See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/264894129

Finite Element Prediction of Progressively Formed Conical Stockpiles

Article

CITATION READS
1 49

4 authors:

Jun Ai Jian-Fei Chen

Formerly The University of Nottingham, and The University of Edinburgh Queen's University Belfast
18 PUBLICATIONS   364 CITATIONS    198 PUBLICATIONS   5,462 CITATIONS   

SEE PROFILE SEE PROFILE

J. Michael Rotter J. Y. Ooi

The University of Edinburgh The University of Edinburgh
231 PUBLICATIONS   3,663 CITATIONS    126 PUBLICATIONS   1,919 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

T-MAPPP https://www.t-mappp.eu/ View project

Size effect of RC beams shear-strengthened with FRP View project

All content following this page was uploaded by Jian-Fei Chen on 04 September 2014.

The user has requested enhancement of the downloaded file.

 Finite Element Prediction of Progressively Formed
Conical Stockpiles

J. Ai, J.F. Chen, J.M. Rotter, J.Y. Ooi

Institute of Infrastructure and Environment, Joint Research Institute for Civil and Environmental
Engineering, School of Engineering, The University of Edinburgh,
Edinburgh EH9 3JL, Scotland, U.K.

Abstract: Conical piles of granular solids can be found in many industrial sites. These piles are
usually progressively formed by depositing from above. A classic question concerning such simple
piles is the observation that the pressure distribution beneath the pile shows a marked local
minimum beneath the apex which is counter-intuitive as this should be the location expected to
have the maximum pressure. Numerous experimental, analytical and computational studies have
been conducted to investigate this classical problem over the last few decades, but a
comprehensive understanding of the problem remains elusive. A number of recent finite element
simulations of the pile have considered the effects of construction history, plasticity and stress-
dependence of modulus of the granular solids. Whilst a pressure dip beneath the apex has been
predicted, significant uncertainties remain about the effects of these factors on the pressure dip
and their interaction.

This paper presents the finite element modelling of a conical stockpile using Abaqus. The effect of
construction history was realized by simulating the progressive formation of the conical pile. This
was achieved by discretising the final geometry of the stockpile into multiple conical layers and
then activating each layer sequentially. The effects of the elastic and plastic parameters were
explored. The results show that a pressure dip may or may not be predicted depending on the
constitutive model and the values for the model parameters. The study also shows that modelling
the conical pile in one single step does not produce the pressure dip. It further shows that the
central pressure dip is predicted using a relatively small number of layers and the magnitude of
the dip is not sensitive to increasing number of layers, which is in contrast with one previous
study.
Keywords: Sandpile, Stockpile, Stress distribution, Pressure dip, Pressure dependent modulus,
Progressive layering, Incremental construction.

1. Introduction

The behaviour of granular solids has attracted much attention of researchers from many
communities such as Applied Mechanics, Geotechnical Engineering, Chemical Engineering,
Materials Handling, Agricultural Engineering and Geophysics. The storage and handling of
granular materials is essential to many industries (Nedderman, 1992). Where the material is held

2009 SIMULIA Customer Conference 1

in very large quantities, it is often stored in a stockpile, formed by dumping the solid (e.g. coal and
mineral ore) to form a pile whose overall shape is typically conical, but may be prismatic,
depending on the method of placement (Figure 1). The solid is often recovered from the stockpile
using a conveyor beneath its centre. In such a case, the structure containing the conveyor needs to
withstand the pressures exerted by the stockpile. Consequently one key aspect of stockpile design
is to determine the pressure pattern beneath it. The experimental finding that there is a significant
local reduction in pressure (Figure 2) beneath the apex of the pile below the values one might
expect can have a strong impact on the design requirements. This reduction in pressure which is
commonly known as the “sandpile” problem has mostly been studied in the past as an interesting
scientific anomaly, but the stockpiles make it of considerable economic importance.

Figure 1. A typical industrial stockpile

6 6 330
Vertical pressure profile
5 Surface profile
5 275
Vertical pressure (kPa)
Vertical pressure (kPa)

4
4 220

Height (mm)
3
3 165
2
2 110
1
1 55
0
h=0.2m h=0.3m h=0.4m
h=0.5m h=0.55m hmax 0 0
-1
-1 -0.5 0 0.5 1 0 0.2 0.4 0.6 0.8 1
Radial position (m) Normalized radius r/R

a) Smid and Novasad (1981) b) Ooi et al. (2008)

Figure 2. Vertical base pressure underneath a granular pile

Despite extensive studies by both the physics and engineering communities over several decades,
a comprehensive understanding of the counter-intuitive phenomenon of the pressure dip remains

2 2009 SIMULIA Customer Conference

elusive. Good reviews of previous analytical, numerical and experimental studies of the problem
can be found in Savage (1997, 1998) and Cates et al. (1998). Among these studies, very few
adopted the finite element method (FEM), even though it is very powerful and flexible in dealing
with complex loading and boundary conditions. This paper investigates the feasibility of adopting
the FEM to model the sandpile problem and explores how the observation from such modelling
may provide further insight into the physical mechanism of the problem from a continuum point of
view.
It has been observed experimentally that the pile construction history is an important factor in the
occurrence of central pressure dip beneath the stockpile (Vanel et al., 1999; Geng et al., 2001).
For a sandpile formed by distributed deposition (“raining procedure”), no dip was found while for
a concentrated deposition (also termed “localized source” or “funnel procedure”), a pronounced
dip was observed. It has also been observed that base deflection (e.g. Trollope, 1957; Lee and
Herington, 1971) and spatial variation of material stiffness (as introduced in Savage, 1997) can
have a significant influence on the extent of the pressure dip. The present study is concerned with
the general case that a conical pile is formed on a rigid flat and rough base by concentrated
deposition. Finite element simulations of such a general case have been conducted by several
researchers. These studies are summarised as below.
Savage (1998) reported results from elastic and elastic-rigid plastic finite element computations
for wedges and cones using three different finite element analysis (FEA) packages including
Abaqus. They adopted an elastic-rigid plastic Mohr-Coulomb model and modelled the pile using
the 8-node quadrilateral element. Many calculations were undertaken to investigate the effects of
the internal friction angle , dilation angle , Poisson’s ratio , cohesion c and elastic modulus E
of the granular solid. It was found that the results were relatively insensitive to all of these
parameters except . The predicted vertical base pressure distribution showed no central dip and
was almost indistinguishable from the active limit state solution. Simulations using purely elastic,
Drucker-Prager and Drucker-Prager/Cap constitutive models produced similar predictions.
Anand and Gu (2000) conducted elastic-plastic calculations of a static conical granular pile with
an angle of repose of =31.5° using the “double-shearing” constitutive model through a user-
material subroutine in Abaqus/EXPLICIT. Two sets of parameters were investigated: one with a
constant internal friction angle of 31.5º, the other with a mobilized internal friction angle which
evolved from 0º to 30.0º with strain hardening. The former produced a state with no plastic
deformation and the vertical stress distribution showed a peak under the apex. The latter generated
a fully plastic state and the vertical stress distribution showed a pronounced dip under the apex of
the conical pile. They concluded that the cause of the dip was the nonhomogeneous plastic strain
occurred during the formation of a sand pile, which resulted in a nonuniform internal friction
coefficient. The largest plastic shear strain concentrated in a wide inclined band which lies slightly
below the pile surface.
Al Hattamleh et al. (2005) also argued that strain localization is the main cause of the pressure dip.
In their model, the construction of the granular heap was simulated by incrementally layering in
five stages. They adopted a double-slip formulation of “double-shear-type” constitutive model
which is similar to that of Anand and Gu (2000) but permits the user to assign the orientations of
initial slip lines. Very pronounced stress dips were predicted in all cases except the case of

2009 SIMULIA Customer Conference 3

homogeneous state where the initial slip orientations equal ±/4±/2. They predicted localized
vertical plastic strain around the apex with the rest of the pile in an elastic state.
Tejchman and Wu (2008) analysed the pressure distribution under both prismatic and conical
sandpiles using a micro-hypoplasticity model which considers the effect of the direction of
deformation rate. The construction of the pile was simulated in ten stages with two different
methods: horizontal layers and inclined layers, namely the raining procedure and funnel
procedure. The results were in qualitative agreement with the experimental results reported by
Vanel et al. (Vanel et al., 1999) and the numerical results reported by Al Hattamleh et al. (2005).
Jeong (2005) conducted an extensive FE study on sandpiles using several loading and boundary
conditions. In contrast to the other recent studies that adopted complicated constitutive relations, a
simple elastic-rigid plastic Mohr-Coulomb model was deployed. A significant effort was invested
in developing the “incremental construction” scheme in which the pile was constructed in many
stages, which has been used by the silo research community (e.g. Yu, 2004) and Geotechnical
Engineering (e.g. Clough and Woodward, 1967; Kerry Rowe and Skinner, 2001). This procedure
was shown to be capable of producing a central pressure dip underneath a conical pile. Jeong
(2005) also observed that the results are very sensitive to the number of construction layers
adopted. A larger number of construction layers predict a bigger pressure dip. The plastic zone in
the final stage was predicted to occupy the majority of the pile except the part near the base and
the tail ends of the conical surface.
Although some FE simulations of stockpile reported above have successfully predicted a pressure
dip, they differ considerably in both the analysis procedure and the final distribution of plastic
strain. Some adopted very complicated constitutive models (Anand and Gu, 2000; Al Hattamleh et
al., 2005; Tejchman and Wu, 2008), while others used more general elastic-rigid plastic model
(Jeong, 2005). Strain localization was argued in two of these studies to be the origin of stress dip
(Anand and Gu, 2000; Al Hattamleh et al., 2005), while incremental construction was shown to be
a key issue in another two studies (Jeong, 2005; Tejchman and Wu, 2008). There might be some
links between these two mechanisms, but they are still unclear.
This study attempts to evaluate the capability of simple constitutive models in predicting the
pressure dip in conical stockpiles and to investigate in detail the effect of incremental construction
scheme on the pressure profile. The evolution of stress distribution during progressive formation
of the stockpile is also shown. The results provide further insight on the potential mechanisms
responsible for the pressure dip using such modelling scheme.

2. Reference test data

The stockpile tests conducted by Ooi et al. (2008) with mini iron pellets centrally poured on a
rigid base were used as reference data in this study. The pellets are approximately spherical and
have a relatively uniform bulk density that is relatively insensitive to packing: the loosest and
densest bulk densities achieved in control tests being 2260 and 2370 kg/m3. The internal angle of
friction for the pellets was measured to be 34º using a direct shear tester. Five repeat tests were
conducted producing a mean pile radius at the base of Rp = 554mm and an average angle of repose
of 29.0º. Free-field pressure cells were used to record the normal base pressures underneath the

4 2009 SIMULIA Customer Conference

pile. Figure 2b shows the normal base pressure distribution at the final stage averaged from 5
repeat tests, where a pressure dip in the centre is clearly evident. The average height above the
base at different radial positions is also shown.

3. Finite element modelling

3.1 Constitutive models

In this study, relatively simple elastic and elastic-rigid plastic constitutive relations are used in
modelling the sandpile. These include linear elasticity (LE), pressure-dependent elasticity (PDE),
linear elasticity with Mohr-Coulomb rigid-plasticity (LEMC) and pressure-dependent elasticity
with Mohr-Coulomb rigid-plasticity (PDEMC). By adopting simple elastic-plastic models, the
role of elastic and plastic parameters in producing the numerical solution can be explored more
clearly.
The PDE model used is a Janbu-type relation (Janbu, 1963; Chen and Mizuno, 1990) which is
expressed as:
m
Et  p
 K   (1)
Pa  Pa 
where Et is the tangent modulus of elasticity of the granular solids, Pa is the atmospheric pressure
(101.3 kPa), p=(1+2+3)/3 is the mean pressure, and K and m are experimentally determined
parameters. Because such a PDE relation is not readily available in Abaqus, it was implemented as
a solution-dependent modulus based on the LE model through the user-subroutine USDFLD. Note
that the LE model in Abaqus only accepts the secant modulus. Consequently, Equation 1 is
transformed to the following form in terms of the secant modulus and implemented in Abaqus:
m
Es  p
 K s   (2)
Pa  Pa 
where:

K s  K 1  m  (3)

Because the routine USDFLD provides access to material point quantities only at the start of each
numerical time increment, the solution clearly depends on the time increment size or the number
of time increment because the material properties remain constant during each increment.
Numerical calibration tests were conducted to ensure that this PDE relation was correctly
implemented. A frictionless uniaxial compression test with a radius of r=1.0m and a height of
z=1.0m was modelled (Fig. 3a). The parameters were chosen as Ks =100 and m=0.4. To avoid
numerical difficulties caused by zero elastic modulus at the beginning of the compression, a small

2009 SIMULIA Customer Conference 5

initial elastic modulus of E0=0.5MPa was adopted. In this uniaxial compression test, the relation
of vertical stress z and vertical strain z for an elastic material can be derived as:

m
 z / pa  1    2 2 
 K s   / 1   (4)
z  31     1  
where is the Poisson’s ratio of the solid. The comparison between the input PDE relation and the
computation outputs using different number of time (loading) increments Nt is shown in Figure 3b.
It is shown that the output curve from the explicit solution approaches the input curve quickly
when number of increments increases. Similar convergence tests were also performed for the pile
simulations to ensure accurate implementation of the PDE nonlinear elastic treatment.

0.05
Vertical stress (×Pair ) Input
Nt=7
0.04 Nt=13
Nt=50
Nt=250
0.03

z 0.02

r 0.01

0
0 0.1 0.2 0.3 0.4 0.5 0.6
-3
Vertical strain (× 10 )

a) Uniaxial compression test b) Input versus predicted stress-strain

relationship

Figure 3. Verification of Janbu elasticity user-subroutine

3.2 Problem configuration

Assuming axisymmetry, the conical sandpile was simulated as a triangle in two dimensions. The
final sandpile geometry was further discretised into triangular elements using the quadratic 6-node
triangular axisymmetic element CAX6. The only load considered is the self-weight of the solids.
The bottom boundary of the pile was fixed in both vertical and horizontal directions, representing
a rigid and completely rough base. The simulation process was treated as a static problem so the
effect of inertia was neglected.

6 2009 SIMULIA Customer Conference

3.3 Incremental construction scheme

The effect of construction history due to the progressive loading of the conical pile was explored
by modelling the progressive formation (or incremental construction) of the conical pile. This was
achieved by discretising the final geometry of the pile into many conical layers and then activating
each layer sequentially, starting from the bottommost layer. This incremental construction process
was implemented in Abaqus by using the element removal and reactivation technique through the
Model Change keyword. A sketch of the incremental construction with FE mesh arrangements is
shown in Figure 4 where the final geometry of pile is divided into several construction layers
(denoted by alternative dark and light grey layers). Each construction layer may contain one
(Figure 4a) or more layers (Figure 4b) of elements. As a result, the total number of layers of
element Nel is a multiple of number of construction layers NIC. In the present study, the sensitivity
of numerical results to both these values ranging from 1 to 60 has been explored.

a) FE mesh with Nel=NIC =5 b) FE mesh with Nel=15, NIC =5

Figure 4. Sketch of incremental construction for sandpile

3.4 Input parameters

Unless stated otherwise, the input parameters adopted in all the simulations are listed in Table 1
based on the experimental data described above. The density was chosen as the minimum value
from the control tests. This is a reasonable assumption because the particles in a stockpile undergo
avalanching during the formation process leading to a relatively loose packing. As suggested by
Jeong (2005), the particles are in a state of constant volume condition during avalanches, which
corresponds to a Poisson’s ratio of around 0.5. As a result, a large Poisson’s ratio of 0.45 was
chosen. The pellets tested were dry and non-cohesive, but a small value of cohesion of 1 Pa is
assumed to avoid numerical difficulties.

Table 1. Input parameters for sandpile simulations

Density Angle of repose Angle of internal friction Angle of dilation Cohesion
    c
3
2260 (kg/m ) 29° 34° 20° 1 (Pa)
Elastic modulus Poisson’s ratio Initial elastic modulus Ks m
E  for Janbu relation E0
2.0 (MPa) 0.45 0.5 (MPa) 100 0.4

2009 SIMULIA Customer Conference 7

4. Key results

Two groups of simulations were conducted: one without incremental construction and the other
adopting incremental construction. Figures 5a and b show respectively the predicted vertical base
pressure distributions underneath the pile for the two groups. The pressure has been normalised by
the hydrostatic pressure under the apex at the base p=H where H is the height of the pile. When
the whole sandpile was analysed as a single layer (Fig 5a), none of the four constitutive models
produced a central dip. This result concurs with the conclusion by Savage (1998). The base
pressure is lower than the hydrostatic value towards the centre (at radius r=0) and slightly higher
than the hydrostatic value elsewhere, satisfying the global equilibrium in the vertical direction. It
is interesting to note that the linear elastic LE model and the non-linear elastic PDE model
produced very similar results whilst the two elastic-plastic models LEMC and PDEMC also
produced very similar results, but the introduction of plasticity has further increased the shedding
of the load away from the centre.

1 1
Hydrostatic Hydrostatic
0.9 0.9
Normalized vertical pressure
Normalized vertical pressure

LE LE+IC
0.8 PDE 0.8 PDE+IC
0.7 LEMC 0.7 LEMC+IC
PDEMC 0.6 PDEMC+IC
0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Normalized radial position Normalized radial position

a) Without incremental construction b) With incremental construction

Figure 5. Vertical base pressure from different constitutive relations (Nel=NIC =20)

The inclusion of incremental construction produced different effects in each constitutive model, as
shown in Figure 5b. The linear elastic-plastic LEMC model produced a shallow dip in pressure
whilst the non-linear elastic-plastic PDEMC model produced the most pronounced dip. Both
elastic models (LE and PDE) did not predict any dip in pressure. The simulation using incremental
construction appears to give rise to a larger shedding of the load away from the centre, thus giving
rise to the manifestation of a pressure dip. The results also support the proposition that material

8 2009 SIMULIA Customer Conference

plasticity is a requirement for predicting the sandpile phenomenon and the progressive loading
history during sandpile formation also plays a vital role.

1 1
Hydrostatic Hydrostatic
0.9 0.9
Normalized vertical pressure

Normalized vertical pressure

NIC=5 NIC=5
0.8 NIC=10 0.8 NIC=10
0.7 NIC=20 0.7 NIC=20
NIC=40 NIC=40
0.6 0.6
NIC=60 NIC=60
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Normalized radial position Normalized radial position

a) Central point included b) Central point excluded

Figure 6. Effect of number of construction layers on vertical stress distribution

with PDEMC (Nel=NIC)

It is noted that the computed base pressure at the centre of the pile on the axis of symmetry shows
a larger value that disrupts the smoothness of each curve, except the linear elastic case. Further
calculations were carried out exploring different number of construction layers and element layers
to investigate the origin of this problem. It was observed that this larger pressure at the central
node is always confined to only within the first column of elements next to the central axis. The
source of the problem may thus be associated with the computation of the nodal stress in the
elements containing nodes where zero radial displacement is imposed and stepwise incremental
construction scheme is used. This issue is further illustrated in Figure 6a which shows five
calculations with increasing number of construction layers. The nodal pressure at the centre was
always predicted to be larger no matter how many incremental layers were used. This issue is
under further investigation. Figure 6b shows the same results as Fig. 6a with the central point of
data omitted. It is evident that this very local occurrence does not influence the overall prediction,
so in the rest of this paper, the central point will not be plotted. It should be noted that no previous
FEA studies have reported this phenomenon.
Figure 6b reveals that the shape of predicted pressure distribution and size of central pressure dip
are not sensitive to the number of construction layers in the range tested in this study (NIC = 5 to
60). This is contradictory to the observation of Jeong (2005) where the size of the dip increases
with an increasing number of construction layers. In another independent study of progressive
filling silo wall pressure using a similar incremental scheme, Yu (2004) concluded that the

2009 SIMULIA Customer Conference 9

calculated peak wall pressure converged quickly and was not sensitive to the number of filling
layers used, which is in agreement with the observation here.
Since the nonlinear elastic-plastic PDEMC model produced the best prediction, the rest of this
paper will focus on the results using the PDEMC constitutive model with incremental
construction. Figures 7-9 show the FEA results using 40 layers of elements and 40 construction
layers (NIC=Nel=40). Figure 7a shows the vertical stress along horizontal paths at different heights
in the pile at the final stage of construction. The FEA predicted that the central dip in vertical
stress also exists within the pile but it reduces quickly from the base upwards. This conclusion is
consistent to that drawn by Anand and Gu (2000). Figure 7b shows the evolution of the normal
base pressure during the incremental simulation process. It is clear that the pressure dip is
experienced throughout the pile formation process from the very beginning.

1 1
Hydrostatic Hydrostatic
0.9 0.9
Normalized vertical pressure

Normalized vertical pressure

h=0 Layers=5
0.8 h=0.16H 0.8 Layers=10
h=0.33H Layers=20
0.7 0.7
h=0.49H Layers=30
0.6 h=0.65H 0.6 Layers=40
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Normalized radial position Normalized radial position

a) Final vertical stress at different height b) Base pressure at different stages

during construction

Figure 7. Vertical stress along horizontal paths using PDEMC elastic-plastic

model

The contours of the vertical stress v and the mean pressure p=(1+2+3)/3 for the whole pile
are shown in Figures 8a and b respectively. Because the tangent elastic modulus is dependent on
the mean pressure according to Eq. 1, the variation of the elastic stiffness in this model is directly
related to the mean pressure. The stiffness is predicted to be increasing with depth as one would
expect. More importantly, the largest stiffness at each level is some radial distance away from the
centre, giving rise to a softer central core surrounded by stiffer surrounding regions. The analysis
has thus identified an arching mechanism arising from the stress dependency of the bulk stiffness
in which the vertical load is attracted to the stiffer zone, away from the softer zone.

10 2009 SIMULIA Customer Conference

 a) Vertical stress b) Mean pressure

Figure 8. Stress distribution in pile (Pa)

1
Hydrostatic
0.9 Test (Ooi et al., 2008)
Normalized vertical pressure

0.8 FEA (PDE+IC)

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
Normalized radial position

Figure 9. Comparison between FEM predictions and test results

The predicted normal base pressure distribution is compared with the experimental result in Figure
9. The FEA predicted a smaller dip than observed in the experiments. One of the possible causes
for his discrepancy is that the test piles were slightly rounded at the top due to the impact of the
pouring pellets whilst a perfect conical pile was assumed in the numerical simulations, resulting in
a smaller apex height in the actual pile than in the numerical simulation. It is also possible that a
better match can be produced by varying some of the input parameters including the dilation
angle, the Poisson’s ratio and the Janbu elastic parameters. Further parametric investigation is
being undertaken and will be reported elsewhere.

2009 SIMULIA Customer Conference

11
5. Conclusion

A finite element analysis of a conical sandpile has been presented and compared with
experimental observations. The key aspects of modelling a sandpile using the finite element
method have been discussed and the outcomes for several elastic and elastic-plastic models, with
and without incremental layer construction scheme, have been presented. The chief conclusions
of the study are:
1. Incorporating plasticity and simulating the progressive construction of the sandpile are both
necessary for the finite element method to predict the classic sandpile pressure distribution
where a significant dip exists beneath the apex of the pile.
2. Whilst the FE calculations have produced a reasonable prediction of the pressure distribution,
they under predict the size of the central dip. A closer match may be achieved by modelling
the actual shape of the test pile and adjusting the input parameters.
3. The size of the pressure dip and the overall pressure profile are found to be insensitive to the
number of construction layers used. This contrasts with the observation of Jeong (2004)
where the dip was reported to become larger as the number of layers increases.
4. The largest stiffness was predicted to be some radial distance away from the centre, giving
rise to a softer central core surrounded by stiffer surrounding regions. The analysis has thus
identified an arching mechanism arising from the stress dependency of the material stiffness
in which the vertical load is attracted to the stiffer zones, away from the softer central zone.

6. Acknowledgement

We acknowledge the support from the Scottish Funding Council for the Joint Research Institute
with the Heriot-Watt University which is a part of the Edinburgh Research Partnership in
Engineering and Mathematics (ERPem), and from the EPSRC (grant GR/T23541). J. Ai was
further supported by a University of Edinburgh Scholarship.

7. References

1. Al Hattamleh, O., B. Muhunthan, and H. M. Zbib, "Stress distribution in granular heaps using
multi-slip formulation," International Journal for Numerical and Analytical Methods in
Geomechanics, 29(7), pp. 713-727, 2005.
2. Anand, L., and C. Gu, "Granular materials: constitutive equations and strain localization,"
Journal of the Mechanics and Physics of Solids, 48(8), pp. 1701-1733, 2000.
3. Cates, M. E., J. P. Wittmer, J. P. Bouchaud, and P. Claudin, "Development of stresses in
cohesionless poured sand," Philosophical Transactions: Mathematical, Physical and
Engineering Sciences (Series A), 1998(1747), pp. 2535-2560, 1998.

12 2009 SIMULIA Customer Conference

 4. Chen, W.-F., and E. Mizuno, Nonlinear analysis in soil mechanics: theory and
implementation. Elsevier Science Publishers, Amsterdam, the Netherlands, 1990.
5. Clough, W., and J. Woodward, "Analysis of Embankment stresses and deformations," ASCE
J. Soil Mech. and Found. Div., 93(4), pp. 529-549, 1967.
6. Geng, J., E. Longhi, R. P. Behringer, and D. W. Howell, "Memory in two-dimensional heap
experiments," Physical Review E, 64(6), pp. 060301, 2001.
7. Janbu, N., "Soil compressibility as determined by oedometer and triaxial tests," Eur. Conf.
Soil Mech. Found. Eng., Wiesbaden, Germany, 19-25, 1963.
8. Jeong, H. Y., "Numerical simulation of stresses under stockpiled mass over ground with or
without a loadout tunnel," Ph.D. Thesis, The University of Western Ontario, 2005.
9. Kerry Rowe, R., and G. Skinner, "Numerical analysis of geosynthetic reinforced retaining
wall constructed on a layered soil foundation," Geotextiles and Geomembranes, 19(7), pp.
387-412, 2001.
10. Lee, I. K., and J. R. Herington, "Stresses beneath granular embankments," Proceedings of the
1st Australia-New Zealand conference on geomechanics 1, Melbourne, 291-296, 1971.
11. Nedderman, R. M., Statics and Kinematics of Granular Materials. Cambridge University
Press, Cambridge, U.K., 1992.
12. Ooi, J. Y., J. AI, Z. Zhong, J. F. Chen, and J. M. Rotter, Eds., Progressive pressure
measurements beneath a granular pile with and without base deflection. Structures and
granular solids: from scientific principles to engineering applications, CRC Press, London,
2008.
13. Savage, S. B., "Problems in the statics and dynamics of granular materials," Powders and
Grains 97, Balkema, Rotterdam, Netherlands, 185, 1997.
14. Savage, S. B., "Modeling and granular material boundary value problems," Physics of Dry
Granular Media, Kluwer Academic publishers, 25-96, 1998.
15. Smid, J., and J. Novosad, "Pressure distribution under heaped bulk solids," Proceedings of
1981 Powtech. Conf., Ind. Chem.Eng.Symp., 63, 1981.
16. Tejchman, J., and W. Wu, "FE-calculations of stress distribution under prismatic and conical
sandpiles within hypoplasticity," Granular Matter, 10(5), pp. 399-405, 2008.
17. Trollope, D. H., "The systematic arching theory applied to the stability analysis of
embankments," Proc. 4th Int. Conf. Soil Mech. Found. Engng, London, 2, pp. 382-88, 1957.
18. Vanel, L., D. Howell, D. Clark, R. P. Behringer, and E. Clement, "Memories in sand:
Experimental tests of construction history on stress distributions under sandpiles," Physical
Review E, 60(5), pp. R5040, 1999.
19. Yu, S. K., "Finite element prediction of wall pressures in silos," Ph.D. Thesis, Built
Environment Research Unit, The University of Wolverhampton, Wolverhampton, U.K., 2004.

2009 SIMULIA Customer Conference

13

View publication stats