approximate depth to the water table,
faction analysis for a well-documented case study of is uncompacted. Underlying these reclamations are
gravelly reclamation from the port of Wellington, New Pleistocene deposits consisting of weathered allu
Zealand. In the advanced analyses, CPT-based relation vium, colluvium, and marine deposits. The grey
ships are used to calibrate the constitutive model for wacke basement rock under these deposits is at
liquefaction evaluation of the gravelly reclamations. 100-150 m depth.
DOI: 10.1201/9781003308829-133
894
samples, results of which are summarized in
Figure 2b, which indicates the fill consists of 60% to
80% gravels (i.e., > 2 mm particle size) with the
remaining 20% to 40% of the mixture consisting of
silt and sand fractions. The fines content (FC) is typ
ically between 5% and 15%.
Figure 1. Aerial view of CentrePort (shaded in red), an old

buried seawall, locations of two strong motion stations, and
the location of one CPT (CPTu042). The inset shows the
location of Wellington and the source and rupture propaga
tion of the 2016 Kaikōura earthquake (modified from
Cubrinovski et al., 2018).
Figure 2. (a) Gravelly ejecta observed after the Kaikōura

earthquake at the Thorndon reclamation (Cubrinovski
2.2 Liquefaction performance during the 2016 et al., 2017), and (b) grain-size distribution of the Thorndon
Kaikōura earthquake reclamation gravelly fill (Dhakal et al., 2020a).
On 14 November 2016, the Mw7.8 Kaikōura earth
quake occurred in the South Island of New Zealand
(location depicted in the inset of Figure 1). The clos 2.3 In-situ site investigations
est distance from the causative faults to CentrePort
was approximately 60 km, resulting in relatively Comprehensive site investigation following the Kai
long duration but moderate amplitudes of ground kōura earthquake characterizes the subsurface soils
shaking (i.e., estimated horizontal peak ground at CentrePort by using over 100 CPTs, soil samples
accelerations, PGA, of 0.2-0.3g at the port). collected from over 30 boreholes, and shear-wave
The Kaikōura earthquake triggered widespread velocity profiles measured using direct-push and sur
liquefaction that severely damaged port buildings face wave methods (Cubrinovski et al., 2018;
and wharves (Cubrinovski et al., 2017). The most Dhakal et al., 2020a; Dhakal et al., 2020b). The trace
severe damage occurred in the Thorndon reclamation of measured cone tip resistance (qc) and calculated
where up to 1 m horizontal movement and over half soil behavior type index (Ic) of a representative CPT
a meter of vertical settlements were observed. Large profile from the Thorndon reclamation (CPTu042)
volumes of gravelly ejecta with thickness up to 150 are shown in Figures 3a and 3b, respectively.
200 mm were also observed on the ground surface The gravel-sand-silt fill from 3 m to 14.7 m depth
(e.g., Figure 2a). Reconnaissance efforts included show characteristic values of qc (6–8 MPa) and Ic
collection and index testing of ejected gravelly (1.9–2.3) throughout depth despite its complex soil
895
composition. The relatively low penetration resist However, several challenges and limitations in the
ance and intermediate values of Ic for soils contain simplified liquefaction evaluation have been identi
ing large amounts of gravel imply that these fills fied for the gravelly reclamations (Cubrinovski et al.,
comprising gravel-sand-silt mixtures display soil 2019). For example, there are several areas in other
behavior type typical for silty sand and reflects gravel reclamations north of the old buried seawall
a dominant influence of the finer fractions (sand and (depicted in Figure 1) with poor estimates of lique
silt) in the soil matrix. This in turn implies that the faction damage (Dhakal et al., 2020a). Furthermore,
finer sand-silt fractions will have a governing role in the simplified procedures provided much smaller
the seismic response of the fill, despite the fact that degree of variation in the ground performance as
the gravel-size particles are dominant by volume compared to actual observations (Dhakal et al.,
(Cubrinovski et al., 2019). The coarser gravel par 2020b). Therefore, additional insights are needed
ticles only have a minor influence, as manifested in from more rigorous dynamic site response analyses
the observed occasional spikes in qc (associated with to enhance our understanding of soil behavior and
low Ic values). evolution of liquefaction in these reclamations.
3 SIMPLIFIED LIQUEFACTION ANALYSIS
The CPT data is used to evaluate liquefaction trig

gering using the Boulanger and Idriss (2014) pro
cedure, which compares the earthquake-induced
cyclic stress ratio (CSR) to the cyclic resistance
ratio (CRR) of the soil to estimate a factor of
safety against liquefaction triggering. The conven
tionally adopted criterion of Ic < 2.6 is used to
identify soils susceptible to liquefaction. The lique
faction assessment is conducted assuming a level
ground condition for a Mw7.8 event with PGA =
0.25 g which represents the geometric mean of the
shaking induced by the 2016 Kaikōura earthquake
(Cubrinovski et al., 2018). The groundwater level
is carefully considered in the analyses including
effects of tidal fluctuations using data from nearby
piezometers to estimate the water level at the loca
tion of the CPTs, at the time of the earthquake
(Dhakal et al., 2020b). The probability of liquefac
tion triggering of PL = 50% is used for the back- Figure 3. Profiles of CPTu042 (a) cone tip resistance (qc),
analysis of this case history instead of the deter (b) soil behavior type index (Ic), and (c) cyclic stress ratio
ministic value of PL = 16%, which is convention (CSR) and cyclic resistance ratio (CRR) calculated using
ally used in forward analysis and design. The FC the Boulanger and Idriss (2014) procedure under the seis
is approximated as 15% in the analysis. mic demand of the 2016 Kaikōura earthquake.
Figure 3 shows the computed profiles of CRR and
CSR alongside the qc and Ic data for the representative 4 EFFECTIVE STRESS ANALYSIS
Thorndon reclamation CPT. The results in Figure 3c
indicate CRR of the reclamation is well below the seis 4.1 Numerical model
mic demand imposed by the 2016 Kaikōura earthquake
(i.e., the CSR). Hence, the CPT-based simplified pro A fully coupled nonlinear effective stress analysis of
cedure indicates that liquefaction should have been trig saturated soil is used to perform 1D site response
gered throughout the depth of the Thorndon gravelly analysis of the Thorndon reclamation soil profile
reclamation in CentrePort. As widespread liquefaction shown in Figure 3. In the analysis, an input earth
effects (e.g., soil ejecta, vertical settlements, and hori quake motion is applied at the base of a 1D soil
zontal ground movements) were observed at the port column, a finite element model comprised of 2D
after this event, the simplified CPT-based liquefaction quadratic elements constrained to deform in simple
triggering procedure provides results that are consistent shear mode of deformation. A key requirement in
with the field observations. This agreement reflects the such analyses is the use of a realistic ground motion,
fact that the sand and silt fractions in the soil matrix for the site and earthquake of interest, to input at
have a governing influence during the earthquake, a reference subsurface layer (Ntritsos et al., 2021).
which the CPT captures. On this basis, general applic Rigorous scrutiny of deconvolved motions from sev
ability of CPT-based simplified liquefaction assessment eral nearby strong motion station (SMS) sites have
procedures to the complex gravelly reclamations at shown that the deconvolved motion from a SMS site
CentrePort could be justified. on native deposit 590 m from the location of the
896
analyzed soil profile (i.e., VUWS; location shown in Following the determination of the simplified pro
Figure 1) is an appropriate motion for input in 1D file, layers with qc1Ncs < 170 and Ic < 2.6 are then
analysis (Dhakal et al., 2022). Note that while the deemed potentially liquefiable, whereas all other
entire deconvolved ground motion (i.e., Figure 4b) is layers are considered as non-liquefiable. The adopted
used as a base input motion in the dynamic analysis, modelling approach for the liquefiable and non
the simplified analysis only uses the PGA of the liquefiable layers is summarized in Figure 5d and
motion (indicated in Figure 4a) recorded at the sur explained in detail in the subsequent section.
face of the closest SMS site 330 m from the location
of the analyzed soil profile (CPLB; location shown
4.3 Constitutive model calibration
in Figure 1) in conjunction with the earthquake mag
nitude. By considering the entire ground motion This study utilizes the Stress-Density model, which is
time history, the dynamic analysis can provide an elastic-plastic constitutive model specifically
insights on the timing, onset, and evolution of lique tailored for analysis of soil liquefaction problems
faction throughout the depth of the fill. (Cubrinovski and Ishihara, 1998a; 1998b). The model
is capable of accurately simulating highly nonlinear
stress-strain behavior of liquefiable soils under both
monotonic and irregular cyclic loading. A key feature
of the model is that it utilizes the state-concept
approach for modelling the effects of density and con
fining stress and therefore can consistently represent
the behavior of sand at any density and confining
stress by using the same set of material parameters.
A key requirement of the model is to accurately
simulate the accumulation of shear strains and conse
quent development of excess pore-water pressures
under cyclic loading. This ability needs to be demon
strated through a series of element-test simulations,
which are also used to calibrate the key response
characteristic of the constitutive model that will allow
to accurately simulate target liquefaction resistance
curves (LRCs). Here, LRC represents the combination
of CSR and number of cycles (Nc) required to cause
Figure 4. (a) PGA of the surface recorded horizontal ground liquefaction or typically 5% double-amplitude strain,
acceleration at CPLB used as input for simplified liquefac
tion assessment, and (b) the entire horizontal ground acceler
for a specific density and vertical effective stress of
ation time history of the deconvolved motion at VUWS used a given soil. Target LRCs can be obtained either from
as input in the effective stress analysis. laboratory tests on soil samples or based on empirical
relationships developed within the simplified trigger
ing methods (e.g., Boulanger and Idriss, 2014). The
latter approach is employed in this study, in which
the Stress-Density constitutive model is calibrated to
4.2 Simplified CPT-based soil profiles simulate LRCs derived from the Boulanger and Idriss
Unlike the simplified methods, in which a nearly con (2014) simplified procedure, with calibration details
tinuous record of the CPT trace can be used for provided in (Ntritsos and Cubrinovski, 2020). Follow
liquefaction calculations, discretization of the profile ing such calibration of the model, the only input
based on its stratification is required for the numer required from the user for the analysis is a CPT pro
ical model of the effective stress analysis. This study file with discretized layers of qc1Ncs and Ic.
illustrates the use of the CPT in identifying the strati Figure 6 depicts the target LRCs based on Bou
fication for this purpose. The nearly continuous CPT langer and Idriss (2014), shown with solid lines, for
profile is first discretized into several distinct layers five different densities (i.e., qc1Ncs values), whereas
or depth intervals over which the CPT data can be the Stress-Density model simulations of the LRCs
approximated by constant values of the clean-sand are shown with open symbols. It illustrates that the
equivalent penetration resistance (qc1Ncs) and Ic. In Stress-Density model can accurately simulate LRCs,
this study, the algorithm developed by Ntritsos and or stress-strain behavior and excess pore water pres
Cubrinovski (2020) is adopted for the determination sure development, for soils with a wide range of
of the simplified soil profile. The resulting simplified qc1Ncs (i.e., from very loose to dense soils).
soil profile is illustrated in Figure 5 with the red solid Non-liquefiable layers are modelled with the
line. The layer discretization is fine enough to allow Stress-Density model’s pore-water pressure gener
detection of thin seams of liquefiable soils while ation feature turned off using shear modulus and
keeping the simplified qc1Ncs and Ic profiles as close damping relationships with shear strain as target
as possible to the actual, more irregular, CPT traces. curves in the element test simulations.
897
Figure 5. Profiles of CPT (a) cone tip resistance (qc), (b) clean-sand equivalent cone tip resistance (qc1Ncs), and (c) soil
behavior type index (Ic) with measured data (shown in black) and simplified (layered) data (shown in red), and (d) associ
ated modelling approaches for a liquefiable layer (shown in red) and non-liquefiable layers (shown in green).
the start of the dynamic analysis (computational

time). A second, deeper layer (11.5-13 m) liquefies
several seconds later. The onset of liquefaction in
these two layers reduces the demand in the deposit
as illustrated with the decrease in amplitude and loss
of high-frequency content in the acceleration time
histories. This in turn reduces pore pressure build-up
and development of liquefaction in the remaining
part of the profile.
These results are sensitive to key modelling deci
sions made by the user. For example, the shallow
liquefied layer in Figure 7 includes a very thin soil
layer at 7 m depth with Ic near the liquefaction sus
ceptibility threshold of 2.6, which may have higher
liquefaction resistance than that adopted in the ana
lysis. Also, the intensity of the input ground motion
Figure 6. Semi-empirically derived target liquefaction will have a significant effect on how excess pore
resistance curves (solid lines; Boulanger and Idriss, 2014), water pressures and dynamic interactions develop
and Stress-Density model simulated liquefaction resistance during the deposit. The seismic effective stress ana
curves (open symbols; Ntritsos and Cubrinovski, 2020) for lysis provides means to investigate important mech
different clean-sand equivalent resistances (qc1Ncs).
anisms of the liquefaction response which are
beyond the capacity of the simplified procedures.
4.4 Analysis results
Figure 7 shows the results of the analysis for the 5 CONCLUSIONS
selected CPT profile (CPTu042) including distribu
tion of maximum shear strains and pore-water pres This study illustrates the use of the CPT for liquefaction
sure ratios at different time sections, throughout the assessment of reclaimed gravelly soil in the port of Wel
depth of the profile. Computed acceleration time his lington using simplified and advanced methods. CPT
tories at four depths are also shown. The results indi data obtained in the fills composed of 60–80% gravels
cate that liquefaction occurs first in a shallower layer and 20–40% sand-silt fractions indicate the dominant
(6.5-8 m depth) at approximately 40 seconds after influence of the finer sand-silt matrix in the fill with
898
Figure 7. Results from the 1D effective stress analysis performed for CPTu042 showing (a) clean-sand equivalent cone tip
resistance (qc1Ncs) profile, (b) soil behavior type index (Ic) profile, (c) maximum shear strain (γmax) profile, (d) excess pore-
water pressure ratio (ru) profiles at different computational times, and (e) acceleration time histories at four depths of interest.
relatively low tip resistances (qc = 6–8 MPa) and soil Cubrinovski, M., Bray, J.D., de la Torre, C., Olsen, M.,
behavior type values (Ic = 1.9–2.3) characteristic of Bradley, B.A., Chiaro, G., Stocks, E., Wotherspoon, L.
sand-silt mixtures. The CPT-based simplified assessment & Krall, T. (2018). Liquefaction-Induced Damage and
result in relatively low cyclic resistances that are CPT Characterization of the Reclamation at CentrePort
below the demand of the 2016 Kaikōura earth Wellington, Bulletin of the Seismological Society of
quake, which is generally consistent with observa America, 108(3).
tions. Advanced seismic effective stress analysis is Cubrinovski, M., Ntritsos, N., Dhakal, R. & Rhodes, A.
also conducted through careful modeling of 1D soil (2019). Key aspects in the engineering assessment of
layers using a sophisticated constitutive model cali soil liquefaction, Proceedings of Seventh International
Conference on Earthquake Geotechnical Engineering,
brated to CPT-based semi-empirical liquefaction Rome, Italy, June 17- 20.
resistance curves. Unlike the simplified analysis Dhakal, R., Cubrinovski, M., Bray, J.D. & de la Torre, C.
which only provides a pseudo-static factor of safety, (2020a). Liquefaction Assessment of Re-claimed Land
the effective stress analysis provides further details at CentrePort, Wellington, Bulletin of the New Zealand
beyond the capability of simplified methods, such Society for Earthquake Engineering, 53(1), 1–12.
as the timing and onset of liquefaction, and dynamic Dhakal, R., Cubrinovski, M. & Bray, J.D. (2020b). Geo-
interaction between different layers. technical Characterization and Liquefaction Evaluation
of Gravelly Reclamations and Hydraulic Fills (Port of
Wellington, New Zealand), Soils and Foundations, 60
REFERENCES (6), 1507–1531.
Dhakal, R., Cubrinovski, M. & Bray, J. (2022). Input
Boulanger, R.W. & Idriss, I.M. (2014). CPT and SPT Ground Motion Selection for Site Response Analysis at
Based Liquefaction Triggering Procedures, Report No. the Port of Wellington (New Zealand), Proceedings of
UCD/CGM-14/01, University of California, Davis. Fourth International Conference on Performance-Based
Cubrinovski, M. & Ishihara, K. (1998a). Modelling of sand Design in Earthquake Geotechnical Engineering, Bei
behaviour based on state concept, Soils and Founda jing, China, July 15- 17.
tions, 38(3), 115–27. Ntritsos, N. & Cubrinovski, M. (2020). A CPT-based effect
Cubrinovski, M. & Ishihara, K. (1998b). State concept and ive stress analysis procedure for liquefaction assessment,
modified elastoplasticity for sand modelling, Soils and Soil Dynamics and Earthquake Engineering, 131, 106063.
Foundations, 38(4), 213–25. Ntritsos, N., Cubrinovski, M. & Bradley, B.A. (2021).
Cubrinovski, M., Bray, J.D., de la Torre, C., Olsen, M., Challenges in the definition of input motions for foren
Bradley, B.A., Chiaro, G., Stocks, E. & Wotherspoon, L. sic ground-response analysis in the near-source region,
(2017). Liquefaction Effects and Associated Damages Earthquake Spectra, 87552930211001376.
Observed at the Wellington CentrePort from the 2016 Robertson, P.K. & Wride, C.E. (1998). Evaluating cyclic
Kaikoura Earthquake, Bulletin of the New Zealand Soci liquefaction potential using the cone penetration test,
ety for Earthquake Engineering, 50(2), 152–173. Canadian Geotechnical Journal, 35, 442–459.
899
© 2022 the Author(s), ISBN 978-1-032-31259-0
Concept design of a new CPT module for direct in situ measurement

of p-y soil responses
A. Diambra, J. Creasey, J. Leonet, A. Conn, E. Ibraim & G. Mylonakis
University of Bristol, UK
D.J. White, B. Cerfontaine & S.M. Gourvenec

D. Igoe
Trinity College Dublin, Ireland
ABSTRACT: Due to the rapid expansion of the offshore energy market, driven by the installation of wind
turbines founded on single- or multi-piled foundations, the design and optimisation of laterally loaded piles has
attracted enormous interest in the last decade. Current industrial design practice, for offshore and onshore piles,
typically employs lateral load-displacement springs (p-y) to model the soil response under serviceability, ultim
ate or fatigue limit state scenarios of this foundation type, supplemented by moment and axial springs. Correl
ations based on CPT data are commonly used to determine soil properties and, in turn, the stiffness and
resistance of the p-y reaction curves, in advance of detailed laboratory testing that follows later in the project
schedule. To extend the potential for in situ testing to support lateral pile design, this paper presents the novel
idea of complementing a CPT device with a new module capable of probing the soil in such manner that the
monotonic and cyclic p-y soil response (including its evolution during the foundation design life) can be directly
measured in the field. The stress and strain fields induced by such a module resemble those of a miniature lat
erally loaded pile element and differ from circular or flat cavity expansions induced by pressuremeters or dilat
ometers. The new device will be developed in a collaborative research project ‘ROBOCONE’ financed by the
UK and Irish research councils. The background thinking and initial conceptual design of the device, including
the review and selection of appropriate motion mechanisms and instrumentation, is presented in this manuscript.
1 INTRODUCTION It is customary geotechnical practice to relate the

parameters of lateral resistance-displacement soil
Modern society demands more efficient infrastructure springs (p-y curves), used in the design of laterally
under stricter and more challenging operational and loaded piles, directly to CPT data or to soil strength
climatic conditions, such as offshore energy structures and stiffness parameters inferred from CPT data
in harsh environments and urban developments in (Truong and Lehane, 2014). These procedures are
overly congested areas. The loading conditions are almost invariably based on empirically derived
complex and time variant, and the mechanical proper- correlations.
ties of the ground can vary spatially and temporally. In The extension of the kinematic range of in situ geo
this setting, a ‘whole-life’ design approach (Gourve- technical tests, via modern robotics, could permit the
nec, 2020; White et al. 2022), acknowledging the evo- application of kinematic mechanisms and strain histor
lution of these actions and responses is important. ies to the ground that mimic more closely the stress
Developing new technologies and methods to charac- paths expected from the geotechnical infrastructure,
terise the whole-life engineering properties of the such as a pile. This could allow the direct measurement
ground is essential to meet the required safety, resili- of the soil ‘p-y’ response and its whole design life evo
ence and efficiency of geotechnical design and civil lution, if appropriate stress and/or strain histories are
infrastructure. imposed.
The CPT has been the primary platform and This paper presents the initial feasibility study
manipulator for 50 years for in situ geotechnical char- and background thinking for the future development
acterisation of ground properties. CPT data have also of a new additional ‘p-y’ modular section which can
underpinned the design of many geotechnical struc- complement the standard CPT equipment. The work
tures, including onshore and offshore pile foundations. is part of a new collaborative research project
DOI: 10.1201/9781003308829-134
900
involving the University of Bristol, University of sediments. The present project aims at transferring
Southampton and Trinity College Dublin, financially some of the RIGSS capabilities to deeper soil layers.
supported by the UK-Ireland research councils.
3 OVERVIEW OF THE PROPOSED

2 CONTEXT AND NOVELTY OF THE P-Y MODULE
PROPOSED DEVELOPMENT
3.1 Main principles
Since its early development in the 1950s at the Dutch
Laboratory for Soil Mechanics in Delft, the CPT has A diagram of the working mechanism for the pro
been considerably upgraded in the testing and control posed p-y module is shown in Figure 1. A module
procedure as well through the inclusion of new sen capable of horizontal translation is included behind
sors. The latter, for example, include seismic, electric the standard cone penetrometer. This device
resistivity and acoustic transducers (i.e. Campanella mimics the load and displacement history imposed
et al. 1986; Shinn et al. 1998; Houlsby and Ruck, by a laterally loaded pile element, so that the
1998). module response can be upscaled for pile design or
Alongside sensor development, the kinematic used to determine constitutive soil parameters. The
range of the CPT has also been expanded. The cone development of appropriate theoretical frameworks
pressuremeter combines a standard cone penetrom for the interpretation of the module response will
eter with the capability of a standard pressuremeter, be also a fundamental contribution of the forth
enabling the application of cavity expansion kinemat coming research project. The aim is to be able to
ics on surrounding soil (Houlsby & Withers, 1988). derive full monotonic and static reaction curves
A series of friction sleeves with torsional load and and their evolution after application of particular
axial load sensing capabilities and with varying surface stress histories.
texture have also been included behind a standard cone
penetrometer through recent development at the Geor
gia Institute of Technology (Martinez & Frost, 2018).
The system is coupled with a series of pore pressure
sensors, enabling the measurement of excess pore
water pressure ahead and after each axial and/or tor
sional friction sleeve. The device can monitor both
quasi-static and cyclic behaviour enabling an insight
into cyclic strength degradation as well as liquefaction
potential.
Other attempts have been made to impose cyclic
vertical motions to simulate the cyclic response of Figure 1. Diagram of the ‘p-y’ CPT module and whole life
axially loaded piles or to evaluate the cyclic lique stress histories that could be imposed.
faction potential of soils (Diambra et al. 2014; Hos
seini-Sadrabadi et al. 2019). Vibrational movements
have also been imposed by Stähler et al. (2018) to The difference between the displacement field and
assess the liquefaction resistance of soil deposits. soil reactions induced by the proposed p-y module and
None of the proposed kinematic mechanisms has those produced by pressuremeters or flat dilatometers
considered the possibility to explore the horizontal is shown in Figure 2. The proposed p-y module
(‘p-y’) soil resistance, mimicking the soil loading induces an asymmetric bulb of pressure creating a flow
imposed by a laterally loaded pile. This could be of soil around its cylinder, similarly to a laterally
achieved through an additional module behind the loaded pile.
cone, similar to the torsional frictional sleeves.
Through modern robotics, the p-y module could
impose complex stress histories involving cycles of
loading and unloading, similar to laboratory element
testing. Relative to laboratory testing, this would elim
inate the challenges of sample disturbance and re
capturing stress states as well as the time-consuming
and expensive operations of drilling and sampling as
well as the laboratory testing itself.
The developments within the RIGSS (Remote
Intelligent Geotechnical Seabed Surveys Joint Indus
try Project (White et al, 2017, 2022) demonstrate the Figure 2. Cross-section of displacement field and soil reac
benefits of introducing robotic and smart testing in tions induced by ‘p-y’ module (left), pressuremeter (middle)
the geotechnical characterisation of shallow seabed and flat dilatometer (right).
901
3.2 Application of whole-life stress histories 5 DEFORMATION AND CAPACITY
REQUIREMENTS
Robotic control and a precise actuation system is
necessary in order to enable the application of
complex load histories, mimicking whole-life 5.1 Required capacity
stress histories as shown in Figure 1. Geotechnical The pushing force required to displace and bring the
design recognises that the strength and stiffness of soil to failure can be determined by inversely apply
ground, and therefore the performance, safety and ing the p-y theory to the translating module. The
resilience of our infrastructure, evolves through horizontal force exerted from the soil and, therefore.
time under the exposure to in-service loadings. For the required force to move a typical 44mm diameter
example, offshore facilities on soft marine sedi (i.e. 15cm2 cone) is shown in Figures 3 and 4 for
ments can experience transient falls in strength, both sands and clays for varying depth and
potentially by a factor of 2, but long-term gains by undrained shear strength, respectively. Both sets of
a factor of 4 due to loading episodes and consoli graphs show these forces for differing heights of the
dation (Zhou et al. 2019). Driven piles, for module. The required force for the sand is estimated
example in sand and chalk, can also exhibit similar using the p-y curve proposed by Reese et al. (1974),
levels of short-term degradation and long-term assuming a soil friction angle of 40°. The required
recovery depending on the loading history (Jardine force for the clay is estimated using both the Mat-
and Standing, 2012; Ciavaglia et al. 2017). These lock et al. (1970) p-y curves for soft clay and the
advanced aspects of ground behaviour are progres Reese et al. (1975) p-y curves for stiff clay. The rele
sively entering into engineering practice and so it vant movements are on the order of 0.1 diameters
is timely for in situ testing protocols to also
advance in parallel, to provide the necessary site
characterisation.
4 MAIN DESIGN REQUIREMENTS AND

CONSTRAINTS
The initial background thinking and design of the

module has revealed a number of design require
ments and challenges to be addressed in order to
achieve the successful development of the ‘p-y’
module. Among others, the following challenges
have been recognized in this early stage of develop Figure 3. Required Horizontal Force vs Depth for sand to
ment. The module mechanical arrangement requires invoke lateral displacement of CPT module.
the following characteristics:
• Application of straight-line horizontal motion to
the module. This may include a mechanism that
transforms vertical movement or rotary motion
(from a motor or other drive unit) into straight
line horizontal movement.
• Ability to provide appropriate force capacity to
overcome the soil resistance.
• Ability to offer adequate precision and resolution
to be able to measure the stiffness of the soil over
the required strain range.
• Sufficient robustness.
• Capable of hosting sensors to measure soil
response, including pore water pressures. Figure 4. Required Horizontal Force vs Undrained Shear
• Resilient against soil becoming trapped between Strength for clay to invoke lateral displacement of CPT
the moving parts. module.
Include sensors to measure soil pressure, module
movements (over the very small to large displacement
5.2 Displacement resolution and control
domain) and generated pore water pressure. Some of
the above points are further elaborated in the Stiffness information over the shear strain range 0.01
following. 1% are typically important for design and can be
902
Table 1. Potential mechanical systems to use on the ROBOCONE ‘p-y’ module (based on Hricko, 2014 and ENIM, 2020).
Mechanism Description Strengths (+)/Weaknesses (-)
Peaucellier Lipkin Planar linkage transforming rotary to straight- + Compact

line motion. + Exact straight-line
——————–
- Complex revolute joints
- Fragile mechanism with high stress
Scott Russell Linkage transforming vertical to horizontal + Compact

motion. Consists of three links all equal with + Exact straight-line
a rolling or sliding connection. ——————–
- Weakness at revolute joints
- Rapid wear
Hoekens A four bar linkage converting rotational to + Compact

straight line motion. ——————–
- Approximate straight line
- Weakness due to revolute joints
Watt Three bars bolted together in a chain. The three + Compact

bars pivot around two bolts on the central bar to ——————–
move a central point in a near straight line. - Approximate straight line
Chebyshev Linkage consisting of three bars that converts + Compact

rotational to straight line motion for a central ——————–
point. - Approximate straight line
Roberts Linkage consisting of five bars that converts + Compact

rotational to approximate straight line motion. ——————–
- Approximate straight line
Slider crank A four link mechanism with three revolute + Exact straight line
joints and one sliding joint. Rotation of the + Resilient
crank drives linear motion. ——————–
- High friction
- Bulky due to the sliding pair
(Continued )
903
Table 1. (Cont.)
Mechanism Description Strengths (+)/Weaknesses (-)
Screw nut Very simple mechanism that converts rotational + Exact straight line
to linear motion and torque to a linear force. + Resilient
+ High forces
——————–
- High wear rate
- Needs accurate adjustments.
Rack-pinion A circular gear (pinion) connected to a linear + Exact straight line

gear (pinion) to convert rotational to straight + Produces high forces
line motion. + Accurate movement
——————–
- High friction
- Fragile
Hydraulic pistons A hydraulic cylinder or motor that acts as + Can provide high energy and force
a mechanical actuator used to give + Easily controlled using valves/levers.
a unidirectional force through a unidirectional ———————
stroke. - Requires maintenance to avoid leaks
and fluid contamination.
obtained from tests featuring unload-reload loops. 7 FIRST TRIALS AND ANALYSES OF
This cyclic loading can be applied with the proposed STRAIGHT-LINE MECHANISM
p-y module in addition to monotonic loading to inves
tigate a wide range of soil properties. The resolution 7.1 Scott Russell mechanism
threshold of 0.01% is slightly too high to detect the
small strain elastic stiffness, G0, and if possible The Scott Russell linkage consists of three links, all
a lower resolution should aim to be achieved. Preci equal length that are used to form a right-angle
sion of 1 μm is typical of pressuremeter expansion and change of motion from vertical to horizontal linear
will be targeted in these developments. motion as shown in Figure 5. As Point A moves ver
tically upwards, Point C moves horizontally assum
ing a rolling or sliding connection.
6 REVIEW AND SELECTION OF STRAIGHT
LINE MECHANISMS
Table 1 provides a review of available mechanical

systems to transform a vertical or rotary motion into
straight-line horizontal motion. Table 1 also con
siders the potential use of a hydraulic horizontal
pushing system. A schematic, description and
account of strengths/weaknesses are listed for each
option.
Figure 5. Scott Russell mechanism that can be used in
The main outcome from this preliminary assess ‘p-y’ module.
ment is that many mechanical systems do not pro
vide a perfect straight-line mechanism. As such An initial 3-D printed prototype shown in
these mechanisms have been discarded. Figure 6 has been fabricated for this mechanism
It appears that the Scott Russell mechanism is one type. Pulling up the vertical bar in tension triggers
of the few mechanisms that converts perfect vertical an outward movement of the cone while avoiding
to horizontal motion. Its simplicity makes it a viable buckling of the main vertical bar. Several Scott-
option for the proposed module, although there are Russell mechanisms in series linked to a single pull
concerns about its robustness. A second option could ing bar can be included to increase the system
be the use of a hydraulic system with micro-pistons. robustness and stability, although it appears that this
These two options are analyzed in greater detail in mechanism may be suitable for low strength soils or
the following section.
904
at shallow depths. It should be noted that the force the p-y module. The hydraulic system can provide
conversion efficiency depends on the inclination of high forces using simple and miniature design. The
the bar AC in Figure 5, and it decreases as the orien direction of oil flow, the pressure and flow rate
tation of the bar moves away from 45°. within the circuit can be controlled using valves and
hydraulic systems. This allows for more accurate
control of the moving module than a mechanical
system, although accuracy and precision will finally
depend on the actual piston design.
A preliminary diagram of the p-y module with the
horizontal hydraulic piston actuation is shown in
Figure 7. A series of miniature pistons is placed
within the fixed part of the cone, above and below
the module. When oil flows inside hole 1 within the
top section, pistons force the bar to move the middle
section horizontally, representing the test position.
Then, when oil flows in hole 2, pistons move the bar
and the middle section back to its standard position.
Similar to the initial design of the Scott Russell mech
anism, a required horizontal force of 5 kN is assumed
for an initial sizing of the hydraulic pistons. A maximum
fluid pressure of 12 MPa in the pistons is also assumed.
Using fundamental pressure formula and force equilib
Figure 6. Initial 3-D printed prototype of the ‘p-y’ module rium, the minimum diameter of the pistons can be calcu
with the Scott-Russell mechanism. lated for varying number of pistons to provide the
required horizontal force. Results are provided in
For the first development of a real cone,
Table 2.
a functioning range in sands up to 10 m depth and in
clays with an undrained shear strength up to 75 kPa
can be assumed. According to the preliminary capacity
requirements from Figures 3 and 4 with a module
height of 200 mm, maximum overall horizontal forces
up to 5 kN are expected for these conditions.
Considering the loss of efficiency as the inclination
of the bar AB departs from the 45° (Figure 5),
a vertical pull-out force equal to 10 kN could be
assumed for preliminary sizing of the vertical pulling
rod. This would result in a 6 mm diameter steel rod
assuming a yield tensile strength for steel of 350 MPa.
A hydraulic actuation system can be used to pull the
rod up and down in order to trigger the lateral move
ment of the p-y module. This may be placed within the
fixed parts of the cone, either above or below the
module. A cylindrical system with a diameter of
35 mm and an allowable fluid pressure of 12 MPa
would be sufficient to provide the required 10 kN.
If four Scott Russell mechanisms are considered,
the necessary horizontal force Fc (Figure 5) at each
linkage would be 1.25 kN. The connection of the
Scott Russell mechanism should be designed for
a such force range.
The dimensions derived above may pose chal
lenges considering the standard cone dimensions.
Slight geometrical modification to the cone diameter
and module height may prove beneficial if the Scott
Russell mechanical system is adopted.
7.2 Hydraulic piston system

A miniature horizontal hydraulic drive system has
also been considered where oil flow can be used to
control pistons that trigger the horizontal motion of Figure 7. Diagram of ‘p-y’ module using hydraulic pistons
system.
905
It appears that a configuration with four pistons of Diambra, A., Ciavaglia, F., Harman, A., Dimelow, C.,
12 mm diameter may be appropriate and sufficient. Carey, J. and Nash, D.F., 2014. Performance of cyclic
Two pistons can be placed in the top section and two cone penetration tests in chalk. Géotechnique Letters, 4
in the bottom section of the cone. This configuration (3), pp. 230–237.
ensures that the soil pressure acting on the outer sur Ecole Nationale d’Ingénieurs de Metz (ENIM). (2020).
face is evenly distributed over an appropriate number Transmission de puissance
of pistons. It should be noted that using four pistons Gourvenec, S. (2020). Whole-life geotechnical design:
What is it? What’s it for? So what? And what next?:
of 17 mm diameter will provide a twofold increase in Keynote lecture. Proc. 4th Int.l Symposium on Frontiers
horizontal force, suggesting that this system may be in Offshore Geotechnics. DFI. pp. 206–246
a good candidate for future upscaling to larger forces. Hosseini-Sadrabadi, H., Celeste, F., Chareyre, B., Dano, C.,
Sibille, L. and Riegel, P., 2019 Interpretation of a cyclic
Table 2. Calculated minimum piston diameter, D, for vary Cone Penetration Test (CPT) under saturated conditions:
ing number of pistons, n, to provide required horizontal force. numerical and experimental approaches.
Houlsby, G. T., & Ruck, B. M. (1998). Interpretation of sig
Max pressure (MPa) No. pistons D (mm) F (kN) nals from an acoustic cone penetrometer. Geotech. Site
Characterization, 2, 1075–1080.
12 2 17 5.44 Houlsby, G. T., & Withers, N. J. (1988). Analysis of the cone
12 4 12 5.43 pressuremeter test in clay. Geotechnique, 38(4), 575–587.
12 6 10 5.65 Hricko, J. (2014). Straight-Line Mechanisms as One Build
ing Element of Small Precise Robotic Devices. Applied
12 8 8.5 5.44
Mechanics and Materials, 96–101.
Jardine, R. J., & Standing, J. R. (2012). Field axial cyclic
loading experiments on piles driven in sand. Soils and
8 CONCLUSIONS foundations, 52(4), 723–736.
Marchetti, S. Totani, G. Calabrese, M and Monaco, P.
This paper has presented an initial feasibility study (1991). P-y curves from DMT data for piles driven in
and background thinking for the future development clay. DFI, Piling & Deep Foundations, 263–272.
Martinez, A., & Frost, J. D. (2018). Undrained behavior of
of a novel ‘p-y’ modular section to be attached
sand–structure interfaces subjected to cyclic torsional
behind a standard cone penetrometer. The module shearing. J. Geotech. & Geoenvironmental Eng., 144(9),
will enable in situ probing of the soil similar to 04018063.
a laterally loaded pile element while also imposing Matlock, H. (1970, April). Correlation for design of lat
complex whole-life stress histories and explore how erally loaded piles in soft clay. In Offshore Technology
soil response changes during the design life. Conference. OnePetro.
A number of challenges and design requirements Reese, L.C, Cox, W.R and Koop, F.D. (1974). Analysis of
have been identified in this paper. Available mechan laterally loaded piles in sand. Offshore Technology Con
ical systems to produce the desired horizontal trans ference. OnePetro.
Reese, L.C, Cox, W.R, Koop, F.D. (1975). Field testing and
lation of the module have been reviewed and the use
analysis of laterally loaded piles in stiff clay. 7th Annual Off
of hydraulic actuation with multiple systems appear shore Technology Conference, Houston. 672–690.
a good candidate for future developments. Shinn, J. D., Timian, D. A., Morey, R. M., Mitchell, G.,
The full development of the proposed module and Antle, C. L., & Hull, R. (1998). Development of a CPT
the interpretative framework is the main objective of deployed probe for in situ measurement of volumetric
a collaborative research proposal ‘ROBOCONE’ soil moisture content and electrical resistivity. Field
involving the University of Bristol, University of Analytical Chemistry & Technology, 2(2), 103–109.
Southampton and Trinity College Dublin. Stähler, F. T., Kreiter, S., Goodarzi, M., Al-Sammarraie,
D., & Mörz, T. (2018). Liquefaction resistance by static
and vibratory cone penetration tests. Cone Penetration
Testing 2018 (pp. 591–597).
ACKNOWLEDGMENTS Truong, P., & Lehane, B. (2014). Numerically derived
CPT-based py curves for a soft clay modeled as an elas
The authors would like to acknowledge the financial tic perfectly plastic material. In 3rd Int. Symp. on Cone
support from the Engineering and Physical Sciences Penetration Testing (pp. 975–982).
Research Council (EPSRC - Ref: EP/W006235/1) White, D. J., Doherty, J. P., Guevara, M., & Watson, P. G.
and Science Foundation Ireland (SFI - Ref: 21/ (2022). A cyclic py model for the whole-life response of
EPSRC/3787). piles in soft clay. Computers and Geotechnics, 141, 104519.
White, D. J., Stanier, S. A., Schneider, M. A.,
O’Loughlin, C. D., Chow, S. H., Randolph, M. F., … &
REFERENCES Chow, F. C. (2017). Remote Intelligent Geotechnical
Seabed Surveys–Technology Emerging from the RIGSS
Campanella, R. G., Robertson, P. K., & Gillespie, D. (1986). JIP. OSIG 8th Int. Conf.Proceeding (Vol. 1214,
Seismic cone penetration test. In Use of in situ tests in geo No. 1222, pp. 1214–1222). SUT.
technical engineering (pp. 116–130). ASCE. Zhou, Z., O’Loughlin, C. D., White, D. J., & Stanier, S. A.
Ciavaglia, F., Carey, J., & Diambra, A. (2017). Time- (2020). Improvements in plate anchor capacity due to
dependent uplift capacity of driven piles in low to cyclic and maintained loads combined with
medium density chalk. Géotech. Letters, 7, 90–96. consolidation. Géotechnique, 70(8), 732–749.
906
© 2022 the Author(s), ISBN 978-1-032-31259-0
The use of dynamic probing tests and cone penetration tests to verify the
effectiveness of expanding polyurethane resin injections for ground
improvement
A. Dominijanni
Politecnico di Torino, Italy
M. Gabassi
Uretek Italia Spa, Italy
A. Minardi
Geotechnical Engineering Ltd, Italy
S. Pavan
Uretek Italia Spa, Italy
ABSTRACT: Injection of expanding polyurethane resins is a popular method to improve both the stiffness
and the shear strength of the ground below existing foundations. The effect of the polyurethane resin expan
sion is to increase the soil confining stress and density around the injection holes. An estimation of the hori
zontal stress and volumetric strain changes that are induced within the ground is derived from the theory of
cavity expansion in elasto-plastic materials. A series of case-histories is presented to document the feasibility
of different in-situ tests to evaluate the achieved ground improvement. The tests have been performed before
and after the injection of polyurethane resins and the obtained results have been compared with theoretical
predictions. The considered investigation methods include the dynamic probing tests and the cone penetration
tests. The preliminary results that have been achieved using an experimental miniature cone penetration test
are also illustrated. The advantages and limitations of different test methods are discussed and practical indi
cations for conducting such verifications of polyurethane resin injection effectiveness are provided.
1 INTRODUCTION of treatment is different from conventional compac

tion grouting techniques due to the physical and
The use of polyurethane expanding resins is mechanical processes governing the expansion of the
a widespread technique adopted to solve ground- expanding polyurethane resin in the subsoil. The
related engineering problems. Due to their significant cavity expansion theory represents a relatively
swelling capacity, polyurethane resins are currently simple and reliable approach for modelling the
deployed successfully in the following cases (Domini changes in the stress state and the density of the soil
janni and Manassero, 2014): induced by the expansion of the resin (Yu, 2000;
ASCE, 2010; Dominijanni and Manassero, 2014).
- filling and stabilization of underground cavity;
Despite the possibility of using well-established
- reduction of soils hydraulic conductivity; design methodologies, the assessment of the success
- heaving of pavement and foundations in settle of the ground treatment with expanding polyurethane
ments problems; resins still represents a major challenge. When the
- ground improvement and compaction. ground treatment is performed before the construc
Ground improvement and compaction with tion of structures and buildings, it is possible to
expanding polyurethane resins aim to increase soil carry standard in-situ tests, such as cone penetration
density and soil mechanical parameters such as stiff tests (CPT), to evaluate the achieved ground
ness and strength. This type of treatment, often per improvement. However, when the treatment is per
formed below existing foundations, can be formed below existing structures, the use of standard
considered part of the treatment methods related to in-situ testing equipment might not be feasible due
compaction grouting techniques. However, this type to the space required for this type of equipment.
DOI: 10.1201/9781003308829-135
907
A solution to this issue is provided by the per
formance of reduced scale in situ testing, such as
mini-CPT and small dynamic probing tests. This
paper illustrates some examples of using these
types of testing equipment, besides the adoption of
CPT testing, to evaluate the effectiveness of the
ground improvement with expanding polyurethane
resin.
2 GROUND TREATMENT WITH

POLYURETHANE RESINS
Polyurethane resins are obtained from the exo

thermic reaction occurring by the mixing of
a polyol and an isocyanate. The swelling cap
acity is the key feature of this material, and it is
responsible for a volume increase under uncon
fined conditions (null mechanical stress) of
about thirty times (Dei Svaldi et al., 2005;
Buzzi et al., 2008). When the resins are injected
underground, the expansion process occurs in Figure 1. Schematic layout of the two types of tubes usu
confined conditions leading to a compaction of ally adopted for the injection of polyurethane resins, along
the soil surrounding the injection point until with the expansion process of the injected resin.
a mechanical equilibrium between the resin and
the soil is achieved. The injection can be per Instead, under undrained conditions, the soil is not
formed at different depths using either several allowed to compress, and the mean effective stress
injection tubes with single-point injection at the decreases due to the formation of excess pore pres
bottom end of the tube or a single tube with sures. During the subsequent consolidation phase, the
several lateral injection points (multi-point). In soil follows an unload/reload compression curve until
both cases, the injections are performed with the original mean effective stress is reached. However,
small tubes (external diameter of about 10 because of the higher rigidity of the unload/reload
15 mm) and equipment, reducing the treatment’s compression curve with respect to the curve for first
impact on the existing structure as much as loading, the amount of compression is lower than that
possible. achieved under drained conditions.
From a design point of view, the cavity expan Therefore, the use of polyurethane has to be firstly
sion theory represents an efficient tool that can considered either in coarse-grained soil or in par
be used by engineers (Yu and Houlsby, 1991; tially saturated fine-grained soil.
Shrivastava et al., 2018). In this framework, The evaluation of the soil geotechnical parameters
spherical cavities expansion can be considered is often performed with in-situ tests, such as
when single point injection tubes are adopted for dynamic and static penetration tests, using empirical
the treatment. On the other hand, when multi- correlations available in the scientific literature.
point injection tubes are used, it is possible to Moreover, penetration tests may also be adopted to
consider a cylindrical cavity expansion evaluate the efficiency of the treatment by compar
(Figure 1). This design approach allows engineers ing the results of the tests performed before and after
to estimate the amount of a specific type of resin the treatment in the soil surrounding the injection
to be injected to obtain a given increase of the points. The role of in-situ testing is, therefore, funda
soil density and mechanical parameters. mental not only for the design of the ground treat
The efficiency of the treatment depends on several ment but also for the assessment of the treatment
factors, such as the type of soil, the presence of efficiency.
groundwater, the soil stiffness and strength, the
injection layout, the amount and type of injected
resin. When compaction is the aim of the treatment, 3 IN-SITU PENETRATION TESTS
the efficiency of the treatment is higher under
drained conditions (Kovacevic et al., 2000). Beside the standard CPT test, the use of smaller
In fact, under drained conditions, the mean effect equipment is often required to evaluate the effective
ive stress increases monotonically, and the soil com ness of the ground treatment with polyurethane
presses along a compression curve for first loading, resins. In the following two subsections, a brief over
with a consequent increase in density, stiffness and view of the miniature cone penetrometer (mini-CPT)
shear strength. and the medium-light dynamic probing penetrometer
908
(DPM 30-20) are introduced. These types of pene- 4 CASE STUDIES
trometers have the advantage of being easy to trans
port, install, and use. In this section a series of case studies are presented,
where the assessment of the effectiveness of the
ground treatment with polyurethane resins is evalu
3.1 Mini CPT
ated by performing standard CPT, mini-CPT and
The miniature cone penetrometer is a reduced scale DPM 30-20 in situ tests. The examples refer to
cone with a diameter of 18 mm and an apex angle ground treatment works related to resin injections
equal to 60°. Examples of the use of the mini cone below shallow foundations of residential buildings.
penetrometer can be found in Squeglia and Lo Presti
(2010), where this equipment has been adopted to
4.1 Case a)
develop an innovative methodology for the evalu
ation of the compaction degree of earth works, and The first case study refers to the ground treatment per
in de Vries et al. (2018) where a comparison with formed below the plate foundation of a residential
standard CPT tests is illustrated. The cone is pushed building that experienced differential settlements
downward in the soil at a constant rate equal to during construction. During the design phase, an in-
2 cm/s by an electric motor. The tip resistance (qc) is situ testing investigation composed of four CPT tests
evaluated by a load cell placed above the penetrom was carried out. The map of the building is presented
eter with maximum load of 5 kN and accuracy of in Figure 2, along with the location of the four CPT
5 N. tests. Figure 3 illustrates the tip resistance (qc) profile
The performance of mini-CPT tests is not obtained from the tests CPT-1 and CPT-2. From this
common for soil geotechnical characterization. outcome, the foundation soil can be roughly divided
Therefore, this test method suffers of a lack of in a shallow layer of granular soil (mainly sand) with
empirical correlations to estimate geotechnical thickness equal to 3 m overlaying a deeper layer of
parameters, as well as a direct comparison with the fine-grained soil with a thickness equal to 7 m.
standard CPT test. To face this issue, mini-CPT
tests are often carried out along with standard CPT
or, in case this is not possible, with dynamic prob
ing tests.
3.2 Dynamic penetrometer DMP 30-20

Dynamic probing tests (DP) consist of driving into
the soil a steel cone located at the end of a set of
driving rods by blowing the upper end of the rods
with a specific hummer (Cestari, 2012). DP tests can
be classified in different category (light, medium,
heavy, super-heavy), depending on the hammer
weight, falling height, cone size and geometry, rods
size, etc.
In the framework of ground treatment with poly
urethane resins, the use of a medium-light penetrom
eter, the DPM 30-20, is particularly considered in
Italy. This equipment is based on the measurement
of the blow number required for a penetration of the
cone equal to 10 cm (N10). Although this equipment
is not included in any standard, its main features are
similar to the standard light and medium dynamic Figure 2. Map of the building along with the position of
penetrometer. The mass of hammer is 30 kg and fall the four CPT tests (n.1 to n.4) performed before construc
ing height is 20 cm. The cone has a base diameter tion, the treated are with resin injections, and the four CPT
tests (n.5 to n.8) performed after ground treatment.
equal to 35.7 mm and an apex angle equal to 60°
(Cestari, 2012).
Although the possibility to correlated the DPM As illustrated in Figure 2, the ground treatment
30-20 test results (N10) to the standard penetration was performed on the north side of the building,
test (NSPT), the main disadvantage of this type of test which was subjected to major settlements. The injec
is represented by the absence of the measurement of tion points were placed on a regular squared grid
the skin friction during the penetration of the cone. with spacing equal to 1 m, and they were performed
Therefore, the measured penetration resistance might first in correspondence of the building perimeter and
be overestimated due to the missing contribution of then below the building. Moreover, the resin was
the skin friction. injected first at a depth between -5 and -7 m with
909
Figure 3. Comparison of tip resistance (qc) profile between the CPT tests performed before (CPT-1) and after (CPT-5 and
CPT-6) the ground treatment with polyurethane resins.
multipoint injection tubes. A second set of injection combination of multi-point tubes and single point tube
was then carried out at a shallower depth between -2 is adopted to treat a thickness of the ground equal to
and -5 m with single-point injection tubes every 1 m. 3 m (from -1.3 m to -4.3 m). The injection points
A second series of in-situ tests, composed of four along the perimeter are spaced of 1 m.
standard CPT, was performed after the ground treat Figure 4 shows the map of the building along
ment to evaluate the achieved improvement. The with the location of the in-situ tests carried out to
position of these tests is illustrated in the Figure 2 assess the efficiency of the treatment. Due to the
where the tests are numbered from 5 to 8. In particu presence of the existing structures, mini-CPT tests
lar, the CPT-5 and CPT-6 are carried out close the were adopted and performed in three different points.
treated area, while the CPT-7 and CPT-8 are further In each point, two tests were carried out one before
and, therefore, less influenced by the injections. and one after ground treatment. The two tests of each
Figure 3 shows a comparison between the tip resist point were located close to each other to minimize
ance profiles obtained from the test performed before the possible influence of different ground conditions.
the treatment (CPT-1 is taken as a reference) and the
tests performed after treatment (CPT-5 and CPT-6). In
the shallow soil layer below the foundation (down to
-3 m depth), a pronounced increase of the tip resistance
is highlighted by both CPT tests (CPT-5 and CPT-6).
The average value of qc referred to this layer is higher
than 10 MPa. On the other hand, in the in the depth
range between -3 m and -7 m, the increase of the tip
resistance exhibited by the two CPT tests is less pro
nounced. On average, the qc value obtained from the
two tests carried out after treatment is more than twice
the average value measured before treatment.
Figure 4. Map of the building illustrating the location of

4.2 Case b) the mini-CPT tests performed evaluate the efficiency of the
The second case study refers to the work carried out ground treatment.
for the ground consolidation below an office building
with dimensions equal to 15x7 m. The injections are Figure 5 illustrates the tip resistance profiles of
performed along the building perimeter underneath the the three mini-CPT tests performed before ground
foundation starting from of a depth equal to 1.2 m. A treatment. From these findings, the soil can be
910
mainly classified as sand-silt mixture. In particular, mini-CPT tests performed before and after treat-
the test S5 exhibited the lower tip resistance, high- ment. In Figure 6 is illustrated an example obtained
lighting the presence of silt/clayey silt below a depth from the mini-CPT S5 and S6. The graph clearly
equal to -3 m. highlights the increase of tip resistance measured at
depths corresponding to the resin injection.
In the range between -1.3 m and -2.8 m the tip
resistance increased from 1.6 MPa to 4.0 MPa, while
in the range between -2.8 m and -4.3 m the tip resist
ance increased from 0.6 MPa to 2.4 MPa.
4.3 Case c)
The last case study illustrates the treatment with poly
urethane resins of the ground foundation below an
existing residential building. The treatment involved
just part of the building as illustrate in the Figure 7.
The injections were performed below the building
foundation located at -1.2 m depth and along the per
imeter of the building, with a spacing equal to 1 m,
using multi-point tubes. The thickness of the soil layer
treated is 2 m, from -1.2 m (depth of the foundation) to
-3.2 m. The ground below the foundation is mainly
composed of a mixture of loose fine soil and gravel.
In this example, the efficiency of the ground treat
ment was evaluated with dynamic probing tests, adopt
ing the medium-light penetrometer DPM 30-20. As
illustrated in the Figure 7, two DPM 30-20 tests were
carried out: P1 before the treatment and P2 after the
Figure 5. Tip resistance (qc) profile of the three mini-CPT
treatment.
tests (S1, S3, S5) performed before the ground treatment.
Figure 7. Map of the building along with the location of

the treatment zone and the location of the DPM 30-20 tests
performed before (P1) and after (P2) the ground treatment.
Figure 8 shows the results of both the two DPM 30

20 tests. From the results of the test P1, two soil layers
were identified: (i) from -1.2 m to -2.3 m with an aver
age N10 value equal to 7 and (ii) from -2.3 m to
-3.3 m with an average N10 value equal to 2.4. After
the treatment, the performed DPM 30-20 test high
lighted an increase of the average N10 value for both
layers. In particular, the layer 1 exhibited an average
value equal to 27, while the layer 2 exhibited an aver
age N10 value equal to 21.
A design tool based on the cavity expansion theory
in a finite medium (Uretek, 2021) was used to assess
Figure 6. Comparison of the tip-resistance profiles the achieved ground improvement in terms of increase
obtained from the mini-CPT tests S5 and S6 performed in relative density and to verify the amount of resin
before and after ground treatment. injected. A perfect elasto-plastic Mohr-Coulomb model
with a non-associated flow rule was adopted for the
The evaluation of the ground treatment is carried soil, considering two sets of parameters for the two dif
comparing the tip resistance profiles from two ferent layers (Table 1). As the spacing of the injection
911
treatment can be assessed by performing in-situ tests
with static and dynamic penetrometers. This paper pre
sents a series of case studies where standard CPT,
mini-CPT, and DPM 30-20 tests were adopted.
The main advantage of the standard CPT test is
the possibility of using the test results not only for
treatment performance evaluation but also for the
geotechnical characterization of the subsoil. How
ever, it is often difficult to adopt such equipment
when the treatment is carried out inside an existing
structure.
The reduced size of the mini-CPT and DPM 30
20 equipment represents their main advantage. On
the other hand, the lack of well-established empirical
correlations to estimate soil geotechnical parameters
from these tests is their main drawback.
Based on the examples presented in this work, all
three in-situ tests can be regarded as suitable tools to
assess the efficiency of the ground treatment with
expanding polyurethane resins.
Figure 8. Comparison of the test results in terms of N10

REFERENCES
obtained from the tests performed before (P1) and after ASCE (American Society of Civil Engineers), 2010. Com
(P2) the ground treatment. paction ground consensus guide, ASCE Standard
ASCE/G-I 53-10. ASCE.
holes was equal to 1 m, the radial distance of the fixed Buzzi, O., Fityus, S., Sasaki, Y., e Sloan, S., 2008. Struc
(no-displacement) boundary of the medium was set ture and properties of expanding polyurethane foam in
equal to 0.5 m, i.e. half of the injection hole spacing. the context of foundation remediation in expansive soil.
Mechanics of Materials 40,1012–1021.
The average increase in relative density obtained from
Cestari, F., 2012. In situ geotechnical tests. Pàtron
the theoretical analyses was 12% for layer 1 (from editore.
31% to 43%) and 21% for layer 2 (from 17% to 38%). de Vries, G. T., Laban, C., & Bliekendaal, E., 2018. Com
The total amount of resin required to obtain such parison of mini CPT cone (2 cm2) vs. normal CPT cone
improvement resulted in being equal to 18.3 kg, which (10 cm2 or 15 cm2) data, 2 case studies. In Cone Pene
is in good agreement with the amount of resin injected tration Testing 2018 (pp. 683–688). CRC Press.
in-situ (16.9 kg). This example highlights the potential Dei Svaldi, A., Favaretti, M., Pasquetto, A., e Vinco, G.,
of the cavity expansion theory to simulate ground 2005. “Modellazione analitica del miglioramento del
behavior when treated with expanding polyurethane terreno attraverso iniezioni di resina ad alta pressione
d’espansione.” 6th International Conference on Ground
resins.
Improvement Techniques, Coimbra, Portugal.
Dominijanni, A., Manassero, M., 2014. Consolidamento
Table 1. Parameters of the soil adopted for the dei terreni con resine espandenti: guida alla progetta
calculations. zione. McGraw-Hill.
Kovacevic, N., Potts, D. M., & Vaughan, P. R., 2000. The
Layer 1 2 effect of the development of undrained pore pressure on
the efficiency of compaction grouting. Geotechnique, 50
Unit weight [kN/m3] 18 18 (6),683–688.
Young modulus [MPa] 10 6 Shrivastava, N., Zen, K., Shukla, S. K., 2017. Modeling of
Poisson ratio [-] 0.25 0.25 compaction grouting technique with development of
Shear strength angle [°] 34 34 cylindrical cavity expansion problem in a finite medium.
International Journal of Geosynthetics and Ground
Cohesion [kPa] 0 0
Engineering, 3(4),1–12.
Dilatancy angle [°] -5 -8 Squeglia, N., & Lo Presti, D. C. F., 2010. Use of mini CPT
Max void ratio [-] 0.65 0.65 to evaluate degree of compaction in fine-grained soils.
Min void ratio [-] 0.30 0.30 In CPT’10. 2nd International Symposium on Cone
Initial void ratio [-] 0.541 0.589 Penetration Testing, Huntington Beach, California.
Uretek, 2021. SIMS 2.0 (Soil Improvement Modelling
Software) – Training manual.
5 CONCLUSIONS Yu, H. S., Houlsby, G. T., 1991. Finite cavity expansion
in dilatant soils: loading analysis. Geotechnique,
Expanding polyurethane resins represent a valid alter 41(2),173–183.
native to conventional grout for ground improvement Yu, H. S., 2000. Cavity expansion methods in geomecha
and compaction. The effectiveness of the ground nics. Springer Science & Business Media.
912
© 2022 the Author(s), ISBN 978-1-032-31259-0
Evaluation of flow liquefaction susceptibility of a sandy-silt tailings

using the CPTu
M.P. dos Santos Junior & R. César Gomes
Universidade Federal de Ouro Preto, Brazil
S.G. Silva Ribeiro

GeoFast – Geotechnical Modelling, Brazil
B.G. Delgado
CONSTRUCT-GEO, Faculdade de Engenharia da Universidade do Porto, Portugal
ABSTRACT: Flow liquefaction is a behavior observed in saturated or nearly saturated geomaterials that
show a strain softening response during undrained shear, most common in very loose sands and silts as well
as very sensitive clays. In the mining industry, flow liquefaction is a subject of high relevance due to the geo
technical characteristics of the tailings. Many methodologies have been developed to evaluate the susceptibil
ity to flow liquefaction using in-situ tests, especially the CPTu test, since it provides high accuracy and good
repeatability. This paper compares four methodologies based on the CPTu test to evaluate the susceptibility to
flow liquefaction: i) Plewes et al. (1992), ii) Olson (2001), iii) Shuttle & Cunning (2008) and iv) Robertson
(2016). The results obtained highlight the need to correct the original contour suggested by Olson (2001) for
medium and high compressibility geomaterials. Furthermore, all other methods evaluated presented similar
results in the overall classification.
1 INTRODUCTION
Casagrande (1975) showed that loose sands tend to

contract during drained shear while dense sands tend
to dilate under the same condition until it reaches the
critical void ratio. If sheared under an undrained
condition loose sands will generate positive excess
porewater pressure, reducing the effective stress and,
therefore, its shear strength, while the opposite
would occur for dense sands.
Flow liquefaction is a behavior observed in satur
ated on nearly saturated soils that show a strain soften
ing response in undrained shear due to its contractive
behavior under drained conditions, most common in
very loose sands and silts as well as sensitive clays.
As shown in Figure 1 presented by Robertson &
Wride (1998), sand-like soils that are looser than the
critical state will generate positive excess porewater
pressure and show a strain softening behavior in
undrained shear, illustrated by the Strain Softening
Condition (SS). If the state prior to shear is close to the
critical state, these soils can experience a limited con
traction, exhibited as a Limited Strain Softening (LSS)
condition in Figure 1, also referred in the literature as
quasi-steady state condition. If the soil is denser than Figure 1. Typical undrained monotonic behavior of sand-
the critical state, however, it will tend to exhibit like soils in triaxial compression tests - Robertson & Wride
(1998).
DOI: 10.1201/9781003308829-136
913
a Strain Hardening (SH) behavior and the undrained and Equation 7, as a function of the slope of the crit
strength will be greater than the drained strength. ical state line.
2 SUSCEPTBILITY TO FLOW LIQUEFACTION
This paper is going to evaluate the susceptibility of

a sandy-silt tailings to flow liquefaction using the
results of a Cone Penetration Testing (CPTu) per
formed in the interior of a Tailings Storage Facility
(TSF). Four methodologies will be used herein to where Μ is the critical friction ratio (Μ=qc/p’c) and
evaluate the state of the tailings: i) Plewes et al. λ10 is the slope of the critical state line (CSL) meas
(1992), Olson (2001), Shuttle & Cunning (2008) and ured in log10 p’ - e space. It was assumed Μtc=1,45
Robertson (2016). herein, using the average value of the range sug
gested by Jefferies & Been (2016).
2.1 Plewes et al. (1992)

2.2 Olson (2001)
Plewes et al. (1992) suggested a relationship
between the slope of the critical state line (λ10) and Olson (2001) performed an extensive evaluation of 33
the normalized friction ratio (F or Fr) as shown by case histories of liquefaction flow failures and devel
Equation 1. oped a comprehensive procedure to evaluate i) lique
faction susceptibility, ii) triggering of liquefaction and
iii) post-triggering/flow failure stability.
In order to evaluate the susceptibility to flow
liquefaction, Olson (2001) suggested the use of the
contour proposed by Fear & Robertson (1995) based
on the Standard Penetration Test (SPT). To convert
the boundary from SPT to CPT, Olson (2001) used
a relationship qc/N60=0.60, typical of clean sands, as
where: shown in Figure 2. The contour proposed by the
author to distinguish between contractive and
fs = sleeve friction resistance
qt = corrected cone resistance
σv0 = total vertical stress
Once the slope of the critical state line (λ10) is

determined the state parameter (ψ) can be calculated
using the equation suggested by Shuttle & Cunning
(2007).
Where Qp is the tip resistance normalized by the

mean affective stress (p’0) and Bq is the pore pres
sure ratio, defined as shown below.
where:
u2= pore pressure measured behind the cone
u0 = in situ pore pressure
As a screening-level assessment, Jefferies & Been Figure 2. Boundary separating contractive from dilative
(2016) suggested that the effective inversion coeffi behavior suggested by Olson (2001) to evaluate the suscep
cients, kB and m,
B could be determined using Equation 6 tibility to flow liquefaction.
914
dilative behavior is represented by Equation 8, based associated with the ratio of qc/N60=0.6. Many investi
on the corrected tip resistance (qc1) and the vertical gators (Burland & Burbidge 1985; Jefferies & Davies
effective stress (σ’v0). 1993; Robertson & Campanella 1983; Schnaid &
Odebrecht 2012; Stark & Olson 1995) demonstrated
that the ratio qc/N60 is a function of soil type (usually
expressed as a function of the mean grain size – D50).
This paper suggests the use of the ratio qc/N60 to
account for the effect of compressibility and Equa
The corrected CPT tip resistance, qc1, is obtained as tion 11 is suggested for this purpose.
follows:
where qc is the measured cone resistance and Cq is

the CPT-based overburden correction factor sug- where δ = Ratio qc/N60.
gested by Kayen et al. (1992), defined by To evaluate the susceptibility to flow liquefaction
Equation 10. it is suggested the use of Equation 12.
Therefore, β>0 will indicate that (qc1)contour

>(qc1)measured for the same vertical effective stress,
where Pa is one atmosphere of pressure in the same which indicates that the point in the profile is in
units as σ’v0. a contractive state.
Olson (2009) discussed the effect of compressibil
ity on the evaluation of liquefaction susceptibility and
suggested the use of the slope of the critical state line 2.3 Shuttle & Cunning (2008)
(λ10) to update the original boundary suggested by Shuttle & Cunning (2007) presented a very detailed
Olson (2001) as shown by Figure 3. According to the study using finite element analysis with cavity expan
author, the original boundary should only be used for sion theory and the NorSand Model to evaluate the
low compressibility soils (λ10~0.03) which is liquefaction potential of a very loose silt tailings (Rose
Creek silt tailings). Following discussions with Peter
K. Robertson, Shuttle & Cunning (2008) presented
a contour to distinguish between contractive and dila
tive behavior using the soil behavior chart suggested
by Jefferies & Davies (1991), as presented in Figure 4.
Since the authors did not present the equation cor
responding to the suggested contour, it is suggested
herein the use of Equation 13 to be used to represent
this boundary.
where,
To characterize the behavior of the soil with depth the

parameter α will be used, defined by Equation 15.
Figure 3. Contours suggested by Olson (2009) to separate Following this definition, positive values of α will
contractive from dilative conditions for soils of low, indicate that [Q(1-Bq)+1]Contour>[Q(1-Bq)+1]Measured,
medium and high compressibility. meaning that the measured value is located below the
915
where,
where n≤1.0.
The soil behavior type index, Ic, was first pro
posed by Jefferies & Davies (1993) who recognized
that the boundaries between the soil behavior type
zones could be approximated by concentric circles
whose radius indicates the soil behavior type index.
Robertson & Wride (1998) modified the definition of
Figure 4. Boundary separating contractive from dilative
behavior suggested by Shuttle & Cunning (2008) - Jefferies
Ic to apply to the Robertson (1990) chart, as defined
& Been (2016). by Equation 19.
envelope shown in Figure 4 and, therefore, is suscep

tible to flow liquefaction.
Robertson (2016) suggested the use of the contour
CD=70 (also shown in Figure 5) to differentiate soils
that are contractive and dilative at large strains. The
CD=70 boundary combines two different criteria: i)
Qtn,cs=70 for sand-like soils and OCR=4 for transi
tional and clay-like soils. Equation 20 is suggested
by the author to represent the contour CD=70.
2.4 Robertson (2016)
Robertson (2016) updated the CPT-based normalized
soil behavior type (SBTn) classification system pro
posed by Robertson (2009) to use behavior-based
instead of textural-based descriptions, as presented
in Figure 5.
3 RESULTS
The tailings evaluated herein is a by-product of the

mining of iron ore in the state of Minas Gerais/
Brazil. The assessment of the tailings included the
collection of disturbed samples to evaluate: i) the
grain size distribution curve, ii) the water content
(w), iii) the specific gravity of soil solids (Gs), iv)
the unit weight (γ and γd), v) the liquid limit (LL)
and vi) the plastic limit (PL).
The grain size distribution curves of the samples
collected are shown in Figure 6. In terms of mean
values, roughly 79% of the tailings corresponds to
silt-sized particles, 4% to clay-sized particles and the
remaining (17%) is fine sand. The fines were non-
Figure 5. Updated CPT-based normalized soil behavior plastic and the mean values of the unit weight was
type chart proposed by Robertson (2016). γ=23.4 kN/m³ and γd=19.4 kN/m³.
The collected samples were also used to deter
The author updated the soil behavior type index to mine the water content (w) and these values were
use a hyperbolic shape, as indicated by Equation 16. converted to void ratio (e) using Equation 21. The
916
was divided into three main regions: i) Region I:
Predominantly dilative behavior (0-16m), ii) Region
II: Interbedded layers of contractive and dilative
behavior (16-27m) and iii) Region III: Predomin
antly contractive behavior (below 27m).
The results of the Fear & Robertson (1995) bound
ary suggested by Olson (2001) is shown in Figure 9.
As can be observed, using this method most of the
tailings profile is classified as contractive. In fact, only
Figure 6. Grain size distribution curves of the sandy-silt

tailings according to the ASTM D422-63 (2007).
mean value of water content was 21.4% and the void
ratio was 0.93.
The specific gravity (Gs) was determined following

the procedures of the ASTM D854-14, to give
a mean value of 3.8.
The normalized parameters (Qt, Fr and Bq) from
the CPTu performed on the sandy-silt tailings is Figure 8. Evaluation of the susceptibility to flow liquefac
shown in Figure 7. As can be seen, the initial portion tion using a) Robertson (2016), b) Shuttle & Cunning
of the sounding did not show excess porewater pres (2008) and c) Plewes et al. (1992).
sure up until 20m. Furthermore, below 27m the tail
ings exhibit a very low tip resistance and high Bq
values. It is also important to notice the high hetero
geneity on the tailings profile due to the interbedded
layers of clay-like and sand-like behavior.
In order to evaluate the susceptibility of the tail
ings to flow liquefaction the contours suggested by
Robertson (2016), Shuttle & Cunning (2008) and
Plewes et al. (1992) are presented in Figure 8. Based
on the behavior observed in the sounding, the profile
Figure 9. Results of the susceptibility to flow liquefaction

Figure 7. Normalized CPTu parameters of the sounding using the Fear & Robertson (1995) contour suggested by
performed in the sandy-silt tailings. Olson (2001).
917
It is also important to emphasize that all these
methods should be used as a screening-level assess
ment. Moreover, there are instances where the
methodologies discussed herein can yield to differ
ent conclusions regarding the state of the soils and
further investigations (including laboratory tests)
would certainly be an effective tool to guide
towards a better understanding of the behavior of
the geomaterials.
REFERENCES
ASTM D422-63(2007)e2. 2007. Standard Test Method for
Particle-Size Analysis of Soils (Withdrawn 2016).
Figure 10. Evaluation of the susceptibility to flow liquefac
ASTM International, West Conshohocken, PA.
tion using a) Olson (2001) corrected for qc/N60=0.07, b)
ASTM D854-14. 2014. Standard Test Methods for Specific
Shuttle & Cunning (2008) and c) Plewes et al. (1992).
Gravity of Soil Solids by Water Pycnometer. ASTM
International, West Conshohocken, PA.
the first 2m of the profile (σ’v0<50kPa) was deemed to Burland, J.B. & Burbidge, M.C. 1985. Settlement of Foun
be in a dilative state. dations on Sand and Gravel. In Proceedings of the Insti
The limitations of the original contour suggested by tute of Civil Engineers (London). pp. 1325–1381.
Olson (2001) for medium and high compressibility Casagrande, A. 1975. Liquefaction and cyclic deformation
soils was discussed herein in the section 2.2. Equa of sands: A critical review. In Proceedings of the Fifth
Panamerican Conference on Soil Mechanics and Foun
tion 12 was applied in this paper to account for the
dation Engineering. pp. 79–133.
effect of compressibility and it was determined the Fear, C.E. & Robertson, P.K. 1995. Estimating the
ratio qc/N60 that would be necessary for the boundary undrained strength of sand: a theoretical framework.
suggested by Olson (2001) to yield results similar to Canadian Geotechnical Journal, 32(5): 859–870.
the other methodologies evaluated. As can be seen in doi:10.1139/t95-082.
Figure 10, it was necessary to use a ratio of qc/N60 Jefferies, M. G. & Been, K. 2016. Soil Liquefaction:
=0.07 (a reduction by a factor of 8.6 in relation to the A Critical State Apporach.
original contour) to calibrate the Fear & Robertson Jefferies, M.G. & Davies, M.P. 1991. Soil classification by
(1995) boundary for use with the sandy-silt tailings. the cone penetration test: Discussion. Canadian Geo-
technical Journal, 28(1): 173–176. doi:10.1139/t91-023.
Jefferies, M.G. & Davies, M.P. 1993. Use of CPTu to Esti
4 CONCLUSION mate Equivalent SPT N60. Geotechnical Testing Jour
nal, 16(4): 458–468. doi:10.1520/gtj10286j.
Kayen, R.E. Mitchell, J.K. Seed, R.B. Lodge, A. Nishio, S.
This paper presented an evaluation of the suscepti & Coutinho, R. 1992. Evaluation of SPT, CPT and shear
bility to flow liquefaction of a sandy-silt tailings wave-based methods for liquefaction potential assess
using four different methodologies: i) Plewes et al. ment using Loma Prieta data. In Proceedings of the 4th
(1992), ii) Olson (2001), iii) Shuttle & Cunning Japan-U.S. Workshop on Earthquake Resistant Design
(2008) and iv) Robertson (2016). of Lifeline Facilities and Countermeasures of Soil
The results of the study showed that for the tail Liquefaction. pp. 177–192.
ings evaluated herein the contours suggested by Olson, S.M. 2001. Liquefaction Analysis of Level and
Plewes et al. (1992), Shuttle & Cunning (2008) and Sloping Ground Using Field Case Histories and Penetra
tion Resistance.
Robertson (2016) yielded similar results (Figure 8).
Olson, S.M. 2009. Strength Ratio Approach for Liquefac
The original boundary suggested by Olson (2001) tion Analysis of Tailings Dams. In Proceedings of the
resulted in a very conservative assessment (Figure 9). University of Minnesota 57th Annual Geotechnical
Equation 12 was used herein to account for the effect Engineering Conference. pp. 37–46.
of compressibility and the ratio qc/N60=0.07 was used Plewes, H.D. Davies, M.P. & Jefferies, M.G. 1992. CPT
to correct the contour to yield results similar to the based screening procedure for evaluating liquefaction
other methodologies evaluated. susceptibility. In Proceedings of the 45th Canadian Geo-
The results of the study are useful to highlight the technical Conference.
importance of the correction of the boundary sug Robertson, P.K. 1990. Soil classification using the cone
penetration test. Canadian Geotechnical Journal, 27(1):
gested by Olson (2001) for soils of medium and high
151–158. doi:10.1139/t90-014.
compressibility (specially tailings). Due to its inher Robertson, P.K. 2009. Interpretation of cone penetration
ent limitations, the authors do not recommend the tests - A unified approach. Canadian Geotechnical Jour
use of the contour suggested by Olson (2001) to nal, 46(11): 1337–1355. doi:10.1139/T09-065.
evaluate the susceptibility to flow liquefaction in Robertson, P.K. 2016. Cone penetration test (CPT)-based
soils of medium to high compressibility. soil behaviour type (SBT) classification system — An
918
update. Canadian Geotechnical Journal, 53(12): Shuttle, D.A. & Cunning, J. 2007. Liquefaction potential of
1910–1927. doi:10.1139/cgj-2016-0044. silts from CPTu. Canadian Geotechnical Journal, 44(1):
Robertson, P.K. & Campanella, R.G. 1983. Interpretation 1–19. doi:10.1139/T06-086.
of Cone Penetration Tests. Canadian Geotechnical Jour Shuttle, D.A. & Cunning, J. 2008. Reply to the discussion
nal, 20(4): 718–745. by Robertson on “Liquefaction potential of silts from
Robertson, P.K. & Wride, C.E. 1998. Evaluating cyclic CPTu.” Canadian Geotechnical Journal, 45(1): 142–145.
liquefaction potential using the cone penetration test. doi:10.1139/T07-119.
Canadian Geotechnical Journal, 35(3): 442–459. Stark, T.D. & Olson, S.M. 1995. Liquefaction resistance
doi:10.1139/t99-102. using CPT and field case histories. Journal of Geotech
Schnaid, F. & Odebrecht, E. 2012. Ensaios de Campo nical Engineering, 121(12): 856–869. doi:10.1061/
e suas aplicações à Engenharia de Fundações. (ASCE)0733-9410(1995)121:12(856).
919
© 2022 the Author(s), ISBN 978-1-032-31259-0
Prospects on data mining approach for pile geotechnical design utilizing

CPT and CPTu records: Case study: AUT database
A. Eslami & S. Heidarie Golafzani
Amirkabir University of Technology, Tehran, Iran
S. Moshfeghi
University of British Columbia (UBC), Vancouver, Canada
ABSTRACT: A comprehensive database has been compiled including CPT soundings performed adjacent
to pile load test and related geotechnical information, namely AUT (Amirkabir University of Technology):
Geo-CPT&Pile Database. In this paper, after a brief review of existing CPT and pile databases, the specifica
tions of the updated AUT: Geo-CPT&Pile Database as well as different categories of data are presented. Sub
sequently, several procedures developed using this database from different geotechnical aspects are reviewed
and introduced. These implementations are extended for appraisal of currently used CPT-based methods for
pile geotechnical design by focusing on methods screening, uncertainty- reliability measures, and Perform
ance-Based Design (PBD) approach regarding resistance factors. After all, an algorithm is presented including
a formulated procedure of pile geotechnical design by means of smart database collections, reproduction of
CPT profile, assortment of competitive methods, integration of geotechnical aspects i.e., capacity, load-
displacement performance and resistance distribution. Correspondingly, via realizing prospects on implemen
tation of major aspects leads towards optimum pile performance-based design upon a data mining approach.
1 BACKGROUNDS in their correlations while others do not, or some

methods have been develop based on total stress ana
The application of databases in geotechnical engin lysis (TSA) while some were developed based on
eering returns to long decades ago since researchers effective stress analysis (ESA) or empirical correl
tried to develop correlations among in-situ or lab ations. Accordingly, various assessment criteria have
measured parameters and design geotechnical been introduced to measure accuracy, precision and
parameters. Cone penetration testing (CPT) is one of efficiency of these predictive methods (Eslami et al.
reliable in-situ testing with less measurement error 2019a).
in comparison to others (Phoon and Kulhawy 1999). The application of CPT and pile databases cannot
It provides continuous records with depth and due to be limited to performance assessment of predictive
its similarity to piles, a broad range of correlations methods and these databases can provide us with
have been made for prediction of ultimate pile bear more information.
ing capacity (Eslami and Fellenius 1997). Piles are Present study introduces a comprehensive geo
important foundation systems in geotechnical engin technical engineering database in the field of CPT
eering facilitating construction in offshore and and pile. This database named as AUT:Geo
onshore areas or even in abnormal subsoil conditions CPT&Pile database has had different applications in
or extreme loads imposed by superstructure in high- the pile engineering such as reliability-based assess
rise buildings and skyscrapers. ment of drilled displacement piles (Moshfeghi &
CPT-based methods differ in many areas such as Eslami 2019), assessing the performance of CPT-
the compiled CPT and pile database regarding the pile based methods in predicting axial bearing capacity
characteristics and the subsoil conditions where these of helical piles, reproduction and realization of CPT
piles were installed, model assumptions and simplifica records (Jamshidi Chenari et al. 2018), predicting
tions, input parameters for predicting unite side and the load displacement behavior of driven piles
toe resistances, data processing for excess pore water regarding CPT records, and uncertainty appraisal of
pressure in back shoulder of CPT cone, failure pattern CPT-based methods regarding statistical, probabilis
around the pile toe and so on. For instance, some tic and reliability-based criteria (Eslami et al. 2020,
methods consider loading directions or friction fatigue Heidarie Golafzani & Eslami 2021) and the
DOI: 10.1201/9781003308829-137
920
importance of relevant data-based approach in geo Valikhah et al. proposed a new analytical-
technical pile design (Eslami & Heidarie Golafzani numerical method to estimate axial load-
2020). Eventual1y, a CPT and pile data-based displacement behaviour of driven piles in granular
approach is introduced for pile design regarding the soils using CPT records. Implementing the method
researches have been done. of stress characteristics, they analysed the stress field
below and around the pile and in effect, the failure
mechanism. This failure mechanism has then been
2 AUT: GEO-CPT&PILE DATABASE; AN used by implementation of the kinematical approach
EFFICIENT TOOL IN DATA MINING FOR of the limit analysis to compute the displacement
GEOTECHNICAL PILE DESIGN field (Eslami et al. 2019a).
The application of artificial intelligence has been
The AUT:Geo-CPT&Pile database has been com developed in geotechnical engineering in recent
piled from well-published and documented geotech years. Implementing group method of data handling
nical engineering sources and includes 600 records type neural networks optimized using genetic algo
of pile loading tests along with the results of adja rithms, Ardalan et al. (2009) estimated the pile unit
cent cone or piezocone penetration tests (Moshfeghi shaft resistance.
et al. 2015; Eslami et al. 2019b). Eslami et al. (2020b) by employing a database
This database was primarily aimed to assess of instrumented pile load tests as well as CPT
the performance of different CPT-based methods. records correlated the pile unit shaft resistance
Moshfeghi & Eslami (2018, 2019) compiled with CPT sleeve friction. They proposed an ana
a database of fourty-three and seventy-six records lytical-empirical procedure for estimating pile
of driven piles installed in sandy soils and their shaft capacity considering scale effects. Factors
adjacent CPT records from AUT:Geo-CPT&Pile such as mechanism and rate of penetration, size
database and studied the effect of different criteria effect (i.e., length and diameter), friction fatigue
for interpretating static pile load test records to are taken into account in this approach.
select the most consistent approach with the CPT- Heidarie Golafzani et al. (2020a, b) compiled
based methods. Among the four selected criteria, a database of sixty driven piles installed in different
the Brinch Hansen 80% criterion and the load at soil types and studied the performance of different
the displacement of 10% of the pile diameter approaches, i.e., static analyses, SPT-based methods
were the two most consistent criteria with the and CPT-based methods including twelve methods.
CPT-based approaches. They also assessed the In their studies, they considered different statistical,
performance of different CPT-based methods con probabilistic and reliability-based criteria, including
sidering wasted capacity index (WCI) and cost mean and coefficient of variation, best fitted line,
optimization regarding safety factor. Assessments 20% accuracy level, cumulative distribution function
indicate that the German (Kempfert & Becker (i.e., P50 and P90-P50), confidence interval, root mean
2010), LCPC (Bustamante & Gianeselli 1982), square error (RMSE) and efficiency ratio (i.e., the
Meyerhof (1983), UniCone (Eslami & Fellenius ratio of load and resistance factor (LRFD) resistance
1997) and UWA-05 (Lehane et al. 2005) methods factor to the model parameter). The model parameter
have shown the most efficient predictions at their was defined as the ratio of measured to predicted
optimum factor of safety. bearing capacity. They concluded that modified first
Askari Fateh et al. assessed the performance of ten order second moment (FOSM) is less time-
direct CPT-based methods for helical piles via consid consuming and complicated, and leads to statistically
ering thirty-seven cases of helical piles installed in identical resistance factors to first order reliability
different soil types. Also, the accuracy of two differ (FORM) method and Monte Carlo simulations
ent assumptions of failure mechanism around helical (MCS). They also compared the performance of the
piles was examined through comparing the predicted considered approaches and methods via radar charts
axial bearing capacity and measured ones in static and deduced that CPT-based methods perform better
pile load tests. Finally, they suggested a new CPT- than two other approaches.
based method to estimate the bearing capacity of hel Eslami & Heidarie Golafzani (2020) stated that
ical piles (Eslami et al. 2019a). selection of predictive methods for pile geotech
Jamshidi et al. (2018) developed an algorithm for nical design has a pivotal role in an optimum and
realization of CPT data based on non-stationary site-specific design. Results indicated that methods
random field theory. The proposed algorithm imposes prioritized by statistical, probabilistic and reliabil
soil layering alongside inherent soil variability based ity-based criteria attain higher resistance factor in
on Eslami and Fellenius (1997) soil classification load and resistance factor design (LRFD) or lower
chart. After detection of soil layering based on the safety factor in allowable stress design (ASD)
simplified proposed approach, the statistical charac approaches. They emphasized that the global safety
teristics of each soil layer are defined as multi-criteria factor cannot stands alone against all uncertainty
functions, assembled into the non-stationary auto- sources and methods attaining similar safety fac
covariance matrix and the routines continue in Monte tors, do not result in the same probability of failure
Carlo scheme for production of CPT records. and reliability index (i.e., β).
921
3 GEOTECHNICAL DATA-BASED PILE number of testing and comprehensive geotechnical
DESIGN site investigation.
If the number of tests is minimum or tests are
As reviewed, CPT and pile databases support the geo unavailable, it is referred to the database for either
technical designer with invaluable information about gaining extra information via smart selection accord
the performance of predictive methods for special pile ing to the available akin sites or data production gen
types installed in a particular project site/with definite erally such as CPT data or specifically such as P-Δ
characterizations or even for a wide range of pile types and resistance distribution if necessary. These proced
installed in various site locations and conditions. Fur ures constitute the first stage of data collection.
thermore, databases can be mined for extra information The second stage is about processing the axial pile
aiding in geotechnical pile design. Figure 1 illustrates bearing capacity predictions. Various methods and
the steps for this new suggested/proposed approach. approaches result in a wide range of predictions.
In this approach, it is necessary to compile Regardingly, applying miscellaneous evaluation cri
a database of geotechnical information about various teria including statistical and probabilistic criteria and
sites including pile load test results along with adja risk, reliability and efficiency-based criteria, leads to
cent CPT records to the investigated piles. selection of appropriate methods according to the
In the first step, the engineer should evaluate the available site conditions, codes and local information.
site conditions in terms of its available geotechnical Indeed, this stage is dedicated to method screening
information and categorizes it as no testing, minimum leading to optimum geotechnical design. Eventually,
Figure 1. Geotechnical data-based pile design algorithm.
922
the understanding of the geotechnical engineer is
improved and an optimum site-specific design
approach is attained.
4 ENGINEERING IMPLEMENTATION
A database of sixty driven piles was gathered from

the AUT:Geo-CPT&Pile database with their adjacent
CPT records worldwide to illustrate the application of
the proposed data-based geotechnical pile design
approach. Most of the piles are installed in clayey
soils (i.e., about 45% of the piles in the compiled
database), followed by sandy soils (i.e., about 32%)
and mixed soils (I.e., about 23%). According to what
explained in the earlier section, for an optimum geo
technical pile design, it is appropriate to perform
a comprehensive in-situ and lab tests to characterize
the subsoil conditions and further consider it. In the
case, the number of tests was limited or was not avail
able for the current project, the geotechnical practi
tioner can gather extra information about the project
site by searching through the available geotechnical
site investigation reports for the nearby projects or
a by smart selection from CPT and pile databases.
From the investigated database a driven pile was
selected. This circular concrete pile was installed in
sandy soils of Norway with embedment length of
15.5 m and diameter of 280 mm. By searching for
similar site and pile conditions, a concrete square
pile installed in Sweden was selected with embed
ment length of 12.8 m and width of 235 mm.
Figure 2 provides the soil behavior classification
(SBC) for these two close sites. As it is presented,
the most majority of soil types are silt-sand and
sand-gravel regarding the Eslami and Fellenius
(1997) soil behavior classification.
Figure 3 illustrates the static pile load test (PLT)
results for these two piles. Both piles behaved simi
larly under compression loading. As stated earlier, in
case insufficient geotechnical data, Jamshidi Chenari
et al. (2018) proposed a non-stationary algorithm for Figure 2. Soil stratigraphy; (a) soil behavior classification
reproduction and realization of CPT records regard for the two sites (Eslami and Fellenius 1997), (b) the sub
ing soil stratigraphy. soil layering in Sweden, (c) the subsoil layering in Norway.
Figure 4 compares the reproduced CPT records,
i.e., qc and fs and their analogous real ones for the
two investigated sites in Norway and Sweden. The 1983), Meyerhof (1983), UniCone (Eslami and Felle
geotechnical engineer can reproduce these CPT nius 1997), German (Kempfert and Becker 2010),
records with different inherent soil variability chan and Modified UniCone (Niazi and Mayne 2016) have
ging from low to high in soil characteristics reflect led to close predictions. However, other investigated
ing the subsoil condition. methods, including Schmertmann (1978), Dutch (de
Figure 5 compares the performance of different Ruiter and Beringen 1979) andFugro-05 (Kolk et al.
CPT-based methods for these two piles via model par 2005) resulted in less accurate predictions due to the
ameter and it is the ratio of measured to predicted differences mentioned earlier. Regardingly, finding
bearing capacity, i.e., Qm/Qp. The site conditions for matching piles with similar dimensions and embed
these two considered piles, influence the performance ment length installed in close or near close site condi
of predictive methods appropriately in a similar way. tions, enhances the geotechnical engineer to have an
For instance, if the acceptable limit for Qm/Qp optimum site-specific design by selecting appropriate
changes from 0.8 to 1.2 as is shown in Figure 5, criteria and implementing multi-criteria decision-
methods such as LCPC (Bustamante and Gianesselli making models.
923
Figure 3. Pile load test results for the two considered piles.
Figure 5. The performance of different CPT-based methods
for these two considered piles.
5 CONCLUSIONS
Uncertainties are an inseparable part of geotechnical

engineering, and many attempts have been made to
consider their influence on geotechnical designs in the
last decades. Characterizing the subsoil conditions is
one of the essential steps in geotechnical designs, and
CPT by producing continuous and reliable records has
a crucial role in determining inherent soil variability.
One of the primary applications of CPT soundings is
predicting axial pile bearing capacity due to its simi
larity to the pile. In this regard, several methods have
been developed based on CPT records. However,
these methods result in a wide range of predictions
according to the embedded uncertainties. To overcome
the difficulties of selecting an appropriate predictive
method, CPT and pile databases assist the geotech
nical engineer. Accordingly, AUT:Geo-CPT&Pile
database was introduced, and its implementation in
various pile engineering issues was reviewed. How
ever, the data mining of such a database is not limited
to performance assessment of predictive methods, and
an algorithm for data-based pile design approach is its
other prospect. In this approach, with the aid of avail
able data in the compiled database, the designer can
search for the matching site and pile conditions in the
absence of necessary and sufficient site geotechnical
information or even reproduce and realize CPT
records. After gathering adequate CPT and pile data,
various predictive methods can be assessed via statis
tical, probabilistic, and reliability-based criteria. Even
tually, the methods’ screening procedure is finalized
Figure 4. Reproduced CPT records for the two site condi via considering codes and local info, and the superior
tions, (a) the Swedish site, (b) the Norwegian site. methods are selected. Correspondingly, the final pile
924
geotechnical design will be an optimum pile perform Capacity; Static Analysis, SPT and CPT-Based
ance-based design upon a data mining approach. The Methods. Geotechnical and Geological Engineering,
Application of this algorithm was presented through pp.1–19.
an example. Heidarie Golafzani, S., and Eslami, A., 2021. Uncertainty
Appraisal of CPT-based Methods for Axial Pile Bearing
Capacity. In Proceedings of the 46th Annual Conference
REFERENCES on Deep Foundations, Deep Foundation Institute (DFI),
Las Vegas, Nevada, October 12-15, 12p.
Ardalan, H., Eslami, A. and Nariman-Zadeh, N., 2009. Jamshidi Chenari, R., Kamyab Farahbakhsh, H., Heidarie
Piles shaft capacity from CPT and CPTu data by polyno Golafzani, S., and Eslami, A., 2018. Non-stationary
mial neural networks and genetic algorithms. Computers realisation of CPT data: considering lithological and
and Geotechnics, 36(4),pp.616–625. inherent heterogeneity. Georisk: Assessment and Man
Bustamante, M. & Gianeselli, L., 1982. Pile Bearing Cap agement of Risk for Engineered Systems and Geoha
acity Prediction by Means of Static Penetrometer CPT. zards, 12(4) 265–278.
Proceedings of the 2nd European Symposium on Pene Kolk, H.J., Baaijens, A.E. and Senders, M., 2005. Design
tration Testing, 493–500. criteria for pipe piles in silica sands. In Design criteria
de Ruiter, J. and Beringen, F., 1979. Pile foundations for for pipe piles in silica sands (pp. 711–716). CRC Press/
large North Sea structures. Marine Georesources & Balkema.
Geotechnology, 3(3) 267–314. Kempfert, H.G. and Becker, P., 2010. Axial pile resistance
Eslami, A. & Fellenius, B. H. 1997. Pile Capacity by of different pile types based on empirical values. In
Direct CPT and CPTu Methods Applied to 102 Case Deep Foundations and Geotechnical In-Situ Testing, pp.
Histories. Canadian Geotechnical Journal, 34, 886–904. 149–154.
Eslami, A., Moshfeghi, S., Heidari, S., and Valikhah, F., Lehane, B.M., Schneider, J.A. and Xu, X., 2005. The
2019 a. AUT:Geo-CPT&Pile database updates and UWA-05 method for prediction of axial capacity of
implementations for pile geotechnical design, Geotech driven piles in sand. Frontiers in offshore geotechnics:
nical Engineering Journal of the SEAGS & AGSSEA, ISFOG, pp.683–689.
50(3) 74–90. Meyerhof, G. G., 1983. Scale Effects of Ultimate Pile
Eslami, A., Moshfeghi, S., MolaAbasi, H. and Eslami, M. Capacity. Journal of Geotechnical Engineering, 109,
M., 2019 b. Piezocone and Cone penetration test (CPTu 797–806.
and CPT) applications in foundation engineering. But Niazi, F.S. and Mayne, P.W., 2016. CPTu-based enhanced
terworth-Heinemann. UniCone method for pile capacity. Engineering Geol
Eslami, A., Heidarie Golafzani, S., and Moshfeghi, S., ogy, 212, pp. 21–34.
2020a. CPT and pile database for performance-based Phoon, K.-K. & Kulhawy, F. H., 1999. Evaluation of Geo-
design of pile axial bearing capacity. In Proceedings of technical Property Variability. Canadian Geotechnical
the 45th Annual Conference on Deep Foundations, Journal, 36, 625–639.
Deep Foundation Institute (DFI), Oxon Hill, Maryland, Moshfeghi, S., Eslami, A., and Hosseini, S.M.M., 2015.
October 13-16, 12p. AUT-CPT&Pile database for piling performance using
Eslami, A., Lotfi, S., Infante, J.A., Moshfeghi, S., and CPT and CPTu records. In Proceedings of the 40th
Eslami, M.M., 2020b. Pile shaft capacity from cone Annual Conference on Deep Foundations, Deep Foun
penetration test records considering scale effects. Inter dation Institute (DFI), Oakland, California, October 12
national Journal of Geomechanics, ASCE, 20(7), 13 p. 15, 10p.
Eslami, A. and Heidarie Golafzani, S., 2020. Relevant Moshfeghi, S. and Eslami, A., 2018. Study on pile ultimate
data-based approach upon reliable safety factor for pile capacity criteria and CPT-based direct methods. Inter
axial capacity. Marine Georesources & Geotechnology, national Journal of Geotechnical Engineering, 12(1)
pp.1–14. 28–39.
Heidarie Golafzani, S., Jamshidi Chenari, R., and Moshfeghi, S. and Eslami, A., 2019. Reliability-based
Eslami, A., 2020a. s, Georisk: Assessment and manage assessment of drilled displacement piles bearing capacity
ment of Risk for Engineered Systems and Geohazards, using CPT records. Marine Georesources & Geotechnol
DOI:10.1080/17499518.2019.1628281. ogy, 37(1) 67–80.
Heidarie Golafzani, S., Eslami, A. and Jamshidi Schmertmann, J.H., 1978. Guidelines for cone penetration
Chenari, R., 2020b. Probabilistic Assessment of Model test: performance and design (No. FHWA-TS-78-209).
Uncertainty for Prediction of Pile Foundation Bearing United States. Federal Highway Administration.
925
© 2022 the Author(s), ISBN 978-1-032-31259-0
Evaluating mitigation of kinematic moments of precast driven piles in

liquefiable layers using pre- and post-CPTu tests
K. Fakharian, D. Mohtashamamiri, K. Behroozian & M. Bahrami
Department of Civil & Environmental Engineering, Amirkabir University of Technology, Tehran, Iran
T. Bahrami & I.H. Attar

Pars GeoEnviro Inc., Tehran, Iran
ABSTRACT: Precast piles are driven to 23 m depth to improve the capacity of foundations located on satur
ated liquefiable sands on top underlain by a thick soft to medium stiff clay. Analyses show excessive kine
matic moment at borderline of the two layers during liquefaction. It was decided to drive additional shorter
piles (12 m) to mitigate liquefaction hoping that kinematic moments shall be reduced within allowable limits.
CPTu tests were carried out before and after pile driving. The results show that qc did not change considerably
in the top saturated silty sand layer as well as the clayey soil underneath. The friction sleeve, however,
increased more significantly due to radial displacement and compaction/consolidation attributed to cavity
expansion during pile driving. Analyses showed that pile arrangement was not sufficient to overcome lique
faction during earthquake. However, safety factor increased sufficiently to reduce kinematic moment within
the allowable limits. The details of the site operations, CPTu data before and after pile driving, analyses and
interpretations are presented and discussed.
Keywords: Pizeocone Penetration Test (CPTu), Liquefaction, Precast pile, Kinematic moment
1 INTRODUCTION effects was analyzed by Blaney et al. (1976), Kagawa

& Kraft (1980), Dobry & O’Rourke (1983), Nikolaou
Soil deposits at a site subjected to an earthquake may et al. (1995), Luo & Murono (2001) and others. Deter
experience increases in pore water pressure and time- mination of the kinematic forces acting on the pile
dependent vertical and lateral ground movements. As depends on the liquefaction potential and soil param
a consequence of the earthquake-induced ground eters before and during liquefaction.
movements, piles and other deep foundations will be Conventional experiences have held that driving
subjected to two sources of additional lateral loading: displacement piles into loose and medium dense sand
will densify the soil volume close to the pile. Some
a. Inertial Forces/Moments: forces and moments researchers including Meyerhof (1959), Nataraja &
that are induced in the piles because of the accel Cook (1983), Bement & Selby (1997), Gianella et al.
erations generated within the structure by the (2015), Stuedlein et al. (2016) and Stuedlein & Gia
earthquake. nella (2017) and Rhyner (2018) have studied the densi
b. Kinematic Forces/Moments: forces and bending fication effects caused by driving piles into granular
moments that are induced in the piles because of soil. By changing the soil properties due to pile driving
the ground movements resulting from the earth and subsequent alterations in the soil liquefaction
quake. Such movements will interact with the piles potential, the amount of kinematic forces on the pile
and because of the difference in stiffness of the will change. Similarly, some have evaluated the
piles and the moving soil, there will be lateral stres effects of pile driving on variations of stress state
ses developed between the pile and the soil, result
and consolidation surrounding the pile shaft (e.g.,
ing in the development of shear forces and bending Khanmohammadi & Fakharian, 2018; Fakharian &
moments in the piles. Khanmohmmnadi, 2022). Fakharian et al. (2014)
Tajimi (1969) and Penzien (1970) were among the evaluated the effect of surcharge pressure of the reac
first researchers to study the problem through using tion system on the load-movement response of piles.
analytical and numerical approaches, respectively. Fol This paper presents a case study conducted to evalu
lowing these early efforts, the problem of kinematic ate the effects of pile driving on loose, clean and silty
DOI: 10.1201/9781003308829-138
926
sand properties. The subsoil condition before pile driv ultimate bending moment. It was decided to drive add
ing was characterized using CPTu test as well as bore itional shorter piles (12 m) as “compaction piles” to
holes and soil sampling. Square precast concrete piles mitigate the liquefaction hoping that the kinematic
were designed and driven down to 23 m of embedment moments shall be reduced within the allowable limit of
depth as the main piles or bearing piles. Analytical cal the pile section. Figure 3 shows the bearing and com
culations showed excessive kinematic moments at the paction piles layout in the trial field.
borderline of the two layers when liquefaction occurs. After driving the bearing and compaction piles,
It was decided to drive additional shorter 12 m piles four CPTu tests were performed to evaluate the
(compaction or improvement piles) to mitigate the changes in soil condition at locations surrounded by
liquefaction hoping that the kinematic moments shall piles and beyond. All tests were performed about 45
be reduced within the allowable limits of the pile sec days after pile driving. The location of the post-pile
tion. After driving the main piles and consolidation driving tests is shown in Figure 3. CPT-10 and CPT-
piles, hereinafter referred to as bearing and compaction Ref are outside of the piling area and in a way indi
piles, another 4 CPTu were conducted to see how the cates the initial state (in situ) of the soil. CPT-ZA3
pile driving had contributed to the soil improvement and CPT-ZA4 are located between the bearing and
between the piles. The site geological condition, CPTu compaction piles, and CPT-ZA5 is located between
data before and after pile driving, kinematic moment compaction piles.
analyses before and after pile driving are carried out
and interpretations are presented and discussed.
2 FIELD DESCRIPTION
2.1 Subsoil conditions

The trial field is located in Jask approximately 250 km
south-east of Bandar-e-Abbas city in Hormozgan prov
ince, on coastal area of Gulf of Oman, Iran. Project
area extends North-South, with an apparent length of
1600 m and width of 400 m. The soil stratification of
the trial zone consists of three different layers
(Figure 1):
Layer (I): The layer mostly consists of sand and
cohesionless silt and classified as medium dense to
dense silty sand. This layer has been observed from Figure 1. Geotechnical cross-section of the trial field in
ground surface down to 8 m depth. GWT is Jask, on Coastlines of Sea of Oman.
1.5 m deep any lies in Layer I.
Layer (II): Second layer consists of firm to stiff
clay and silty clay. This layer underlying layer (I)
starts from the depth of 8 m and in some areas has
extended to a depth of 28 m.
Layer (III): This Layer is situated below
the second layer (deeper than 28 m) and consists of
medium dense to very dense sand and silty sand.
Figure 2 shows the distribution zones of clay, silt
and sand with depth. In fact the red line with square
points shows the border of clay-silt and the yellow line
with triangle points shows the border of silt-sand.
Down to 8 m, soil dominantly constitutes sand and silt,
and the percentage of clay is less than 40%. From
depth of 8 to about 26 m, the clay fraction is between
60 to 80%.
2.2 Main and compaction piles layout

Square 400 × 400 mm precast concrete piles were
designed and driven down to 23 m of depth as the
“main piles”. Ultimate bending moment of piles is 143
kNm. During liquefaction, however, analytical calcula
tions show that kinematic moment at the borderline of
the two layers (Layer I and Layer II) is greater than the Figure 2. Grain fraction proportions with depth.
927
2.3 Pre- and post-driving trial tests classification results into the zone of the soils pre
senting the overconslidated or cemented type.
Results of pre- and post-driving CPTu tests are shown
in Figure 4. The results after pile driving indicates
that no considerable changes are observed in qc in
both layers (I) and (II) till depth 12 m, equivalent to
the length of compaction piles. The friction sleeve fs,
however, has increased significantly. The pore pres
sure variations show that original in situ condition has
induced higher pore pressure from starting the clayey
soil of Layer II (CPT-10 and CPT-Ref). However,
after pile driving, both ZA3 and ZA4 points situated
in between the piles have generated negative to hydro
static values of u down to 12 m of depth equivalent to
the depth of compaction piles. This is understood to
be attributed to compaction of sand as a result of
increase in radial stress around the pile shaft, contrib
uting to dilative response during shearing and hence
generation of negative PWP as CPT loading is rapid
hence considered as undrained. Friction ratio Rf has
significantly increased within the 12 m of the compac
tion piles and even down to 15 m.
Figure 3. “Main” and “compaction” piles layout at trial field.
As shown in Figure 2, the soil has a significant per

centage of fine-grained at most depths. Especially from
depth of 8 m, the percentage of sand in the soil is zero
indicating that the soil has very low compactibility, and
as a result, qc has not increased significantly.
Pile penetration induces soil compaction and con Figure 4. Pre- and post-driving CPTu results- trial field in
solidation which is manifested by an increase in Jask Oil Terminal project.
horizontal effective stress, i.e., it causes the in-situ
earth pressure coefficient, K0, to increase. Sleeve
friction is defined as the product of horizontal effect 3 INTERPRETATIONS
ive stress and the friction coefficient at the soil–CPT
sleeve interface. Assuming the coefficient of friction 3.1 Assessment of soil liquefaction potential
between the soil and CPT sleeve is constant, sleeve
friction increases due to the increase in horizontal A procedure suggested by Robertson & Wride (1998)
effective stress (normal to sleeve). is used to evaluate the potential for cyclic softening
The analysis of CPTu results using Robertson and cyclic liquefaction. The procedure used the basic
(1990) soil classification chart (Figure 5) shows that methodology, developed by Seed & Idriss (1971),
the pile driving process has dragged the CPTu calculating cyclic stress ratio (CSR) induced by the
928
for CPT-ZA3 and CPT-ZA4, the cyclic softening FS
is above 1 compared to the cyclic liquefaction FS of
CPT-Ref which is lower than unity.
In depths below 9 m, calculation of potential for
cyclic softening is performed, considering the dom
ination of clay-like behavior of the soil. The uniform
form of the calculated safety factor profile for the
CPT-Ref (generally lower than 1) has changed due
to the installation of the compaction piles down to
about 12 m. This shows the effectiveness of driving
compaction piles and raising FS to greater values.
Figure 5. Robertson (1990) soil classification of pre- and

post-driving CPTu results.
earthquake, and cyclic resistance ratio (CRR) of the

soil. If CSR is greater than CRR, then liquefaction
can occur. The proposed procedure estimates CRR
based on the “sand-like” or “clay-like” classification
identified by factor Ic=2.6. The updated procedure is
well established in 2009 by Robertson (2009). Sand-
like behavior is allocated to soils having Ic<2.5. Such
soils have the potential for cyclic liquefaction. Clay-
like soils have Ic>2.7 in which cyclic softening
potential is recognized. The transition from sand-like
to clay-like behavior generally occurs when 2.5<Ic
<2.7. Fine- grained soils transition depends on Atter
berg Limits and Plasticity Index.
Figure 6 shows the evaluation of “cyclic liquefac Figure 6. Changes of FS for “liquefaction” and “cyclic
tion” and “cyclic softening” potential for CPT-Ref, softening” for pre- and post-driving CPTu results.
CPT-ZA3 and CPT-ZA4. Down to 6 m depth, Ic
value is lower than 2.5 where the probability of
3.2 Assessment of kinematic loading effects
cyclic liquefaction is significant. CPT-Ref, CPT-ZA3
and CPT-ZA4 show cyclic liquefaction factor of The kinematic moment is evaluated using a method
safety (FS) lower than 1 between 4 to 6 m. Consid proposed by Nikolaou et al. (2001). In this method,
ering the effects of the main and compaction pile pile is modeled as a beam on a dynamic Winkler
driving, CPT-ZA3 and CPT-ZA4 demonstrate the foundation (BDWF), with homogenous, isotropic,
cyclic liquefaction safety factor is higher than the and linearly elastic soil in each layer and a constant
CPT-Ref in this layer (average FS of 0.74 at ZA3 soil-damping ratio. They found that induced
point versus average FS of 0.45 at the Ref. point) moments were maximum at interfaces between
Despite the higher FS, the calculations still show layers of different stiffness, and then performed
that cyclic liquefaction potential in this zone still a series of calculations to determine the bending
exist. However, about 60% increase in FS has con moment at the interface between the two layers.
tributed to significant reduction of the kinematic The method does not consider any particular condi
forces induced in the piles. tion about the thickness of the two layers and is
The soils between 6 to 9 m are classified as transi therefore, applicable for any depth of the layer
tion soils. As the main and compaction piles pene interface.
trate through the soil, the CPT-ZA3 and CPT-ZA4 The method proposes a reduction factor η to be
have shown higher sleeve friction, fs. Thus, the soil applied to the maximum steady-state pile bending
behavior categorized in the transition zone has turned moment within the frequency domain to arrive at the
into clay-like behavior dominated by the potential for corresponding peak value in the time domain. In the
cyclic softening. As illustrated in the subjected zone current study, the parameter η is equivalent to 0.37.
929
The reduction of the maximum kinematic bending
moment is resulted from: (1) increase in fs (due to
increase in lateral stress in soil) and hence reduction
where Mres is the bending moment developed under in liquefaction potential of the soil, (2) reduction of
resonant conditions and it is calculated using the fol the liquefied layer thickness.
lowing relation: The presented results and the differences in CPTu
parameters between the initial in situ condition and
after the pile driving sequences are indications of
effective improvement role of the driven main and
compaction piles.
in which d is pile diameter, L pile length, Ep 4 CONCLUSIONS

Young’s modulus of pile, E1 Young’s modulus of
upper layer,Vs1 average shear wave velocity in upper A case study is presented on the effect of driving
layer,Vs2 average shear wave velocity in lower layer, “main bearing” and “compaction or improvement”
ρ1 the mass density of upper layer and h1 is the piles in the silty sand to silty clay strata at a study
thickness of upper layer. field having high seismic and liquefaction/cyclic
Shear stress τc which is proportional to the actual softening potential. CPTu tests were performed both
shear stress is likely to develop at the interface, as prior and after driving the piles. Analyses were per
a function of the free-field acceleration at the soil formed to evaluate the maximum kinematic
surface, as: moments at the interface of the two sand-like (top)
and clay-like (bottom) layers. On the basis of the
presented results, the main conclusions are summar
ized below:
• The qc increased slightly in the top saturated
Predictions obtained using the above formula are layer as well as the clayey soil underneath till
shown in Figure 7 for the Ref point and ZA3. The depth of 12 m, equivalent to the length of com
magnitude of kinematic moment is significantly paction piles.
lower in the test results of post-driven point, CPT • The friction sleeve (and hence friction ratio),
ZA3. This is understood to be attributed to the effect however, increased more significantly due to
of the main and compaction piles driven down to radial displacement and compaction/consolidation
23 m and 12 m, respectively. attributed to cavity expansion surrounding the
After driving the compaction piles, the thickness pile during driving.
of the liquefied layer has reduced from 8.5 m • Liquefaction analyses show that the pile arrange
(left graph of Figure 7) to 5.5 m (right graph). ment was not sufficient to overcome the liquefac
tion for the earthquake of the site with 0.4g
acceleration and M=7.
• However, the factor of safety increased suffi
ciently to reduce kinematic moments within the
allowable limits.
REFERENCES
Bement, R. & Selby, A. 1997. Compaction of granular
soils by uniform vibration equivalent to vibrodriving of
piles. Geotechnical & Geological Engineering 15(2):
121–143.
Blaney, G. W. 1976. Dynamic stiffness of piles. In Proc. 2nd
Int. Conf. Numer. Meth. Geomech: 1001–1012. Blacksburg.
Dobry, R. & O’rourke, M. (1983). Discussion of Seismic
response of end-bearing piles. Journal of Geotechnical
Engineering 109(5): 778–781.
Fakharian, K., Meskar, M. & Mohammadlou, A.S. 2014.
Effect of surcharge pressure on pile static axial load test
results, International Journal of Geomechanics, ASCE
14(6): 40241–40249.
Fakharian, K. & Khanmohammadi, M. 2021. Effect of
OCR and pile diameter on load-movement response of
Figure 7. Schematic of kinematic loading value at inter piles over time embedded in clay, International Journal
faces between layers for a) CPT-Ref, b) CPT-ZA3. of Geomechanics, ASCE, (in press)
930
Gianella, T. N., Stuedlein, A. W., & Canivan, G. J. 2015. Penzien, J. 1970. Soil-pile foundation interaction. Earth
Densification of liquefiable soils using driven timber quake engineering: 349–381.
piles. In 6th International Conference on Earthquake Robertson, P. K. 1990. Soil classification using the cone
Geotechnical Engineering. Christchurch, New Zealand. penetration test. Canadian Geotechnical Journal 27(1):
Kagawa, T., & Kraft, L. M. 1980. Lateral load-deflection 151–158.
relationships of piles subjected to dynamic loadings. Robertson, P. K. & Wride, C. 1998, Evaluating cyclic
Soils and Foundations 20(4): 19–36. liquefaction potential using the cone penetration test.
Khanmohammadi M. & Fakharian K. 2018. Numerical Canadian Geotechnical Journal 35(3): 442–459.
simulation of soil stress state variations due to mini-pile Robertson, P. K. 2009. Interpretation of cone penetration
penetration in clay. International Journal of Civil Engin tests—a unified approach. Canadian geotechnical jour
eering, Transaction B: Geotechnical Engineering 16(4): nal 46(11), 1337–1355.
409–419. Seed, H. B., & Idriss, I. M. 1971. Simplified procedure for
Luo, X. & Murono, Y. 2001. Seismic analysis of pile foun evaluating soil liquefaction potential. Journal of the Soil
dations damaged in the January 17, 1995 South-Hyogo Mechanics and Foundations division 97(9): 1249–1273.
Earthquake by using the seismic deformation method. Stuedlein, A. W. & Gianella, T. N. 2017. Effects of driving
International Conferences on Recent Advances in Geo- sequence and spacing on displacement-pile capacity.
technical Earthquake Engineering and Soil Dynamics. Journal of Geotechnical and Geoenvironmental Engin
San Diego, California eering 143(3): 06016026.
Meyerhof, G. 1959. Compaction of sands and bearing cap Rhyner, F. C. 2018. Densification of granular soils by pile
acity of piles. Journal of the Soil Mechanics and Foun driving and implications for evaluation of liquefaction.
dations Division 85(6): 1–29. In IFCEE 2018: 284–300. Orlando, Florida
Nataraja, M. S. & Cook, B. E. 1983. Increase in SPT Stuedlein, A. W., Gianella, T. N., & Canivan, G. 2016.
N-values due to displacement piles. Journal of geotech Densification of granular soils using conventional and
nical engineering 109(1): 108–113. drained timber displacement piles. Journal of Geotech
Nikolaou, A., Mylonakis, G., & Gazetas, G. 1995. Kine nical and Geoenvironmental Engineering 142(12):
matic bending moments in seismically stressed piles. 04016075.
State University of New York, Buffalo, Report NCEER Tajimi, H. 1969. Dynamic analysis of a structure embedded
95-0022, National Center for Earthquake Engineering in an elastic stratum. Proc. 4th World Conf. on Earth
Research quake Eng. Santiago.
931
© 2022 the Author(s), ISBN 978-1-032-31259-0
Verification of 3D FEM analysis of ground improvement works using

CPT test results
G.A. Faour & K.N. Khouri
NSCC International Ltd, Abu Dhabi, United Arab Emirates
ABSTRACT: The cone penetration test (CPT) has been widely used as one of the most well-known and
reliable performance tests for ground improvement works. A comprehensive CPT testing campaign was con
ducted for the proposed ground improvement works of a highly technical project in Oman. The ground
improvement works were evaluated based on the results of pre and post improvement CPT testing. CPT was
the basis for choosing rigid inclusions (RI) and Vibro-compaction (VC) as the most suitable and economical
ground improvement techniques for this project. The acceptance criteria based on the Post improvement CPT
results incorporated the liquefaction potential and both allowable immediate and long-term settlements. The
degree of ground improvement was evaluated by verifying that the final test results in the field met the project
design criteria. It was concluded that CPT testing can be accurately correlated with multiple ground improve
ment acceptance criteria including liquefaction and settlement.
1 INTRODUCTION
1.1 Project description

A Development Company are constructing a
1.50 million square meters, world class multi cluster
entertainment and leisure park in Barka, Oman.
NSCC International was contracted for the design,
construction and verification of the ground improve Figure 1. Pre-CPT campaign - RDE area.
ment works. The design was conducted to provide
a cost-effective solution to mitigate liquefaction and
achieve both immediate and long-term settlement
criteria. The project was divided into two main
areas: Retail, Dining, Entertainment (RDE) area and
Theme Park area (TP). This paper elaborates on the
design of the ground improvement schemes and their
verification using CPT testing for these two areas.
1.2 Site investigation program

A series of one hundred additional cone penetration
tests (CPT) (to compliment the borehole test results
provided by the tender stage soil investigation report)
were undertaken to provide a detailed evaluation of
Figure 2. Pre-CPT campaign - Theme Park area.
the subsurface conditions prior to construction for
both RDE and TP areas. Figures 1 and 2 illustrate the
approximate locations of the conducted boreholes and
CPTs for both the RDE and TP areas respectively. interchangeable layers of siltstone and sandstone. The
sand layer was typically about 6.0m in thickness with
occasional interlayering of a cohesive silt zone up to
1.3 Subsurface conditions 1.0m thick encountered at 1.50m below the existing
The soil conditions generally consisted of sandy ground level. Another silt layer was found at a depth
layers overlying silt layers which are underlain by of 6.50m below the existing ground level with
DOI: 10.1201/9781003308829-139
932
a thickness of 2.0m. The underlying siltstone and sand that a soil layer of approximately 1m in thickness
stone layers were typically located at about 8.0m in and susceptible to liquefaction was found at about
depth. Figure 3 shows the variation of the cone pene 2.0m depth below the existing ground level; which
tration resistance values (qc) with depth for the differ may be underlain by thin discontinuity lenses of
ent soil strata encountered in both RDE & TP areas. liquifiable soil within a zone from about 4.0m to
The groundwater table was encountered at about 4 6.5m in depth.
to 4.50m depth below the existing ground level In addition to the above, settlement analyses were
(-2.00m MSL); however, a conservative design performed for all RDE foundations by checking both
groundwater table level of 0.00m MSL was considered immediate and long-term settlements. Based on the
in the design and the analysis to allow for seasonal calculated settlement values and the Factor of Safety
variations. against liquefaction, it was deemed necessary to imple
ment ground improvement techniques for foundations
underlain by these potentially liquifiable zones to meet
the project performance criteria. Several ground
improvement methods were evaluated to eliminate the
susceptibility of liquifiable soil and to eliminate exces
sive settlements. Based on the evaluation of the differ
ent ground improvement methods, the vibro
compaction method was selected to address both
settlement and soil liquefaction concerns under all the
RDE structures.
2.2 Theme Park (TP) area

The same analyses were repeated for the Theme Park
area by analysing the pre-CPTs illustrated in
Figure 2. The results of the analyses indicated that
there is no risk of liquefaction for this area except for
a few structures (Entry, Citadel, and Pond 11 & 12).
However, our calculations show excessive long-term
settlements for all structures founded in this area due
to the presence of a consistent thick silt layer. Based
on the above analyses, vibro-compaction was chosen
to mitigate liquefaction, followed by rigid inclusions
Figure 3. Variation of qc with depth. to minimise the foundations settlements. For struc
tures which are not prone to liquefaction, only rigid
inclusions were adopted. The following section
depicts the full liquefaction and settlement analyses
2 GEOTECHNICAL EVALUATION established for one of the structures in the Theme
Park area, namely, the Citadel & Entry structure.
2.1 Retail, Dining, Entertainment (RDE) area
Pre-CPT analyses were carried out to assess the 3 AN APPLICATION CASE TO CITADEL AND
liquefaction susceptibility and to estimate the result ENTRY STRUCTURE
ant immediate and long-term settlement of the
encountered soils under the RDE structures. The 3.1 Structure properties
methodology to estimate the liquefaction potential
requires an estimate of the cyclic stress ratio (CSR) Figure 4 portrays the overall geometry of Citadel &
profile caused by the design earthquake and the Entry structure along with the CPTs conducted in
cyclic resistance ratio (CRR) of the ground. If the this area. The total area of this structure is 3,976m2
CSR is greater than the cyclic resistance ratio with a design total service loading of 125kPa.
(CRR), liquefaction can occur. The CSR is usually
estimated based on potential triggering shear stresses 3.2 Liquefaction and settlement assessment
resulting from a 1 in 475-year design earthquake
exerting a firm ground (i.e., Bedrock) peak horizon Both liquefaction and settlement analyses were imple
tal acceleration of 0.15g. In our case this gave mented using the results of the conducted pre-CPTs.
a potential ground surface peak horizontal acceler The results of the analyses show the presence of soil
ation of 0.23g, due to soil amplification. A simplified layers susceptible to liquefaction. To mitigate liquefac
method to estimate CSR was developed by Seed and tion risk under this structure, Vibro-compaction was
Idriss (1971) based on the maximum ground acceler performed by adopting a 3.0m triangular grid to an
ation at the site. The results of the analyses revealed improvement depth of 8.50m. The grid spacing was
933
Figure 4. Citadel and entry layout.
chosen based on the results of the trial campaign con Figure 5. Liquefaction results using CLiq software.
ducted in that area. The effectiveness of the grid spa
cing was verified using the results of the zone load
tests (ZLT) with four CPTs executed at the four corners
of the ZLT. PLAXIS 3D analysis was carried out to following response spectrum curve and time history.
compare the theoretical ZLT settlement results with the UBC3D-PLM Sand model is an effective stress
measured ones. elasto -plastic model which can simulate the lique
faction behavior of sands and silty sands under seis
mic loading. (Tsegaye (2010), Petalas & Galavi
3.3 Liquefaction analysis using CLiq (2012)).
The effectiveness of liquefaction mitigation under the
Citadel & Entry structure was corroborated by carry
ing out a post testing campaign as shown in Figure 4.
A pair of post treatment CPTUs were performed at 1/
3 the distance between two Vibro-compaction points
and at the centre of the triangle formed by three
vibro-compaction points. Evaluation of liquefaction
potential was assessed using CLiq software developed
by Geologismiki using Robertson (NCEER R&W
1998, 2009). Each post-CPT pair was checked for
liquefaction susceptibility by considering an earth
quake magnitude of 6.0, a PGA of 0.15g at bedrock
and a Factor of Safety against liquefaction of 1.25.
The results of the liquefaction analysis for all post-
CPTs conducted in the area are shown in Figure 5.
The results confirm that liquefaction is mitigated by
the vibro-compaction technique with a minimum
factor of safety of 1.25 except for thin confined local
ized layers with thickness of approximately 0.30m.
These thin layers are classified by Robertson (2011)
as transition zones and are not susceptible to liquefac
tion. The liquefaction analyses resulted in minimum
seismic vertical and horizontal settlements.
3.4 Liquefaction analysis using PLAXIS 3D

The above results were checked by carrying out
liquefaction analysis using PLAXIS by adopting the Figure 6. Earthquake parameters for PLAXIS liquefaction
UBC3D – PLM Sand Model and by defining the analysis.
934
The Factor of Safety against liquefaction using
PLAXIS was calculated as the ratio of the excessive
pore pressure changes and the initial effective pres
sure. Plaxis liquefaction analysis resulted in a FoS
of 1.51 which is greater than the empirical factor of
safety calculated using CLiq software thus confirm
ing that the liquefaction mitigation criteria were
achieved. The advantages of the dynamic analysis
using UBC Sand is the ability to simulate the
excess pore water pressure generation in the soil, as
well as the ability of checking the amplification,
liquefaction/plastic points.
3.5 Settlement analysis using Taspie

The settlement analyses of the post-CPTs indicate that
the settlement criteria were not achieved by Vibro Figure 7. Load transfer mechanism (Taspie Software).
compaction only due to the resultant settlement
exceeding 25mm. Therefore, it was decided to install
The above figure reveals that the settlement cri
450mm diameter rigid inclusions, 10.0m in length in
teria was achieved with an associated settlement of
the vibro-compacted ground, to ensure the calculated
approximately 25mm. The rigid inclusion settlement
settlement be less than the allowable (25mm) under
was approximately 14mm with a maximum stress of
125kPa loading. The rigid inclusions were overlain
4.55Mpa located at the neutral axis (-2.80m MSL).
with a 40cm thick load transfer platform (LTP). The
LTP plays an indispensable role in transferring the
loads to the toe of the rigid inclusions and to uniformly 3.6 Settlement analysis using PLAXIS 3D
transfer the settlement under the structure. The post-
To validate the estimated settlements by Taspie soft
CPTs were used to derive the geotechnical parameters
ware, a full PLAXIS 3D finite element analysis was
of the in-situ soil. As for the LTP, a compacted gravel
conducted for Citadel & Entry structure as shown in
layer was considered. Layering and material param
Figure 8 below. The rigid inclusions were defined as
eters of the LTP and foundation soils are listed in
embedded beams and the slab on grade as plate
Table 1. The Ground level elevation is at 3.0m MSL
elements. Soil properties illustrated in Table 1 were
and the ground water level is at 0.00m MSL.
defined using the Mohr-Coulomb material model.
Table 1. Material parameters and foundation soils.
Thickness γ E φ
Layer # (m) (kN/m3) (Mpa) (Deg) ν
LTP (Gravel) 0.5 19.50 50.00 38 0.25

MD Sand 1 0.58 18.00 37.50 33 0.35
MD Sand 2 0.62 18.00 53.40 33 0.35
MD Sand 3 1.00 18.00 27.00 33 0.35
Silt 0.44 16.50 6.50 26 0.35
MD Sand 4 0.94 17.50 36.70 33 0.35
VD Sand 0.42 19.00 76.00 34 0.35
Figure 8. PLAXIS 3D View (Citadel & Entry).
* γ: unit weight, ν: Poisson’s ratio, E: elastic modulus, φ’:
drained friction angle, MD: Medium dense, VD: Very dense The above PLAXIS 3D analysis has shown
a maximum settlement of 19mm as depicted in
Figure 9. Moreover, the same load transfer mechanism
The RI design is performed using Taspie by shown in Figure 7 using the Taspie software was also
adopting a unit cell model. This software follows the created by the PLAXIS 3D run as demonstrated in
general recommendations for the design and con Figure 10.
struction of rigid inclusions in ASIRI (2012).
Figure 7 displays the resultant settlement along
3.7 Measured vs theoretical results
with the stresses exerted on both soil and rigid inclu
sions for one of the post-CPTs pair conducted in the According to Chapter 8 of ASIRI guidelines, static
Citadel & Entry area. load tests on an isolated inclusion in compression
935
Table 2. Measured versus theoretical comparison.
RI (Settle RI (Settle RI (Settle

Testing ment- ment- ment-
Static Load Taspie) PLAXIS Measured)
Load Test (kN) (mm) 3D) (mm) (mm)
STA-07 1273 15.50 13.77 3.960

STA-08 1116 19.50 14.11 3.880
STA-03 1222 16.00 14.80 8.060
Table 2 shows that the theoretical settlements

derived from both Taspie and PLAXIS 3D analyses
Figure 9. Settlement distribution. are conservative compared to the measured
settlements.
3.8 Conclusion
The CPT testing technique has proven to be an
invaluable tool for calibrating and verifying the
effectiveness of both Vibro-compaction and rigid
inclusions techniques, particularly in its ability to
identify the strength of the silt layers. The CPT test
ing used in both the PLAXIS 3D and Taspie soft
ware runs have resulted in conservative settlement
calculations which favourably compare with the
measured static load test settlements. Moreover,
CPT testing has proven to be well correlated with
the liquefaction acceptability criteria as calculated
by carrying both static and dynamic liquefaction
analyses using both CLiq and PLAXIS software
respectively.
REFERENCES
ASIRI 2012. Recommendations for design, construction,
and control of foundation over soils reinforced by rigid
inclusions, Presses des Ponts.
Boulanger, R. W., and Idriss, I.M. (2005). New criteria for
distinguishing between silts and clays that are suscep
tible to liquefaction versus cyclic failure. Proc., Tech
nologies to Enhance Dam Safety and the Environment,
Figure 10. Load Transfer Mechanism (PLAXIS 3D). 25th Annual United States Society on Dams Conf.,
USSD, Denver, 357–366.
Boulanger, R. W., Meyers, M.W., Mejia, L.H., and Idriss, I.
should be conducted. Three static load tests were con M. (1998). Behavior of a fine-grained soil during Loma
ducted in the Citadel & Entry area as shown in Prieta earthquake. Can. Geotech. J., 35 (1), 146–158.
Figure 4 by applying a testing load equivalent to the Hor B. et al. 2015. A 3D FEM analysis on the performance
sum of the head load acting on the rigid inclusion and of disconnected piled raft foundation. Japanese Geotech
the negative skin friction acting on the RI from its nical Society Special Publication 2 No.34, 1238–1243.
head to its neutral plane. Hor B. et al. 2017. Ground improvement using rigid inclu
The effectiveness of rigid inclusions in the field sion for the foundation of LNG tanks. Proceedings of
response of Citadel/Entry foundation was assessed by the 19th International Conference on Soil Mechanics and
Geotechnical Engineering, Seoul 2017.
comparing the measured settlements with the theoret Kirsch, K. (1985). Over 50 years of deep vibratory compac
ical calculations using both Taspie and PLAXIS 3D tion: Milestones of German geotechnique. Geotechnik,
software. Special Issue, Deutsche Gesellschaft fur Erd -und
The following table enumerates the resultant Grundbau, Essen, Germany.
settlement of the static tests compared with the the Kirsch, F. and Sondermann, W. (2001). Ground improve
oretical calculations. ment and its numerical analysis. In proceedings of the
936
XCth ICMFE, Istanbul, Turkey. A.A. Balkema, Rotter Schmertmann, J.H. (1978). Guidelines for cone penetration
dam, the Netherlands. test, performance, and design. Report FHWA-TS-78
NCEER (1997). Proceedings of the NCEER Workshop on 209,145. US Federal Highway Administration, Wash
Evaluation of Liquefaction Resistance. Technical Report ington, DC.
NCEER-97-0022. Tsegaye, A. (2010). Plaxis liquefaction model. external
Petalas, A., Galavi. V. (2012). Plaxis liquefaction model report. PLAXIS knowledge base.
ubc3d-plm. PLAXIS knowledge base. West, J.M. (1976). The role of Ground Improvement in
Robertson et al. (1983). Interpretation of cone penetration Foundation Engineering: Ground Treatment by Deep
tests. Canadian Geotechnical Journal, 20 (4), 718–733. Compaction. The Institution of Civil Engineers,
Robertson, P.K. (1990). Soil Classification using CPT. Can London, UK.
adian Geotechnical Journal, 27(1),151–158. Wehr, J. and Sondermann, W. (2013). Deep vibro tech
Robertson, P.K. (2011). Automatic software detection of niques. In Kirsch, K. and Bell, A. (eds.) Ground
CPT transition zones.” Geotechnical news, June, 2011. Improvement. CRC Press, Taylor & Francis Group,
Seed, H.B., and Idriss, I.M. (1982). Ground motions and Boca Raton, FL.
soil liquefaction during earthquakes, Earthquake Engin Yoshimi, Y. (1980). Protection of structures from soil lique
eering Research Institute, Berkeley, Calif. faction hazards. Geotechnical Engineering, 11.
937
© 2022 the Author(s), ISBN 978-1-032-31259-0
A comprehensive design procedure for pile groups in liquefiable soils

Teleios Srl – Società di Ingegneria, Castel Maggiore, Bologna, Italy
ABSTRACT: Among the various strategies to mitigate the effects of soil liquefaction one of the most effect
ive is the design of deep foundations. The designs of piles in liquefiable soils are too often concerned with the
only axial bearing capacity, addressed by simply neglecting pile resistance in the liquefiable layer. This
approach is inadequate to properly face the complexity of the problem. In the present paper we intend to
examine this theme throughout its multiple aspects. We have synthesized a design procedure of analysis that
has been already applied in some projects in different areas of Emilia-Romagna region characterized by high
risk of liquefaction. The proposed design procedure is based on the most updated theories and design refer
ences concerning piles in liquefiable soil, such as those of Cubrinovsky, Olson & Stark, Rollins, Bhattacharya,
Madabhushi and others, primarily referring to CPTu.
1 INTRODUCTION the constitutive laws that mathematically represent

the mechanical response of the soil.
When designing a structure interacting with the soil In the case of high risk of liquefaction, the ge
in a site subjected to a high risk of liquefaction otechnical model valid under seismic conditions sig
during earthquake, there are many solutions that can nificantly differs from the static one because of liq
be evaluated to reduce and mitigate this risk. uefaction effects on soil properties. The liquefiable soil
For foundations resting on liquefiable soil one of layer is described below in terms of both modified
the most effective solutions is the use of piles. But, stiffness and strength parameters.
in this case, their design has to properly account for
the effects induced on piles by liquefaction.
2.1 Stiffness parameters
The phenomena involving a pile under seismic
cyclic loading in presence of liquefiable soils are As regards stiffness parameters we referred to the
complicated. There are different important aspect to theory proposed by Cubrinovski et al. (2009). In
be accounted for: the change in shaft resistance, the a simplified 3-layers model, in which the central one
reduction of base capacity also in deeper layer not is potentially liquefiable, the pile is modelled as
directly liquefacting, the equilibrium instability due a beam connected to a series of springs representing
to the loss of lateral support from liquefied soil and the lateral stiffness of the soil.
the modification of the geotechnical model under As expected, the stiffness offered by the liquefied
liquefaction, directly conditioning the analyses of soil (k2) is significantly lower than the one of the
pile groups with geotechnical numerical models. same non-liquefied soil (k1). The results observed in
In order to properly face the design of deep foun full-scale tests on piles show that the stiffness deg
dations on liquefiable soils we have resumed the radation factor β2 = k1/k2 typically varies in a range
principal bibliographic studies with the aim to build of 1/50 ÷1/10 for cyclic liquefaction (Figure 1). In
a comprehensive design procedure. In the paper we our design procedure we chose to refer to the lower
will describe in detail the various step of this proced bound, i.e. to β2 = 1/50.
ure and we will also propose a real design example The degradation factor β2 is applied to the stiff
in which we have already applied this process. ness parameters of the liquefiable soil layer, in par
ticular to the initial tangent value of soil elastic
modulus within the numerical Boundary Element
2 GEOTECHNICAL MODEL Method analyses perfomed adopting a non-linear
hyperbolic constitutive model. The above theory
In a geotechnical design the first, and maybe most (Cubrinovski et al. 2009) was originally developed
important, step is the construction of an accurate and for “p-δ” curves methods: we extended the same
reliable geotechnical model, i.e the parameters and approach to numerical BEM analyses.
DOI: 10.1201/9781003308829-140
938
When liquefaction occurs, the degree of growth
of the pore pressure can be described by the intersti
tial pressure ratio ru:
According to Equation 2, soil shows liquefaction

when ru = 1. Liquefaction can easily occur in superfi
cial sandy layers. Usually, indeed, deeper coarse
grained soil layers show a higher resistance to lique
faction because their higher density and also because
the greater effective pressure. So, when deep founda
tions are designed to resist to liquefaction, it can
occurs that pile toes are placed in non-liquefiable
deeper sandy layer. But also in these deeper layers
Figure 1. Degradation of stiffness in the liquefied layer pore water pressure can increase, so ru > 0 and so
observed in full-size test on piles. From Cubrinovski et al. soil resistance decrease. This fact, not immediately
(2009). perceivable, has to be properly taken into account in
designing piles.
2.2 Strength parameters Starting from the model for the base resistance of
a pile (Vesic 1972) and considering the effects due to
The liquefiable soil is described in terms of strength the increase of pore water pressure, Knappet &
parameters, in seismic conditions, via its residual Madabhushi (2008b) showed how the tip resistance
undrained strength su(LIQ), as proposed by Olson & of piles in liquefiable soil is related to its correspond
Stark (2002). ing value in static conditions by the relation reported
in Equation (3):
The formulation proposed in Equation 1 is valid Where ru has been defined in Equation 2, ϕ is the
for CPT tests. As in can be seen, it is structured to angle of shearing resistance of the coarse grained
define an interval of values with an amplitude of soil in which the pile toe is placed while Qbase,E and
0.03 MPa. Many researchers have studied post- Qbase,S are the tip bearing capacity of the pile,
seismic conditions: residual undrained strength was respectively, in seismic and static conditions.
evaluated on the basis of the deformed configuration From Equation 3 it can be seen how the resistance
of the foundation after the earthquake. is related to ru but this factor is not simple to evalu
From Bowen & Cubrinovski (2008) it can be seen ate. It would be necessary to perform advanced site
how, in case of cyclic liquefaction, it is safe to refer effect analyses, accounting also for the liquefaction
to the mean value of the interval. That is what we of the soil. This cannot be done in ordinary practice,
chose to apply in the design procedure. so a simplified approach has been developed.
Given a liquefiable soil layer, in which liquefac
tion occurs (ru = 1) it is assumed that excess pore
3 PILE AXIAL CAPACITY pressure remain constant for depth greater than its
Once defined the geotechnical model, next step con

sists in evaluating the pile axial capacity. Many dif
ferent approaches can be used: correlations with
strength parameters of the soil, direct correlations
with in-situ soundings (CPT & CPTu, DMT, SPT,
etc.), pile load tests. As concern the design proced
ure, we have focused on direct methods based on
CPTu: in particular, the method proposed by Eslami
& Fellenius (1997) improved by Niazi (2013).
3.1 Base capacity

The axial capacity of piles in liquefiable soils is well Figure 2. Effective stress conditions around piles for full
illustrated by Madabhushi et al. (2009). liquefaction until zL depth. From Madabhushi et al. (2009).
939
thickness zL. Considering the growth of effective Anyway, experimental tests (Knappet 2006) have
stresses with depth as almost linear, this lead to shown how critical loads are not so small as those
a bilinear shape of the factor ru. Once the trend of ru predicted by theoretical formulations. This evidence
and the geometry of the pile are known it is easy to has been explained with the fact that liquefied soil
evaluate the interstitial pressure ratio at pile tip level has a small, but not null, stiffness neglected in the
Lp. It can be derived (see Figure 2). theoretical model. So Eulero’s expression has been
updated by Madabhushi et al. (2009), accounting for
finite, small, stiffness of the liquefied layer by the
factor ru,base, as reported in Equation 7:
3.2 Shaft capacity

Madabhushi et al. (2009) have extended their studies
also to the shaft bearing resistance of a pile, obtain Where E and I are elastic modulus and moment of
ing the formulation reported in Equation 5, that fol inertia of the section of the pile, h is the length of the
lows a philosophy similar to Equation 3. pile crossing the liquefiable layer and β is a factor
accounting for the fixity at both ends of the pile.
(β∙h) represent the equivalent length of the unsup
ported pile tract. Introducing the radius of gyration
of the section rg, the slenderness ratio λ is defined:
Where Qshaft,E and Qshaft,S are the shaft bearing cap
acity of the pile evaluated, respectively, in seismic
and in static conditions.
So, in a more general way, it is possible to define
the total bearing capacity of a pile in a liquefiable
soil Pbc from Equation 6. Bhattacharya & Lombardi (2012), basing on data
collected from real cases, have defined an admissible
domain, bounded by a slenderness ratio λ = 50
(dashed black line in Figure 3), separating deep
foundations that have shown good performances
from those who have not. From Figure 3 it can be
seen how the choice of the limit value of slenderness
According to Equations 5 and 6, the shaft λ = 50 can be excessively precautionary. For that
resistance of a pile in a liquefiable soil can be reason we propose to refer to a limit value equal to
significantly reduced when the factor ru increase, λ = 75 (continuous red line in Figure 3).
but it is not strictly equal to zero. Also other In reality piles can suffer buckling for axial load
researches (Rollins 2015), basing on experimental lower than the one predicted by Equations 7 also
data, have demonstrated that the contribution of because imperfections not accounted in theoretical
the liquefiable layer to shaft resistance is not models, inducing geometrical second order problem
null, but can decrease to about 50% of its corres (P-δ effect). Named δ0 the displacement induced by
ponding static value. Anyway, in the present the horizontal action (earthquake), the total lateral
design procedure we propose to completely neg
lect the contribution of the liquefiable soil layers
to the shaft resistance of the piles.
4 PILE INSTABILITY
Liquefaction causes the loss of lateral support of the

soil to the piles and the subsequently significant vari
ation of lateral stiffness of the foundation. In these
conditions piles can face a crisis for buckling.
Pile instability due to liquefaction has been stud
ied by Bhattacharya (2003) and Bhattacharya et al.
(2004) considering a simplified model in which
liquefied soil has no strength and stiffness. The pile
crossing this layer will behave like an axially
loaded column. Under these hypotheses the critical
load Pcr corresponding to the loss of the elastic Figure 3. Admissible domain: measured performances of
equilibrium is given by the well-known Eulero’s deep foundations after real earthquakes. From Bhattacharya
expression. & Lombardi (2012).
940
displacement δ in presence of a vertical action P will 5 CALCULATION METHODS
be greater. Equation 9 describes the normalized
trend of the amplification factor δ/ δ0. Designing a deep foundation on liquefiable soil is
complex and, as seen in the previous sections, many
aspects have to be properly considered to correctly
face the problem. So it is clear how also the adoption
of a correct method to analyze the pile group is fun
damental in order to not nullify the benefits of the
The normalized movement δ/ δ0 is almost linear adopted advanced theories, as previously reported.
for values of vertical load ψ = P/Pcr ≈ 0.30 so it can We used numerical BEM analyses with the soft
be assumed: ware Repute, developed by Geocentric (Bond &
Basile 2010, Basile 1999), adopting for the soil a non
linear hyperbolic constitutive model. In that way pile
group effects and other phenomena, as pile plasticiza
tion and shadowing, can be properly accounted.
The procedures described in the previous section
As suggested by Bhattacharya & Lombardi mainly refers to single piles. Instead we propose to
(2012), and also as considered in the proposed extend these analyses to pile groups by adopting cor
design procedure, it is adopted ψ = 0.35 as limit. rect calculation methods, such as the one mentioned
According to Equations 6 and 7 both the total above, in addition to the definition of a correct
axial bearing capacity of the pile Pbc and its critical geotechnical model, accounting for liquefaction (see
load Pcr depend from factor ru. So critical values of section 2).
ru can be derived, corresponding to reaching these
two limiting conditions.
Following this approach Madabhushi et al. (2009) 6 DESIGN EXAMPLE
describe a series of graphs in which, depending on
pile geometry (diameter D0 and elastic properties E, The design procedure that we propose to design
I) and required factor of safety FOS, the domain of a piled foundation on liquefiable soil has been
use of the pile is defined (see Figure 4). These described in detail by theoretical point of view in the
graphs can be derived for the specific design case previous sections of the paper. Now we want to
considering the liquefaction conditions of the site, describe how this procedure has already been applied
and can be enriched by also plotting the hyperbolic is some real case in which the authors designed deep
trend of ru,base, in that way also accounting for the foundations to mitigate the liquefaction risk.
thickness and depth of the liquefiable layer. The two major works in which we adopted our
The piles have to be anyway verified against their design procedure are the enlargement and seismic
structural strength because of the stresses (bending retrofit of the Cento (Ferrara, Italy) sports hall and the
moment and shear) due by the inertial and kinematic construction of the new Pavillion 37 as part of the
interaction with the superstructure. revamping project of the fair quarter of Bologna
(Italy). Due to the space available in the paper, we will
only describe the first of these two projects.
6.1 Investigation campaign and geotechnical

characterization
A first investigation campaign, consisting in four cone
penetration tests with piezocone CPTu and seismic
piezocone SCPTu, two flat dilatometer DMT tests, geo
physical tests and laboratory ones, has been firstly car
ried out. From these data an high risk of liquefaction
was found out so it was decided to perform an integra
tive campaign to better go insight the liquefaction prob
lem. Four CPTu and dynamic laboratory tests (resonant
column RC and cyclic triaxial TX CYC), on samples
taken from two boreholes, have been performed.
Liquefaction risk has been assessed with different
Figure 4. Design chart for a solid circular RC pile in loose
approaches: from CPTu (Robertson 2009), from
sand (Dr= 35%): D0= 0.5m, FOS= 2. From Madabhushi DMT (Monaco et al. 2005), from dynamic lab tests.
et al. (2009). In the blank areas the pile is verified, in those The liquefaction potential index LPI, evaluated
identified by “S” the pile encounter a crisis by bearing cap adopting the Sonmez (2003) approach, results ranging
acity while in those identified by “I” the crisis is because of from the various investigations from about 8 to 18,
instability. identifying an high risk of liquefaction.
941
Liquefaction is induced by a layer of loose sands
that extends from about 1 to 6 meters of depth from
ground level. The geotechnical model in seismic
condition accounting for liquefaction, reported in
Table 1, has been defined according to what has
been described in section 2.
Table 1. Geotechnical model in seismic conditions,

accounting for liquefaction.
Undrained Frinction Tangent

strength Su angle modulus
Layer From To [m] [kPa] ϕ [°] E0 [MPa]
Liquefi
able sand 1.0 6.0 1 - 2
Clay 6.0 11.0 30 - 90
Clay 11.0 18.0 75 - 105
Figure 5. Foundation plan of the sport hall. In red the two
Sand 18.0 26.0 - 32 135
seismic cores are highlighted.
Clay 26.0 - 75 - 150
defined the geotechnical model valid in seismic con

ditions under liquefaction, the pile ultimate axial
6.2 Description of the intervention
capacity have been defined.
The project primarily consists in the construction of Pile base capacity in static conditions has been
a new grandstand: this element, as enlargement of evaluated adopting the formulation proposed by
the existing sport hall, contains two reinforced con Berezantsev (1965), because pile tip in placed in the
crete cores that are the principal part of the bracing lower sandy layer, finding out a value of Qbase,S =
system adopted to increase seismic resistance of the 1190 kN. Considering the development of excess
whole structure. These cores absorb the great pore pressures also in the lower sandy layer, it does
amount of the seismic forces because they are not reach liquefaction but the tip resistance decrease.
designed to support and to brace the roof of the According to Equation 3 a ratio of 0.86 between the
hall. Since there where no limits on this side for seismic and static value of base resistance has been
operating machines, we chose to adopt FDP (Full evaluated, corresponding to a value of 0.238 for the
Displacement Piles) piles: due to the particular factor ru,base (Eq. 4). This means that in seismic con
shape of the drilling tip the pile is realized without ditions the pile base resistance decrease to:
removing soil, so also a positive densifying effect is
Qbase,E = 0.86 ∙ 1190 ≈ 1000 kN (11)
induced.
In the opposite side of the hall there is the existing Pile shaft capacity has been calculated via a direct
tribune, supported by an RC frame. Also this zone correlation with data from CPTu adopting the
has been involved in the seismic retrofitting, and the method of Eslami & Fellenius (1997) improved by
existing shallow foundation have been reinforced Niazi (2013). We referred to the 6 CPTu performed
with micropiles designed according to the proposed on site, finding out values of Qshaft,S ranging from
procedure. Here they will not be discussed. 1390 kN to 1778 kN, with a mean value of 1600 kN.
As described in § 3.2, in our proposed design pro
cedure the shaft capacity is completely neglected in
6.3 Foundation analysis
the liquefiable layer. Adopting the same calculation
Given the loads acting on foundation from super procedure of the static case, in seismic conditions we
structures and defined the geotechnical properties found out values of Qshaft,E ranging from 1070 kN to
of the soil, each foundation has been analyzed 1448 kN, with a mean value of 1263 kN.
adopting a numerical approach: BEM analyses have The BEM analyses of the piled foundation of the
been carried out with the software Repute (see sec core (Figure 6) give axial forces on piles as reported
tion 5). in Table 2. According to the design criteria defined
For each seismic core a deep foundation consist by the Italian code NTC pile axial capacity check is
ing in 24 FDP piles, diameter 600mm, with a length satisfied in static condition with a maximum exploit
of 22.0 meters (depth of pile tip from g.l.) has been ation of 81% and in seismic condition at 92%.
adopted. The pile cap have dimensions 12.10 × 9.20 The last step of the design involve the check of piles
meters and is 1.2m thick. The overall foundation against instability. It can be considered that FDP piles
plan is reported in Figure 5. are rigidly connected at the top to the pile cap. The
Following the design procedure described in the same fixity can be considered at the base, because the
previous sections of the paper, after have being liquefiable sandy layer is near the ground and the piles
942
Table 2. Maximum axial forces on piles from BEM REFERENCES
analyses.
Basile, F. 1999. Non-linear analysis of pile groups. Proc. .n
Action u.m. Static Seismic of Civil Eng., Geotech. Eng., Vol. 137, No. 2, April,
pp 105-115.
Maximum compressive force [kN] 1047 964 Berezantsev, V. G. 1965. Design of deep foundations.
Maximum tensile force [kN] - 15 Bhattacharya, S. 2003. Pile instability during earthquake
Maximum horizontal shear force [kN] 81 79 liquefaction. PhD thesis, University of Cambridge, UK.
Maximum vertical settlement [mm] 6.9 5.9 Bhattacharya, S., Madabhushi, S. P. G., Bolton, M. D.
2004. An alternative mechanism of pile failure during
Maximum horizontal displacement [mm] 0.5 11.2
seismic liquefaction. Geotechnique 54 (3): 203-213.
Bhattacharya S., Lombardi D. 2012. Sul comportamento
sismico delle fondazioni su pali in terreni liquefacibili.
Rivista Italiana Di Geotecnica 1/2012.
Bond, A. J., Basile, F. 2010. Repute 2.0, Software for pile
design and analysis. Reference Manual, Geocentrix Ltd,
United Kingdom, 49p.
Bowen, H. J., Cubrinovski, M. 2008. Pseudo-static analysis
of piles in liquefiable soils: parametric evaluation of
liquefied layer properties. Bulletin of the New Zealand
Society for Earthquake Engineering, Vol. 41, No. 4,
December 2008.
Cubrinovski, M., Ishihara, K. & Poulos, H. 2009. Pseudo-
static analysis of piles subjected to lateral spreading.
Figure 6. Repute calculation model of the piled foundation
Special Issue Bulletin of NZ Society for Earthquake
of the core.
Engineering.
De Alba, P. A. 1983. Pile settlement in liquefying sand
continue below in the non-liquefiable clayey layers. deposit. J. Geotech. Enggn. 109 (9): 1165-1179.
Eslami, A., Fellenius, B.H. 1997. Pile capacity by direct
Nevertheless, as a precaution, we assumed the equiva
CPT and CPTu methods applied to 102 case histories.
lent length (β∙h) as twice the thickness of the liquefiable Can. Geotech. J. 34 (6): 886–904.
layer (equal to 5.0m, see Table 1). Considering the elas Knappet, J. A. 2006. Piled foundations in liquefiable soils:
tic properties of the cross-section of the piles (E∙I = accounting for axial loads. PhD thesis, University of
159.04 MNm2), the interstitial pressure ratio (ru,base = Cambridge, UK.
0.238) and limiting the critical load by the factor ψ = Knappet, J. A., Madabhushi, S. P. G. 2008a. Liquefaction
0.35 (see Eq. 10) the ultimate axial load can be evalu indiced settlement of pile groups in liquefiable and lat
ated as Pult ≈ 96900 kN, more than 100 times the max erally spreading soils. J. Geotech. Enggn. 134(11).
imum axial force acting on piles under seismic loading Knappet, J. A., Madabhushi, S. P. G. 2008b. Designing
against pile tip bearing capacity failure in liquefiable
(see Table 2). The slenderness ratio λ = 67 is lower
soils. Proc. 2nd BGA Int. conf. on foundations 2: 1237
than the assumed upper admissible limit of 75. 1246.
The piled foundations, designed according to the Madabhushi, G., Knappett, J., Haigh, S. 2009. Design of
proposed procedure, satisfy all the safety criteria. pile foundations in liquefiable soils. London Imperial
College Press.
Monaco, P., Marchetti, S., Totani, G. & Calabrese, M.
7 CONCLUSIONS 2005. Sand liquefiability assessment by Flat Dilatometer
Test (DMT). Proc. XVI ICSMGE, Osaka, 4, 2693-2697.
In the paper a design procedure to correctly deal Niazi, F. S. 2013. An Update on Pile-CPTu Direct Correl
with the design of pile groups on liquefiable soil has ations. Ports 2013, ASCE.
been presented. The procedure provides guidance as Olson, S. M.& Stark, T. D. 2002. Liquefied strength ratio
from liquefaction flow failure case histories. Can. Geo-
concern the geotechnical model (modified strength
tech. J. 39, 629–647.
and stiffness parameters under liquefaction), the pile Robertson, P. K. 2009. Interpretation of cone penetration
axial capacity (shaft and base components) and the tests – a unified approach. Canadian Geotechnical Jour
pile instability. Each step of the procedure is based nal, 46: 1337-1355.
on specific theories from various authors, in order to Rollins, K.M., Hollenbaugh, J.E. 2015. Liquefaction
base the design on solid scientific bases. Induced Negative Skin Friction from Blast-induced
Cone penetration tests are fundamental in many Liquefaction Tests with Auger-cast Piles. 6th Int. Conf.
steps of the procedure: just think to the geotechnical on Earth. Geotech. Enging. Christchurch, New Zealand,
characterization and the pile capacity evaluation. 1-4 Nov. 2015.
Sonmez, H. 2003. Modification to the liquefaction potential
The procedure is intended as a guide for designers
index and liquefaction susceptibility mapping for a
who have to face the design of a deep foundation on liquefaction-prone area (Inegol-Turkey). Environ. Geol
liquefiable soil. This is a very sensitive problem, so ogy 44 (7), pp. 862–871.
we hope the guide can help geotechnical engineers Vesic, A. S. 1972. Expansion of cavities infinite sol mass.
in this challenging work. J. Geotech. Engng. 98 (3): 265-298.
943
© 2022 the Author(s), ISBN 978-1-032-31259-0
Deep foundations of the new Pavilion 37 – fair quarter Bologna

Teleios Srl – Società di Ingegneria, Castel Maggiore, Bologna, Italy
Vincenzo Colella
Geofondazioni Ingegneria e Lavori Srl, Martellago, Venezia, Italy
Giampaolo Cortellazzo
Università di Padova, Padova, Italy
Andrea Dei Svaldi

Desam Ingegneria e Ambiente s.r.l., Mogliano Veneto, Treviso, Italy
ABSTRACT: The present paper concerns the design of driven precast tapered piles, that have been used
within the project of the new Pavillion 37 of the fair quarter of Bologna. A particular focus will be reserved to
the evaluation of the axial bearing capacity of these elements and to the prevision of their load-settlement
curve. The geotechnical characterization, as well as the prediction of the axial resistance of the piles, has been
mainly based on Cone Penetration Test with piezocone CPTu. The previsions will be compared with the
results of a series of static load tests performed during the construction.
1 INTRODUCTION deep foundation was basically justified by the high

loads transmitted by the roof, together with the geo
The present paper concerns the design of deep foun technical conditions of the site.
dation for a new exhibition hall, especially as regard The whole project concern also two other adjacent
the prediction of the axial bearing capacity of the buildings (the so-called “Mall” and “East entrance”),
piles and of their load-settlement curve. not object of the present paper, whose foundations
The framework of the project and the geometry have been designed similarly.
will be described first, following with the description
of the geotechnical campaign and the soil character
ization. Then we will assess the project prediction by
comparison with the results of the static load tests
performed on-site during construction.
All these evaluations are based on CPTu test, as
described in the following.
2 PROJECT DESCRIPTION
Within the revamping project of the fair quarter of

Bologna, the new Pavilion 37 rises as flagship of the
exhibition centre with its remarkable opening roof
and its planimetric dimensions of 80 x 184 meters
(Figure 1). The steel structure of the roof stands on 8
reinforced concrete towers, whose foundations repre
sent the most important and complex geotechnical
elements of the whole project.
The choice was to adopt deep foundation for all
the RC towers, consisting in driven precast tapered Figure 1. Typical deep foundation of the RC towers sup
piles arranged in groups (Figure 2). The adoption of porting the roof of the pavilion.
DOI: 10.1201/9781003308829-141
944
3 PILE DESCRIPTION The geotechnical characterization is mainly based
on the interpretation of the performed CPTu. They
Driven precast tapered piles were used in the project. are very effective in recognizing soil behavior and
The choice was due to different aspects, both geo strength and stiffness parameters. In Figures 3, 4 and
technical and construction-related: 5 some of the principal results from the interpret
ations are showed.
– Driven piles have good geotechnical perform
ances. In addition, because of the installation pro
cess, an increase in density is induced in the soil.
– They are rapid to install and it is possible to con
trol the performance of piles during driving.
– In soft soils or in presence of water neither pre
drilling nor hole support are required.
– There is no excavation material, so the organiza
tion of the construction site is clean and easier.
The adopted piles have a diameter of 53/26 cm
(head/toe), a taper of 1.5 cm/m and a length of 18
meters. The piles were precasted in factory with
a specific centrifugation process.
Figure 3. Point resistance and soil index vertical profiles

from the interpretation of one of the performed CPTu.
Figure 4. Friction angle and undrained strength vertical pro

files from the interpretation of one of the performed CPTu.
Figure 2. Typical tapered precast pile with length L= 18m.

Vertical sections and reinforcements.
For driving were used hydraulic hammers with

weight of 6 and 7 tons. The driving energy was 36
kNm and the settlement during driving was meas
ured in the range from 5 to 7 mm/blow.
4 GEOTECHNICAL CHARACTERIZATION
The site was characterized by performing an investiga

tion campaign consisting mainly in on-site soundings.
Four cone penetration tests with piezocone CPTu,
a borehole (with soil sampling and SPTs) and geophys Figure 5. Confined and initial shear moduli vertical profiles
ical MASW and HVSR tests have been carried out. from the interpretation of one of the performed CPTu.
945
An alternation of coarse-grained and fine-grained
material was found out, described by the following
geotechnical model (Table 1).
Table 1. Geotechnical model.
Undr. Effective Friction Tangent

From - To strenght cohesion angle φ modulus
Layer [m] cu [kPa] c' [kPa] [°] E0 [MPa]
Sandy
0.0 8.0 - 0 32-34 200-250
silt
Clay 8.0 15.0 60-70 5 27-30 120-170
Sandy
15.0 20.0 - 0 38-40 ≥ 500
gravel
Figure 6. Tapered pile axial capacity evaluation from
Sandy CPTu adopting the Eslami & Fellenius method.
to
20.0 35.0 70-80 5 27-30 300-350
clayey
silt
Table 2. Pile axial ultimate capacity from CPTu.
Soil stiffness was described in terms of the initial tan
Shaft capacity [kN] Base capacity [kN]
gent modulus because the deep foundations have been
analysed adopting a numerical BEM (Boundary Element Eslami&Fellenius Togliani
Method) approach with the software Repute, developed
by Geocentrix (Bond & Basile 2010, Basile 1999).
CPTu Taper Taper
sounding Shaft benefit Shaft benefit E & F Togliani
5 BEARING CAPACITY OF PILES
CPTu 1 1322 794 1306 843 340 385
In the design phase, pile axial capacity has been pre CPTu 2 1593 956 1536 1180 367 416
dicted adopting calculation methods directly correlated CPTu 3 1502 1201 1527 1354 394 441
to CPTu. CPTu 4 1364 1228 1556 1480 344 378
Then, during construction, pile capacity has also
been evaluated by interpreting the measured load- acceptance testing, under a maximum vertical load
settlement curves derived from the pile load tests, in approximately equal to 1.5 time the exercise load
order to validate the design predictions. (Figure 7). The results obtained from acceptance test
ing are well aligned to the design prediction from
5.1 Analytical evaluation of pile capacity direct correlations with CPTu, considering also that
these load tests have not kept to failure but were
We referred to different methods based, directly or intended to verify the correct installation and perform
indirectly, on CPTu tests. The first is the method pro ance of the piles.
posed by Eslami & Fellenius (1997). In this case the Moreover during the construction of a nearby
beneficial effect of tapering has been taken into building two static load tests were performed, on the
account by increasing the shaft capacity by a factor same type of pile, reaching loads of 2000 kN and
that, as from Nordlund (1963) and Horvath (2002), 2400 kN. These last two load tests have been inter
varies between about 1.5 for clayey-like soil layer to preted according to the approach proposed by Chin
2.0 for sandy ones. An example is reported in Figure 6. (1970) to evaluate the ultimate capacity of the tested
The second method is the one proposed by pile, in Figure 8 an example of the interpretation is
Togliani (2010). This method, in particular, is specific proposed.
for tapered piles and is able to directly evaluate both
the shaft capacity of the pile, related to its side area,
and the component related to the tapering benefit. 6 LOAD-SETTLEMENT CURVES PREVISION
The axial ultimate capacity of the tapered piles,
predicted applying the previous methods, is reported The load-settlement curves, representing the
in Table 2, the values are referred to a limit settle response of the single pile, have been predicted
ment of about 10 mm. adopting an analytical model (Randolph & Wroth
1978), two numerical approaches by BEM analysis
5.2 Load tests interpretation with the software Repute and by FEM analysis with
the software Plaxis and adopting a mixed approach
Four static load tests have been performed on the L= based on the cavity expansion theory (Dei Svaldi
18 m tapered piles during construction, as part of the et al., 2012).
946
a lower stiffness with respect to them, giving to the
foundation design a certain margin of safety.
On the basis of the load-settlement curve for the
single pile, a calibration analysis of the model of the
deep foundation of the RC tower was carried out, by
evaluating with different methods the settlements
under the action of the vertical load (15785 kN). The
foundation consists of a reinforced concrete raft
1.6 m thick resting on 28 tapered piles (Figure 2).
For the calibration has been adopted the following
methods: the formulation proposed by Mandolini
(1994), the BEM method and the traditional multi
layer one-dimensional method, where the pile-soil
Figure 7. Acceptance testing curves.
system was considered as a layer with stiffness
modulus (Table 3) evaluated through an interaction
analysis between pile, plate and soil based on the
concept of efficiency (Fleming et al., 1992). Table 4
reports the results of the calibration analysis.
Table 3. Equivalent stiffness modulus of soil-pile system.
ηw 0.280 efficiency
Kp 525 MN/m single pile stiffness
Kpg 148 MN/m pile group stiffness
Eeq 1108 MPa
Table 4. Settlement of RC tower under vertical load.
Settlement (mm)
Figure 8. Interpretation of the static load test n.1 with the
Chin method. Mandolini
BEM 3.94
All these predictions were based on the param Equivalent stiffness 3.8
eters of the geotechnical model obtained from CPTu
tests (Table 1).
After carrying out the load tests, the measured Once calibrated the BEM method was used to
curves have been compared to the predicted ones, find study the behaviour of the RC tower under the differ
ing out what is graphically represented in Figure 9. ent load combinations in compliance with the Italian
The design prediction does not deviate much from Technical Standards for Buildings (DM 17.01.2028).
to the experimental measures, and generally show An example of settlements and pile reactions
under the action of ultimate state loads, obtained
from BEM analysis, are shown in Figure 10.
Figure 9. Load-settlement curves: comparison between Figure 10. RC tower - settlements and reactions of the
design predictions and load tests. piles obtained from BEM analysis under USL loads.
947
7 CONCLUSIONS Bond, A. J., Basile, F. 2010. Repute 2.0, Software for pile
design and analysis. Reference Manual, Geocentrix Ltd,
The interpretative criteria of the CPTu tests allow to United Kingdom, 49p.
reliably define the stratigraphy and the mechanical Dei Svaldi, A., Mazzucato, A. 2012 – Studio dei pali di tipo
characteristics of the soils at the basis of the founda cilindrico e troncoconico infissi in terreni alluvionali
tion design. Research Report, University IUAV of Venice.
For a complete design of the piled rafts it is not Eslami, A., Fellenius, B.H. 1997. Pile capacity by direct
only important to calculate the bearing capacity of CPT and CPTu methods applied to 102 case histories.
Can. Geotech. J. 34 (6)886–904.
piles but also to define their behaviour through the
Horvath, J. S. 2002. Static Analysis of Axial Capacity of
construction of the load-settlement curve. In this Driven Piles in Coarse-Grain Soil. Integrated Site Char
context, it is possible to use complex numerical acterization and Foundation Analysis Research Project.
models but also simpler methods which, in any case, Report No. CGT-2002-1.
allow to obtain usable results for the design. Mandolini, A. 1994. Cedimenti di fondazioni su pali. Ph.
A final aspect to highlight is that in order to D. Thesis, Department of Geotech. Eng. University of
obtain reliable results in complex analyses with Napoli Federico II.
numerical methods, the calibration by comparing the Nordlund, R.L. 1963. Bearing capacity of piles in cohesion-
results of simple analyses with other calculation less soils. Journal of the Soil Mechanics and founda
methods is relevant. tions division, America society of civil engineers,
New York,N.Y., USA, Vol. 89, No. SM3 , pp. 1–35.
Randolph, M.F., Wroth, C.P. 1978. Analysis of deform
REFERENCES ations of vertically loaded piles. Journal of Geotech
nical Engineering, ASCE , vol. 104, GT12 , 1465–1488.
Basile, F. 1999. Non-linear analysis of pile groups. Pro Togliani, G. 2010. Pile Capacity Prediction using CPT
ceedings of the Institution of Civil Engineers, Geotech Case History. 2nd Int. Symp. on Cone Penetration Test
nical Engineering, Vol. 137, No. 2, April , pp 105–115. ing, Huntington Beach, CA, USA, May 2010.
948
© 2022 the Author(s), ISBN 978-1-032-31259-0
Consolidation settlement of coastal areas of the Emilia-Romagna region

from cone penetration tests
D. Giretti
University of Bergamo, Bergamo, Italy
V. Fioravante
University of Ferrara, Ferrara, Italy
L. Perini & L. Calabrese

Geological, seismic and soil survey of the Emilia-Romagna Region, Bologna, Italy
ABSTRACT: This paper describes an attempt to develop and calibrate a simple tool to forecast the natural
subsidence of coastal Holocene deposits. An empirical correlation has been calibrated between the penetration
resistance measured by a static cone penetrometric tests and i) the normal compression line NCL of fine grained,
coastal deposits ii) the current void ratio and iii) the void ratio on the NCL at the site effective stress. The com
parison between the site void ratio and the normally consolidated void ratio can be used to roughly estimate if
the soil is overconsolidated, normally consolidated or under-consolidated/structured and, in the latter case, if the
fine layer is in the condition of further developing consolidation settlement to reach the self-weight equilibrium.
1 INTRODUCTION A tool has been calibrated which allows to derive

from a qc profile and for fine grained layers two rele
The vulnerability of coastal areas to climate change vant void ratio values: the current site void ratio ec and
and, in particular, to sea level rise, is a cause for the void ratio the soil would have at the site effective
alarm at an international level. The geological survey stress if normally consolidated eNC. The comparison
group of the Emilia-Romagna Region (SGSS), in between these values roughly indicates if the soil is
Italy, has carried out a study to highlight the effect of overconsolidated, normally consolidated or under-
on-shore subsidence on the potential sea transgres consolidated/structured and, in the latter case, allows
sion, and the action of subsidence has resulted to estimate the possible consolidation settlement the
decisive in some areas of the coast, where the altimet layer could still develop to reach the self-weight
ric altitude is close to zero. The origins and possible equilibrium.
evolution of subsidence need to be evaluated to limit The tool has been validated for the pilot sites and
its impact on the risk of inland sea transgression. In used to estimate the expectable natural subsidence of
this context, a study has been carried out to estimate fine grained deposits at selected sites of the coastal
the natural component of subsidence linked to the areas where CPTUs were available.
compaction of recent coastal deposits of fine grained
soils. The research has been based on three main
components: i) two near shore pilot sites, well charac 2 GEOLOGICAL SETTING AND REFERENCE
terized from a geotechnical point of view, where sub SITES
sidence rate is monitored by means of assestimetric
measures; ii) several site where both CPTUs and The study area belongs to the outer sector of the
oedometric tests on fine grained samples were avail Pliocene-Quaternary Apennine foredeep (Ricci et al.
able, iii) a large database of CPTUs carried out in 1986), progressively affected by recent phases of
sites spread all along the Emilia-Romagna coast. compressive deformation of the Apennine chain.
The results of CPTUs and oedometric tests on The Quaternary succession is over 2000 m thick and
Holocene fine grained soils have been used to link records the gradual filling of the highly subsiding
the cone penetration resistance qc to i) the compres basin. The land lowering due to tectonics is still
sion index Cc; ii) the altitude e1 of the active and maximum rates have been estimated
1-D compression line in the e-logσ’v plane defined at about 1 mm/y (Cuffaro et al. 2010). The most recent
σ’v = 1 kPa; iii) the current void ratio ec. succession (Middle Pleistocene-Holocene) is
DOI: 10.1201/9781003308829-142
949
characterized by the sedimentary cyclicity consisting by Holocene deposits of clays, silts and sands. Gorino
of the metric alternation of sandy and highly com stratigraphy is representative of the coastal deposits
pactable fine levels, induced by the glacial-eustatic originated by the Po river typical of the northern
fluctuations of the late Quaternary. The upper 30-40 sector of the regional coasts; Lido di Classe soil pro
metres of this succession, testify the transgressive- file is typical of the southern sector coastal deposits,
regressive Holocene cycle (Amorosi et al. 2003; Ste originated by the activity of Apennine rivers.
fani and Vincenzi 2005,) characterized by paralic Both sites were characterized by means of continu
and marine deposits overlying the Pleistocenic allu ous core drilling and undisturbed sampling, CPTU
vial plain in the lower portion and, in the central and and SCPTU tests, oedometric tests on undisturbed
upper portion, by a thick sedimentary wedge depos sample of fine soil. The boreholes were equipped
ited by the deltaic and beach prograding systems. with settlement gauges, anchored to the Pleistocene
From a lithological point of view, in the littoral deposits deeper than 30 m and cemented to the hole,
sector, the shallowest succession is constituted by for monitoring the subsidence of the deposit over
coastal and delta sands in the upper part and by thick time. The analyzes carried out so far have shown in
prodelta and beach fine deposits in the lower part. The both sites a progressive lowering of about 1 mm/year.
natural compaction of these recent fine deposits may be Gorino site (qc profiles in Figure 2) is character
responsible for the observed subsidence (Teatini et al. ized within the first 5-6 m of depth from the ground
2011), that reaches maximum rates of 12 mm/year. surface by a predominantly sandy layer with frequent
In the study area, sediment deposition during pro- alternations of finer soils, followed by a clayey silty
gradation took place through complex phases of feed stratum up to 27-29 m of depth. This layer, made of
ing, switching and abandonment of the ancient Po delta inorganic and medium to high plasticity fines, resulted
branches; this has generated variable depositional his mainly normally consolidated from oedometric tests
tory even for sediments that are very similar in and is followed by Pleistocene sandy deposits. The
appearance. ground water table is in average 0.5 m deep.
At Lido di Classe, a superficial 8 - 9 m thick layer
of clean sands is followed by a dense alternation of
silty sands, sandy and clayey silts up to about 24 m of
depth, then by silty clays. The fine grained layers, char
acterized by low plasticity, are slightly over consolidate
from the ground surface to 9 m of depth, overconsoli
dated from 9 to 15 m of depth, normally consolidated
at higher depths. The ground water table is 1.7 m deep.
Figure 1. Location of pilot sites and available data: corings

(green dots), CPTUs (blue dots) and selected tests (pink dots).
As part of the studies aimed at evaluating the nat

ural subsidence of the superficial plio-quaternary
deposits of the coastal areas, SGSS has selected two Figure 2. CPTSs at pilot sites, a) Gorino, b) Lido di Clsse.
pilot sites evidenced in Figure 1 at the localities of
Gorino (Ferrara Province) and Lido di Classe
(Ravenna Province), where it is possible to exclude In both sites, oedometric tests have been repeated
the occurrence of subsidence of anthropogenic origin. also on reconstituted specimens. In some cases, the
Both sites are characterized in the first 30 m of depth undisturbed oedometric curve lied above the intrinsic
950
compression line (Burland 1990), indicating pres Despite the significant dispersion in Figures 6 and
ence of structure, possibly due to flocculated depos 7, and acknowledging the rather low correlation
ition occurred in conditions of high sedimentation coefficients, a power function and an exponential
rates in the delta front and proximal prodelta envir function have been drawn to try to link Cc and ec to
onment. These results are consistent with the geo the normalized cone resistance:
logical and depositional history of the sites.
In both sites, the void ratio at the depths of
undisturbed sampling, computed as a function of
the water content and specific gravity and assuming
complete saturation, resulted in some cases higher
than the void ratio derived from the oedometric
curve at the specific vertical stress, indicating still
in progress consolidation phenomena, accordingly
with the estensimetric measures, or presence of
structure.
In addition to the 2 pilot sites, 5 extra sites were
considered in this study, 2 located in the norther
sector of the regional coast at the locality of Porto
Garibaldi, 3 in the southern sector, in the area of the
city of Ravenna (Figure 1), all characterized by simi
lar geological origin and stratigraphy as the two pilot
sites. For all the sites considered, CPTUs and oedo
metric tests on undisturbed samples were available.
Figure 3 reports the grain size fractions and plastic
index as a function of depth measured on the undis
turbed samples overall analyzed.
3 EMPIRICAL CORRELATION
From each oedometric curve available, the compres

sion index Cc (slope of the normal compression line
in the e - logσ’v plane) and the void ratio e1 (altitude
of the normal compression line at σ’v = 1 kPa) have
been derived. Figure 4 shows the Cc measures vs.
depth; in Figure 5 e1 is plotted versus Cc. A linear
relation exists between Cc and e1, which can be
expressed as:
The site void ratio at the sampling depth has been

derived as a function of the measured water content
w and specific gravity Gs as:
where the degree of saturation S has been

assumed equal to 1.
From the CPTU carried out nearest to the borehole
from which the undisturbed samples tested in oed
ometer came, an average qc value was derived at the
depth of sampling. Figures 6 and 7 show the compres
sion index Cc and the site void ratio ec plotted versus Figure 3. Grain size fraction and plastic index, a) Norther
the average qc normalized over the atmospheric pres sector, b) Southern sector.
sure pa. Cc is in the range 0.25 – 0.5 when qc/pa is
between 5 and 15 and decreases as the normalized These tentative relations allow, given a generic qc pro
cone resistance increases, indicating a decrease of file, to derive, for fine grain layers, Cc, ec and e1 accord
compressibility as the soil resistance increases. ing to eqs. 1, 3 and 4. The computed site void ratio ec,
951
compared with the normally consolidate void ratio eNC,
allows to establish if the layer is normally consolidated,
overconsolidated or underconsolidated; eNC can be com
puted at a certain depth, known e1 and Cc, as:
For the pilot sites, Cc and e1 are measured quan

tities; ec was computed via eq. 2 from w and Gs
measured on undisturbed samples.
Figure 8 compares, for the site of Gorino:
– the profile of eNC_lab computed via eq. 5 and using
e1 and Cc from laboratory measures; Figure 5. Altitude of the normal compression line e1 vs. Cc.
–the site void ratio ec_lab at sampling depth from eq. 2;
– the profile of eNC_CPT computed via eq. 5 and using order of 50 cm. This value is obtained as integral of
e1 and Cc estimated from qc via eqs. 3 and 1; the difference Δe = ec_lab – eNC_lab, at the depths
–the site void ratio ec_CPT at sampling depth from eq.4. where eNC_lab > ec_lab.
Regarding to the profile eNC_CPT (obtained via
Eq. 4) it can be noted that although it differs from
eNC_lab profile, it is in relation to ec_CPT similarly to
the lab profiles. In this case, the possible consolida
tion settlement is of the order of 74 cm.
Figure 6. Compression index Cc as a function of qc/pa.
Figure 4. Compression index as a function of depth.
The discontinuities which characterise the eNC_lab

profile (black line) are due to change in Cc and e1
derived from direct measures on undisturbed sam
ples. On the other hand, the profile eNC_CPT (green
line) is more regular, as it mirrors the qc profile. The
comparison between ec_lab value and the eNC_lab pro
Figure 7. Site void ratio ec as a function of qc/pa.
file shows that from 6 to 10 m the soil is slightly
over-consolidated; between 10 and 14 m the ground
is normally consolidated; at greater depths the cur 4 LARGE SCALE APPLICATION
rent void ratio is greater than the normally consoli
dated value, indicating either that the ground on site The procedure described above was applied to CPTUs
is still subject to consolidation compression or that it selected from the SGSS database and located in the
is slightly structured. In the first case, subsidence two sectors of the regional coast evidenced in Figure 1
settlement due to the one-dimensional compression (blue dots). Figure 9 shows for one of the test sites
of the layer present between 14 and 27 is of the (where a silty clay layer, with frequent sandy
952
intercalations, is present from 14 m to 30 m of depth
from the ground surface, topped by a sandy deposit),
the measured tip resistance qc, lateral friction fs, inter
stitial overpressure u2, the estimated hydrostatic profile
u, the profiles of the normal consolidated void ratio
eNC_CPT and the current void ratio ec_CPT. From the
large scale application resulted that the settlements
estimated for the norther sector are lower than those
expected in the southern area (average value of
0.38 m in the Ferrara area and 0.58 m in the Ravenna
area).
Figure 8. Gorino site: computed and measured void ratio.
Higher thickness of the layers of fine-grained soil Figure 9. Application of the correlation to a CPTU.
returns higher estimated settlements. The local
values of estimated consolidation settlement have 5 CONCLUSION
been interpolated over the areas of interest to derive
a map of expectable subsidence. The results of a significant number of CPTU and
The function was tested to the CPTs in the northern laboratory tests (particle size analyzes, Atterberg
coastal sector, between Casalborsetti and Lido di limits, oedometric tests) carried out in coastal sites
Volano. The potential settlement values for each point characterized by the presence of recently deposited,
were interpolated with the inverse distance weighting fine-grained soil deposits, have been analyzed and
method, obtaining the map shown in Figure 10. The spa interpreted to calibrate a semi-empirical correlation
tial distribution and the range of settlements are consist between the compressibility of fine grained soils and
ent with the subsoil nature, characterized by a marked the results of static penetrometric tests.
lithological heterogeneity. The lowest settlements are The interpretation of all the available data allowed
expected near the Reno river mouth and between Porto the elaboration of correlations between:
Garibaldi and Lido delle Nazioni where the delta and – the tip resistance qc measured by a CPTU and the
the coastal sandy deposits are thick. On the other hand, compression index Cc measured during the oedo
in the areas affected by the highest settlements, prodelta metric tests;
clays, lagoonal silts and marsh peat are prevailing. – the resistance qc and the site void ratio ec;
The map provides important information to inter – the compression index Cc and the altitude e1 of
pret the subsidence monitoring data and it also sup the normal compression line in the e-logσ’v plane.
ports spatial planning, highlighting the areas most
vulnerable to coastal risks, such as marine submer These correlations, even if characterized by sig
sion due to land lowering. nificant dispersion, have been used to elaborate
953
expected subsidence map, which provides important
information to interpret the subsidence monitoring
data and supports spatial planning, highlighting the
areas most vulnerable to coastal risks, such as
marine submersion due to land lowering.
It should be pointed out that
– the estimated settlements is probably an upper
limit, as it’s not possible to distinguish undercon
solidation from structuring, and its course over
time is not known;
– if the difference between the current site void
ratio and the normal consolidated void ratio were
due to cementation or the presence of a structure,
in the absence of variation in the lithostatic con
ditions, the soil would be stable and would not
show spontaneous subsidence;
– the calibrated correlations are based on a limited
sample, albeit significant, of geotechnical data
that must be integrated to allow for a refinement
of the proposed equations;
– nevertheless, the database employed is the first attempt
of systematical collection, analysis and interpretation
of compressibility data in area strongly affected by
subsidence of natural and anthropic origin;
– in the definition of the empirical correlations, only
coastal sites located along the regional coast were
taken into consideration, therefore the calibrated
equations could only have local validity and should
be used only in contexts similar to those studied;
– the calibrated equations should be used only for
large-scale areal considerations; for local assess
ments at the scale of the artefact, alternatives to
the necessary geotechnical investigations on site
and in the laboratory cannot be considered;
Figure 10. Map of expectable settlements in the coastal – refinement of calibrated correlations through the
sector between Casalborsetti (RA) and Lido di Volano (FE). execution of test fields and site and laboratory
geotechnical investigations is desirable.
a procedure that allows to evaluate, starting from the
results of a CPTU, the site void ratio profile ec of the REFERENCES
fine-grained layers and the voids ratio profile that the
fine soil would have if normally consolidated, eNC. Amorosi, A., Centineo, M.C., Colalongo, M.L., Pasini, G.,
The comparison between the two profiles allows to Sarti, G., Vaiani S.C. 2003. Facies architecture and
establish if the fine soil layer is: latest Pleistocene- Holocene depositional history of the
Po Delta (Comacchio area), Italy. J. Geol., 111, 39–56.
– overconsolidated (ec < eNC); Burland, J.B. 1990. On the compressibility and shear
– normally consolidated (ec ≈ eNC) or strength of natural clays. Géotechnique, 40(3), 329–378
– it has yet to reach the condition of equilibrium due Cuffaro, M., Riguzzi, F., Scrocca, D., Antonioli, F.,
to its own weight, or the soil has a structure or Carminati, E., Livani, L., Doglioni, C. 2010. On the
microcementation (ec > eNC); at present it is not pos geodynamics of the northern Adriatic plate. Rend. Fis.
Acc. Lincei, 21 (Suppl.1):S253–S279.
sible to distinguish the two different conditions. Ricci, Lucchi, F., 1986. The Oligocene to Recent foreland
In the first two cases the layer is stable and no basins of the Northern Apennines. In (P.A. Allen &
subsidence is expected in the absence of lithostatic P. Homewood eds): Foreland Basins. IAS Spec. Publ.,
load variations. In the third case, the difference 8, 105–139.
Stefani, M., Vincenzi, S. 2005. The interplay of eustasy, cli
between the current void index and the normal con mate and human activity in the late Quaternary depos
solidated theoretical one allows to estimate any itional evolution and sedimentary architecture of the Po
settlements that the deposit has yet to/could show Delta system. Marine Geology 222-223, 19–48.
due to consolidation or destructuring. Teatini, P., Tosi, L., Strozzi, T. 2011. Quantitative evidence
The procedure elaborated has been used to com that compaction of Holocene sediments drives the pre
pute the possible subsidence settlement at selected sent land subsidence of the Po Delta, Italy. Journal of
sites along the Emilia-Romagna coast and to derive Geophisics Research, 116, B08407.
954
© 2022 the Author(s), ISBN 978-1-032-31259-0
A comparative study on liquefaction assessment of Rajarhat area of Kolkata

by using different approaches
Abhipriya Halder
Department of Civil Engineering, Narula Institute of Technology, India
Kaustav Das, Saptarshi Nandi & Kaushik Bandyopadhyay

Department of Construction Engineering, Jadavpur University, India
ABSTRACT: Liquefaction potential assessment is of particular significance to geotechnical engineers in order

to evaluate the risks for construction of high rise buildings, railway embankments, canal, road embankments, earth
dams etc. in areas susceptible to liquefaction. Investigations were made at Rajarhat area of Kolkata, India by
means of CPT, DMT, SPT and SDMT measurements. The main purpose of carrying out these tests is to grab
a relatively rare opportunity to compare the results of all these methods for arriving at a conclusive evidence in
predicting liquefaction. It was found that the results of all the in-situ techniques were relatively consistent in pre
dicting liquefaction within a depth range of 0 to 17.0 m below the ground surface. CPT, DMT, SPT and SDMT
are basically field-based in-situ tests for measurements of important geotechnical parameters and also provide reli
able data for assessment of liquefaction potential of a site. In recent times Finite Element Method based (FEM),
a powerful viable alternative tool is available to the geotechnical engineers for application in various fields. In the
present investigation, an attempt has been made to study the assessment of liquefaction based on the numerical
simulation using Finite element software in addition to all the field techniques mentioned before. The main object
ive of the present study is to evaluate the correlation between factor of safety against liquefaction using different
in-situ test approaches as well as the FEM tool. Finally, the paper presents the correlation between the depth and
factor of safety for this particular Rajarhat area of Kolkata for understanding the vulnerability of this area to
liquefaction.
Keywords: Dilatometer and seismic dilatometer test, cone penetration test, standard penetration test,
Finite Element Method, liquefaction potential
1 INTRODUCTION the vertical stress and it results in the loss of shear

strength of saturated soil mass. This particular phe
1.1 Background nomenon is known as liquefaction. Therefore, esti
mation of liquefaction susceptibility of soil is
Soil plays a vital role for the proper installation of
important prior to any construction of structures for
footing. Footings are placed for transferring the
avoiding the future vulnerability to any calamity.
super-structure load to soil. Hence, proper findings
out of geotechnical parameters including liquefac
tion susceptibility of sub-soil maintains an important
1.2 Objectives
role for the stability of the soil. If it is not accounted
properly, the entire stability of structure can be The aim of this study is to assess the liquefaction
adversely affected due to failures of the sub-soil. susceptibility and evaluate the factor of safety
Soil liquefaction is a phenomenon occurred due to against liquefaction by using different in-situ test
vibration (generally during earthquake) to the sub approaches (i.e., SPT, CPT, DMT and SDMT) as
soil and results to the failure of sub-soil. During well as the FEM (finite element method) tool (i.e.,
earthquake cyclic loading is triggered and passes NovoLIQ and Plaxis 2D software). Finally, present
through the sub-soil. Therefore rise of pore water the variation of factor of safety (FOS) along depth
pressures takes place through sub-soil in addition to for Rajarhat area of Kolkata.
DOI: 10.1201/9781003308829-143
955
2 MATERIALS AND METHODS 5. Computation of Factor of safety (FOS) from SPT,
CPT and DMT tests data.
2.1 Liquefaction susceptibility criteria of Rajarhat 6. Plaxis 2D and NovoLIQ software were used as
a finite element tools.
Excessive pore water pressure generation is the
cause of liquefaction. It is directly connected to the
compositional characteristics of the soil i.e., fine
3 RESULTS AND DISCUSSIONS
content, gradation, plasticity index etc.
Generally, cohesion less soils (sand, silty sand or
sandy silt) undergo liquefaction but this phenomenon 3.1 Estimation of soil property
may be also occurred on plastic & cohesive silty The in situ investigation was done by two numbers of
clay and sensitive clay (Updike et al., 1988; Kramer, boreholes of up to 30 m depth in which the standard
1996). Wang (1979) suggested Chinese Criteria to penetration test (SPT) were performed. The cone
evaluate liquefaction susceptibility based on the penetration tests (CPT), Flat Dilatometer tests (DMT)
earthquake observations in China. and seismic Dilatometer test (SDMT) were performed
There are four criteria which might be considered up an average depth of 20m. Undisturbed soil samples
to evaluate liquefaction susceptibility for the fine were collected and unconsolidated undrained (UU) tri
grained sediments (Wang, 1979; Kramer, 1996). axial tests were performed. In addition, the grain size
These criteria are given below distribution tests were performed on collected soil
1) Fraction finer than 0.005 mm ≤ 15%
samples to obtain the fine fraction of the subsoil.
2) Liquid limit, LL ≤ 35%
Two numbers of CPT (CPT1 and CPT2) tests were
3) Natural Water Content ≥ 0.9LL
conducted out up to a depth of 19.0m below the
4) Liquidity Index ≤ 0.75
ground level, at Rajarhat, Kolkata. The CPT tests
were carried out by using the attached CPT assembly
However, many researchers have followed some provided with Pagani TG 63-150 penetrometer. The
varying criteria on estimating the liquefaction sus CPT tests were performed at 2cm/s penetration rate
ceptibility which unless if properly addressed, may vertically downward. On every 200mm depth interval
cause the extensive damage when silty or clayey readings were taken.
soils containing more than 15% clay size particles Two numbers of DMT tests i.e., DMT1 and
are found (Bray et al., 2004; Bray and Sancio, 2006). DMT2 (aligned to the straight line with the CPT
On the other hand, based on water content (wc), points), were carried out by giving 500mm spacing
liquid limit (LL) and plasticity index (PI), Bray and between the respective CPT tests points. At one of
Sancio (2006) had proposed a new compositional the DMT points (i.e., DMT2), seismic tests (SDMT)
criteria obtained from the results of cyclic triaxial were performed to evaluate shear wave velocity (Vs)
test to determine liquefaction susceptibility: by striking a hammer (≈10kg) to a steel plate placed
1) Highly Susceptible towards liquefaction: PI < 12 at the ground level under the base of the penetrom
and (wc/LL) ≥ 0.85. eter (Pagani TG-63/150) stand aligned perpendicular
2) Moderately Susceptible towards liquefaction: to the plane of DMT blade. This seismic test was
12 < PI < 18 and 0.85> (wc/LL) ≥ 0.8. performed on every 500mm depth interval starting
3) Non-Susceptible towards liquefaction: PI > 18 from 1m below the existing ground level. This seis
and (wc/LL) < 0.8. mic test was executed during the Dilatometer Test.
The shear wave velocity (Vs) measurement was
The subsurface condition of Kolkata, Rajarhat done with the aid of seismic module attached to the
site specify that the sedimentary deposits under control unit. One number of DMT test and one
lying the city involve predominantly of grain size number of seismic dilatometer test (SDMT) were
favorable for liquefaction and this area is formed also performed up to a depth of 18.0m below ground
on coarse grained artificial non engineered fill. So level at this site.
the site may be susceptible to soil liquefaction. CPT1 and DMT1 points were located near the
[Nath et al., 2018] Borehole1 (BH1) and the rest test points (i.e., CPT2
and DMT2) were located near the Borehole 2 (BH2),
as shown in Figure 1.
2.2 Methods
Different Methods used for liquefaction susceptibil 3.2 Determination of earthquake magnitude from
ity analysis in this study were given below: PEER GROUND MOTION DATABASE
1. Estimation of Soil property. Authentic ground motion database is needed as an
2. Determination of earthquake magnitude from important parameter of earthquakes that is likely to
peer ground motion database. occur in Rajarhat area of kolkata. To get the data his
3. Determination of CSR. tory of earthquake, it was collected from the website
4. Determination of CRR from SPT, CPT and DMT of the PEER GROUND MOTION DATABASE.
tests data. From the previous earthquake database results, it
956
Where,
amax = peak horizontal acceleration on the surface
of soil caused by earthquake
g = gravitational acceleration
σv = vertical overburden stress
σ′vo = effective vertical overburden stress
rd = coefficient of stress reduction
3.4 Determination of CRR
3.4.1 Determination of CRR from SPT

Seed & Idriss (1983 -1985) proposed CRRM=7.5
equation as:
Figure 1. Location plan of the testing points.
Where:
(N1)60cs: Corrected Value of N SPT
3.4.2 Determination of CRR from CPT

CRR from CPT had been calculated using this
following graph:
Figure 2. Photographs of the test progress.
was observed that the earthquake magnitude in the

study area is 6.5 MW (Moment Magnitude) which
was used as boundary earthquake magnitude for
liquefaction susceptibility analysis of the study area.
As per Indian standard (i.e., IS 1893 (Part-1):
2016), the zone wise Peak Ground Acceleration
(PGA) is mentioned. On that basis Kolkata is classi
fied into the zone III having PGA of 0.3g. Figure 3. Graph of CRR Vs (qc1N)cs recommended by
Robertson and Wride (1998).
3.3 Determination of CSR Where:
The CSR value had been determined using Equa (qc1N)cs: Value of corrected cone resistance
tion-1, given below. The components of the equation

are horizontal earthquake peak acceleration on the 3.4.3 Determination of CRR from DMT
soil surface (amax), gravity, stress reduction coeffi Cyclic Resistance Ratio (CRR) was calculated from
cient, and overburden stress value. All variables are horizontal stress index (Kd) values using the equa
formulated in the following equation: tion proposed by Monaco et al (2005).
957
3.5 Computation of factor of safety (FOS) from
SPT, CPT, DMT tests data
Furthermore, CRR value was compared to the value
of CSR of each layer of soil for evaluating the value
of FOS. FOS was plotted against depth, showing the
value of FOS for each test depth (i.e., 200 mm).
Below is a graph of the variation of factor of safety
along depth for SPT tests.
Figure 6. Variation of FOS with depth for DMT1, CPT1,

SPT1 tests points.
Figure 4. Variation of CRR with depth for DMT1, CPT1,

SPT1 tests points.
Figure 7. Variation of FOS with depth for DMT2, CPT2,

SPT2 tests points.
Figure 5. Variation of CRR with depth for DMT2, CPT2,

SPT2 tests points.
3.6 NovoLIQ and PLAXIS 2D software were used

as finite element tools
3.6.1 NovoLIQ software

NovoLIQ finite element based software had been
used for liquefaction analysis based on SPT and Figure 8. Variation of FOS with depth at the test location
SDMT tests data.The following graphs show the of BH-01 based of SPT.
variation of FOS with depth
A Plate had been assumed to be placed on the
3.6.2 Plaxis software soil. The properties of the plate is given below.
Plaxis 2D finite element based software had been deff (effective depth) =150mm,
used for liquefaction analysis based on DMT tests EA = 4107919.18 kN/m for isolated footing
data. EI = 7702.3 kN m2/m for isolated footing
958
Figure 9. Variation of FOS with depth at the test location Figure 11. Settlement of 2m × 2m isolated footing when
of BH-02 based on SPT. an earthquake of magnitude 6.50 is applied.
4 CONCLUSION
From the above result and discussion it can be con

cluded that
• For SPT, CPT and DMT, the graph of factor of
safety with depth showing a similar pattern upto
an average depth of 11.0m but in case DMT the
graph is showing an irregularity upto an average
depth of 4.0m. Soil description estimated from
DMT test shows that there is a fill layer of 4.0m
at the test location. Hence it can be concluded
that the liquefaction analysis done by DMT is
more realistic in nature with respect to CPT
and SPT.
• The liquefaction Analysis done by the finite elem
ent software ‘NovoLIQ’ based on SPT data
Figure 10. Variation of FOS with depth at the test location shows a similar pattern with coventional
of SDMT. approach. But when liquefaction analysis is done
based on SDMT data, it shows that the soil is
On these plates a static load of 85 kPa with liquefiable upto a depth of 14.0m f. Hence it can
a eartquake of magnitude of 6.5, corresponding be concluded that the liquefaction analysis based
settlement of the foundation soil profile has been on SDMT data is conservative in nature.
estimated. The value of the settlement had been • The liquefaction Analysis had been done by the
compared with the findings of Tokimatsu, K., & finite element software ‘Plaxis 2D’ for 2m × 2m
Seed, H. B. (1987)., Evaluation of settlements in isolated footing based on DMT data. It shows
sands due to earthquake shaking paper. a settlement of 169mm when an earthquake of
magnitute 6.50 is applied on the foundation. This
value of settlement suggested that the foundation
3.7 Discussion will fail due to the liquefaction effect according
In this paper, the value of factor of safety against to the table (Tokimatsu et al., 1987) given below.
liquefaction were estimated from in-situ test’s results Table-1
(i.e, CPT, DMT and SPT) and had been plotted for
respective tests depths. Also the computed factor of Layer Thickness (ft) Settlement (in)
safety against liquefaction from SDMT and SPT
tests had been compared and plotted by using finite 1 5 0.13
element based software (i.e., Novoliq). 2 5 0.22
Apart from, the settlement of a shallow founda 3 10 0.67
tion had been calculated by using Plaxis2D software 4 10 0.77
against earthquake magnitude of 6.5 and compared 5 10 0.86
with the allowable limit proposed by Tokimatsu, K., 6 10 0.72
& Seed, H. B. (1987).
959
Hence, more research is required for liquefaction Kramer, S. L. (1996). Geotechnical earthquake engineer
analysis of soil based on insitu tests in Kolkata. ing. Pearson Education India.
Wang W (1979) Some findings in soil liquefaction. Water
ACKNOWLEDGMENTS Conservancy and Hydroelectric Power Scientific
Research Institute, Beijing.
The estimation of settlements was carried out by Bray, J. D., Sancio, R. B., Riemer, M. F., & Durgunoglu, T.
(2004, January). Liquefaction susceptibility of fine-
using PLAXIS2D V20 finite element based software grained soils. In Proc., 11th Int. Conf. on Soil Dynamics
received from BENTLEY EDUCATION for and Earthquake Engineering and 3rd Int. Conf. on
research purposes. Their kind permission to use the Earthquake Geotechnical Engineering (Vol.1, pp.
software is gratefully acknowledged. 655–662). Stallion Press, Singapore.
The liquefaction analysis was carried out by using Bray, J. D., & Sancio, R. B. (2006). Assessment of the
NovoLIQ finite element based software received liquefaction susceptibility of fine-grained soils. Journal
from A ROCSCIENCE COMPANY for research of geotechnical and geoenvironmental engineering, 132
purposes. Their kind permission to use the software (9), 1165–1177.
Nath, S. K., Srivastava, N., Ghatak, C., Adhikari, M. D.,
is gratefully acknowledged.
Ghosh, A., & Ray, S. S. (2018). Earthquake induced
DMT tests were performed with the equipment liquefaction hazard, probability and risk assessment in
received from Studio Prof. Marchetti s.r.l., Rome, the city of Kolkata, India: its historical perspective and
Italy. The Cone Penetration tests were carried out by deterministic scenario. Journal of Seismology, 22(1),
TG63/150 static/dynamic penetrometer provided by 35–68.
Pagani Geotechnical Equipment, Calendasco, IS:1893:2016, “Criteria for Earthquake Resistant Design of
Italy, (https://www.pagani-geotechnical.com/). Sin Structures,” Bureau of Indian Standards, New Delhi,
cere thanks are extended to these companies. 2016. [Online]. Available: https://archive.org/details/
Field boreholes and Laboratory tests were carried 1893Part12016. [Accessed: 28-May-2017]
out by Continental Consultants, Kolkata. Contri Robertson, P. K., & Wride, C. E. (1998). Evaluating cyclic
liquefaction potential using the cone penetration test.
butions from Mr. B.N. Basak (Chief Consultant) Canadian geotechnical journal, 35(3), 442–459.
from Continental Consultants, Kolkata is highly Monaco, P., Marchetti, S., Totani, G., & Calabrese, M.
acknowledged. (2005). Sand liquefiability assessment by flat dilat
ometer test (DMT). In Proceedings of the 16th Inter
REFERENCES national Conference on Soil Mechanics and
Geotechnical Engineering (pp. 2693–2698). IOS
Updike, R. G., Egan, J. A., Moriwaki, Y., Idriss, I. M., & Press.
Moses, T. L. (1988). A model for earthquake-induced Tokimatsu, K., & Seed, H. B. (1987). Evaluation of settle
translatory landslides in Quaternary sediments. Geo ments in sands due to earthquake shaking. Journal of
logical Society of America Bulletin, 100(5), 783–792. geotechnical engineering, 113(8), 861–878.
960
© 2022 the Author(s), ISBN 978-1-032-31259-0
Kriging analysis on CPTU data from offshore wind farm

Rongjie He & Jinhui Li*
Department of Civil and Environmental Engineering, Harbin Institute of Technology (Shenzhen), China
Shaoli Yang
Norwegian Geotechnical Institute (NGI), Oslo, Norway
Ben He
Key Laboratory for Far-shore Wind Power Technology of Zhejiang Province; PowerChina Huadong Engineering
Corporation Limited, (HDEC), Hangzhou, China
ABSTRACT: In recent years, offshore wind power has been widely developed. Because of the com
plexity of marine geology and the large spatial variability of soils, it is necessary to obtain
CPTU data before the foundation design. However, how to reasonably interpret the CPTU data will
greatly affect the reliability and safety of the offshore windfarm foundation. In addition, there is
a need to predict soil conditions at the locations without CPTU data. In this study, based on the
available CPTU data of offshore wind farm in east China sea, kriging method is used to interpret the
CPTU data, and the soil conditions in a large area can be predicted. This method can be used to
develop ground model for a large offshore wind farm, and estimate the relevant soil parameters
based on CPTU data. The analysis methods can provide reference for other projects of offshore wind
farm.
Keywords: offshore wind farm, CPTU, kriging method, marine geology
1 INTRODUCTION the locations without CPTU data has become an

important engineering problem which needs to be
Offshore wind power has been widely used as solved.
the clean energy in many countries. Dinh and Based on the above background, taking the
Nguyen (2019) proposed that offshore wind CPTU data of offshore wind farm in east China
power has the advantages of higher full load sea as an example, this study used the Kriging
hours per year, longer service life and higher method to predict the CPTU data in the unknown
rotor speed comparing with onshore wind power. area, and then used the Robertson method
In practical engineering, foundation design should (Robertson and Wride, 1998) to obtain the soil
be carried out before the installation of offshore classification of the whole seabed profile. The
wind turbine. predicted results can reach the 95% confidence
Marine soil has the characteristics of large spa interval in statistics. The algorithm can provide
tial variability (Dasaka and Zhang 2012; Ching and reference and guidance for the foundation design
Phoon 2013; Li et al. 2014; Lloret-Cabot et al. of offshore wind turbine.
2014). Before the foundation design of the offshore
wind turbine, it is necessary to obtain the CPTU
data of relevant soil layers. Then the obtained 2 INVESTIGATION SITE OF OFFSHORE
CPTU data can be interpreted to get soil param WIND FARM
eters. For example, Robertson method (Robertson
and Wride, 1998) can be used to determine the clas The site of the offshore wind farm is in the east
sification of marine soil layer based on the CPTU China sea. The positions of the boreholes are shown
data. However, due to the high cost of offshore in Figure 1. This site contains 27 boreholes. The
work, the CPTU tests in a large area are usually drilling depth of each borehole is more than 35 m,
limited. Therefore, how to predict soil conditions at and CPTU data is measured every 0.02 m along the
DOI: 10.1201/9781003308829-144
961
depth. In Figure 1, the interval of boreholes in differ
ent columns is 11000 m which is too large, so 15
boreholes in the same column are selected for ana
lysis as shown in the frame. The measured CPTU
data contain cone penetration resistance (q), sleeve
friction (f) and pore pressure (u).
According to the soil samples at the site, the actual
soil layers of boreholes no. 50, no. 72 and no. 77 have
been already known which are shown in Figure 2. It
can be seen from Figure 2 that the actual seabed geo
logical conditions are generally divided into four
layers: Ooze clay, clay, silty clay, and silty sands. The
soil layers in the first 25 m are relatively homogeneous,
and they are Ooze clay. The soil layers beyond
25 m vary from location to location. Figure 3 shows
the typical CPTU data along depth of borehole no. 50.
It can be seen that the trend of the curve at the first Figure 3(a). Measured cone resistance along depth below
25 m is increasing linearly, and the trend of the curve is seafloor
much more complicated beyond 25 m.
Figure 3(b). Measured sleeve friction along depth below

seafloor
Figure 1. The position of the boreholes.
Figure 3(c). Measured pore pressure along depth below

seafloor.
Figure 3. The variation of CPTU data with depth in bore-
hole no. 50.
Figure 2. Soil conditions of known boreholes (Borehole Based on CPTU data, the classification of soil
no. 50, no. 72 and no. 77). layer can be obtained. There are many traditional
962
soil classification methods, such as Jefferies method points in region D, whose position coordinates
(Jefferies and Davies, 1991), Olsen method (Olsen are x1, x2, x3 to xn, respectively, and the corres
and Mitchell, 1995), Robertson method (Robertson ponding observation values are Z(x1), Z(x2), Z(x3)
and Wride, 1998) and so on. Liu et al. (2013) pro to Z(xn), respectively. Then the formula of the
posed that the Robertson method is suitable for the estimated value at unknown position x0 in region
Chinese soil classification. Therefore, the Robertson D is shown in Equation 7,
method is used for soil classification in this study.
The calculation formulas are shown from Equation 1
to Equation 6,
where λi is the weight coefficient of Kriging interpol

ation. The values of λi should be known. The semi
variogram is used to calculate the λi. The
semivariogram of the observation data can be calcu
lated by Equation 8,
where Nh is the number of the observation points

and h is the separation distance between different
points. Based on the semivariogram values of the
observation points, the Gaussian model and expo
nential model are usually used to fit the trend of the
semivariogram which are shown in Equation 9 and
where Qtn is normalized cone penetration resistance;
Equation 10,
Fr is normalized sleeve friction; Bq is normalized pore
pressure; σv0 is the total overburden stress; σ0v0 is
effective overburden stress; u0 is equilibrium pore
pressure; qt is corrected cone penetration resistance;
a is the net area ratio between load cell support diam
eter, d, and cone diameter, D; Ic is soil classification
index. The soil types are attributed from Ic which is
shown in Table 1.
where C0 is nugget; C is partial sill and a is range.
Before fitting the semivariogram, it is essential that
Table 1. The soil classification based on Robertson the data is stationary; that is, the mean and covari
method. ance of the data depend only upon separation, not on
absolute location. If the data are non-stationary,
Ic Soil type treatment must be given to transform the data to
a stationary set by removing the deterministic com
Ic>3.6 Ooze clay ponent called the trend, and the stationary residual
2.95<Ic,<3.6 Clay random component is then analyzed.
2.60<Ic,<2.95 Silty clay After fitting the semivariogram, the weight coeffi
2.05<Ic,<2.60 Sandy silt cient λi can be calculated based on Equation 11,
1.31<Ic,<2.05 Silty sands
Ic<1.31 Dense sand
3 KRIGING METHOD
Kriging interpolation (Liu et al., 2016) is an opti

mal linear unbiased interpolation method. For where γij is the modelled semivariogram values based
ordinary Kriging, there is a group of observation on the distance between the two samples pertaining to
963
the ith and jth locations, and μ is the Lagrange param
eter. The formula for calculating the variance is
shown in Equation 12,
where σOK is the OK standard variance of the Kri

ging method. Assuming that the parameter is nor
mally distributed with a mean value z0 (estimated
value) and a standard variance σOK, the 95% confi
dence interval of the estimation is [z0-1.96σOK,
z0+1.96σOK] (Ang and Tang, 2007).
Taking the CPTU data in depth 10 m for analysis
Figure 4(a). The semivariogram of cone resistance
as an example, the fitting equations of the determin
istic component are shown in Equation 13, Equa
tion14 and Equation 15, respectively,
where qtrend, ftrend and utrend are trend values of cone

penetration resistance, sleeve friction and pore pres
sure, and x is the longitude coordinate of the obser
vation points.
After removing the deterministic component, the
stationary residual random components are used to Figure 4(b). The semivariogram of sleeve friction
fit the semivariogram. In this case study, comparing
with Gaussian model, exponential model has higher
fitting goodness. As a result, the exponential func
tion is used to fit the semivariogram. The number of
lag distances is 105 in total. Taking about 200 m as
a group, the lag distances are divided into 14 groups.
The lag distances of 14 groups are averaged to
obtain the best fitting results and they are shown in
Figure 4.
Cheon and Gilbert (2014) indicated that it is reason
able to have a horizontal range of more than 1000 m in
offshore engineering. Based on the results, the range of
the three residuals are all 120 m which are reasonable.
Then, the CPTU data in unknown positions can be pre
dicted based on the modelled semivariogram.
4 PREDICTION OF SOIL TYPES
From Figure 2, there are mainly four soil layers in the Figure 4(c). The semivariogram of pore pressure
seabed. According to the calculation, different soil Figure 4. The semivariogram of CPTU data in depth 10 m.
layers have different modelled semivariograms. As

a result, four modelled semivariograms were used to
analyze the depths below seafloor at depth of 1m to In order to verify the accuracy of the Kriging
25 m, 26 m to 29 m, 29 m to 31 m, and 31 m to 35 m, method, the soil types at boreholes no. 50 and no. 77
respectively. were predicted. The predicted results are shown in
964
Figure 5 and Figure 6. The results were compared
with the actual soil types at the two boreholes. It can
be seen that the predicted soil types are basically
consistent with the actual soil types. The results
prove that the Kriging method is reasonable.
B1 is the first unknown position without CPTU
data which needs to be predicted. The longitude and
latitude of B1 are 120.556 degree and 27.14 degree,
respectively. For the depth of 15 m, 27 m, 30 m and
35 m, the estimators of cone penetration resistance of
B1 are 0.521 MPa, 1.213 MPa, 2.105 MPa and 3.290
MPa; the estimators of sleeve friction of B1 are 6.7
kPa, 18 kPa, 21.3 kPa and 61.3 kPa, and the estim
ators of pore pressure of B1 are 0.414 MPa, 0.601
MPa, 0.808 MPa and 0.822 MPa. The 95% interval
of confidence is shown in Table 2 and the predicted Figure 5. The comparison between prediction value and
soil types are shown in Figure 7. For Table 2, it just actual data in borehole no. 50.
means that the value will be within this interval.
Table 2. The predicted CPTU data in B1 (95% confidence

interval).
Depth (m) q (MPa) f (kPa) u (MPa)
15 [0.521, 0.521] [6.690, 6.705] [0.411, 0.418]

27 [1.106, 1.321] [17.3, 18.7] [0.572, 0.630]
30 [0, 8.794] [20.80, 21.8] [0.443, 1.173]
35 [0, 31.725] [0.056, 0.067] [0.176, 1.469]
B2 is the second unknown position without CPTU

data which also needs to be predicted. The longitude
and latitude of B2 are 120.592 degree and 27.14
degree, respectively. For the depth of 15 m, 27 m,
30 m and 35 m, the estimators of cone penetration
Figure 6. The comparison between prediction value and
resistance of B2 are 0.520 MPa, 1.186 MPa, 2.038
actual data in borehole no. 77.
MPa and 3.066 MPa; the estimators of sleeve friction
of B2 are 6.8 kPa, 17.3 kPa, 21.5 kPa and 57.3 kPa,
and the estimators of pore pressure of B2 are 0.414
MPa, 0.643 MPa, 0.760 MPa and 0.827 MPa. The
95% interval of confidence is shown in Table 3. For
Table 3, it just means that the value will be within this
interval.
According to the soil classification criteria, the
predicted soil types are the same with that at location
B1, which can be seen in Figure 7.
Table 3. The prediction CPTU data in B2 (95% confi

dence interval).
Depth (m) q (MPa) f (kPa) u (MPa)
15 [0.52, 0.52] [6.79, 6.805] [0.410, 0.417]

27 [1.081, 1.291] [16.6, 18.02] [0.622, 0.665]
30 [0, 8.636] [0.021,0.022] [0.415, 1.104]
35 [0, 30.797] [0.052, 0.062] [0.189, 1.465]
Based on the CPTU data of boreholes, the seabed Figure 7. The prediction of soil classification in B1 and B2
profile can be predicted and the results are shown in (95% CI).
965
work was supported by the Natural Science Founda
tion of China [grant number 51979067] and Shen
zhen Technology Innovation Project [grant number
JCYJ20210324121402008].
REFERENCES
Ang, A.H.S. & W.H. Tang (2007). Thin layer effects on the
CPT qc measurement. Canadian Geotechnical Journal,
42 (5), 1302–1307.
Ching, J. & K.K. Phoon (2013). Probability distribution
for mobilized shear strengths of spatially variable
soils under uniform stress states. Georisk, 7 (3),
Figure 8. The prediction soil layer of seabed profile. 209–224.
Cheon, J.Y & R.B. Gilbert (2014). Modeling spatial vari
ability in offshore geotechnical properties for
Figure 8. It can be seen that the seabed profile reliability-based foundation design. Structural Safety. 49
mainly has four soil layers containing ooze clay, (2014), 18–26.
Dasaka, S.M. & L.M. Zhang (2012). Spatial variability of
clay, silty clay and sandy silt. In the first 25 m, the
in situ weathered soil. Geotechnique, 62 (5), 375–384.
soil is homogeneous which is ooze clay. Beyond Dinh, V.N. & H.X. Nguyen (2019). Design of an offshore
25 m, the soil conditions become much more com wind farm layout. Lecture Notes in Civil Engineering,
plex. The general conditions of soil layer are consist 2019, 18, 233–238.
ent with the actual soil conditions. It means that Jefferies, M.G. & M.P. Davies (1991). Soil classification by
Kriging method can predict the soil conditions at the cone penetration test: Discussion[J]. Canadian Geo-
unknown positions effectively, which can be used to technical Journal, 28 (1), 173–176.
guide the foundation design of offshore wind turbine. Li, J.H., Huang, J., Cassidy, M.J., & R. Kelly (2014).
Spatial variability of the soil at the Ballina National
Field Test Facility. Australia Geomechanics, 49 (4),
41–47.
5 CONCLUSIONS Li, J.H., Cassidy, M.J., Huang, J., Zhang, L., & R. Kelly
(2016). Probabilistic identification of soil stratification.
In this study, the offshore wind farm in east China Geotechnique, 66 (1), 16–26.
sea is analyzed as a case study. The results indicate Liu, S.Y., Cai, G.J., & H.F. Zhou (2013). Practical soil clas
that Robertson method is suitable for the soil classi sification methods in China based on piezocone penetra
fication in the east China sea and Kriging method tion tests. Journal of Geotechnical Engineering, 35 (10),
can predict the soil conditions in unknown positions 1765–1775 [in Chinese].
with 95% interval of confidence. The algorithm can Lloret-Cabot, M., Fenton, G.A., & M.A Hicks (2014). On
the estimation of scale of fluctuation in geo-statistics.
be used to provide guidance and reference for the
Georisk, 8 (2), 129–140.
foundation design of offshore wind turbine. Olsen, R.S. & J.K. Mitchell (1995). CPT stress normaliza
tion and prediction of soil classification. Proceedings of
International Symposium on Cone Penetration Testing,
ACKNOWLEDGMENTS CPT’95, Linköping, Sweden, SGI Report 3:95, 2,
257–262.
This study is supported by Key Laboratory for Far- Robertson P.K. & C.E. Wride (1998). Evaluating cyclic
shore Wind Power Technology of Zhejiang Province liquefaction potential using the cone penetration test.
and Norwegian Geotechnical Institute (NGI). This Canadian Geotechnical Journal, 35 (3), 442–459.
966
© 2022 the Author(s), ISBN 978-1-032-31259-0
Evaluation of CPT-based design method for offshore pile

B. Huang
School of Architecture and Civil Engineering, Huizhou University, China
E.J. Bittar
Y. Zhang
The University of Newcastle, Australia
X. Fu
School of Civil Engineering, Wuhan University, China
ABSTRACT: In offshore engineering, it’s difficult to obtain undisturbed samples. CPT-based design
method of pile has become recommending method. In this paper focusing on offshore pile foundation engin
eering, the typical CPT-based design methods are introduced. Various design methods are evaluated in respect
of friction fatigue, pile loading direction, and the plug ratio of open-ended pile in capacity contribution, and
the determination of design parameters. CPT-based design methods in the clay are compared with API
method. For the typical stratum in the China East Sea, the monopiles of an offshore wind farm are analysed.
The calculated capacities of monopiles with different methods are compared with the measured value, and the
reliability of methods are evaluated. At last according to the vertical loading condition of the offshore mono-
pile, reasonable suggestions for the current design methods and parameters determination are given.
1 INTRODUCTION Among the methods applicable to sand, all but

API-00 are based on CPT. It is noted that the cone
At present, the popular methods for the vertical bearing tip resistance has not been corrected by pore pressure.
capacity of pile foundations in offshore engineering When using these methods, the difference between
are either based on the empirical formula of CPT tip the calculated results before and after the correction
resistance or are directly related to soil parameters, should be fully considered. A unified CPT-based
such as undrained strength, yield stress, sensitivity, method has been developed by the JIP group of
internal friction angle, etc. The CPT-based design UWA, NGI, Imperial College, Fugro, BP, Delft Uni
method of pile has gradually replaced design methods versity of Technology, University of Texas, DNV·GL
based on lab tests or previous experience, and it has and Lloyd’s Register EMEA (Lehane et al., 2020).
become the preferred method for offshore pile founda The methods applicable to clay include API-00,
tion. It is generally believed that the cone tip resistance Fugro-96, NGI-05, ICP-05, UWA-13 and Fugro-10
obtained by CPTu is more reliable than the sleeve fric methods, of which API-00, Fugro-96, and NGI-05 are
tion, which is the main design method parameter. based on the undrained strength of clay. The ICP-05
Table 1 summarizes the current design methods for method is based on the tip resistance of CPT and is not
offshore piles. corrected by pore pressure. UWA-13 and Fugro-10 are
Fellenius (2020) pointed out that all CPT-based based on the test results of CPTu with the tip resistance
pile design methods are established under basically of CPT corrected by pore pressure. A new Unified
the same geological conditions for specific areas. CPT-based method is also being developed by the JIP
That is to say, each method is based on limited piles group.
and soil tests. If geological conditions are not simi
lar, these methods are not suitable. Following the
completion of a Joint Industry Project, a unified 2 DESIGN METHOD
database (Lehane et al., 2017) has been developed
which includes ICP, NGI, Fugro, UWA and other 2.1 Unified CPT-based method in sand
databases.
The unit shaft friction may be calculated from:
DOI: 10.1201/9781003308829-145
967
Table 1. Pile design methods referred. 2.2 UWA-13 method in clay
Soil type Design method Reference The unit shaft friction is calculated from:
Sand Unified CPT-based Lehane et al. (2020)

method
Sand API-00 API (2000)
Sand Fugro-05 Kolk et al. (2005)
Sand ICP-05 Jardine et al. (2005) where qt is the total cone tip resistance at depth z, h is
Sand NGI-05 Clausen et al. (2005) the relative distance above pile tip, R* is the equiva
Sand UWA-05 Lehane et al. (2005) lent radius = (R2 – Ri2)0.5, R is the outer radius of pile,
Clay API-00 API (2000) and Ri is the inner radius of pile.
Kolk & Van Der Velde The unit end bearing follows the recommenda
Clay Fugro-96
(1996) tions of Jardine et al. (2005) for undrained loading
Van Dijk & Kolk and assumed that the open-ended piles are always
Clay Fugro-10
(2010)
plugged:
Clay ICP-05 Jardine et al. (2005)
Clay NGI-05 Karlsrud et al. (2005)
Clay UWA-13 Lehane et al. (2013)
Where qt is the average total cone tip resistance over

±1.5D at pile tip.
The method ignores the effects of reversed end
where σ’rf is the radial effective stress at failure, σ’rc bearing or base suction for piles in tension.
is the radial effective stress after installation and
equalization, Δσ’rd is the change in radial effective
stress due to loading stress path (dilation), and f/fc is 2.3 Fugro-10 method in clay
1 in compression and 0.75 in tension. The unit shaft friction is calculated from:
The radial effective stress after installation and
equalization may be calculated as:
where qc is the cone tip resistance, Are is the effect where qn is the net cone tip resistance at depth z, σv0
ive area ratio, Are=1-PLR·(Di/D)2, PLR is soil plug is the total vertical stress at depth z, σ’v0 is effective
length ratio, h is the relative distance above the pile vertical stress at depth z, h is the relative distance
tip, D is the outer diameter of pile, Di is the inner above pile tip, and uL is unit length to render the
diameter of pile. expression dimensionless (= 1.0 m).
The change in radial effective stress due to load The unit end bearing (qb) and reversed end bear
ing stress path (dilation) may be estimated as: ing (qrb) may be calculated from:
where σ’v is the vertical effective stress.

where qn,avg is the average net cone tip resistance
The unit end bearing may be calculated from:
over ±1.5D at pile tip, and ub is the hydrostatic pres
sure at pile tip (in kPa).
3 EVALUATION OF SOIL PLUG

Where Bqc is the average cone tip resistance within
a zone 1.5D above and below the pile tip in rela Lehane et al. (2017) suggested in the absence of
tively homogeneous sands. any better method, Equation (12) was therefore
968
employed for the calculation of plug lenth ratio in API-00 method as the preferred recommended
sand and clay. This is consistent with field test. method in the new version of the API specification.
According to statistical analysis of 71 piles’ data
base, the Unified CPT-based method in sand has the
lowest coefficient of variation among all the design
methods of pile in sand (Lehane et al., 2020).
Unified CPT-based method, UWA-13 and Fugro
where PLR is soil plug length ratio, and Di is the pile
10 methods all take into account the friction fatigue
internal diameter expressed in metres.
effects caused by pile construction.
Lehane et al. (2020) proposed a new empirical
Through a large number of database examinations
formula for soil length plug ratio:
for different design methods, the reliability statistics
analysis results are shown in Table 2. There are 49
piles, with diameters ranging from 0.2 to 0.8m. Com
pared with the pile design method in sand, the calcu
lation results of pile capacity in clay have a larger
where dCPT is the diameter of standard static cone variation coefficient. This is because there are more
penetration, 35.7mm. design parameters for piles in clay compared to sand,
Based on the unified database (Lehane et al., 2017), and they are also more complicated and diverse.
the relationship between the soil plug length ratio and Among the CPT-based design methods, the variation
the pile inner diameter is plotted together, as shown in coefficient of the UWA-13 method is smaller than
Figure 1. It can be seen that there is no obvious differ that of the Fugro-10 method. The API-00 method is
ence in the law of soil plugging in clay and sand, that a bit conservative and Fugro-10 method is sometimes
is to say, the relationship between soil plugging and unsafe compared to UWA-13 method.
soil properties is not significant, only related to the
inner diameter of pile. Equations (12) and (13) reflect
that soil plugging varies with pile diameter with the Table 2. Method uncertainties for piles in clay (Lehane
same trend, especially in the range of 0.4~1.2m pile et al., 2017).
diameter. However, in terms of small diameter piles,
equation (13) is closer to the actual cases. Qc/Qm
Method No of sample μw σw CoVw
API-00 23 0.73 0.29 0.40

UWA-13 43 0.99 0.48 0.49
Fugro-10 43 1.17 0.69 0.59
In Table 2, Qc is the calculated capacity, Qm is the

measured capacity, μw is weighted mean, σw is
weighted standard deviation, and CoVw is coefficient
of variation.
The comparison of open-ended pile design
methods in clay is shown in Table 3. It is believed
Figure 1. Relationship between plug length ratio and pile that the soil inside large-diameter steel pipe pile is in
inner diameter. soil coring mode, and there is no soil plug occurring.
The API-00 method considers that the total bearing
capacity is equal to the sum of the outer shaft resist
The typical pile diameter to wall thickness ratio of ance, the inner shaft resistance and the base resist
steel pipe piles is 30-60. For offshore piles with a pile ance of the annulus area, and use the same unit shaft
diameter greater than 1.5m, the soil plug ratios calcu friction to calculate the inner and outer shaft resist
lated by equations (12) and (13) are both greater than ance. Neither the UWA-13 method nor the Fugro-10
0.96. So for steel pipe piles with a diameter of larger method directly mentions the inner shaft resistance
than 1.5m, there is no soil plug formed or the soil plug in the calculation. Instead, they use the equivalent
length ratio is equal to one which is in coring mode. base resistance to calculate total bearing capacity of
the pile, where the base area is equal to the area of
the outer diameter circle. The base resistance used
4 REVIEW OF DESIGN METHODS FOR by the soil core is a reduction of the unit end bearing
AXIAL CAPACITY OF PILES of the annulus area or closed-ended pile. The reduc
tion factor in the Unified CPT-based method is 24%
The Unified CPT-based method of driven pile in for sand, and in the UWA-13 method, it’s about 40%
sand proposed by the JIP group has replaced the for clay.
969
Table 3. Comparison of design methods for open-ended
pile.
Method Soil type Capacity calculation
API-00 Clay If Qb,plug<Qs,inner, plugged,

sand Qtotal=Qs,outer+Qb,plug+Qb,ann;
If Qb,plug≥Qs,inner, unplugged,
Qtotal=Qs,outer+Qs,inner+Qb,ann.
Unified Sand Unplugged is the same as
CPT-based plugged,
method Qtotal=Qs,outer+Qb,plug+Qb,ann.
UWA-13 Clay Always plugged,
Qtotal=Qs,outer+Qb,plug+Qb,ann.
Fugro-10 Clay Unplugged is the same as
plugged,
Qtotal=Qs,outer+Qb,plug+Qb,ann. Figure 2. Profiles of tip resistance, shaft friction and pore
pressure in CPTu test.
In Table 3, Qtotal is total bearing capacity, Qs, inner

is inner shaft resistance, Qs, outer is outer shaft resist resistance as the parameter which is scientific and rea
ance, Qb, plug is base resistance of the soil plug, and sonable. The pile end is buried 69 m deep and is
Qb, ann is base resistance of the annulus area. located in the supporting layer of silty sand.
When the pile is subjected to a tension load in The undrained strength of clay is determined by
sand, the Unified CPT-based method uses the empirical expression based on CPT on basis of
a reduction factor of 0.75 to calculate the shaft resist Nk=20 inferred by SHANSEP method (Ladd &
ance, without considering the base resistance. When Foott, 1974).
the pile is subjected to a tension load in clay, none of
the methods considers the reduction of the shaft 5.2 Field test of offshore piles
resistance. Fugro-10 propose the unit end bearing
when the pile is in tension, while UWA-13 method The pile outer diameter is 1.8 m, the total pile length
doesn’t consider the base resistance of the pile in is 93 m, and the wall thickness is 20~40 mm. The
tension. API-00 does not explicitly propose IHC S800 hydraulic hammer is used for pile driving.
a calculation of the base resistance in tension. There is no slipping phenomenon during the pile
driving process, and the pile’s embedded depth is
69 m. After pile driving, the soil plug/core is
5 CASE ANALYSIS OF OFFSHORE PILE 1.45 m slightly above the mudline, and the soil plug
ratio of the steel pipe pile which has an inner diam
eter of about 1.7 m is 1.0, which is a soil coring
5.1 Soil conditions
mode.
A field test of large diameter steel pipe piles was car The vertical static compression test was carried
ried out in an offshore wind farm project. The site is out 44 days after the pile was driven using the slow
about 20 km offshore, with the seabed topography of maintained loading method, then rested for 25 days,
the total site area not varying much, and water depth then for the same test pile the vertical static tension
at 8~12 m. As for regional geological information, test was conducted using the slow maintained load
the local Quaternary sediments within the explor ing method. Figure 3 shows the comparison between
ation depth, the upper part is the silty clay and clay the axial compression and tension test results of the
deposited in the Holocene littoral facies, and the pile. The axial compression test of a single pile
lower part is the silty clay and silty sand deposited in achieves the destruction of the soil around the pile,
the Upper Pleistocene estuary-littoral facies. Accord and the ultimate axial compression capacity is 22
ing to the soil layer classification method proposed MN; the axial tension test of the single pile also
by Robertson et al. (1986), the site stratum distribu achieves failure, and the ultimate axial tension cap
tion is shown in Figure 2. acity of the single pile is 18.7 MN. When the pile
CPTu cone tip resistance, shaft friction resistance, head displacement is small, the compression load-
and pore pressure test results are also shown in displacement curve is closer to the tension test
Figure 2. It can be seen that the cone tip resistance of curve, and the compression curve is slightly higher
the clay increases linearly with depth from 4~50 m, than the tension curve; as the pile head displacement
indicating a relatively uniform soil layer. However, gradually increases, the discrepancy between com
the shaft friction resistance changes with depth more pression and tension becomes larger and larger.
discretely. Therefore, pile design methods only use tip After roughly 40 mm, both reach the limit state, and
970
the difference is also the largest. This is because the Table 5. Comparison of measured results and calculated
contribution of the base resistance of the pile in com results of tension capacity of different design methods.
pression increases as the displacement of the pile
increases, while the end resistance can be ignored in Shaft Base Total Calculated/
tension; in addition, at the depth of 50~74.7 m is all resistance resistance resistance Measured
silty sand, and the shaft tension friction in sand is Method (MN) (MN) (MN) (%)
also reduced by 25% compared with compression
Measured NA 0 14.2 NA
(Lehane et al, 2020).
API-00 16.6 0 16.6 117.0
Fugro-10 12.3 0 12.3 87.1
UWA-13 12.9 0 12.9 91.2
Note: The tension capacity measured deducts the weight of

pile and soil plug (4.6 MN). The shaft resistance and base
resistance in sand use the Unified CPT-based method.
Figure 3. Comparison between axial compression and ten

sion load-displacement curves.
5.3 Comparison between design methods

Comparisons of measured and calculated results of
capacity with different design methods for compres
sion and tension are shown in Table 4 and Table 5. It
can be seen in Table 4 that all design methods are
safer in calculating compression capacity, resulting
in about 79%~95% of the measured value. From the
calculated result of tension capacity it can be seen
that in tension, the API-00 method based on Figure 4. Shaft friction distribution with depth during
undrained strength are 17% higher than the meas ultimate bearing calculated by different methods: (a) com
ured value, while the CPT-based design methods are pression; (b) tension.
9%~13% lower than the measured value.
The UWA-13 method and Fugro-10 method,
which are completely based on CPT, are safe in The compression capacity calculated by the UWA
the calculation of compression and tension capacity. 13 method is 81% of the measured value, and the
Fugro-10 method is 79%. The UWA-13 method cal
culated tension capacity is 91% of the measured
Table 4. Comparison of measured results and calculated value and Fugro-10 method is 87%. Therefore, from
results of compressive capacity of different design this project case, UWA-13 method is slightly better
methods. than Fugro-10 method.
Figure 4 is the shaft friction distribution with
Shaft Base Total Calculated/ depth during ultimate bearing calculated by different
resistance resistance resistance Measured methods, compared with measured friction. The
Method (MN) (MN) (MN) (%) shaft friction of clay calculated by API-00 method is
the greatest. The CPT-based methods calculate
Measured NA NA 26.6 NA
a smaller value but are very similar to each other.
API-00 18.1 7.0 25.1 94.5
However, the trends of shaft friction changing with

Fugro-10 13.9 7.0 20.9 78.6
depth are almost the same.

UWA-13 14.5 7.0 21.4 80.8
The measured shaft friction is the sum of the

Note: The compression capacity measured takes into inner and outer friction, and the two parts cannot be
account the weight of pile and soil plug (4.6 MN). The distinguished, while the shaft friction calculated by
shaft resistance and base resistance in sand use the Unified each design method is only the outer friction. There
CPT-based method. fore, for the UWA-13 and Fugro-10 methods with
971
more reliable calculation results, it is obvious that REFERENCES
the measured total friction is greater than the calcu
lated outer friction, especially in sand below 50 m. It API. 2000. API RP 2A-WSD: Recommended Practice for
fully indicates that the contribution of the inner fric Planning, Designing and Constructing Fixed Offshore
tion in the sand is more significant, and relatively the Platform-Working Stress Design, 21st Edition. API.
contribution of the inner friction in the clay is smal Washington, DC.
Clausen, C. J. F., Aas, P. M., & Karlsrud, K. 2005. Bearing
ler. So the internal resistance of the soil core of the
capacity of driven piles in sand, the NGI approach. In
pipe pile is mainly exerted in the lower part of pile, Proceedings of the 1st International Symposium Fron
which increases the complexity of the calculation of tiers in Offshore Geotechnics. Perth, Western Australia.
the internal resistance. Therefore, each design pp. 677–682.
method adopts the equivalent base resistance of soil Fellenius, B. H. 2020. Basics of foundation design, elec
plug or core to calculate the internal resistance tronic edition.
which is a method worth promoting. Jardine, R., Chow, F., Overy, R., & Standing, J. 2005. ICP
design methods for driven piles in sands and clays.
Thomas Telford, London, UK.
Karlsrud, K., Clausen, C. J. F., & Aas, P. M. 2005. Bearing
6 CONCLUSIONS
capacity of driven piles in clay, the NGI approach. In
Proceedings of the 1st Int. Symp. on frontiers in offshore
This paper introduces and evaluates the CPT-based geotechnics, Perth, WA, Australia. Taylor & Francis,
design methods for offshore pile under vertical load London, UK, vol. 1, pp. 775–782.
which are currently popular in the industry. The Kolk, H. J., & Der Velde, E. 1996. A reliable method to
main conclusions are as follows: determine friction capacity of piles driven into clays. In
Proceedings of Offshore Technology Conference, Hous
(a) The Unified CPT-based method is recommended ton, Texas. Pp.337–346.
for full-scale offshore pipe in sand. For clay, the Kolk, H. J., Baaijens, A. E. & Senders, M. 2005. Design
UWA-13 method and Fugro-10 method based on criteria for pipe piles in silica sands. 2005. CRC Press/
CPT are recommended compared to API-00 Balkema,711–716.
method. Ladd, C. C., & Foott, R. 1974. New design procedure for
(b) Through the case analysis of large-diameter stability of soft clays. Journal of Geotechnical and
steel pipe piles, the UWA-13 method and Fugro Geoenvironmental Engineering, vol 100, GT7,
763–786.
10 method based entirely on CPT are conserva Lehane, B. M., Schneider, J. A., & Xu, X. 2005. The
tive in the calculation of compression and ten UWA-05 method for prediction of axial capacity of
sion capacity in which the calculated bearing driven piles in sand. In Proceedings of the 1st Int. Symp.
capacity is 79%~92% of the measured value. on frontiers in offshore geotechnics, Perth, WA, Austra
The UWA-13 method is marginally better than lia. Taylor & Francis, London, UK, vol. 1, pp. 683–689.
Fugro-10 for the case in this paper, and Fugro Lehane, B. M., Li, Y., & Williams, R. 2013. Shaft capacity
10 method is sometimes unsafe. of displacement piles in clay using the cone penetration
(c) It is difficult in offshore engineering to deter test. Journal of Geotechnical and Geoenvironmental
Engineering, 139(2),253–266.
mine the undrained strength of the intact clay
Lehane, B. M., Lim, J. K., Carotenuto, P., Nadim, F.,
sample. The strength parameters in different Lacasse, S., Jardine, R. J., & Van Dijk, B. F. J. 2017. Char
soil layers and different depths can be deter acteristics of unified databases for driven piles. In Pro
mined by the relatively mature SHANSEP ceedings of the 8th International Conference of Offshore
method, but this method is only suitable for Site Investigation and Geotechnics OSIG, London, UK.
low sensitivity, unnaturally cemented and low Society for Underwater Technology, vol 1, pp. 162–191.
structured cohesive soil. Lehane, B., Liu, Z., Bittar, E., Nadim, F., Lacasse, S.,
Jardine, R. J., … & Morgan, N. 2020. A new CPT-based
axial pile capacity design method for driven piles in
sand. In Proceedings of the 4th International Symposium
ACKNOWLEDGEMENTS on Frontiers in Offshore Geotechnics, Austin, Texas,
USA. American Society of Civil Engineers.
The first author gratefully acknowledges the finan Robertson, P. K., Campanella, R. G., Gillespie, D., &
cial support by Natural Science Foundation of China Greig, J. 1986. Use of piezometer cone data. In Pro
(Grant No. 51978540) and China Scholarship Coun ceedings of in Use of in situ tests in geotechnical engin
cil (201906275010). The authors also express sin eering. ASCE, pp. 1263–1280.
Van Dijk, B. F. J., & Kolk, H. J. 2010. CPT-based design
cere thanks to Professor Barry Lehane, who assisted
method for axial capacity of offshore piles in clays. In
with the interpretation of CPT-based methods. We Proceedings of the International Symposium on Fron
also acknowledge the assistance provided by Dylan tiers in offshore geotechnics II, Perth, Australia. Taylor
Mo and Tommy Le. & Francis Group, London, pp. 555–560.
972
© 2022 the Author(s), ISBN 978-1-032-31259-0
Application of CPT based 3DFE approach for estimating monopile

damping in sand
D. Igoe & M.B. Mohammed
Trinity College Dublin, Ireland
ABSTRACT: Monopiles are the most popular foundation type supporting offshore wind turbines. The use of
3DFE modelling to analyse the pile response has grown significantly in recent years, supported by the introduc
tion of the PISA framework and the development of easy to use commercial software such as PLAXIS Mono-
pile Designer. In sands, the small-strain hardening soil (HSS) model is widely adopted for monopile design,
partially due to its relatively simple set of soil input parameters. However to date there is no widely used
approach for determining the required soil parameters and approaches often differ across the industry. Igoe and
Jalilvand (2020) developed an approach to derive the HSS soil model parameters directly from CPT data. This
procedure has been validated by the authors for monotonic loading against a database of large-scale field tests
including PISA field tests in the Dunkirk marine SAND. This paper examines the use of the 3DFE approach
under cyclic loading to determine soil damping for monopiles. The results from the 3DFE analysis are com
pared to field test data and show a good match with the damping estimated under two-way cyclic loading.
1 BACKGROUND 3DFE analyses undertaken for a range of pile geom

etries and load eccentricities. The commercial soft
1.1 Introduction ware, Plaxis Monopile Designer, has been developed
to allow industry practitioners to simply apply the new
The global offshore wind market has grown rapidly
PISA design model (Brinkgreve et al. 2020). The
over the past decade with an average 30% increase
developments from the PISA project have accelerated
per year since 2010 (International Energy Agency,
the use of 3DFE for monopile design in industry. Des
2019). Monopile foundations, which are large diameter
pite the widespread use of 3DFE analysis for monopile
steel tubes driven into the ground, represent around
design, there is no industry standard approach to deriv
80% of all offshore wind turbine foundations installed
ing the required soil input parameters. Igoe and Jalil
to date and will likely remain the most common foun
vand (2020) developed an approach to derive the HSS
dation solution for offshore wind for at least the next
soil model parameters directly from CPT data, which
10 years. The geotechnical design of monopiles is gov
is used in this paper.
erned by the response to lateral loads and overturning
Offshore wind turbines are flexible structures which
moments. Initially, monopiles were designed using the
are dynamically sensitive and offshore loading is pri
industry standard API p-y method (API RP2GEO
marily cyclic and dynamic in nature. While significant
2014), which was adopted from the oil and gas indus
improvement in monopile design was achieved
try. However, in recent years it became evident that
through the PISA project, the project focused on
the API p-y method was not suitable for the design of
monotonic loading, to define ‘backbone curves’, which
large diameter monopiles (Kallehave et al. 2015). The
could be extended in the future to account for cyclic
recently completed PISA project (Burd et al. 2020a,b,
and dynamic effects. Accurately calculating the damp
Byrne et al. 2020a,b), was formed to develop new
ing of the structure is critical to prevent excessive
design methods for large diameter monopiles. The
fatigue damage over the lifetime of the structure. Con
PISA project proposed a new approach for the geo
sequently, the PISA “Soil Damping Project” was
technical design of monopiles which was calibrated
a supplementary piece of field testing within the PISA
from a suite of 3DFE analysis, which were validated
project, focused on obtaining the pile-soil response
against large-scale field tests. The new approach,
under two-way cyclic lateral loading at very low load
termed the ‘PISA design model’, is an enhanced form
amplitudes. A review of monopile damping is pro
of the p-y method, which is extended to include add
vided in Malekjafarian et al. (2021). This paper pre
itional components of soil resistance which are signifi
sents the development of a 3DFE modelling approach,
cant for low slenderness piles (2 < L/D < 6). Site
using CPT data to derive the soil parameters, in order
specific soil reaction curves can be calibrated from
DOI: 10.1201/9781003308829-146
973
to estimate monopile damping. The results from the 3 3D FE MODELING
3DFE are compared to the damping values measured
from the PISA field tests. 3.1 Soil model
Pile DM1 was modelled in 3DFE using the commer
2 PISA DAMPING FIELD TESTS cially available finite element software Plaxis 3D v21.
The soil elements are modelled as ten-node tetrahe
The PISA project field tests in sand were undertaken dral elements. For sand deposits, Igoe and Jalilvand
at a test site at Loon Plage, near Dunkirk in northern (2020) developed a procedure to calculate the soil
France. The Dunkirk site consists of a normally con model input parameters for the HSS soil model dir
solidated dense to very dense sand, with a top ectly from Cone Penetration Test (CPT) data. The
3 m comprising very dense fill material, below new approach adopted some widely used CPT correl
which is a dense natural sand. The ground water ations and was validated against a database of lateral
level was at about 5.4 m depth. The site was chosen load field tests. This procedure offers a consistent
for its extensive pre-existing site investigation his approach to develop all the required soil input param
tory, including in-situ field characterisation alongside eters and offers improved predictions of the piles
substantial suites of laboratory testing (Byrne et al monotonic response compared with the current state
2020b). The pile test program was designed to pro of the art Plaxis Monopile designer recommendations
vide a high-quality dataset which could be used for (Panagoulias, et al 2018).
validation of 3DFE modelling. The damping tests The CPT data was first used to derive the sand
involved undertaking 2-way cyclic loading on relative density using the expression proposed by
a single pile, designated DM1. The geometry for pile Kulhawy and Mayne (1990) as follows:
DM1 is provided in Table 1 below. The load test
setup for 2-way loading is described in Byrne et al
2020c (see Figure 1). The main loading is applied by
a hydraulic actuator located at a height, h,
10 m above the ground. The hydraulic actuator
reacts against a much larger (2.0 m) diameter reac
tion pile. To apply two-way loading and to ensure
that the hydraulic loading system is always in ten
sion, a back-stay system with a dead load was used.
Although the back-stay imposes a very small add
itional vertical load to the test pile due to the angle
of loading, this has a negligible effect on the lateral
response. Results from the PISA Damping Project
were presented in Beuckalaers (2017) and are used
for comparison with the 3DFE modelling approach
developed in this paper.
Table 1. Pile DM1 geometry.
Test D (m) L (m) L/D t (mm) h (m) where qt is the CPT cone resistance corrected for
pore water effects, pref is a reference stress (atmos
DM1 0.762 3.97 5.21 14 10.02 pheric pressure) taken as ≈100 kPa and n is a stress
exponent which varies between 0 and 1 depending
on soil type and stress level. The Over Consolidation
Ratio (OCR) was determined using the following
formula from Mayne (2001):
The OCR is used in the determination of relative

density, see Eq. (1), but is not input into the plaxis
models directly. Instead the coefficient of lateral
earth pressure, K0, is input into Plaxis, which is cal
Figure 1. Load test setup for DM1 pile (from Byrne et al. culated using the approach proposed by Mayne and
2020b). Kulhawy (1982)
974
material model (Young’s modulus, E, and Poisson’s
ratio, ν). Interface elements are added to the pile
shaft plate (both internal and external surfaces of the
pile). The interfaces have zero thickness and are
composed of 12-node elements which consist of
Igoe and Jalilvand recommend a maximum OCR of pairs of nodes which link the 6-noded plate elements
20 was assumed, to avoid unrealistically high K0 to the soil. The reduction in interface shear strength
values at shallow depths. The soil peak friction angle when slip occurs is accounted for using the strength
was determined using the CPT correlation by Bolton reduction factor Rinter which is selected depending
(1986): on the soil type and available geotechnical data. An
Rinter value of 0.7 was used for the sand-steel inter
face. The vertical model boundaries were set at ±6m
in the direction of loading and 4m in the perpendicu
lar direction, while the bottom boundary extended to
where ’0 cv is the critical state friction angle, was 13.2m as shown in Figure 2.
determined from lab test data at each site (’0cv = 32
degrees can be assumed in the absence of such data) Table 2. Soil Parameters used to model Pile DM1.
and p0 is the mean effective stress. The stiffness
parameters were derived from CPT cone resistance, Depth E50ref Eoedref Eurref f ψ G0ref K0
starting with the small strain shear modulus: [m] [kPa] [MPa] [MPa] [°] [°] [MPa] [-]
0 90.0 91.5 269.9 43.7 15.9 234.2 2.43

1.2 166.9 166.9 500.7 47.9 21.6 331.9 2.27
2.1 77.4 77.4 232.1 42.1 13.7 212.5 1.39
where α = 185 was used for Dunkirk Sand (verified 3.3 87.8 87.8 263.4 42.6 14.4 228.7 1.35
against shear wave velocity measurements). The 3.6 122.1 122.1 366.4 45.2 17.9 276.9 1.55
remaining stiffness inputs required for the analysis 4.2 80.3 80.3 241.0 41.9 13.4 217.2 1.16
were derived from the G0 adjusted for reference 6.3 67.2 67.2 201.7 40.5 11.6 195.9 0.98
stress levels (and therefore indirectly from the CPT 7.8 53.4 53.4 160.2 38.9 9.6 171.4 0.82
cone resistance) as follows (in MPa): 8.2 64.3 64.3 192.8 39.9 10.8 190.8 0.88
where m is the stress exponent taken as 0.5 in cohe

sionless soils.
Full details of the correlations used are provided in

Igoe and Jalilvand (2020). An unsaturated soil unit
weight of 17.7 kN/m3 and the reference shear strain,
γ0:7 ¼ 1:5 x 10-4 was used in the analysis. The
other soil parameters calculated are provided in
Table 2.
3.2 Pile model

The pile is modelled at full scale in half-space
assuming “wished-in-place” condition. The pile is
modelled by shell elements using a linear elastic Figure 2. Plaxis model for pile DM1.
975
3.3 Loading the measure response from the DM1 field test, as
report in Beuckalaers (2017). The calculated damping
In order to simplify the calculations and reduce the
ratios from the load-displacement and moment-
model run time, the loading was applied as a static
rotation responses are compared to the measure
analysis using the ‘plastic’ calculation type. The pile
response from the DM1 field test (as reported in
was loaded applying a prescribed load to the top of
Beuckalaers 2017) and presented in Figures 6 and 7
the pile stick-up (as seen in Figure 2). Four separate
respectively. It is evident that the 3DFE underpredicts
cyclic load levels of 4, 10, 15 and 20 kN were con
the damping measured from the load-displacement
sidered in the analysis. For each load level, 5 cycles
response at low load levels but matches the field tests
were applied, and the 5th cycle loop was used to
well at the higher load levels. For the moment-rotation
obtain the damping ratio from displacement and
response, the 3DFE predicts the damping well at low
rotation response extracted from the results of Plaxis
load levels but overpredicts at the higher load levels.
3D output. Each cycle was applied in three phases.
The first phase applied positive loading, the second
phase the pile is loaded in the opposite direction and
the third phase the load is brought back to zero. The
damping ratio was calculated using a Python code
that determines the area of the loop and also elastic
potential energy and then uses the equation below to
determine the damping ratio:
where Ediss = dissipated energy during a load cycle;

Eel = elastic stored energy. The global monopile
damping is divided into load-displacement damping
ratio ξv which is calculated from the horizontal load-
displacement response and, moment-rotation damp
ing ξψ which is calculated from the moment-rotation
response. The area of the loop is calculated using Figure 3. Hysteretic soil stress-strain response (from
Gauss’s area formula. The elastic energy is calcu Beuckelaers, 2017).
lated by finding the area of the shaded triangle
shown in Figure 3 in which σ indicates the horizon
tal force or the moment and ε indicates the displace
ment or the rotation. It is important to note that in
most cases the hysteresis loops do not close in the
unload-reload loops. For that case, the initial point at
which the loop starts is changed to the final point so
as to close the loops for the damping ratio calcula
tion. In order to determine the total damping from
both the horizontal displacement and rotational com
ponent, the dissipated energy and elastic energy
from both components are added together.
Figure 4. Load-Displacement response from 5th cycle of

4 RESULTS AND DISCUSSION each load level from 3DFE model of pile DM1.
The results from the 3DFE modelling of pile DM1 are

presented below. The load-displacement response and A comparison of the total damping ratios is pro
moment-rotation response from the 5th cycle for each vided in Figure 8. It is evident that overall the 3DFE
load level is shown in Figures 4 and 5 respectively. As is able to provide a reasonably good match for the
expected, the higher load levels result in larger areas damping ratio measured in the field tests at low load
enclosed in each cycle. The damping for each load levels. At higher load levels the 3DFE overestimates
level was calculated from these loops and compared to the total damping measured.
976
Figure 8. Comparison between 3DFE and field test total
damping ratios (field test data from Beuckelaers 2017).
Figure 5. Moment-Rotation response from 5th cycle of

each load level from 3DFE model of pile DM1. 5 CONCLUSIONS
This paper presents the results from 3DFE modelling

of monopile field experiments to determine founda
tion damping. The small-strain hardening soil (HSS)
model was adopted along with a CPT based
approach to calibrate the soil parameters developed
by Igoe and Jalilvand (2020). The CPT based
approach for deriving the soil input parameters had
previously been validated against monopile field
tests under monotonic loading. This paper examines
the use of the 3DFE approach under cyclic loading
to determine the monopile soil damping. The results
from the analysis are compared to field test data and
show a good match with the damping estimated
under two-way cyclic loading, particularly at low
load levels. At higher load levels the 3DFE tended to
over-predict the damping compared with the meas
ured values. The authors are currently undertaking
Figure 6. Comparison between 3DFE and field test damp further cyclic monopile field tests which can be used
ing ratios from load-displacement response (field test data for validation of the 3DFE damping approach.
from Beuckelaers 2017).
ACKNOWLEDGEMENTS
This publication has emanated from research con

ducted with the financial support of the Sustainable
Energy Authority of Ireland (SEAI) grant number
19/RDD/511.
REFERENCES
API (2014). “API RP2GEO: Geotechnical and Foundation
Design Considerations.”
Benz, T., Schwab, R., and Vermeer, P. (2009). “Small-
strain stiffness in geotechnical analyses.” Bautechnik,
86(SUPPL. 1), 16–27.
Beuckelaers, W.J.A.P. (2017) Numerical Modelling of Lat
erally Loaded Piles for Offshore Wind Turbines. Univer
sity of Oxford Ph.D. Dissertation. (January 2017),
Figure 7. Comparison between 3DFE and field test damp 485–491.
ing ratios from moment-rotation response (field test data Bolton, M. D. (1986). “The strength and dilatancy of
from Beuckelaers 2017). sands.” Géotechnique, 36(1), 65–78.
977
Brinkgreve, R. B. J., Engin, E., and Swolfs, W. M. (2018). Byrne, B.W., McAdam, R.A., Beuckelaers, W.J.A.P.,
“Plaxis 3D Materials Manual.” Burd, H.J., Gavin, K., Houlsby, G.T., Igoe, D.,
Brinkgreve, R., Engin, E., and Engin, H. (2010). “Valid Jardine, R.J., Martin, C.M., Potts, D.M., Taborda, D.M.
ation of empirical formulas to derive model parameters G. and Zdravković, L. (2020). “Cyclic laterally loaded
for sands.” Numerical Methods in Geotechnical Engin medium scale field pile testing for the PISA project”,
eering, (June), 137–142. Proceedings to the Fourth International Symposium on
Brinkgreve R, Lisi D, Lahoz M, Panagoulias S. (2020) Frontiers in Offshore Geotechnics in Houston, USA.
“Vali-dation and Application of a New Software Tool Igoe. D. and Jalilvand, S. (2020). “3D finite element mod
Implement-ing the PISA Design Methodology”, Journal elling of monopiles in sand validated against large scale
of Marine Science and Engineering, 8(6):457. https:// field tests.” Proceedings to the Fourth International
doi.org/10.3390/jmse8060457 Symposium on Frontiers in Offshore Geotechnics in
Burd, H.J., Taborda, D.M.G., Zdravković, L., Byrne, B.W., Houston, USA.
Houlsby, G.T., Gavin, K., Igoe, D., Jardine, R.J., Kirsch, F., Richter, T., and Coronel, M. (2014). “Geotech
Martin, C.M., McAdam, R.A., Potts, D.M., (2020), nische Aspekte bei der Gründungsbemessung von Off
“PISA Design Model for Monopiles for Offshore Wind shore-Windenergieanlagen auf Monopfählen mit sehr
Turbines: Application to a Dense Marine Sand”, Geotech großen Durchmessern.” Stahlbau Spezial 2014 –
nique, 70 (11), pp. 1048–1066, 10.1680/jgeot.18.p.277. Erneuerbare Energien, 83 (SUPPL. 2), 61–67.
Burd, H.J., Beuckelaers, W.J.A.P., Byrne, B.W., Gavin, K., Kulhawy, F. H., and Mayne, P. W. (1990)., Manual on Esti
Houlsby, G.T., Igoe, D., Jardine, R.J., Martin, C.M., mating Soil Properties for Foundation Design.” EPRI
McAdam, R.A., Muir Wood, A., Potts, D.M., Skov EL-6800.
Gretlund, J., Taborda, D.M.G. and Zdravković, L. Mayne, P. W., and Kulhawy, F. (1982). “K0 - OCR relation
(2020). New data analysis methods for instrumented ships in soils.” Journal of Geotechnical Engineering
monopile field tests, Geotechnique, 70(11), pp. Division, 108(6), 851–872.
961–969, doi.org/10.1680/jgeot.18.PISA.002. Malekjafarian, A., Jalilvand, S., Igoe, D., Doherty, P., (2020),
Byrne, B., MacAdam, R., Burd, H. and Houlsby, G. (2015). Foundation damping for offshore wind turbines on mono-
“New design methods for large diameter piles under lat pile supports: a review, Marine Structures, Volume 77,
eral loading for offshore wind applications.” Proceed May 2021, doi.org/10.1016/j.marstruc.2021.102937
ings to the Third International Symposium on Frontiers Minga, E. and Burd, H. (2019). “Validation of the PLAXIS
in Offshore Geotechnics in Oslo, Norway. MoDeTo 1D model for dense sand”.
Byrne, B.W., Houlsby, G.T., Burd, H.J., Gavin, K., Igoe, D., Page, A.M., Skau, K.S., Jostad, H.P. & Eiksund, G.R.
Jardine, R.J., Martin, C.M., McAdam, R.A., Potts, D.M., (2017) A New Foundation Model for Integrated Ana
Taborda, D.M.G. and Zdravković, L., (2020), “PISA lyses of Monopile-based Offshore Wind Turbines.
Design Model for Monopiles for Offshore Wind Tur Energy Procedia. [Online] 137, 100–107. Available
bines: Application to a Stiff Glacial Clay Till”, 70 (11), from: doi:10.1016/j.egypro.2017.10.337.
pp. 1030–1047, Geotechnique,10.1680/jgeot.18.p.255 Panagoulias, S., Brinkgreve, R. and Zampich, L. (2021)
Byrne, B.W., McAdam, R.A., Burd, H.J., Beuckelaers, W.J. “Plaxis Monopile Designer CE V21 Manual”.
A.P., Gavin, K., Houlsby, G.T., Igoe, D., Jardine, R.J., Robertson, P. K., and Cabal, K. (2014). “Guide to Cone
Martin, C.M., Muir Wood, A., Potts, D.M., Skov Penetration Testing for Geotechnical Engineering - 6th
Gretlund, J., Taborda, D. and Zdravković, L. (2020). Edition”.
“Monotonic lateral loaded pile testing in a stiff glacial clay Schnaid, F., and Yu, H. S. (2007). “Interpretation of the
till at Cowden”, 70 (11), pp. 970–985, doi.org/10.1680 seismic cone test in granular soils.” Géotechnique, 57
/jgeot.18.PISA.003. (3), 265–272.
978
© 2022 the Author(s), ISBN 978-1-032-31259-0
The use of CPTU for driven piles designed in a backfilled opencast ‘marl
hole’ in an important post-industrial revolution area within the UK
D. Illingworth & C. Burton
Piledesigns Limited, UK
L. Dhimitri & D. Ward

In Situ Site Investigation, UK
P. Shelton
Phil Shelton Geotechnical Consultancy Limited, UK
ABSTRACT: The Etruria Formation has historically been an important resource for the pottery and brick-
making industry in the Stoke-on-Trent area and was mainly exploited using opencast methods – or ‘marl
holes’ as they were commonly known. These ‘marl holes’ some of which have been reported to be more than
100m deep, have been subsequently backfilled generally with non-organic pottery, brickmaking and domestic
waste. This paper explores the redevelopment of land over a backfilled ‘marl hole’ at a site in Hanley, Stoke
on-Trent, Staffordshire, UK, for residential properties. The development posed a number of challenges for the
design of the foundations due to the historical legacy of the site. Driven steel tubular piles were chosen as
a suitable system as there is precedent with this foundation solution, which has been frequently adopted for
buildings within the area. However, some of the key challenges for this site included the presence of ‘high
walls’ along the edges of the opencast works where driven piles could be deviated off the sides along with the
uncertainty regarding socket lengths of piles driven into the underlying solid strata. To establish a 3D ground
model for the ‘marl pit’, underlying strata, piezocone tests (CPTU) were seen as the most cost-effective and
practicable method of site characterisation. Using this data, piles were installed, test/production piles were
installed, driven to a set/penetration into the bedrock criteria and then compared to this ground model.
1 INTRODUCTION productive Coal Measures at the centre of the coal

field and the Etruria Formation forming the outline,
This paper discusses the use of piezocone tests, particularly along the eastern limb of the syncline
CPTU for characterizing a backfilled opencast site (Rees et al., 1998).
for pile design. It is no coincidence that the six towns forming the
city of Stoke (Tunstall, Burslem, Hanley, Stoke,
1.1 Site description Fenton and Longton) are aligned along the eastern
The site is situated to the south of Hanley, which forms syncline limb above the Etruria Formation.
one of the six towns that make up the City of Stoke-on Unlike the North Staffordshire coal mining, the
-Trent, in North Staffordshire, United Kingdom. Etruria was generally exploited by opencast
The proposed development is made up of a series methods – or ‘marl holes’ as they were called. The
of low-rise (2–3 storey) self-contained residential outcrop of the Etruria, along the eastern limb of the
units. While the current topography shows syncline forms a line of many marl excavations of
a generally flat ground surface, this conceals the his varying sizes. At Fenton, less than 10 km south
torical legacy of this part of Stoke. south-east of Hanley, the marl pit is reported to have
been more than 100m deep.
The British Geological Survey (BGS) no longer
1.2 Comments on the geology regarding the recognises the term ‘marl’ and so now the Etruria
proposed development Marl is officially known as the Etruria Formation.
Published information from the BGS suggests this
The geological structure of the North Staffordshire proposed development site is underlain by Glacial
Coalfield, presented in Figure 1, is a syncline, plun Devensian deposits and in turn Carboniferous bed
ging towards the south-south-west, with the rock. These are mainly formed of a sequence of the
DOI: 10.1201/9781003308829-147
979
Furthermore, there have been many anecdotal
reports of industrial waste being dumped within the
marl holes during times when the controls on waste
disposal were less rigorous than now.
The following photograph presented in Figure 2
shows the nature of the excavations from a nearby
well documented marl hole at Daisy Bank. While
these pits were active, the high walls were steep prob
ably between 45° and 50° overall, but locally up
to 60°.
Figure 1. Geological map of North Staffordshire, modified

from Millott (1937).
Etruria Formation, which is of Middle Carboniferous

age, from 308 to 319 million years ago. It comprises
of mudstones, siltstones and sandstones that overlay
part of the North Staffordshire productive Coal
Measures (Pennine Upper Coal Measures).
Figure 2. Air photos depicting the full extent of the Daisy
Reference to historical data indicates the site sits
Bank Brickworks, English Heritage (1927).
partly over an old marl pit. Data for an adjacent site
indicates that the main extent of this old pit, known as
the Mousehole Marl Pit, was to the north-east of the In respect of future development over the marl
current development. The remainder of the site con holes, probably the worst case is construction over
sidered in this paper also appears to straddle the side the marl hole sides, often termed the ‘high walls’.
of this old pit.
The Etruria Formation has a few unique proper
ties that have made it a very important resource for 2 INTERFACE DELINIATION BETWEEN THE
the pottery and brickmaking industry. Firstly, it has MADE GROUND AND MUDSTONE
a relatively high natural carbonate content common
with non-marine semi-lacustrine ‘marl’ deposits. 2.1 Ground conditions
However, the source material for the Etruria was Field work was undertaken in two phases. The first
eroded material from an older Midlands volcanic phase took place in September 2019 where 9 boreholes
sequence. Therefore, it has a relatively high iron were performed, six to shallow depths to nearly 6m
content that gives rise to the natural colour vari and three deeper boreholes to a maximum of nearly
ations. While the formation is mainly of clay and silt 24m. On the deepest borehole, groundwater was
grain size, the Etruria also includes regular sandstone encountered at 16.5m. Ground conditions at the bore-
layers termed ‘esplays’. These are laterally discon hole locations comprised a layer of made ground over
tinuous with few recognised as named horizons. laying natural strata considered to be Upper Coal
No published records are available for the Mouse- Measures Formation.
hole Marl Pit, which is known to extend beneath the Most of this initial investigation terminated in the
proposed development. Following the exploitation of made ground and no Glacial deposits were found.
the Etruria Formation, the marl holes were backfilled The made ground was mainly granular loose to
with non-organic pottery and brickmaking waste. medium sandy gravel with occasional cohesive
Much of the backfilling was undertaken during the material present in thin layers, typically soft to firm
1960s, 1970s and 1980s (or before), when domestic clay. Mudstone was only encountered in a couple of
waste was largely ash, glass and metal goods and it boreholes found below the made ground and was
is expected that these products are also present. described as very weak rock.
980
Figure 3. Typical CPTU profile for this site.
During the second phase of site investigation, 52

CPTUs were carried out to a maximum depth of
31.5m.
2.2 CPTU interpretation

The delineation of the interface between the made
ground and the mudstone was assessed by the cone
resistance, qc, friction sleeve, fs, porewater pressure,
u2 and inclination. Figure 3 represents a typical
CPTU profile from this site.
The typical site level was 145.5mAOD. CPTU
results indicate bedrock from between 3-30m depth.
Although CPTUs provided a site coverage, it was
found that there is no evidence for bedrock between
128 and 124mAOD.
The increase in qc and fs with greater depths were
immediately noticed when cone was penetrating the
mudstone. Results from 42 CPTUs with a penetration
of more than 1m into this strata suggest that within
mudstone qc > 5.5 MPa, fs > 250 kPa and u2 has
a significant drop to negative measured values. The
negative porewater pressure indicates suction.
Smoother inclination measurements are another indi
cation that Mudstone is a uniform deposit.
To allow further review of the CPTU results Figure 4. Mudstone results plotted on Robertson 1990
within the mudstone, the data points recorded were SBT chart.
plotted in the Robertson 1990 soil behaviour type,
SBT chart, as presented in Figure 4.
The majority of results for Etruria Formation fall in develop the ground model for this site, which is also
zones 8-9, sand to clayey sand and very stiff fine included in the paper.
grained, respectively. It is also noticed that some A simplified graph of friction ratio, Rf (%) for
results plot in zones 3-4, which represent clay and silt mudstone to show the variation of results is pre
mixtures. sented in Figure 6. Black graphs present the min
Figure 5 presents a typical CPTU cross section for imum and maximum Rf within the deposit,
this site. The data shows that the rockhead was meanwhile the red graph presents the averaged
encountered at shallower depths in the south west of results. It is seen that this parameter may vary
the site. In the eastern boundary of the site, mudstone greatly in mudstone and cannot be trusted as the
occurs at greater depths. These results are used to main parameter to interpret it.
981
Figure 5. Cross section showing the mudstone coming in at a shallower depth to the south and much deeper to the north of
the site.
Although CPTU provided a good site coverage, it

was found that there were no probes relating to
datum levels of between 128 and 124mAOD. Conse
quently, CPTU data were provided within two zones
and between approximate levels of 141 to
128mAOD and 124 to 115mAOD. These are
referred to as the upper and lower zones. The strati
graphic plots indicate distinct differences between
the upper and lower zones. Although the results
show some consistency several distinct layers can
also be seen. In particular a more cohesive layer,
with typical Rf = 6-8% between 140-138mAOD, and
then some more sandy horizons averaging Rf =1%
between 133-130mAOD. The lower zone indicates
significantly more variation with values of Rf vary
ing from as low as 1% to as high as 9%.
After reviewing the scatter in these results it is
suggested that the big variation in Rf results within
the deposit can be mainly explained by lithological
variations within the marl.
2.3 Discussion of CPTU results

Findings from CPTU tests regarding the delineation
of the interface between the made ground and the
mudstone for this site are in line with historical facts
and previous experiences from investigations of
backfilled opencast ‘marl holes’. Figure 6. Friction Ratio (Rf) variation.
982
Among 52 tests carried out in total, only a few
tests were refused at shallow depths, due to obstacles
encountered in the first meters of backfilled material.
42 CPTU tests penetrated through the mudstone, 18
of which with a penetration of more than 2m into
this material, which allowed us to review the results
and the findings of this paper were crucial for the
ground model and foundation design. Etruria Forma
tion, interpreted as mudstone behaves as very stiff
fine grained material, plotting in zone 9 in Robertson
1990 SBT when qc>5.5 MPa and fs >250 kPa and as
sand to clayey sand when qc >15 MPa and fs >500
kPa, plotting in zone 8.
Porewater pressure can be another important meas
ured parameter to interpret Mudstone, due to the
sudden drop in the interface with made ground and
large suction developed as the penetration continues.
Another important indication of the penetration in mud-
stone is the inclination in both directions, which
becomes uniform soon after the start of penetration Figure 7. Marl Hole Contour Profile 2D.
through this deposit.
3 GROUND MODEL
The ground model for this site was developed based

on CPTU results that had penetrated into the natural
strata beneath the made ground.
For some tests where sensible assessment of this
rockhead boundary was unclear, or the CPTU test
appeared to have deviated or refused on an obstruc
tion within the made ground, a repeat test was car
ried out.
Based on the CPTU results a model of the pit was
constructed using the contouring package Surfer.
Surfer interpolates irregularly spaced XYZ data into
a regularly spaced grid. The grid data can then be
interpreted using gridding and mapping modelling Figure 8. Marl Hole Contour Profile 3D.
tools to produce 2D or 3D maps such as contour,
shaded relief and surface maps. In addition, calcula
tions such as bilinear interpolation can be carried out western edge of the marl pit the average gradient
where the Z value at the XY location can be calcu appears to be about 50 degrees although locally this
lated from the nearest four grid nodes. In Figures 7 is indicated as increasing to almost 65 degrees.
and 8 the marl hole contour profiles is shown, These back analysed gradients are comparable with
including the pile positions, shown in red. those previously indicated for marl pits in the area.
The initial Surfer plots in Figures 7 and 8 provide From the results of the CPTU tests a suitable
a profile of the western high wall of the marl hole. ground model showing the profile of the made
This model was used to check that installed piles ground/bedrock interface across the site has been
have a suitable penetration into the natural strata. established in relation to the proposed pile locations.
In reviewing the updated ground model, the data The CPTU testing has provided data which indicates
shows that rockhead occurs at 144-143mAOD along a good profile for the interface, and which looks rea
the western boundary of the site. There is a fairly sonable when comparing all the available informa
shallow gradient towards the east until typical rock- tion. Where CPTU results looked anomalous, they
head falls to 141-140mAOD. Thereafter the plot have been discounted.
shows a steeper incline representing the western In reviewing and analysing the CPTU data, at least
edge of the marl hole. Rockhead falls to about 123 1-2m bedrock penetration was typically considered
122mAOD and then thereafter the gradient is more sufficient to provide confidence that the probe attained
gradual towards the eastern site boundary. a suitable penetration into the natural material.
Nearer the eastern site boundary rockhead levels This surfer ground model has been used to review
are typically 120mAOD although with some further the toe depths of the installed piles against the inter
reduced levels. Reviewing the steeper part of the polated mudstone levels. It may be noted that the
983
surfer ground model has been considered as more ground model, which allowed a suitable piling solu
reliable than particular pile driving records. tion to be installed for the site.
From reviewing 42 CPTUs, it may be suggested
that Robertson 1990 normalized SBT chart is a good
4 FOUNDATIONS classification method to interpret the mudstone in an
accurate way. It provides more consistent results,
The foundation proposal was to adopt top driven indicating the behaviour of the deposit as overconso
tubular steel piles with a 178mm diameter tubular lidated cemented weak rock, plotting data points for
steel section considered. The driving system this deposit in soil zones 8-9.
employed was using a 5.5 tonne hammer. It may also Possible inconsistencies in CPTU based classifi
be noted, based on experience, that it was expected cation should be attributed to physical complexity,
that the proposed newly installed piles would which may affect the overall behaviour of the
achieve a penetration at least comparable to the deposit especially in the upper zone, where some
CPTU tests. soils show a more clay-like behaviour and some
From the piling works a typical penetration of others a more sand-like behaviour.
approximately 3 to 4m into natural strata was
achieved. This was considered appropriate as it is
anticipated that for the proposed piles, using an effi REFERENCES
cient energy input, a greater penetration compared to
the CPTU tests, should be achieved. Lunne, T. et al. 1997. Cone Penetration Testing in Geotech
nical Practice. Spon Press, London, UK.
As noted in monitoring the installation data all
Millott, J. O’N. 1937. The Coal Seams of North Stafford
piles used within the foundation system showed rea shire. Physical and Chemical Survey of the National
sonable penetration into the mudstone. Only one pile Coal Resources No. 39, Department of Scientific and
indicated an anomaly, taken as possible deviation, Industrial Research.
and which was subsequently replaced. Robertson, P.K. et al. 1986. Use of Piezometer Cone Data.
Proceedings of the ASCE Speciality Conference In Situ
‘86ʹ: Use of In Situ Tests in Geotechnical Engineering.
5 CONCLUSIONS ASCE (American Society of Civil Engineers),
New York, NY, USA.
Rees, J.G. et al. 1998. Geology of the country around
CPTU tests were used to provide a cost effective
Stoke-on-Trent: Memoir for 1:50000 geological sheet
method to profile and establish a ground model for 123 (England & Wales). Memoirs of the Geological
the backfilled opencast marl hole in Stoke on Trent. Survey of Great Britain, England and Wales.
There was an initial concern on using CPTU to get Robertson, P.K. 1990. Soil Classification Using the Cone
to depths due to equipment limitations to penetrate Penetration Test. Canadian Geotechnical Journal 27(1):
through hard strata and rocks. 151–158.
At the end of the site investigation program, Robertson, P.K. and Cabal, K.L. 2015. Guide to Cone
CPTUs provided effective information for the Penetration Testing 6th Edition. Gregg Drilling & Test
ing Inc, Signal Hill, CA, USA.
984
© 2022 the Author(s), ISBN 978-1-032-31259-0
Consolidation settlement prediction using cone penetration testing

M. Kermani
Golder Associates Ltd., Montreal, Canada
F. Esford
Golder Associates Ltd., Vancouver, Canada
ABSTRACT: A reduction in the local groundwater table surrounding a mine, in the Abitibi-Témiscamingue
region of Quebec, Canada, has led to dissipation of porewater pressure and displacement of the ground surface. The
mechanical behavior of the site’s fine-grained soil has been studied in the laboratory (consolidation tests) and in-situ
tests, particularly cone penetration tests (CPT). Additionally, displacements of the ground surface have been moni
tored for over 10 years. The consolidation settlements due to the groundwater table drawdown were calculated based
on tip resistance values measured during the CPT. By comparing the calculated magnitudes and trends of vertical
displacement with measured values, the consolidation calculation method using CPT data was validated. At some
locations, there was a good match between the calculated and measured trends. At other survey monuments, the
measured displacements were greater than calculated values. It was concluded that additional mechanisms were con
tributing to the displacements. The possible influence of these mechanisms varies spatially within the studied site.
1 INTRODUCTION displacement monitoring to assess and estimate the

consolidation settlements. Using cone penetration
1.1 Problem statement testing data, the validity of the Fellenius (2011)
method for estimation of the tangent modulus based
The studied area is a mine site in the Abitibi-
on CPT tip resistance values, and the Janbu (1967)
Témiscamingue region of Quebec, Canada. During
method for estimation of consolidation settlements
mine production, conditions both underground and
for the studied site are evaluated. To estimate the ini
on surface of the mine began to change. The ground
tial and final vertical effective stresses, the porewater
water level declined and inflows to the underground
pressure before drawdown and actual porewater pres
workings increased.
sure were considered. The vertical and lateral coeffi
Subsequently, investigations were carried out to
cients of consolidation were estimated from
understand the conditions and mechanisms associ
laboratory oedometer tests and in-situ CPT dissipa
ated with the ground surface movements.
tion tests. Then by using Terzaghi’s one-dimensional
As mining progresses, the owner continues to moni
consolidation theory, settlement versus time plots
tor the area, including measuring: surface displace
were generated. For locations where it was estimated
ments, bedrock deformations, elevation of the phreatic
that primary consolidation was achieved by 95%,
surface (through monitoring wells, vibrating wire piezo
secondary compression settlements were also added
meters, and CPT dissipation tests.), water infiltration
to the settlement prediction time plots. Note that in
into the underground workings, blast monitoring, and
this article all stresses are effective stress, unless
micro-seismic activity. Data is regularly reviewed and
otherwise noted.
potential interactions between mining activities and site
conditions are assessed by a multidisciplinary team.
For the sake of this study, samples, in-situ test 1.3 Studied site’s stratigraphy
data, and displacement monitoring data recorded at
The studied site is in the Abitibi-Témiscamingue
fixed survey monuments have been provided by
region of Quebec province in Canada. The strati
Agnico Eagle Mines Limited (AEM).
graphic profile generally consists of (from top to
bottom):
1.2 Objectives
‒ A desiccated layer of firm to stiff, brown, clay or
This paper makes use of the gathered data during the silty clay, typically 2 m in thickness. The desic
past 10 years, particularly the in-situ testing and cated layer is associated with the annual
DOI: 10.1201/9781003308829-148
985
fluctuation in the phreatic level and freeze-thaw increases. Janbu proposed an empirical equation for
cycles. the change in modulus with the stress level.
‒ A soft to firm, sensitive, grey, lacustrine (Lake
Barlow-Ojibway clay), varved silty-clay to clay,
of low to high plasticity, with a water content
close to or above the liquid limit. Varves of silty where: Mt = tangent modulus; σr = reference stress;
clay to clay are reported to vary between 2 and σ= effective stress in the direction of strain; m =
20 mm thick and silt between 1 and 15 mm modulus number; α= dimensionless stress exponent.
thick. The undrained shear strength typically Both m and α depend on the material type and can
varies from 14 to 45 kPa, and sensitivity varies be obtained from laboratory tests on high quality
from 5 to 40. The clay is primarily saturated. It samples or in-situ tests (CPT in this study).
can reach thicknesses up to 30 m.
‒ A cohesionless, grey, silt and/or sand layer or
mixture of silty sand or sandy silt, with occa
sional cobbles. The thickness varies from 2 to
25 m. The silt/sand was initially reported as
loose to compact and saturated however, due to
the reduction of the groundwater level, the
majority of the silt/sand layer in the vicinity of
the mine is now unsaturated. As a result, the
compactness of this material has increased. It is
now reported to be compact to dense at depth.
‒ A compact to very dense till unit that is variable
in composition is discontinuously present. In
general, the till is well graded and composed of
sand and gravel, some silt and some clay, and
occasional cobbles and boulders.
‒ Bedrock.
Figure 1. Vertical displacement data from surveying com
bined with satellite data at a specific location.
2 METHODOLOGY This approach has been used to calculate the

magnitude of future, primary consolidation settle
2.1 Field observations ments (after the date each CPT was pushed). These
CPTs were generally pushed after partial or full
The field observation data on the displacements drawdown of the water table had occurred, which
include survey monuments and inclinometers. As in turn altered the initial in-situ conditions. There
most of the survey monuments were installed after fore, this approach was found to be of interest
the onset of displacements, they have not captured because it considers the soil conditions, as well as
the complete deformation history. Therefore, data porewater pressure at the time the CPT was pushed
from the satellite imagery analysis (InSAR) at the and allows estimating the settlements going forward
location of each survey monument was also used to with a higher precision.
supplement the measurements. Figure 1 shows an Two different equations are derived from Janbu
example of how the displacement from satellite (1967) method, one for the fine-grained material, and
(PALSAR and TSX satellite data) measurements one for granular materials. The settlement of the clay
has been added to survey monument data to esti layer is calculated according to the following equation:
mate the total displacement since the onset of
movements. Conversely, when both sets of data
were available, the surface measurements were
used to verify the precision of the satellite data.
Note the satellite displacement data is measured
along the line of sight. The PALSAR data had
and the settlement of the silt and sand layer are cal
a grid size pixel spacing of 15 m and TSX had
culated according to the following equation:
a 6 m spacing.
2.2 Janbu (1967) tangent modulus method

The tangent modulus method was developed by
Janbu in the 1960’s upon the assumption that the soil Where: ε = strain induced by an increase in effective
modulus of material can be estimated by dividing stress; mr = recompression modulus number; m =
stress by strain. The modulus decreases as the stress modulus number; σ0 = initial effective stress (kPa);
986
σ1 = final effective stress (kPa), and σp = preconsoli saturated and wet unit weights are neglected, the
dation stress (kPa). final effective stress after dissipation of pore pres
In order to obtain an upper bound for the settle sure would equal the current total stress (the red
ments, a simplifying assumption was made that the dashed line). For each CPT, the red curve showing
soil was in a normally consolidated state (OCR=1). the preconsolidation pressure was found using CPT
By this assumption, the modulus number m remains data. The preconsolidation pressure was calculated
the only compressibility material property to esti for each 10 cm increment of the soil using the fol
mate the consolidation settlement. As explained in lowing correlation:
the next section, Fellenius (2011) has proposed an
empirical equation that correlates the modulus
number to the CPT tip resistance.
where σv0 is the vertical total stress and the parameter
2.3 Fellenius (2011) method Nσt is assumed to be 3.4 for Eastern Canada clays
(Demers and Leroueil, 2002). Then the values were
Fellenius (2011) suggests the following equation for validated by laboratory oedometer tests and field vane
the estimation of the modulus number based on tests (FVT). Equation 6 was used to estimate the pre
basic CPT data: consolidation pressure from the FVTs (Mesri, 1988).
where SuðmobÞ is the undrained shear strength mobil

ized on the failure surface in the field. According to
where: a = empirical modulus modifier which Bjerrum (1973), the shear strength from FVT is
depends on soil type; qt = unadjusted cone resistance close to SuðmobÞ for clays with plasticity indexes
(but corrected for pore pressure); σv = the vertical close to 20%, which is the case for the studied site.
effective stress; σr = the reference stress, here 100
kPa; and K0 = the coefficient of lateral earth pres
sure at rest. The term 3=ð1 þ 2K0 Þ varies between
1.2 and 1 for a soil with K0 between 0.5 and 1. This
term was assumed to be unity, which provides con
servative values for the modulus number.
The modulus modifier, “a” was estimated based
on the Soil Behaviour Type (SBT) determined from
the CPT data (Table 1).
Table 1. Modulus modifier values based on SBT (from

Massarsch et al.,1997).
SBT (Jeffries and Davies, 1993) Modulus modifier, a
0, 1 and 2 3
3 5
4 12
5 15
6 22
7 28
8 35
Figure 2 schematically shows the state of stresses Figure 2. Schematic of the state of stresses considered for
in the clay unit at the location of a given CPT. The consolidation calculations.
total and effective stresses at the moment the CPT
was pushed are shown in the figure. The soil unit The consolidation settlement is a function of the
weight was estimated using Shelby tube samples, as difference between the actual and final effective stress
well as interpretation of CPT data. The pore pressure (the difference between the green line and red dashed
at the moment the CPT was pushed needs to be con line. For the zone between the actual effective stress
sidered for the calculation of the actual effective and the preconsolidation pressure, the calculations
stress. CPT dissipation tests provided reliable values were done using the recompression modulus mr (indi
for the calculations. If the difference between cated by the grey arrows), and for the zone
987
between preconsolidation pressure line and final where S is the magnitude of secondary compression,
effective stress, the modulus number was used, as pre- Cα is the secondary compression index, L0 is the ini
sented in equation 1 (indicated by the orange arrows). tial thickness of the consolidating layer, tp is the dur
ation of the primary consolidation stage, e0p is the
void ratio at the beginning of secondary compression,
2.4 Distribution of settlement through time
and t is the time interval of interest that begins from
As explained in the previous sections, the magnitudes the start of consolidation. Cα was estimated using one-
of settlements were calculated using the Janbu (1967) dimensional laboratory consolidation test results on
and Fellenius (2011) methods. To estimate the distri- high-quality large diameter samples carefully collected
bution of the calculated settlement through time, the from the same geologic formation at another site in the
rate of consolidation needed to be considered. The region.
rate of consolidation was calculated according to Ter- The determination of the beginning of secondary
zaghi’s well-known one-dimensional consolidation compression has always been controversial among
theory. The time factor Tv was calculated from the geotechnical engineers. When calculating secondary
below equation: compression settlement during a specific period of
time (between times t1 and t2 ), when the secondary
compression has started before t1 , the following
equation can be derived from equation 8:
where t is the time to obtain a certain degree of consoli

dation, Hdr is the length of the longest drainage path,
and cv is the vertical coefficient of consolidation. The Therefore, the time of the beginning of secondary
consolidation ratio U (%) was calculated using empir- compression does not influence the results.
ical equations (Holtz et al., 2011) as a function of Tv .
Considering the varved nature of the clay, the coeffi

cient of consolidation in the horizontal direction is 3 RESULTS
clearly higher than the vertical direction. In order to
adopt a representative coefficient of consolidation for Figure 3a shows the initial and final effective vertical
clay, both laboratory oedometer tests and CPT dissipa- stresses along with the preconsolidation pressures for
tion tests were considered. A slower rate, cv =13.6 m2/ each 10 cm soil increment. The preconsolidation pres-
yr, was determined from average vertical coefficients sures from CPT were validated with laboratory oed
of consolidation from several laboratory oedometer ometer test, as well as field vane tests (FVT). There is
tests. A faster rate, cv = 63.1 m2/yr, was adopted from generally a good agreement between CPT, FVT and
average radial coefficients of consolidation based on oedometer test results. At depths between 6 m and
the interpretation of CPT dissipation tests. Plotting the 9 m, FVT tends to overestimate the σ0p , possibly
estimated settlements versus measured displacement because of higher plasticity of clay in these depths. To
during 10 years of monitoring showed that the slower carry out the consolidation calculations, the preconso
rate provided a better match with the measurements. lidation pressure from CPT data was used.
This is coherent with the mechanism of consolidation As shown in equations 2 and 3 and Figure 2, the
in the site, which is ground water table drawdown and compression of clay is a function of the effective
the downward drainage of water. If the consolidation stress change from the preconsolidation pressure to
calculation method presented in this article is to be the final effective stress, and the modulus number.
used for other sites, the consolidation mechanism and Figure 3b shows the modulus numbers calculated
the direction of drainage needs to be considered. For using equation 4 based on the CPT data for each
instance, if wicked drains are installed, radial coeffi- 10 cm of soil increment. An advantage of the con
cient of consolidation could be more representative of solidation settlement calculation method presented
site conditions. in this article is that at each elevation a specific soil
modulus is considered and the variability throughout
2.5 Secondary compression the soil column is taken into account. A step change
is observed in the calculated modulus numbers at
Where the consolidation was estimated to have been depth 11 m. This is because the SBT changes from 3
complete by 95%, the secondary compression settle- to 1 at this depth, meaning that the soil becomes
ments were also added to the consolidation settle- more plastic, and the modulus modifier, a, changes
ments, using the conventional method: from 5 to 3 (see Table 1).
Considering the equation 2 and relating the modu
lus number to the conventional consolidation test
parameters, the modulus number can be estimated
directly form oedometer test results using the follow
ing equation:
988
The consolidation calculations estimate the settle
ments, from the moment the CPT was pushed and
considers the geotechnical properties of the encoun
tered soil at the time of the test. For each CPT the
where Cc is the compression index and e0 is the initial settlement vs. time predictions are presented with
two different consolidation rates (faster and slower)
void ratio. The modulus numbers estimated based on
consolidation test results are shown along with the considering two different coefficients of consolida
values from CPT tests (Fellenius 2011 equation) in tion, as explained in section 2.4.
Figure 3b. The laboratory modulus numbers are slightly The 2011 CPT with the slower rate has captured
the general trend of the settlements from the test
greater than CPT values. But there is generally a good
match between field and laboratory values. Note that date. However, the measured total ground displace
for the consolidation calculations, Fellenius (2011) was ment has exceeded the calculated total settlement.
The higher cv (from CPT) appears to overestimate
used to estimate the compression modulus m in clay.
The recompression modulus number mr in clay, as well the rate of consolidation. The consolidation calcula
as both moduli for silt/sand units were estimated based tions are carried out for the 2017 CPT with the
on laboratory consolidation tests. updated pore water pressure regime, and soil geo
Figure 3c shows the cumulative ultimate settlement technical characteristics at the time the cone was
calculated for each depth. As expected, the majority pushed. It appears that the 2017 CPT data has cap
of settlements occur at the depths where the clay pre tured well the ground displacements for the years
consolidation pressure is close to the current effective following the test. The 2017 CPT was located closer
stress (6 m to 20 m). Additionally, it can be observed to the survey monument (15 m vs. 25 m for the 2011
that greater settlements are calculated between depths CPT). The fact that the 2011 CPT provided results
11 m and 20 m, where the modulus numbers are which deviated from the measured values could be
smaller compared to depths 6 m to 11 m. attributed to the spatial variability of the stratigraphy.
Figure 4 shows the consolidation calculation vs. Note that for this location, based on the dissipation
time for two of the studied CPTs from the date each tests, the 95% primary consolidation has not been
CPT was pushed, along with the measured vertical dis attained. Therefore, the secondary compression
placement at a nearby survey monument. The survey settlements are not considered.
monument readings show some variations due to The mismatch between measured and predicted
freeze-thaw cycles but minimum values (fall readings) settlements was observed for some monuments. The
show the vertical displacement trend and magnitude.
Figure 3. Example of calculation of a) initial and final state of stresses, b) modulus number, and c) ultimate settlement.
989
discrepancy was concluded to be associated with, of a site where a decline in the phreatic surface has
but not limited to factors such as the following: led to consolidation settlements was presented. The
Stratigraphic variation: The CPTs were some results showed that the method can capture the con
times up to 50 m away from the reference survey solidation settlement-time curves. A discussion is
monument. Several drilling and site investigation presented on the reasons why for some survey monu
programs on the site have shown that the stratig ments, the match between measured and predicted
raphy is highly variable. So, the stratigraphy encoun settlements was less accurate.
tered in the CPT may not necessarily be the same as This method can be used to create a contour map of
the subsurface layers below the survey monument. the predicted consolidation settlements for a site with
Further phreatic surface drawdown: The overall low stratigraphic variability. It is recommended to
ground water table at the site was assumed to remain space out the CPTs to cover the zone of interest, and
constant from the moment the study was carried out. to carry out more CPTs close to the sensitive structures
Any additional drawdown of the water table (in the to obtain more reliable data. A combination of CPTs
partially drained areas) could increase the estimated and traditional boreholes is recommended to collect
settlements. high quality samples for laboratory testing to validate
Lateral movements: Lateral movements could the calculations parameters. As CPTs can be carried
contribute to the vertical displacements by reducing out faster and at a lower cost, they should be used to
the soil confinement. capture the variability in site stratigraphy.
Development of suction in soil: It was assumed in It should be noted that the studied case is different
this study that the final porewater pressure in eleva from conventional consolidation problems where
tions above the phreatic surface will become zero. adding a load forces the porewater out of the soil
Soil-Water Characteristic Curve testing indicated that matrix and leads to an increase in the effective stress.
most of the clay layer will remain in a saturated state In the studied case, the increase in effective stress is
in long-term. However, considering that the phreatic caused by a decline in groundwater table. Therefore,
surface is below the bottom of the clay layer, it is pos as the groundwater table is below the bottom of the
sible that negative pore pressures (suction) develop in clay layer, negative porewater pressure (suction)
clay, which would increase the final effective stresses could be generated in the clay which may lead to
beyond the estimated values (higher than the actual greater effective stress and larger deformations. This
total stress). Higher final effective stress would lead aspect is presently being evaluated by measuring
to greater deformations in clay, than what is calcu potential suction in fine-grained soil layers on the site.
lated considering a zero final porewater pressure. The
installation of vibrating wire piezometers in clay
would allow estimating the vertical effective stress ACKNOWLEDGEMENTS
with more precision.
The authors acknowledge the contribution of Agnico
Eagle Mines Limited (AEM) in providing the infor
mation, support and guidance in the studies associ
ated with this article. Further the authors would like
to acknowledge the contributions of a former col
league, Gerd Janssen, in developing this approach
for estimating the settlements.
REFERENCES
Fellenius, B. H., 2011. Basics of foundation design. Elec
tronic edition. www. Fellenius.net, 362 p.
Holtz, R. D., Kovacs, W. D. Sheahan, T. C. 2011. An
introduction to geotechnical engineering. Second
Edition.
Jeffries, M.G. and Davies, M.P. 1993. Use of CPTu to Esti
mate Equivalent SPT N60, Geotechnical Testing Jour
nal, ASTM ,16: 458–168
Massarsch, K.R., Westerberg, E., and Broms, B.B., 1997.
Footings supported on settlement-reducing vibrated soil
Figure 4. Example of measured and predicted settlement nails. 14th, Hamburg 97, Vol. 3, pp. 1533–1539.
vs. time plots. Mesri, G. 1988. A reevaluation of SuðmobÞ ¼ 0:22σ0p using
laboratory shear tests. Canadian Geotechnical Journal.
Vol. 26, 162–164
4 CONCLUSIONS Demers, D., & Leroueil, S. 2002. Evaluation of preconsoli
dation pressure and the overconsolidation ratio from
In this article, a consolidation settlement calculation piezocone tests of clay deposits in Quebec. Canadian
method using CPT data was presented. A case study Geotechnical Journal, 39(1), 174–192.
990
© 2022 the Author(s), ISBN 978-1-032-31259-0
3D FE derivation of CPT based soil reaction curves for monopile lateral

static design in sand
Louis-Marin Lapastoure
Department of Civil, Structural & Environmental Engineering, Trinity College Dublin, Dublin, Ireland
Gavin and Doherty Geosolutions Ltd, Dublin, Ireland
David Igoe
Department of Civil, Structural & Environmental Engineering, Trinity College Dublin, Dublin, Ireland
ABSTRACT: This paper aims to offer novel correlation between all 4 components of PISA soil reaction
curves for monopile modelling and Cone Penetration Testing (CPT) data. This approach requires minimum user
inputs and offers a consistent approach to develop site-specific soil reaction curves when CPT data are made
available. The procedure developed by Igoe and Jalilvand was used to derive small-strain hardening (HS-small)
soil model parameters from a range of CPT profile. HS-small parameters are then used to model the monopile
response to static lateral loading in sand using Plaxis 3D. This procedure has been validated by the authors
against a database of large-scale field tests including PISA tests in the Dunkirk marine sand. Soil reaction
curves were directly extracted from the Plaxis models then applied into Timoshenko beam elements to match
the 3D FE response and validated the extraction process. It is demonstrated that current CPT based correlations
do not match the shape nor the ultimate reaction of 3D FE extracted soil reaction curves. It is also shown that
all four components of soil reactions are required to accurately model monopiles which are short and rigid.
1 INTRODUCTION 2019). It offers the option to extract site-specific soil

reaction curves from 3D finite element models. How
1.1 Context ever, this can be seen as computationally costly and
requires careful calibration of the 3DFE soil input
Monopiles are the main foundation type supporting
parameters.
about 80% of the installed offshore wind turbines to
Soil strength and stiffness parameters are often
date in Europe (Wind Europe, 2020). The traditional
determined using correlations with Cone Resistance
industry design approach for monopile was adopted
Testing (CPT) results, and therefore researchers have
from the oil and gas industry. The ‘p-y’ approach
been looking at correlating soil reaction curves with
recommended in the main design standards (API
CPT directly. Broms (1964) originally proposed rela
2011, DNV 2013) was validated against a rather
tionship between ultimate lateral reaction, pu, as
small database of long slender piles with diameter
a function of the cone resistance, qc, in silica sands
smaller than 1 m. On the contrary, monopiles are
as per equation (1) where D is the pile diameter, p’
now designed with a low slenderness ratio (L/D ≈ 3)
is the mean effective stress and pa is a reference
and a large diameter up to 10 m. Since 2014, it is
stress taken as 100 kPa.
warned that the API approach has “not been cali
brated for monopiles with larger diameters and are in
general not valid for such monopiles” (DNV 2014).
The recently completed PISA project was devel
oped with the aim of improving on the traditional API
design approach for monopile. In the PISA design
model, pile-soil interactions are not limited to the dis Suryasentana and Lehane (2014) developed new
tributed lateral reaction (p-y curves) but also include p-y formulations for piles in sand based entirely on
distributed moment, base shear and base moment 3D finite element modelling, as opposed to back ana
(Burd et al. 2020). The PISA framework has now been lysis of small scale instrumented laterally loaded pile
implemented into the commercial software PLAXIS tests as was the traditional approach. Equation (2)
Monopile Designer (formerly MoDeTo) and is being below has been calibrated against FE models using
commonly used in the industry (Minga and Burd the hardening soil model (HS) in Plaxis 3D
DOI: 10.1201/9781003308829-149
991
foundation and validated against lateral pile field 2.1 Model set up
tests with diameters of about 0.6 m and pile penetra
Monopiles are modelled at full scale and half space
tions of 17 m (L/D > 28). γ denotes the soil unit
using the commercially available finite element pack
weight and Z the depth at which the soil reaction is
age PLAXIS 3D (see Figure 2). They are modelled
calculated.
with linear elastic plate elements with Young’s modu
lus of 210 GPa and Poisson’s ratio of 0.3. Since the
lateral response is the main concern, the self-weight of
the monopile is omitted. The piles are assumed wished
in-place with no effect of the installation taken into
account.
However, these CPT based correlations are not This study is limited to monopile modelling in
compatible with the PISA framework and, as only sand. The hardening soil model with small-strain stiff
the p-y response is considered, they might not be ness (HS-small) is an improvement of the HS model
suitable for modelling monopiles (similarly to API). used by Suryasentana and Lehane (2014). It has been
successfully used to accurately model monopile lat
eral response in sand (Igoe and Jalilvand 2020), and
1.2 Proposed approach thus is considered here. Vertical interfaces are added
It is proposed here (see Figure 1) to use 3D finite between the pile shaft and the soil to allow for differ
element modelling to derive CPT based correlations ential displacements, to introduce an interface
of soil reaction curves including distributed lateral strength reduction factor (set to 0.7 here) and to allow
reaction, distributed moment, base shear and base extraction of soil distributed reactions (distributed lat
moment (as per PISA framework). eral load and distributed moment). At pile toe, an
Soil parameters required as inputs for the 3D FE additional horizontal interface is added to allow for
are directly derived from CPT profiles as per the extraction of base reactions (base shear and base
approach proposed by Igoe and Jalilvand (2020). moment) but with no strength reduction considered.
This approach requires minimum user inputs and has The lateral loading as a result of wind and wave
been validated against a database of lateral pile field action is applied as a prescribed displacement at
tests including PISA tests in the Dunkirk marine a height, e, above mudline. Burd et al. (2020) showed
sand. that there is negligible effect of the loading eccentri
All components of soil reaction are extracted from city on soil reaction curves. Hence the value of e is
the 3D FE models for a range of pile geometries and taken as 60 m (a mean value to represent both wave
CPT profiles. CPT based correlations are calibrated and wind dominated scenarios). The lateral loading is
from this database of soil reaction curves and by applied in the y direction. The size of the domain is
matching the 3D FE and 1D FE mudline responses. taken as 12D in the direction of loading and 4D in the
The approach is then validated against large scale perpendicular direction as recommended in the
pile field tests (the same database used by Igoe and PLAXIS Monopile Designer manual (Panagoulias
Jalilvand 2020). et al. 2021), where D is the monopile diameter. The
depth of the model is here set to 80 m with 16 soil
layers of 5 m thickness each. This is to ensure the soil
domain is large enough to avoid boundary effects.
Figure 1. Flowchart of the proposed approach.
2 PLAXIS 3D FINITE ELEMENT MODELLING
This section presents how the 3D finite element

models are set up in PLAXIS 3D and how results are
being post-processed. Figure 2. Description of the monopile model in Plaxis 3D.
992
2.2 Soil parameters 2.3 Monopile geometry
Igoe and Jalilvand (2020) developed an approach to To allow for the calibration of the CPT based cor
derive all of the required HS-small input parameters relations, a range of monopile geometries is con
using widely used CPT correlations. All correlations sidered. Literature showed that there is negligible
used are summarised in Table 1. The approach has effect of the pile wall thickness and load eccentri
been validated against a database of large scale field city on the soil reaction curves so these are kept
tests including PISA tests in the Dunkirk marine constant as t = D/110 and e = 60 m, respectively
sand and offers better predictive accuracy metrics (representative values based on experience). Out
than correlations suggested for the Plaxis Monopile side diameters (D) of 6 m, 8 m, 10 m and 12 m are
Designer (Panagoulias et al. 2021). considered. Slenderness ratios (L/D) of 2, 3, 4, 5
By re-arranging the correlations presented in and 6 are considered. This results in pile embed
Table 1, synthetic CPT profiles for constant sand ded length (L) ranging from 12 m to 72 m. Hence
relative density are generated. Relative densities of a total of 20 geometries are considered for each
30% (loose), 50% (medium dense), 70% (dense) soil profile covering recent, current and future
and 90% (very dense) are considered. At this stage, design.
only normally consolidated state is considered with
OCR set to 1. A constant ratio sleeve friction to
2.4 Soil reaction curves extraction
cone resistance, fs/qc, of 1% is assumed. The con
stant volume friction angle is taken as 32 degrees Unlike Suryasentana and Lehane (2014), soil reac
for all profiles. The soil unit weight is taken ranging tion curves are not calculated from the derivative
from 17 kN/m3 to 20 kN/m3 depending (linearly) of shear and bending moment diagrams. The
on relative density. Finally, the small strain shear approach briefly presented in the PLAXIS Mono-
modulus profiles are estimated from the cone resist pile Designer manual (Panagoulias et al. 2021) has
ance with α set to 185 as recommended by Igoe & been preferred as it allow for extraction of all
Jalilvand (2021). components of the soil reaction.
Distributed lateral load and distributed moment
Table 1. Correlations used by Igoe & Jalilvand (2021).
are calculated from integration of the normal and
tangential stresses acting on monopile shaft (verti
Correlations References cal interface). Soil reactions are integrated along
1 m (ΔZ) intervals as shown on Figure 3 and as per
Eref
Brinkgreve et al. equations (3) and (4). Resulting lateral forces, Fy,
50 ¼ Eur =3
ref
(2010) and vertical forces, Fz, are calculated at each of the
Eref ref Brinkgreve et al. 6 Gaussian stress points of each interface elements
oed ¼ E50 (2010) as per equations (5) and (6), respectively. Where σn ,
ref
ur ¼ 0:00464 · E0
Eref
1:724
Modified after Kirsh τ1 and τ2 are the stresses directly extracted from
Where E0 ¼ 2ð1 þ ÞG0ref
ref
et al. (2014) Plaxis, X and Y are the coordinates of the stress
m ¼ 0:5 Benz et al. (2009)
point, D is the pile diameter, A is the area of the
’0 ¼ ’0 cv þ 3ðDR ð10 - ln p0 0 Þ - 1Þ Bolton (1986) interface element and w is the weight of the stress
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi point (0.11 for stress points 1 to 3 and 0.22 for
DR ¼ Qtn
0:15
Kulhawy & Mayne stress points 4 to 6).
305xOCR
(1990)
f nf n
qt -σv0 pref n Roberston & Cabal
Qtn ¼ pref σ0 v0
0 - (2014)
n ¼ 0:381 · Ic þ 0:05 σprefv0 - 0:15

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
ð3:47 - log10 Qtn Þ þ
Ic ¼
ð1:22 þ log10 Fr Þ2
Fr ¼ qt -σv0
fs
0 0
sin ’ - sin ’ cv
sin ψ0 ¼ 1-sin ’0 sin ’0 cv
Brinkgreve et al.
(2018)
� �1=3
G0 ¼ α qt · σ0 v0 · pref Schnaid and Yu
(2007)
γ0:7 ¼ 1:5 · 10-4 Benz et al. (2009)

f nm
Gref
pref Brinkgreve et al.
0 ¼ G0 p0 0 The base shear and base moments are calcu
(2018)
lated from the horizontal interface. This is not
0 sin ’0
K0 ¼ ð1 - sin ’ Þ x OCR Mayne & Kulhawy detailed here as the approach is very similar. All
(1982)
calculations are repeated for each load increment
993
Figure 3. Integration of distributed lateral reaction and
distributed moment.
in order to build the full reaction curves. These

are then saved in a formatted .csv file for
later use.
With 4 relative densities considered and 20 monopile

geometries for each soil profile, the database consists
in 80 3D FE models. On average, it took 1 hour to
set up, run and process each model on a relatively
high capacity computer (Intel(R) Xeon(R) W-1270P
CPU @ 3.80GHz, 32 GB DDR4). The computation
times were greatly reduced thanks to the automation
Figure 4. Comparison of 1D FE and 3D FE monopile lat
of most aspects of the PLAXIS modelling through eral responses for case RD70%_D10L30.
the PYTHON scripting interface. The database has
been recently completed and this section presents
and discuss preliminary results obtained.
model. The initial stiffness and ultimate capacity are
defined at a mudline displacement of D/10000 and
3.1 Comparison 1D FE – 3D FE responses
D/10, respectively. From experience, monopiles are
In order to make sure soil reaction curves are being subjected to strict serviceability requirements leading
correctly extracted from PLAXIS, 3D FE and 1D FE to low utilization in ULS GEO checks (≈ 1/2). When
monopile response are compared. the load partial factor of ≈1.35 and material partial
For the sake of conciseness, only one model with factor of 1.15 is considered, the SLS load is approxi
pile diameter of 10 m, embedment length of 30 m in mately defined as 1/3 of the ultimate capacity. One
70% relative density sand (case RD70%_D10L30) is may consider that the 1D FE with p-y only may be
presented in Figure 4 but all other models show considered overly conservative with underestimation
similar results. Both load-displacement (top of the of the initial stiffness and the ultimate capacity by
figure) and moment-rotation (bottom of the figure) 16.4% and 21.1%, respectively. The “SLS” rotation
curves compare relatively well when all reaction is overestimated by 47.7%. When adding other com
components are considered in the 1D FE model. ponent of soil reactions to the 1D FE model, the
However, when only p-y curves are being con match is much better with only a 4% underestimation
sidered, the 1D FE response is found to be consider of initial stiffness and ultimate capacity. The “SLS”
ably softer than the 3D FE response. This is in rotation is only overestimated by 6.9%.
agreement with findings from the PISA project This quantitative assessment has been repeated for
(Byrne et al. 2015) and confirms the need of CPT all 80 models. Only results pertaining to the estima
based correlations including all components of pile- tion of ultimate capacity are presented in Figure 5
structure interactions and not limited to p-y curves but the other two metrics show the same trend. For
only. a typical slenderness ratio (L/D) of 3, the pile ultim
To add to this qualitative assessment, Table 2 ate capacity is underestimated by more than 20% if
compare initial stiffnesses, SLS rotations and ultim only ‘p-y’ curves are used in 1D FE while the error
ate capacities from 1D FE and 3D FE for the same is reduced to about 5% when all components are
994
Table 2. Comparison of 1D FE and 3D FE initial stiff et al. 2020) and Broms (1964). Input parameters
nesses, SLS rotations, and ultimate capacities for case required for API (friction angle) and PISA rule (rela
RD70%_D10L30. tive density) are calculated as per approach set out in
Table 1. Broms (1964) approach is directly based on
Ultimate the synthetic CPT profiles presented in section 2.2.
Initial Stiffness SLS Rotation Capacity Figure 6 Highlights the need to not only correctly
Model [MN/m] [deg] [MN] match the ultimate reaction but also accurately
3D FE 244.4 0.450 43.9
describe the shape of the reaction curves. Broms
1D FE 204.2 0.664 34.6
(1964) overly simplify the problem by considering
py only (- 16.4 %) (+ 47.7 %) (- 21.1 %) that the full reaction is mobilised at any displacement.
1D FE 234.5 0.481 42.1 In the API (2011), a hyperbolic tangent function is
all curves (- 4.0 %) (+ 6.9 %) (- 4.0 %) considered with the initial stiffness being a function of
sand peak effective friction angle and depth. Although
both approaches underestimate the ultimate capacity
they will results in stiffer response in most practical
cases (until a displacement of about 3% of pile diam
considered. For any L/D, the error is always larger eter). On the contrary, the initial part of the PISA rule
when considering p-y curves only which suggests the curve match relatively well with 3D FE. Although the
need for the other component of soil reaction (distrib ultimate capacity is underestimate again, they will
uted moment, base shear and base moment). For the results in similar response in most practical case.
p-y only scenario there is a clear trend for the errors In the PISA framework each of the reaction
to reduce when L/D increases. This is in agreement curves (distributed lateral reaction, distributed
with the results of Byrne et al. (2015) and it explains moment, base shear and base moment) are fitted
why approaches validated for slender piles (API with 4 parameters: the initial stiffness, the curvature,
2011, Broms 1964, Suryasentana and Lehane 2014) the ultimate reaction and the displacement at which
had satisfactory results for slender piles. However, the ultimate reaction is reached. This results in
for current monopile design with L/D of as low as 3, a total of 16 functions to correlate with CPT.
these approaches are no longer applicable.
It is not yet clear why the errors with all curves
included tend to slightly increase with L/D. How
ever, the errors are deemed small enough (typically
less than 5%) and the extracted soil reaction curves
are deemed satisfactory.
Figure 6. Comparison of distributed lateral reactions

(p-y curves) at 4.5 m depth for case RD70%_D10L30.
4 ALTERNATIVE APPROACH
To date, no satisfactory CPT based correlations could

be achieved due to a number of limitations with the
Figure 5. Comparison of ultimate capacity relative errors. current approaches. The main one (see sketch on
Figure 7) is the lack of reaction close to pile point of
3.2 Comparison soil reaction curves rotation due to insufficient displacement. Although
the pile is pushed until failure (mudline displacement
For the same case (RD70%_D10L30), Figure 6 shows larger than 10% of the pile diameter), the ultimate
the distributed lateral reaction curves (p-y curve) at reactions are only reached for the top first 30%-40%
a depth of 4.5 m as extracted from the 3D FE model of the pile length. No reaction at all is recorded close
and calculated as per API (2011), PISA rule (Burd to point of rotation (at about 70% of pile length).
995
Comparison of 3D FE extracted curves with exist
ing approach shows that the ultimate distributed lat
eral soil reaction seems to be currently under
estimated, but this may not necessarily translate into
an under-estimation of the pile response depending
on the curve stiffness (and the account of other soil
reaction components).
Current work is focusing on the derivation of
CPT based correlations for all soil reaction curves
including distributed lateral load, distributed
moment, base shear and base moment. However, no
satisfactory results could be obtained so far.
It is being investigated if the current 3D model-
ling approach should be revised. The current
Figure 7. Limitation of the current rotation approach and approach involves rotation of the monopile around
proposed alternative approach. of point of rotation. There is no pile displacement
close to the point of rotation and hence no soil reac
tions making most of the data extracted from the 3D
FE models un-usable. An alternative approach is
An alternative approach is being considered where being considered where the pile is being laterally
the pile would be laterally translated rather than translated in order to record soil reaction along the
rotated. This would allow for full mobilisation of whole length of the monopile rather than only the
reaction at any depth. Some components of the soil top 30%-40%.
reaction may not be capture (distributed moment,
base moment) and will still need to be extracted from
the original models or correlated with the other com ACKNOWLEDGEMENT
ponent (distributed lateral reaction, base shear).
On-going works are focusing on comparing The research conducted in this publication was
p-y curves extracted from both approaches to make funded by the Irish Research Council Postgraduate
sure they are comparable. Employment-based Programme under grant number
EBPPG/2019/4.
5 CONCLUSION
REFERENCES
This paper presents a novel approach to derive CPT
based correlation of soil reaction curves including API (2011). “API RP2GEO: Geotechnical and Foundation
distributed lateral reaction (p-y curves) but also dis Design Considerations.”
Benz, T., Schwab, R., and Vermeer, P. (2009). “Small-
tributed moment, base shear and base moment. This
strain stiffness in geotechnical analyses.” Bautechnik,
approach relies on finite element modelling in Plaxis 86(SUPPL. 1), 16–27.
3D and the CPT based correlation of hardening soil Bolton, M. D. (1986). “The strength and dilatancy of
model with small strain stiffness (HS-small) param sands.” Géotechnique, 36(1),65–78.
eters developed by Igoe and Jalilvand (2020). Brinkgreve, R. B. J., Engin, E., and Swolfs, W. M. (2018).
A total of 80 FE model are considered based on “Plaxis 3D Materials Manual.”
a range of pile geometry and synthetic CPT profiles Brinkgreve, R., Engin, E., and Engin, H. (2010). “Valid
for relative densities ranging from 30% (loose sand) ation of empirical formulas to derive model parameters
to 90% (very dense sand). for sands.” Numerical Methods in Geotechnical Engin
eering, (June), 137–142.
All soil reaction curves are directly extracted
Burd, H., Taborda, D., Zdravković, L., Abadie, C.,
from Plaxis 3D. Soil reactions are calculated from Byrne, B., Houlsby, G., Gavin, K., Igoe, D., Jardine, R.,
integration of stress at the pile-soil interface rather Martin, C., McAdam, R., Pedro, A. and Potts, D.
than derivation of pile shear force and bending (2020). “PISA design model for monopiles for offshore
moment profiles with depth. These are incorporated wind turbines: application to a marine sand.” Géotechni
in a 1D FE solver and the obtained responses is com que, 70:11, 1048–1066.
pared with 3D FE to ensure the reaction curves are Byrne, B. MacAdam, R., Burd, H. and Houlsby, G. (2015).
correctly extracted. Both initial stiffness, ultimate “New design methods for large diameter piles under lat
capacity and rotation under representative SLS loads eral loading for offshore wind applications.” Proceed
ings to the Third International Symposium on Frontiers
are found to match well. By comparing the 1D FE
in Offshore Geotechnics in Oslo, Norway.
responses when only p-y curves are considered, it is DNV (2013). DNV OS-J101: Design of offshore Wind Tur
showed that existing approaches relying on bine Structures.
p-y curves solely are not satisfactory for monopile DNV (2014). DNV OS-J101: Design of offshore Wind Tur
modelling with slenderness ratio (L/D) of about 3. bine Structures.
996
Igoe. D. and Jalilvand, S. (2020). “3D finite element model- Petrasovits, G., and Award, A. 1972. “Ultimate lateral
ling of monopiles in sand validated against large scale field resistance of a rigid pile in cohesionless soil.”
tests.” Proceedings to the Fourth International Symposium Proc., 5th European Conf. on SMFE 3, The Spanish
on Frontiers in Offshore Geotechnics in Houston, USA. Society for Soil Mechanics and Foundation,
Kirsch, F., Richter, T., and Coronel, M. (2014). “Geotech 407–412.
nische Aspekte bei der Gründungsbemessung von Off Prasad, Y. V. S. N., and Chari, T. R. 1999. “Lateral capacity
shore-Windenergieanlagen auf Monopfählen mit sehr of model rigid piles in cohesionless soils.” Soils Found.,
großen Durchmessern.” Stahlbau Spezial 2014 – 392, 21–29.
Erneuerbare Energien, 83 (SUPPL. 2), 61–67. Robertson, P. K., and Cabal, K. (2014). “Guide to Cone
Kulhawy, F. H., and Mayne, P. W. (1990). „Manual on Esti Penetration Testing for Geotechnical Engineering - 6th
mating Soil Properties for Foundation Design.” EPRI Edition”.
EL-6800. Schnaid, F., and Yu, H. S. (2007). “Interpretation of the
Mayne, P. W., and Kulhawy, F. (1982). “K0 - OCR relation seismic cone test in granular soils.” Géotechnique, 57
ships in soils.” Journal of Geotechnical Engineering (3),265–272.
Division, 108(6),851–872. Suryasentana, S., and Lehane, B. M. (2014). “Numerical
Minga, E. and Burd, H. (2019). “Validation of the PLAXIS derivation of CPT-based p–y curves for piles in sand”,
MoDeTo 1D model for dense sand”. 64(3)
Panagoulias, S., Brinkgreve, R. and Zampich, L. (2021) Wind Europe (2020). “Offshore wind in Europe: key trends
“Plaxis Monopile Designer CE V21 Manual”. and statistics.”
997
© 2022 the Author(s), ISBN 978-1-032-31259-0
CPT-based assessment of densification induced by stone column installation

F. Marchi, E. Zambianchi, A. Boschi, A. Mastrangelo & G. Marchi
ENSER Srl, Faenza, Italy
G. Gottardi & L. Tonni

UNIBO, Bologna, Italy
ABSTRACT: The use of CPTs for liquefaction susceptibility assessment has become increasingly common
in design practice. In the past, the design of densification soil improvement methods against liquefaction has
been commonly based on experience, large scale field tests and, possibly, validation after the intervention.
Nevertheless, a design method to assess quantitatively the densification effect induced by stone column instal
lation can be developed following a workflow based on CPTs, in order to estimate the efficiency of the soil
improvement and to check the compliance with safety requirements before accessing the site. As by common
practice, the assessment of liquefaction susceptibility is performed by means of CPT data and related empir
ical methods. The design method proposed in this paper is based on the estimation of modifications induced
by the soil densification technique, easily detectable by the CPT profile, in order to carry out a liquefaction
susceptibility analysis. The approach is applied in particular to a well-documented case study in Northern
Italy (Bondeno, Ferrara province), providing a consistent dataset of piezocone tests carried out before and
after soil treatment through stone columns, as part of a blast-induced liquefaction experiment. In this way,
a clear comparison between pre- and post- intervention conditions along the CPT profiles with depth can be
performed, thus allowing a validation of the design methodology.
1 INTRODUCTION susceptibility assessment have been developed and val

idated in the last decades (e.g. Seed and Idriss 1971,
Among the ground improvement solutions available to Robertson and Wride 1998, Idriss and Boulanger
mitigate the liquefaction hazard, densification is prob 2008, Boulanger and Idriss 2014).
ably the most common and widespread. In particular, The results of the design workflow will be
vibratory compaction methods are a common and applied to the data coming from the field tests per
effective form of densification for cohesionless soils formed in connection with a full-scale controlled
(Castro 1969), as proven by extensive research (e.g. blast test in Bondeno (FE), Italy. A detailed geo
D’Apollonia 1954, Mitchell 1981, Baez 1995, Adalier technical investigation of the site is presented by
and Elgamal 2004, Wissmann et al. 2015, Vautherin Amoroso et al. (in press).
et al. 2017, Amoroso et al. 2018). In spite of its large
application in the last decades, design methods to
quantify and predict the efficiency of the improvement 2 THEORETICAL WORKFLOW
are not commonly reported in technical manuals.
Soil improvement has been mainly calibrated In case of real scale field test, common practice sug
through real scale field tests; such type of tests pre gests following a number of steps:
sents, however, some issues to face. Real scale tests 1. performing investigations on the natural soil
can be difficult to perform in a design phase, since (CPT in the present study, PRE index will be
often the executor of the treatment has not been used to address these data),
chosen yet or the permission to operate in the area is 2. assessing liquefaction susceptibility on the data
not ready. Finally, real scale tests are expensive and collected,
need to be designed too. 3. performing the soil improvement treatment,
In the following, a workflow will be described in 4. performing new investigations (POST-M index
order to obtain a prediction of the improvement effi will be used to address these data) in order to
ciency. The proposed workflow is based on the CPT compare them with the pre-treatment ones,
data. In fact, CPTs are cost effective site investigations 5. assessing liquefaction susceptibility on the base
and several simplified procedures for liquefaction of post-treatment data collected.
DOI: 10.1201/9781003308829-150
998
A soil improvement is recognized to be successful
when the liquefaction susceptibility detected in the
post-treatment phase in neglectable or tolerable.
The proposed workflow will follow the same steps. where Ar is the substitution ratio defined as the area of
Point 4 will be done trying to predict a theoretical the treating element divided by the area of the treated
post-treatment CPT data (POST-P index will be used mesh, as also explained in Figure 1. For the Ar values
to address this data), modifying artificially the cone generally considered in practical applications, equation
resistance qc and sleeve friction fs measurements, and (5) results in values of e0,cs,POST-P within the range of
finally assessing liquefaction susceptibility on the base values typically assumed by the void ratio.
of this predicted post-treatment test. In the following,
the proposed procedure for the determination of the
POST-P CPT parameters will be presented in detail.
With reference to the different CPT measurements
(qc,PRE, fs,PRE), the cone resistance is assumed as the
main parameter for the procedure. In particular the
procedure will be applied to the normalized cone
penetration resistance for clean sands, Qtn,cs-PRE,
defined as per Robertson & Wride (1998). The densi
fication treatment in fact has its proper application
on clean sands. Accordingly, it is then possible to
back-calculate qt,cs-PRE and qc,cs-PRE from Qtn,cs-PRE,
according to the following relationships:
Figure 1. Definition of the Ar parameter.
where pa = atmospheric pressure; σv0 = vertical

Based on the improved void ratio, the procedure
total stress; σ0v0 = vertical effective stress; u = pore can be applied backward, giving:
pressure; a = cone area ratio.

Using the Jamiolkowsky et al. (2001) correlation,
it is then possible to obtain the relative density
(DR,cs,PRE) and then the void ratio of the untreated
soil (e0,cs,PRE):
By applying again equations (1) and (2), it is then

possible to back-calculate qt,POST-P and qc,POST-P. The
procedure presented above only involves the param
eters defined from cone penetration resistance,
where C0 = 17.68; C1= 0.50; C2 = 3.10 according whereas the sleeve friction resistance has not been
to Jamiolkowsky et al. (2001) whereas emax,cs and touched yet. A disconnected modification in one of
emin,cs are assumed in this study equal to 1 and 0.4 these two parameters would lead to a change in the
respectively, in agreement with relevant ranges of soil behavior type index (Ic), that is mainly related to
the maximum and minimum void ratio shown by the soil class. Assuming that the improvement will
Cubrinovski & Ishihara (2002). not change the soil type, fs,POST-P can be back calcu-
On the basis of the value assumed by the void lated from the improved cone penetration resistance.
ratio, it is possible to quantify the improvement A new profile of the CPT can be then predicted
induced by the densification treatment. In particular, (qc,POST-P, fs,POST-P). Ordinary procedures to assess
the void ratio of the soil after treatment, e0,cs,POST-P, liquefaction susceptibility can be performed on the
can be inferred from a correlation originally pro- new profile in order to evaluate the efficiency of the
posed by Mitchell (1981) to predict sand compaction soil improvement treatment on a theoretical base.
due to pile installation. Accordingly, e0,cs,POST-P is An additional evaluation has to be performed to
given by: assess the suitability of soils for vibratory compaction,
999
since it is documented in literature that only soils with As reported in Amoroso et al. (in press), the Bon
low fines content react properly to vibratory compac deno test site subsoil consists of a silty-clayey hard
tion (Massarsch, 1991). topsoil in the upper 3.5m, with an average plasticity
index (PI) of 20%, followed by a non-plastic silty
sand with fine content typically in the range 25-35%
3 VERIFICATION ON BLAST TEST DATA and therefore classified as SM.
Suitability for vibratory compaction of CPT01_PRE
Bondeno blast test site offered the occasion to verify has been evaluated following Massarsch (1991);
if the theoretical framework gives a good prediction Figure 2 shows that the cone penetration data points, in
of the soil improvement efficiency. The site has been terms of tip resistance qc and friction ratio FR (calcu
widely investigated in natural conditions, before lated as 100 · qc/fs), fall in the domain of compactable
treatment (Phase I), after the soil improvement materials.
(Phase II) and after the blast-induced shaking (Phase Figure 3 provides a comparison between CPT01
III). The soil improvement consisted of a number of and CPT01bis, carried out at a distance of 1.5m. Pro
4 x 4 columns in quadrangular grid (2m center-to files of qt, soil behaviour type index Ic, relative dens
center spacing) of Rammed Aggregate Piers (RAP) ity DR, fine content FC, Factor of Safety against
columns (Saftner et al., 2018), each 9.5m long and liquefaction SF and Liquefaction Potential Index LPI
with a final diameter of 0.5m (Ar ≈ 5%) (Amoroso are provided. The plot clearly shows where and how
et al., in press). the soil improvement affected the degree of compac
Among the available cone penetration tests car tion of the natural soil, taking into account that RAP
ried out in the test site, CPT01 performed in Phase length was 9.5m.
I (also referred in the following to as CPT01_PRE) Liquefaction analyses have been performed fol
and CPT01bis performed in Phase II (also referred lowing the procedure suggested by Boulanger &
to as CPT01bis_POST-M) have been chosen to test Idriss (2014).
the procedure described above.
4 FINAL RESULTS
Following the theoretical workflow presented in sec

tion 2, data from CPT01_PRE test have been elabor
ated to produce a design prediction of the treated soil.
The prediction will be referred to as CPT01_POST
P. A comparison between measured and predicted
profiles is presented in Figure 4.
The comparison in terms of qt shows a very good
agreement from 5.5m to 8.0m in depth. In the lower
part of the treatment (from 8.0m to 9.5m) the effi
ciency of the soil improvement seems to decrease;
this can be related to the influence of the end depth
of the treatment. Getting closer to the tip of the col
umns, the basic hypothesis of two-dimensional
behavior in the horizontal plane, on which eq. (5)
relies, can no longer be considered as strictly valid.
Three-dimensional effects probably have some influ
ence in the soil volume, causing a decrease in the
densification effect.
In the range between 3.4m and 5.3m, the agree
ment is still quite good, although the predicted qt
Figure 2. Soil classification for assessment of deep com
profile shows values higher than those measured.
paction based on CPT (Massarsch, 1991).
The same comments apply to the values of DR.
The evaluations related to liquefaction susceptibil
In particular: ity, expressed in terms of SF and LPI, show similar
trend, though the suggested workflow seems to slightly
‒ CPT01 is the test performed before the soil overestimate the efficiency of the soil improvement.
improvement treatment (qc,PRE); Considering the result on LPI, the distribution along
‒ CPT01 will be modified following the workflow depth suggests the same differences discussed above.
exposed in section 2 in order to predict the post- In particular, from 5.5m down to the end of the test the
treatment cone resistance profile (qc,POST-P); agreement between prediction and measured data is
‒ CPT01bis is the test carried out after the installa extremely good. More significant differences are
tion of the RAPs, providing the profile of the observed in the range 3.4m - 5.3m depth. Results from
measured cone resistance (qc,POST-M). comparison between CPT01_POST-P and
1000
Figure 3. Comparison between pre-treatment (CPT01_PRE) and post-treatment (CPT01bis_POST-M) investigations.
Figure 4. Comparison between pre-treatment (CPT01_PRE) and post-treatment investigations, as measured in

CPT01bis_POST-M and predicted in CPT01_POST-P.
1001
Table 1. Numerical evaluation of predictions. that, close to the column tip, the efficiency of the
treatment will gradually decrease to zero, while in
z Soil qt DR SF the method described in this paper the densification
m MPa % effect is solely quantified on the horizontal displace
ment of the soil.
CPT01_PRE 6.39 56 0.66
7.09 57 0.80
CPT01bis_POST-M
3.4-5.3 (+11%) (+3%) (+20%) ACKNOWLEDGEMENTS
9.65 69 1.59
CPT01bis_POST-P
(+51%) (+24%) (+139%) The study presented in the paper was based on the
CPT01_PRE 8.94 60 0.76 data collected within a Blast Test experiment, carried
CPT01bis_POST-M
12.33 69 1.85 out with the financial support of Istituto Nazionale di
5.8-8.0 (+38%) (+15%) (145%) Geofisica e Vulcanologia (INGV), CIRI Edilizia
13.45 73 2.76 e Costruzioni - University of Bologna, Geopier®
CPT01bis_POST-P
(+50%) (+22%) (+264%) Foundation Company, and Releo srl.
CPT01_PRE 7.66 57 0.71
9.86 63 1.36
CPT01bis_POST-M
3.4-9.5
(+29%) (+11%) (+91%) REFERENCES
11.55 70 2.20
CPT01bis_POST-P (+51%) (+23%) (+209%) Adalier, K. & Elgamal, A. 2004. Mitigation of liquefaction
and associated ground deformations by stone columns.
Eng. Geol., 72(3-4): 275–291.
Amoroso, S., Rollins, K.M., Monaco, P., Holtrigter, M., &
Thorp, A. 2018. Monitoring ground improvement using
CPT01bis_POST-M, taking CPT01_PRE as reference, the seismic dilatometer in Christchurch, New Zealand.
are summarized in Table 1. The percentage increment Geotech. Test. J., 41 (5): 946–966. https://doi.org/
of qt, DR and SF after treatment is given in brackets. 10.1520/GTJ20170376.
Amoroso, S., Martínez, M.F.G., Monaco, P., Tonni, L.,
Gottardi, G., Rollins, K.M., Minarelli, L., Marchetti, D.,
5 CONCLUSIONS Wissmann, K.J. Comparative study of CPTU and
SDMT in a silty sand liquefaction-prone site improved
by Rammed Aggregate Piers and subject to controlled
The theoretical workflow proposed in this paper is blasting. Journal of Geotechnical and Geoenvironmen
a guideline for the quantification of soil improve tal Engineering (in press).
ment effectiveness in densification treatment against Baez, J.I. 1995. A design model for the reduction of soil
liquefaction. Following the procedure, it is possible liquefaction by vibrostone columns. Ph.D dissertation,
to obtain a design prediction of the soil improvement Univ. of Southern California.
based only on CPT data. This simple approach is Boulanger, R.W. & Idriss, I.M. 2014. CPT and SPT based
very useful in a design phase where in situ tests are liquefaction triggering procedures. Rep. No. UCD/
commonly available for liquefaction assessment, but CGM-14/01. Davis, CA: Center for Geotechnical Mod
eling, Dept. of Civil and Environmental.
it is hardly possible to carry out full-scale field tests. Castro, G. 1969. Liquefaction of sands. Ph.D. Dissertation,
The application of the workflow on a case-study Harvard University.
in Bondeno (FE) allowed to evaluate the perform Cubrinovski, M. & Ishihara, K. 2002. Maximum and min
ance of the method by comparing predictions with imum void ratio characteristics of sands. Soil and Foun
verification tests. The overall result gives a good dations, Vol.42, No.6, 65–78.
agreement in the lower liquefiable layer, though fur D’Appolonia, E. 1954. Loose sands - their compaction by
ther analyses need to be performed in order to verify vibroflotation. In Proc., Symp. on Dynamic Testing of
local differences with depth. In addition, only one Soils: 138–162. West Conshohocken, PA: ASTM
CPT was available for the verification of the work- International.
Idriss, I.M. & Boulanger, R.W. 2008. Soil liquefaction
flow, verifying the approach on other test couples
during earthquakes. Report No. MNO-12. Oakland, CA:
performed in different sites with different soil condi Earthquake Engineering Research Institute.
tions is needed. Jamiolkowski, M., Lo Presti, D.C.F., Manassero, M., 2001.
The workflow seems to slightly overestimate the Evaluation of Relative Density and Shear Strength of
soil improvement beneficial effects. This can be Sands from CPT and DMT. ASCE Geotechnical Special
related to a variety of factors. A general efficiency Publication No. 119: 201–238.
factor is a possible solution in order to take into Massarsch, K.R., 1991. Deep Soil Compaction Using Vibra
account different aspects related to the installation tory Probes. In Robert C. Bachus, Ed., American Society
method. The procedure would also need to be for testing and Material, ASTM, Symposium on Design,
Construction, and Testing of Deep Foundation Improve
extended through the implementation of efficiency ment: Stone Columns and Related Techniques, ASTM
factors accounting for marginally compactable soils Special Technical Publication, STP 1089: 297 319.
(with higher fine and/or plastic contents) and for the Mitchell, J.K. 1981. Soil improvement: state-of-the-art. In
reduction of soil treatment efficiency approaching Proc. 10th Int. Conf. on Soil Mechanics and Foundation
the treatment tip. It is in fact reasonable to consider Engineering, Vol. 4: 509–565.
1002
Robertson, P.K. & Wride, C.E. 1998. Evaluating cyclic Vautherin, E., Lambert, C., Barry-Macaulay, D., &
liquefaction potential using the cone penetration test. Smith, M. 2017. Performance of rammed aggregate
Can. Geotech. J., 35(3): 442–459. piers as a soil densification method in sandy and silty
Robertson, P.K. & Cabal, K.L., 2015. Guide to Cone Pene soils: experience from the Christchurch rebuild. In
tration Testing for Geotechnical Engineering, 6th edition. Proc., 3rd Int. Conf. on Performance-based Design in
Saftner, D.A., Zheng, J., Green, R.A., Hryciw, R. and Earthquake Geotechnical Engineering.
Wissmann, K.J. 2018. Rammed aggregate pier installa Wissmann, K.J., van Ballegooy, S., Metcalfe, B.C.,
tion effect on soil properties. P. I. Civil Eng. Ground Dismuke, J.N., & Anderson, C.K. 2015. Rammed
Impr. 171 (2), 63–73. aggregate pier ground improvement as a liquefaction
Seed, H.B. & Idriss, I.M. 1971. Simplified procedure for mitigation method in sandy and silty soils. In Proc.,
evaluating soil liquefaction potential. J. Geotech. Engrg. 6th Int. Conf. on Earthquake Geotechnical
Div., ASCE, 97(9): 1249–1273. Engineering.
1003
© 2022 the Author(s), ISBN 978-1-032-31259-0
Geopier Impact technology for liquefaction risk mitigation based on CPTu

investigations
G. Martinez & K.J. Wissmann
Geopier Foundation Company, Davidson, USA
M. Franceschini, E. Bandiera & F. Fiorelli

Teleios Srl-Società di Ingegneria, Castel Maggiore, Bologna, Italy
ABSTRACT: The present paper is focused on the Impact Rammed Aggregate Pier system, a ground
improvement technology developed by Geopier Foundation Company. From extensive in-situ tests also
including blast tests, Impact elements have proven to be very effective at densifying in-situ granular soils and
increasing the factor of safety against liquefaction. The intent of the paper is to verify the methods that are
used to evaluate liquefaction mitigation with Impact piers. One method to evaluate the mitigation of liquefac
tion risk for design of Impact piers is to estimate the increase in CPT tip resistance qc expected after installa
tion of the Impact elements. In the paper we will consider CPTu tests performed before and after Impact
installations in various sites worldwide to attempt to verify the accuracy of formulas from literature and to try
to understand which aspects can be refined in future studies. The future aim is to possibly modify the formulas
to account for local soil conditions in the Emilia-Romagna region in Italy.
1 INTRODUCTION estimating densification based on Piezocone Penetration

Tests (CPTu). One method available in literature will be
Among the various technologies for soil improvement, presented and its reliability will be assessed by means
stone columns are a method extensively used all of comparison with pre and post-installation CPTu tests
around the world with different technological variants. performed in various sites around the world.
In addition to their demonstrated capacity to The future aim of the authors is to develop correl
increase soil resistance and to reduce foundation settle ations specific to the local soil conditions in the Emilia-
ments against static loads, the increasing interest in Romagna region in Italy, where the authors have already
soil improvement methods for geotechnical seismic designed several projects for liquefaction mitigation.
engineering has led to an advance in the study of
gravel columns for liquefaction mitigation purposes.
The proposed paper focuses on a specific type of 2 RAMMED AGGREGATE PIER ELEMENTS
gravel column, the Rammed Aggregate Pier (RAP)
Impact technology developed by Geopier Foundation Rammed Aggregate Pier elements are a ground
Company. The Impact technology is a displacement improvement technology developed over the last 25
system that produces highly compacted gravel col years by Geopier Foundation Company. The RAP
umns together with vibro-densification of the in-situ installation method creates a densified column of
granular soils. Successfully used in several hundred aggregate surrounded by a stiffened matrix soil, with
sites worldwide, the Impact system has also proven to the aim of providing settlement control and allowing
be very effective at densification and controlling lique for higher bearing capacities.
faction compared to other methods such as timber RAP installation methods can be classified as either
piles and reinforced gravel rafts (EQC, 2015). “drill and fill”, such as the GP3 system for non-caving
Since predicting densification is a very useful tool for soils, or “displacement” such as the Impact system.
the design engineer when liquefaction mitigation is
required, the authors recognize the importance of 2.1 Impact system
having reliable predictive correlations and having know The advantage of the Impact system is that it allows
ledge of which regions in the world these correlations for the construction of a Rammed Aggregate Pier
may need to be modified for. The aim of the article is to element by displacement of the in-situ soil, aided by
present the state of the art of predictive methods for the use of a vibratory hammer.
DOI: 10.1201/9781003308829-151
1004
The use of this technology is particularly suitable mandrel and densifies the surrounding granular soil,
in the case of loose sand deposits below the ground thereby reducing this risk.
water table and for soils where it is not possible to Used in a large number of projects worldwide
carry out drilling without supporting the walls of the including Italy, the effectiveness of the system in
drilled cavities. mitigation of liquefaction risk has been demonstrated
The mandrel consists of a steel tube with by both on-site tests and real case histories. In add
a specially designed compaction chamber at the tip for ition, their effectiveness in mitigating liquefaction in
achieving high vertical compaction of the aggregate clean sandy soils has been well documented by Far
and, consequently, lateral expansion of the cavity. rell et al. (2010); Majchrzak et al. (2010); Wissmann
A sacrificial plate is typically used for the initial drive et al. (2015) and Saftner et al. (2016).
of the mandrel, after which the mandrel is filled with The effectiveness of RAP Impact elements for
gravel to begin construction of the pier. The mandrel liquefaction mitigation in siltier soil has been dem
is subsequently raised by about 90 cm, allowing the onstrated from the full-scale data set collected at Bri
aggregate to flow into the displaced cavity, and then ceño Bridge embankment; a site reinforced with
lowered to compact the gravel, forming layers 30/ RAP elements that performed well after the Muisne
35 cm thick. The achievable depths are typically up to Mw 7.8 earthquake (Smith & Wissmann, 2018;
15 m, with constructed diameters ranging from 50 to Amoroso et al., 2020; Salocchi et al., 2020). Finally,
60 cm. The construction methodology has been full-scale blast-induced liquefaction tests were
described in detail by Majchrzak et al. (2010). recently carried out in Bondeno, Italy (Amoroso
et al., 2020; Rollins et al., 2021) to directly evaluate
the effectiveness of RAP Impact elements in mitigat
ing liquefaction hazards in silty sands. Also in this
case, improvement with Impact elements was effect
ive in reducing liquefaction-induced settlements to
acceptable levels in comparison with the untreated
soil.
4 PREDICTION OF LIQUEFACTION
MITIGATION WITH CPTU
From a design point of view, there are formulations

and methods available in literature to quantify the
improvement of the soil matrix after installation of
Figure 1. RAP Impact installation sequence. Impact piers for liquefaction risk mitigation. These
are based on in-situ CPTu investigations, which are
usually the only data provided to the engineer at the
The installation process is quick and clean (dry design stage.
installation), and the displacement process produces A popular method to quantify the improvement of
minimal spoils. The end result of this construction the soil matrix after installation of Impact piers is to
method is a very stiff aggregate column, with the estimate the densification expected after the installa
added benefit of densification of the surrounding tion of the piers. Estimating the increase in CPT tip
sandy matrix soil. resistance qc is a common way of considering densi
fication and this will be the method of focus in this
paper. By using the increased qc, which is dependent
3 LIQUEFACTION MITIGATION WITH RAP on the soil type, depth and other parameters, and by
GROUND IMPROVEMENT ELEMENTS referring to the pre-improvement sleeve friction fs
and soil behavior type index Ic, it is possible to
The Impact technology finds its core applications on evaluate the updated Factor of Safety against
structures transmitting medium to medium-high loads, liquefaction.
such as industrial sheds or buildings of medium Additional methods such as evaluation of Shear
height, tanks, silos, embankments or floorings. Stress Attraction (Green et al., 2008; Rayamajhi
Another specialty use of the RAP Impact technol et al., 2012), the increase in lateral confinement
ogy is its application in reducing liquefaction risk in stress and Cyclic Resistance Ratio CRR (Salgado
loose to medium dense sandy soils below the et al., 1997), and use of the stiffness of the Impact
groundwater table. The high vertical compaction and piers to reduce post-liquefaction deformations
vibratory energy applied during Impact pier installa (Rollins et al., 2021) are beyond the scope of this
tions expands the cavity created by the driven paper.
1005
4.1 New Zealand-based correlation both before and after installation of Impact elements;
and comparisons are made between the predicted qc
The study undertaken by Vautherin et al. (2017), pro
increments and the measured qc increments.
vides correlations for predicting the degree of densi
fication after Impact pier installations. The study is
based on a large database of CPT data spread among 5.1 Christchurch, New Zealand sites
80 sites across Christchurch, New Zealand. The sites
The pre and post-installation CPTs presented
were treated with Impact piers after the full-scale
below were performed during the full-scale trails
trials in 2013 indicated their effectiveness in reduc
carried out in Christchurch, New Zealand in 2013
tion of liquefaction risk (EQC, 2015).
(EQC, 2015). The tests were performed in three
The authors introduced the ratio Qc to indicate
different trial sites (Site 3, Site 4 and Site 6), in
improvement, and is defined as the ratio between the
which Rammed Aggregate Pier (RAP) columns
post treatment qc and the pre – treatment qc.
were installed to 4m depth with a triangular spa
The relationships for Qc in clean sand are given
cing of about 1.50 m, 2.00 m and 1.80 m on-center
by equation 1 and 2 as a function of the effective
resulting in Area Replacement Ratios of approxi
confinement stress σ0v0 .
mately 14.5%, 8% and 5%, respectively (Amoroso

For σ0v0 < 30 kPa:
et al., 2015). In all the sites, pre and post-
installation CPTu tests were performed in both nat
ural and treated soil to investigate the effects of
For σ0v0 > 30 kPa: RAP Impact installations on soil properties, with
post-installation CPTs performed at the center of
Þ a pier group.
Figures 2-4 plot the measured pre and post
In silty soils, with soil behavior type index Ic com installation CPT tip resistance qc and the predicted
prised between 1.8 and 2.4, the correlations proposed qc using Equations 1-5 as well as the soil behavior
for Qc are presented below. The authors of the type index Ic obtained from the CPTu carried out
Vautherin (2017) paper highlight that these three equa in untreated and treated soil. The predicted qc
tions should not be used for design given the large scat values were plotted separately for Ic<1.8 and
ter observed in the data. Ic>1.8.
For 1.8 < Ic < 2.0:
With
With R2= 0.06. For 2.2 < Ic < 2.4:
With R2= 0.03.
Impact piers in most of the sites were installed at an

area replacement ratio (ARR) of 8%. ARR values of 5
to 8% are typically used in liquefaction mitigation
projects.
5 PREDICTION ASSESSMENT
The scope of this paper is to both verify the accuracy

of the densification formulas from literature, and to
attempt to identify areas for improvement for future
studies.
In order to assess the reliability of the correlations Figure 2. Comparison of measured pre and post-installation
available, we have considered CPTu tests performed qc profile with depth and predicted ones according to
in Christchurch, New Zealand and Bondeno, Italy exposed correlations and Ic for site 3.
1006
out to assess the effectiveness of RAP elements for
liquefaction mitigation (Rollins et al., 2021). In the
study, 16 RAP elements were installed to a depth of
9.5 m in a square grid pattern at a spacing of 2m on-
center, resulting in an ARR of 5%. The post-
installation CPT was performed at the center of
a group of four piers. The pre and post-installation
CPT results are provided in Figure 5 below obtained
from Rollins et al. (2021).
Figure 3. Comparison of measured pre and post-qc profile

with depth and predicted ones according to exposed correl
ations and Ic for site 4.
Figure 5. CPTu profile pre and post RAP treatment at the

Bondeno test site (Rollins et al., 2021). Pre-RAP qc data is
plotted in black and post-RAP qc data is plotted in red.
The authors were not able to obtain the CPT data

file to plot all the measured and predicted increases in
qc with depth as was done for the New Zealand sites.
5.3 Composite results

We proceeded to analyze several representative layer
data points in each of the CPTs from all four sites to
compare the predicted and measured increases in
CPT tip resistance qc directly. The points were
selected at depths where the pattern of the peaks and
valleys of the pre and post-improvement qc suggest
that a similar layer was detected in both CPTs. Points
where the post-improvement qc was lower than the
pre-improvement qc were not selected. It is noted that
pre and post-improvement soil behavior type index Ic
values may not be closely matched for the selected
points since the post-improvement qc and sleeve fric
tion fs are different than the pre-improvement values
after densification has occurred.
The data points chosen are tabulated in Table 1.
Figure 4. Comparison of measured pre and post-qc profile Data points for the New Zealand Site 3 were only
with depth and predicted ones according to exposed correl selected between 0m and 1m depth as the stratigraphy
ations and Ic for site 6.
below 1m was more dissimilar between the pre and
post-improvement CPTs. The pre and post-
improvement qc values for the Bondeno site were
5.2 Bondeno test site, Italy
visually approximated using Figure 5 as the CPT data
A full-scale trial for blast induced liquefaction was car files were not available. The data between 3.5 and 6m
ried out in Bondeno, Italy in 2018. The test was carried depth at this site was discarded due to observed issues
1007
with aggregate flow during pier installations and the
corresponding lack of cavity expansion (densification).
It is noted that Figure 5 plots the total cone resistance
qt; though, the difference between qt and qc may be
small for the points selected as most points of the
Bondeno data points have Ic values of less than 1.8.
Table 1. Selected data points for comparison. Pre-improve

ment and post-improvement data tabulated in the same row
were considered to pertain to the same soil layer.
Post-improvement
Pre-improvement data points
data points (Measured)
Figure 6. Comparison of predicted increase in qc with
Depth qc Depth qc measured increase in qc profile for all sites.
Site (m) (MPa) (m) (MPa)
NZ Site 3 0.44 2.70 0.34 4.41

6 DISCUSSION AND FUTURE
0.74 5.17 0.51 8.38
DEVELOPMENTS
NZ Site 4 0.79 8.75 0.99 10.72
1.71 2.37 1.79 3.48
This study indicated that the correlations used for estimat
2.0 1.3 2.14 2.59
ing improvement in qc show generally good agreement
2.5 1.13 2.5 2.33
with post-installation qc measurements, despite the uncer
3.34 8.94 3.56 15.72
tainties and variables that can affect these correlations.
4.0 9.09 4.0 14.28
The degree that a soil layer can be densified,
NZ Site 6 1.5 3.15 1.5 5.12
which can be expressed in terms of increase in the
2.0 5.84 2.0 13.7
tip resistance qc in CPTu tests, is a function of the
2.5 6.96 2.5 14.03
lithology and confinement stress. Little difference in
3.0 7.21 3.0 13.42
densification was evident between the Impact pier
3.5 8.06 3.5 14.22
spacings of 1.5 to 2m used in the Christchurch trials.
4.0 9.12 4.0 15.05
Additional research could be undertaken to consider
Bondeno 6.8 9.9 6.8 15.0
the normalized CPT tip resistance qc1n proposed by
7.0 8.3 7.0 10.9
Boulanger (2003), which automatically eliminates the
7.15 9.3 7.25 15.4
influence of the soil’s effective overburden pressure.
7.75 10.5 7.95 15.4
Moreover, New Zealand case histories show that
8.2 9.9 8.2 14.4
in, certain soils, densification occurred below the
8.6 12.0 8.6 14.0
bottom depth of the columns. Clearly this is not con
sidered in the bibliographical formulas and is an area
that needs an in-depth study.
It is also evident that the greatest accuracy is found
in the tests carried out in Christchurch. This is because
some areas of the city, following the Canterbury Earth
The predicted post-improvement qc values were quake Sequence in New Zealand in 2010-2011, were
subsequently estimated with the Vautherin et al (2017) home to a unique large-scale study on the efficacy of
equations using the pre-improvement qc values in the Impact technology along with other type of ground
Table 1 and the respective effective confinement stress improvement. Thus, a large amount of data and site
σ0v0 and soil behavior type index Ic for each data point measurements on the effect of RAP Impact installation
tabulated. The predicted � and� measured increases in has been derived from the Christchurch area.
CPT tip resistance qc Δqc were subsequently esti The final goal is to develop more accurate tools to
mated and plotted in Figure 6. The results suggest improve the accuracy of RAP Impact pier designs for
good agreement between the Vautherin et al (2017) liquefaction risk mitigation. The authors, together with
correlation and the measured qc values for the New the licensed Geopier installer in Italy, have designed
Zealand sites and suggest that the correlation overpre and constructed several liquefaction risk mitigation
dicts the increase in qc for the Bondeno site soils. It is projects with RAP Impact elements in the Emilia-
possible that aggregate flow negatively impacted Romagna region, including the cities of Ravenna,
densification even below 6m depth at the Bondeno Cervia and Forlì. In these sites, “pre and post” CPTu
site. investigations are being carried out to form part of
1008
a future local database. Eventually, a new general cor Amoroso, S., K. M. Rollins, P. Monaco, & A. Thorp (2015).
relation could be proposed, or the correlations could be Use of SDMT testing for measuring soil densification by
specialized to be specific to the Emilia-Romagna ground improvement. 3rd International Conference on the
region. This is an ambitious goal and is expected to be Flat Dilatometer, Christchurch, New Zealand.
developed over a medium-long term time frame. The Boulanger, R.W. (2003). State normalization of penetration
authors hope to include the pre and post-installation resistance and the effect of overburden stress on lique
CPTu tests performed at Bondeno test site (Figure 5) faction resistance. Proc. 11th international conference
on soil dynamics and earthquake engineering and 3rd
into the future Emilia- Romagna local database. international conference on earthquake geotechnical
engineering. University of California, Berkeley.
EQC (2015). Residential Ground Improvement Main Report.
7 CONCLUSION Findings from trials to manage liquefaction vulnerability,
https://www.eqc.govt.nz/assets/Publications-Resources/
The Impact pier technology developed by Geopier Residential-Ground-Improvement-Findings-from-trials-to
Foundation Company is a ground improvement method manage-liquefaction-vulnerability-report.pdf
employed worldwide for reduction in site requiring Farrell, T.M., K. Wallace, & J. Ho (2010). Liquefaction miti
gation of three projects in California. Fifth International
liquefaction risk mitigation. The displacement and
Conference on Recent Advances in Geotechnical Engin
vibro-compaction construction technique improve the eering and Soil Dynamics and Symposium in Honor of
surrounding sandy soils by means of densification Professor I.M. Idriss, San Diego, California.
which subsequently reduces the risk of liquefaction. Green, R.A., C. G. Olgun, & K. J. Wissmann (2008). Shear
Numerous tests and bibliographic case histories stress redistribution as a mechanism to mitigate the risk
have demonstrated the effectiveness of this technology of liquefaction. ASCE GSP-181, Geotechnical Earth
for liquefaction risk mitigation. In addition, predictive quake Engineering and Soil Dynamics IV.
methods are available for the designer to estimate the Majchrzak, M., T. Farrell, & B. Metcalfe (2010). Innovative
densification of the soil based on lithology. soil reinforcement method to control static and seismic
The main purpose of the article was to test the reli settlements. Contemporary Topics in Ground Modification,
Problem Soils, and Geo-support. Geotechnical Special
ability of these methods by means of post-installation Publication No. 187. ASCE Press, Reston, Va., 313–320.
CPTu data from sites around the world. The predict Rayamajhi, D., T. V. Nguyen, A. Ashford, &
ive correlations appear to be a useful assessment tool R. W. Boulanger (2012). Effect of discrete columns on
to help in predicting liquefaction mitigation using shear stress distribution in liquefiable soil. ASCE Geo-
Impact during the geotechnical design. Nevertheless, Congress, 1908–1917.
further efforts need to be made to extend the existing Rollins, K.M., S. Amoroso, P. Andersen, L. Tonni, &
databases and eventually to specialize them geograph K. Wissmann (2021). Liquefaction mitigation of silty
ically by studying and evaluating site-specific correl sands using Rammed Aggregate Piers based on
blast-induced liquefaction testing. Journal of Geotechnical
ations. In this direction the authors are already
and Geoenvironmental Engineering Volume 147 Issue 9.
carrying out more in-depth studies, focused on design Salgado, R., R. W. Boulanger, & J. K. Mitchell (1997). Lat
phase, that may be the subject of future publications. eral stress effects on CPT liquefaction resistance
correlations. Journal of Geotechnical and GeoEnviron
mental Engineering, Vol. 123, No. 8.
ACKNOWLEDGMENTS Salocchi, A. C.; L. Minarelli, S. Lugli, S. Amoroso,
K. M. Rollins, D. Fontana (2020). Liquefaction source
The authors would like to thank Sara Amoroso for layer for sand blows induced by the 2016 megathrust
providing the data of on-site tests presented here. earthquake (Mw 7.8) in Ecuador (Boca de Briceño).
Journal of South American Earth Sciences. - ISSN
0895-9811. – 103, pp. 1–10.
REFERENCES Smith, M. E. & K. Wissmann (2018). Ground improvement
reinforcement mechanisms determined for the Mw 7.8
Amoroso, S., K. M. Rollins, P. Andersen, G. Gottardi, L.Tonni, Muisne, Ecuador, earthquake. 5th Geotechnical Earth
M. F. Garcia Martinez, K. Wissmann, L. Minarelli, quake Engineering and Soil Dynamics Conference:
C. Comina, D. Fontana, P. M. De Martini, P. Monaco, Liquefaction Triggering, Consequences, and Mitigation
A. Pesci, V. Sapia, M. Vassallo, M. Anzidei, A. Carpena, - GEESDV, Austin, Texas, GSP 290, pp. 286–294
F. Cinti, R. Civico, I. Coco, D. Conforto, F. Doumaz, Vautherin, E., C. Lambert, D. Barry-Macaulay, & M. Smith
F. Fannattasio, G. Di Giulio, S. Foti, F. Loddo, S. Lugli, (2017). Performance of Rammed Aggregate Piers as a soil
M. R. Manuel, D. Marchetti, M. Mariotti, V. Materni, densification method in sandy and silty soils: experience
B. Metcalfe, G. Milana, D. Pantosti, A. Pesce, A. C.Saloc from the Christchurch rebuild. 3rd International Confer
chi, A. Smedile, M. Stefani, G. Tarabusi, & G. Teza (2020). ence on Performance-based Design in Earthquake Geo-
Blast-induced liquefaction in silty sands for full scale test technical Engineering. Vancouver, BC, Canada, No. 215.
ing of ground improvement methods: Insights from Wissmann, K. J., S. van Ballegooy, B. Metcalfe,
a multidisciplinary study. Eng. Geol. 265 (Feb ): 105437. J. N. Dismuke, & C. K. Anderson (2015). Rammed
Amoroso, S., K. M. Rollins, K. Wissmann, & L. Minarelli Aggregate Pier ground improvement as a liquefaction
(2020). Estimation of lateral spreading by SPT, CPTU and mitigation method in sandy and silty Soils. 6th Inter
DMT following the 2016 Mw7.8 Ecuador earthquake. national Conference on Earthquake Geotechnical
Proc. ISC’6 Conference, Budapest, Hungary. Engineering. Christchurch, New Zealand.
1009
© 2022 the Author(s), ISBN 978-1-032-31259-0
Estimating bearing capacity of polar snow using the Cone Penetration Test
(CPT)
A.B. McCallum & G. White
School of Science, Technology and Engineering, University of the Sunshine Coast, Sippy Downs, Australia
ABSTRACT: Increased access to the polar regions requires increased infrastructure. Design and construction
of this infrastructure can only be achieved by accurately estimating surface bearing capacity. We reviewed con
temporary methods for determining bearing capacity in soils and applied these to data obtained from almost
100 cone penetration tests conducted in Antarctica. Numerous direct and indirect methods exist to enable esti
mation of surface bearing capacity in polar snow, but the preferred method is the rate-controllable, friction-
sleeve equipped, cone penetration test (CPT). Application of this and similar techniques is essential to ensure
the correct design and construction of infrastructure in ever-more trafficked polar regions.
1 INTRODUCTION In this paper we validate direct derivation of bear

ing capacity in polar snow from rate-controllable,
As our climate changes, contestation for polar friction-sleeve equipped, cone penetration test data.
resources is increasing (Dodds and Hemmings,
2015). This results in increased polar operations and
the need for increased supporting infrastructure. 2 BACKGROUND
White and McCallum (2018) reviewed the design
and construction of pavements of snow and ice, Many authors have examined the derivation of bearing
including measures of snow bearing capacity. capacity and/or settlement of geomaterials, either dir
Abele (1990) says that although the strength of ectly or indirectly. Direct measurement of bearing cap
snow can be determined by various methods, there acity is derived directly from CPT data after
are three primary measurement techniques: surface consideration is given to differences in the cone shape,
loading, sample strength testing, or probing, in size and penetration rate, by applying modification
which a penetrative device is forced into the snow factors. In contrast, indirect measurement includes
obtaining a proxy for snow strength. Of these, prob measuring material parameters derived from CPT
ing is the preferred means of assessing in situ data, and then converting to bearing strength via stand
strength because it allows the profile of the snow ard equations. The indirect approach is the common
bearing strength to be measured on undisturbed and conventional method for polar snow. These indir
in situ material (McCallum, 2013). ect methods use the application of standard bearing
Numerous instruments have been used historic capacity theory (Terzaghi, 1943), cavity expansion
ally to probe snow, such as the Swiss Rammsonde theory (Yu, 2000) and various numerical models.
(Haefeli, 1936) or the Standard Penetration Test The measurement of snow bearing capacity is
(SPT) or Drop Cone Penetrometer (DCP); more informed by the practices applied to other geomater
recently lightweight penetrometers have been ials, such as soil. Many approaches for direct bearing
developed for alpine snow (predominantly ava capacity derivation in soils are described in Lunne
lanche) work (Schneebeli and Johnson, 1998). (1997), including those by Schmertmann (1978),
However, because polar snow is a rate-sensitive Meyerhof (1976) and Tand (1995). Additionally,
material (McCallum, 2017a) a robust, hydraulic Eslami and Gholami (2006) presented an analytical
ally-driven, constant-rate penetrometer, such as model to directly derive ultimate bearing capacity
that used for the Cone Penetration Test (CPT), is from cone resistance and Lee and Salgado (2005) pro
preferable for assessing physical characteristics of posed a method to directly estimate the bearing cap
hard polar snow, to depth (~5 m), from which snow acity of circular footings on sands based on the cone
bearing capacity might be estimated. The existence penetration test.
of a friction-sleeve further increases the usefulness Mayne, who has written extensively on the applica
of this tool (McCallum, 2017b). tion of both methods (e.g., Mayne (2020)) noted in
DOI: 10.1201/9781003308829-152
1010
a recent report (Dagger et al., 2018), that many tech The US Army Cold Regions Research and Engin
niques (particularly indirect techniques) are now obso eering Laboratory (CRREL) has used an indirect
lete and calculation of bearing capacity from cone method to estimate bearing capacity for polar infra
penetration test data is now the preferable method. structure such as roads, runways and buildings.
Having briefly examined these techniques in soil, White and McCallum (2018) recently reviewed this
we now briefly inspect similar methods that have approach; an example nomogram from which neces
been considered for polar snow. sary snow hardness/strength criteria must be met for
various aircraft is shown (Figure 1).
3 HISTORICAL ESTIMATION OF BEARING

CAPACITY IN POLAR SNOW
Abele (1963) examined the correlation between snow

compressive strength and penetration resistance, as
measured by the Rammsonde. Furthermore, Mellor
and Smith (1966) examined the impact of tempera
ture on Rammsonde measurements in homogeneous
snow. They showed that Rammsonde hardness varied
with snow temperature similarly to unconfined com
pressive strength. Salm (1971) examined the rheo
logical behaviour of snow under high stresses and
Mellor (1975) extensively examined snow mechanics,
including the rheology, elastic and viscoelastic behav
iour, creep behaviour, failure and shear strength.
Mellor (1975) noted that the derivation of strength Figure 1. Required hardness (or strength) of a snow pave
measurements for snow from Rammsonde penetra ment for various wheel load conditions (from Abele, 1990).
tion resistance data is somewhat ‘arbitrary’ but upon
further examination (because of the utility of the
Rammsonde) he concluded that correlation between For surface polar infrastructure, Blaisdell and
ram resistance and uniaxial compressive strength in Weale (2006) examined bulk and differential settle
snow is important, but variations in snow density and ment into the snowpack of a foundation on a deep
microstructure complicate the relationship, meaning snowpack at the South Pole (stratigraphy was not
a universal correlation is unlikely to be found. described) and Geduhn and Enss (2006) assessed the
Barthelemy (1975) summarised snow-road con service life of Neumayer Station, Antarctica, defining
struction techniques and examined the conversion of failure in terms of settlement and establishing a stress
Rammsonde penetration resistance to hardness amount to be avoided in order to deny the onset of
values, and additionally related evident relationships tertiary creep (McCallum, 2012). Aside from White
between Ramm hardness, confined shear strength and McCallum’s review of ice and snow runway
and expected bearing potential. pavements (White and McCallum, 2018), limited
More recently, Petrovic (2003) briefly examined additional consideration of bearing capacity deriv
a wide array of snow and ice mechanical properties ation (direct or indirect) for polar snow is apparent.
and Shapiro et al. (1997) in re-reviewing the state of
snow mechanics presented an indirect calculation for
settlement based on snow density but highlighted 4 CONTEMPORARY ESTIMATION OF
that estimating the natural snow’s bearing capacity is BEARING CAPACITY IN POLAR SNOW
not easily determined, again because of variations in
both density and microstructure/bonding. Models are Because snow bearing strength is rate-sensitive, tem
being developed to assist in such estimation, but perature dependant and variable with depth/location,
necessary empirical data for complex snow load McCallum (2012) postulated that a hydraulically-
response does not exist (Shapiro et al, 1997). driven, constant-rate, friction-sleeve equipped pene
Land and Harrison (1995) examined the stress- trometer, could be useful to assess bearing capacity of
deformation behaviour of snow of various densities polar infrastructure, such as snow roads and runways.
and Scapozza and Bartelt (2003) used a specially McCallum conducted almost 100 CPTs in Antarctica
constructed triaxial testing device to examine the and presented initial postulations on the direct and
behaviour of snow of different density at different indirect derivation of surface bearing capacity for polar
strain rates and confining pressures, also examining snow, in layered and nonlayered situations (McCallum,
the work hardening that occurs after initial failure. 2012). He focussed on direct methods for layered and
So, historically, the estimation of bearing capacity in unlayered snow, in terms of both ultimate (elastic
snow has proven difficult, either directly or indir deformation only) and allowable (plastic deformation/
ectly, because of the variability in snow density and settlement occurs) bearing capacity. Typical CPT data
microstructure, which also changes with temperature. considered (Figure 2) and McCallum’s postulations
1011
where A and α are parameters; h is the initial sample
height; Δh is settlement; γo is initial specific weight
and γ is final specific weight, with parameters of
state A and α depending on microstructure, initial
density, temperature and loading rate.
Abele and Gow (1976) examined settlement of
compacted snow due to compressive stress and estab
lished stress-deformation (pressure-sinkage) relation
ships that depended on initial density and temperature.
Shoop (1993) examined field terrain characteriza
tion for vehicle trafficability on snow using the Ramm
sonde and other devices, from which bearing capacity
could be indirectly calculated; specifics were not
given.
Figure 2. Example CPT data from which McCallum Fellin and Lackinger (2007) used cohesion values
(2012) derived his relationships, for snow of particular ini obtained from rapid unconfined uniaxial compres
tial density.
sion tests to design strip footings for cable car
towers founded upon glacial ice, and Masterson
comprehensively examined ice bearing capacity and
construction (Masterson, 2009).
Table 1. Methods to derive BC from CPT data.
Lee (2010) modelled plate indentation in snow
BC (kPa) for qcavg using FEA (ABAQUS), calculating an initial yield
Method Relationship strength in compression using the Drucker–Prager
of 2 MPa
cohesion (Pa) and snow friction angle. Lee back-
Ultimate homogeneous qcavg/∼ 3 ~ 667 calculated these data from indentation tests via an
(Method 1) optimization procedure.
Ultimate homogeneous qcavg/∼ 2.2 ~ 909 McCallum (2012) deliberated extensively on the
(Method 2) derivation of strength and bearing capacity (Table 1)
Allowable qcavg/∼1.5 ~ 1330 from CPT in polar snow, presenting an equation to
homogeneous derive ultimate (elastic) bearing capacity (MPa) for
Ultimate layered Depends on N/A homogeneous snow of initial density ∼ 450 kg m−3
strata (Equation 2):
Allowable layered Depends on N/A
strata
Allowable homoge- qcavg/∼ 2.4 ~840
neous (direct)
Where M equals 5.3 for snow of initial density ∼

450 kg m−3 (McCallum, 2012). McCallum (2012)
(Table 1) are presented; qcavg is average tip resist also built on Johnson’s work (Johnson, 2003) to
ance (MPa). show how his Statistical Micromechanical Theory of
We now briefly further examine the direct and Cone Penetration in Granular Materials might be
indirect estimation of bearing capacity in both homo applied to estimate shear strength from CPT for the
geneous and layered snow. cohesive granular material, polar snow.
McCallum (2013) presented preliminary relation
Homogeneous snow ships to derive snow density from both CPT tip resist
ance and CPT sleeve friction. Strong qualitative
Homogeneous snow can be considered elastic (Poulos relationships were evident, but quantitative relation
and Davis, 1974) prior to initial failure (McCallum, ships were not significantly examined. Additionally, he
2012) and ultimate or allowable bearing capacities can presented initial deliberations on deriving snow shear
be estimated either indirectly or directly. strength from CPT tip resistance data; but this deriv
ation relied on knowing initial snow density values.
Indirect Methods Further to this work, McCallum (2014) applied
Kartashov (1965) examined the relationship between relationships developed by Robertson (2010) for soil,
settlement and pressure, for snow of varying density to show that a statistically significant relationship
and determined a power-law relationship (Equation 1): exists between CPT sleeve friction and snow density.
1012
Direct Methods resistance to penetration, but retrieved data are rela
Irwin (1991), though controlled laboratory experi tive, not absolute, and a measure of neither snow
ments in manufactured snow, derived an expression strength nor bearing capacity is attained.
for footing load (bearing capacity) based on pres- Most recently, Haehnel et al. (2019) examined the
sure-bulb depth, shear stress and compressive stress, Phoenix compacted snow airfield near McMurdo
derived directly from Rammsonde hardness data, for Station, Antarctica, to determine strength necessary
snow of variable density, temperature, ageing and to support a C-17 aircraft. They defined strength
grain size. This built on previous work in natural depth profiles in terms of uniaxial compressive
snow (Irwin, 1989) and incorporated volumetric strength, determined via numerous variable-rate
compaction of the snow under the footing and shear penetrative methods, including the Rammsonde
resistance around the cylindrical perimeter. (RAM), Dynamic Cone Penetrometer (DCP) and the
McCallum (2012) built on this theory and devised Russian Snow Penetrometer (RSP). They used elas
a direct derivation of allowable bearing capacity for tic modulus estimated from snow density (kg m-3)
polar snow (Table 1) from CPT data by incorporat (Shapiro et al., 1997) as input into a back-calculation
ing empirically observed differences in cone shape, tool “FAA Backcalculation” (BAKFAA V. 2.0*)
size and penetration rate, also incorporating compac from which Major principal stresses versus depth
tion ahead of the penetrating cone. were predicted, for different aircraft at varying snow
McCallum (2018b) also applied Einav’s work densities. Subsequently, they generated recom
(Einav, 2007a; 2007b) on breakage mechanics mended structural design strengths for the runway, to
theory to postulate on the estimation of pile end a depth of 1 m, in terms of unconfined uniaxial com
bearing capacity from polar snow CPT data. pressive strengths, derived from penetrometer
strength indices for both the RSP and RAM (e.g.,
Non-homogeneous snow Equation 3 for deriving unconfined uniaxial com
pressive strength from RAM data (Abele, 1990).
In the more-complicated case of non-homogeneous
snow, less indirect and direct methods have been
examined.
Indirect Methods No additional contemporary indirect methods to

Sturm et al. (2004) used a micromechanical theory derive bearing capacity in non-homogeneous polar
to convert high-resolution micro-penetrometer data snow are known to exist.
to snow bond strength and Geduhn and Enss (2006)
examined the stress dependent creep behaviour of Direct Methods
snow and ice to assess the bearing capacity and McCallum (2012) examined existing direct methods
settlement of columns used to support Germany’s for soil such as Meyerhof (1976) and Schmertmann
Antarctic Station Neumayer III. They used an ideal (1978) to estimate what the ultimate bearing capacity
ised depth-density curve to represent actual site of snow might be as a percentage of CPT tip resist
stratigraphy. ance. Although he considered the direct manipulation
To estimate the extent of vertical stress transfer of CPT data due shape, size and penetration rate to
within layered snow, estimation of elastic modulus is derive bearing capacity in homogeneous snow (previ
necessary. Marshall (2005) used ABAQUS to model ously discussed), he did not examine the direct deriv
surface loading within a layered snowpack. He ation of surface bearing capacity from CPT data in
refined a statistical-micromechanical model (Johnson layered non-homogeneous polar snow.
and Schneebeli, 1999) to define a macroscopic elas No other research is known to exist that considers
tic modulus based on mini-penetrometer data, con the derivation of bearing capacity in layered polar
cluding that this method could prove valid for snow directly from CPT data.
(larger diameter) CPT data.
McCallum (2012) subsequently derived macro
scopic elastic modulus measurements directly from 5 APPLICATION OF CONTEMPORARY SOIL
CPT tip-resistance (for snow of assumed density), METHODS TO POLAR SNOW
obtaining results consistent with values reported by
Mellor (1975), concluding that the results were strain The limited research on direct snow bearing strength
rate dependant. McCallum went on to examine ultim measurements, coupled with the low reliability of
ate and allowable bearing capacity of layered polar indirect measurement, indicates there is significant
snow, using CPT-derived modulus data, subsequently advantage in applying contemporary soil bearing
generating plots of vertical stress diminution with strength methods, such as those proposed by Dagger
depth using the FEM package Pdisp (Oasys Ltd, (2018) and Lehane (2019).
2009). Lehane (2019) examined how CPT end-resistance
Subsequent to McCallum’s work the Avatech SP1 (qt) can be used to directly predict bearing capacity
and SP2 Smartprobes (Avatech, 2014) were devel for driven and bored piles in sand, silts and clays,
oped to measure snowpack stratigraphy and and shallow foundations on sand; he also examined
1013
shaft friction correlations (similar to McCallum’s Examining Robertson’s direct equation, for
work in snow (McCallum, 2018)). Contemporan a surface footing, suggests that (depending on footing
eously, the behaviour of shallow foundations on B/D etc.) bearing capacity equals ~0.5. qc(av). This is
snow is of most interest. Lehane (2019) deduced very similar to McCallum’s direct derivation of ~0.42.
a direct relationship between CPT tip resistance and qc(av). Again, application of a direct soil method for
shallow foundation settlement (Equation 4). cohesive soils to polar snow appears useful.
Dagger (2018) emphasise that traditional (indirect)
techniques are no longer required because direct
methods from CPT exist for square, rectangular and
circular shallow footings. They propose robust direct
where q0.1 equals the applied bearing pressure where methods for all soils and propose a general direct CPT
foundation settlement divided by foundation width method, but derivation is particular to footing width
equals 0.1, and qc, average equals average cone tip and accepted settlement, and is not easily comparable
resistance. This relationship is very similar to that here.
derived by Mayne (2012) where ultimate surface
footing bearing capacity equals 18% of mean CPT
tip resistance. It is interesting how both these formu 6 CONCLUSION
lations are of a similar order to McCallum’s initial
formulation (Table 1; McCallum (2012). Many techniques, both indirect and direct, exist to esti
However, snow is not sand, and although mate surface bearing capacity in soils; far less research
Lehane’s derivation (from Terzaghi’s initial formula has been conducted into snow, a cohesive, rate-
tion (Terzaghi, 1956)) considers the effects of hori sensitive material, whose evolving microstructure
zontal stress, soil compressibility/stiffness, friction depends on initial density and ongoing temperature
angle and consolidation, it does not incorporate gradients.
cohesion, which snow immediately gains upon for Indirect techniques for snow have been derived
mation (Szabo, 2007). Boufrina (2018) also exam (McCallum, 2012 etc.), but in soils, such techniques
ined design of shallow foundations directly from are no longer preferable, because robust direct-
CPT data but they also did not consider cohesion. derivation methods now exist (Dagger, 2018).
Robertson (2010) briefly considered direct and Direct methods derived for snow, such as those
indirect estimation of shallow foundation bearing postulated by McCallum (McCallum, 2012), agree
capacity from CPT data in cohesive soils (Equa well with accepted methods for cohesive soils
tions 5 & 6 respectively). Indirect: (Robertson, 2010). Therefore, the preferred method
for deducing surface bearing capacity of polar
snow, is through application of rate-controllable,
friction-sleeve equipped, cone penetration test
(CPT) data.
Where qf is ultimate bearing capacity, Nc is
a function of footing width and shape, su is ACKNOWLEDGEMENTS
undrained shear strength (corrected for Bjerrum’s
correction (Bjerrum (1973)), γ is soil unit weight and McCallum’s CPT research in Antarctica was sup
D equals depth of footing. Direct: ported by the Menzies Foundation, Lankelma
Limited (UK) and the British Antarctic Survey.
REFERENCES
Where qf is ultimate bearing capacity, Ksu varies Abele, G., 1963. A correlation of unconfined compressive
from 0.3 to 0.6 depending on footing B/D and shape, strength and ram hardness of processed snow, USA
and soil OCR and sensitivity, γ is soil unit weight CRREL Technical Report 85.
and D equals depth of footing. Abele, G., 1990. Snow Roads and Runways, USA CRREL
For a surface footing, application of the indirect Monograph 90-3.
equation suggests a bearing capacity of ~5.su. This is Abele, G. and Gow, A. J., 1976. Compressibility Character
very close to McCallum’s derived relationship between istics of Compacted Snow, USA CRREL Report 76-21.
CPT tip resistance and shear strength for polar snow of Bjerrum, L. 1973. Problems of Soil Mechanics and Con
struction on Soft Clays and Structurally Unstable Soils.
density 400 kg m-3 of ~4.5.su (McCallum, 2012).
In Proc., Eighth International Conference on Soil Mech
Robertson’s indirect relationship is directly applicable anics and Foundation Engineering, Moscow, Vol.3, pp.
to snow, for snow of density 407 kg m-3. Contempor 111–159.
ary indirect methods to estimate surface footing bear Dagger, Saftner, Mayne, 2018. Cone Penetration Test
ing capacity for cohesive soils, can probably be used Design Guide for State Geotechnical Engineers, Minne
for polar snow. sota DoT.
1014
Dodds, K. and Hemmings, A.D., 2015. Polar Oceans: Sov McCallum, A. B., 2013. CPT: A valuable tool for investi
ereignty and the Contestation of Territorial and gating polar snow, NZ Journal of Hydrology, 52(2).
Resource Rights McCallum, A. B., 2014. Direct estimation of snow density
Eslami, A. and Gholami, M., 2006. Bearing capacity ana from CPT, Proceedings 3rd International Symp. on CPT.
lysis of shallow foundations from CPT data, Proceed McCallum, A. B., 2017. CPT in Antarctic firn: an intro
ings of the 16th International Conference on Soil duction to interpretation, Journal of Glaciology, 60
Mechanics and Geotechnical Engineering. (219).
Haefeli, R., 1936. “Beitrge zur gologie der schweiz geo McCallum, A. B., 2017. Assessing mass balance with the
techn.”. Serie-Hydr, 3. cone penetration test, Journal of Glaciology, 63 (239).
Irwin, G. J., Mohamed, A. M. O., Alammawi, S. and McCallum A. B., 2018. Applying breakage mechanics
Yong, R. N., 1991. Prediction of load carrying capacity theory to estimate bearing capacity from CPT in polar
of deep snow by rammsonde hardness, Journal of Terra- snow, Proceedings 4th International Symp. on CPT.
mechanics, Vol. 28. McCallum, A. B. and Wiegand, A., 2018. Simple Method
Johnson, J. and Schneebeli, M., 1999. Characterizing the for Estimating Snow Strength Using CPT Sleeve Friction
microstructural and micromechanical properties of Data. Journal of Cold Regions Engineering, Vol. 32 (4).
snow, Cold Regions Science and Technology. Mellor and Smith, 1966. Strength studies of snow, USA
Kartashov, S. N., 1965. Mechanical properties of snow and CRREL Research Report 168.
firn”, in: “Symposium international sur les aspects Meyerhof, G.G., 1976. Bearing capacity and settlement of
scientifiques des avalanches de neige”, 69, Union de pile foundations (11th Terzaghi Lecture).
Geodesie et Geophysique internationale. Petrovic, J.J., 2003. Mechanical properties of ice and snow
Lee J., and Salgado R., 2005. Estimation of bearing cap Journal of Materials Science, 38 (1-6).
acity of circular footings on sand based on Cone Pene Poulos, H. G. and Davis, E. H., 1974. Pile Foundation Ana
tration Test, Journal of Geotechnical and lysis and Design, University of Sydney.
Geoenvironmental Engrg 131 (4). Schmertmann, J. H., 1978. Guidelines for cone penetration
Lehane, B. M., 2019, CPT-based design of foundations, E. test: performance and design, Technical Report, US Fed
H. Davis Memorial Lecture (2017). eral Highway Administration.
Shapiro, L. H., Johnson, J., Sturm, M. and Blaisdell, G.L., Schneebeli, M., and Johnson, J. B., 1998. “A constant-
1997. Snow Mechanics. Review of the State of Know speed penetrometer for high resolution snow
ledge and Applications, USA CRREL Technical Report stratigraphy”. Annals of Glaciology, 26, pp. 107–111.
97-3. Shoop, S., 1993. Terrain Characterization for Trafficability,
Lunne, T., Robertson, P. and Powell, J., 1997. Cone Pene USA CRREL Report 93-6.
tration Testing in Geotechnical Practice, Soil Mechanics Szabo, D. and Schneebeli, M. 2007. Subsecond sintering of
and Foundation Engineering, Vol. 46. ice, Applied Physics Letters, Vol. 90 (15).
Marshall, H. P., 2005. Snowpack spatial variability: Tand, K.E., Warden, P.E. and Funegård, E.G. (1995). Pre
towards understanding its effect on remote sensing dicted-measured bearing capacity of shallow footings on
measurements and snow slope stability, PhD Thesis, sand. Proceedings, Intl. Symp. on Cone Penetration
University of Colorado. Testing, Vol. 2.
Mayne, P. W., 2020. The 26th Széchy Lecture: Use of in- Terzaghi, K., 1943. Theoretical Soil Mechanics, John
situ geotechnical tests for foundation systems. Proceed Wiley & Sons.
ings of the Széchy Károly Emlékkonferencia, published White, G. and McCallum, A B., 2018. Review of ice and
by the Hungarian Geotechnical Society, Budapest: snow runway pavements, International Journal of Pave
12–73. ment Research and Technology, Vol. 18 (3).
McCallum, A. B., 2012. Cone Penetration Testing in Polar Yu, H. S., 2000. Cavity Expansion Methods in Geomecha
Snow, PhD thesis, University of Cambridge. nics, Springer Science + Business Media.
1015
© 2022 the Author(s), ISBN 978-1-032-31259-0
The use of CPT based metamodels to predict the performance of offshore

anchor piles
A Mentani & L. Govoni
F. Bourrier
French National Research Institute for Agriculture, Food and Environment, Grenoble, France
ABSTRACT: The paper presents the development of metamodels for the prediction of the load-
displacement response of steel piles driven in sand subjected to pull-out. Two metamodels are created for the
evaluation of the tensile capacity and initial stiffness of the pile. They were developed based on the outcomes
of a finite element testing campaign, employing models of parameters derived from the tip resistance of cone
penetration tests. Two hundreds finite element simulations, which included various soil-pile configurations,
were required to calibrate accurate metamodels. Assessment of the procedure was carried out with reference
to available data on a model pile and related cone penetration test results. The approach relies on particularly
simplified finite element models, but it can be extended to accommodate modelling features of higher com
plexity. The results find application to the design of offshore piles used as anchors for floating structures.
1 INTRODUCTION displacement curve (De Nicola & Randolph 1993; van

tol & Broere 2006; De Gennaro et al. 2008).
The paper focuses on the drained, static load- The implementation and calibration of these
displacement response of steel open-ended piles models can be, however, a complex and computa
subjected to pull-out. In the offshore environment, tionally onerous task. To overcome this limitation,
tensile loading conditions may become critical for metamodelling techniques can be employed, as they
piles employed as anchor foundations, particularly allow to store the results of finite element analyses in
when used with vertical or taut line moorings. In the simple mathematical functions, which have the
context of offshore wind exploitation, these solutions advantage of an easy implementation and low com
offer an attractive alternative to catenaries, as they putational cost (Sudret 2008).
may allow to contain the area over which a floating In this paper, metamodels are developed to predict
wind farm would extend, aiding with a reduction of some behavioural features of piles driven in
the investment costs (Castro-Santos & Diaz-Casas a homogeneous sand bed. Building up on the experi
2016). ence matured in the context of the CPT-methods over
The tensile response of offshore piles is traditionally the last decade, a simple CPT-based Finite Element
estimated using the shaft load-transfer curve approach, (FE) modelling strategy is adopted to investigate the
combined with ultimate shaft friction prediction pile response when subjected to a tensile load through
methods. The development of load-transfer curves a parametric study. A Polynomial Chaos Expansion
dates back to the Fifties and several formulations are (PCE) metamodel (Xiu & Karniadakis 2002) is built
now available as comprehensively reviewed in Bohn from the results obtained by the FE study. The predic
et al. (2017). In sand, the tensile capacity is now esti tion capacity of the developed metamodel is then
mated according to cone penetration test-based proced assessed with respect to selected data included in the
ures (CPT-methods), which predictive performance ZJU-ICL experimental database of piles driven in sand
was assessed in Schneider et al. (2008). The approach (Yang et al. 2015).
is very accurate and its implementation straightfor
ward, however, uncertainties may arise when selecting
the most suitable formulation among those available 2 FE SIMULATION PROGRAMME
(Foglia et al. 2017, Schmoor et al. 2018).
Finite element or finite difference models A FE parametric study was carried out, in which
can be also employed to describe the pile tensile load- an upward vertical displacement was applied to
DOI: 10.1201/9781003308829-153
1016
a whished in place model pile. They were total stress,
small-strain and static analyses and the software suite
Abaqus FEA (ABAQUS 2014) was used to the scope.
2.1 Details of the FE models

The FE models used in the parametric study
involved a pile foundation of diameter D, length
L and wall thickness t. The pile is subjected to
a drained axial pull-out test from a uniform sand
deposit, that is characterised by a constant value of
the relative density Dr and a constant effective unit
weight (γ’ = 10 kN/m3). The models were axial-
symmetric, with zero displacement boundaries set
at a distance of 15D from the pile shaft and
10D down the pile tip. A sensitivity study was car
ried out which showed that, to avoid any conver
gence issues, a very fine uniform mesh was
required in the vicinity of the pile (Figure 1).
Assuming a fully plugged failure, the pile was
modelled as a solid, deformable element, obeying
to a linear elastic constitutive law. It features
a uniform cross section. The equivalent density and Figure 1. Distribution of the mesh size along the FE
models geometry and applied boundary conditions.
elastic properties were calculated to account for the
section geometry on the pile weight and axial
deformation.
The soil response was modelled as linear-elastic (Schneider et al. 2008). Piles were wished in place,
and perfectly plastic, failing according to the therefore the effects of installation on the soil stress
Mohr-Coulomb criterion. Model parameters are state prior to loading was implemented to ensure that
derived from an artificial cone tip resistance the radial stress on the pile was, at any soil depth,
(qc,FE). The particular form of the trend is gener that predicted by Jardine et al. (1998)
ated according to the relation given by Jamiolk
owski et al. (2003) as this was also used in the
interpretations made in the ZJU-ICL database
(Yang et al. 2015).
where
2.2 Sampling and results

where pa = atmospheric pressure and σ’v0 = in situ
Five independent input variables were considered in
vertical effective stress.
the design of the FE test programme. The pile was
The soil’s Young modulus was prescribed to vary
described by three variables, whose range was estab
with the artificial cone tip resistance according to
lished to encompass the geometries encountered in the
Robertson (2009)
ZJU-ICL experimental database. Two variables were
used for the definition of the soil model: the relative
density and the modulus factor, αE, which were
where
Table 1. Input variables for the FE test programme.
applies to the net tip resistance and Ic = soil behav Input variable Range
iour type index. Soil peak strength and dilation
angles were implemented in the FE models accord Pile diameter D [m] 0.20 – 1.00
ing to well-established Bolton (1986) correlation. Pile slenderness L/D [-] 10 – 70
The critical state interface friction angle was taken Pile wall thickness ratio D/t [-] 10 – 100
constant (δcv = 29°), as it is generally done for steel Soil density Dr [%] 40 – 100
driven piles in case interface tests are not available Soil modulus factor αE [-] 3 – 10
1017
allowed to vary within a realistic range for clean sands and output variables and it approximates the complex
(i.e., Ic = 1.31 – 2.05). All the input variables are col and implicit function defined by the emulated model
lected in Table 1, along with their domain of variation. (this model is either deterministic or random). They
The FE analyses were conducted with certain com are generally grouped into classification and regression
bination of the input variables by using the Latin types. When the aim is to predict a continuous target
Hypercube Sampling technique (LHS, McKay et al. variable, as in the case examined in this paper, the
1979). According to the LHS, each input variable regression type, such as the PCE, is to be used. In this
range is divided into intervals of equal probability. The work, the open-source Python package OpenTURNS
number of intervals is equivalent to the sample dimen (Baudin et al. 2016) was used to build the MMs.
sion and the location of the design point (i.e. the com
bination of inputs) is taken randomly within the
3.1 Details of the PCE
interval. This method allows for an optimum coverage
of the input variable domain and the sample size can A given model is described by a vector X in which
be easily increased. a finite number of input random variables are gath
In this study, the five input variables were first ered. The response vector Y, which collects the
combined to create a LH sample of size 50 (S50), output quantities, can be represented as the applica
which was increased to 100 (S100) and then to 200 tion of a mathematical model to the input vector.
(S200). In Figure 2, the results of the 200 simulations The PCE is an algorithm which approximate this
are shown in terms of normalised vertical force function, and the chaos representation of the
(V/(γ’DL2)) and displacement (w/D). Two outputs response vector is defined as the linear combination
variables were identified along the curves, the nor of selected multivariate orthonormal basis, Ψk(Z),
malised tensile capacity (Vult/(γ’DL2)) and the nor and their corresponding coefficient αk as repre
malised initial stiffness (Kt/(γ’L2)), which was sented by
evaluated as the initial tangent to the curve.
The input combinations and the resulting outputs
were then used for the development and calibration
of the PCE metamodels. To validate the metamodels
further FE analyses were performed on a new
sample of size 50 (Sval). with Z obtained by applying an isoprobabilistic
transform to the input vector (Z = T(X)).
The choice of the family of orthonormal basis
3 DEVELOPING METAMODELS (e.g., Legendre, Hermite, Krawtchouk) depends on
the distribution type of the input variables, which
A metamodel (MM) or surrogate model is the model are rescaled by the isoprobabilistic transform into
of a model, and metamodelling is the process of gener common distribution types (e.g., uniform, normal,
ating such MMs. A metamodel is an explicit mathem binomial). The following step consists on the deter
atical algorithm representing the relation between input mination of the coefficients, αk, associated to each
polynomial basis. These coefficients are estimated
according to a suitable regression strategy (Sudret
2008). The most common are the least squares
strategy that minimise the quadratic error between
the model response and the polynomial approxima
tion, and the integration strategy, which uses the
inner product rules, thanks to the orthogonality and
normality property of the polynomial basis.
3.2 Calibration and validation of the PCE

Two MMs were calibrated using the results of the
FE test programme as follows. The combinations of
the five input variables listed in Table 1 were col
lected in the input vector X, and the selected outputs
(i.e., the normalised pile tensile capacity load and
tangent initial stiffness) represented two response
surfaces (i.e., Y). A uniform distribution of the input
variable was selected as the most suitable to be
applied to sample created with the LHS technique.
Consequently, the Legendre orthonormal polynomial
Figure 2. Results of the FE test programme in terms of
basis were chosen as associated to this distribution
dimensionless force and displacement.
type. As for the evaluation strategy to compute the
1018
polynomials coefficients, the least squares method
was selected as it was shown to provide more accur
ate results, if compared to the integration strategy.
The calibration (i.e. the identification of coeffi
cient αk) of the two MMs was carried out using the
results of the FE test programme originated by the
created samples (S50, S100, S200). Accordingly, three
MMs (MM50, MM100, MM200) were created for each
of the two outputs (Vult/(γ’DL2), Kt/(γ’L2)) to explore
the influence of the sample size on their accuracy. To
the aim, the validation set of input-output combin
ations (Sval) was used, with the predictive coeffi
cient, Q2, defined by
where N is the size of the validation sample (N=50)

and Var(Y) is the variance of the FE model outputs.
The MMs predictions are compared with the
results of the FE testing campaign with reference
to the two considered outputs: the tensile capacity
(Figure 3a) and the initial stiffness (Figure 3b).
Different markers are used to identify the MM’s
predictions built on different sample sizes. Some
scatter is observed in the prediction of MM cali
brated with the smaller sample size (MM50), par
ticularly at low and high output values and this is
particularly evident for the prediction of the ten
sile capacity. The increase in the sample size,
reduces the error at either end of the output distri
bution, with MM200 ensuring an excellent accur
acy, consistent for both the outputs and estimated
to be larger than 0.98.
4 ASSESSMENT OF THE PROCEDURE
4.1 Experimental data

The data used to explore the approach potential as Figure 3. Accuracy of the MMs in predicting: (a) the ten
a predictive tool were selected among those avail sile capacity; (b) the initial stiffness.
able in ZJU-ICL experimental database. The data
base was developed with the scope of validating the
CPT-methods for axial pile capacity. Therefore, the 1. For a further and likely estimate of the results, the
results of the pile loading tests are always accompan expression suggested by ISO standards (ISO, 2016)
ied by the relevant CPT tip resistance profiles. for the implementation of the CPT-methods was also
Among all, the data of a pile subjected to pull-out employed
was chosen. The selection was made to ensure that
the foundation and the soil had characteristics con
sistent with the FE models used in the calibration
procedure and that are describable through the pro
posed set of input. The adopted pile was steel, open-
ended driven, the soil was uniform, dense, fine to
medium flandrian marine sand. The qc profile at the where p’m = effective mean in situ stress.
test location is shown in Figure 4a. The profiles of relative density estimated with
The qc data were processed to estimate the rela equations 1 and 7 are shown in Figure 4b, along
tive density at the pile location according to equation with their average values. Application of equation 1
1019
to the two average relative density values returns the experimental stiffness (about 380 MN/m) is slightly
artificial tip resistance profiles inserted in Figure 4a, underestimated by the FE models, which predicted an
which are implemented in the FE models according average value of 310 MN/m and 270 MN/m with αE
to the procedure described in section 1.1. equal to 5 and 7, respectively. A better fit could be
Table 2 collects the input data used for the FE and obtained with a larger value of the modulus factor.
MMs. These includes the experimental pile geom To assess the ability of the MM to reproduce the
etry, the average relative densities, and two values of experimentally observed behaviour, the most accurate
αE, corresponding to possible upper and lower bound MMs were used (MM200). The predictions for the dif
for Ic, estimated using the information on the test ferent input combinations of Table 2 are inserted in
site available in Jardine et al. (2006). Table 3.
These results are compared with those predicted
with the FE models in Figure 3 (triangular markers),
showing consistency of the MM200 accuracy. Combin
Table 2. Input data for FE and MM.
ing the outputs of the two MMs a bi-linear response
D [m] L/D [-] D/t [-] Dr [%] αE [-] can be drawn and a direct comparison with the experi
mental load-displacement curve can be pursued, as
0.457 42.23 33.8 72.04; 76.25 5 depicted in Figure 6. As the MMs were built to predict
72.04; 76.25 7 selected behavioural features, they were not expected
to capture the entire curve, but to provide a good esti
mation of the initial experimental stiffness and tensile
capacity. Capacity values well compared with the
results of API and NGI methods: 1450 kN and 1559
kN, respectively. A slightly higher estimate was
observed when compared to the prediction of the
UWA, ICP and Fugro approaches, respectively 1304
kN, 1310 kN and 1100 kN.
The results were obtained using a very simplified
FE strategy and produced encouraging results.
A better implementation of the qc profile as input
Table 3. Predictions of MM200.
Dr = 72%; Dr = 72%; Dr = 76%; Dr = 76%;

Outputs αE = 5 αE = 7 αE = 5 αE = 7
Vult [kN] 1597 1478 1641 1433

Kt [MN/ 350 326 280 261
Figure 4. a) CPT data at the pile test location(redrawn m]
from Yang et al. 2015 and artificial profile): (a) evolution
of the cone resistance qc and (b) relative density Dr depend
ing on depth.
4.2 FE models and MM prediction

The data in Table 2 were employed first to assess the
FE strategy presented in section 1.1. This validation
plays a crucial role in the development of MMs, as
their performance relies upon the robustness of the
mechanical models they stem from.
Four FE analyses were performed (as detailed in
Table 2) and the results are shown to compare well
with the experimental data in terms of load-
displacement curves (Figure 5). A close approxima
tion of the initial stiffness and non-linearity prior to
failure are observed, with the experimental capacity
falling in the rather narrow band defined by the
results obtained with the two sets of FE simulations
performed with the two different estimates of relative
density. The average FE capacity is 1443 kN, very
close to the experimental data (1450 kN). The initial Figure 5. Experimental and FE load-displacement curves.
1020
computational cost and are therefore suitable to para
metric studies, which can be, in turn, interpreted in
a probabilistic framework. The procedure, here pre
sented in its essential steps, can be further extended
to accommodate modelling features of higher com
plexity, increasing the number of input variables and
can be employed to predict other behavioural
aspects, increasing the numbers of outputs.
ACKNOWLEDGEMENT
This work forms part of the activities of the project

SEAFLOWER, which has received funding from the
European Union’s Horizon 2020 research and innov
ation programme, under the Marie Skłodowska-
Curie grant agreement No 891826.
Figure 6. Experimental and MM200 load-displacement REFERENCES

curves.
ABAQUS 2014. Analysis User’s Manual. Version 6.14,
Dassault Systemes Simulia, Inc.
Baudin, M., Lebrun, R., Iooss, B., Popelin, A-L. 2016.
variable should be pursued for a more reliable pro OpenTURNS: an industrial software for uncertainty
cedure. A possible way could be to follow the quantification in simulation. In R. Ghanem, D. Higdon
approach recently proposed by Cai et al. 2021. & H. Owhadi (eds.) Handbook of uncertainty quantifica
tion. Springer International Publishing.
Bohn, C., Lopes dos Santos, A., Frank, R. 2017. Develop
5 CONCLUDING REMARKS ment of axial pile load transfer curves based on instru
mented load tests. Journal of Geotechnical and
The paper has presented the development of a novel Geoenvironmental Engineering 143 (1):04016081.
Bolton, M.D. 1986. The strength and dilatancy of sands.
procedure for the prediction of the tensile response of Géotechnique 36(1): 65–78.
steel displacement piles in sand. The approach has Castro-Santos, L. & Diaz-Casas, V. 2016. Floating Off
made use of a metamodelling technique, employed to shore Wind Farms. Switzerland: Springer International
store the results of a FE test programme with the aim Publishing.
of make them available for preliminary design Cai, Y., Bransby, F., Gaudin, C., Uzielli, M. 2021.
purposes. A framework for the design of vertically loaded piles in
The FE models were simple, with soil parameters spatially variable soil. Computers and Geotechnics 134:
easily identifiable from CPT results. The modelling 104140.
choices largely relied on the recent advance in the Foglia, A., Wefer M., Forni, F. 2017. Large-Scale Experi
ments and Load Transfer Analysis of Piled Foundations
prediction of the axial capacity of offshore piles in Supporting an Offshore Wind Turbine. Offshore Site
sand with the CPT-based methods, thanks to which, Investigation Geotechnics, 8th International Conference
available empirical correlations, had been validated Proceeding, pp. 1078-1083(6).
on a large database of model tests. De Nicola, A. & Randolph, M.F. 1993. Tensile and com
The results obtained, although preliminary, have pressive shaft capacity of piles in sand. Journal of Geo-
shown that: technical Engineering 119(12): 1952–1973.
De Gennaro, V., Frank, R., Said, I. 2008. Finite element
– a simple CPT-based FE modelling strategy can analysis of model piles axially loaded in sands. Rivista
produce results which compare well with Italiana di Geotecnica 2: 44–62.
experimental data, encouraging further ISO (International Organization for Standardization) 2016.
validation; 19901-4: Petroleum and natural gas industries – specific
– MMs can be built using a relatively small requirements for offshore structures. Part 4: Geotech
number of FE simulations and provide very nical and foundation design considerations. Geneva,
Switzerland.
accurate results over wide domains of input
Jamiolkowski, M.B., Lo Presti, D.F.C., Manassero, M.
variables; 2003. Evaluation of relative density and shear strength
– inputs of MMs can be easily identified based on of sands from CPT and DMT. Soil Behaviour and Soft
the interpretation of CPT data and produces Ground Construction. ASCE, Geotechnical Special Pub
good estimate of the experimental response. lication 119: 201–238.
Jardine, R.J., Overy, R.F., Chow, F.C. 1998. Axial capacity
Overall, the implementation of MMs is rather of offshore piles in dense North Sea sands. Journal of
straightforward, avoiding the laborious FE modelling Geotechnical and Geoenvironmental Engineering 124
activities. MMs also run at a negligible (2): 171–178.
1021
Jardine, R.J., Standing, J.R. & Chow, F.C. 2006. Some capacity of driven piles in siliceous sands. Journal of
observations of the effects of time on the capacity of Geotechnical and Geoenvironmental Engineering 134
piles driven in sand. Géotechnique 56(4): 227–244. (9): 1227–1244.
McKay, M.D., Beckman, R.J., Conover, W.J. 1979. Sudret, B. 2008. Global sensitivity analysis using polyno
A comparison of three methods for selecting values mial chaos expansions. Reliability Engineering and
of input variables in the analysis of output from System Safety 93: 964–979.
a computer code. Technometrics 21(2): 239–245. van Tol, A.F. & Broere, W. 2006. Modelling the bear
Robertson, P.K. 2009. Interpretation of cone penetration ing capacity of displacement piles in sand. ICE
tests – a unified approach. Canadian Geotechnical Jour Proceedings Geotechnical Engineering 159(3):
nal 46: 1337–1355. 195–206.
Schmoor, K.A., Achmus, M., Foglia, A., Wefer, M. 2018. Xiu, D. & Karniadakis, G.E. 2002. The Wiener-Askey
Reliability of design approaches for axially loaded off Polynomial Chaos for stochastic differential
shore piles and its consequences with respect to the equations. SIAM Journal on Scientific Computing 24
North Sea. Journal of Rock Mechanics and Geotech (2): 619–644.
nical Engineering 10(6): 1112–1121. Yang, Z.X., Jardine, R.J., Guo, W.B., Cow, F. 2015. A new
Schneider, J.A., Xiangtao, X., Lehane, B. 2008. Database and openly accessible database of tests on piles driven
assessment of CPT-based design methods for axial in sands. Géotechnique Letters 5: 12–20.
1022
© 2022 the Author(s), ISBN 978-1-032-31259-0
Spatial interpolation of sparse PCPT data to optimise infrastructure design

M.P. O’Neill, M.F. Bransby, J.P. Doherty & P. Watson
ARC Research Hub for Transforming energy Infrastructure through Digital Engineering (TIDE), Oceans Graduate
School, The University of Western Australia, Australia
ABSTRACT: In an offshore setting the geotechnical data available to infrastructure designers is usu
ally sparse, and judgement is required in using information from sampled locations to estimate design
parameters at unsampled locations. Recent interest in data-centric methods has seen advances in the
interpolation of sparse data via statistical and analytical approaches. This paper demonstrates the imple
mentation of one such approach, applying Bayesian Compressive Sensing and Markov Chain Monte
Carlo techniques to sparse two-dimensional PCPT data. Through a simplified case study, the paper high
lights how the method incorporates estimation uncertainty and its associated impact on the geotechnical
design of a representative foundation.
1 INTRODUCTION Recent improvements in off-the-shelf computational

capability have seen growing interest in the application
1.1 Background of data-centric methods to analyse and use measured
data, with recent publications highlighting advances in
In an offshore project, the number of piezocone
the interpolation, and in some cases extrapolation, of
penetrometer tests (PCPTs) performed either for
data via statistical and analytical approaches (e.g. Zhao
a development layout or along a given alignment
et al. 2020, Cai et al. 2018, Rahman et al. 2021, Shi &
(e.g. a subsea pipeline or power cable) may be
Wang 2021, Wang et al. 2021). Such approaches pro
limited by technical constraints and high costs – and
vide a tool which can be used to question how design
hence data is almost always sparse. In addition, infra
profiles are selected from various types of SI data and
structure layouts may change and investigated loca
quantify whether these profiles result in appropriate
tions may no longer align. Therefore, geotechnical
installation and operational reliability when incorpor
designers are often required to employ a significant
ated into code-based design.
degree of engineering judgement in using the avail
able data (at ‘sampled’ locations) to estimate PCPT
profiles, and by extension soil design parameters, at 1.2 Purpose of study
untested (or ‘unsampled’) locations. The consequen
This paper describes a study to investigate the poten
tial risks to fabrication and installation costs and sat
tial benefits and consequent challenges of using
isfactory in-service performance can be significant.
advanced statistical methods (that considers proxim
From a geotechnical perspective, the intrinsic bal
ity of SI data to the designed infrastructure) to select
ance in such judgement is the trade-off between (i)
inputs for geotechnical design. One particular
the risk of underestimating uncertainty at any point
method (Zhao et al. 2020) uses a two-dimensional
and therefore giving an ‘unconservative’ design pro
(2D) Bayesian compressive sensing-Markov Chain
file which exposes the designed infrastructure to risk
Monte Carlo (hence termed ‘BCS’) approach, which
of failure to install or to fail during its service life,
is compared to a simpler ‘standard’ approach using
versus (ii) the cost of overestimating uncertainty at
an offshore field PCPT dataset.
any point and generating a ‘conservative’ design
The study incorporates the BCS and standard
profile which will lead to increased fabrication and
analysis results into a simple geotechnical design
installation costs. Such trade-offs have generally
assessment of a representative foundation. By com
been achieved using engineering judgement, with no
paring the findings of the foundation assessment,
formal methodologies (apart from anecdotal records
the study highlights the benefit an advanced statis
of installation risk or in-service failure) available to
tical approach provides in addressing uncertainty
assess which side of the balance general practice lies
(risk) in the estimation of soil strength for founda
for different types of site investigation (SI) data,
tion design.
seabed conditions and infrastructure.
DOI: 10.1201/9781003308829-154
1023
2 FIELD PCPT DATA 2.2 Sampled data case scenarios
For this study three ‘sampled data’ case scenarios
2.1 Overview were assessed:
The field PCPT data were obtained at a deep-water • Case 1 considered qnet-f from all five PCPTs as
site offshore north-western Australia. The sediments input;
across the site comprised high plasticity carbonate • Case 2 considered qnet-f from PCPT1/3/4/5 as
muddy silts and silty muds. For this study the five input and excluded PCPT2 (x = 38 m);
PCPTs located along the example survey line (PCPT1 • Case 3 considered qnet-f from PCPT1/3/5 as input and
to PCPT5) highlighted on Figure 1 were considered. excluded PCPT2 (x = 38 m) and PCPT4 (x = 108 m).
The tests were conducted at approximately regular
intervals along the line, with a horizontal distance (x) For Case 2 and Case 3 the excluded PCPTs were
of 142 m separating PCPT5 from PCPT1. Data were considered to be unsampled (unseen) data.
recorded at regular depth increments (Δz) of 0.02 m
3 BCS ANALYSIS OF FIELD PCPT DATA
The BCS method proposed by Zhao et al. (2020) for

assessing sparse geotechnical data allows for the
formal treatment of uncertainty in the estimation of
geotechnical parameters at unsampled locations. The
method is relatively fast, computationally efficient
and well suited to non-stationary data (i.e. multi-
layered soil stratigraphies with spatially varying soil
properties and soil layer boundaries).
Figure 1. PCPT locations.

3.1 BCS analysis parameters
A 2D BCS analysis was undertaken for each case,
Profiles of the measured (field) net cone resistance extending across the 142 m long survey line and
(qnet-f) and excess pore pressure ratio (Bq-f) versus down to 20 m depth, using horizontal and vertical
depth below mudline (z) obtained at the five PCPT spatial increments of Δx = 1.0 m and Δz =
locations are presented on Figure 2 down to z = 0.02 m respectively. The analysis adopted natural
20 m. A cursory review of the data indicates four soil log values of the input qnet-f data to avoid negative
types (units; U1 to U4) exhibiting similar overall estimated net cone resistance (qnet-e) values and to
trends with depth across the survey line down to reflect the generally log-normally distributed nature
20 m depth, as illustrated on Figure 2. It is noted that of the sampled qnet-f data (see Lacasse 1994, Grif
across the survey line (relative to x) qnet-f is reasonably fiths et al. 2009).
uniform with no significant ‘outliers’. For sites with Each BCS analysis comprised 10,500 iterations.
higher variability the estimation of soil strength would Results from the first 500 iterations were discarded,
require a greater amount of input (measured) data. as these were assumed to represent the numerical
stabilisation phase of the analysis. In order to ensure
the final results were statistically independent, the
results from every 20th iteration (from the remaining
10,000 iterations) were saved while those from the
other iterations were also discarded. Hence, each
BCS analysis comprised 500 independent qnet-e sets
across the full 20 m by 142 m z-x cross-section.
The study did not account for measurement uncer
tainty in the PCPT data (this will be addressed as
part of future/ongoing development of the approach).
3.2 BCS analysis results

Results from the Case 1 BCS analysis are presented on
Figure 3 as z-x heat maps showing the mean and coef
ficient of variation (COV) of the estimated net cone
resistance (qnet-m and qnet-cov respectively) determined
from the 500 independent qnet-e cross-sections, with the
PCPT locations (input qnet-f data) high-lighted. The
Figure 2. PCPT measured net cone resistance & excess
pore pressure ratio. qnet-m map shows a similar soil unit structure across
1024
the survey line as that inferred from the measured
data (see Figure 2). The qnet-cov map provides an
informative illustration of the predictive functionality
of the BCS method. The COV at the PCPT locations
is near-zero; this is to be expected, since at these
locations qnet is measured (known; i.e. qnet-f) and the
corresponding estimated values are considered to be
reliable. Conversely, in between the PCPTs the COV
is relatively high, since at these locations no informa
tion on qnet-f is available. Note the COV increases as
the horizontal distance from the nearest sampled
location also increases.
Figure 4. Case 1, Case 2 & Case 3 BCS analyses – mean

estimated & P10-P90 estimated net cone resistance at
8 m depth.
Figure 3. Case 1 BCS analysis – mean & COV of estimated

net cone resistance.
Figure 4 provides a useful demonstration of the

overall structure of the BCS analysis results. The
figure shows plots of qnet-m and the 10th and 90th per
centile ([P10, qnet-p10] and [P90, qnet-p90] respect
ively) estimated net cone resistance versus horizontal
distance along the survey line (x) at a depth (z) of
8 m for the three cases. On each plot the locations of Figure 5. Case 2 & Case 3 BCS analyses – measured, mean
the PCPT measured qnet-f included in the respective estimated & P10-P90 estimated net cone resistance at PCPT2.
analysis input are highlighted. The qnet-m,
qnet-p10 and qnet-p90 profiles essentially represent near- prediction error for PCPT2 (expressed as the average
smooth continuous functions of x. At the PCPT loca absolute value of [qnet-m – qnet-f]/qnet-f) was 16.9 %.
tions the qnet-f values constrain the range of qnet-e, Furthermore, with the exception of a handful of spe
while as the horizontal distance from the nearest cific abrupt changes in the measured qnet-f, the
PCPT increases the qnet-e range also increases (as qnet-p10 and qnet-p90 profiles bound the qnet-f profile
reflected by the widening range between well.
qnet-p10 and qnet-p90). A similar set of profiles at x = 38 m are presented
Further insight into the BCS analysis functionality on Figure 5b for the Case 3 analysis. The overall
may be obtained by interrogating estimated net cone agreement between qnet-m and qnet-f is similar to that
resistance versus depth profiles at specific locations observed for Case 2. For Case 3 the average normal
along the survey line. Figure 5a shows the measured ised prediction error for PCPT2 was 17.1 %.
(qnet-f) profile for PCPT2 (x = 38 m), together with the
mean, P10 and P90 estimated profiles from the Case 2
BCS analysis at x = 38 m. In this scenario where the 4 FOUNDATION DESIGN ASSESSMENT
PCPT2 measured qnet-f was excluded from the BCS
analysis input, we can assess the performance of the A simplistic design assessment of a representative
BCS method in predicting qnet-f. The mean estimated foundation was undertaken which utilised the results
qnet-m profile shows generally reasonable agreement of the BCS analyses (in the form of estimated soil
with the measured profile, although locally there is strength along the survey line z-x cross-section), in
some mismatch. Comparing the mean estimated and addition to an alternative estimate of soil strength
measured profiles, the Case 2 average normalised based on a more standard approach.
1025
4.1 Standard assessment of field PCPT data where At = caisson wall tip area), friction across the
outer caisson wall (fo) and friction across the inner
A ‘standard’ assessment of field PCPT data across
caisson wall (fi = α su-i Ai, where su-i = average
a specified survey area or line for the purpose of
undrained shear strength across the inner caisson
foundation design often involves assuming that the
wall, Ai = inner caisson wall area). In all cases it
seabed forms a single soil zone in the area of interest
was assumed Nc = 9 and α = 0.6, while su-b, su-o and
(covering the planned infrastructure and investigated
su-i were determined assuming su = qnet/Nkt where
locations). The available data is compiled and
Nkt = cone factor = 15.
simple percentile (or quantile) profiles of the data
versus depth are generated. These profiles are nor
mally independent of horizontal location and apply 4.3 Foundation design example
right across the survey area or line.
A set of foundation design assessment results is pre
For this study the standard assessment was based
sented on Figure 7 in terms of Vcap and Vinst versus
on the quantile regression (QR) ‘zonation’ approach
L for D = 2 m and a caisson centre horizontal loca
(Uzielli et al. 2019). An example of this approach is
tion (xc) of 38 m (coincident with PCPT2). The Vcap
illustrated on Figure 6, which shows the Case 1 (five
and Vinst values were calculated considering soil
PCPTs) measured qnet-f versus depth data. Included on
strengths based on the Case 1 and Case 3 BCS esti
the same figure are the corresponding 10th and 90th
mated qnet values (P10 for Vcap, P90 for Vinst) and
quantile (Q10 and Q90 respectively) representative
Case 3 QR estimated qnet values (Q10 for Vcap, Q90
design profiles of the Case 1 qnet-f data. The QR pro
for Vinst). Since the PCPT2 qnet-f data was included
files were generated considering the same soil units
in the Case 1 analysis input, the Case 1 BCS-
inferred from the field PCPT data (see Figure 2).
estimated Vcap and Vinst values for D = 2 m at x =
38 m closely reflect the soil strength profile derived
4.2 Foundation design scenario directly from the PCPT2 measured qnet-f.
Figure 7a shows for all values of L the Case 3
The simple deterministic foundation design scenario
BCS-P10 Vcap is lower than the Case 3 QR-Q10 Vcap,
considered a circular caisson foundation with nominal
which in turn is generally lower than the Case 1 BCS
outer diameter (D) values of 2 m and 10 m, an embed
P10 Vcap. For Case 3 at x = 38 m (midway between
ment length (L) ranging between 5 m and 20 m, and
the closest PCPTs, PCPT1 and PCPT3) no information
a skirt wall thickness (t) defined by D/t = 50.
on qnet-f is available, and therefore the BCS estimated
The design considered the foundation in-place
strength range is sufficiently wide (about the mean,
vertical capacity and installation resistance. The cap
reflecting higher uncertainty) such that the Case 3
acity (Vcap) was simplistically assumed to comprise
BCS-P10 strength is less than the Case 3 QR-Q10
the sum of end bearing across the full caisson base
strength (and the Case 1 BCS-P10 strength, which is
(qb = Nc su-b Ab, Nc = bearing capacity factor, su-b =
assumed to be similar to the PCPT2 qnet-f).
average undrained shear strength across the caisson
The opposite trend is shown on Figure 7b for
base, Ab = caisson base area) and skin friction across
Vinst, implying the Case 3 BCS-P90 strength is
the outer caisson wall (fo = α su-o Ao, α = adhesion
greater than the Case 3 QR-Q90 strength (and the
factor, su-o = average undrained shear strength across
Case 1 BCS-P90 strength, which again is assumed to
the outer caisson wall, Ao = outer caisson wall area).
be similar to the PCPT2 qnet-f).
The installation resistance (Vinst) was simplistic
Considering a scenario represented by Case 3 and
ally assumed to comprise the sum of end bearing
a design requirement to select an appropriate L for
across the caisson wall tip area (qt = Nc su-b At,
a D = 2 m caisson at x = 38 m, the results presented
Figure 7. Case 1 & Case 3 – example foundation vertical

Figure 6. Case 1 – Q10-Q90 design net cone resistance. capacity & installation resistance versus embedment length.
1026
on Figure 7 highlight the implications of the decision Table 1. Foundation design assessment results.
on if the caisson sizing should adopt an estimated soil
strength based on the available (in-hand) data, or Found Case Reprsnt. Average Average
whether additional data (in the form of a PCPT) Centre Sep. Capacity Install. Resist.
should be acquired at x = 38 m. If it were decided not Dist.
to acquire additional data, the design may adopt the xc srep Ratio Rcap Ratio Rinst
relatively conservative Case 3 BCS estimates for Vcap

and Vinst as these account for the relationship between m m D=2m D=10m D=2m D=10m
location (x) and uncertainty in the estimated soil
strength. Conversely, if the additional PCPT was per 72 1 0.0 1.18 1.15 0.82 0.84
formed, the designers could adopt the less conserva 2 0.0 1.18 1.15 0.82 0.85
tive Case 1 BCS Vcap and Vinst estimates. If no extra 3 0.0 1.16 1.13 0.81 0.83
PCPT was performed and the designers adopted the 55 1 12.0 0.99 0.98 0.98 0.98
simpler traditional approach, the resulting QR-based 2 16.2 0.96 0.96 1.06 1.06
estimate of soil strength would underrate uncertainty 3 16.2 0.95 0.95 0.98 0.98
and lead to a design that may be unconservative (by 38 1 0.0 1.20 1.14 0.86 0.88
an amount represented by the difference between the 2 25.3 0.95 0.95 1.10 1.10
Case 3 QR and BCS estimates for Vcap and Vinst). 3 25.3 0.87 0.86 1.09 1.09
4.4 Overview of foundation design assessment

A summary of the foundation design assessment find
ings is provided in Table 1 in terms of the capacity
ratio (Rcap) equal to the BCS estimated Vcap divided
by the QR estimated Vcap, and the installation resist
ance ratio (Rinst) equal to the BCS estimated Vinst div
ided by the QR estimated Vinst. By normalising the
BCS-based capacity and installation resistance values
against their QR counterparts, the Rcap and Rinst results
highlight the ability of the BCS method to account for
location-specific confidence in soil strength. Note the
Rcap and Rinst values contained in Table 1 are average
values for 5 m ≤ L ≤ 20 m. Results are presented for
xc = 72 m, 55 m and 38 m, D = 2 m and 10 m, and
for the three different cases.
The same results are plotted on Figure 8 in terms
of average Rcap and Rinst values versus the corres
ponding representative horizontal separation distance
of the caisson centre from the nearest PCPT locations
(srep = [s1 s2]/[s12 + s22]°.5, where s1 and s2 = horizon Figure 8. Foundation average capacity & installation
tal distance from the caisson centre to the immedi resistance ratios versus representative horizontal separation
ately adjacent PCPTs situated each side of the caisson distance from nearest PCPTs.
centre location; values of srep are listed in Table 1).
Included on the plots are approximate trendline fits to resistance is generally slightly higher than the cor
the results. responding QR estimated value (Rinst > 1.0). Hence,
The results indicate a general trend of Rcap this suggests for high srep the QR approach tends to
decreasing and Rinst increasing with increasing srep slightly underestimate soil strength uncertainty.
(i.e. greater separation between the caisson and The results listed in Table 1 also show that for
known soils data). For caissons centred about a PCPT caissons centred about a PCPT location (srep = zero),
location (srep = zero), the BCS estimated capacity is as the caisson outer diameter (D) increases Rcap
shown to be higher than the corresponding QR esti decreases slightly while Rinst increases slightly. This
mated value (Rcap > 1.0), while the BCS estimated reflects the increasing level of uncertainty in the
installation resistance is lower than the corresponding BCS estimated strength across the caisson footprint
QR estimated value (Rinst < 1.0). Hence, these results as it expands beyond the PCPT location. For cais
suggest for srep = zero the QR approach tends to over sons situated away from the closest adjacent PCPT
estimate the level of soil strength uncertainty. locations (relatively high srep) the opposite trend is
Conversely, for caissons situated some distance expected, although the results indicate minimal
away from the closest PCPT locations (high srep), change in Rcap and Rinst with increasing D. This sug
the BCS estimated capacity is shown to be slightly gests the foundation must be situated within some
lower than the corresponding QR estimated value limiting distance from the nearest adjacent PCPT for
(Rcap < 1.0), while the BCS estimated installation D to influence Rcap and Rinst.
1027
5 CONCLUSIONS such as “do we need to acquire additional geo-data?”,
“where should we locate our infrastructure?” and
5.1 Interpolation of PCPT data and impact on “what size does our infrastructure need to be?”.
foundation design
The study outlined in this paper provides a demon ACKNOWLEDGEMENTS
stration of the application of an advanced statistical
approach to interpolating sparse field PCPT data (spe This research is supported by the ARC ITRH for
cifically net cone resistance) to improve geotechnical Transforming energy Infrastructure through Digital
design inputs for a representative foundation. The find Engineering (TIDE, http://TIDE.edu.au) which is led
ings of the foundation design assessment (summarised by The University of Western Australia (UWA),
in Table 1 and Figure 8) provide a concise quantitative delivered with The University of Wollongong and
example of the benefit offered by such approaches in a number of Australian and international research
addressing the spatial variation in soil strength uncer partners, and funded by the Australian Research
tainty caused by the layout of an SI. This benefit Council, INPEX Operations Australia, Shell Austra
encapsulates both the opportunity to rationally reduce lia, Woodside Energy, Fugro Australia Marine, Wood
foundation size requirements when soil strength uncer Group Kenny Australia, RPS Group, Bureau Veritas
tainty is low (near SI locations), and the need to fully and Lloyd’s Register Global Technology (grant No.
account for risk by increasing foundation size require IH200100009). The second author holds the Fugro
ments when soil strength uncertainty is high (further Chair in Geotechnics at UWA, whose support is grate
from SI locations). fully acknowledged. The fourth author leads the Shell
Chair in Offshore Engineering research team at
UWA, which is supported by Shell Australia.
5.2 Next steps – interpolation of data
As part of future/ongoing development of the appli REFERENCES
cation of advanced statistical approaches to interpol
ate sparse data, a near-term goal will be to Cai, Y., Li, J., Li, X., Li, D. & Zhang, L. 2018. Estimating
investigate options for incorporating additional soil resistance at unsampled locations based on limited
CPT data. Bulletin of Engineering Geology & the Envir
sources of geodata into approaches such as the Zhao onment 2019, 78:3637–3648, https://doi.org/10.1007/
et al. (2020) BCS method featured in this paper. One s10064-018-1318-2.
potential source is geophysical data which may Griffiths, D.V., Huang, J. & Fenton, G.A. 2009. Influence
inform the spatial alignment of soil unit boundaries of spatial variability on slope reliability using
and the identification and selection of soil unit types 2-D random fields. Journal of Geotechnical & Geoen
and possibly soil classification properties. vironmental Eng., Vol. 135, No. 10, pp. 1367–1378.
The BCS approach may also be expanded to Lacasse, S. 1994. Reliability & probabilistic methods.
a three-dimensional (3D) spatial framework. Such Proc. 13th Int. Conf. Soil Mechanics & Foundation
analyses would require significantly more computa Engineering, New Delhi, India, pp. 225–227.
Rahman, Md.H., Abu-Farsakh, M.Y. & Jafari, N. 2021.
tional time and resources. However, in situations
Generation & evaluation of synthetic cone penetration
where data is available over a specific spatial foot test (CPT) data using various spatial interpolation tech
print or survey area, and for the scenario where the niques. Canadian Geotechnical Journal, 58: 224–237
position of future infrastructure within the footprint (2021), http://dx.doi.org/10.1139/cgj-2019-0745.
changes, such a capability would be beneficial. Shi, C. & Wang, Y. 2021. Non-parametric machine learning
methods for interpolation of spatially varying non-
stationary & non-Gaussian geotechnical properties.
5.3 Next steps – applications of interpolated data Geoscience Frontiers, 12 (2021) 339–350, https://doi.
As demonstrated by the foundation design assess org/10.1016/j.gsf.2020.01.011.
Uzielli, M., Zei, M. & Cassidy, M.J. 2019. Probabilistic
ment, the potential geotechnical engineering applica
assignment of design undrained shear strength using
tions of statistical analysis methods which consider quantile regression. Proc. 7th Int Symp. Geotechnical
spatial distribution of seabed uncertainties are signifi Safety & Risk, pp. 188–193.
cant. A possible future application involves using pre Wang, Y., Shi, C. & Li, X. 2021. Machine learning of geo
dicted soil properties at unsampled locations as input logical details from borehole logs for development of
to a suite of forecasting-costing-optimisation tools, high-resolution subsurface geological cross-section &
which may be used to facilitate planning of future geotechnical analysis. Georisk: Assessment and Man
geotechnical SIs and inform real-time decision- agement of Risk for Engineering Systems & Geohazards,
making for current SIs. Similarly, in the context of an https://www.tandfonline.com/action/showCitFormats?
doi:10.1080/17499518.2021.1971254.
offshore setting the tools may be employed in the
Zhao, T., Xu, L. & Wang, Y. 2020. Fast non-parametric
optimisation of infrastructure design (e.g. pipelines, simulation of 2D multilayer cone penetrometer test
cables, platforms, offshore wind turbines and subsea (CPT) data without pre-stratification using Markov
structures). The goal will be to use these tools in Chain Monte Carlo simulation. Engineering Geology,
a project setting to undertake quantitative cost-benefit 273 (202 0) 1 05670 , https://d oi.org / 10.101 6/j.
assessments in order to provide answers to questions enggeo.2020.105670.
1028
© 2022 the Author(s), ISBN 978-1-032-31259-0
Use of CPT as a soil mixing verification tool: Some practical

observations
A. O’Brien & I. Murray
GE Solutions Consulting Ltd., Whitburn, UK
ABSTRACT: Mass soil mixing is a versatile ground improvement technology for marginal and brownfield
sites. Dry soil mixing is relatively common in the UK for the improvement of ground with very wet and/or
organic materials. Wet mixing is less commonplace and involves introduction of a fluid grout with simultan
eous rotavating of the soil with a mixing tool. This paper presents the results of CPT investigations into wet
mixed soils with a view to use the results as a means to verify the strength and consistency of the mixed
materials. A high volume of data has been extracted from two active sites. CPT soundings are found to be
purposeful in terms of demonstrating the overall improvement effect and integrity of the mixed soil volume
spatially and with depth. Existing CPT testing of mass mixed soil is very limited and this testing has allowed
the mixed soil to be compared to existing CPT soil behaviour indexes to give an insight into their usefulness
for characterising mixed material and its behaviour.
1 INTRODUCTION nature of the soil mixing process is such that the entire
matrix of the soil is de-structured with hydration &
Mass wet soil mixing involves disaggregation over cementation subsequently forming an entirely new
large areas / volumes of soil using a rotavating tool. macro- and microscopic condition. The mixed mater
Mixing usually takes place within discrete “cells” and ials would fall outside of the normal soil behaviour
the mixing depth is typically limited up to 5-7m type / index categorization, which is obviously limited
depending on the application and native soil condi to natural materials. However, if found to be reliable,
tions. The rotavating tool spins at high revolutions the use of CPT to re-parameterise the mixed material
per minute (in the order of 80-90rpm) and grout is would have significant technical and commercial bene
injected under medium pressure (typically up to fits including spatial quality control, estimating spatial
50bar). The result is a completely fluidised cell, hom variability of the mixed area and development of geo
ogenising the native soils and engendering the technical parameters for the mixed soil, thereby miti
required strength with curing time. gating the need for high-frequency laboratory testing.
Post-construction behaviour is usually assessed via It is noted that the use of CPT in soil mixing
limited in situ testing, such as plate bearing tests sup applications is currently typically limited to deep col
plemented with laboratory strength testing of samples umns (e.g., Puppala et al., 2005; Ilander et al., 1999),
recovered during construction, typically to establish which are necessarily more focused and higher inten
a specified unconfined compressive strength. However, sity mixing efforts and this is a distinction which
because of the nature of the mixing process, there is must be made for mass mixing, which attempts to
scope for small inclusions of unmixed material to productively improve larger soil volumes, typically
remain where mixing effort is not adequate. Cone to a lower strength than deep soil mixed columns.
penetration testing is a practical means to establish the
strength and consistency of the mixed soil body, par
ticularly where pre-construction and post-construction 2 THE SITES
soundings are undertaken to allow quantification of the
improvement effect. Two sites have been examined in this study,
The use of the cone penetration test for these pur denoted as BWB and JSR – both are active sites in
poses needs careful consideration if the data is to be western Scotland. Site BWB was predominantly
used for anything more than a comparison of pre firm non-organic lightly overconsolidated horizons
construction and post-construction conditions. The of sandy, silty clay or sandy silt with occasional
DOI: 10.1201/9781003308829-155
1029
lenses of silty sand. The groundwater regime is ill-
defined on this site and is potentially subject to
a hydraulic gradient, but is known to be within the
mixed horizon. The minimum mixed depth for the
BWB site is 6m.
Site JSR was characterised by very soft silty clay
with inclusions of organic materials and occasional
bands of organic materials up to 1m in thickness. The
organic materials are nominally described as amorph
ous peat. The site is low-lying between glacial drum
lins. The groundwater table is, again, ill-defined but
typical equilibrium water level is within the mixed
horizon. The minimum mixed depth for the JSR site
is 4.0m.
2.1 Mass mix procedure

For both sites, a target mixing depth was estab
lished and achieved. Cement was added via
a neat cementitious grout (CEM I only) with
added cement proportions measured between 12
15% of the bulk unit weight of the native soil.
This is a typical proportion for this type of
mixing in order to control both material bulking
and overall injected volume. Curing time post-
mixing varied in a narrow range of 25-35 days
when the CPTs were undertaken at both sites.
3 COMPARISON OF PRE- AND

Figure 1. BWB site tip resistances.
POST-CONSTRUCTION CPT DATA
All CPT data in this study was collected using

15cm2 subtraction piezocones manufactured by
Geomil Equipment B.V. The data procedures were
consistently in accordance with EN ISO 22476
1:2012 and all CPTs achieved Application Class 2,
noting that the soil type prohibited achievement of
Application Class 1 as a practical matter.
Porewater pressure measurements were measured
at the cone shoulder (u2) position.
3.1 Tip resistance

Corrected cone tip resistances for both sites are pre
sented in Figures 1 & 2 for the BWB and JSR sites
respectively. Based on mean corrected tip resistance,
post-construction improvement in tip resistance was
estimated as a multiplier of 6.11, with a standard
deviation of 2.71 for the BWB site (mean pre
construction tip resistance of 1.16MPa, standard
deviation of 0.518MPa). The equivalent multiplier
for the JSR site is 4.86 with a standard deviation of
2.27 (mean pre-construction tip resistance of
0.30MPa, standard deviation of 0.11MPa).
3.2 Sleeve friction

Sleeve friction for both sites are presented in Fig
ures 3 & 4 for the BWB and JSR sites respectively.
Based on mean sleeve friction, post-construction
improvement in sleeve friction was estimated as Figure 2. JSR site tip resistances.
1030
a multiplier of 5.33 with a standard deviation of 1.94
for the BWB site (mean pre-construction sleeve fric
tion of 50.5kPa, standard deviation of 17.93kPa).
The equivalent multiplier is 4.55 with a standard
deviation of 2.17 for the JSR site (mean pre
construction sleeve friction of 12.77kPa, standard
deviation of 5.08kPa). A statistical summary of the
primary measured parameters are presented in
Table 1.
3.3 Dynamic porewater pressure

Dynamic porewater pressure profiles are presented in
Figures 5 & 6 for the BWB & JSR sites respectively.
A rational statistical examination of these profiles is
not possible; however, a number of observations can
be made. Pre-construction dynamic porewater pres
sure, noting that cone saturation was tightly controlled
and dissipation tests were run in order to assure satur
ation of the cone where appropriate, is invariably
negative/approaching cavitation pressure or inconsist
ently and modestly positive (u2 where positive is typ
ically less than 50kPa).
Post-construction, large dynamic porewater pres
sures are generated, particularly for the BWB site.
The large porewater pressures are generated generally
towards the middle of the mixed horizon and are
observed to revert towards pre-construction behaviour
at the base of the mixed horizon. Cavitation pressures
were established over some of the mixed horizon at Figure 3. BWB site sleeve friction.
the BWB site and more consistently towards the base
of the mixed horizon at the JSR site. There are no
noted operational reasons for this occurrence.
3.4 Friction ratio

Comparison of pre- and post-post construction friction
ratio observes a degree of inverse proportionality.
Figure 7 presents a comparison of friction ratio pre
and post-construction based on average values. It is
observed that where the pre-construction friction ratio
is initially low (<4%), there is a tendency for post-
construction friction ratio to be elevated. Where the
Table 1. Statistical summary of measured parameters for

BWB & JSR sites.
Site BWB Site JSR

Measured
parameter Stat. Pre Post Pre Post
Corrected Mean 1.160 6.469 0.301 1.298

cone resist- Median 0.990 7.066 0.294 1.221
ance (MPa) St. Dev. 0.518 2.331 0.105 0.441
COV 0.447 0.360 0.350 0.340
Sleeve fric- Mean 50.502 258.010 12.766 49.394
tion (kPa) Median 47.346 273.448 12.264 47.356
St. Dev. 17.934 87.605 5.077 10.389
COV 0.355 0.340 0.398 0.210
Figure 4. JSR site sleeve friction.
1031
pre-construction friction ratio is >5%, there is
a tendency for the post-construction friction ratio to be
reduced. Between 4% and 5%, there is a transition
zone where friction ratio could be either reduced or ele
vated when compared to pre-construction levels.
3.5 Soil behaviour index / type

Normalised soil behaviour indexes (Ic), calculated
based on the work of Robertson (2009), for both
sites are presented in Figures 8 & 9 below for the
BWB & JSR sites respectively using routine proced
ures for normalising the input parameters. The soil
behaviour indexes are observed to undergo a marked
shift post-construction.
Examining soil behaviour type more broadly, the
shifts are in the main from normally to lightly over-
consolidated fine-grained soils towards very stiff fine-
grained soils, capturing the cemented structure. More
recent soil behaviour index methods (e.g., Robertson,
2016) would have categorised the mixed soil as Clay-
like / Dilative or Transitional / Dilative with estimates
of the lower Bq parameter indicative of increasing
microstructure (cementation, bonding). The SBT charts
for the BWB & JSR charts are presented respectively
in Figures 10 & 11, while the mean shift in soil behav
iour type is shown in Figure 12.
Figure 5. BWB site dynamic porewater pressure (u2).

4 OBSERVATIONS
The direct measurements of cone tip resistance and

sleeve friction are observed to increase post-mixing.
Mean tip resistance increased by between 558% (site
BWB) and 430% (site JSR) for respective mean pre
construction tip resistances of 1.16MPa and 0.3MPa.
Similarly, mean sleeve friction increased by between
511% and 387% for respective mean pre
construction sleeve friction of 50.5kPa and 12.8kPa.
The CPT is observed to be purposeful in demonstrat
ing that achieved improvement is less at the base of
the mixed horizon at the time of the measurements.
The naïve coefficient of variation (noting that
mean values were used in these calculations on the
basis of consistent sample sizes) reduces for the
measurement of tip and sleeve resistance over both
sites, though not markedly, mainly owing to the lack
of stratification over both sites i.e., there are no dom
inant horizons in the pre-mixed zone. The magnitude
of the reduction in variation is most pronounced in
the measurement of sleeve friction for the JSR site
where the pre-construction variation was greatest of
the parameters measured (0.398 pre-construction,
0.210 post construction).
There is marked facility to set up high excess pore-
water pressures post-construction. Pre-construction
soils expressed dilative behaviour and cavitation result
ing in negative u2 measurements. However, post-
construction, and most pointedly at the BWB site, very
high excess porewater pressures were measured.
Figure 6. JSR site dynamic porewater pressure (u2).
1032
Figure 7. Pre- / post-construction friction ratio comparison.
Figure 9. Ic for JSR site.
extract from this chart is the trajectory of the soil

behaviour type which is towards a very stiff and fine-
grained categorisation which is indicative of the
cemented nature of the material. As noted, more
modern SBT methods would identify the materials as
cohesive or transitional and dilative. However, the
magnitude of dynamic porewater pressures established
Figure 8. Ic for BWB site.
As noted, the friction ratio measurements appear to

show a trend of inverse proportionality with the pres
ence of a transitional zone of between 4%-5% where
post-construction estimate has been observed to be
both higher and lower than pre-construction estimates.
Soil behaviour index/type were found to markedly
shift post-construction using traditional methods (see
Figure 12). The SBT chart shows indicatively the aver
age shift pre- to post-construction. The main point to Figure 10. Soil behaviour type chart for BWB site.
1033
5 RECOMMENDATIONS FOR FUTURE
WORK
It is clear that current SBT methods are effective

in capturing the effect of the mixing effort, how
ever, even with more modern methods which con
sider soil fabric / microstructure more actively,
current SBT methods are not optimal for a final
categorisation. Key to this is the observation
regarding the contractive / dilative behaviour and
the marked facility to set up substantial excess
porewater pressures. It appears justified that the
assumption of undrained behaviour is reasonable
for design incorporating wet soil mixing of this
nature and to that end, extraction of undisturbed
samples for more detailed laboratory (triaxial)
testing would be purposeful to establish an appro
priate Nk factor for mixed soil.
As noted, the efficacy of the mixing toward the
base of the mixed horizon is less than through the
Figure 11. Soil behaviour type chart for JSR site. main body of the mixed horizon. However, there
are a number of other effects that should be inves
tigated as contributory here. For example, the
oscillation of the mixing tool, which is typically
excavator-mounted in mass mix applications,
could yield small unmixed inclusions which would
distort CPT results at the base. Also, the exother
mic effects of curing merit further consideration in
strength achievement in particular, how the ambi
ent ground temperature can be contributory to sup
pression or acceleration of curing rate which
would be most pronounced at the native soil inter
face. CPT (with thermal measurement capability)
would be purposeful in investigating this in situ as
extracted samples can be subject to storage and
temperature variations that are not representative
of the in-situ condition.
REFERENCES
Puppala, A.J, Bhadriraju, V. & Porbaha, A. 2005. SPT and
CPT based methods to address shear strength of deep
mixed soil cement columns. Proceedings of the 16th
International Conference on Soil Mechanics and Geo-
Figure 12. Soil behaviour type chart for both sites showing technical Engineering, Millpress Science Publishers/IOS
trajectory of mean shift in behaviour. Press 1257–1260
Ilander, A., Halkola, H., Lahtinen, P. & Kettunen, A. 1999.
EuroSoilStab – Kivikko test embankment – Construc
post-mixing is not consistent with dilative material tion and research Proceeding of Dry Mix Methods for
categorisation. Deep Soil Stabilisation. Bredenberg, Holm & Broms
While sub-optimal as a categorisation tool, soil (eds), Balkema, Rotterdam 347–354
behaviour index may have some utility as a measure EN ISO 22476-1:2012. Geotechnical investigation and test
of the integrity of the mixed soil, particularly where ing – Field testing – Part:1: Electrical cone and piezo
the expected behaviour of the soil pre-construction is cone penetration test
Robertson, P.K. 2009. Interpretation of cone penetration
well defined. This is because the soil behaviour
index takes into account both tip and sleeve resist nal, 46: 1337–1355.
ance and can be stress-normalised. With regard to Robertson P.K. 2016. Cone penetration test (CPT)-based
stress normalisation, variation in the bulk density of soil behaviour type (SBT) classification system — an
the mixed material when compared to the native soil update. Canadian Geotechnical Journal, 53:
bulk density, should be accounted for. 1910–1927
1034
© 2022 the Author(s), ISBN 978-1-032-31259-0
Liquefaction damage assessment using Bayesian belief networks

L. Paolella, A. Baris & G. Modoni
University of Cassino and Southern Latium, Italy
R.L. Spacagna & S. Fabozzi

CNR IGAG, Area della Ricerca Roma 1, Italy
ABSTRACT: The seismic and liquefaction risk assessment implies introducing methods based on different
hypotheses and dealing with different levels of uncertainty affecting the whole process from triggering to surfi
cial manifestation. In this context, soft computing methods, like Bayesian Belief Networks (BBN) and artificial
intelligence algorithms, provide the logic framework for cause-effect relationships and the statistical statement
to manage uncertainties. Taking advantage of the significant amount of geotechnical data and post-earthquake
surveys, an application of BBN versus the forecasting of liquefaction-induced ground damage is proposed con
sidering three main shocks of the 2010 – 2011 Christchurch (New Zealand) Earthquakes Sequence. The BBN
algorithms are firstly employed to identify significant variables and learn the relationships among them, then
a direct and graphical link between input and target data is created. The quantitative validation of the built
architecture enables to advantageously queried the net to predict the result of new datasets.
1 INTRODUCTION site-specific measurements like cone penetration test

(CPT) data for probabilistic site characterization.
The chain phenomena describing the liquefaction A detailed study has been carried out by Tesfamar
process involve numerous and complex relationships iam (2013) about liquefaction risk that calibrates four
that rule the origin, propagation, surficial manifest different BBN structures and defines a procedure to
ation, and induced structural and infrastructural assess the liquefaction risk at the regional and single
damage. In the field of Earthquake Engineering, the building levels. Tang et al. (2018) compare ANN
cause-effect relationships starting from the probability technology and BNNs in describing the liquefaction
of occurrence of a predefined intensity measure have ground severity demonstrating that the Bayesian
been translated into the PEER convolutive integral model achieves better accuracy for each damage
(Cornell and Krawinkler, 2000), which develops the state. Taking advances from previous experiences
performance-based earthquake engineering approach. and the considerable amount of geotechnical data,
In recent years, the development of soft comput a BN model for liquefaction prediction is proposed
ing methods and the progress in artificial intelligence in this work. The best structure is obtained after the
A.I. provide robust and reliable instruments capable processing of ≈9ʹ000 CPTs available from the New
of dealing with large amounts of data in a reasonable Zealand Geotechnical Database and testing each per
time, controlling the quality of results and quantify formance against three main events of the 2010-2011
ing uncertainties. Artificial Neural Networks (ANNs) Canterbury Earthquake Sequence. In a preliminary
and Bayesian Belief Networks (BBNs) have been analysis, the back analysis enables the reconnaissance
advantageously used in various engineering applica of the critical layer, i.e., the sandy stratum most
tions to identify the most significant variables, learn likely to undergo liquefaction during the 2010-2011
the relationships and dependencies among them, and Christchurch earthquake sequence. The obtained
link the input data to the target. After scrutinizing the results have shown that, if opportunely guided, the
new data collection campaigns following the 2012 model can relate pre-defined representative variables
Emilia (Italy) Earthquake sequence, Paolella et al. to liquefaction ground evidence, the latter available
(2019) developed artificial neural networks and from specific post-earthquake surveys. In
Monte Carlo simulations to relate the Ishihara-based a subsequent step, more detailed evaluations have
geotechnical model to the observed ground liquefac been performed on such a critical layer, and a set of
tion in the municipality of San Carlo. Ching and site-specific soil fragility functions is proposed.
Phoon (2017) developed a method based on the Uncertainties at different levels are accounted for
sparse Bayesian learning (SBL) approach to analyze through statistical and probabilistic terms, displaying
DOI: 10.1201/9781003308829-156
1035
and controlling each variable. In conclusion, an alter Proven that the probability of a random variable
native approach to large areas studies (Spacagna depends on his parents, a Bayesian Network can be
et al., 2021) and traditional liquefaction severity indi generalized as shown in Eqn. 2:
cators often evaluated deterministically is presented
by developing a set of fragility functions.
2 BAYESIAN BELIEF NETWORKS
A Bayesian Network falls in the category of probabil 2.1 Bayesian Networks for liquefaction hazard
istic graphical modeling (PGM) technique that is to In general, the assessment of liquefaction hazard
compute uncertainties by using the probability con moves from subsequent steps, which translate the
cept (Pearl, 1988). It is represented as a directed acyc soil propensity to liquefy, the triggering analysis, and
lic graph (DAG) which contains a set of nodes and the liquefaction-induced ground deformation (Bird
links, arches, relating parent nodes to the children et al., 2005). Therefore, a BN model for liquefaction
ones. A directed acyclic graph evaluates the uncer hazard requires introducing three types of nodes: 1)
tainty of an event occurring based on the conditional input nodes that include soil parameters (e.g., relative
probability distribution of each random variable. density, plasticity, fine contents), site conditions
A conditional probability table is used to represent (groundwater depth, thickness and position homoge
this distribution of each variable in the network. neous layers, distance from geological features like
To understand the meaning, it is necessary to rivers/paleochannels) and seismic intensity measures
introduce the inference algorithm that is based on (peak ground acceleration, Magnitude, duration, epi
the Bayes theorem and conditional independence as central distance); 2) state nodes which combine input
follows: into intermediate variables (i.e., the classification of
soil susceptibility, the probability of triggering the
phenomenon), and 3) output nodes describing the
severity of liquefaction-induced ground observations.
Hu et al. (2016) provided an example of a net that
constructed a model for liquefaction potential evalu
It introduces joint probability, a measure of two ation considering 12 factors. In this study, a revised
events happening simultaneously i.e., P(A|B), and the approach is proposed, to link geotechnical and geo
conditional probability of an event B, which is the logical susceptibility to free field liquefaction ground
probability that event B occurs given that an event evidence via the Arias Intensity, without evaluating
A has already occurred. The Bayesian Networks satisfy traditional liquefaction severity indicators, i.e., simpli
the Local Markov Property, stating that a node is con fying an intermediate step that unavoidably introduces
ditionally independent of its non-descendants, given its noises.
parents. In the example of Figure 1, P(D|A, B) is equal After selecting representative variables for lique
to P(D|A) because D is independent of its non faction, several net models are automatically gener
descendent, B. This property aids us in simplifying the ated and tested. With this regard, updating the net
Joint Distribution. The Local Markov Property leads us calculates the probability of having a particular com
to the concept of a Markov Random Field, which is bination of input variables given the evidence, allow
a random field around a variable that is said to follow ing to determine the size, position, and strength
Markov properties. characterization of the most likely layer experiencing
liquefaction during the considered Earthquake
sequence. On this layer, more detailed studies are car
ried out. The development of a probability model for
liquefaction occurrence considers different crust
thickness, Hc, thick, Hl, and average resistance, mean
crr, of the potentially liquefiable layer. The maximum
likelihood criterion is applied to fit the histograms
data to lognormal functions (Baker, 2013).
2.2 Management of uncertainty and Validation

Among the advantages of the proposed method there,
is an immediate and reliable graphic tool that displays
input variables and their relationships. Uncertainties
in data estimates and the validity of basic assumptions
(like the three-layer hypothesis) can be managed at
Figure 1. Scheme of a sample Bayesian Network. different levels. In addition, the probabilistic model
1036
allows practitioners to make predictions for future Christchurch area, around 9ʹ000 CPT profiles with
situations with a certain level of confidence. Valid a depth greater than 10m have been considered. In
ation criteria are required to assess the reliability of a preliminary phase, the Equivalent Soil Profile (ESP)
the obtained results. The metrics used in this study method defined by Millen et al. (2020) is applied to
are borrowed from the binary validation methods determine the liquefaction susceptibility. This criterion
Lusted (1971), which introduces the concept of converts a CPT profile into a three-layered equivalent
Receiver Operative Characteristic Curves obtained by one, defined by Hc, Hl and crr of the potentially lique
combining specificity and sensitivity; the Area Under fiable layer, with an error term used to confirm the
the Curve (AUC) is a global proxy of the estimate consistency with the hypothesis of three-layered pro
quality. In addition, the Kappa statistics method is file. The following analysis include three variables
used to assess the agreement between the actual and (i.e., crust thickness, liquefiable thickness, and resist
expected results (Witten and Frank, 2005). Kappa ance), which define 22 homogeneous soil classes for
statistics is defined in Equation 3: Pa is the fraction of liquefaction susceptibility taken from Millen et al.
agreement and Pe is the fraction of random agreement (2020) criterion. The groundwater table is not expli
used to correct for values. Pa is a summation of the citly accounted since it is already considered in evalu
diagonal values of the confusion matrix. ating the crust thickness. In addition, the ESP normed
error representing a check factor for the consistency
with the 3-layered profile hypothesis is considered. On
the other hand, the distance from riverbeds is assumed
as a proxy for geological susceptibility. Seismic hazard
is characterized for three earthquakes among the main
events of the 2010-2011 C.E.S., i.e., the Sept 2010
Mw7.1 Darfield earthquake, the Mw6.2 Feb 2011, and
3 THE CASE STUDY OF CHRISTCHURCH the Mw6.0 Jun 2011 Christchurch Earthquakes. For
(NEW ZEALAND) each of them, the Arias Intensity is evaluated elaborat
ing the records of Central Christchurch strong motion
3.1 The 2010-2011 Canterbury Earthquake stations. These scenarios are modeled through the fault
Sequence distance and the Arias Intensity selected in place of the
pair PGA-magnitude since it embeds the intensity and
The city of Christchurch (≈370ʹ000 inhabitants in duration of the shaking. On the other hand, the lique
2011), in the Canterbury Region of the South Island faction ground observation is classified as “YES” and
of New Zealand, was repeatedly struck by earth “NO”. The subsequent diagnostic inference showing
quakes during the 2010-2011 seismic sequence the correlation between each variable and the observed
known as Christchurch Earthquake Sequence (C.E.S) liquefaction is reported in Figure 2. Just as an example
The most noticeable were: the Mw 7.1 Darfield event of traditional severity indices, the van Ballegooy et al.
of September 4th 2010, the (Mw 6.2) Christchurch (2014) Liquefaction Severity Number (LSN) which
Earthquake of February 22nd 2011, resulting in 185 combines triggering with a hyperbolic weight function,
fatalities and diffuse devastation to dwellings and is presented in Figure 2.
infrastructures and the Mw 6.0 June 13th 2011. Lique
faction played a significant role in causing the
removal of 900ʹ000 tons of liquefied soil, the demoli
tion of 8ʹ000 buildings (Cubrinovski, 2013; Tonkin &
Taylor, 2013), and an economic loss of NZ$30 billion
only on the residential sector (NZ Parliament).
3.2 Creation of databases and variables

management
Taking advantage of the significant amount of geo
technical data, a general framework to develop BBN
is developed. The strategy to reach such goals
include: i) database construction and preliminary
processing of CPT data; ii) sensitivity and correl
ation analyses; iii) automated training of different
nets and validation test. In addition, a probabilistic
Figure 2. Sensitivity analysis of selected variables versus
model is derived from the output of the structure liquefaction ground observations: a) non-liquefiable crust
showing the best performance. thickness (m); b) thick of potentially liquefiable layer (m);
The database creation includes the collection and c) mean crr; d) normed error (Millen er al., 2020); e) dis
homogenization into a standardized format of many tance of investigated profile from riverbeds (m); e) epicen
CPTs from the New Zealand Geotechnical Database. tral distance (km); g) Arias (1970) Intensity (cm/s); h) LSN
Out of 30ʹ000 stratigraphies available on the whole (van Ballegooy et al., 2014).
1037
Although each variable of Millen et al. (2020) 3.3 Training and validation of Bayesian Belief
method seems to correlate to liquefaction poorly, if Networks for liquefaction
considered alone, from Figure 2, it is intuitive to
To generate a BBN for liquefaction assessment, the
observe that liquefaction occurrence increases with
local score metrics are considered for structure
proportionally with the seismic shaking (see Arias
learning. Following the typical steps in assessing
Intensity) and close to riverbeds. The herein ana
the liquefaction-induced permanent ground deform
lyzed seismic scenarios remark that the relationship
ation, seven structures of Bayesian Belief Networks
between seismic moment magnitude Mw and
are trained and tested comparing the performance
observed liquefaction is here strongly affected by the
versus the liquefaction prediction capability for the
epicentral distance that made the Mw6.0 (located at
Sept 2010 Mw7.1 Darfield earthquake, the Mw6.2
approximately 5km from the City Center) and
Feb 2011, and the Mw6.0 Jun 2011 Christchurch
Mw6.2 February event, whose epicenter was around
Earthquakes. Bearing in mind the results displayed
6-7km South of Christchurch, more severe than the
in Figure 3, these networks architecture is built in
Mw7.1 Darfield event (epicenter 45km West of
the “Genie Academy” environment (Genie, 2020
Christchurch).
https://www.bayesfusion.com/) by introducing dif
In the following analysis, the existing depend
ferent search algorithms. PC (Spirtes et al., 1993)
encies among input variables are evaluated by cal
uses independences observed in data (established
culating the Pearson coefficient; the resulting
employing classical independence tests) to infer the
correlation matrix is plotted in Figure 3, whose
structure that has generated them and is the most
coefficient font size is proportional to the correl
adequate for continuous datasets; Naive Bayes and
ation found. The massive amount of raw data, the
its improved version, i.e., Tree Augmented Naïve
lack of a predefined standard in the format, in con
Bayes (TAN) and Augmented Naïve Bayes (ANB),
junction with a certain level of subjectivity con
Bayesian search (BS), and Greedy thick thinning
nected to post-earthquake rapid surveys, results in
(GTT), respectively defined by Cooper and Herko
a non-negligible noise affecting the whole dataset.
vits (1992), and Cheng et al. (1997). Additionally,
Therefore, Figure 3 shows a partial moderate/
a background knowledge is provided to the PC
strong correlation between parameters introduced
algorithm to build an expert-guided network based
to quantify liquefaction susceptibility, which can
on engineering judgment. The expert-guided model
be merged into one variable (i.e., the ESP soil
is displayed in Figure 4. The net is based on the PC
class) and between Arias Intensity and liquefaction
algorithm customized on the phenomenon know
severity indicators like LSN. However, to reduce
ledge. In particular, the combination of Hc, Hl, and
the intermediate steps resulting in error, in the fol
crr provides the equivalent soil profile class by
lowing process Arias Intensity is directly assumed
introducing the normed error as the first control
as Engineering Demand Parameter for a given soil
factor. The geological susceptibility is separately
configuration, providing an alternative approach to
considered; therefore, the distance from the riv
traditional liquefaction severity indicators. In add
erbed is considered and modeled as an independent
ition, the preliminary classification of variables
variable. On the right side of the net, the epicentral
applied by Millen et al. (2020) has been increased
distance and Arias Intensity can be observed; the
to investigate other geometric configurations
net learned both a dependency among each other
better.
and liquefaction ground evidence. Conditional and
prior probabilities are specified and, in turn,
employed to perform belief updating and extract
posterior beliefs. The quality measure can be
judged with several criteria like the Bayesian
approach or minimum description length (Bouck
aert et al., 2011). The score of the whole network
can be decomposed as the sum (or product) of indi
vidual node scores in a way that enables local scor
ing and searching methods. The performance of
each algorithm is summarized in Table 1 that shows
the result of the 5-fold cross-validation test, which
splits the whole dataset into 5 panels and estimates
how the model is expected to perform when used to
make predictions on data not used during the train
ing of the model. The AUC, in conjunction with the
K statistics, is used to rank the best learning struc
ture. Although a perfect match exists when
K statistics is equal to 1.0, realistically, for a site
Figure 3. Correlation matrix of the considered variables investigation on a regional scale, a K=0.5 match is
for liquefaction. more appropriate (Demshar, 2020).
1038
Table 1. Summary of validation results for the above
curves have been derived on these configurations.
defined Bayesian networks. A fragility curve like the one shown in Equation 4
evaluates the probability of observing liquefaction
SCORE given an intensity measure (IM) idealized by a typical
lognormal distribution:
BBN AUC K-stat. OSR PRECISION
PC 0.81 0.442 0.72 0.70

Naïve Bayes 0.79 0.431 0.72 0.72
ANB 0.77 0.458 0.73 0.71
TAN 0.82 0.447 0.72 0.71
BS 0.81 0.421 0.72 0.56
GTT 0.81 0.460 0.73 0.69 where denotes the Gaussian cumulative distri
Expert-guided* 0.81 0.461 0.73 0.70 bution function, IMm is the median distribution and β
is the logarithmic standard deviation. Even though
* Ranked as the best structure because of the highest that minor literature exists about the modeling of
K-statistics and the physical accounting of cause-effect soil liquefaction vulnerability with this approach
relationships governing the liquefaction phenomenon. (Geyin and Maurer, 2020), fragility curves are gener
ally adopted in procedures to assess seismic and
liquefaction risk on buildings (Fotopoulou et al.,
Once the general performance is evaluated 2018), road and embankments (Syner-G, 2013) and
through the AUC, the features of the Christchurch pipelines (Liu et al., 2015; Baris et al., 2020). Fragil
critical layer have been searched by setting the evi ity functions are developed as an extension of deter
dence of liquefaction and discarding those profiles ministic models, allowing practitioners to make
not ascribable to the three-layered model (Paolella previsions linked to probabilistic seismic hazard ana
et al., 2020). It is found that, for the considered seis lyses. After the experience of Baker (2013) in struc
mic scenarios, the critical layer for liquefaction tural modelling, the maximum likelihood method is
is shallow (Hc < 3m in 60% of cases), mid-size to here applied to reach the appropriate data fitting.
large (Hl > 3m in 81% of cases) and can be modeled Assuming that the number of liquefaction/no lique
with a crr < 0.30 (93%of cases) (Figure 4). faction observations from each experiment is inde
pendent of observations from other experiments, the
probability of observing zj liquefaction occurrence in
nj motions having IM = xj is provided by the bino
mial distribution (Equation 5).
� �
P zj evidence in nj experiments
where pj is the probability that a ground motion

with IM = xj will trigger liquefaction. The maximum
likelihood is thus implemented to find the fragility
function capable of predicting pj with the highest
Figure 4. Back analysis of Christchurch earthquake scen
probability of fitting experimental data. When ana
arios with the selected expert-guided net. lysis data are obtained at multiple IM levels, we take
the product of the binomial probabilities (from Equa
tion 5) at each IM level to obtain the likelihood for
the entire data set.
3.4 A probabilistic model for liquefaction
triggering analysis
The automated processing of such a number of CPT
data and the availability of post-event damage surveys
provides a unique chance to test the performance of
current criteria and develop new models. About the where m is the number of IM levels and Π is
latter task, the probability of observing liquefaction a product over all levels.
given the critical layer has been better investigated by The pairs of IMm and β obtained for each soil con
splitting the liquefiable thickness range into three clas figuration by maximizing Equation 6 are reported in
ses, i.e. 3 < Hl ≤ 5m, 5 < Hl ≤ 10m, 10 < Hl ≤ 15m Table 2; in addition, the maximum observed value of
1039
arias intensity is indicated, meaning that discretion the randomness of seismic source, spatial variability,
should be used in using the obtained fragility functions and error propagation (Paolella et al., 2020). The
out from the suggested range. Figure 5 displays both Bayesian Network-based back analysis located the
the suit of functions showed in Table 2 and the inter Christchurch critical sandy layer at a relatively small
polated experimental points; a graphical comparison depth (Hc < 3m), and a very low relative density is
among each other is presented in Figure 5d. It reflects found on it (crr <0.3). The following analysis defined
the positive correlation between thick of liquefiable a set of soil fragility curves that couple the liquefac
layer Hl and liquefaction occurrence for a given tion susceptibility of such a critical layer to the seis
shaking. micity of the area via the Arias Intensity measure.
Their applicability should respect the Arias Intensity
range reported in Table 2, requiring additional ana
lyses and judgment if this value is exceeded.
REFERENCES
Arias A., 1970: “A measure of earthquake intensity”. Seis
mic design for nuclear power plants, R. J. Hansen, ed.,
MIT Press, Cambridge, Mass.
Baker J. W., 2013: “Efficient analytical fragility function
fitting using dynamic structural analysis”, Earthquake
Spectra.
Baris A., Spacagna R. L., Paolella L., Koseki J. &
Modoni G., 2020: “Liquefaction fragility of sewer pipes
derived from the case study of Urayasu (Japan)”. Bullet-
tin of Earthquake Engineering, Springer B.V. 2020.
Bird, J., Crowley, H., Pinho, R., Bommer, J.; 2005:
“Assessment of building response to liquefaction
induced differential ground deformation”. Bulletin of
the New Zealand Society for Earthquake Engineering,
38-4, Dec. 2005, 215–234.
Bouckaert R.R., 2011: “WEKA Manual for version 3-7-5”.
Figure 5. Probability of observing liquefaction manifest Cheng, J., David A. B. & Weiru L., 1997: “An Algorithm
ations given the AI, for each profile configuration: a) for Bayesian Belief Network Construction from Data”.
3 < Hl ≤ 5m; b) 5 < Hl ≤ 10m; c) 10 < Hl ≤ 15m; d) general Proceedings of AI & Statistics, pages 83-90.
comparison. Cornell, C.A., and Krawinkler, H. 2000. “Progress and
Challenges in Seismic Performance Assessment”. PEER
Center News, 3, 1–3.
Table 2. Summary of fragility function parameters. Ching J., Phoon K.K., 2017: “Characterizing uncertain
site-specific trend function by sparse Bayesian
H1(m) IMm β Arias Intensity (cm/s) range learning”. ASCE Journal of Engineering Mechanics,
143(7), 04017028.
3-5 142.3 1.03 < 350 Cooper G. F., and Herskovits E. A., 1992: “A Bayesian
method for the induction of probabilistic networks from
5-10 123.8 1.02 < 350
data”. Machine Learning, 9:309–347.
10-15 102.6 1.06 < 350 Cubrinovski M., 2013: “Liquefaction-Induced Damage in
the 2010-2011 Christchurch (New Zealand)
Earthquakes”. In: Proceedings of the 7th International
Conference on Case Histories in Geotechnical Engineer
4 CONCLUSIONS ing, 29 Apr–4 May, Chicago, Illinois.
Demshar J., 2020: “Development of Liquefaction Hazard
After processing ≈30ʹ000 CPTs, seven structures of Map Using a Geostatistical Method”. University of Min
Bayesian Networks are tested against the liquefaction nesota, ProQuest LLC.
ground evidence induced by three main shocks of the Fotopoulou S., Karafagka S., Pitilakis K., 2018, Vulnerabil
2010 – 2011 Christchurch Earthquake Sequence. The ity assessment of low-code reinforced concrete frame
validation criteria ranked the “expert-guided” Net as buildings subjected to liquefaction-induced differential
the best architecture. It accounts for the cause-effect displacements, Soil Dynamics and Earthquake Engin
relationships ruling the phenomenon introducing new eering 110 (2018) 173–184.
GeNIe, (2020). BayesFusion, LLC. Retrieved from http://
variables that better describe the geological/geotech www.bayesfusion.com/.
nical susceptibility (e.g., epicentral and riverbeds dis Geyn M., and Maurer B. W., 2020: “Fragility functions for
tances) not considered by traditional liquefaction liquefaction-induced ground failure”. Journal of Geo-
severity indicators. The robustness of the proposed technical and Geoenvironmental Engineering, © ASCE,
method herein is quantified by an AUC equal to ISSN 1090-0241.
0.81, i.e., ≈15% higher than traditional indices, which Hu, J., Tang, X., and Qiu, J.: Assessment of Seismic lique
cannot manage all the uncertainties connected with faction potential based on Bayesian network constructed
1040
from domain knowledge and history data, Soil Dyn. Spirtes P., Glymour c., and R. Scheines. 1993: “Causation,
Earthq. Eng., 89, 49–60, 2016. Prediction and Search”. Springer Verlag, Berlin.
Liu M., Giovinazzi S. and Lee P., 2015: “Seismic fragility SYNER-G, 2013: “Systemic Seismic Vulnerability and
functions for sewerage pipelines”. (ASCE), Pipelines Risk Analysis for Buildings, Lifeline Net-works and
Conference 2015, 23-26 August, Baltimore (MD), USA. Infrastructures Safety Gain”. ISBN: 978-92-79-33135-0.
Lusted L.B., 1971, Signal detectability and medical deci DOI: 10.2788/23242. Web-site: http://www.vce.at/
sion making. Science 171: 1217–1219. SYNER-G/files/project/proj-overview.html
Millen M., Viana da Fonseca A., Quintero J., Ferreira C., Tang X., Bai X., Hu J. H., Qiu J., 2018: “Assessment of
Oztoprak S., Bozbey I., Oser C., Aysal N., Kosic M., liquefaction-induced hazards using Bayesian networks
Logar J., 2020: “Equivalent soil profiles to integrate based on standard penetration test data”. Natural Haz
in situ tests results and soil-structure interaction in lique ards and Earth System Sciences 18(5):1451–1468.
fiable soils. The Adapazari case-study”. Bull Earthq Eng Tesfamariam S., 2013: “Seismic risk analysis using Bayes
(Special issue). ian belief networks”. In Handbook of seismic risk ana
Paolella L., Salvatore E., Spacagna R. L., Modoni G., lysis and management of civil infrastructure systems,
Ochmanski M., 2019: “Prediction of Liquefaction S. Tesfamariam and K. Goda editors, Woodhead Pub
Damage with Artificial Neural Networks”, Atti 7ICEGE lishing Limited, pp.175–208.
2019. Tonkin & Taylor, Ltd. (2013): “Liquefaction Vulnerability
Paolella L., Spacagna R. L., Chiaro G., Modoni G., 2020: Study”. Report to Earthquake Commission. Tand
“A simplified vulnerability model for the extensive T ref. 52020.0200/v1.0, prepared by S. van Ballegooy
liquefaction risk assessment of buildings”. Bullettin of and P. Malan, available at https://canterburygeotechnical
Earthquake Engineering, 2020. database.projectorbit.com.
Pearl J., 1988: “Probabilistic Reasoning in Intelligent Sys van Ballegooy S., Malan P., Lacrosse V., Jacka M.E.,
tems: Networks of Plausible Inference”. San Francisco, Cubrinovski M., Bray J.D., O’Rourke T.D., Crawford S.
California: Morgan Kaufmann Publishers, Inc. A., Cowan H., 2014: “Assessment of
Spacagna R.L., Porchia A., Fabozzi S., Cesarano M., Liquefaction-Induced Land Damage for Residential
Peronace E., Romagnoli G., 2021: Seismic liquefac Christchurch”. Earthquake Spectra (30) No. 1: pages 31–
tion assessment in Calabria region in Southern 55, February 2014.
Italy: a geostatistical approach at regional and sub Witten I.H., Frank E., 2005: “Data mining: Practical
regional scale. International Journal of Geosciences Machine Learning Tools and Techniques. Burlington,
(accepted). MA: Elsevier pp. 143–185.
1041
© 2022 the Author(s), ISBN 978-1-032-31259-0
Geotechnical zoning of deltaic and alluvial soils of Guayaquil (Ecuador)

using CPT and Nkt calibration based on FVT
Subterra, Ecuador
R. Luque
Geosísmica, Ecuador
ABSTRACT: This document summarizes the results obtained from geotechnical zoning analysis in the city
of Guayaquil and its surroundings, which stratigraphy consists mostly of alluvial and estuarine deltaic soils,
by compiling over 600 CPTs and 800 boreholes. Products generated include a map of rigid stratum depth with
high impedance contrast as well as spatial distribution of liquefaction susceptibility using a simplified meth
odology. Furthermore, shear strength su peak and remolded have been evaluated by comparing CPT-borehole
pairs and 26 sites with CPT and electric FVT. It has been found that su peak from laboratory tests adjusts well
to the FVT su remolded and fs, suggesting sample disturbance. It has been verified that Nkt values measured
with FVT are lower than average reported values in previews studies, possibly due to diatoms identified in
scanning electron microscope (SEM). Fr based estimates appear to yield an upper limit of Nkt factor in Guaya
quil soft clays, while su maps for 5 m soil segments are provided from CPT-based Nkt values.
1 INTRODUCTION estimate liquefaction potential of granular soils. While

considering testing procedures, CPT estimates provide
Guayaquil is located on the Ecuadorian coast, and a more accurate and efficient alternative. However, it
has a flat relief in most of its extension. Geologic is imperative to calibrate CPT geotechnical parameters
ally, the soils correspond to the Holocene with in unconventional soils, with measurements obtained
a large alluvial plain and estuarine deltaic deposit, from other in situ and laboratory tests (Mayne, 2007b;
located at the foot of the Chongón-Colonche Moun Robertson, 2009; Robertson, 2012). Therefore, this
tain Range. Due to the geological characteristics of report analyzes: 1) back-calculated Nkt values from
Guayaquil, a large percentage of sites include soils FVT measurements and 2) a simplified geotechnical
susceptible to liquefaction or profiles with a thick zoning of Guayaquil considering highly compressible
upper layer of soft clay and organic material. More top stratum, shear strength su, and susceptibility to
over, the tectonic environment rich in dissolved liquefaction (LSN, LPI and settlements).
silica, favors the existence of diatoms, elements that
provide a characteristic geotechnical behavior.
Geotechnical investigation in Guayaquil has been 2 IN-SITU INVESTIGATION
predominantly conventional drilling and sampling,
from the 1950s to the present, but in the last decade, The database considered for developing a com-
emphasis has been placed on in situ testing, especially pressible-top-layer zoning map consists of 1432
through CPTu soundings. In this document, the experi soundings: 615 CPTs (95% down to refusal condi
ence from 615 CPTs in Guayaquil and the surrounding tions), and 817 boreholes from several companies.
area is synthesized along with 143 electric field vane In addition, 26 FVT profiles were evaluated,
tests (FVT), carried out in an effective area of delivering 143 su values (Figure 1). SEM images
450 km2. In situ results have been integrated with were taken at 14 sites to corroborate the existence
existing boreholes. of amorphous microstructures.
The common state of practice in Guayaquil is to For CPTu tests, a standard 10 cm2 cone, compres
perform unconfined compression tests or use handheld sion type, with 150 cm2 sleeve was used. The pore
devices (torvane, penetrometer) to determine pressure filter (u2 position) is made out of bronze,
undrained shear strength su in clays, and measure N60 and saturated with silicone oil. The piezocone is
(standard penetration number, corrected) in order to pushed at 2 cm/s with a 15-ton Pagani TG-63
DOI: 10.1201/9781003308829-157
1042
equipment recording measurements every cm. FVTs cases, also near boreholes. Vane measurements were
have been carried out by pushing the vane housing taken between 1 and 17 m (avg. 5.6 m) while
with the same penetrometer and further advancing recorded peak values vary between 10 and 240 kPa
50x100 mm vanes, 25 cm from penetrated depth. (avg. 64 kPa). Different CPT parameters at FVT
Torque is applied from the surface at one-meter inter depth location were compared for trend evaluation:
vals with Geomil GVT-100 tester. The data acquisi corrected cone resistance (qt), normalized cone
tion system records the resistance as a function of resistance (Qtn), normalized friction ratio (Fr), nor
angular rotation of the vane from initial position until malized pore pressure ratio (Bq), normalized SBTn
peak and remolded resistance is measured. Index (Ic). Tip values of qt range between 0.1 and
4.4 MPa, those of Ic between 2.1 and 3.8, Bq
between -0.3 and 1.0, and Fr between 1% and 11%.
The bearing factor Nkt for cone tip resistance was
back-calculated first, using a single qt value at vane
test depth and secondly, with an average qt value
±25 cm from FVT depth, to consider lenses variabil
ity across the distance between both tests. The lower
the standard deviation of qt (σqt), the closer both Nkt
values are. Hence, by filtering measurements with
σqt 40:2 MPa, 95% of Nkt values range between 3
and 12 (Nkt-avg = 7.4).
There are limitations regarding the procedure and
interpretation of the vane failure mechanism, as
explained by Bjerrum (1972), Ladd & Foott (1977),
among others. Even when considering a correction
factor μ ¼ 0:80, according to Bjerrum, for an plasti
city index (PI) of 50% (conservative in most data
points analyzed), estimated Nkt values are in the
range of 4 to 15, with an average of 9.3. Thus, rela
tively low values of Nkt should be expected in the
upper compressible stratum of Guayaquil.
While no correlation was detected with Fr, inversely
proportional trends have been identified with Ic and Bq
and directly proportional with Qtn and OCR. However,
none of the correlations provides an acceptable adjust
ment coefficient. Figure 2 illustrates the correlation
between Nkt and Qtn, which could be considered the
Figure 1. Location of soundings. best trend found among normalized parameters.
A preliminary classification of each sounding was

necessary to determine profile types and potential
geotechnical problems. It was found that 38% of
profiles have a top soft clay stratum with variable
thickness from 3 to 30 m, 24% presented susceptibil
ity of liquefaction in the first 20 m and 11% of pro
files include both conditions. Also, 10% of profiles
presented organic material content or more than
30 m of soft clay while 10% of soils consisted of
high impedance contrast within the top 25 m. This
totals 93% of the profiles in which an exhaustive
geotechnical analysis is required on most medium
risk projects. In 83% of soundings where a thick soft
clay layer or loose sand is present, is convenient to
use CPTu for a better identification of sandy intercal
ations and suitable characterization of soft layers,
both in terms of strength and compressibility. Figure 2. Back-calculated Nkt values as a function of Qtn.
3 Nkt MEASUREMENT Furthermore, the Nkt range has been analyzed

graphically among FVTsu and net cone resistance
Piezocone, along with vane testing, was performed ðqnet ¼ qt -σv Þ considering 136 measurements with
at 26 different sites, less than 1.5 m apart and in 7 σqt 51 MPa. Three sub-datasets corresponding to
1043
different geotechnical zones (106 values) have been tests for preliminary and somewhat conservative esti
compared: alluvial, deltaic-estuarine of Guayaquil mates in Guayaquil alluvial and deltaic soils. In add
and Durán. The qnet-su relation is shown in Figure 3 ition, an upper Nkt limit of 12 is proposed for the NC
with slopes between 8 and 10, which corresponds to clay deposit, while site specific CPT-FVT pairing
plausible Guayaquil Nkt values. along with laboratory tests would be necessary for
evaluating site specific behavior. Therefore, an
Figure 4. Comparison between su values from CPT, FVT,

unconfined compression and torvane tests.
Figure 3. Correlation between qnet and su from field vane

tests: (a) 136 measurements with σqt 51 MPa; (b) three
sub-datasets for geotechnical zones.
Based on evaluated trends the following can be

concluded: 1) Nkt values are low compared to aver
age values of conventional soils and 2) a correlation
for Nkt with normalized parameters such as Bq or Fr
(Mayne & Peuchen, 2018; Robertson, 2012), seem
unfit for the available data from Guayaquil.
Although laboratory tests are commonly used to
determine su in Guayaquil, sample quality is poor in
many cases. According to Lunne et al. (1997) criterion,
based on void ratio initial (eo) and void ratio variation
ðDe Þ, 50 consolidation tests were analyzed resulting in
90% of samples having De =e0 40:04 and 60% > 0.07,
so the quality of sampling is generally low or medium.
In addition, when comparing CPTs and boreholes from
40 sites, laboratory su peak tends to adjust to CPT fs,
and similarly to in-situ remolded su at sites where FVT
was also available. As an example, Figure 4 illustrates
this matter in two profiles, suggesting resistance loss
due to samples disturbance.
Hence, Robertson Nkt-Fr correlation has proven to Figure 5. Spatial distribution of estimated su in Guayaquil
be more appropriate than readily available laboratory from 0 to 5 m and 5 to 10 m soil segments.
1044
Figure 6. SEM images from samples obtained in alluvial and estuarine deltaic deposits of Guayaquil (South East-D1,
Durán and Kennedy-D3).
interpolation has been performed in ArcGIS for aver worth noting that fs values lower than 5 kPa have
age values of estimated su in 5 m soil segments, except been reported in several CPTs, demonstrating the
for segments which are predominantly granular soils presence of very soft and possibly sensitive soils.
(>55% SBTn 5-8). Figure 5 shows maps generated for Further research with additional testing is required to
0-5 m and 5-10 m segments. establish a more specific range of St by geotechnical
zones.
Previous studies indicate that St is inversely pro
4 PRESENCE OF MICROFOSSILS portional to Nkt (Robertson, 2012; Mayne & Peu
chen, 2018), which would support the low range
Diatom microfossils have been found in several values of Nkt obtained in Guayaquil. Nevertheless,
marine and lacustrine deposits, such as those in no relationship between measured sensitivity and Nkt
Japan, Mexico and Colombia. The presence of dia has been observed, possibly due to the difficulty of
toms in two estuarine locations in Guayaquil has determining the exact remolded value in highly sen
been widely discussed (Vera, 2014). sitive soft soils or due to the need of precisely deter
Diatoms are eukaryotic unicellular algae, with mining diatom content, which has not been the case.
a frustule composed of SiO2 that fossilizes after the
death of the microorganism. They are found in marine
environments with dissolved silica, very common in 5 GEOTECHNICAL ZONING OF
places with high tectonic activity (Díaz-Rodríguez, COMPRESSIBLE STRATUM AND SOIL
2011). Diatoms modify the static and dynamic LIQUEFACTION
response of soils. A higher diatom content increases
porosity, compressibility and dilatation (Díaz- Other parameters evaluated from the database were: 1)
Rodríguez et al., 1998; Shiwakoti & Tanaka, 2002). thickness of top compressible layer or depth of rigid
Besides, su increases proportionally with diatom con stratum, 2) liquefaction potential, 3) SBTn Index Ic, 4)
tent (Wiemer & Kopf, 2017). Diatoms may cause relative density Dr of granular materials, 5) OCR
a deviation from existing Nkt trends per normalized and 6) coefficient of consolidation cv (Paredes, 2020).
CPT parameters and also provide an explanation for This report illustrates zoning maps with regards to the
encountered relatively low Nkt values. compressible layer and soil liquefaction potential.
Abundant presence of diatoms and framboidal A preliminary qualitative and quantitative evalu
pyrites has been found in sites with low Nkt. In 13 ation of the rigid stratum depth was performed to
out of 14 sites, SEM images identified a wide range determine the first and foremost impedance contrast
of diatoms from different species. Figure 6 shows with the upper clay deposit. Supported by dozens of
the microstructure identified in three different sec pile driving analyses, this stratum would correspond
tors. As diatoms exist in soil microstructure, and to a pile bearing capacity of 2500 kPa for a 50 cm
consequently higher su-peak values are measured, it is square section. For defining such stratum, Qtn must be
reasonable to obtain higher sensitivity values (St = greater than 100 (N60 greater than 45 blows in sands)
su-peak/su-rem) as well, compared to a diatom-free soil and thickness greater than 2 m. If the identified layer
matrix. Although measured St values from FVT was less than 2 m thick and located above a second
range between 1.1 and 12.8, with an average of 2.5, soft clay layer, then the following layer with Qtn and
several tests show 5 to 10 kPa oscillation around thickness greater than 100 and 2 m, respectively, was
remolded values. So, reported St average could be considered. The thickness variation of the upper com
higher, perhaps close to 4 considering lowest meas pressible stratum has been calculated by means of
ured values. At such low values it is difficult to a global interpolation process with local adjustment
accurately determine soil sensitivity. Besides, it is and iterative finite differences (topo to raster), which
1045
allows generating smooth curves and considering
boundaries. Figure 7 illustrates the interpolation map,
with depth contours between 0 and 50 m.
Figure 8. Spatial distribution of LPI and LSN.
Figure 7. Rigid stratum depth map from 1432 soundings

(615 CPTs). Rock outcrop (Benítez & Vera, 2006).
This map provides a rapid identification of depths

in which pile installation may complicate if appro
priate measures are not considered. Also, it consti
tutes a preliminary drainage or compressibility
boundary with regards to settlements analyses.
Although, the presence of soft to firm clays greater
than 20-meter thick has been confirmed in most of
the estuarine area, certain sectors south-west of the
city are noteworthy, where dense soils have been
identified at shallower depths as well as sandbars.
Due to the abundance of soft soils in Guayaquil,
this map results beneficial for the construction indus
try and real state. Based on the map, an exploration
campaign could be targeted towards potential geo
technical issues and project risks.
In an effort to develop liquefaction potential maps,
the factor of safety FS was calculated for each sound
ing in Guayaquil database, considering a seismic
demand of Mw=7.5 and PGA=0.4 g (according to the
PSHA disaggregation carried out by Beauval et al.,
2018). Subsequently, LPI (Iwasaki & Tokida, 1981) Figure 9. Spatial distribution of free field post-liquefaction
and LSN (Tonkin & Taylor Ltd, 2013) indices were settlements down to 20 m and complete profile.
calculated as well as post-liquefaction vertical free
field deformations for the first 20 m (Sett @ 20m), and Results indicate that 20% of CPTs have LPI > 10,
for the entire profile (Zhang, Robertson, & Brachman, 8% LSN > 20, 20% Sett@20m > 15 cm and 35%
2002). Figures 8 and 9 illustrates the interpolated Sett in the entire profile > 15 cm. The low percent
maps. age of moderate LSN is due to the strong influence
1046
of upper layers with high fines content that, in most Beauval, C., Mariniere, J., Yépez, H., Audin, L.,
cases mitigate damage at ground surface. Zones Nocquet, J.-M, Alvarado, A., Baize, S., Aguilar, J.,
where estimated post-liquefaction deformations in Singaucho, J., & Jomard, H. (2018). A New Seismic
the entire profile is greater than 15 cm, even though Hazard Model for Ecuador. Bulletin of the Seismological
Sett@20m is lower, are especially relevant for pile Society of America 108 (3). doi:10.1785/0120170259
design projects to ensure the neutral plane is below Bjerrum, L. (1972). Embankments on Soft Ground. Proc.
liquefiable layers. ASCE Espc. Conf. on Performance of Earth and Earth-
Supported Structures.
According to LSN and LPI, sectors with highest Díaz-Rodríguez, J., Lozano-Santa Cruz, R., Dávila-
risk and damage associated with liquefaction are Alcocer, V., Vallejo, E. & Girón, P. (1998). Physical,
those located north, in the alluvial zone. However, chemical, and mineralogical properties of Mexico City
low LSN values south-east of Isla Mocolí and at the sediments: a geotechnical perspective. Canadian Geo-
first five kilometers of Ave. Puntilla-Samborondón technical Journal 35 (4): 600–610.
are notable, since before mapping a high liquefac Díaz-Rodríguez, A. (2011). Diatomaceous soils: monotonic
tion potential would have been expected due to its behavior. International Symposium on Deformation
geotechnical location in alluvial zones. On the con Characteristics of Geomaterials. Seúl: Civil Engineer
ing Department, National University of Mexico. DOI:
trary, for estuarine deltaic areas where predomin
10.13140/2.1.3322.5606
antly thick clayey layers are expected, some soil Iwasaki, T., Tokida, K., & Tatsuoka, F. (1981). Soil Lique
profiles are prone to liquefaction due to the pres faction Potential Evaluation with Use of the Simplifified
ence of sandy layers in the top 20 m. Procedure. International Conferences on Recent
Advances in Geotechnical Earthquake Engineering and
Soil Dynamics. 12.
6 CONCLUSIONS Ladd, C., & Foott, R. (1974). New design procedure for
stability of soft clays. J. of the Geotech. Eng. Div., 100
A database of 615 CPTs, 26 FVT profiles and 817 (GT7), 763–786.
boreholes have been used to perform a geotechnical Lunne, T., Berre, T., & Strandvik, S. (1997). Sample dis
turbance effects in soft low plastic Norwegian clay.
zoning of 1) top compressible stratum, 2) liquefac Symposium on Recent Developments in Soil and Pave
tion potential, and 3) undrained shear strength su for ment Mechanics: 81–102.
the city of Guayaquil and surroundings. Mayne, P. (2007b). Invited Overview Paper: In-situ test
Diatoms have been identified in the soils of calibrations for evaluating soil parameters, Characteriza
Guayaquil, and the presence of these microstructures tion & Engineering Properties of Natural Soils, Vol. 3
possibly relates to the low Nkt values measured in (Proc. IS-Singapore), Tay-lor & Francis Group, London:
FVT. Although a valid trend is not observed between 1602–1652.
Nkt and Fr-Bq, there is a direct proportionality with Mayne, P., & Peuchen, J. (2018). Evaluation of CPTu Nkt
cone factor for undrained strength of clays. Proceedings
Qtn. Average values of Nkt varying between 8 and 10
of the 4th International Symposium on Cone Penetration
have been found for different geotechnical zones. Testing (CPT’18) (pp. 423–429). Delft, (ND): CRC Press.
There is high variability and heterogeneity in the Paredes, J. (2020, noviembre). Evaluación de parámetros
alluvial and estuarine soil deposits from Guayaquil, geotécnicos de los depósitos de suelos ubicados en la
based on different geotechnical zoning maps. More llanura aluvial y en el complejo deltaico estuarino de
over, the database highlights the suitability of CPTu Guayaquil. Master’s Thesis. Escuela Superior Politéc
investigation in soft compressible soils given that nica del Litoral (ESPOL), Guayaquil, Ecuador.
drainage conditions are better determined when Robertson, P. (2009a). Interpretation of cone penetration
detecting a sandy intercalation that may have not tests – a unified approach. Canadian Geotechnical Jour
been characterized while sampling. Likewise, CPT nal 46, 1337–1355.
Robertson, P. (2012). Interpretation of in situ tests - some
liquefaction analysis is more reliable since it would insights. J.K. Mitchell lecture, Proceedings of ISC’4,
consider sandy or clayey lenses that might aggravate (pp. 3–24). Recife, Brazil.
or alleviate foundation damage. Finally, pile driving Shiwakoti, D., Tanaka, H., Tanaka, M., & Locat, J. (2002).
analysis with CPT has been highly beneficial in sev Influences of diatom microfossils on engineering proper
eral construction projects for Guayaquil soil deposits, ties in soils. Soils and Foundations Vol. 42 No. 3, 1–17.
by modeling with great precision intercalations of Tonkin & Taylor Ltd. (2013). Liquefaction Vulnerability
dense soils that hinders pile penetration while estimat Study. Report to Earthquake Commission, ref.
ing shaft resistance accurately. Further research hopes 52020.0200/v1.0, prepared by S. van Ballegooy and
to provide detailed field results from case studies. P. Malan.
Vera, X. (2014). Seismic Response of a Soft, High Plasti
city, Diatomaceous Naturally Cemented Clay Deposit.
Doctoral Thesis. University of California, Berkeley.
REFERENCES Wiemer, G., & Kopf, A. (2017). Influence of diatom micro
fossils on sediment shear strength and slope stability.
Benítez, S., & Vera, X. (2006). Estudio Geológico de la Geochemistry, Geophysics, Geosystems 18, 333–345.
ciudad de la ciudad de Guayaquil. Informe final de Zhang, G., Robertson, P., & Brachman, R. (2002). Estimat
Investigación y Estudio del comportamiento dinámico ing licuefaction-induced ground settlements from CPT
del subsuelo. IIFIUC, Universidad Católica Santiago de for level ground. Canadian Geotechnical Journal, 39,
Guayaquil. 1168–1180.
1047
© 2022 the Author(s), ISBN 978-1-032-31259-0
Use of piezocone with dissipation tests CPTu_Δu, in tailings dams in

Mexico: Case history
J.L. Rangel-Núñez
Universidad Autónoma Metropolitana, Plantel Azcapotzalco, Mexico
E. Ibarra-Razo & R. Flores-Eslava

inGeum Ingeniería, Mexico
ABSTRACT: This paper describes and analyzes several aspects observed during the exploration of mine tailings
deposits in Mexico using the CPTu_Δu, test: high ground temperatures, interstratification, chemical reaction, dissi
pation curves with negative pore pressure, and S-wave profile. It was observed that the use of sensors with tempera
ture compensation for the qT and fs measurements produces satisfactory results if temperature changes are not
abrupt. For the case of the dynamic pore pressure sensor u2, where no temperature compensation was made, the
measurements are within the range mentioned in the standard except in the case when the temperature increases to
values of 50°C or higher. The presence of high interstratification between thin layers of hard and soft soils, with
different degrees of saturation, leads to errors in the measurement of the tip resistance and can produce deviations
from the verticality of the cone, making it necessary to carry out corrections. The different types of dissipation
curves observed in the tailings dams studied, where the soils are partially saturated, are generally the same as those
described by Sully et al. (1999), with the particularity that in some cases, the equilibrium pressure reaches negative
values. Finally, the S-wave profile obtained with correlations reproduces the general trend of the measured S-wave
profile. Nevertheless, it was not possible to capture the variations in this type of deposit. In the case history described
in this paper, the maximum values were not reproduced. Therefore, in this type of deposit, it is advisable the direct
measurement of the shear wave and to use the smaller intervals of measurement according to the technique used.
1 INTRODUCTION One of the techniques used to characterize soils in

tailings dams
� is the seismic� piezocone with dissipation
The design, execution and interpretation of an testings sCPTu-Δu test ; however, it has been
exploration campaign require experience and know observed that for this type of deposit there are factors
ledge of both geology and exploration techniques to that make it difficult to perform the test and their
achieve satisfactory results. The case of tailings interpretation.
dams is a particular challenge because this type of This paper describes and analyzes several aspects
deposit presents peculiarities, namely: they are very observed during the exploration of mine tailings
young, i.e., compaction or consolidation is some deposits in Mexico using the sCPTu-Δu test, that is:
times in the process; chemical processes may still be high ground temperature, interstratification, chemical
developing that can affect measurements, mainly in reaction with the soil and evaluation of the dissipa
gold mines; the different sequences and forms of tion test and s-wave velocity profile.
deposition using different types of debris induce In deposits where it is difficult to use any field
a highly variable heterogeneity and anisotropy in the exploration techniques, as in the case of tailings dams,
mechanical properties of the deposit. the execution of a piezocone test requires to following
All the mentioned aspects have an impact on the procedures that ensure the quality of the tests in order
geotechnical exploration campaign. For example, the to eliminate factors that hinder its interpretation and
number of stages of the exploration campaign could finally to reliably obtain both the stratigraphic profile
increase, as well as, the field and laboratory tests. and the geotechnical model of the deposit.
Undisturbed soil sampling is difficult, it is not possible In order to validate exploration with CPTu-Δu in
to use the correlations commonly employed to estimate tailings deposits, it is advisable to observe the fol
mechanical properties and it is often necessary to carry lowing aspects: verification of zeros at the beginning
out modifications of the processes of execution of the and end of the test, performing repeatability tests
field tests, such as increasing the rotation speed in more frequently, changing filters and verifying their
shear vane tests. saturation, evaluating tip and shaft continuously,
DOI: 10.1201/9781003308829-158
1048
measure verticality, measure temperature, control registered to validate the temperature compensation
driving speed and in general carry out continuous already described and in the case of the pore pres
maintenance of the equipment. sure, the objective was to observe the influence of
the temperature in their measurements.
Figure 1 shows the results obtained in the first
2 FACTORS AFFECTING THE experiment for each of the variables studied. it can
INTERPRETATION OF THE PIEZOCONE be seen that for the qc and fs sensors, the measure
TESTS ment variations are very small, falling within the
interval error indicated in ASTM D5778-12. Regard
A variety of CPTu-Δu and sCPTu-Δu tests have ing the pore pressure, u2, where the measurements
been carried out in several Mexican tailings dams were performed without compensating temperature,
where aspects that have hindered the performance the results indicate variations of about 8kPa for low
and interpretation of these tests have been observed. temperatures and 13kPa for high temperatures.
Some of them, which are particular aspects of Although, although they are small values, for a class
exploration in tailings dams, are described below. 1 cone the value determined for high temperatures is
in the limit according to the ASTM standard. On the
2.1 High ground temperature other hand, this temperature effect can induce errors
for the interpretation of dissipation tests, so it is
During the exploration of a tailings dam of a gold advisable to provide the cone with compensating
mine in northern Mexico, where high ambient temper temperature in the tailing deposit are expected.
atures arise, the piezocone temperature before driving
was 25°C, and during the test, it gradually increased
to a maximum value of 50°C, in other words, twice
the ambient temperature. High temperatures can
affect the measurements of the parameters qT and fs
when the equipment does not have compensating tem
perature sensors. If the cone has compensating tem
perature sensors, according to the manufacturer, the
sensors can stabilize the measurements if:
▪ The temperature gradient should be uniform as
the piezocone is driven, i.e. not present abrupt
temperature changes.
▪ The compensation is performed for temperature
variations or changes in the piezocone of up to
50°C (temperature variation in the piezocone with
respect to that adopted at ambient temperature).
For a temperature variation of up to 50°C,
a change in the voltage output of the load sensor
of up to 10 microvolts is compensated.
▪ The output voltage of the load sensor for the full-
scale output (FSO=75 MPa) is 8.7 mV, so that 10
microvolts correspond to 0.115% of the output volt
age for the full scale; in terms of stress, it represents
a compensation of approximately 86.20 kPa for
a maximum temperature change or variation of 50°
C (1.72 kPa/°C), so that for the case studied where
there was an increase of 25°C, there would be
a compensation of 43.1 kPa.
However, for the pore pressure sensor, the equip
ment does not perform any temperature compensa Figure 1. Effect of temperature on pore pressure (u2), tip
tion, so that the measurements of u2 and Δu may be (qc) and shaft (fs) sensors.
erroneous. In this sense, two experiments were pro
posed to evaluate these effects using corrections for The second experiment consisted of evaluating
qT and fs. the repeatability of the piezocone testing under field
The first experiment consisted of gradually conditions. To do this, two CPTs were executed, two
increasing the temperature of the water in which the meters apart for each other, in an area where variable
piezocone was submerged under controlled condi temperatures between 41.5 and 44°C were recorded
tions, from 4°C to 45°C, while recording the vari during the extraction stage. The temperature in the
ations in the zero readings of the tip, shaft and pore piezocone was measured with a digital thermometer.
pressure sensors: the first parameters (qT and fs) were Figure 2 shows the results obtained in both boreholes
1049
and it can be seen that this temperature gradient gen this, interstratification makes it difficult to interpret dis
erated during the driving of the equipment does not sipation tests because drainage and soil saturation con
affect the measurements, considering an ambient ditions are often different, so it is common to observe
temperature of 25°C at the beginning of the driving. that dissipation curves have different and unusual
shapes.
2.3 Dissipation curves

Dissipation tests were carried out using piezocones
with u sensor located at position 2 (u2). The dissipation
tests were executed at various sites of tailings dams,
where there are clays and silty clays and no ground
water level. Figure 4 shows the main types of dissipa
tion curves observed in the tailings dams studied:
a. With monotonic decay up to the equilibrium pres
sure, u0, either positive or negative.
b. A maximum is present and the equilibrium pres
sure may be positive or negative.
c. With monotonic growth up to the equilibrium
pressure.
Figure 2. Repeatability survey.
2.2 Interstratification
The presence of high interstratification between thin
layers of hard and soft soils, with different degrees
of saturation, causes sudden changes in the tip resist
ance. Moving from a hard to a soft stratum leads to
errors, low values in the measurement of the tip
resistance (Figure 3), making it necessary to carry
out corrections (Boulanger and DeJong, 2018) and
the instrumentation must guarantee low hysteresis in Figure 4. Main types of dissipation curves observed in the
order to obtain trustworthy data. studied tailings dams.
According to the criteria presented by Sully et al.

(1999), monotonic decreasing curves with a positive
value are of type I and II and occur in normally con
solidated fine soils; those with a maximum value are
of type III if the initial pressure is higher than the
equilibrium pressure, but if the initial pressure is
lower, they are of types IV and V. According to this
classification, all types of curves are present in the
tailings dams studied, with the particularity that there
are negative equilibrium pore pressures, although
with small magnitude, between 0kPa and -50kPa.
In the literature, methods have been proposed to
determine the value of the pressure increase Δu=ui-u0
Figure 3. Alternation between hard and soft lenses or layers generated during piezocone driving and determined
produces reductions on qT values.
in dissipation tests following the nomenclature shown
in Figure 5 (Burns and Mayne, 1998; Sully et al.,
Likewise, since the measurement of the parameters 1999; Paniagua et al., 2016; Imre et al., 2018), how
qT and fs is carried out at a given time but not at the ever, these methods are only applicable for type I to
same depth, and the alternation in the stiffness of the III curves. For the other cases, it is difficult to deter
soil can produce deviations in the verticality of mine the value of Δu due to the number of factors
the cone, it is necessary to make depth corrections. involved: fastening of the driving rods, position of
Therefore, the interpretation of the records is not imme the sensor, drainage and saturation conditions of the
diate since post-processing is required. In addition to soil, soil type, to mention the most important ones.
1050
Mexico City, where the groundwater level is superfi
cial, and although the deposit is saturated, there are
significant drops in pore pressures. Curves with
behavior similar to that observed in the tailings dams
are observed, but there is a case where u2 starts at
negative values.
2.4 Chemical reaction

It is common that after driving the cone, the probe
gets dirty, mainly in fine soils, and has to be cleaned
to start a new drilling, but a relevant aspect that was
observed in drillings on mining deposits, was the gen
eration of an impregnated oxide layer in the piezocone
Figure 5. Nomenclature used for conventional dissipation after the execution of a probe (Figure 7), which is dif
curves. ficult to remove with the process that is commonly
done even using a brush with steel bristles. One of the
Considering the hypothesis that the dynamic pore important characteristics of the deposits where this
pressure u2 is given by eqs 1 and 2 (Burns and aspect was observed was that there was a constant
Mayne, 1998), and given that the octahedric compo emanation of gaseous ammonia (NH3) generated as
a consequence of the leaching process. The most feas
nent produces only positive increases in pore pres
ible and efficient solution to remove this oxide layer
sure, the shear component, which depends on the
dilative/contractive behavior of the soil, would be was to immerse the removable parts of the piezocone,
one of the factors responsible for the suction gener such as the shaft jacket and the apex in acetic acid
(CH3 CO OH), as well as to clean the fixed parts with
ated in the driving process for the dissipation curve
types III to V. The decrease of the initial pore pres the same compound and with the help of a toothbrush.
sure in the dissipation test can also be caused by the
abrupt stop of the driving (Kurup and Tumay, 1995),
by the actual drainage conditions, by the saturation
conditions of the soil, by the movement of the bars
after stopping the driving, among other factors.
As previously mentioned, all the tests were per

formed in tailings deposits where there is no ground
water level, so the soils are partially saturated. For
comparison purposes, Figure 6 shows dissipation
tests carried out in the clays and silty clays of
Figure 7. Piezocone with oxide induced by the minerals
that make up the tailings deposit.
2.5 S-wave velocity profile

The tailings dams are formed by deposits of very
young soils, some of them still in the process of com
paction or consolidation, where the definition of the
stratigraphic profile and the evaluation of the mechan
ical properties of each stratum is difficult, mainly
because there is heterogeneity, both vertical and hori
zontal, and it is also difficult to obtain unaltered sam
ples. Therefore, in this type of deposits, field
techniques are used, especially sCPTu-Δu and VST
tests, because it is possible to define the stratigraphic
profile and determine, either directly or indirectly, the
Figure 6. Main types of dissipation curves observed in
mechanical properties of each unit, particularly
clayey soils in Mexico City.
1051
parameters such as modulus of stiffness at small same depth, and the alternation in soil hardness can
deformations (G0), permeability coefficient (k) and produce deviations from the verticality of the cone, it
undrained shear strength (su), in order to carry out sta is necessary to carry out depth corrections. Therefore,
bility analyses. the interpretation of the logs in this type of terrain is
The velocity profiles obtained at one site from the not immediate, since post-processing is required.
sCPTu are presented in Figure 8. In this case, the The different types of dissipation curves observed
determination of the S-wave velocity profile is per in the tailings dams studied, where the soils are par
formed with two procedures: direct measurement by tially saturated, are generally the same as those
means of the seismic modulus and by means of described in Sully et al (1999), with the particularity
empirical correlations. It is observed that, although that in some cases the equilibrium pressure has nega
the correlation generates satisfactory results, it was tive values.
not able to reproduce the alternation of values, espe A case is described in which a chemical reaction
cially the maximum values. occurs in the piezocone during its driving, which
may affect the readings. Ammonia generated as
a consequence of the leaching process induces oxi
dation, so it is necessary to perform continuous
maintenance of such equipment, and for the type of
deposits studied vinegar proved to be the best sub
stance for cleaning the cones.
Finally, velocity profiles are determined for two
sites of tailings dams through correlations with the
value of qT and the one determined with the seismic
modulus. It is observed that the profile obtained with
correlations reproduces the general trend but it is not
possible to capture the variations of the shear wave
velocity in this type of deposits, in the particular case,
the maximum values are not reproduced, therefore, in
Figure 8. Determination of the S-wave profile by correl this type of deposits it is advisable the direct measure
ations and with the seismic modulus. ment of the shear wave and to use the smaller inter
vals of measurement according to the technique used.
3 CONCLUSIONS
REFERENCES
This paper describes and analyzes several aspects
observed during the exploration of mine tailings ASTM D5778–12 “Standard Test Method for Electronic
deposits in Mexico using the CPTu-Δu test: high Friction Cone and Piezocone Penetration Testing Of
ground temperatures, interstratification, chemical Soils”.
reaction, dissipation curves, and S-wave profile. Boulanger R.W. and DeJong J.T., (2018), “Inverse filtering
High temperatures can affect measurements of the procedure to correct cone penetration data for thin-layer
parameters qT, fs, u2, and Δu when the equipment and transition effects”, Penetration Testing 2018,
does not have compensators, but if they do exist, it is Hicks, Pisano and Peuchen (Eds), Delft University of
Technology, The Netherlands, 25–44
required that temperature changes are not abrupt so Burns S.E. and Mayne P.W. (1998), “Monotonic and
that the sensor can stabilize during the measurement. dilatory pore pressure decay during piezocone test in
For this reason, experiments were carried out in this clay”, Canadian Geotechnical Journal 35,
type of deposits using equipment with compensators, 1063–1073.
finding that the variations of qT and fs measurements Kurup P.U. and Tumay M.T. (1995), “Piezocone dissipation
were found to be within the error range indicated in curves with initial excess pore pressure variation”, Proc.
the standards (ASTM D5778-12). Regarding the Int. Symposium on Cone Penetration Testing, CPT´95,
dynamic pore pressure, u2, where the measurement Linköping, Sweden, 195-200.
was performed without a temperature compensator, Imre E., Schanz T., Bates L. and Fityus S. (2018), “Evalu
ation of complex and/or short CPTu dissipation test”,
the results indicate variations of about 13kPa, which, Cone Penetration Testing 2018, Hicks, Pisano and Peu
although small, can induce errors for the interpret chen (Eds), Delft University of Technology, The Nether
ation of dissipation tests. lands, 351–357
The second factor analyzed was the presence of Paniagua P., Carrol R., L´Heureux J.S., and Nordal S.
high interstratification between thin layers of hard and (2016), “Monotonic and dilatory excess pore water dis
soft soils, with different degrees of saturation. Moving sipation in silt following CPTu at variable penetration
from a hard to a soft stratum leads to errors in the rate”, Proc. IS Osaka-Engineering Practice and Per
measurement of the tip resistance, making it necessary formance of Soft Deposits, Osaka, Japan, 147–152
to carry out corrections (Boulanger and DeJong, Sully P.J., Robertson P.K., Campanella R.G., and
Woeller D.J. (1999), “An approach to evaluation of field
2018). Also, since the measurement of qT and fs CPTu dissipation data in over consolidated fine-grained
parameters is performed at a given time, but not at the soils”, Canadian Geotechnical Journal 36, 369–381.
1052
© 2022 the Author(s), ISBN 978-1-032-31259-0
Assessment of the spatial variability of a Croatian flood embankment using

the cone penetration test
C. Reale
University of Bath, UK
M.S. Kovačević & M. Bacic

University of Zagreb, Croatia
K.G. Gavin
Delft University of Technology, The Netherlands
ABSTRACT: Understanding how soil varies spatially is necessary in order to accurately quantify the reliabil
ity of geotechnical infrastructure. For long linear infrastructure such as flood embankments, incorporating vertical
and horizontal scales of fluctuation can have a significant impact on stability assessments. This paper presents
preliminary results and discussion from a field test designed to determine the vertical and horizontal scales of
fluctuation of a Croatian flood embankment. A series of 15 CPTUs were carried out over a 200m length of the
embankment with a Multi-channel Analysis of Surface Waves (MASW) survey done on the same section. CPT
spacing was designed specifically to determine horizontal variation with multiple CPTs carried out in close prox
imity to each other. There was significant variation in soil stratigraphy over the embankment section with pock
ets of increased strength and stiffness showing up in the MASW and CPT results. This paper discusses dealing
with horizontal correlation in challenging deposits and presents initial findings from the underlying sand layer.
1 INTRODUCTION stochastic but it is convenient to model it as such, as

information is limited(Phoon, 2008). If a dataset is not
Soil properties vary as a function of space and time. stationary, then estimation of soil statistics may be
Quantifying exactly how they vary is of upmost subject to bias. To prevent this, the dataset should
importance in determining the capacity of geotech undergo transformation to achieve stationarity, data
nical structures(De Gast, Vardon and Hicks, 2020). decomposition is frequently adopted to this end.
This is particularly important for aged linear infra Phoon et al (2003) describe a modified Bartlett
structure, which underwent less rigorous design than hypothesis test that can be used to check stationarity.
its modern counterparts and has a lower safety If one is concerned about the average properties
margin as a result (Reale et al., 2016; Reale, Xue within some volume of soil (e.g. the average shear
and Gavin, 2017). It is infeasible and inadvisable to strength or average resistance of a material) then areas
replace such infrastructure en-masse, moreover it is of high value balance areas of low value so that the
likely unnecessary given that the structures have variance of the average goes down as the volume of
remained stable so far. Instead, a better approach is soil mobilised becomes larger. Point variations such
to quantify the uncertainties present in the material in as those listed by (Phoon & Kulhawy, 1999) are typ
order to accurately assess the risk of failure. Such an ically much higher than spatially averaged variations.
approach could be applied consistently with a risk- Spatial averaging therefore reduces uncertainty. The
based decision methodology to decide which assets net result of which is lower failure rates which are
to improve or replace (Reale, Xue and Gavin, 2016). more consistent with those observed in reality (Phoon,
Random field theory considers soil properties at 2008). Depending on the scale of fluctuations
a given location as random variables. Within a zero involved spatial averaging can have a significant
mean stationary random field, the spatial variation impact on the stability of linear infrastructure.
between one point and another in the same structure Random field theory has been successfully
can be described by their correlation structure. Statis used to generate one dimensional, two dimen
tically modelling spatial variability presumes sional and three dimensional geological models
a stationary random field. Such an assumption may be (Zhu and Zhang, 2013; Lloret-Cabot, Fenton and
hard to achieve in reality given that soil is not really Hicks, 2014) and applied to various geotechnical
DOI: 10.1201/9781003308829-159
1053
problems such as bearing capacity of foundations
(Fenton and Griffiths, 2003; Srivastava and Babu,
2009),water flow (Renato et al., 2006), two
dimensional slope stability (Srivastava, Babu and
Haldar, 2010; Santoso, Phoon and Quek, 2011;
Tabarroki, Ahmad and Banaki, 2013; Li et al., where j = 0,1,…, n-1 with n being the number of
2014), three dimension slope stability (Hicks and data points, τj ¼ jDτ is the lag distance between the
Spencer, 2010), scour (Prendergast, Reale and two points in question where Dτ is the distance
Gavin, 2018) and suction caisson design between two adjacent points, μ is the estimated
(Remmers et al., 2019). mean, σ is the standard deviation and X is the
This study describes the initial findings from random soil property in this case tip resistance.
a series of CPTs investigating the horizontal and ver A Markov correlation function (Lloret-Cabot,
tical variability of a flood embankment in Croatia, Fenton and Hicks, 2014; Kasama and Whittle, 2015)
the findings will be utilised in a wider study to inves was used to approximate the spatial correlation
tigate the stability of the flood defenses and to assess structure, see Equation 4. It is important to note, that
their potential for liquefaction. many correlation functions exist and the choice of
correlation function will depend on the goodness of
fit achieved with the underlying correlation structure.
2 METHODOLOGY The scale of fluctuation θ, is varied until the correl
ation structure obtained from Equation 3 is described
Decomposition can be used to investigate a CPTs by the correlation function i.e. until the difference
underlying spatial correlation structure where a trend between ^ρðτÞ and ρðτÞ is negligible.
function is fitted to and extracted from a dataset
using least squares or some similar approach. This
removes any underlying trend from the data leaving
behind some fluctuating component. The correlation
structure of this fluctuating component can then be
determined.
After removing any discernible trend, the soil For horizontal spatial variation the Equation 4
property (in this case qc) for a normal distribution holds except there is a greater distance between
can be described by Equation 1. measurement points and the spacing between points
is unlikely to be uniform. In practice this compli
cates the process as it becomes difficulty to deter
mine when like is being compared with like. The
authors compared the top of the sand layer in each
CPT as if they occurred at the same depth and deter
where μ is the mean value described at some mined horizontal correlations across CPT using a 1m
depth z using Equation 2, σ is the standard deviation moving window therein.
at the same depth and G is a matrix containing
n spatially correlated normal random processes of
zero mean and unit variance which account for the 3 TEST SITE
spatial correlation structure of the soil.
The CPT test location is located next to
a embankment in central Croatia, in Orle Munici
piality, around 25 km from Zagreb. The embank
ment is part of a flood defence network which
where ai is the mean trends value at the beginning protects the wider area from the influence of the
of the ith layer, bi is the slope of the trendline in Odra and Sava rivers. CPTs were performed over
question and z is the depth. a 200m length at the following spacings [0 m,
After removing the linear depth trend of each qc 2 m, 5 m, 10 m, 25 m, 50 m, 75 m, 100 m,
profile in the dataset, the standard deviation of the 125 m, 150 m, 175 m, 190 m, 195 m, 198 m,
detrended tip resistances can be calculated. Divid 200 m], see Figure 1, to a depth of 15 m. The
ing an individual detrended tip resistance by its first soil layer at the site was a clay of variable
respective standard deviation transforms the tip thickness, with deeper deposits of between 7m
resistance variation into the standard normal space. and 8m found on either end of the 200m test
i.e. it produces normal random fields with a mean length. Layer thickness reduced to approximately
of zero and a standard deviation of 1. Variations 2m in the middle of the test length. Underneath
within these normal random fields can be used to the clay was a dense sand deposit to great depth.
estimate the spatial correlation structure b
ρðτj Þ of the The full set of CPT traces demonstrated a lot of
tip resistance with depth or horizontal distance, see variability, particularly within the sand layer and
Equation 3. at transition depths, see Figure 2.
1054
Figure 1. CPT traces and MASW were performed along
the blue line adjacent to the embankment.
Figure 4. CPT traces from 10 to 15m with the mean qc

profile shown in red.
4 RESULTS
The CPT traces in the lower sand layer were

detrended, normalised and the underlying correlation
structure was determined using the procedure
described in section 2. Equation 4 was then fitted to
the underlying structure by optimizing the scale of
fluctuation. The vertical scale of fluctuation θV was
found to be 0.82 m with the 95% confidence intervals
ranging from 0.78 m to 0.85 m. A strong goodness of
Figure 2. 15 CPT traces across the site. fit with achieved using the Markov correlation function
with an R-square of 0.9358 and an RMSE of 0.0528.
The same procedure was followed for horizontal
Multi-Channel Analysis of Surface Waves correlation with intermediate points interpolated to
(MASW) was also performed at the site, see Figure 3,
facilitate curve fitting. Initial results indicated that
which corroborated the assumed stratigraphy showing
the horizontal scale of fluctuation was approximately
deeper deposits of low shear wave velocity at either 8.39 m with a 95% confidence interval range of
end of the test length. The presence of localized 7.45 m to 9.33 m.
deposits of increased stiffness can also be seen within
the sand layer, demonstrating the variability of the
material.
Due to the complex layering present in the upper
clay layer the initial spatial variability interrogation
is focused in the lower sand layer from a depth of 10
to 15m where the soil behaves more consistently
across the site, see Figure 4.
Figure 3. Shear wave velocities from MASW performed at

the site, demonstrating the variability in the depth of the Figure 5. Vertical correlation structure found in the lower
sand layer across the site. sand layer.
1055
REFERENCES
Fenton, G. A. and Griffiths, D. V (2003) ‘Bearing-capacity
prediction of spatially random c – ϕ soils’, Canadian Geo-
technical Journal, 40(1), pp. 54–65. doi: 10.1139/t02-086.
De Gast, T., Vardon, P. J. and Hicks, M. A. (2020) ‘Assess
ment of soil spatial variability for linear infrastructure
using cone penetration tests’, Géotechnique, pp. 1–15.
doi: 10.1680/jgeot.19.sip.002.
Hicks, M. and Spencer, W. (2010) ‘Influence of heterogen
eity on the reliability and failure of a long 3D slope’,
Computers and Geotechnics, 37(7–8), pp. 948–955. doi:
10.1016/j.compgeo.2010.08.001.
Kasama, K. and Whittle, A. J. (2015) ‘Effect of spatial variabil
ity on the slope stability using Random Field Numerical
Limit Analyses’, Georisk: Assessment and Management of
Risk for Engineered Systems and Geohazards, 10(1), pp.
Figure 6. Initial horizontal correlation structure found in 42–54. doi: 10.1080/17499518.2015.1077973.
the lower sand layer. Li, D.-Q. et al. (2014) ‘Effect of spatially variable
shear strength parameters with linearly increasing
mean trend on reliability of infinite slopes’, Struc
tural Safety, 49, pp. 45–55. doi: 10.1016/j.strusafe.
2013.08.005.
5 DISCUSSION AND CONCLUSIONS Lloret-Cabot, M., Fenton, G. and Hicks, M. (2014) ‘On the
estimation of scale of fluctuation in geostatistics’, Geor
This paper presents initial results from a site inves isk: Assessment and Management of Risk for Engineered
tigation to investigate spatial variability in Croatia. Systems and Geohazards, 8(2), pp. 129–140. doi:
Initial work was focused on the underlying sand 10.1080/17499518.2013.871189.
layer which behaved more consistently than the Phoon, K.-K., Quek, S.-T. and An, P. (2003) ‘Identification
of Statistically Homogeneous Soil Layers Using Modi
finer surface deposits. Initial results suggest than fied Bartlett Statistics’, Journal of Geotechnical and
horizontal variability is an order of magnitude Geoenvironmental Engineering, 129(7), pp. 649–659.
greater than the vertical variability. However sig doi: 10.1061/(asce)1090-0241(2003)129:7(649).
nificant uncertainty exists in determining the hori Phoon, K. (2008) Reliability-based design in geotechnical
zontal correlation structure. One issue stems, from engineering: computations and applications. Edited by
the variability in layer depth across the site, as the K. Phoon. Taylor and Francis.
boundary of the sand layer is inclined. This makes Phoon, K. and Kulhawy, F. (1999) ‘Characterization of
it difficult to correlate “like with like” across the geotechnical variability’, Canadian Geotechnical Jour
site as merely correlating CPTs at the same depth nal, 36(4), pp. 612–624.
Prendergast, L. J., Reale, C. and Gavin, K. (2018)
could result in correlating across layer boundaries. ‘Probabilistic examination of the change in eigen
To overcome this the authors considered the top of frequencies of an offshore wind turbine under pro
the sand layer in each CPT as the start point of the gressive scour incorporating soil spatial variability’,
analysis and used a moving window of 1m to deter Marine Structures, 57, pp. 87–104. doi: 10.1016/j.
mine horizontal correlations below those point. As marstruc.2017.09.009.
there is no continuous measurement in the horizon Reale, C. et al. (2016) ‘Multi-modal Reliability Analysis of
tal direction, there needs to be a methodology to Slope Stability’, Transportation Research Procedia, 14,
ensure that the correct data is being used to deter pp. 2468–2476. doi: 10.1016/j.trpro.2016.05.304.
mine the mean autocorrelation behavior horizon Reale, C., Xue, J. and Gavin, K. (2016) ‘System reliability
of slopes using multimodal optimisation’, Géotechnique,
tally. Different averaging procedures will be 66(5), pp. 413–423. doi: 10.1680/jgeot.15.P.142.
investigated, while different window sizes to deter Reale, C., Xue, J. and Gavin, K. (2017) ‘Using Reliability
mine the effect of sample size on scales of fluctu Theory to Assess the Stability and Prolong the Design
ation. The analysis will also be rerun to consider Life of Existing Engineered Slopes’, in Risk Assessment
the start point at intermediate CPT locations to and Management in Geotechnical Engineering: from
check consistency across the site. Some consider Theory to Practice, ASCE Geotechnical Special Publi
ation also needs to be paid to stationarity. This is cation in Memory of the Late Professor Wilson H. Tang,
difficult to ensure in the vertical direction, but pp. 61–81. doi: 10.1061/9780784480731.006.
much more so in the horizontal direction where Remmers, J. et al. (2019) ‘Geotechnical installation design
of suction buckets in non-cohesive soils: A
there are much more limited discrete measure reliability-based approach’, Ocean Engineering, 188, p.
ment points which makes the implementation of 106242. doi: 10.1016/j.oceaneng.2019.106242.
stationarity checks such as Bartlett statistics chal Santoso, A., Phoon, K. and Quek, S. (2011) ‘Effects of soil
lenging. This will be investigated moving spatial variability on rainfall-induced landslides’, Com
forward. puters & Structures, 89(11–12), pp. 893–900.
1056
Srivastava, A. and Babu, G. (2009) ‘Effect of soil vari Tabarroki, M., Ahmad, F. and Banaki, R. (2013) ‘Determin
ability on the bearing capacity of clay and in slope ing the factors of safety of spatially variable slopes mod
stability problems’, Engineering Geology, 108(1–2), eled by random fields’, Journal of Geotechnical and
pp. 142–152. Geoenvironmental Engineering, 139(12), pp. 2082–2095.
Srivastava, A., Babu, G. and Haldar, S. (2010) ‘Influence Zhu, H. and Zhang, L. (2013) ‘Characterizing geotechnical
of spatial variability of permeability property on steady anisotropic spatial variations using random field theory’,
state seepage flow and slope stability analysis’, Engin Canadian Geotechnical Journal, 50(7), pp. 723–734.
eering Geology, 110(3–4), pp. 93–101. doi: 10.1139/cgj-2012-0345.
1057
© 2022 the Author(s), ISBN 978-1-032-31259-0
Prediction of bearing capacity and settlement using penetrometer design

method for shallow foundation and load transfer curves
P. Reiffsteck & F. Szymkiewicz
GERS-SRO, Université Gustave Eiffel, IFSTTAR, Marne la Vallée, France
M.A. Benz Navarrete & T.A. Luong

Research and Development, Sol-Solution, Riom Cedex, France
ABSTRACT: This paper presents a simple, economic, and sound design method that could be easily imple
mented worldwide, based on a new generation version of Dynamic Cone Penetrometer Test (DCPT) devices
giving access to the whole cone resistance versus penetration depth of the cone curve, to predict the bearing
capacity, as well as to estimate the load – settlement curve of a shallow foundation which is partly inspired
from the French Penetrometer design method. A methodology is then proposed in this study for the concep
tion of shallow foundation based on load transfer curves obtained from DCPT.
1 INTRODUCTION measurements made during the dynamic penetrometer

driving to obtain a dynamic cone load-penetration
Spread foundations are the most extensively used curve at cone-soil interface, which is qualitatively
geotechnical structures in civil engineering world, as comparable to the expansion curve obtained by the
they are simple to construct and the most cost- pressuremeter as well as settlement curve obtained
effective choice for support of a building furthermore. during shallow foundation loading test.
These foundations are implemented for all kinds of This paper then introduces a methodology for the
structure from small individual buildings to bridges. calibration of the transfer curves from the DCPT
For the conception of shallow foundations including tests, therefore, to predict the load – settlement curve
the evaluation of bearing capacity and its settlement, the for the conception of the shallow foundations.
conventional methods such as c and ’ method (Skemp
ton, 1951) (Terzaghi, 1943) and the one-dimensional
approach often overestimate limit load and underesti 2 DYNAMIC CONE LOADING TEST (DCLT)
mate settlement, which leads to a non-conservative
design or differential settlement, both being detrimental 2.1 Description of DCLT apparatus
to the structures (Poulos & Carter, 2001). The instrumented dynamic cone penetrometer (DPT)
Static and dynamic cone penetration tests has been designed and based on the same functional
(respectively, CPT and DCPT) are the techniques in principle as the P.A.N.D.A. (from French Pénétro
the ground investigation and foundation design, mètre Autonome Numérique Assisté par ordinateur).
which are widely used in many countries around the It is an instrumented lightweight DPT which is
world. However, in conventional procedures, these driven with variable energy (Gourvès & Barjot, 1995)
tests only allow a fairly assessment of soil deform provided by the operator by means of a hand-hammer
ation since no direct relationship between load and mass. For each blow, the energy supplied as well as
settlement (stress strain relation) is obtained during total penetration is directly measured (Benz Navarrete,
the test. Thus, the development of a new test proto 2009).The new version of P.A.N.D.A. used in this
col and equipment of Cone Penetrometer Test in its work, includes new sensors and the general principle is
dynamic version are improved not only to ensure the based on wave equation solution. A general description
advantages of economy and simple operation, but of the device as well as the theoretical implemented
also improve the measurement capacity to derive approach is presented by (Benz Navarrete, et al., 2021).
various soil parameters and increase their reliability. After each hammer blow, instrumented anvil
The dynamic cone loading test (DCLT), developed records strain, acceleration, and displacement. An
very recently in France (Benz, et al., 2013) (Escobar example of raw measurements performed during the
Valencia, et al., 2013), consists of analyzing the test is shown in Figure 1.b.
DOI: 10.1201/9781003308829-160
1058
Figure 1. (a) Dynamic penetrometer P.A.N.D.A.3 (Benz
et al. 2013 and (b) raw measurements (acceleration, force,
and displacement).
Penetrometer is composed of rods with a diameter

of 14 mm and 500 mm in length and of overflowing
conical tips with a cross-section of 2 or 4 cm2
(respectively, 15.9 or 22.5 mm in diameter). In
accordance with the ISO-22476-2 Standard (2005),
the apex angle of cones is 90°. To avoid the effects
Figure 2. Example of (a) force decoupled waves in meas
of skin friction on the rods, the use of overflowing urement point xA, and reconstructed cone signals (b) vel
conical tips (having a larger diameter than that of the ocity and displacement; and (c) cone stress.
rods) has proved to be sufficient. However, jacking
or mud injection can be implemented in necessary
(ISO 22476-2).
2.2 Getting DCLT curves

The derivation of mechanical parameters is based on
pile dynamic, rapid shocks, Split Hopkinson Pres
sure Bars (SPHB) as well as rock percussion bore-
hole tests (Rausche, 1970; Karlsson, et al., 1989;
Bussac, et al., 2002; Omidvar, et al., 2014; Omidvar,
et al., 2015).
Figure 2 shows for each blow, by supposing no
skin friction along the rods and from measurements
carried out in the upper end of rod string (close to
the anvil), elementary downward and upward waves Figure 3. Field Dynamic Cone Loading (DCLT) curves
(Jossigny, France).
are decoupled. From that, force Fp(t), velocity vp(t)
and settlement sp(t) are then computed at cone/soil
interface (Benz, et al., 2013) (Escobar Valencia, curve can be modelled by a perfectly elasto-plastic
et al., 2013) and (Tran, et al., 2019). Dynamic Cone and dashpot model, where total soil resistance qd(t) is
Load-penetration (DCLT) curves are obtained by the sum of pseudo-static qps, damping qdyn(t) and iner
plotting cone resistance qd(t) as a function of its tial qin resistances (Eq. (1)) which can be neglected
settlement sp(t) (Figure 3). due to the low mass of penetrometer. Pseudo-static
Cone resistance qd(t), shown in Figure 3, represents resistance qps is displacement dependent and mod
the total soil resistance under dynamic loading. DCLT elled by a non-linear elastic perfectly plastic law
1059
while qdyn(t), is penetration rate dependent and mod ○ Limit resistance assessment is obtained after
elled with a radiation dashpot (Eq. (2)). inertial correction, qlim = Fp(t)* - mp.a(t0);
where mp is the mass of penetrometer and
a(t0) is average acceleration at t0.
In this way, knowing Es and qlim, DCLT curves
can be modeled by a nonlinear elastic plastic
model as the Simplified Hyperbolic one pre
sented below.
_ and u are, respectively, acceler
In Eq. (2), ü, u,
ation, velocity, and displacement at penetrom
eter’s cone/soil interface. Qd is the total strength 2.3 Simplified hyperbolic model adjustment to
and Ap the cone section. cb and kb are the dash- DCLT
pot and spring constant (El Naggar & Novak, The possibility to establish the cone resist
1996), defined by: ance–penetration curve of DCLT is to fit
a predetermined curve type suitable for repro
ducing the test from the experimental curves,
with the choice set on the characteristic
features of the curve which are the modulus E s
and the limit resistance q lim . A similar
development has been proposed by (Elhakim,
2005).
To reproduce the curves in a systematic way
based on the parameters that are the cone resistance
Where, R(H) is equivalent radius introduced by and the dynamic modulus, we will use a hyperbola
(Holeyman, 1988) and computed according to formulation. The final version homogenizes the
(Novak & Beredugo, 1972; El Naggar & Novak, terms of the formula and systemizes the determin
1994). Gs, ρs and vs are respectively shear modulus, ation of the Ai parameters, closer to the method pro
density, and Poisson’s ratio of soil. posed by (Baud & Gambin, 2008), the
Plastic resistance or soil limit resistance qlim is deformation ε:
obtained from velocity, displacement, and strength
signal according to the Unloading Point Method pro
posed by (Middendorp et al. 1992; Hölscher et al.
2012) and explained by (Brown 1994, 2016).
Moreover, spring constant or dynamic stiffness
kdyn is determined from DCLT experimental meas
urements by mean of FRF transfer functions (Davis
1975; Paquet and Briard 1976; Tran et al. 2019). For
the case of a rigid embedded foundation subject to
transient dynamic loading, it is assumed that at low
frequencies, kdyn is close to kb. Knowing kb and
Where, x ¼ qqd2 a non dimensional term, where q2 is
assuming Poisson’s ratio vs, shear modulus Gs can
the inflection point at the end of the linear part, cor
be obtained from Eq. (3). Moreover, Elastic modulus
responding to the conventional definition of
(Es) can be then obtained by applying the theory of
pLM ¼ 1:7pf M . We can therefore estimate x with
elasticity. Other elastic parameters can be deduced.
a value around
� 0,6 �to 0,7.
In practice, DCLT curves are analyzed as follows: α ¼ qs22hEe ¼ 1 - 2 hbe Cf , a non dimensional term,
– Assessment of dynamic stiffness kb from FRF with Cf ¼ 0:79 corresponding to the form coefficient
curves, of the rigid circular
� cone � of P.A.N.D.A, an influence
– Determination of shear modulus Gs (Eq. (3)). depth he ¼ πR 4 1-
2
(Butterfield & Banerjee,
– Elastic modulus (Es) assessment from Gs and νs. 1971), b and R being respectively diameters and
– Determination of soil limit resistance qlim from radius of the cone.
DCLTs curves and Eq. (2) according to UPM β is equal to 0:95α, this ratio allowing to decrease
method: the initial slope, in order to guide the hyperbola with
○ From cone velocity signal vp(t), find the an upward concavity.
moment (t0) where vp(t) is zero after the blow. Figure 4 shows the adjustment curves modelled
○ Confirm that the t0 moment coincides with max by this version, based on data of DCLT tests that
imum tip displacement sp(t) is at its maximum. are the tip resistance, qd and the dynamic modu
○ From strength signal Fp(t), computed average lus, Edyn at the depth from 1 m to 1.25 m, in
value Fp(t)* at (t0 ± dt) moment (with dt Jossigny site and at the depth from 1.75 m to
≈0,1ms). 2 m, in Cran site.
1060
5 and real load pressure anticipated up to qnet
obtained with equation 6. Results obtained with the
proposed method are compared with settlement pre
dicted by Schmertmann method as proposed by
Eurocode 7 part 2.
Figure 5. Principle of the similitude between shallow foun

dation loading test and DCLT.
4 FOUNDATION LOADING TESTS
4.1 Loading procedure

Figure 4. Curve adjusted in (a) Jossigny from 1 to
1.25 m and (b) Cran from 1.75 to 2 m. Several loading tests of B = 1; 0.71 m square and
0.6 m circular footings have been performed on sev
eral sites. Ground investigation using P.A.N.D.A
3 PROPOSED DESIGN METHOD (DCLT), PMT, CPT, SPT and other techniques
helped to define the ground model.
3.1 Bearing capacity Foundations were loaded, using load steps of
a 30-minutes duration, until failure occurred, with an
Bearing capacity of the foundation is computed average number of 10 steps. The Figure 6 and 7
using the French penetrometer design direct method show schematically the equipment used for shallow
with slight modification (AFNOR, 2013). foundation loading tests.
With the qnet the net pressure under the foundation,

q0 the total vertical stress at the base of the founda
tion after construction, kd the penetrometer bearing
capacity coefficient, qde the equivalent cone resist
ance, De and B respectively, the equivalent embed Figure 6. Schematic shallow foundation loading test.
ment depth and the width of foundation and kd0 , a,
b, c and d the coefficients tabulated according to the
type of soil and the shape of the foundation.
3.2 The load-settlement curve approach

As shown in Figure 5, a similar approach to the Load
Settlement Curve Method proposed by Briaud has
been developed to access the full load settlement
curve (Briaud, 2007). The principle is based on the
similarity of the cone resistance – penetration curve
acquired by the DCLT at each hammer blow with the
loading – settlement curve of the shallow foundation.
This transformation based on full-scale tests and
numerical simulations is achieved using the equation Figure 7. Foundation loading test in Cran.
1061
4.2 Sites presentation and tests performed
Table 1 summarizes information on two sites of dif
ferent soil types where foundation and DCLT tests
have been performed. Based on the results of several
shallow foundation load test for each site, a mean
curve can be fitted (by least-squares method).
Table 1. Sites and number of DCLT test.
Site Soil type DCLT test

Quantity DCLT curves /ml
Jossigny Silty 6 130 c/ml

Cran Marine clay 9 128 c/ml Figure 9. Distribution curves of the experimental and com
puted bearing capacity ratio for (a) square 1x1m footing in
Figure 8 presents the mean loading – settlement Jossigny and (b) circular D = 0.6 m footing in Cran.
curve for Jossigny site for the case of 1m square
foundations and two experimental curves for Cran prediction of bearing capacity, for Jossigny as well
site for the case of a 0.6 m circular footing. as Cran sites, especially the results from DCLT
method.
Figure 10 shows settlement prediction, by
4.3 Tests results and comparison to prediction four design methods: the French Application
The bearing capacity prediction using the pressure- Standard of EC7 (NF P94-261), the Load
meter and penetrometer methods, proposed in the
French national application standard of Eurocode 7
NF P94-261 for shallow foundation design, is pre
sented in Figures 9 and 10.
Figure 8. Mean curve of foundation load tests performed on

Jossigny (Square 1x1m) and Cran (Circular D = 0,6m) sites.
The cumulative distribution curves of ratios

between calculated and measured bearing capacity
coefficients are compared with Normal law distribu
tion function.
The DCLT method results superimposed well on
the ones obtained using pressuremeter method as
proposed by NF P94-261. The safety level of the
bearing capacity is represented by the deviation from Figure 10. Comparison of experimental and predicted settle
an average value of 1. Figure 9 shows that the pre ment curves. For (a) 1x1 m square foundation on silt in Jossigny
diction of these methods tends to a conservative and (b) D = 0,6 m circular foundation on marine clay in Cran.
1062
Settlement Curve Approach (LSCA) proposed by cone penetration test for. Journal of Rock Mechanics
(Briaud, 2007), the Simplified Hyperbolic (SH) and Geotechnical Engineering, p. 31.
model for PMT and the Schmertmann method, Benz, M., Escobar, E., Gourvès, R. & et al, 2013. Dynamic
for a square footing in Jossigny and circular measurements of the penetration test - Determination of
footing in Cran. We also note that the values of the tip’s dynamic load-penetration curve. In: 18th Inter
the respective conventional bearing capacity of national Conference on Soil Mechanics anf Geotech
soil in Jossigny and in Cran are respectively nical Engineering: Challanges and Innovations in
Geotechnics, ICSMGE 2013.
about 36 and 6 tons. Briaud, J.-L., 2007. Spread Footings in Sand: Load Settle
Figure 10a, compares the mean experimental ment Curve Approach. Journal of Geotechnical and
curve (Figure 8) to computed settlement. The pre Geoenvironmental Engineering, pp. 905–920.
dicted settlement obtained with NF P94-261 and Brown, M., 2016. Design methods based upon rapid pile
Schmertmann methods and the SH model for PMT load tests. ISSMGE - ETC 3 Int Symp Des Piles Eur.
well match the experimental mean curve, in the first Brown, D., 1994. Evaluation of Static Capacity of
half the loading phase. However, the LSCA method Deep-Foundations from Statnamic Testing. Geotech Test
underestimates the settlement. One drawback of this J 17, p. 403–414.
Bussac, M., Collet, P., Gary G & Othman, R., 2002. An
method is the non-zero settlement for a loading close
optimisation method for separating and rebuilding
to zero. one-dimensional dispersive waves from multi-point
Figure 10b shows the case of marine clay in Cran. measurements. Application to elastic or viscoelastic
The NF P94-261 method gives an acceptable settle bars. J Mech Phys Solids 50, pp. 321–349.
ment prediction while the two others present a non Butterfield, R. & Banerjee, P., 1971. A rigid disc embedded
conservative result. in an elastic half space. Geotechnical Engineering, pp.
When the loading is over 50% of bearing cap 2: 35–52.
acity, the four methods all underestimate the settle El Naggar, M. & Novak, M., 1994. Non-Linear Model for
ment of the real foundation. The calculated curve of Dynamic Axial Pile Response. J Geotech Eng 120, p.
the LSCA method is always farthest from the experi 308–329.
El Naggar, M. & Novak, M., 1996. Nonlinear analysis for
mental mean curves while NF P94-261 method is dynamic lateral pile response. Soil Dyn Earthq Eng 15,
the closest prediction. p. 233–244.
Elhakim, A., 2005. Evaluation of shallow foundation dis
placements using soil small-strain, Atlanta: s.n.
5 CONCLUSIONS Escobar Valencia, E., Benz Navarrete, M., Gourvès, R. &
Breul, P., 2013. Dynamic Cone Penetration Tests in Granu
This paper presents a method to predict bearing cap lar Media: Determination of the tip’s Dynamic Load-
acity of shallow foundations from the cone resistance Penetration Curve. AIP Conf Proc 1542, pp. 389–392.
and dynamic modulus of the newly developed Gourvès, R., 1979. Méthode et dispositif pour la mesure in
situ de caractéristique de déformation des sols. p. 14.
Dynamic Cone Loading Test. A database of tests has Gourvès, R. & Barjot, R., 1995. Le pénétromètre dynamique
been collected to validate the proposed design method léger PANDA. In: 11ème Congrès Européens de Mécani
whose results show a good agreement with the real que des sols et des travaux de fondations, pp 83–88.
foundation loading test results. However, the enrich Holeyman, A., 1988. Modelling of dynamic behaviour at
ment of more data from dynamic penetrometer and the pile base. Proc Third Int Conf Appl Stress Theory to
foundation loading tests in different sites is needed to Piles, p. 174–185.
better define the model factor of this method. Hölscher, P., Brassinga, H., Brown, M. & et al, 2012.
Rapid Load Testing on Piles Interpretation Guidelines,
Boca Raton, Florida: CRC Press Taylor & Francis
ACKNOWLEDGEMENTS Group.
Karlsson, L., Lundberg, B. & Sundin, K., 1989. Experi
This project was carried out within the framework of mental study of a percussive process for rock
collaborative research program Emerg3R grant. fragmentation. Int J Rock Mech Min Sci 26, pp. 45–50.
Middendorp, P., Bermingham, P. & Kuiper, B., 1992. Stat
namic load testing of foundation piles. In: Proceedings
REFERENCES of the Fourth International Conference on Application
of Stress Wave Theory to Piles, pp 581–582.
AFNOR, 2013. Justification of geotechnical work - National Novak, M. & Beredugo, Y., 1972. Vertical vibration of
application standards for the implementation of Eurocode embedded footings. J Soil Mech Foudation Div ASCE
7 - Shallow foundations - NF P94-261. p. 160. 98, p. 1291–1310.
Baud, J.-P. & Gambin, M. P., 2008. Homogenising MPM Omidvar, M., Iskander, M. & Bless, S., 2014. Response of
Tests Curves by Using a Hyperbolic Model. Geotech granular media to rapid penetration. Int J Impact Eng
nical and Geophysical Site Characterization, Proc. 66:, pp. 60–82.
ISC’3, 1–4 April. Omidvar, M., Malioche, J., Bless, S. & Iskander, M., 2015.
Benz Navarrete, M., 2009. Mesures dynamiques lors du Phenomenology of rapid projectile penetration into
battage du pénétromètre Panda 2, s.l.: University of granular soils. Int J Impact Eng 85, p. 146–160.
Clermont-Auvergne. Pan, C., Zhang, R., Luo, H. & Shen, H., 2016. Baseline correc
Benz Navarrete, M., Pierre, B. & Roland, G., 2021. Appli tion of vibration acceleration signals with inconsistent initial
cation of wave equation theory to improve dynamic velocity and displacement. Adv Mech Eng 8, pp. 1–11.
1063
Paquet,J. & Briard, M., 1976. Controle Non Destructif Des Terzaghi, K., 1943. Die Berechung der Durchlaessig
Pieux En Beton. Ann Inst Tech Batim Trav Publics, p. 49–80. keitsziffer des Toneseaus dem Varlauf der
Poulos, H. & Carter, J., 2001. Foundations and retaining hydrodynamischen spannugser schinungeen. Sit
structures - research and practice. Proc Int Conf Soil zungsberichte de Akadennie der wissehsahaften,
Mech Found Eng, p. 80. 132, (Abt., II a).
Rausche, F., 1970. Soil response from dynamic analysis Tran, Q., Benz Navarrete, M. & Breul, P., 2019. Soil
and measurement on piles, s.l.: Case Western Reserve dynamic stiffness and wave velocity measurement
University. through dynamic cone penetrometer and wave analysis.
Skempton, . A. W., 1951. The bearing capacity of clays. Geotech Eng XXI Century Lessons Learn Futur chal
London, Building Research Congress, Div. I. lenges NP López-Acosta al 1–8.
1064
© 2022 the Author(s), ISBN 978-1-032-31259-0
Probabilistic analysis of gyttja undrained strength from CPTU data for

slope stability analysis
S. Rios, L. Sousa & A. Viana da Fonseca
CONSTRUCT-GEO, Faculty of Engineering, University of Porto, Porto, Portugal
P. Milheiro-Oliveira
CMUP, Faculty of Engineering, University of Porto, Porto, Portugal
O. Hededal
COWI A/S, Kongens Lyngby, Denmark
ABSTRACT: The design of an artificial peninsula near Copenhagen led to an extensive experimental cam
paign where a significant number of CPTU tests were performed. The CPTU data presented a significant dis
persion due to material heterogeneity, which motivated a probabilistic analysis of the soft soil undrained
strength. For that purpose, a probabilistic density function of the undrained strength ratio was estimated and
introduced as input in a limit equilibrium slope stability analysis. This enables the evaluation of the probabil
istic density function of the factor of safety, the probability of failure and the reliability index instead of
a single deterministic factor of safety. This information can be used by the designer or project owner to
choose between different solutions based on the required level of risk.
1 INTRODUCTION during its intended life with appropriate degrees of reli

ability and in an economic way sustain all actions and
1.1 Probabilistic analysis in geotechnical influences likely to occur during execution and use,
engineering and remain fit for the use for which it is required”.
As the deterministic scenario may underestimate the
The main concern of an engineer when designing
risk, it is recommended to adopt a reliability-based
a certain structure is its safety. In geotechnical engin
design instead of the usual factor of safety. Basic reli
eering, designers have to deal with the uncertainties
ability measures include probability of failure and reli
related to the geological materials, the variability
ability index. These can be used to verify the margin
existing in nature, and simplified models. The conven
of safety defined as the difference between resistance
tional way of coping with these uncertainties, is to
and load. Reliability index refers to the number of
use a deterministic factor of safety. A factor of safety
standard deviations between the mean safety margin
of 1.5, for example, is often used to account for the
(M) and failure, being defined as:
combination of uncertainties in the ground, in the ana
lysis parameters and the calculation method. There is
the general perception that a design with a safety
factor higher than 1.5 has to be safe. However,
a factor of safety of 1.5 represents a spectrum of fail where SD is the standard deviation of the safety
ure probabilities, which depend on the uncertainties in margin.
the analysis. As shown by Lacasse and Nadim (1996) The reliability index, β, can be related to the fail
a design with a high factor of safety can have higher ure probability (pf) by the expression:
failure probability than another with a lower factor of
safety, if the involved uncertainties are large. This
means that a higher factor of safety does not imply
a smaller risk, because it is affected by the uncertain where � represents the cumulative distribution func
ties in the analysis (Lacasse et al., 2019). tion of the standardised normal distribution.
At present, Eurocodes (Eurocode 0, CEN, 2002) are This work presents an attempt of applying
based in the statement that”a structure shall be a probabilistic approach to describe the undrained
designed and executed in such a way that it will, strength ratio in a slope stability situation.
DOI: 10.1201/9781003308829-161
1065
1.2 Lynetteholmen project
The case study presented in this work concerns the
construction of an artificial peninsula, built outside
the Copenhagen harbor, in an area called Lynette
holm. This artificial peninsula will span over 282
hectares over the North of Refshaleøen and will be
delimited to the north by Nordhavn and to the west
by the Trekroner. The main purpose for its construc
tion is to prevent the city of Copenhagen from the
occurrence of floods and storm surges. In fact, it is
predicted that by the end of 2100, the sea level will
rise between 54 cm (under the “Paris agreement”)
and 74 cm (“business as usual” scenario), according
to the Geological Survey of Denmark and Green
land. Lynetteholm is expected to be completed in
2070. The hope is that the first homes can begin
building in 2035.
The artificial peninsula will be mainly built with
hydraulic fill contained by earth supporting struc
tures such as sheet pile walls and embankments.
Then perimeter design layout will be arranged with
sand and rock in the front to form beaches.
2 SITE CONDITIONS
Figure 1. Gyttja thickness map.
2.1 Site geology
Formerly below the Scandinavian Ice Shield, the
area has now an upper layer of filling material result
ing mainly from dredging, although ashes, slags,
rubble, concrete rubble and other man made compo
nents can be included. Mixed with the fill or immedi
ately below, post-glacial marine deposits are found
consisting mainly in gyttja. On top of the gyttja
layers, slag that was thrown overboard from steam
ships is frequently found. Gyttja is a very soft and
odorous soil that originates from remains of plants
and animals rich in fats and proteins. The material
consists of both organic matter and inorganic (clay,
silt, fine sand, shells or limescale). During its forma
tion, varying quantities of inorganic material are
mixed with the organic and when examined, a lower
loss of filament and water content is found which is
common for peat (Larsson, 1990).
2.2 Geotechnical data

COWI carried out extensive geotechnical studies in
order to obtain factual geological, geotechnical and
hydrogeological data for the area. The geotechnical
studies included 301 CPTU tests and 191 boreholes.
Based on boreholes and CPTU, a thickness map was
developed (Figure 1) indicating the thickness of soft
postglacial deposits, including gyttja, peat and clay.
The thickness of the organic post-glacial deposits is
greatest in the former meltwater valleys, where up to
6-9 m thick layers of gyttja and peat were taken. Figure 2. Corrected cone resistance (qt) evolution with
CPTU data (Figure 2) revealed low values of the depth for all CPTU data.
corrected cone resistance (qt) defined as follows:
1066
qt ¼ qc þ u2 ð1 - aÞ ð3Þ Table 1. Undrained strength obtained in the triaxial tests.
where qc is the tip cone resistance, u2 is the pore Triaxial N0. CU; TC (kPa)
pressure measured behind the cone, and a is the area
BH_P_34_1 4.2 12.1 0.35
factor calculated as the ratio of the cross sectional
BH_P_34_2 14.3 41.0 0.35
area of the load cell, An, by the projected area of the
BH_P_34_3 58.5 186.3 0.31
cone, Ac, (a=An/Ac). In this case, two different cones
BH_P_33_1 16.3 28.3 0.58
were used for the CPTU campaign. A number of 193
BH_P_33_2 21.8 58.3 0.37
tests were conducted by a piezocone with a=0.6 and
BH_P_33_3 60.5 205.4 0.29
106 tests with a=0.8.
BH_P_25_3.1 6.5 12.0 0.54
Figure 2 plots qt values with depth, ranging in the
BH_P_25_3.2 16.5 44.0 0.38
interval of 0 to 0.6 MPa, where a significant disper
BH_P_25_3.3 60.5 185.0 0.33
sion is observed. To better understand the type of
BH_P_25_8.1 16.0 33.0 0.48
material present, the results were analysed by the
BH_P_25_8.2 84.0 217.0 0.39
unified approach proposed by Robertson (2009)
BH_P_25_8.3 63.5 214.0 0.30
which have indicated a clayey/silty behavior with no
BH_P_84_1 13.5 29.3 0.46
relevant sensitivity (Figure 3)
BH_P_84_2 22.0 44.0 0.50
BH_P_84_3 47.5 195.0 0.24
3.2 Undrained strength from empirical

correlations
Mayne and Peuchen (2018) have proposed an empir
ical correlation to obtain the parameter Nkt from the
Bq parameter of the CPTU interpretation (equation 5)
Bq being given by equation (6) where qnet is the net

cone resistance (qnet = qt – σvo),
Nkt can then be used to calculate the undrained

strength using the following well-known relation:
Figure 3. CPTU data plotted in the Robertson (2009) soil

behavior chart.
Considering only the CPTU tests performed near
3 UNDRAINED STRENGTH the boreholes where the samples for the triaxial tests
were retrieved and the CPTU data at the depths at
3.1 Laboratory data which those samples were taken, the following
Based on 5 boreholes, 15 valid anisotropically consoli results were obtained (Table 2).
dated undrained triaxial compression tests (CAUC) It is clear from this data that much higher
were conducted to determine the gyttja undrained undrained strength ratios ðCu =σ0 v0 Þ were obtained
shear strength (Table 1). from the CPTU data interpreted with Mayne and
From that data summarized in Table 1, the Peuchen (2018) correlation (Table 2) in comparison
evaluation of the undrained shear strength (CU,TC) with the laboratory data presented in Table 1. These
was plotted against the vertical effective stress values are also very high taking into account that the
after the consolidation, σ0v , obtaining the following soil under study concerns a normally consolidated
linear regression: material where an undrained strength ratio between
0.2 and 0.4 was expected. This might be explained
by two main reasons: i) this empirical correlation
has not been developed for gyttja; ii) the site is very
1067
Table 2. Undrained strength obtained in the CPTU tests
by the Mayne and Peuchen (2018) correlation.
Nkt Cu (kPa) Cu =σ0v0
14.3 9.5 0.91

15.6 8.3 0.77
15.7 8.3 0.86
13.3 17.8 0.80
13.9 19.5 0.90
13.5 17.8 0.82
17.5 7.6 0.86
15.6 9.4 0.88
14.8 10.9 0.93
13.3 18.2 0.66
13.1 15.4 0.59
13.8 15.9 0.60
11.7 23.7 1.04
12.0 21.4 1.02 Figure 4. Histogram of all the undrained strength ratio values.
12.2 23.3 1.05
heterogeneity of the material and the presence of man-
made materials, namely slag, mixed with the soft soil.
heterogeneous where it is possible that gyttja is
mixed with other man-made materials.
4 PROBABILISTIC ANALYSIS
3.3 Proposed approach
For the data presented in Figure 4, a probabilistic
In these conditions, an average Nkt value was adopted density function was fitted using a Lognormal distri
considering only the five CAUC tests performed with bution (equation (8)),
lower effective stresses (<60 kPa) considered more
representative of the effective stresses acting at gyttja
according to Figure 1. To avoid differences between
effective stresses of CPTU tests and the effective
stresses of CAUC test, the undrained strength coming
from CAUC tests was calculated with equation (4) where σ is the Lognormal standard deviation (or
for the depth at which the CPTU data was taken. The shape parameter), m is the Lognormal mean (or scale
results are presented in Table 3 from where an aver parameter) and θ is the offset of the Lognormal mean
age Nkt value of 19.2 was taken (Sousa, 2020). (or the location parameter). This case, the following
estimates of the parameters were obtained to adjust
the histogram of Figure 4: σ=0.852, m=0.720 and
θ=0.04, as presented in Figure 5.
Table 3. Nkt values obtained in 5 CAUC tests.
cu,TC avg qnet,avg

Borehole (kPa) CPTU (MPa) Nkt
7.7 0.133 17.2

BH_P_25 CPT_A_235
12.5 0.208 16.6
BH_P_33 11.2 CPT_A_193 0.237 21.2
BH_P_34 8.1 CPT_A_179 0.135 16.7
BH_P_84 11.3 CPT_P_84 0.277 24.5
Then, the undrained strength obtained in all the 80

000 CPTU measurements was calculated with this Nkt
value using equation (7). The results are plotted in
Figure 4, which represents the histogram of all the
undrained strength ratio values.
It is clear that, even using a quite high Nkt value,
there are still a significant number of undrained
strength ratios above 0.4. This may be associated to the Figure 5. Lognormal fit to the undrained strength ratio.
1068
5 SLOPE STABILITY ANALYSIS 6 CONCLUSIONS
The probabilistic distribution of the undrained Geotechnical engineering design is progressively

strength as described above was included in a slope moving from a deterministic approach to a prob
stability analysis using the limit equilibrium software abilistic approach. This enables the consideration
SLOPE/W from GEOSTUDIO®. The aim of this of natural variability and the uncertainties associ
work was to analyze the gyttja stability during its ated to geotechnical parameters. It also enables to
excavation. The retaining support structures built in quantify the probability of failure and the reliabil
the perimeter of the area to contain the hydraulic fill, ity index associated to a certain design, which can
will be founded in the layers below gyttja. For this be used by designers or project owners to decide
reason, the gyttja will be excavated in those areas. upon different options depending on the required
Slope stability analysis are needed because submar level of risk. In this work, the large amount of
ine landslides caused by failures in the gyttja layer CPTU data performed for the construction of an
can cause tsunamis and/or disturbances in the sub artificial peninsula close to Copenhagen was used
marine environment with impact on its biodiversity. to perform a probabilistic analysis of the soft soil
In these conditions, limit equilibrium analysis undrained strength ratio. The probabilistic density
using the Morgenstern-Price method were performed function of this variable was then introduced in
considering the gyttja undrained strength as a a slope stability analysis by the limit equilibrium
random variable whose distribution is the one indi method. The results indicated that for a slope of
cated in equation (8). The probabilistic procedure 9 m high and 45º of inclination, the reliability
used herein was the Monte Carlo method, integrated index associated to the probability of failure was
in SLOPE/W, considering a total amount of 500 MC 1.45, which is quite low. This means that although
samples. This was applied to a slope of 9 m high and a deterministic factor of safety of 1.4 was obtained,
45º of inclination, as illustrated in Figure 6. the reliability associated with the studied design
is low.
ACKNOWLEDGMENTS
This work was financially supported by: Base Fund

ing - UIDB/04708/2020 of the CONSTRUCT - Insti
tuto de I&D em Estruturas e Construções - funded
by national funds through the FCT/MCTES
(PIDDAC) and by UTD/MAT/00144/2019. It also
received European Commission funds from the
ERASMUS program. The authors express gratitude
Figure 6. Slope used in the gyttja stability analysis.
to By og Havn A/S for permission to use geotech
nical data from the Lynetteholmen project.
The results of the probabilistic distribution of the
factor of safety (FS) are indicated in Figure 7, where
a minimum FS of 1.43 and a maximum FS of 9.40 REFERENCES
were obtained with an average of 2.64. In addition,
a reliability index of 1.45 was obtained for this case. Casagrande, A. (1965). The role of the ‘calculated risk’ in
earthwork and foundation engineering. 91.
CEN (2002). Eurocode 0 - basis of structural design. Euro
pean Committee for Standardization, Brussels
Khan, F., & Malik, A. (2013). Probability and sensitivity
analysis of the slope stability of Naulong dam. Pakistan
Journal of Engineering and Applied Sciences, vol 13,
54–64.
Lacasse, S., Nadim, F. 1996. Uncertainties in characterizing
soil properties. Uncertainty in the Geologic Environ
ment: From Theory to Practice (Uncertainty ‘96), ASCE
GSP 158 (49–75).
Lacasse, S., Nadim, F., Liu, Z.Q., Eidsvig, T.M.H., Le, T.
M.H., Lin, C.G. (2019). Risk assessment and dams –
Recent developments and applications. Proceedings of
the XVII ECSMGE- Geotechnical Engineering founda
tion of the future, Reykjavik, Iceland
Larsson, R. (1990). Behavior of organic clay and gyttja.
Rapport-Statens geotekniska institut, 38.
Figure 7. Probabilistic density function of the factor of Mayne, P., & Peuchen, J. (2018). Evaluation of CPTU Nkt
safety. cone factor for undrained strength of clays. Proceedings
1069
of the Cone Penetration Testing Conference, Delft, Robertson, P.K. (2009). Interpretation of cone penetration tests
Netherlands, 423–429. —a unified approach. Can. Geotech. J., 46(11), 1337–1355
Peck, R. B. (1969). Advantages and Limitations of the Sousa, L. (2020). Probabilistic undrained strength evalu-
Observational Method in Applied Soil Mechanics. ation of soft soil for slope stability design. MSc Thesis
Géotechnique, 19(2), 171–187. doi:10.1680/geot.1969. presented to the Faculty of Engineering of University of
19.2.171 Porto, Portugal
1070
© 2022 the Author(s), ISBN 978-1-032-31259-0
Monitoring ground improvement using in situ tests in Guayaquil, Ecuador

F. Ripalda, D. Falquez & D. Besenzon
ESPOL Polytechnic University, Escuela Superior Politécnica del Litoral, ESPOL, Department of Earth Sciences,
Guayaquil, Ecuador
R. Luque
Geosísmica, Guayaquil, Ecuador
F. Illingworth
Subterra, Guayaquil, Ecuador
S. Amoroso
University of Chieti Pescara, Pescara, Italy
Istituto Nazionale di Geofisica e Vulcanologia, Italy
ABSTRACT: The present work describes the use of the seismic dilatometer test (SDMT) and the piezocone
test (CPTu), to assess the effects of ground improvement at a wastewater treatment plant in Guayaquil, Ecua
dor. The ground improvement consisted of 15 m-long, 0.55 m-diameter and 2 m-spacing stone columns built
with vibro-replacement technique. The tests were carried out both in natural and in treated soils to compare
the variation of the geotechnical parameters in the analyzed deposits. The results show specific sensitivity of
the DMT over the CPTu tests to the ground improvement into the layer composed of sands and sandy silts,
while VS values show a limited increase in the treated area.
1 INTRODUCTION resistance qc1N, using vibro-replacement stone col

umns in combination with vertical drains in deposits
Ground improvement involves different techniques to with FC < 65%. Mitchell & Wentz (1991) showed
modify soil response under different conditions. a 100% increase for the corrected cone resistance for
Ground modification performance is based on assess overburden stress, (qc1), and a 45% increase for the
ing problematic soils, liquefaction potential, soil SPT corrected penetration resistance, (N1)60, when
instability, insufficient bearing capacity and excessive comparing pre and post-treatment results in soil
settlement, seepage. U.S. Army Corps of Engineers layers with FC < 55%. Vibro-replacement stone col
(1999). umns installation may have the double beneficial
Mitchell (2008) discussed the applications and effect to cause densification of the surrounding soil
limitations of these densification methods, and the during installation and facilitate the dissipation of the
author noted that the degree of improvement given by excess of pore water pressure developed during an
the deep dynamic compaction, vibro-compaction and earthquake, by providing a shorter path of drainage,
blasting is greater in clean sands since it decreases as Adalier & Elgamal (2004).
the fines content (FC) increases. Nevertheless, several Therefore, the effectiveness verification of the
studies have documented mitigation works using vari improvement using in situ tests becomes relevant
ous FC values (including rather high percentages), since these investigations allow a quick assessment,
highlighting an improvement given by the vibro which compares selected geotechnical parameters
replacement stone columns, Mackiewicz & Camp obtained before and after the treatment. Numerous
(2007) used an improvement index (Ii), given by the authors (Schmertmann 1986, Mackiewicz & Camp
ratio between the cone resistance (qc) after and before 2007, Mitchell 2008, Monaco et al. 2014, Bała
the treatment minus one, to provide an improvement chowski & Kurek 2015, Wotherspoon et al. 2015,
of 0.3 < Ii < 2.8 for FC < 5%, and of 0 < Ii < 1.6 for Massarsch & Fellenius 2019, Massarsch et al. 2020)
15% < FC < 40%. (Luehring et al. 2001) showed an evaluate the change of the soil characteristics using
increase of 95% for the corrected SPT blow count different in situ tests and their parameters: SPT blow
(N1)60, and 180% for the normalized corrected cone count NSPT in the standard penetration test (SPT),
DOI: 10.1201/9781003308829-162
1071
horizontal stress index KD and constrained modulus
M in the flat dilatometer test (DMT), corrected cone
resistance qt in the piezocone penetrometer test
(CPTu). Other studies (e.g., Wotherspoon et al.
where σ0v0 is the vertical effective stress.
2015, Hwang et al. 2017, Comina et al. 2021) have
To estimate OCR in sands the approximation by
applied shear wave velocity VS in the geophysical
Monaco et al. (2014) was used:
measurements provided by invasive or non-invasive
tests (e.g., seismic piezocone SCPTu, seismic dilat
ometer SDMT, down-hole DH, cross-hole CH,
multichannel analysis of surface waves MASW).
Moreover, several research discusses the change in
the at-rest lateral earth pressure coefficient K0, the 3 IN SITU TESTS
overconsolidation ratio OCR and the ratio M/qt
when monitoring the densification effectiveness and The results presented in this study belong to a trial
the lateral stress increase. A combination of CPT and site located within a wastewater treatment plant
DMT tests is performed to estimate the parameters (WTP) in Guayaquil, Ecuador. Figure 1, shows the
mentioned above, as suggested in previous studies location where CPTu and SDMT tests were per
(e.g., Baldi et al. 1986, Marchetti et al. 2001, Amor formed before the stone columns (SC) installation,
oso et al. 2018, 2020, Massarsch et al. 2020) natural soil (NS), and after SC installation, treated
The present study describes the effects of ground soil (TS), up to 16-20 m depth. NS soil testing is
improvement using SDMT and CPTu tests. In this identified as CPTu1_NS and SDMT1_NS, while sur
regard, CPTu and SDMT tests and VS measurements veys after SC installation are detected as CPTu2_TS
were executed in natural and treated soils, to com and SDMT2_TS. The SC were in a staggered
pare the geotechnical parameters, to assess liquefac arrangement with 2 m spacing. Additional informa
tion before and after treatment. tion regarding the NS condition was obtained from
the borehole, SPTs (SPTP3_NS) and CPTu tests
(CPTu14_NS) performed during the WTP
2 COMBINATION OF SDMT AND CPTU FOR construction.
MONITORING GROUND IMPROVEMENT
Single-parameters derived from SDMT and CPTu

tests can detect the modification in soil characteris
tics due to improvement works. As stated by various
authors (e.g. Schmertmann 1986, Bałachowski &
Kurek 2015, Amoroso et al. 2018, 2020, Massarsch
& Fellenius 2019, Massarsch et al. 2020), these
parameters can be identified in the horizontal stress
index KD and the constrained modulus M from
DMT, the corrected cone resistance qt (or the cone
resistance qc) and the relative density DR from CPT.
KD is directly derived from the corrected DMT
membrane lift-off pressure reading and contains
information about the stress history of the soil. Con
currently M is a function of the three DMT inter
mediate parameters (horizontal stress index KD,
dilatometer modulus ED and material index ID). qt
(or qc) is a direct measurement from CPT, while DR
is usually based on correlations as a function of the
cone resistance and effective stress, Juang et al.
(1996). According to previous ground improvement
studies related to densification techniques (e.g. Mas
sarsch & Fellenius 2002, 2019, Massarsch et al.
2019), the horizontal stress also increases after com
paction, making KD (and therefore M) more sensi
tive than qt (and consequently DR) to detect the
modifications induced by the treatment. Figure 1. Location of the water treatment plant at “Las
Moreover, the coupled CPT-DMT parameters, Esclusas”, in-situ tests and stone columns.
such as K0, OCR, can help identify the treatment
effectiveness in sandy soils. The present research For the execution of CPTus and SDMTs, the shal
estimated K0 using the more recent relationship pro low compacted fill layer (≈ 0.6-0.8 m thick) was
posed by Hossain & Andrus (2016): removed, to prevent damage to the geotechnical
1072
equipment. Table 1 summarizes the basic informa the silty sand to sandy silt layer (Ic < 2.6) is located.
tion of the in-situ tests used for verifying the ground However, for some depth intervals between 4 and
improvement effectiveness. The ground-water table 10 m, qt (≈ 4.2-6 m, 8-9 m, 9.6-10.4 m depth), and
(GWT) fluctuations, at the trial site, were strongly DR (≈ 4.5-6 m, 8-9 m depth) values in the NS are
influenced by the Guayas river tide, following the somewhat higher than in TS. This rise is observed
Navy Oceanographic and Antarctic Institute meas when the Ic increases in the TS, and it behaves more
urements, INOCAR (2021). like fine-grained soil. Figure 3 also compares the
CPTu-DR values with the ones evaluated from SPT,
Skempton (1986). The DR SPT-based values in the
NS are in good agreement with the related CPTu
Table 1. Summary information of the in situ tests at the ones from ≈ 6 to 8 m depth, while between 8 and
trial site. 11 m depth, the SPT-based method overpredicts the
relative density. The SPT-based overestimations of
Field test Depth (m) GWT depth* (m) DR can be attributed to the lens of silt mixtures
detected only by CPTu and SDMT.
SPTP3_NS 19.0 2.0
The DMT parameters, were calculated using the
CPTu14_NS 20.8 2.7
Marchetti et al. (2001) formulae. The equilibrium
SDMT1_NS 20.4 3.4
pore pressure, u0, obtained from the third DMT pres
SDMT2_TS 20.6 3.4
sure reading (p2) into the sandy layers, well deter
CPTu1_NS 17.6 3.8
mined the GWT location. The effectiveness of the
CPTu2_TS 19.0 3.8
treatment is noticeable from ≈ 2 to 6 m depth, by
* Note: Measured from the ground surface post filling. looking at KD and M and profiles (Figures 2, 3); in
this depth range, ID > 0.6 predominate in both soil
conditions. The increase in KD profile is clearly
defined in this depth interval and a 52% increment is
observed after the treatment. The shear wave vel
3.1 Geotechnical description
ocity VS (Figure 3) also provides some increase after
Figure 2 summarize the soil profile in both soil condi improvement, but limited between 4 and 6 m. A spe
tions using CPTu and SDMT tests at the Guayaquil cific lateral soil heterogeneity is distinguishable in
trial site. Beneath the shallow fill, the soil is variable, the NS and TS, ID profiles between ≈ 6 to 8 m depth:
but four clearly defined layers can be observed. The the TS exhibits a fine-grained soil behavior, consid
first layer is approximately 2 m thick and varies from ering the lower ID values (0.3 < ID < 1.2 correspond
silt to clay, as described by: the soil behavior index ing to silty clay to silt), while the NS of the same
(Ic) profile that intercalates between 2.6 and 3.4, the layer results mostly silty-sandy (1.2 < ID < 2.3). This
material index (ID) values which are between 0.2 and response helps to understand why for the same depth
1.1. Underlying this layer, loose to medium dense interval, the horizontal stress index KD and the con
sand mixtures (2MPa < qt < 8 MPa; 2 < KD < 9) are strained modulus M are much lower despite the SC
present with a maximum depth of ≈ 10 m. These non- installation. The analysis of CPTu and DMT com
plastic sands and silty sands are mainly characterized bined parameters is displayed in Figure 3 to monitor
by Ic < 2.6 and ID > 1.2. A lens of variable thickness, ground improvement effectiveness. The OCR and K0
comprised of silt mixtures (2.6 < Ic < 3.0, 0.6 < ID < estimations were performed both in fine-grained and
1.1) is present within the sandy layer between ≈ 7 and incoherent soils. Specifically, for ID < 1.2, OCR and
10 m depth. Finally, below 10-11 m depth, normally K0 were estimated by DMT using Marchetti et al.
to moderately overconsolidated clays are encountered, (2001) formulae, while for sandy layers (Ic < 2.6 and
according to OCR approximation by Marchetti ID > 1.2) the combined CPT-DMT approach was
et al. (2001). This finding associates the following used according to Equation 2 from Monaco et al.
DMT and CPTu parameters: 2.2 < KD < 3.3, with (2014) for OCR and to Equation 1 from Hossain &
3.1 < Ic < 3.9, 0.2 < qt < 2.0 and 0.2 < ID < 0.6. Andrus (2016) for K0. The OCR and K0
profiles detect the effectiveness of the SC treatment
between ≈ 2.6 and 6.6 m depth. Below 6.6 m, the NS
4 GROUND IMPROVEMENT and TS trend remains unchanged despite the SC
installation.
Figures 2 and 3 show the variation of the CPTu, Table 2 summarizes the average test results of
DMT and combined (CPTu and SDMT) parameters the single and combined parameters in the layer
in natural soil (NS) and treated soil (TS), estimated where the increase was better noticed for Ic < 2.6
according to Robertson & Cabal (2015). The relative and ID > 1.2, approximately between 3.2 and
density (DR) estimation is based on the correlations 6.6 m depth. The improvement was calculated by
proposed by Kulhawy & Mayne (1990). Ic profiles relating the difference between TS and NS to NS
present a very slight variability of the soil before and results, expressed as a percentage. The CPTu con
after treatment, which makes quite comparable the ventional indicators of improvement show an
data within the depth of the SC improvement where increment of 6% for qt and 7% for DR, while for
1073
Figure 2. CPTu and SDMT basic parameters in both soil conditions (natural and treated soil) at the Guayaquil trial site.
Figure 3. CPTu and SDMT estimated parameters in terms of DR, VS, M, OCR and K0 at the Guayaquil trial site.
Table 2. Summary of average parameters (pre and post 5 CONCLUSIONS

treatment) between 3.2 and 6.6 m depth: qt, DR, KD, M, VS,
OCR, K0. Despite the length of the SCs, the effectiveness of
the treatment resulted noticeable only between 3.2 to
qt DR VS 6.6 m depth, where the sand mixtures where detected
(MPa) (%) KD M (Mpa) (m/s) OCR K0
by in situ tests. Below this layer a lens of silt mix
NS 5.0 42.5 5.0 43.6 121.2 3.1 1.3 tures, with higher FC (up to 46%) approximately
TS 5.3 45.7 6.2 61.9 152.8 6.1 1.5 between 7 to 10 m depth, and of a cohesive soil
Inc. 6.0 7.6 22.4 42.0 26.1 98.4 15.8 layer, from 10-11 m depth, were identified.
(%) The evaluation of the soil improvement between
3.2 to 6.6 m depth was mainly detected by using the
combined CPTu and SDMT parameters, with a 98 %
increment in OCR and 15% increment in K0. The
the SDMT parameters, KD increased 22%, M 42% relatively low increment in K0 can be attributed to
and VS 26%. For the combined CPTu and SDMT the high initial K0 condition in NS (K0 ≈ 1.27), as
parameters, K0 increased just 16%, while OCR already noticed by Schmertmann (1985). In the
increased 98% CPTu based effectiveness assessment, qt and DR
1074
have a similar increase (6% and 7.5% respectively), inocar.mil.ec/web/index.php/productos/tabla-mareas#
although the NS and TS were related to quite homo busqueda-de-datos-de-mareas (accesed 2021 Mar 22).
geneous subsoil, as detectable looking at Ic. Further Juang, C., Huang, X., Holtz, R. & Chen, J. (1996) ‘Deter
more, SDMT single parameters, KD, M, VS, mining relative density of sands from CPT using fuzzy
provided a more evident SC improvement, even still sets’, Journal of Geotechnical Engineering, 122(1),
limited, 22%, 42% and 26% respectively. Therefore, pp. 1–6.
at the Guayaquil trial site the densification provided Kulhawy, F. H. & Mayne, P. W. (1990) Manual on estimat
ing soil properties for foundation design. (No. EPRI-EL
by the SCs resulted merely perceived by the CPTu -6800). Electric Power Research Inst., Palo Alto, CA
tests probably also due to the lateral soil variability. (USA); Cornell Univ., Ithaca, NY (USA). Geotechnical
Engineering Group.
ACKNOWLEDGEMENT Luehring, R., Snorteland, N., Stevens, M. & Mejia, L.
(2001) Liquefaction Mitigation of a Silty Dam Founda
tion Using Vibro-Stone Columns and Drainage Wicks:
Special thanks to Studio Prof. Marchetti (Italy) for A Case History at Salmon Lake Dam. Water Oper.
kindly providing the SDMT apparatus. Manage. Bull. (198): 1–15
Special thanks also to Hidalgo e Hidalgo S.A. for Mackiewicz, S. M. & Camp, W. M. (2007) ‘Ground Modi
sharing information for the present research. fication: How Much Improvement?’ Geo-Denver 2007,
Denver, Colorado, United States, p.9.
Marchetti, S., Monaco, P., Totani, G. & Calabrese, M.
REFERENCES (2001) The Flat Dilatometer Test (DMT) in Soil Investi
gations— A Report by the ISSMGE Committee TC16.
Adalier, K. & Elgamal, A. (2004) ‘Mitigation of liquefac Proceedings of In Situ 2001, International Conference
tion and associated ground deformations by stone on In Situ Measurement of Soil Properties, Bali, Indo
columns’, Engineering Geology, 72(3–4), pp. 275–291. nesia. ISSMGE, London, UK, p.42.
Amoroso, S., Rollins, K., Monaco, P., Holtrigter, M. & Massarsch, K. & Fellenius, B. H. (2019) ‘Evaluation of
Thorp, A. (2018) ‘Monitoring ground improvement vibratory compaction by in situ tests’, ASCE Journal of
using the seismic dilatometer in Christchurch, New Geotechnical and Geoenvironmental Engineering, 145
Zealand’, Geotechnical Testing Journal, 41(5), pp. (12), pp. 1–15.
946–966. Massarsch, K., Wersäll, C. & Fellenius, B.H. (2019) ‘Hori
Amoroso, S., Rollings, K., Andersen, P., Gottardi, G., zontal stress increase induced by deep vibratory
Tonni, L., García Martínez, M., Wissman, K., et al. compaction’, in Proceedings of the Institution of Civil
(2020) ‘Blast-induced liquefaction in silty sands for Engineers: Geotechnical Engineering, pp. 1–26.
full-scale testing of ground improvement methods: Massarsch, K., Wersäll, C., Fellenius, B.H.,
Insights from a multidisciplinary study’, Engineering Bałachowski, L., Kurek, N. & Konkol, J. (2020) ‘Dis
Geology, 265, p. 105437. cussion: Horizontal stress increase induced by deep
Bałachowski, L. & Kurek, N. (2015) ‘Vibroflotation Con vibratory compaction’, Proceedings of the Institution of
trol of Sandy Soils Using DMT and CPTU’, in The 3rd Civil Engineers: Geotechnical Engineering, 173(4),
International Conference on the Flat Dilatometer, pp. pp. 370–375.
185–190. Massarsch, K. & Fellenius, B. H. (2002) ‘Vibratory com
Baldi, G., Belloti, R., Ghioma, V., Jamiolkowski, M., paction of coarse-grained soils’, Canadian Geotechnical
Marchetti, S. & Pasqualini, E. (1986) ‘Flat Dilatometer Journal, 39(3), pp. 695–709.
Tests in Calibration Chambers’, in Proc. In Situ ’86 Mitchell, J. K. (2008) ‘Mitigation of liquefaction potential
ASCE Spec. Conf. on Use of In Situ Tests in Geotechnical of silty sands’, in Symposium Honoring Dr. John
Engineering. Virginia Tech, Blacksburg, pp. 431–446. H. Schmertmann for His Contributions to Civil Engin
Comina, C., Mandrone, G., Arato, A., Chicco, J., Duò, E. eering at Research to Practice in Geotechnical Engineer
& Vacha, D. (2021) ‘Preliminary Analyses of an Innova ing. New Orleans, Louisiana, United States, pp.
tive Soil Improving System by Sand/Gravel Injections– 433–451.
Geotechnical and Geophysical Characterization of Mitchell, J. K. & Wentz, F. J. (1991) Performance of
a First Test Site’, Engineering Geology, 293, p. 106278. Improved ground. During the Loma Prieta Earth
Hossain, A. M. & Andrus, R. D. (2016) ‘At-Rest Lateral quake, UCB/EERC-91/12, Earthquake Engineering
Stress Coefficient in Sands from Common Field Research Center, University of California, Berkeley,
Methods’, Journal of Geotechnical and Geoenvironmen p. 100.
tal Engineering, 14(12), pp. 06016016–1–06016016–5. Monaco, P., Amoroso, S., Marchetti, S., Marchetti, D.,
Hwang, S., Roberts, J., Stokoe, K., Cox, B., van Totani, G., Cola, S. & Simonini, P. (2014) ‘Overconsoli
Ballegooy, S. & Soutar, C. (2017) ‘Utilizing dation and Stiffness of Venice Lagoon Sands and Silts
Direct-Push Crosshole Seismic Testing to Verify the from SDMT and CPTU’, ASCE Journal of Geotechnical
Effectiveness of Shallow Ground Improvements: and Geoenvironmental Engineering, 140, pp. 215–227.
A Case Study Involving Low-Mobility Grout Columns Robertson, P. & Cabal, K. (2015) Guide to Cone Penetra
in Christchurch, New Zealand’, in Proceedings of tion Testing for Geotechnical Engineering. 6th edn.
Grouting 2017. Honolulu, Hawaii, pp. 415–424. Signal Hill, CA: Gregg Drilling & Testing.
INOCAR (2021) Instituto Oceanográfico y Antártico de la Schmertmann, J., Baker, W., Gupta, R. & Kessler, K.
Armada - Tabla de mareas puertos del Ecuador (1986) ‘CPT/DMT quality control of ground modifica
(Oceanographic and Antarctic Institute of the Navy tion at a power plant’, in Specialty Conference-In Situ
Table of tides of ports of Ecuador) . See https://www. ’86. Blacksburg, VA, pp. 985–1001.
1075
Schmertmann, J. H. (1985) Measure and Use of the In Situ 1110-1-185. Available at: http://usacetechnicalletters.
Lateral Stress, Practice of Foundation Engineering. tpub.com/ETL-1110-1-185/ETL-1110-1-1850002.htm
A Volume Honoring Jorj O. Osterberg. The Department (Accessed: 6 October 2020).
of Civil Engineering, Northwstern University, pp. Wotherspoon, L., Cox, B., Stokoe II, K., Ashfield, D. &
189–213. Phillips, R. (2015) ‘Utilizing Direct-Push Crosshole
Skempton, A. W. (1986) ‘Standard penetratrion test proced Testing to Assess the Effectiveness of Soil Stiffening
ures and the effects in sands of overburden pressure, Caused by Installation of Stone Columns and
relative density, particle size, ageing and Rammed Aggregate Piers’, in Proceedings of the
overconsolidation’, Géotechnique, 36(3), pp. 425–447. 6th International Conference on Earthquake
U.S. Army Corps of Engineers (1999) Guidelines on Geotechnical Engineering - 6ICEGE. Christchurch,
Ground Improvement for Structures and Facilities, ETL New Zealand.
1076
© 2022 the Author(s), ISBN 978-1-032-31259-0
Plate anchor capacity estimation through CPT tip resistance in sand

A. Roy & S.H. Chow
Department of Infrastructure Engineering, the University of Melbourne, Australia
ABSTRACT: Reliable estimation of plate anchor uplift capacity in sand through analytical and empirical
equations is often complicated due to uncertainties in estimation of soil properties required in the equations.
In order to address this uncertainty, this study proposes a correlation to estimate plate anchor vertical uplift
capacity in sand based on cone tip resistance measured from cone penetrometer tests (CPT). The correlation
was established using a database of reported centrifuge experiments on circular, rectangular and strip anchors
in loose and dense silica sand at various embedment depths and g-levels, along with the corresponding centri
fuge CPTs performed in the same testing boxes. The centrifuge cone tip resistances were also depth-corrected
to remove the effect of shallow embedment. Through regression analyses, the correlation between plate
anchor capacity and cone tip resistance in dimensionless form was developed, with different coefficients fitted
for circular, rectangular and strip anchors respectively.
1 INTRODUCTION Pucker et al. 2013) and pile foundations (Schneider et al.

2010) over the last two decades, but is yet to be explored
Plate anchors are routinely used onshore as foundations in uplift capacity problems.
for structures subjected to high uplift or lateral loads. This paper explores the test results from an elab
With a recent global thrust in harnessing energy from orate set of reported centrifuge experiments (vertical
cleaner renewable energy sources, plate anchors are also uplift tests) on circular, rectangular and strip anchors
likely to find increased application in mooring offshore with companion cone penetrometer tests in different
renewable energy devices owing to their cost- and cap densities at various embedment depths, g-levels and
acity-effectiveness relative to other anchors and piles. OCR levels. The cone tip resistance (qc) measured in
Such renewable energy devices are likely to be located the centrifuge were depth-corrected to remove the
in much shallower water where the seabeds of interest effect of shallow embedment. A correlation between
include coarser-grained deposits. Hence, reliable estima plate anchor capacity and cone tip resistance is then
tion of plate anchor uplift capacity in coarser-grained developed using a regression analysis on the data
soils will be of interest for geotechnical practitioners. base of centrifuge anchor tests.
One of the most widely used soil exploration methods
in coarser-grained offshore deposits is cone penetration
tests (CPT). As undisturbed core sample collection for 2 EXPERIMENTAL DATABASE DETAILS
coarser-grained soils requires specialised equipment and
becomes expensive for offshore sites, CPT data is often The anchor database comprised of 91 uplift tests
the only source of geotechnical information available across 20 sand samples reported from two different
during the early planning stages for offshore projects. centrifuge studies:
Considering that behaviour in sands is highly dependent • Tests reported by Roy et al. (2021a,b) on strip, rect
on initial fabric, stress levels and relative density (RD) angular and circular plate anchors in silica sand of
(Been et al. 1991, Gajo & Wood 1999), a continuous relative density, RD ~ 70% and RD ~ 45% (details in
CPT profile can macroscopically capture such effects Table 1 and Table 2). These tests were conducted at
with respect to the change in soil stratigraphy. As com gravitation accelerations ranging between 20g and
pared to a traditional ‘operational’ friction angle (φ) 100g, and embedment ratios (H/B or H/D, where
based approach, there is great value in direct application H is depth of the plate anchor, B is the anchor width
of cone tip resistance (qc) profiles in design approaches, and D is the anchor diameter) of 2 to 6 to investi
because it will significantly improve reliability in predic gate the behaviour of plate anchors under different
tion by eliminating uncertainties in estimation of a back stress levels (hence at different g) and load inclin
analysed φ. Such qc based design approaches have been ation. For the current study, only vertical uplift tests
reported for bearing capacity problems on footings and from the respective samples are included in the
spudcans (Lee & Salgado, 2005, Liu & Lehane 2020, database.
DOI: 10.1201/9781003308829-163
1077
• Tests reported by Hao et al. (2018) on circular (R) with aspect ratio 2:1, circle (C), helix (H)) as listed
plates and single helices in sand of RD = 85-95% in Tables 1-3 and Figure 1. All these tests were con
at 20g (details in Table 3) at anchor embedment ducted in a fine sub-angular silica sand, commonly iden
ratios (H/D) in the range 2 to 10; tified as the UWA Silica sand with properties as
summarised in Table 4. The samples for the centrifuge
tests were prepared using air pluviation technique and
anchor testing setup can be found in the respective
Table 1. Anchor database in loose sand (after Roy et al. papers.
2021a).
Sample No., RD Anchor Table 2. Anchor database in loose sand (after Roy et al.
(%) (and Anchor H/B or factor, 2021 a,b).
γðkN=m3 Þ code g OCR H/D Nγ
Sample No., RD Anchor
ST3 30 1 1.95 2.28 (%) (and Anchor H/B or factor,
ST2 30 1 1.95 2.14 γðkN=m3 Þ code g OCR H/D Nγ
D1, 76.72 (16.61) ST1 50 1 1.95 2.18
R2 30 1 2.00 3.49 ST2 30 1 1.75 1.31
R3 50 1 2.00 3.37 ST3 50 1 1.75 1.33
R1 100 1 2.05 3.38 L1, 42.8 (15.68) ST1 100 1 1.75 1.26
R3 30 1 1.80 2.01
ST2 30 1 2.70 2.51
R2 50 1 1.80 1.90
ST3 50 1 2.70 2.30
R1 100 1 1.85 1.90
R2 30 1 2.70 3.73
D2, 73.5 (16.52)
R3 50 1 2.95 3.82 ST2 30 1 2.75 1.56
R1 100 1 2.68 3.25 ST1 50 1 2.75 1.60
C1 33.3 1 1.80 4.18 ST3 100 1 2.75 1.56
C2 66.6 1 1.80 3.43 L2, 48.9 (15.84) R1 30 1 2.8 1.98
R2 20 1 2.75 2.82
ST1 30 1 3.95 4.17
R5 20 1 1.90 2.29
ST2 50 1 3.95 3.99
R4 33.3 1 1.83 1.72
R 30 1 4.00 7.11
D3, 75.28 (16.58) 3 C1 33.3 1 1.90 2.61
R1 50 1 4.00 6.34
R2 100 1 4.00 6.15 ST2 30 1 3.60 2.38
C1 33.3 1 2.70 7.32 ST3 50 1 3.85 2.35
C2 66.6 1 2.70 6.68 ST1 75 1 3.70 2.25
L3, 45.7 (15.76) R3 30 1 3.80 3.93
ST1 30 1 3.85 4.01
R2 50 1 3.80 3.86
ST2 50 1 3.96 3.60
R1 100 1 3.73 3.75
R 30 1 3.81 5.98
D4, 75.14 (16.57) 1 C1 33.3 1 2.53 4.70
R3 50 1 3.95 5.89
C2 66.6 1 2.53 4.70
R2 100 1 3.85 5.59
C1 33.3 1 2.53 6.31 L4, 47.35 (15.8) R4 20 1 1.83 2.13
C2 66.6 1 2.57 6.12 C3 20 1 1.90 2.39
D5, 69.53 (16.41) R4 20 1 1.80 2.96 L5, 47.37 (15.79) R4 20 1 2.87 3.33
C3 20 1 1.93 4.33 C3 20 1 2.77 4.19
D6, 71.1 (16.45) R5 20 1 3.03 4.29 L6, 48.67 (15.83) R4 20 1 4.07 4.51
C3 20 1 2.90 8.28 C3 20 1 3.90 6.04
D7, 74.35 (16.54) R4 20 1 3.77 6.55 C6 30 1 5.78 9.73
C7 30 3 5.92 9.92
C5 30 1 5.94 14.46
C5 30 5 5.95 12.27
C4 30 3 5.9 17.22
C4 30 5 5.85 11.40
C7 30 5 5.87 18.21 L7, 50.21 (15.87) R8 30 1 5.78 6.46
R6 30 1 6.05 8.95
D9, 78 (16.65) R7 30 3 5.92 6.95
R5 30 1 5.8 8.58
R6 30 5 5.93 8.97
R8 30 3 5.9 9.1
R5 30 5 5.9 8.69
R7 30 5 5.93 9.2
Companion cone penetrometer tests were also

These uplift tests involved anchors with different conducted in these reported studies at 20g using
geometry (strip (ST) with aspect ratio 8:1, rectangular a 10 mm diameter cone penetrometer (dcone).
1078
Figure 2 presents qc profiles measured across: (a) the not possible due to the space constraints in the strong
13 sand samples reported in Roy et al. (2021) at boxes; as such Equation (1) was subsequently used in
20g performed at the start of testing, and (b) qc pro such cases to establish the respective qc profile. The
files from the 13 sand samples in Hao et al. (2019). validity of such an approach is confirmed by the good
These qc profiles are used to estimate the relative agreement between the measured and simulated qc
density (RD) of the samples using a qc - RD relation zm/dcone profiles using Equation (1) at two different
proposed by Roy et al. (2019) for the UWA silica RD and g-levels in a separate CPT study as shown in
sands, that accounts for effective vertical stressðσ0v ), Figure 3.
relative density and normalised penetration depth
(zm /dcone):
where the constants for UWA Silica sand were

obtained as C0 = 8.5, C1 = 0.89, C2 =3.3, C3 = 0.36,
C4 = 1049.2 RD7.25, C5 = 3.7 respectively. ΔC1 oc and
ΔC2 oc are constants that account for the decreased
compressibility of the sands under OC condition, the
details of which are explained in Roy et al. (2019). It Figure 1. Model plate anchors and piezocone (after Roy
is worth noting that the effect of shallow embedment et al. 2021a, b).
on qc for centrifuge CPTs are incorporated through
a varying depth factor (fD) in Equation 1. Using Equa
tion 1, relative density at the anchor locations was
back-analysed as RD = 41 – 51% for the loose sand
and RD = 69 – 76% in the dense sand reported by
Roy et al. (2021a). For the tests reported by Hao et al
(2019) on circular plates, the RD was ~ 85 - 96%.
Table 3. Anchor database in very dense sand at 20g and

OCR = 1 (after Hao et al., 2019).
Relative density (%)
(unit weight) (kN/m3) Anchor code H/D qa (kPa) Nγ
96.2 (17.07) H 3 182.29 8.9

86.7 (16.87) H 4 340.12 12.6
86.4 (16.86) H 6 861.68 21.3
96.2(17.07) H 6 966.73 23.6
90 (16.94) H 7.5 1285.68 25.3
86.4 (16.86) H 8 1407.82 26.1
96.4 (17.07) H 8 1731.80 31.7
Figure 2. Measured qc profiles in loose, dense (Roy et al.,
88.8 (16.91) H 9 1991.16 32.7 2021a) and very dense sand (Hao et al., 2019).
96.1 (17.07) H 9 2156.46 35.1
96.2 (17.07) H 9 2070.69 33.7
96.4 (17.07) H 10 2465.23 36.1 Table 4. Physical properties of UWA silica sand.
90 (16.94) H 10.5 2169.90 30.5
86.7 (16.87) H 2 78.28 5.8 Soil properties UWA Silica sand
85.8 (16.84) C 3 181.87 9.0
85.8 (16.84) C 6 860.86 21.3 Angularity Sub-rounded to sub-angular
85.8 (16.84) C 9 1879.34 31.0 Maximum void ratioa 0.789
Minimum void ratioa 0.512
Specific gravityb 2.67
For anchor tests conducted at different g-levels or
Uniformity coefficient (U) 1.73
OCRs within the same sample (see samples D1-D7,
Mean particle size (d50) 0.2 mm
D9 in Table 1 and L1-L7 in Table 2), a qc profile at
the respective g-level or OCR was required to gener a Maximum and minimum void ratios calculated as per
ate the correlate to anchor capacity. However, CPT AS 1289.5.5.1
tests at all g-levels as those of the anchor tests were b Specific gravity calculated as per AS 1289.3.5.1
1079
3 ESTABLISHING NΓ - qc CORRELATIONS failure mechanism. As per reported studies by Bolton
et al. (1999) and Liu & Lehane (2020), the critical nor
3.1 Anchor capacity factor malised depth governing the transition from shallow to
a deep mechanism for a cone penetration test usually
The mobilised peak anchor capacity (qa) in this paper is ranges between 5 and 15. So, cone profiles from centri
reported in dimensionless form as peak anchor factor fuge samples would produce a smaller qc value at shal
(Nγ ¼ qa =σ0vo ; where σvo
0
is the effective overburden low depths than the corresponding prototype samples
stress). The qa and Nγ for each anchor test is tabulated in at similar stress and density levels. Hence, correction
Tables 1-3. The Nγ is typically attained within normal is necessary to obtain an equivalent qc value under the
ized displacement ðδ=B or δ=DÞ of 15% in medium effect of similar stress and density levels but independ
dense (RD = 48%) and 7-12% % in dense sand (RD = ent of penetrating distance. This can be readily worked
70%), as reported in Roy et al. (2021a,b). At a given H/ out using the correlations by use of depth factor fd = 1.
B, the highest Nγ is observed from circular plates The depth-corrected qc profiles simulated using fd
whereas the least is observed from strip anchors. = 1 in Equation (1) is shown in Figure 4 for two dif
ferent densities at two different stress levels. These
profiles show that the effect of depth correction on qc
is more pronounced as the g-level and the RD
increases. For the measured (uncorrected) centrifuge
qc profiles in Figure 4, a shallow failure mechanism
results in an increased concave upwards curvature at
the shallower penetration depths. In contrast, the
simulated (depth-corrected) profiles are convex-
shaped at the shallow penetration depths. The simu
lated and measured profiles tend to merge with each
other at z/dcone value greater than 8.
3.3 Obtaining average qc values

The anchor tests reported in Table 1-3 were conducted
at stress levels ranging from 20g to 100g. In order to
obtain a correlation between Nγ and qc, a representative
qc value in the influencing domain was necessary at the
corresponding g level of the anchor test. Numerical stud
ies have confirmed that (Al Hakeem and Aubeny, 2019,
Hao et al. 2014), for plate anchor uplift under drained
conditions, the failure mechanism significantly mobilises
a soil on top of the anchor plate up to a distance of
2D or 2B. Based on this evidence, a representative qc
value (termed as qc,avg) was taken as the average over
Figure 3. Comparison of measured and simulated qc - zm/ a distance of 2D or 2B over the top of the anchor using
dcone profiles at different RD and g-levels. the depth-corrected simulated qc profiles obtained by
correlations in Equation 1. This zone encompasses the
distance required to mobilise peak Nγ for all anchor
3.2 Correcting qc for shallow embedment effects
types and embedment ratios.
The qc profiles from centrifuge samples undergo shal
low penetration and need to be corrected for effects of 3.4 Correlation for anchors
shallow embedment (Bolton et al. 1999). For example,
if the results for a prototype anchor having D = Figure 5 presents the average cone tip resistance
0.6 m at H/D of 2 was to be simulated, a model anchor (qc,avgÞ vs. mobilised peak anchor capacity (qa) data at
plate with diameter of 30 mm (as reported in the data all g levels for circular anchor. The data shows qa lin
base) would have to be embedded at a sample depth of early with an increase in qc,avg for all anchor types.
60 mm at 20g. With centrifuge penetrometers having However, at a particular value of qc,avg, qa values from
dcone = 10 mm, the anchor placement depth would be loose sand are observed to higher than those in dense
located within a penetrating distance of 6dcone, result sand. Now, in order to inter-compare qa at similar
ing in a shallow failure mechanism for the penetrom qc,avg values in loose and dense sand, an anchor would
eter. For corresponding prototype anchors having D = have to be located at a much deeper depth (or higher
0.6 m, it would have to be embedded at a depth of σ0v ) for the loose sand than for the dense sand. This
1.2 m in field (for H/D = 2); a penetrometer in such suggests that there appears to be a significant contribu
conditions would have penetrated by ~33dcone (consid tion of σ0v in governing peak Nγ. This hints that a stress
ering a standard cone diameter of 35.8 mm) to reach normalised qc,avg value would be necessary to reliable
embedded depth for the anchor, resulting in a deep correlation.
1080
Figure 4. Comparison of measured cone profiles and simu
lated cone profiles using fD = 1 at different g-levels in loose
and dense sand.
In order to capture the effects of stress on the

penetration resistance, Sharp et al. (2010) suggested
the use of a normalised tip resistance term qc1N :
where Pa is the atmospheric pressure, σ0v is the effect-

ive overburden stress and m is a stress exponent term
ranging from 0.3 to 0.8 that better accounts for the
effects of σ0v for a given problem. Trial iterations with
using m = 1 across all densities (i.e. without using any
RD dependent normalisation), showed that insufficient Figure 6. Correlation of Nγ with qc1N in loose and dense
overlap across all datasets, thus hindering the devel sand for (a) strip anchors, (b) rectangular anchors and (c)
circular anchors.
opment of a unique correlation. This suggested that
a density dependent m value needs to be used to
ensure sufficient overlap between the datasets. The value of m was therefore evaluated using
an optimisation routine to produce the greatest
overlap in loose and dense sand data by
a relative shift in the abscissa of the Nγ - qc1N
plot. This was achieved iteratively resulting in
a value of m = 0.6 in loose sands (RD ~45%),
0.4 in dense sands (RD ~70%) and 0.39 in very
dense sands (RD ~90%). More anchor experi
mental data at RD < 40% would be required to
draw an unambiguous conclusion on the precise
values of m in very loose sand. In absence of
such data, it is assumed that m would take an
upper bound of 0.8 at very low RD values,
therefore the exponent m used to obtain qc1N can
be expressed as:
Figure 5. Variation of peak anchor capacity with average cone
tip resistance in loose and dense sands for circular anchors. m ¼ 0:41ðRD ð%Þ=100Þ-0:24 ð3Þ
1081
Applying the optimized m in Equation (2), the cor REFERENCES
relation between qc1N and Nγ is established as shown
in Figure 6 and presented in Equation (4): Been, K., Jefferies, M. G. & Hachey, J. 1999. The critical
state of sands. Géotechnique 41(3): 365–381. doi:
10.1680/geot.1991.41.3.365.
Bolton, M. D., Gui, M.W., Garnier, J., Corte, J.F.,
Bagge, G., Laue, J. & Renzi, R. 1999. Centrifuge cone
penetration tests in sand. Géotechnique 49(4): 543–552.
where fu is a coefficient dependent on anchor shape doi: 10.1680/geot.1999.49.4.543.
and is equal to 0.0396, 0.0784 and 0.2 for strip, rect Gajo, A. & Wood, D. M. 1999. A kinematic hardening con
angular and circular plates respectively. The R2 stitutive model for sands: the multiaxial formulation.
values for the respective correlations for strip, rect International Journal for Numerical and Analytical
angular and circular plate anchors are 0.776, 0.26 Methods in Geomechanics 23: 925–965.
and 0.881 respectively. The significantly lower R2 of Al Hakeem, N. & Aubeny, C. 2019. Numerical Investiga
0.26 for the rectangular anchors is due to greater tion of Uplift Behavior of Circular Plate Anchors in
Uniform Sand. Journal of Geotechnical and Geoenvir
scatter in the experimental dataset at lower qc1N onmental Engineering 145(9): doi: 10.1061/(asce)
levels and suggests that more experimental data gt.1943-5606.0002083.
would be required to obtain a reliable correlation. Hao, D., Wang, D., O’ Loughlin, C.D. & Gaudin, C. 2019.
The fu values also indicate that the anchor capacity Tensile monotonic capacity of helical anchors in sand:
increases in the order of strip, rectangular and circu interaction between helices. Canadian Geotechnical
lar anchors, which agree with existing studies (e.g. Journal 56(10):1534–1543. doi: 10.1139/cgj-2018-0202.
Murray and Geddes 1987). Hao, D., Fu, S. & Rong, C. 2014. Numerical Analysis of
Uplift Capacity of Circular Plate Anchor in Sand. Jour
nal of geotechnical engineering 19: 18947–18961.
4 CONCLUSIONS Lee, J. & Salgado, R. 2005. Estimation of Bearing Capacity
of Circular Footings on Sands Based on Cone Penetra
tion Test. Journal of Geotechnical and Geoenvironmen
A correlation between anchor capacity factor (Nγ) tal Engineering 131(4): 442–452.
and normalized cone tip resistance (qc1N Þ has Liu, Q. & Lehane, B. M. 2020. A centrifuge investigation
been established based on a collected database of of the relationship between the vertical response of foot
centrifuge tests on strip, rectangular and circular ings on sand and CPT end resistance. Géotechnique 71
anchors at varying g-levels in loose and dense (5):1–11. doi: 10.1680/jgeot.19.p.253.
sands. The centrifuge cone profiles were depth- Murray E.J. & Geddes, J.D. 1987. Uplift of anchor plates
corrected to remove the effect of shallow embed in sand. Journal of Geotechnical and Geoenvironmental
ment and make them inter-comparable with proto Engineering 113(3): 202–215.
type tests. The results indicated that a varying Pucker, T., Bienen, B. & Henke, S. 2013. CPT based pre
diction of foundation penetration in siliceous sand.
stress exponent factor is necessary to account for Applied Ocean Research 41: 9–18. doi: 10.1016/j.
the effect of overburden stress in loose and dense apor.2013.01.005.
sands and to establish a unique correlation. It is Roy, A., Chow, S., O’ Loughlin, C.D. & Randolph, M.F.
desired to validate the proposed Nγ -qc1N correl 2021a. Towards a simple and reliable method for calcu
ation using field anchor and companion CPT lating uplift capacity of plate anchors in sand. Canadian
data. The results can also be used to readily com Geotechnical Journal 58(9): 1314–1333. doi: 10.1139/
pute anchor capacity through CPT profiles when cgj-2020-0280.
the soil fabric has been disturbed due to installa Roy, A., O’ Loughlin, C.D., Chow, S. & Randolph, M.F.
tion effects, which is a distinct advantage when 2021b. Inclined loading of horizontal plate anchors in
sand. Géotechnique: doi: 10.1680/jgeot.20.P.119
compared to other methods relying on in-situ Schneider, J. A., Xu, X. & Lehane, B. M. 2010. End bear
inputs of density and friction angles. ing formulation for CPT based driven pile design
methods in siliceous sands. 2nd International Sympo
sium on Cone Penetration Testing 3(May): 8.
ACKNOWLEDGEMENTS Sharp, M. K., Dobry, R. & Phillips, R. 2010. CPT-Based
Evaluation of Liquefaction and Lateral Spreading in
The authors acknowledge the financial support from Centrifuge. Journal of Geotechnical and Geoenviron
the Australian Research Council Discovery Grant mental Engineering 136(10):1334–1346. doi: 10.1061/
Scheme DP190100914. (ASCE)GT.1943-5606.0000338.
1082
© 2022 the Author(s), ISBN 978-1-032-31259-0
CPT results and installation parameters for CFA piles in pyroclastic soils
G. Russo & M. Ramondini

Department of Civil and Environmental Engineering, University of Napoli Federico II, Italy
A. Vecchietti
Department of Engineering, University of Perugia, Italy
G. Russo
Department of Earth Science, Environment and Resources, University of Napoli Federico II, Italy
ABSTRACT: A large number of bored CFA piles was installed during the construction of a big treatment plant
in Poggiomarino (Naples), where the subsoil is formed by sub-horizontal strata of alluvial soils of pyroclastic
origin interbedded with organic silt layers, overlying the pyroclastic base formation. The installation parameters of
CFA piles (rate of revolution, rate of penetration, torque, etc.) have been recorded during construction. In the paper
the statistical correlation between CPT results and monitoring data of CFA piles installation was analysed. The
statistical correlation allowed evaluating the bearing capacity of each CFA pile based on the monitoring of installa
tion data. The correlation was developed by considering different pile groups located around the CPT verticals, in
order to analyse the sensitivity to the spatial variability of geotechnical subsoil properties. The statistical based
prediction of bearing capacity was validated by considering two load tests on pilot CFA piles at the study site.
1 INTRODUCTION If v=nl, the volume removed is null, the auger

advances into the ground without any removal of the
The CFA or Continuous Flight Auger pile is a rather surrounding soil.
widespread pile type in USA also called auge cast If v=0, the auger turns without advancing and
piles. They are installed by first inserting an auger works as a an archimedean coclea screw by continu
under the combined action of a torque and an axial ously removing soil. To obtain an overall compres
force. The auger retrieval is carried out while con sion effect, it should be Vd 2 Va , and therefore:
crete is pumped through the central stem.
With this technology is possible and is more and d02
more frequent the monitoring of installation param v 2 nl 1 - ¼ vcrit ð3Þ
d2
eters during the pile construction. These parameters
are connected to the step of auger insertion and to
the final stage of auger retrieval and concrete down The second member of the equation (3) represents the
wards pumping. critical penetration speed vcrit, of the auger into the
In the drilling phase, the tube advancement takes ground.
place with an intermediate mechanism between that The paper is dedicated to determine a statistical
of a screw and that of a coclea. Named v the speed correlation between CPT results on one side and
of advancement, n the speed of rotation and l the installation parameters of the CFA piles recorded
pitch of the auger, in a time Δt there is a volume of during the pile construction at the site of the large
displaced soil: treatment plant in Poggiomarino (Napoli, Italy) on
the other side.
A further correlation aimed at the prediction of the
bearing capacity of piles on the basis of the installation
and a volume of soil removed from the auger and recorded parameters has been developed and tested
taken to the surface: using the results of statical design load tests on several
piles instrumented with strain gauges along the pile
shaft
DOI: 10.1201/9781003308829-164
1083
2 GEOTECHNICAL INVESTIGATIONS AND generally higher than 40%, the porosity n ranges
SUBSOIL CONDITIONS between 55% and 65%. The few available lab data on
the shear strength and estimation deriving from the
The whole area has been investigated between 1998 interpretation of CPT tests lead to an overall value of
and 1999, the available in-situ tests include n. 5 the friction angle j0 ¼ 28� for the alluvial soils,
boreholes (20-50m) with undisturbed/disturbed sam j0 ¼ 35� for the base formation of pozzolana.
ples for laboratory tests, 19 CPT (14-30m), 7 CPTU
(16-51m), 4 SCPTU (16-39m).
Two main lithotypes, whose main physical and Table 1. Main physical and mechanical parameters of
mechanical characteristics are reported in Table 1, subsoil layers.
characterize the site (Figure 1): a lacustrine/alluvial
depth f0 c’ Eed
soil constituted by pyroclastic sand and silt with
organic or peat levels, with a thickness of 15÷18 m is Layer [m] [kN/m3] [°] [kPa] [MPa]
found above the Ignimbrite Campana formation,
made of pozzolana (silty sands) and grey tuff levels, Alluvial soil 0.0÷18.0 15.8 28 0 0.5÷5.0
that extends down to 50 m (max. borehole depth). Pozzolana 18.0÷50.0 15.6 35 0 20.0
Cone penetration resistance qc is between 2 ÷ 3 MPa
in the mostly silt, organic levels while qc = 50 ÷ 70
MPa in the sandy levels; in the pozzolana layer, qc
ranges between 5 ÷ 200 MPa. 3 INSTALLATION PARAMETERS OF CFA
PILES AND IN SITU LOAD TESTS
In this study a group of 65 piles with 27 m length and

0.8 m diameter belonging to the biological treatment
tank have been considered, as reported in Figure 2.
Figure 1. Subsoil profile (a) and CPT results (b) of the

Figure 2. Investigated area with details of biological treat
investigated area.
ment tank foundation; the red square evidences the set of
65 CFA piles selected for this study.
Data referring to the alluvial soil are scattered, with
grain size distribution ranging from sandy silt to silty The drilling machines were equipped with Jean-Lutz
sand; the unit weight ranges between 14 and 20 kN/ recording system that allowed for continuous monitor
m3, the water content w=20-60%, with corresponding ing of installation parameters. Rate of revolution, rate of
value of the porosity n=40-60%. Some thin levels penetration and torque with depth were recorded every
with high organic content have been detected at vari 8 cm. Conversion of operating pressure in torque
ous depths, with lower values of (11-12 kN/m3) and (kNm) was performed with reference to a specific rela
higher values of w and n (respectively 80-200% and tionship for the drilling machine provided by the manu
65-80%). The base formation of pozzolana (silty facturer. Torque has been considered as the most
sand) is more uniform: the unit weight ranges relevant installation parameter in this study, for its close
between 14 and 16 kN/m3, the water content w is relation with the subsoil characteristics. An increase of
1084
soil resistance along the depth results in increase of
torque applied for continuing the pile installation and
vice versa. Profiles of average values of torque and box-
plots of statistical parameters with depth for the n. 65
piles considered in the study are reported in Figure 3.
Figure 3. Measured values of torque along depth: average

values and boxplot representation.
Figure 5. Measured values of torque along depth for test

piles.
Table 2. Load tests results.
d L Qmax wmax Pmax Smax
Load test [m] [m] [MN] [mm] [MPa] [MPa]
a* b* c*
pile n.1 0.8 24.0 4.08 75.6 1.55 2.81

Figure 4. Results of piles load tests: axial loads vs. settle pile n.2 0.6 22.5 3.26 81.9 0.89 2.59
ments and axial loads vs. depth curves. pile n.3 0.8 24.1 5.30 22.8 1.36 3.94
Load tests to failure on three instrumented CFA d = diameter; L = length; Qmax = maximum axial load;
piles belonging to the foundations were conducted wmax = maximum settlement; Pmax = maximum tip resist
(Mandolini at al., 2002). Vertical displacements were ance; Smax = maximum shaft resistance.
measured by means of four dial gages and precision
levelling. All the test piles were equipped with the
removable tape extensometers (Bustamante and 4 STATISTICAL ANALYSIS
Jezequel, 1991) whose results were also compared
with vibrating wire gages installed in a central pipe Statistical correlations between the relevant parameters
(Russo, 2005). The main results of the load tests are were investigated with R numerical code (R Core
listed in Table 2, while the axial load-settlement Team, 2013). CPT tip resistance qc and torque Mt
curves (total load Q, shaft load, S and base load P) recorded during the drilling of 65 piles of the bio
and the axial load distributions along the pile shaft logical treatment tank and CPT results in the proximity
as an example are reported for all the tested piles in of the area were first considered. A sensitivity analysis
Figure 4. In Figure 5 is shown the torque registered was performed, aimed at investigating correlations at
during the tested piles. difference distances scales, considering three ranges,
1085
Figure 7. CFA piles: regression functions of qc=f(Mt) for
the three selected set of CPTs.
located within the biological tank foundation, n. 29

CPTs performed within the plant perimeter). For each
group of selected CPTs, the tip resistance qc varies
with depth as shown in boxplots of Figure 6, high
lighting similar trends of qc with depth regardless the
number of data selected, thus confirming the homo
geneity of subsoil profile in the considered area.
Regression analyses of average values of Mt and
qc with depth confirmed that correlation coefficients
ρ for each set of selected CPTs were similar (namely
ρ ¼ 0:63 for n.4 CPTs set, ρ ¼ 0:66 for n.7 CPTs
set, ρ ¼ 0:71 for n. 29 CPTs set), denoting
a positive, moderate correlation between the two
variables. The function qc= f(Mt) was determined by
means of non-linear regression analysis. Polynomial
regression models were used with the constrain that
qc = 0 for Mt = 0. The three functions are reported in
Figure 7. The assessment of the “goodness of fit”
was provided through the absence of trends of resid
uals (not shown), and a close relation between meas
ured qc and fitted qc trends with depth for the three
levels of analysis.
The statistical correlation between Mt and the uni
tshaft resistance s of test piles was also analyzed. The
distribution of s with depth was evaluated from the
measured variation of axial load along the test pile
shaft. The function s = f(Mt) derived by the non-linear
regression analysis between s(z) and measured Mt(z)
is reported in Figure 8.
Figure 6. CPTs tip resistance qc vs. depth: average trends

and boxplots for a) n.4 CPTs, b) n.7 CPTs, c)n.29 CPTs.
from the nearest CPTs around the selected piles up to

all available CPTs in the investigated area (namely n.4 Figure 8. Test piles: regression function of s=f(Mt) from
CPTs located close to the selected piles, n.7 CPTs the results of load tests.
1086
5 PREDICTION OF UNIT SHAFT RESISTANCE closer to the effects of bored piles, may explain
the overestimation of shaft resistance based on
To assess the suitability of inferring CFA pile shaft the correlation function s=f(qc(Mt).
resistance form installation parameters, the derived The correlation function s=f(Mt) seems to be less
relations from statistical analysis were used for cal effective in the estimation of both test piles shaft resist
culating the shaft resistance of selected CFA piles ance, probably due to the limited number of Mt data
and compared with the results of load tests. on test piles available for the statistical regression.
The average shaft resistance of the selected CFA
piles was evaluated following the relation (Viggiani
et al., 2012): 6 CONCLUSIONS
sðzÞ ¼ αqc ðzÞ CFA piles are generally considered as replacement piles
because there is a rather large volume of soil removed
where a is a coefficient varying from 0.02 (qc < 2 during the construction. It has long been recognized
MPa, very loose soil) to 0.07 (qc > 25 MPa, dense however that the careful control of the installation
soil) and the tip resistance qc was evaluated as func parameters via monitoring tools on board of the equip
tion of the average torque Mt(z). ment used to install the pile may positively influence the
The shaft bearing capacity of test piles has been performance of the pile under live loading. Particularly
calculated by integrating along the pile length the the comparison between the advancement velocity, v, of
function s = f(Mt) obtained through the statistical the auger and the critical velocity vcrit allow to evaluate
analysis. at least qualitatively the net effect of the installation of
The shaft resistance inferred from load test results the auger on the surrounding soil.
was considered as benchmark for the predictions of In this paper following the path already outlined
shaft resistances based on the regression functions by the authors in previous published works (Mando
with installation parameters. Only test piles n.1 and lini & Russo, 2008, Mandolini et al. 2012,)
n.3 were considered, having those piles the same a correlation between CFA installation parameters
diameter of CFA piles set considered in the study and soil response during the widely used CPT inves
(i.e. d = 0.8 m). In Table 3 the calculated shaft resist tigation tool has been outlined. The best correlation
ances of CFA piles (both ordinary and test piles) and has been found between the tip resistance qc of the
the shaft resistance inferred from load tests are CPT and the torque Mt. A direct statistical correl
reported. ation between this last value and the shaft unit fric
tion of piles load tested to failure has been also
Table 3. Comparison between calculated and measured established. Finally it is also well known that CPT is
pile shaft resistances. a valid tool to directly obtain bearing capacity of
piles at the design stage. In this paper the load tests
Smax S [s = f(qc(Mt))] S [s=f(Mt)] available carried out on instrumented CFA piles
have allowed to compare shaft capacity derived by
Load test [MPa] [MPa] [MPa] statistical correlations with installation parameters
and that one directly measured during loading tests
a* b* c*
on three CFA piles. Two comparisons have been car
pile n.1 2.52 3.17 3.19 3.90 3.51 ried out: one based on direct correlation with the
pile n.3 3.54 3.73 3.50 3.95 2.95 measured torque on tested pile and the other based
on undirect correlation between the torque and the
* a = n.4 CPTs; b = n.7 CPTs; c = n.29 CPTs shaft capacity passing via the correlation with the
CPT tip resistance. The comparison is indeed rather
satisfactory, particularly for the second type of cor
With reference to the integral shaft resistance relation, showing that another step has been done in
S calculated referring to the function s=f(qc(Mt), the direction of reliable pile performance prediction
the results highlight an overestimation of the on the basis of a careful controlled installation
shaft resistance for pile test n.1, whereas for pile process.
n.3 the estimate can be considered as satisfactory,
whenever the level of the analysis performed. As
shown by Mandolini et al. (2012), test pile n.1 is ACKNOWLEDGEMENTS
characterized by a rate of penetration vP along
the pile shaft lower than the critical value vcrit, The Authors are grateful to Maria Vicario, Sara Ersi
whereas for test pile n. 3 the rate of penetration lia Sansone and Sara Polidoro for their effort in per
is higher than vcrit. The effects of the installation forming the statistical analyses during their Master
procedure of test piles n.1 on the soil mass, thesis works.
1087
REFERENCES Bustamante, M., and Jezequel, J.F. (1991). La mesure des
deformations a l’aide des estensiometeres amovible.
Mandolini A., Ramondini, M., Russo, G., Viggiani, C., LCPC Methode d’essai n°34 LEPC, Paris
2002. Full scale loading tests on instrumented CFA Russo G. 2004 Full load test on instrumented micropiles. Proc.
piles. In Deep Foundations 2002: An International Per of Inst. of Civil Eng. Geotech. Eng. 157 (GE3), 127–135
spective on Theory, Design, Construction, and Perform Mandolini A, Russo G. 2008. Statistical Analysis of CFA piles
ance (pp. 1088–1097). construction. Proc. of 5th International Geotechnical Sem
Viggiani, C., Mandolini, A., Russo, G., 2012. Piles and Pile inar Deep Foundations on Bored and Auger Piles, BAP V,
Foundations. Spon Press. Ghent, 8–10 September 2008, ISBN 978-88-6342-027–2
1088
© 2022 the Author(s), ISBN 978-1-032-31259-0
Validation of CPT-based methods for estimation of footing settlement in

sand
V.A. Sakleshpur, M. Prezzi & R. Salgado
Lyles School of Civil Engineering, Purdue University, West Lafayette, IN, USA
ABSTRACT: A variety of design methods have been proposed to estimate the settlement of footings in sand
using cone penetration test (CPT) results. Yet, due to the limited number of well-documented, high-quality
footing load tests, efforts still need to be made to assess the performance of these methods. This paper pre
sents two case histories of full-scale, instrumented load tests performed on axially loaded, square footings in
silica sand reported in the literature. The performance of the traditional Schmertmann method and more
modern CPT-based methods are evaluated through a detailed comparison between the predicted footing load-
settlement curves and those obtained from footing load test measurements. For the case histories considered,
these modern methods are shown to generally produce more accurate, reliable, and consistent predictions of
footing response in sand than the traditional Schmertmann method.
1 INTRODUCTION of the footing load-settlement curve by an initial

linear component followed by a nonlinear (para
Schmertmann’s method (Schmertmann 1970, Schmert bolic) component (Gavin et al. 2009).
mann et al. 1978) is widely used in foundation engin The reliability and accuracy of design methods in
eering practice to predict the settlement of footings in predicting the settlement of footings in sand can be
sand using cone penetration test (CPT) data. The assessed by comparing the predictions with measured
method is based on the following observations: (1) results from instrumented footing load tests. In this
footing settlement results from vertical strains that start paper, two case histories (Briaud & Gibbens 1997,
at some value at the base of the footing, peak at some Lehane et al. 2008) comprising a total of nine instru
depth below the footing base, and then approach zero mented, full-scale load tests performed on square
below a certain depth, and (2) the representative elastic footings in sand at Perth, Australia (four footings),
modulus of each sublayer within the influence zone of and Texas, USA (five footings), are documented. The
the footing is determined from the representative cone footing load-settlement curves predicted by four
resistance for that sublayer. The genesis and shortcom CPT-based design methods (Schmertmann et al.
ings of Schmertmann’s method have been critically 1978, Lee & Salgado 2002, Gavin et al. 2009, and
reviewed by Pantelidis (2020). The main shortcoming Mayne et al. 2012) were compared with the data
of the method is that it is based on linear elasticity and obtained from the footing load tests.
does not account for modulus degradation with Of these, the method by Mayne et al. (2012) is
increasing footing settlement level. a purely direct CPT-based method, i.e. the method
Over the past two decades, several methods relies only on CPT data (e.g. cone resistance qc) to
have been proposed for estimation of footing the exclusion of other variables. The other three
settlement in sand using CPT results. These methods are indirect or soil property-based methods;
methods may be broadly classified into three cat for example, the method by Schmertmann et al.
egories: (1) methods based on Schmertmann’s (1978) requires the estimation of a representative
approach but with some modifications and Young’s modulus E from qc and the calculation of
improvements (Lee et al. 2008, Lee & Salgado a strain influence factor Iz; the method by Lee & Sal
2002, Mohamed et al. 2013, O’Loughlin & Lehane gado (2002) accounts for the dependency of the ratio
2010), (2) methods that fit an equation to load test E/qc on footing settlement level w, footing size B,
data after normalizing the unit load and settlement and relative density DR; while the method by Gavin
of the footing with respect to cone resistance and et al. (2009) requires the small-strain Young’s modu
footing size, respectively (Mayne et al. 2012, lus E0 as input in addition to qc. A summary of the
Mayne & Dasenbrock 2018, Mayne & Illingworth equations involved in these design methods can be
2010), and (3) methods that approximate the shape found in Sakleshpur et al. (2021a).
DOI: 10.1201/9781003308829-165
1089
2 CASE HISTORIES
2.1 UWA footing load tests (Lehane et al. 2008)
2.1.1 Site description and cone resistance profiles

Lehane et al. (2008) reported the results of four,
instrumented, footing load tests performed at the Uni
versity of Western Australia (UWA) test site at Shen
ton Park, Perth, Australia. The soil profile at the site
consists of 5–7 m of poorly graded, medium-dense,
siliceous dune sand overlying weakly cemented
Tamala limestone. The sand layer is of Holocene age
and was formed from the chemical weathering (dis
solution) of limestone with subsequent erosion, trans
portation, and re-deposition by wind (Lehane et al.
2008, Schneider 2007). The presence of small quan
tities (< 5%) of remnant carbonates from the parent
limestone provides the potential for very light cemen
tation and bonding between the sand particles (Lim
& Lehane 2014). The groundwater table is typically
Figure 1. Cone resistance profiles obtained from four CPT
located at a depth of about 5.5 m, just above the lime soundings performed at Shenton Park [digitized from Schnei
stone layer. The degree of saturation S of the sand der (2007)].
layer is typically less than 15% and varies seasonally
by up to about 5% (Lehane et al. 2004, 2008, Lim &
Lehane 2014). Table 1 summarizes the properties of 2.1.2 Footing dimensions and loading details
Shenton Park sand; the sand particles are sub-angular Table 2 summarizes the width, thickness, and embed
to sub-rounded in shape. ment depth of four square footings constructed inline at
the site. The center-to-center distance between the foot
ings is about 5.5 m for footings 1 and 2, 5.2 m for foot
ings 2 and 3, and 4.1 m for footings 3 and 4 (Schneider
Table 1. Properties of Shenton Park sand (Lehane et al.
2007). The four CPT soundings, CPT-1 to CPT-4, were
2004, Schneider et al. 2008).
performed at a horizontal distance of about 3 m from
Property Unit Value the centerline of footings 1–4, respectively (Schneider
et al. 2008). The soundings were performed about nine
Particle sizes D10, D50, D60 mm 0.21, 0.42, 0.47 days after the footings were constructed and six days
Coefficient of uniformity CU — 2.24 before they were tested (Schneider 2007). The footings
Fines content % <5 were loaded up to a maximum value of about 200 kN
Unit weight γm * kN/m3 16.14–16.63 in increments of 15–20 kN using the reaction provided
Minimum void ratio emin — 0.45 by a 25-tonne CPT truck; each load increment was
Maximum void ratio emax — 0.81 maintained for about 10 minutes (Lehane et al. 2008).
Relative density DR % 35–55
Critical-state friction angle c ** (°) 32 Table 2. Dimensions of Shenton Park footings (Lehane
et al. 2008).
* Based on in situ sand replacement density tests.
** Based on isotropically-consolidated (p’ = 100 kPa) Width Embedment Thickness
triaxial compression tests performed on specimens recon Footing B (m) depth D (m) t (m)
stituted to e0 = 0.60 (DR = 58.3%).
1 1.5 1.0 1.0
2 1.0 1.0 1.0
Figure 1 Shows the cone resistance profiles 3 1.0 0.5 0.5
obtained from four CPT soundings performed at the 4 0.67 1.0 1.0
site. Based on self-boring pressuremeter test
(SBPMT) results, Lehane et al. (2008) stated that the
coefficient of lateral earth pressure at-rest K0 2.1.3 Comparison between predicted and measured
decreases from 0.70 at a depth of 1.3 m to load-settlement curves
a relatively constant value of 0.43 below a depth of Figure 2 compares the load-settlement curves of foot
2.3 m. Using this information, we considered ings 1–4 obtained from the footing load tests at Shenton
a constant K0 value of 0.70 between 0–1.3 m depth, Park with those predicted using the methods of
a linear decrease in K0 from 0.70 to 0.43 between Schmertmann et al. (1978), Lee & Salgado (2002),
1.3–2.3 m depth, and a constant K0 value of 0.43 for Gavin et al. (2009), and Mayne et al. (2012). For aged,
depths greater than 2.3 m (Sakleshpur et al. 2021b). normally consolidated silica sand, the parameter λ in
1090
the Lee & Salgado (2002) method was set to a value of
0.53 (Sakleshpur et al. 2021b), while the E/qc ratio in
the Schmertmann et al. (1978) method was set to
a value of 3.5 (Robertson & Campanella 1989). The
following points should be noted for the Gavin et al.
(2009) method: (1) The normalized yield settlement
level wy/B and small-strain Poisson’s ratio ν0 were set
to values of 0.03% and 0.15, respectively, (2) The
depth profiles of small-strain shear modulus G0
obtained from the results of two seismic cone penetra
tion tests performed at the site were averaged to obtain
a representative G0 profile, and the small-strain Young’s
modulus E0 was then determined using the relationship
E0 = 2G0(1 + ν0), and (3) The effect of time-dependent
settlement (creep) was considered by setting the values
of parameters t and tR as 600 seconds and 100 seconds,
respectively; t = time elapsed since the application of
the load increment, and tR = reference time corres
ponding to the onset of creep settlement.
Figure 2. Comparison between predicted and measured

load-settlement curves at Shenton Park for (a) Footing 1,
(b) Footing 2, (c) Footing 3, and (d) Footing 4.
Although unload-reload tests were performed on

the footings, the unload-reload loops of the measured
load-settlement curves were not included in Figure 2
for simplicity. According to Lehane et al. (2008),
some of these loops were unplanned and arose due
to on-site difficulties with the pump for the hydraulic
jack. It can be seen from Figure 2 that the load-
settlement curves obtained using the Lee & Salgado
(2002) method are either in good agreement (e.g.
footings 2 and 4) or slightly conservative (e.g. foot
ings 1 and 3) with respect to the UWA footing load
test data than those obtained using the other methods
considered in this study.
1091
2.2 Texas A&M footing load tests (Briaud &
Gibbens 1997)
2.2.1 Site description and cone resistance profiles

Briaud & Gibbens (1997) reported the results of five,
instrumented, footing load tests performed at the
National Geotechnical Experimentation Site on the
Texas A&M University Riverside Campus near College
Station, Texas, USA. The soil profile at the site consists
predominantly of medium-dense, silty fine silica sand
of Pleistocene age up to a depth of 11 m. The fines con
tent at the site generally varies from about 8% near the
ground surface to 35% at a depth of 9 m; the fines are
nonplastic. The silty sand layer is overconsolidated due
to the desiccation of the fines and the removal of 0.5–
1.5 m of overburden prior to the construction of the
footings. Below the silty sand layer, there is a very stiff,
marine clay deposit of Eocene age extending down to
a depth of about 33 m. The liquid limit LL and plasti
city index PI of the clay layer are 40% and 21%,
respectively. The groundwater table was observed at Figure 3. Cone resistance profiles obtained from five CPT
soundings performed at Texas A&M [digitized from Briaud
a depth of 4.9 m from the ground surface. Table 3 sum & Gibbens (1997)].
marizes the properties of the silty sand layer at the site.
2.2.2 Footing dimensions and loading details
Table 3. Properties of silty sand layer at Texas A&M Table 4 summarizes the as-built dimensions and
(Briaud & Gibbens 1997). embedment depths of the footings. The footings
were loaded using the reaction provided by four
Property Unit Value 0.91-m-diameter, 21.3-m-long, belled drilled shafts
(with 60° under-reamed bells of 2.7 m base diam
Specific gravity Gs — 2.64–2.66 eter) and one 0.91-m-diameter, 5-m-long, cylindrical
Mean particle size D50 mm 0.15–0.20 drilled shaft. The footings were loaded in increments
Coefficient of uniformity CU — 1.8–2.4 equal to 1/10th of the footing capacity estimated by
Unit weight γm kN/m3 15.28–15.65 Briaud & Gibbens (1999) using traditional bearing
Minimum void ratio emin — 0.62–0.65 capacity calculation methods. Each load increment
Maximum void ratio emax — 0.91–0.94 was typically maintained for 30 minutes. The layout
Relative density DR % 55 of the footings and CPT soundings at the site can be
Critical-state friction angle c * (°) 34.2 found in Briaud & Gibbens (1997) – CPTs 1 and 2
* Based on consolidated, drained triaxial compression test correspond to footings 5 and 4, respectively, while
results. CPTs 5–7 correspond to footings 1–3, respectively.
Figure 3 shows the cone resistance profiles obtained Table 4. As-built dimensions of Texas A&M footings
from five CPT soundings performed at the site. It can be (Briaud & Gibbens 1999).
seen that the cone resistance qc obtained from sounding
CPT-7 is very low (≈ 300 kPa) at a depth of 3 m. The Length L × width B Embedment Thickness
corresponding values of sleeve resistance fs and friction Footing [as-built (m)] depth D (m) t (m)
ratio fs/qc were also reported to be very low at this depth
(Briaud & Gibbens 1997). However, results obtained 1 3.004 × 3.004 0.762 1.219
2 1.505 × 1.492 0.762 1.219
from adjacent CPT soundings (CPT-2 and CPT-6) reveal
that the cone resistance at a depth of 3 m is about 6 3 3.023 × 3.016 0.889 1.346
MPa, which is 20 times greater than that obtained from 4 2.496 × 2.489 0.762 1.219
5 0.991 × 0.991 0.711 1.168
sounding CPT-7. In addition, results obtained from an
SPT boring (SPT-1) adjacent to CPT-7 show that the
SPT blow count at a depth of 3 m is equal to 22 (Briaud
& Gibbens 1997). Therefore, we believe that the very 2.2.3 Comparison between predicted and measured
low qc values observed for sounding CPT-7 near a depth load-settlement curves
of 3 m may not reflect the true soil state below footing Figure 4 compares the load-settlement curves of
3. Accordingly, for a depth of about 2.0–3.5 m, instead footings 1–5 obtained from the footing load tests at
of using the qc profile obtained directly from sounding Texas A&M with those predicted using the methods
CPT-7, we considered the cone resistance to increase of Schmertmann et al. (1978), Lee & Salgado
linearly from 4.2–6.5 MPa within this depth range. (2002), Gavin et al. (2009), and Mayne et al. (2012).
1092
For overconsolidated silica sand, the parameter λ in
the Lee & Salgado (2002) method was set to a value
of 0.91 (Sakleshpur et al. 2021b), while the E/qc
ratio in the Schmertmann et al. (1978) method was set
to a value of 6.0 (Robertson & Campanella 1989).
The following points should be noted for the Gavin
et al. (2009) method: (1) The values of wy/B and ν0
were set to 0.03% and 0.15, respectively, (2) The
depth profiles of small-strain shear modulus G0
obtained from the results of in situ cross-hole tests
performed at the site were averaged to obtain
a representative G0 profile from which the values of
the small-strain Young’s modulus E0 were obtained,
and (3) The effect of time-dependent settlement
(creep) was not included in the analysis of this case
history because the equation to model the creep
response was developed by Lehane et al. (2008) pri
marily for the Shenton Park site.
Figure 4. Comparison between predicted and measured

load-settlement curves at Texas A&M for (a) Footing 1, (b)
Footing 2, (c) Footing 3, (d) Footing 4, and (e) Footing 5.
1093
The data points on the measured load-settlement Lee, J. & Salgado, R. 2002. Estimation of footing settlement in
curves shown in Figure 4 correspond to the footing sand. International Journal of Geomechanics 2(1): 1–28.
settlements obtained at the end of each load increment Lee, J., Eun, J., Prezzi, M. & Salgado, R. 2008. Strain influ
during the load tests; for simplicity, all unload-reload ence diagrams for settlement estimation of both isolated
cycles have been removed from these curves. It can be and multiple footings in sand. Journal of Geotechnical
seen from Figure 4 that the load-settlement curves and Geoenvironmental Engineering 134(4): 417–427.
obtained using all the methods, except that of Schmert Lehane, B.M., Ismail, M.A. & Fahey, M. 2004. Seasonal
dependence of in situ test parameters in sand above the
mann et al. (1978), are in fairly good agreement with the water table. Geotechnique 54(3): 215–218.
Texas A&M footing load test data over a wide range of Lehane, B.M., Doherty, J.P. & Schneider, J.A. 2008. Settle
settlement levels. The load-settlement response obtained ment prediction for footings on sand. Deformation char
using the Schmertmann et al. (1978) method tends to be acteristics of geomaterials; Proc. 4th intern. symp.,
unconservative when the load or settlement level Atlanta, GA, Vol. 1, 133–150. Amsterdam: IOS Press.
increases beyond the range to which the representative Lim, J.K. & Lehane, B.M. 2014. Characterisation of the
Young’s modulus applies. effects of time on the shaft friction of displacement piles
The method by Gavin et al. (2009) is sensitive to in sand. Géotechnique 64(6): 476–485.
Mayne, P.W. & Illingworth, F. 2010. Direct CPT method
the value of relative settlement wy/B up to which the
for footing response in sands using a database approach.
small-strain Young’s modulus E0 is assumed to oper Cone penetration testing; Proc. 2nd intern. symp., Hun
ate. For example, in the UWA case history, if the tington Beach, CA, 9-11 May, Vol. 3, 315–322.
value of wy/B is decreased from 0.03% [the value Mayne, P.W., Uzielli, M. & Illingworth, F. 2012. Shallow
suggested by Gavin et al. (2009)] to 0.005% (Lehane footing response on sands using a direct method based
et al. 2008), the footing load-settlement curves on cone penetration tests. Full-scale testing and founda
obtained using the Gavin et al. (2009) method would tion design: Honoring Bengt H. Fellenius (GSP 227);
be in better agreement with the load test data. In con Proc. GeoCongress 2012, Oakland, CA, 25–29 March,
trast, the method by Mayne et al. (2012) was devel 664–679. Reston, VA: ASCE.
oped based on statistical (regression) analysis of Mayne, P.W. & Dasenbrock, D. 2018. Direct CPT method
for 130 footings on sands. Innovations in geotechnical
a footing load test database; both the UWA and engineering: Honoring Jean-Louis Briaud (GSP 299);
Texas A&M footings were a part of this database. Proc. IFCEE 2018, Orlando, FL, 5–10 March,
135–146. Reston, VA: ASCE.
Mohamed, F.M.O., Vanapalli, S.K. & Saatcioglu, M. 2013.
3 CONCLUSIONS Generalized Schmertmann equation for settlement esti
mation of shallow footings in saturated and unsaturated
The paper presents two case histories of axial load sands. Geomechanics and Engineering 5(4): 343–362.
tests performed on full-scale, instrumented, square O’Loughlin, C.D. & Lehane, B.M. 2010. Nonlinear cone
footings in silica sand. The load-settlement curves penetration test-based method for predicting footing
obtained from the test results for each of the nine settlements on sand. Journal of Geotechnical and
footings were reported. Four CPT-based footing Geoenvironmental Engineering 136(3): 409–416.
design methods were used to predict the load- Pantelidis, L. 2020. A critical review of Schmertmann’s strain
influence factor method for immediate settlement analysis.
settlement curves of the test footings in both case Geotechnical and Geological Engineering 38(1): 1–18.
histories. The estimated load-settlement curves were Robertson, P.K. & Campanella, R.G. 1989. Guidelines for
compared with those obtained from the footing load geotechnical design using the cone penetrometer test
tests to assess the performance of the design and CPT with pore pressure measurement. 4th Edition,
methods. Among the four footing design methods, Columbia, MD: Hogentogler & Co.
the Lee & Salgado (2002) method generally provides Sakleshpur, V.A., Prezzi, M., Salgado, R. & Zaheer, M.
the most accurate load-settlement estimates for the 2021a. CPT-based geotechnical design manual,
case histories considered in this paper. Finally, from Volume 2: CPT-based design of foundations (methods).
West Lafayette, IN: Purdue University (in press).
a practical standpoint, the paper demonstrates the
Sakleshpur, V.A., Prezzi, M., Salgado, R. & Zaheer, M.
applicability of CPT data and CPT-based methods 2021b. CPT-based geotechnical design manual, Volume 3:
for estimation of footing settlement in sand. CPT-based design of foundations (example problems).
West Lafayette, IN: Purdue University (in press).
Schmertmann, J.H. 1970. Static cone to compute static
REFERENCES settlement over sand. Journal of the Soil Mechanics and
Foundations Division 96(3): 1011–1043.
Briaud, J.-L. & Gibbens, R. 1997. Large-scale load tests Schmertmann, J.H., Hartmann, J.P. & Brown, P.R. 1978.
and data base of spread footings on sand. Report No. Improved strain influence factor diagrams. Journal of the
FHWA-RD-97-068. McLean, VA: Federal Highway Geotechnical Engineering Division 104(8): 1131–1135.
Administration. Schneider, J.A. 2007. Analysis of piezocone data for dis
Briaud, J.-L. & Gibbens, R. 1999. Behavior of five large placement pile design. Ph.D. Dissertation. Perth, Austra
spread footings in sand. Journal of Geotechnical and lia: The University of Western Australia.
Geoenvironmental Engineering 125(9): 787–796. Schneider, J.A., Fahey, M. & Lehane, B.M. 2008. Characterisa
Gavin, K., Adekunte, A. & O’Kelly, B. 2009. A field investi tion of an unsaturated sand deposit by in situ testing. Geo-
gation of vertical footing response on sand. Proceedings of technical and Geophysical Site Characterization 3; Proc.
the ICE-Geotechnical Engineering 162(5): 257–267. 3rd intern. conf., Taipei, 1–4 April, 633–638. CRC Press.
1094
© 2022 the Author(s), ISBN 978-1-032-31259-0
Derivation of SRD for driven piles from CPT data

J.A. Schneider
U.S. Army Corps of Engineers, St. Paul, MN, USA
M.F. Randolph & J.P. Doherty

ABSTRACT: Measurements from the cone penetration test are analogous to parameters used in assessing
the static component of the soils pile driving resistance. This paper presents comparisons of pile driving data
from 610 mm open ended pipe piles driven through lake clays into underlying glacial till, with estimates of
driving resistance using the web-based software, IMPACT. IMPACT uses a continuum-based soil model that
explicitly separates internal and external shaft friction in the soil-pile model. CPT based expressions for
evaluating the static resistance to driving include factors that account for end condition, friction fatigue and
differences in the diameter of the CPT and the pile. Evaluation of static tension load test results requires add
itional estimates of setup and relative contribution of driving resistance from the soil plug.
1 INTRODUCTION wave analysis, and pile load testing for a case history
of an open-ended pile in clay. The stress waves are
Data from cone penetration tests (CPT) are ideal for analyzed systematically to provide insight into the
assessing the ‘static resistance to driving’ (SRD) of derivation of the SRD from CPT parameters.
a pile as well as for evaluating long-term axial capacity Dynamic analysis of piles, including drivability
after installation and equalization. Pile-soil interaction studies and stress-wave analysis following a dynamic
parameters can be derived from the cone data, includ test, is presented in this paper, using the web-based
ing values of end-bearing resistance and maximum software, IMPACT. The software allows input of CPT
values of shaft friction relevant in the current vicinity data as the basis for assessing relevant pile-soil inter
of the pile tip. action parameters, with several published recom
The sleeve friction measurement, while conceptu mendations for estimating pile SRD from CPT data
ally similar to pile shaft friction, has complications in being included as options. However, the web-app
application to assessing both long-term static resist allows the user to input their own algebraic expressions
ance and short-term pile driving resistance. This is relating, for example, maximum shaft friction to cone
illustrated in Figure 1 by comparing� average � long tip resistance for any given soil behaviour type, and
term shaft friction from pile load tests τf;avg in clay similarly for the friction degradation functions and the
to average CPT sleeve friction (fs,avg). The ratio of end-bearing resistance. Application of the web-app is
τf;avg =f s;avg ranges from over three to approximately illustrated with an example field study where dynamic
0.5. For low values of fs,avg, setup, or the increase in stress-wave data were measured during driving of an
capacity with time due to equalization of pore water open-ended pile in lake clay overlying glacial till.
pressures tends to control the ratio. For high values of
fs,avg, the ratio is less than unity, implying that friction
fatigue, or the reduction in local friction with con 2 SITE CHARACTERIZATION, FARGO, ND
tinued pile penetration, tends to dominate long term
pile friction when compared to CPT sleeve friction. Data discussed in this paper were collected as part of
For the case of pile driving, evaluation of pile shaft a load test program at the Diversion Inlet Structure, in
friction from CPT sleeve friction would seem to be Fargo, ND, USA. Site characterization included piezo
simpler than attempting to assess setup for evaluating cone penetration tests (CPTU), prebored pressuremeter
the long-term capacity. However, viscous rate effects tests (PMT), standard borings and index testing, as
and separating the contribution of internal as com well as laboratory triaxial and oedometer tests.
pared to external shaft friction for open ended piles Figure 2 shows results of a CPTU performed adja
still require additional analysis and interpretation. This cent to the pile driving locations, along with PMT
paper presents interpretation of pile drivability, stress and index test results collected in the area. The site
DOI: 10.1201/9781003308829-166
1095
and Doherty et al. (2020). Equations based on the
Alm & Hamre (2001) drivability method were used
in this paper to account for friction fatigue. The
method includes friction fatigue that leads to
a reduction in local shaft friction ðτs Þ with height (h)
above the pile tip (in m) according to
The rate of friction fatigue (k) increases with nor

malized cone tip resistance according to
Figure 1. Comparison of ratio of average pile shaft friction

to average CPT sleeve friction (database used discussed in
Schneider 2010).
The soils at the site are predominantly clayey.
The maximum shaft friction is evaluated from CPTU
consisted of approximately 13 m of lake clays under sleeve friction, and the minimum shaft friction is
lain by glacial till. Most CPTs were terminated at the related to normalized cone tip resistance by
top of the till. The CPTU shown had the deepest
penetration in the till, of about 2.5 m. CPT refusal in
the clay till was typically associated with sharp
changes in cone inclination which exceeded 1 degree
over a 1 m push length.
3 PILES AND DRIVING DATA

The end bearing in clays is taken as 60% of the
Two 18.3 m long, 610 mm diameter x 19 mm wall cone tip resistance.
thickness (D/t = 32) open ended pipe piles were Soil stiffness was based on secant unload-reload
driven with a Delmag D46-32 hammer to a depth of modulus from pressuremeter tests. This resulted in
approximately 16.2 m. Each pile was monitored with ratios of G/qb of approximately 30 and G=τs of 200.
a pile driving analyzer (PDA) during installation. Poisson’s ratio was taken as 0.5. Viscous rate effects
One pile (TP-7) was statically load tested after were used to evaluate dynamic shaft friction ðτd Þ
a setup time of 120 days. The second pile, TP-8, was using a power law function of
restruck at 125 days after installation.
Figure 3 shows the collected pile driving data
(blowcount, stroke, and transferred energy) adja
cent to the CPTU cone tip resistance. In addition
to the measured CPTU data, the geometric mean
of tip resistance taken either from 4 equivalent
diameters (Deq) above the cone tip to 2 Deq below where vref is 1 m/s and Dv is the relative velocity
the cone tip, or from 4 diameters (D) above the between the pile and the soil. The parameter β was
cone tip to 2 diameters below, are shown. The taken as unity for all soils. Softer clays used an α of
equivalent pile diameter of 210 mm represents 0.2, while heavily overconsolidated clays (Till) used
a solid circle with the same area of steel as the an α of 0.06.
pile annulus. Average cone tip resistance and Comparisons between measured and calculated
sleeve friction were used in analyses. blowcounts using impact and a continuum soil model
The Delmag D46-32 hammer has a rated energy are shown in Figure 4. Two cases were analyzed: (1)
of 165 kN-m. Hammer efficiencies ranged from 35% with the plug friction taken as 1 percent of the exter
in the lake clays to 45% in the glacial till for TP-7 nal friction, and (2) with the plug friction equal to
and from 40% to 55% for TP-8. the external friction. The small amount of friction for
the first case was primarily to maintain numerical
stability. Minimal to zero plug friction is often used
4 DRIVABILITY ANALYSIS for drivability studies of open-ended piles.
Both cases overpredict blow counts in the upper
The background and workflow for an IMPACT pile lake clays, but show good agreement in the glacial
drivability analysis is discussed by Randolph (2008) till at the end of driving.
1096
Figure 2. Site characterization data from the Fargo Diversion Inlet Structure site.
Figure 3. Pile driving data from the Fargo Diversion Inlet Structure site.
1097
where F is the force, v is the velocity, and Z is the

pile impedance, EA/c. EA is the cross-sectional
rigidity and c is the wave speed in the pile.
Stress wave measurements are presented for three
cases: (i) the end of driving for TP-7 (300th blow);
(ii) the end of driving for TP-8 (350th blow); and
(iii) the beginning of restrike collected 125 days
after driving on TP-8 (6th blow).
The first comparison in Figure 5 for TP-7 uses the
recommended end bearing and shaft friction based on
the Alm & Hamre (2001) method. The match between
the measured and calculated upward traveling wave is
good. However, the method assumes no internal shaft
resistance for a coring case. This has historically been
a reasonable assumption, since commercial wave
equation analysis software does not explicitly model
the soil plug. To logically step through the relationship
between CPT and pile SRD parameters, modeling the
soil plug explicitly becomes more useful.
A series of four analyses of upward traveling
Figure 4. Comparison of Impact predictions of blowcount stress waves are shown in Figures 6 through 9 for
based on CPT data to measured results using Alm & pile TP-8 at the end of driving (EOD). Details of the
Hamre (2001) method, and τf;in ¼ 0:01τf;out and τf;in ¼ τf;out . analyses are:
1. Assume CPT parameters apply directly to the
pile driving analysis with full internal friction,
τs ¼ f s , qb = qt, τp ¼ τs (Figure 6).
2. Essentially remove internal pile plug friction
τp ¼ 0:01τs (Figure 7).
3. Reduce external friction using friction fatigue,
increasing internal friction over lower two diam
eters of the pile (Figure 8).
4. Improve fit, increasing end bearing (Figure 9).
Two more (5th and 6th) analyses are presented for
the TP-8 beginning of restrike (BOR) blow to
explore setup mechanisms, the results of which
Figure 5. Comparison of measured and calculated wave up

for TP-7 assuming resistance distribution from Alm &
Hamre (2001) and τf;in ¼ 10 kPa.
5 SIGNAL MATCHING
Signal matching was undertaken using the down

ward travelling stress-wave as input and comparing
the calculated and measured upward travelling stress
waves (see, for example, Randolph 2003). The
upward travelling wave is

for TP-8 assuming qb = qc, τf ¼f s , and τf;in ¼ τf;out (Step 1).
1098
are compared to static load test response in the driving using CPT parameters, even when no
next section. internal friction from the plug is assumed.
5. Increase external friction to account for setup Friction fatigue was used to reduce external fric
(Figure 10). tion. A greater rate of degradation (k) was needed for
6. Increase internal friction to account for different the till soils as compared to the softer lake clays.
pile response during restrike (Figure 11). Some internal friction was needed for an acceptable
match. A constant internal shaft friction of 700 kPa,
Figures 6 and 7 show a calculated up wave force approximately equal to the CPT sleeve friction, was
greater than the measured up wave force, for
used over the lower two diameters near the pile tip.
a normalized time between approximately 4 and 5. A reasonable match for this case is shown in Figure 8.
This indicates overprediction of soil resistance to The match was slightly improved in Figure 9, by
increasing the annular base resistance from qt to
1.4qt. The higher qb is considered reasonable since

for TP-8 assuming qb = qc, τf ¼ f s , and τf;in ¼ 10 kPa
(Step 2). Figure 9. Improving fit from Step 3 by increasing qb=1.4 qc
(Step 4).

for TP-8 assuming qb = qc, but reducing τf for friction
fatigue and adding some internal friction at bottom of pile Figure 10. Matching TP-8 restrike blow through increasing
(Step 3). external friction.
1099
interpretation of the external shaft capacity from the
restrike blow was approximately 10% less than the
load test capacity.
Table 1. Summary of soil resistance inferred from

dynamic measurements and static load tests.
TP-7 TP-8 TP-8 TP-7

EOD1 EOD2 RS2 SLT
External Shaft 3899 2883 3794 4200

(kN)
Internal Shaft 36 811 2988
(kN)
Base (kN) 257 581 581
Unplugged (kN) 4192 4276 7364
Plugged (kN) 6025 7704 8614
1
Assumed minimal internal friction
2
Analysed using progressive CPT procedure
Figure 11. Improving restrike match for TP-8 by increasing
internal friction.
CPT refusal was met (following increased inclin

ation) prior to the termination tip elevation of the
pile. It is likely that higher resistance layers exist in
the till.
A restrike blow performed on TP-8 was recorded
125 days after pile installation, as shown in
Figure 10. Compared to Figure 9, a significant
increase in measured up wave resistance is shown.
A two-step procedure was used to match the restrike
blow:
1. The external friction was increased to match the
peak up force.
2. The internal friction was increased to improve
the overall fit.
The first estimate of external friction was based
on using CPT sleeve friction directly, with no fric
tion fatigue degradation. The match in Figure 10 was
surprisingly good. This is not to imply that CPT Figure 12. Comparison of external shaft friction from PDA
sleeve friction is always a good match for long term signal matching to tension load test results on TP-7.
pile capacity (see Figure 1). However, the compari
son was reasonable for this pile. Increasing the plug A point of interest from Table 1 comes from
length to 20D, and keeping the internal friction equal looking at the internal shaft friction. For TP-7, the
to the external friction leads to the best match using analysis assumed that essentially all shaft friction
this technique, as shown in Figure 11. is external. This is a common assumption for ana
lysis of open-ended pipe piles, leading to three
follow-on effects: (i) the effects of friction fatigue
6 LOAD TEST COMPARISON are underestimated; (ii) the effects of soil setup
are underestimated; (iii) if plug resistance is
A static load test was performed on TP-7 120 days ignored when interpreting restrikes, tension cap
after pile installation. The load-displacement curve is acity could be significantly overpredicted.
shown in Figure 12. The contributions of internal and
external shaft friction, as well as annular base resist
ance and total plugged and unplugged capacity are 7 CONCLUSIONS
shown in Table 1. The restrike blow for TP-8 showed
an increase in capacity by a factor of approximately This paper focused on analyses that highlight the
1.3 compared to TP-8 at the end of driving. The mechanisms controlling the relationship between
1100
soil-pile resistance and CPT parameters. Insights REFERENCES
related to both soil resistance to driving and static
capacity highlighted effects of friction fatigue and Alm, T. and Hamre, L. (2001). Soil model for driveability
setup at this clayey site. The CPT is an excelled predictions based on CPT interpretations. Proc. 15th Inter
national Conference on Soil Mechanics and Geotechnical
tool to constrain parameters used in pile analyses. Engineering, Istanbul, Turkey, 2, 1297–1302, 2001.
However, additional mechanisms need to be Doherty, J.P., Randolph, M.F. and Schneider, J.A. (2020).
included such that rational decisions can be made Analyzing the driving performance of pile foundations
during installation as well as in relation to long using data driven models. Proceedings, ISFOG 2020:
term capacity. 664–673.
Randolph, M.F. (2003). Science and empiricism in pile
foundation design. Géotechnique, 53(10): 847–875.
Randolph, M.F. (2008). IMPACT: Dynamic analysis of pile
DISCLAIMER
driving. Users’ manual, version 4.2.
Schneider, J.A. (2009). Uncertainty and bias in evaluation
The views expressed are those of the authors and do of LRFD ultimate limit state for axially loaded driven
not necessarily represent those of the U.S. Army piles, The Journal of the Deep Foundation Institute, 3
Corps of Engineers. (2): 25–36.
1101
© 2022 the Author(s), ISBN 978-1-032-31259-0
New method for assessing soil liquefaction resistance using a cyclic cone
penetrometer
A. Sharma, P. Rapanakis, E. Incardona, C. Dano, L. Sibille & B. Chareyre
University of Grenoble Alpes, CNRS, Grenoble INP, 3SR, Grenoble, France
H.H. Sadrabadi
Equatech R&D, Equaterre, Annecy, France
ABSTRACT: This paper introduces a new methodology and machinery to extract soil properties and assess its
liquefaction susceptibility. A mechanical Gouda tip cone penetrometer is used to apply a cyclic load on the soils.
This test aims a load closer to real world liquefaction events, more robustness and a cheaper cost than
a conventional CPTu. Since the cone tip used for the test does not have a pore pressure sensor, nor a friction
sleeve and electrical or optical cables, it is of low maintenance, easy to operate and can be used to drill through
gravel or any dense layer to reach layers of interest without digging a borehole, saving time and money. The
liquefaction resistance is found using the response of soils to cyclic loading on the cone tip. The methodology
also involves measuring the change of tip resistance before and after applying a cyclic load and correlates it to
the volume changes that took place during it. For this, an experimental campaign was carried out in a new cali
bration chamber using a mini cone with stress and pore pressure changes measured independently in the soil
during the cyclic load. The different response to cyclic load of clean Fontainebleu GA39 sand under different
vertical stresses, saturation conditions and created by different sample preparation methods is studied. In turn, the
pore pressure increase and the corresponding number of cycles during the cyclic cone penetration tests are com
pared with similar results from cyclic compressive triaxial tests. This allows us to find a relation between the
loading applied in the two methods to cause a similar increase of pore pressure. Hence, a method is suggested to
compare the loading applied by the cone penetrometer to the Cyclic Stress Ratio(CSR) expected in a liquefaction
event. Eventually, using the CSR and CRR defined, the liquefaction susceptibility of the soils can be calculated.
1 INTRODUCTION developed in the 3SR laboratory to facilitate better

placement of independent pore pressure and total
A new cone penetration testing methodology is intro stress sensors and for better control of saturation con-
duced to predict liquefaction and extract soil properties.
The cone we use for our tests is a mechanical Gouda
tip (Figure 1) which lacks any sensors inside the tip.
Instead, the tip resistance is measured using the force
transmitted by the internal rods and recorded by a force
sensor on the surface. The test uses a cone of standard
10 cm2 surface area for the field experiments, while for
the laboratory experiments, we use a miniature cone of
4 cm2 to minimize boundary effects.
The loading devices, both at the site and in the
laboratory, involves two different driving mechan
isms. One for the constant penetration, to measure the Figure 1. The schematic(top, sizing in mm) and image of
the Gouda tip used for the field test.
tip resistance against penetration at a constant speed
and to reach the desired depth for the cyclic test. The
other is a cyclic jack capable of applying the cyclic ditions. The boundary conditions applied are fixed lat
load to the internal rod of the Gouda tip without eral and bottom boundaries, while the vertical stress
moving the external rods. This jack is constrained in is applied from top using a pressurized membrane
its total drive length (6 cm for laboratory and 10 cm with a hole in the center to allow the cone to enter
for field experiments). For the laboratory experi the calibration chamber. The tip opening is meas
ments, a calibration chamber has been newly ured using the internal measurement of the cyclic
DOI: 10.1201/9781003308829-167
1102
jack and using a LVDT attached to a frame at the The number of cycles required for a given dis
head of the cyclic jack. Under normal driving, the placement, the displacement of tip during the waiting
length of internal rods are chosen to ensure that the time, the rate of displacement during cycles, and the
tip is slightly open (<0.5 cm) and all the tip force is change of tip resistance before and after cycles are
transferred to the force sensor attached at the piston a few significant results being looked into to distin
of the cyclic jack. The cyclic jack acts as a rigid guish between different soil types and extract soil
frame for the internal rods and the force sensor properties, especially the liquefaction susceptibility.
during static driving. During the cyclic loading, the The density of the soil, the stress state the soil exists
cyclic jack applies the load directly on the internal in, are a few of the parameters that can affect the
rods, which moves the front end of the tip without response of the soil and vary these results.
moving the external rods.
3 CHARACTERIZATION OF SAND
2 METHODOLOGY
All the tests were performed on clean Fontainebleau
The traditional CPTu relies on empirical relation GA39 sand whose main properties are summarized in
ships that use the pore pressure, the friction sleeve Table 1.
readings and the cone penetration resistance for pre
dicting liquefaction resistance (Idriss & Boulanger
2006, Robertson 1990). These empirical relations are Table 1. GA39 sand properties, from the supplier SIL
developed from comprehensive databases of tests BECO, France.
done in-situ and in calibration chambers.
However, the dependency on empirical relation SiO2 GS d10 d50 d60
(%) (-) (mm) (mm) (mm) Cu emax emin
ships, the complexity of use and susceptibility of the
sensors to damage and malfunction leads to uncer >99.1 2.56 0.087 0.113 0.122 1.1 1.01 0.56
tainties in CPTu tests. Also, these tests are relatively
expensive to perform, involves equipment with high
maintenance cost and may also require boreholes to
The sand grain shape is sub-angular, sub-rounded
traverse through hard strata. As such, the CPTu test
(Rimoy 2013). The grain size distribution of this
becomes an unpractical solution for a large number
sand falls directly under soils that are traditionally
of geotechnical problems. Besides, the empirical
prone to liquefaction.
relations used to predict liquefaction susceptibility
and other soil properties may need to be adapted for
local soil conditions.
Our novel approach tries to overcome a few of
these shortcomings and provide a cost effective and
reliable method to find liquefiable layers under all
field conditions.
We use a mechanical Gouda tip to apply stress-
controlled cycles on depths of interest. For each test,
the outer rod of the cone penetrometer will be fixed,
and the tip of the cone will be used to apply a cyclic
load in accordance with the patented loading mech
anism on the tip using the internal rods. The max
imum displacement of the tip during each cyclic Figure 2. Grain size distribution for sand prone to liquefac
tion and Fontainebleau GA39 sand.
CPT test is 6 cm in the lab and 10 cm on the field. It
is limited by the dimensions of the Gouda cone or/
and the maximum displacement of the jack applying A number of laboratory experiments were per
the cyclic load. As such, in each cyclic test, we formed to characterize the sand properties, its inter
measure the properties of a layer of depth around action with the cone, and the effect of sample
10 cm while the zone of influence of each test may preparation. Direct shear interface tests between
be as high as 72 cm as is the case in conventional smooth steel used to make the cone and the sand at
CPTu (Rogers 2006). The loading applied to the soil constant normal stress resulted in a friction angle of
can be divided into eight steps which involve two 13.1°. We also performed constant head permeameter
pushes at a constant speed for 1 cm penetration, tests to study the effect of the sample preparation
waiting times after each push while applying method on permeability. Moist tamping specimen
a constant force, and uniform stress cycles after the created more permeable samples than those created
waiting time. The frequency used for the cyclic load by dry pluviation, which were more permeable than
ing was 1 Hz, and the amplitude of cycles was those created by dry tamping. In all cases, the order
decided based on the tip resistance measured during of magnitude of permeability was similar (10-5 m/s)
the initial push. and corresponded to those expected from fine sand.
1103
However, the values varied between 3.4 to 7.1* 10-5 These effects can be caused due to the density differ
m/s. These permeability values correspond to condi ence as well as the difference of method of sample
tions of completely drained conditions during normal preparation. However the shear stress were quite close
cone penetration. (McNeilan and Bugno, 1985). at high axial strain as expected at critical state.
Also, as a part of this study 20 tri-axial tests Also, the effect of back pressure is quite small for
(6 drained, 14 un-drained) were performed to study samples prepared by same method and having simi-
the effect of sample preparation, back pressure, initial
confining stress on very loose(Relative density<15%)
or loose sand specimens (Relative density<35%). For
the drained tests in Figure 3, we see a clear distinction
Figure 3. Varying drained behaviour for different sample Figure 5. Difference of behaviour of moist tamped sample
preparation methods at 200 kPa confining stress. based on initial confining stress.
in volumetric strain for tests done at similar confining lar initial density. As such the specimen in the cali
stresses (200 kPa) based on sample preparation bration chamber which is saturated without a back
method even when all the samples fell in loose or pressure should behave similarly as the one saturated
very loose category. The relative density for the two using a high back pressure.
air poured sample was 25% (±2.5%) and was the min For undrained tests too, the initial density of the air
imum the authors could achieve with this method of poured sample shown in Figure 4 was higher (27% &
35%) for air poured samples as compared to moist
tamped samples (-7% & -1%). The behaviour of the
specimen even under the same initial confining pres
sure (200 kPa) varied depending on the method of
sample preparation and the resulting relative density. It
was observed that even loose air poured samples
dilated (negative pore pressure) hence showing no ten
dency to liquefy. While very loose moist tamped
sample had a development of pore pressure as the axial
strain increased, as such may be prone to liquefaction.
Therefore, we will be able to control the initial
state of the soil sample in the calibration chamber by
either preparing the sample by moist tamping or air
pouring. In particular the control of initial relative
density can be used to create samples that are sus
ceptible to liquefaction or not for this sand.
Lastly, in Figure 5 we can observe the undrained
Figure 4. Varying behaviour of samples during undrained
triaxial tests.
response of four moist tamped samples with relative
density of -4.5±3% when initial confining stress is
changed. We can observe a change of behaviour when
sample preparation. The moist tamped sample resulted initial confining stress changes from 200 to 300 kPa
in much looser samples. The samples discussed in giving us an indication as to what depths on the field
Figure 3 had a relative density of -13.3% (±0.3%) We and what vertical stresses in the calibration chamber
observed that moist tamped loose sample contracted can be used to differentiate between liquefiable and
while shearing but the air poured sample dilated. non-liquefiable cases for these kinds of samples.
1104
4 CYCLIC CPT TESTS IN LABORATORY loading. Thus the test can play a major role in distin
guishing the saturation condition of the soil without
For preliminary tests, we performed cyclic penetration a pore pressure measurement and this is an useful infor
tests on the same Fontainebleu GA39 placed in a cylin mation for calculating the liquefaction probability.
drical container (35 cm Ø and 40 cm depth) at a depth It can also be observed that the penetration of the
tip in dry sample continues for the second round of
cycles almost with the same rate as during the first
round of cycles. However, in the saturated sample, the
rate of displacement continuously decreases during the
first round (between ~20 to ~150 seconds in Figure 6)
and reduces drastically for the second round of cyclic
loading. This occurs even when the maximum and the
minimum amplitude of the two sets of cycles remain
similar. This decrease of the penetration rate is an indi
cation of densification of the saturated sample due to
the first cyclic loading and the second 1 cm push. This
effect is more pronounced in the saturated sample.
Figure 6. The difference of behaviour of saturated and dry

sand to cyclic penetration test.
of 15 cm without applying a confining pressure. We

created the sample by air pouring and observed a rela
tive density of 28±2% and performed tests in dry and
saturated state. We observed that the total tip penetra
tion caused by the cyclic CPT test is less in the dry
state than in the saturated state (Figure 6). However,
the tip resistance measured during static penetration at
Figure 8. Different behaviour of the two saturation condi

tions to static penetration after the cyclic load.
5 FIELD EXPERIMENTS
During the initial investi

Cone Penetration Testing 2022

Uploaded by

Copyright:

Available Formats

Cone Penetration Testing 2022

Uploaded by

Document Information

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

Cone Penetration Testing 2022

Uploaded by

Copyright:

Available Formats

Cone

- Equipment and Procedures

PROCEEDINGS OF THE 5TH INTERNATIONAL SYMPOSIUM ON CONE PENETRATION

Cone Penetration Testing 2022

First published 2022

Library of Congress Cataloging-in-Publication Data

A catalog record has been requested for this book

ISBN: 978-1-032-31259-0 (hbk)

Symposium organizers xix

International advisory board xxiii

Session 1: Equipment and procedures

Italian Geotechnical Association, AGI Alma Mater Studiorum – Università di Bologna

Under the auspices of

ISSMGE TC 102 “In-Situ Testing”

International advisory board

Guido Gottardi – Chair, Università di Bologna, Italy

Glenda Abate – Università di Catania, Italy

B.M. Lehane & E.J. Bittar

S. Lacasse, Z. Liu & F. Nadim

1 INTRODUCTION loading direction, re-testing and loading rate. In add­

2.1 Approaches for evaluation of shaft friction

Bustamante & Gianeselli (1982), for example,

i. Ultimate shaft friction (τf) varies is direct propor­

3.3 Allowance for partial plugging

Capacity estimate for piles for ﬂexible piles in strain-

qb0.1 = [0.12 +0.38Are]qp; for plugged base (expected when

4 CPT METHODS FOR CAPACITY OF

Equation (8) illustrates that the relationship between

5.1 Dependence of τf on loading direction

Fugro-10 (Van Dijk & qnet, h &

The Uniﬁed method in clay is calibrated on

7 LOAD DISPLACEMENT RESPONSE USING

The load-displacement response of piles is normally

7.1 Piles in sand

Although Equation (14) is not a perfect match

Material Point Method simulations of cone penetration and CPT

1 INTRODUCTION 1.2 Methods of cone penetration simulation

Displacement or velocity boundary conditions are

6. Adjust (translate/compress/reset) the back-ground

An objective time rate is required in large-defor­ D/B L U MPM

Small-strain(“elastic”) ν 0.15 compression

4.1 Constitutive model

5.3 Simulation of a real CPT in clayey soil

Shallow Limit unit bearing Computational simu­ REFERENCES

Practical use of shear wave velocity measurements from SCPTU in clays

1 INTRODUCTION considered. The meaning of numerical Vs values

3 SCPTU MEASUREMENT TECHNIQUE

3.1 Traditional approach

2.2 Anisotropy of shear wave velocity

Figure 2. Set-up for CPTU test (a) typical test arrangement

Ideally the test should be carried out according to

3.3 Recent developments in techniques

Material NIBS/EN Poulos (2021)

4 SHEAR WAVE VELOCITY VALUES

Before looking at Vs proﬁles from various countries

5 TYPICAL VS PROFILES FOR CLAYS

5.1 Norwegian clays

5.4 Eastern Canadian clays

5.8 Comparison/link between various clay types

Gmax ¼ SFðeÞðσ0 v σ0 h Þn pa ð1-2nÞ ð4Þ

1 INTRODUCTION loading direction, re-testing and loading rate. In add

i. Ultimate shaft friction (τf) varies is direct propor

An objective time rate is required in large-defor D/B L U MPM

Shallow Limit unit bearing Computational simu REFERENCES

1 INTRODUCTION but useful improvements to traditional CPT equip

Note: 1=high; 2=high to moderate; 3=moderate; 4=moder

Figure 3. Statistical clustering for non-deterministic strati

While goodness-of-ﬁt statistics (typically, the coefﬁ

founded on data from multiple sites. While geotech

where Kc = correction factor for ﬁnes content esti