Computers and Geotechnics 91 (2017) 146160

Contents lists available at ScienceDirect

Computers and Geotechnics

journal homepage: www.elsevier.com/locate/compgeo

Research Paper

A Modified Mohr-Coulomb Model to Simulate the Behavior of Pipelines

in Unsaturated Soils
D.J. Robert
School of Engineering, RMIT University, Melbourne, VIC 3001, Australia

a r t i c l e i n f o a b s t r a c t

Article history: At the present time, it is very common in practice to utilize Mohr-Coulomb model to simulate the soil
Received 5 January 2017 behaviour in the application of soil-pipeline interaction problems. However, the traditional Mohr-
Received in revised form 23 May 2017 Coulomb model is unable to predict the realistic loading that can apply on buried pipes during large
Accepted 10 July 2017
ground deformation. Especially, the linear elastic-perfectly plastic Mohr-Coulomb model is not capable
of simulating the unsaturated soil loading which can result larger than anticipated loading due to suction
induced additional normal force between the soil particles. A user defined unsaturated modified Mohr-
Keywords:
Coulomb model is developed within a generalized effective stress framework considering suction hard-
Pipelines
Unsaturated soils
ening effects to capture the realistic loading induced by unsaturated soil medium. Firstly, the model has
Unsaturated modified Mohr-Coulomb been developed considering microscopic and macroscopic suction hardening mechanisms, and was
model implemented into a commercial finite element program associated with user subroutine written in
Triaxial test FORTAN. Then the model was validated through a series of unsaturated triaxial compression tests con-
Large scale test ducted on the basis of different sand types having various initial conditions. Finally, the model has been
User subroutine applied to simulate the behaviour of pipelines subjected to lateral soil loading in unsaturated soils. The
results revealed that the modified Mohr-Coulomb model has reasonable predictions when compared to
the load-displacement response of pipes obtained from two large scale testing programs. The developed
model can be used to predict the increased strength and stiffness associated with soil suction that
increases lateral loads on pipelines, and thus has widespread relevance for simulating the pipeline
response in unsaturated soils under externally imposed ground movement.
2017 Elsevier Ltd. All rights reserved.

1. Introduction The investigations on the behaviour of pipes subjected to exter-

Pipelines used for the transport of energy and services are
important lifelines in the modern society. The vital role that they
play in the present economy is reflected by many kilometers of
pipelines laid across large geographical areas. They are often bur-
ied underground for protection and support, but large ground
deformations can induce catastrophic failures on the pipeline. Pro-
gressive ground deformations such as caused by landslides or
earthquakes can induce pipeline deformation impacting service-
ability or trigger failures such as buckling or wrinkling. The earth-
quakes occurred in Turkey (Kocaeli), Taiwan (Chi-Chi) and Japan
(Sanriku and Niigataken Chuetsu-oki Earthquake) offer numerous
examples of the damage to the pipelines (Fig. 1). The damage
caused by ground deformation can be further enhanced by the
presence of suction in the unsaturated medium where the pipes
are usually constructed.
E-mail address: dilan.robert@rmit.edu.au
nal loading have been extensively studied for the last 10 decades
[21,40,19,10,4,40,16]. However, little attention is devoted until
recently on the investigations of moisture influence of buried pipes
subjected to large ground deformation. The recent large scale
experiments conducted at Cornell University, US [26,24,25,15,27]
and Tokyo Gas Co Ltd, Japan [29] on pipes buried in unsaturated
soil were able to highlight the significance of considering soil sat-
uration when assessing the pipe responses under large ground
deformation. For example, the results from the large scale tests
revealed that the response of buried pipes under external loading
can significantly be influenced by the partial saturation of soil
depending on the soil type, leading modifications to current design
standards that were developed on the basis of pipes tested in
dry/fully saturated tests. These studies will assist for better
understanding of the behaviour of soil-pipeline system, thus to
provide backbone in the response predictions of buried pipes using
numerical modelling approaches.
The development of numerical analysis and its application to
geotechnical engineering problems over the past 20 years have

http://dx.doi.org/10.1016/j.compgeo.2017.07.004
0266-352X/ 2017 Elsevier Ltd. All rights reserved.
 D.J. Robert / Computers and Geotechnics 91 (2017) 146160
(a) 1994 Sanriku earthquake in
Japan (Koseki et al., 1998)
(c) 1999 Kocaeli earthquake in
Turkey (EERI, 1999)
Fig. 1. Failures of steel pipelines due to earthquakes. (See above-mentioned references for further information.)
provided geotechnical engineers with an extremely powerful anal-
ysis tool [28]. They facilitate to assess the response of pipes consid-
ering wide range of variables in timely and efficient manner. The
most recent research work conducted in the area of numerical
modelling of soil-pipeline interaction problems has been able to
highlight the development of proper numerical tools to capture
the realistic behaviour of pipelines when subjected to ground
movements. For example, Yimsiri et al. [46] have adopted state
dependent hardening/softening model (Nor-Sand [12]) to study
the behaviour of pipelines subjected to lateral ground deformation.
Cheong et al. [9] also used the Nor-Sand model to investigate the
response of elbow pipelines subjected to lateral ground deforma-
tion. Chatterjee [7] has adopted modified Cam-Clay model [36] to
quantify the lateral resistance of pipes in soft soil. Bryden et al.
[6] have used hyperbolic hardening soil model, which can capture
the stress-dependent variation in stiffness to study the effect of
surface loading on buried pipelines. The uses of such advanced soil
constitutive models were effective for predicting the behaviour of
pipelines due to their superior capability of capturing the load
dependent material state changes. However, in most occasions,
they demand comprehensive calibration of soil parameters using
detailed and advanced testing programs prior to use in the desired
application. On the other hand, it is very common in practice to uti-
lize standard material models such as Mohr-Coulomb model to
simulate the soil behaviour in the application of soil-structure
147
(b) 1999 Chi-Chi earthquake in Taiwan
(provided by Tokyo Gas Co. Ltd., Japan)
(d) 2007 Niigataken Chuetsu-oki
Earthquake in Japan (Provided by Tokyo
Gas Co. Ltd., Japan)
interaction problems [8,44,1,37,15]. These models are often chosen
considering its simplicity, ease of use, reasonable computational
time and the high level of understanding among the engineers.
Such models are readily available in commercial finite element
programs so that researchers/practitioners can use them at less
cost. However, they have their own limitations which can signifi-
cantly influence the prediction of buried pipeline response unless
appropriate model changes are incorporated.
The linear elastic-perfectly plastic Mohr-Coulomb model is
unable to predict the realistic loading that can apply on buried
structures during large ground deformation. Especially, the original
Mohr-Coulomb model is not applicable to simulate the unsatu-
rated soil loading which can result larger than anticipated loading
due to suction induced additional normal force between soil parti-
cles. Recent studies showed that soil suction under unsaturated
condition can change the stiffness and strength, affecting the beha-
viour of the pipeline systems subjected to ground deformation
[25,15,38,33,35,31]. Therefore, utilization of traditional Mohr-
Coulomb failure criterion could lead to under-estimation of soil
loads on pipes, thus could result in unexpected pipeline failures
in the events of construction or earthquake induced ground defor-
mations. This study showed that Mohr-Coulomb model could well
be used to predict realistic pipeline loading if it can be modified to
include the suction into the constitutive modelling framework. A
user defined unsaturated modified Mohr-Coulomb model is
148 D.J. Robert / Computers and Geotechnics 91 (2017) 146160

developed within a generalized effective stress framework includ-
ing suction hardening effect. Firstly, the model has been developed
considering microscopic and macroscopic suction hardening
mechanisms, and was implemented into a commercial finite ele-
ment program associated with user subroutine written in FORTAN.
Then the model was validated through a series of unsaturated tri-
axial compression tests conducted on the basis of different sand
types having various initial conditions. Finally, the model has been
applied to simulate the behaviour of pipelines subjected lateral soil
loading in unsaturated soils. The results revealed that the modified
Mohr-Coulomb model has reasonable predictions when compared
to the load-displacement response of pipes obtained from two
large scale testing programs. The developed model can be used
to predict the increased strength and stiffness associated with soil
suction that increases lateral loads on pipelines, and thus has wide-
spread relevance for simulating the pipeline response under exter-
nally imposed ground movement.
2. Mohr-Coulomb model
Mohr-Coulomb model is a simple linear elastic-perfectly plastic
model which is widely used for the design applications in geotech-
nical engineering to simulate material response under monotonic
loading. The model is widely popular in the community for mod-
elling the behaviour of soils due to its relative simplicity and the
requirement of the basic soil properties (such as friction and dila-
tion of soils).
Within the Mohr-Coulomb (MC) framework, the soil is mod-
elled as an isotropic dilatant elastic-perfectly plastic material. Elas-
tic behaviour is modelled assuming linear isotropic elasticity of
which the stress is defined from the elastic strain as;
r DE ee 1 softening is completed. Calibration of cpF (defined by Eq. (4)) is
where r is the stress, DE is the fourth-order elasticity tensor, and ee
is the elastic strain.
The model behaves elastically and obeys Hookes law until the
onset of yielding which is determined by the Mohr-Coulomb yield
criterion. For general states of stress, the yield criterion of the
model in terms of three stress invariants is defined as (p: compres-
sion negative) [2];
q
P ec0 tan w2 Rmw q2  p tan w 3
where
Rmw H; e
41  e2 cos2 H 2e  12
p Rmc p=3; /0
21  e2 cos H 2e  1 41  e2 cos H 5e2  4e
3  sin /0
Rmc p=3; /0
6 cos /0
where /0 and w are the friction angle and the dilation angle mea-
sured in the p  Rmw q plane, respectively. c0 is the initial cohesion
yield stress; H remains as defined in yield criterion; e is a parameter
which is referred to as meridional eccentricity which defines the
rate at which the hyperbolic function approaches the asymptote
(the flow potential becomes a straight line in the meridional stress
plane as the meridional eccentricity becomes zero); and e is a
parameter referred to as the deviatoric eccentricity [2] that
describes the out-of-roundedness of the deviatoric section in terms
of the ratio between shear stress along the extension meridian
H 0 and shear stress along the compression meridian
H p=3.
The original Mohr-Coulomb model can be modified to capture
the strain softening behaviour of the material at large deforma-
tions [32]. The softening behaviour is captured by reducing the
mobilized friction and dilation angle with an increase in plastic
deviatoric shear strains [3] as shown in Fig. 2, where /0max and
/crit are the peak friction angle and critical state friction angle,
respectively, and wmax is the ultimate dilation angle, wres residual
dilation angle, cpF is the plastic deviatoric shear strain when the
required on the basis of direct shear tests in order to obtain a
mesh-independent solution. The procedure introduced by Anasta-
sopoulos et al. [3] can be adopted when deriving plastic shear
strain at the completion of softening cpF based on direct shear test
data. This approach can be effective for obtaining mesh size inde-
pendent solution for soil-pipeline interaction analysis involving
large ground deformations [32,30].

F Rmc q  p tan /0  c0 0 2 dxf  dxp dxp  dxy dxf  dxp

cpF cpp 4
where 16d50 D dFE

1 1
Rmc H; /0 p sinH p=3 cosH p=3 tan /0
3 cos /0 3

/0 is the friction angle (slope of the yield surface in p  Rmc q mob

'
, mob
stress planes)
c0 is the cohesion of the material
 3
H is the deviatoric polar angle defined as cos3H qr
max
'

p 1=3r11 r22 r33

r crit
3
q Sij Sij ;
2
where Sij is the deviatoric stress tensor, defined as Sij rij pdij max
where dij is Kronecker delta.

 1=3  1=3
9 9
r S:S : S Sij :Sjk : Ski
2 2
fp dev
p

The strain response follows a non-associative flow rule. A flow

function is defined with a smooth elliptic function, proposed by Fig. 2. Typical variation of mobilized friction and dilation angles with plastic
Menetrey and William [23] in the deviatoric stress plane as; deviatoric strains (Robert and Soga, 2010).
 D.J. Robert / Computers and Geotechnics 91 (2017) 146160 149

where
dFE - size of the element
D - thickness of the sample in direct shear test
dxy ; dxp and dxf are the horizontal displacements at yielding,
peak and completion of softening, respectively. c '' Effect cu' Effect
3. Modified Mohr-Coulomb model

In the current study, the modified Mohr-Coulomb model (MMC)

is developed within a generalized effective stress framework con-
sidering suction hardening effect. Following equation has been
used in this study to develop the stress-strain framework [39]. Fig. 3. Capillary cohesion effects on granular material.
 0   
r r Sw sdij
ij
net
ij eij
and 5
s ua  uw Sw cohesion in sands. The contributions from c0u and c00 for strength
and dilatancy characteristics of soils can depend on sand type
where r0ij is the Bishops effective stress tensor, rnet
ij is the net stress, (i.e. particle size distribution). The development of any constitutive
s is the matric suction, Sw is the degree of saturation and eij is the models which incorporates the unsaturated soil state should
strain tensor. include both the microscopic as well as macroscopic effect of
The genuine application of Bishops effective stress principle suctions.
simplifies the formulation of the model. Albeit some earlier criti- The yield and potential functions of the model are now gov-
cism (e.g. [13]) for the use of Bishops effective stress concept for erned by the additional state variable Sw . Hence, the extended con-
modelling the unsaturated soil response (in particular volume sistency condition of the modified Mohr-Coulomb model becomes;
change characteristics), recent research highlights
[5,14,45,42,39,18] showed that the inclusion of the product  T
Sw  s in the effective stress formulation makes the hydro- @F @F p @F
dr de dSw 0 11
mechanical coupling straightforward, featuring a direct modifica- @r @ eps s @Sw
tion in effective stress upon a change in suction. In this study, it
where
has been shown that the generalised effective stress framework  
is capable enough to provide accurate model predictions when @F p @F @/0 @F @w @F @c0
p des 0 p p p deps 12
compared with unsaturated soil element testing and large scale @ es @/ @ es @w @ es @c0 @ es
pipeline experiments. The proposed modified Mohr-Coulomb
Further, the variation of microscopic cohesion c0u and unsatu-
model has been developed by extending the linear-elastic perfectly
rated dilation wunsat in the Mohr-Coulomb model are captured as
plastic Mohr-Coulomb model.
a function of water saturation as well as plastic deviator strains. I.e.
The Mohr-Coulomb failure criterion assumes that failure occurs
when the shear stress on any point in a material reaches a value c0u co l 13
that depends linearly on the normal stress in the same plane as
defined in Eq. (6). wunsat wo l 14
sf c0 r0 tan /0 6 where co f Sw ; wo f Sw ; and l f cpdev
8
Mohr-Coulomb strength envelope for unsaturated soils can be >
> 1:0 for 0 6 cpdev 6 cpdev peak
>
>  p p 
obtained by defining the Mohr-Coulomb failure criteria (Eq. (6)) < c c
within the generalised effective stress framework (Eq. (5)); l f cpdev cp dev f cp for cpdev peak < cpdev < cpf 15
>
> v peak f
>
>
de
0
sf c r  ua f Sw ua  uw f  tan / 0
7 : 0:0 for cpf 6 cpdev 6 1:0
Rearranging Eq. (7); Incremental plastic shear strains deps is related to the plastic
0 0 0 potential function via the flow rule (Eq. (16)).
sf c r  ua f tan / Sw ua  uw f tan / 8
@P
Eq. (8) can be further rearranged as; deps k 16
@J
0 0 0
sf c Sw ua  uw f tan / r  ua f tan / where
c0 c00 r  ua f tan /0 9  1=2
1
J r11  r22 2 r22  r33 2 r33  r11 2 r212 r223 r231 
where 6
c0 c0s c0u 10:1 17

and Combining Eqs. (11), (12) and (16);

 T  

c00 Sw ua  uw f tan /0 10:2 @F @F @/0 @F @w @F @c0 @P @F

dr  0 p p p k  dS w
@r @/ @ es @w @ es @c0 @ es @J @Sw
The total cohesion of unsaturated soil can be represented as
18
c0 c00 which includes classical cohesion c0s and unsaturated
apparent cohesion c0u c00 . Unsaturated apparent cohesion results The incremental stress dr is also related to the elastic consti-
due to double suction effect; suction induced hardening occurring tutive matrix DE ; incremental total strain de and incremental
at the microscopic particle contact level c0u and suction induced plastic strain dep . Thus,
macroscopic stress Sw ua  uw  mobilized in terms of shearing
dr DE de  dep 19
resistance. Fig. 3 shows the development of unsaturated apparent
150 D.J. Robert / Computers and Geotechnics 91 (2017) 146160

dep is related to the plastic potential function via the flow rule. The changes in the degree of saturations (i.e. dSw ) were esti-
mated on the basis of volumetric strains under undrained condi-
@P
dep k 20 tion [34]. The matric suctions were estimated from the degree of
@r saturation using soil water retention curves. The water retention
Combining Eqs. (18)(20); curves were fitted with the van Genuchten model [43] for each
density. The numerical model simulates suction induced by satura-
 @F T
@r
DE de @S@Fw dSw tion changes corresponding to soil density changes. The model has
k  T  0
 21 been implemented using FORTRAN subroutine in the commercial
@F @F @/ @w @c0
@r
DE @@Pr  @/0 @ eps
@F
@w p
@ es
@F
@c 0 p
@ es
@P
@J finite element package ABAQUS.
Careful attention was devoted when defining the stress state of
the material after yielding. The approach proposed by Menetrey 4. Tested sands and model calibration
[22] was identified to be effective to keep the stress state of the
material exactly on the yield locus after it reaches yielding. 4.1. Sand description
An elastic predictor step is defined based on the previous con-
verged step so that ei1 ei De. This leads to the definition of The model has been tested for two different types of sands; A
the trial stress tensor (rtrial ) as defined by; coarser sand (called Cornell sand in this manuscript) having speci-
fic gravity, Gs = 2.69; maximum void ratio, emax = 0.70; minimum
rtrial ri DE De 22 void ratio, emin = 0.39, and a finer sand (called Chiba sand in this
A plastic corrector step is performed when the trial stress is out- manuscript) with Gs = 2.65, emax = 0.95 and emin = 0.50. The particle
side the yield surface. The purpose of the plastic corrector step is to size distributions of the two sands are shown in Fig. 4. The grain
restore the consistency by holding the stress state always at the sizes of Chiba sand are uniformly distributed within the range of
yield surface. I.e. ri1 is always on the yield surface if the stress 0.90.08 mm (coefficient of curvature, C c = 1.06; uniformity coeffi-
state is plastic. cient, C u = 2.1), whereas those of Cornell sand (C c = 0.8, C u = 3.3)
are relatively well graded. The percentage of finer particles with
@P grain size smaller than 0.42 mm is about 90% for Chiba sand and
ri1 rtrial  Drpcorrect rtrial  kDE 23
@r approximately 30% for Cornell sand. Hence, Cornell sand is coarser
In this approach, it has been assumed that the stress ri has with a broader distribution of grain sizes, whereas Chiba sand is
finer and well sorted. Due to the higher amount of fine particles
reached the yield surface and that the gradient @@Fr is calculated at
present in Chiba sand, larger amount of energy (suction) is
the stress ri . However if the stress tensor ri is inside the yield
required to withdraw a specified amount of water from the pore
locus and the trial stress is outside the yield surface, then the stress
spaces. Fig. 5 shows the drying retention curves (data obtained
on the yield surface has to be estimated. In order to avoid the
from [32]) for various densities of sands using suction plate tech-
non-linear computation of such contact stress, the truncated
nique [20].
Taylor series can be expanded around the trial stress [22]. The
stress increment to bring the stress state back into the yield surface
has been defined as dradjust . 4.2. Triaxial tests
The new truncated Taylor series can be expanded around the
trial stress rtrial as; A series of unsaturated triaxial tests were performed on the
basis of both the sands at various initial densities and confining
!T
@F trial @F trial p @F trial stresses. Results revealed that the characteristics of two sands
i1 trial
F F dradjust des dSw 0 24 are significantly different, and hence they are important for devel-
@r @ eps @Sw
oping a unified constitutive model that can capture the behaviour
Incremental stress adjustment (dradjust ) is related to the elastic of typical sands under partially saturated conditions. Results
constitutive matrix (DE ), incremental total strain adjustment showed that the maximum dilation angle wmax and the effective
(deadjust ) and incremental plastic strain adjustment (depadjust ) Thus, apparent cohesion c0u are strongly affected by the changes in
water saturation in Chiba sand, whereas such changes are not sig-
dradjust DE deadjust  depadjust 25 nificant in Cornell sands. For example, Fig. 6 shows the variation of
c0u with water saturation for Chiba sand and Cornell sand. There is
Combining Eqs. (23) and (24);
!T
i1 trial @F trial
F F DE deadjust  depadjust
@r
@F trial p @F trial
des dSw 0 26
@ eps @Sw
Stress is adjusted by keeping the total strain adjustment at zero;

i:e: deadjust 0 27

Using Eqs. (16), (18), (20), (24) and (25), kunsat

trial can be derived as;
trial
F trial @F@Sw dSw
kunsat
trial    trial 0  28
trial T trial
DE @@Pr  @F@/0 @@/ep @F@w @@w
trial trial
@F @c0 @Ptrial
@r ep
@F@c0 p
@ es @J
s s

The functional and derivatives terms used in the formulation

are given in the appendix. Fig. 4. Grain size distribution for Chiba & Cornell sands [32].
 D.J. Robert / Computers and Geotechnics 91 (2017) 146160 151

1000
Matric Suction (kPa)

100

10

0.1 d_17.0kN/m3
d_18.0kN/m3
0.01
0 0.2 0.4 0.6 0.8 1
Water Saturation
(a) Chiba sand

1000
Fig. 7. Variation of peak dilation wmax with water saturation for Chiba sand
Matric Suction (kPa)

100 (cd = 1.58 g/cm3) and Cornell sands (cd = 1.64 g/cm3).

10
trend, but with significantly low magnitudes compared to Chiba
1 sand properties. In effect, the strength increase in Chiba sand can
d_13.5kN/m3 be due to both microscopic effective apparent cohesion c0u as well
0.1 d_14.7kN/m3 as due to macroscopic cohesion c00 induced by significant dila-
d_15.6kN/m3 tancy. For Cornell sand, it was found that the magnitude of c0u
0.01 was very small, but a mild contribution to strength is provided
0 0.2 0.4 0.6 0.8 1 by macroscopic cohesion c00 at smaller saturations where there is
Water Saturation high matric suction in the sand.
(b) Cornell sand The strength enhancement characteristics can be illustrated
using Coulomb strength envelope. Fig. 8 shows a schematic dia-
Fig. 5. Soil water characteristic curves of the two sands with different dry density
gram illustrating the increase in strength caused by suction and
(re-plotted from [32]).
apparent cohesion for Chiba and Cornell sand. For Chiba sand, the
strength parameters for a given dry density are /0 , c0u and c00 . On
I II III the other hand, the strength parameters for Cornell sand, for a given
dry density are /0 and c00 . As schematically showed in the figure, the
7 increase in strength for Cornell wet sand is attributed mainly by the
suction induced macroscopic capillary cohesion c00 . The capillary
6
cohesion c0u , which is considered to be due to the suction induced
5 hardening occurring at the microscopic particle contact level, has
an influential effect on the strength characteristics of Chiba sand
c'u (kPa)

4
in addition to c00 . If one could plot the strength envelope for Cornell
3 sand considering the generalised Bishops effective stress frame-
work (dashed line in Fig. 8), this would lie more or less to an iden-
2
tical level with that of the dry sand at the same dry density (dotted
1 line in Fig. 8). Even though the wet sand envelope for Chiba sand lies
at a similar gradient with that of the dry sand envelope at the same
0
0 0.2 0.4 0.6 0.8 1 dry density, the additional strength contribution from c0u shifts the
Saturation wet sand envelope to a higher level (see Fig. 8a).

Chiba sand, rd=1.58g/cm3 Cornell sand, rd=1.64g/cm3

Fig. 6. Variation of effective cohesion c0u with water saturation for Chiba sand and
Cornell sands.
no clear increase of c0u for Cornell sand, however, c0u for Chiba sand
showed a bell-shaped trend with the changes of saturations. Ini-
tially, the value of c0u increases with decreasing water saturation
from full saturation conditions (Sw decreases from zone III to II).
After reaching a peak, the value of c0u then decreases with further
decrease in water saturation. At residual water conditions (i.e. in
Zone I), the contribution of microscopic capillary cohesion is
almost negligible. In the capillary regime (Zone III), the real beha-
viour would be to diminish c0u towards zero at the entry to zone III.
The dilatancy behaviour of Chiba sand has also depicted similar
bell-shaped trend, but peaked around residual saturation levels
(Fig. 7). The dilatancy behaviour of Cornell sand holds similar
4.3. Model calibration
Single element finite element (FE) analyses were performed to
simulate the triaxial tests for the purpose of calibrating the pro-
posed modified Mohr-Coulomb model. The calibration model was
developed using three dimensional 8-noded linear brick reduced
integration element (C3D8R, [2]) with various applied net
confining stresses (10 kPa, 25 kPa & 55 kPa), initial densities
(1.451.58 g/cm3 & 1.641.68 g/cm3 for Chiba and Cornell sand
respectively) and moisture conditions (5.918% and 4% for Chiba
and Cornell sand respectively) in response to the triaxial tests per-
formed. Triaxial compression loading was simulated by applying
vertical displacement on the top surface of the element while
restraining the base surface and one side of the two transverse sur-
faces. The remaining two transverse surfaces were allowed to
deform freely with the applied vertical displacement.
152 D.J. Robert / Computers and Geotechnics 91 (2017) 146160

Wet sand Envelope-Total Stress

Wet sand Envelope-Bishop Stress
Wet sand Envelope-Total Stress
Deviator stress '
Strength increase Wet sand Envelope-Bishop Stress
due to suction, c ' ' '
Strength increase due to
apparent cohesion, cu' Strength increase
due to suction, c ' '
Dry sand Envelope
Dry sand Envelope

Normal Stress Normal Stress

(Total, Bishops effective stress) (Total, Bishops effective stress)

(a) Chiba Sand (b) Cor nell S and

Fig. 8. Material input for /0 ; c00 ; and c0u for the Chiba sand and Cornell sands.

Fig. 9. Behavioural prediction (Strength & Dilatancy) for Cornell sand at 4% moisture content (w) and cd = 1.68 g/cm3.

The results of numerical simulations of the behaviour of Figs. 911 and compared with actual triaxial (TX) test data [see
unsaturated Chiba sand and Cornell sand at initial dry densities = Robert [29] for other predictions of the numerical model in
1.58 g/cm3 and 1.68 g/cm3 respectively are presented in comparison with triaxial test data for Cornell and Chiba sand].
 D.J. Robert / Computers and Geotechnics 91 (2017) 146160
Fig. 10. Behavioural (i.e. strength) prediction for Chiba sand cd = 1.58 g/cm3 at different moisture contents (w).
The results show the deviator stresses and volumetric strains
against axial strain under three net-confining stress levels of
10 kPa, 25 kPa and 55 kPa. The constitutive model predicts realistic
stiffness, peak strength, residual strength and dilation for unsatu-
rated tests at all the confining stress levels to an axial strain of 0.04.
Even though the partial saturation increases the dilatancy at
low moisture contents and low stress levels (which is predicted
well by the model), such behaviour is diminished by larger confin-
ing stresses due to excess densification (confining) effect over
unsaturation. The proposed model is unable to predict such over-
shadowed dilatancy at low moisture and high confining stress
levels (see Fig. 11b). This is because the incapability of the model
to predict the behaviour of unsaturated sands close to the region
of residual saturation at larger stress levels. In other words, it
shows the limitation of the generalised effective stress formulation
applied to the pendular regime of water saturation. However, the
model is capable of predicting the strength and dilatancy of tested
sands at other initial conditions. The calibrated model parameters
for modified Mohr-Coulomb model are showed in Table 1 for Chiba
and Cornell sand under unsaturated conditions.
5. Model application to simulate buried pipe response in
unsaturated soils
5.1. Large-scale physical model experiments
The developed and calibrated modified Mohr-Coulomb model
has been applied to simulate the large scale physical model exper-
153
iments which were conducted to investigate the behaviour of
pipelines subjected to lateral ground movements. Experiments
were conducted at different initial states of sands (dry density
and moisture content) as well as different depths on the basis of
pipelines buried in Chiba sand and Cornell sand. Some results of
the physical model tests are presented here, and further details
(including results) of the large scale tests are reported by Robert
et al. [35], Jung et al. [15], ORourke [25] and Robert [29].
5.2. Numerical simulations of the full-scale soil-pipe tests
FE simulations of the laterally-loaded pipe experiments were
carried out using ABAQUS [2] under plane strain conditions. The
modified Mohr-Coulomb model was implemented as VUMAT in
ABAQUS. To simulate the pipe behaviour under box movement,
the pipe was pulled laterally by imposing equal lateral displace-
ment on all pipe nodes while setting it to move freely in the verti-
cal direction. Considering the fast pipeline loading in low
permeable unsaturated soils, the analyses were conducted under
undrained condition (i.e. volume of water remains constant). The
wall boundaries of the FE model (length = 10 m and height = 2.7 m)
were assumed to be smooth and supported in the normal direction.
Adaptive meshing has been incorporated in the analyses to control
the mesh distortions that result from large deformations of the soil
caused by lateral pipe displacements. The soil and pipe elements
were 4-node bilinear, plane strain, reduced integration with hour-
glass control (CPE4R) elements. The pipe was assumed to be a lin-
ear elastic material (ASTM Grade A-36 steel). Fig. 12 shows the
154 D.J. Robert / Computers and Geotechnics 91 (2017) 146160

Fig. 11. Behavioural (i.e. dilatancy) prediction for Chiba sand cd = 1.58 g/cm3 at different moisture contents (w).

geometry and the mesh discretisation of the FE model used to sim-
ulate the laterally loaded pipeline experiments.
The interaction between the soil and pipe surface was modelled
by assuming that the allowable frictional stress, scrit , is given by
lr0n where l tan /l and /l is the interface friction coefficient.
Slip along the soil-pipe interface occurs when s scrit and the sep-
aration between the soil and pipe occurs when r0n < 0. This beha-
viour was modelled using the finite movement solution available
in ABAQUS [2]. In the current study, /l was set to equal to half
of the saturated peak frictional angle of soil [8,46].
5.3. Model prediction of pipe response
The response of the pipe is discussed on the basis of soil defor-
mation mechanisms and pipe loading. The results of large scale
experiment are compared with the model predictions obtained
from the proposed modified Mohr-Coulomb model (MMC) for the
lateral pipe test based on Chiba sand which was having initial soil
moisture and dry density of 15.2% and 1.61 g/cm3 respectively. The
model responses using linear elastic-perfectly plastic Mohr-
Coulomb model (MC) are also showed in comparison.
The deformation mechanism of the soil in the FE models was
derived from the deviator strain contours at a dimensionless pipe
displacement, d/D, of 0.7, where d is displacement and D is the pipe
external diameter of 114.6 mm. Fig. 13 shows the observed soil
deformation pattern in large scale experiment, and Fig. 14a and b
shows the deviator strain contours from the FE simulations using
MC and MMC models, respectively. Similar to what was observed
in large scale tests, the lateral pipe movement results forming
shear bands 1 and 2 in the FE models. Even though the inclinations
of shear bands formed between MC and MMC models are the same,
 D.J. Robert / Computers and Geotechnics 91 (2017) 146160 155

Table 1
Modified Mohr-Coulomb model input parameters for unsaturated sand tests.

Sand category cd (g/cm3) p (kPa) w E (kPa) m /0 max () /crit ()

Chiba 1.58 10 5.9 15,000 0.35 45.0 33.0
25 5.9 18,500 0.35 45.0 33.0
55 5.9 23,000 0.35 45.0 33.0
1.59 10 8.5 10,000 0.35 45.0 33.0
25 8.5 14,500 0.35 45.0 33.0
55 8.5 20,000 0.35 45.0 33.0
1.60 10 12.4 7000 0.35 45.0 33.0
25 12.4 12,000 0.35 45.0 33.0
55 12.4 17,500 0.35 45.0 33.0
1.60 10 18.0 6000 0.35 45.0 33.0
25 18.0 9600 0.35 45.0 33.0
55 18.0 14,000 0.35 45.0 33.0
1.51 10 12.0 5000 0.35 40.0 33.0
25 12.0 9000 0.35 40.0 33.0
55 12.0 12,000 0.35 40.0 33.0
1.45 10 5.8 8000 0.35 38.0 33.0
25 5.8 9000 0.35 38.0 33.0
55 5.8 10,000 0.35 38.0 33.0

Cornell 1.64 10 3.7 4000 0.35 38.0 34.0

25 3.7 8000 0.35 38.0 34.0
55 3.7 12,000 0.35 38.0 34.0
1.68 10 3.7 6500 0.35 39.0 34.0
25 3.7 10,000 0.35 39.0 34.0
55 3.7 14,000 0.35 39.0 34.0

Depends on H/D

5.0m

2.0m

10.0m

Fig. 12. Geometry and mesh discretisation of the model used for the FE analyses.

Continuous soil heaving

Gap between pipe and Decreasing Depression

Region 1

Region 2

(a) Gap formation (b) Soil depression (c) Soil heave

Fig. 13. Soil deformation mechanisms in the large scale test (based on Chiba sand) (a) Gap formation at smaller pipe displacement, (b) soil depression on the active side and
(c) Group surface heave on the passive side [35].

MC model results thicker shear bands with larger amount of dis-
placed soil than MMC model. This is shown by the up heave ground
displacement plotted in Fig. 15 which shows more ground surface
heave in MC model than in MMC model. Further, model predictions
show no active soil depression at the ground response as was seen
in large scale experiments at small pipe displacements. However,
at large pipe displacements (i.e. 400 mm), the soil behind the pipe
suddenly collapsed (see Fig. 13c for the Chiba sand). This behaviour
could not be observed in the FE model as the pipe displacements in
FE simulations were restrained upto 150 mm as of numerical dif-
ficulties encountered by high mesh distortion resulted at larger
ground displacements.
The proposed model was applied to predict load-displacement
relationships for pipes buried in Chiba sand and Cornell sand.
156 D.J. Robert / Computers and Geotechnics 91 (2017) 146160

SB-2

SB-1 (a) MC model

SB-2

SB-1 (b) MMC model

Fig. 14. Mechanism of shear strain distribution contours of FE analyses using (a) MC model & (b) MMC model at dimensionless displacement (d/D) of 0.7.

0.07 Load-displacement responses of the pipe have also been predicted

Verticle soil displacement (m)

MC model prediction by unsaturated Nor-Sand model (UNS) which was previously

0.06 MMC model prediction developed by the author [35] incorporating the effect of increasing
0.05 apparent tensile strength due to the water meniscus at the particle
contacts as well as enhanced dilation. The model predictions are
0.04
compared with the test data for 4 different Chiba sand experiment
0.03 cases which have different dry density and moisture content of soil
[29]. The predictions from the proposed model are compared to
0.02
Cornell tests (reported in [15]) for different H/D cases, but having
0.01 similar dry densities and moisture contents of sands as most of
the large scale tests conducted at Cornell were largely targeted
0
similar sand initial conditions. A detailed description of the test
0.2 0.3 0.4 0.5 0.6 0.7
-0.01 cases analysed for load-displacement relationships is given in
Normalised distance along the ground surface Table 2.The FE load-displacement relationships are presented in
Figs. 16 and 17 for the Chiba and Cornell sand cases, respectively.
Fig. 15. Vertical soil displacement at the ground surface (top surface) of the FE
It can be seen that the loading responses from the proposed model
models at dimensionless displacement (d/D) of 0.7.
match in terms of stiffness, peak and softening and compare well

Table 2
Large scale test cases (extracted from [29]).

Test case Sand type Pipe diameter (mm) H/D cd (g/cm3) w (Sw) (%)

1-CB Chiba 114.6 5.74 1.52 17.3 (61.8)

2-CB Chiba 114.6 5.74 1.57 15.4 (59.4)
3-CB Chiba 114.6 5.74 1.61 15.2 (62.5)
4-CB Chiba 114.6 5.74 1.32 14.5 (38.2)
1-CN Cornell 124.0 5.28 1.63 4.15 (17.1)
2-CN Cornell 120.0 5.5 1.65 4.91 (20.9)
3-CN Cornell 174.0 6.5 1.58 4.27 (16.4)
4-CN Cornell 174.0 3.5 1.63 4.61 (19.3)
 D.J. Robert / Computers and Geotechnics 91 (2017) 146160 157

Pipe Load (kN/m length)

50 50
MC prediction Experiment MC prediction
40 40

30 30 MMC prediction
MMC prediction UNS prediction
20 20 Experiment
UNS prediction
10 10

0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
Displacement (m) Displacement (m)
(a) F-d plot for case 1-CB (b) f-d plot for test case 2-CB
Pipe Load (kN/m length)

50 50
MC prediction
40 40

30 30
MMC prediction UNS prediction
Experiment MC prediction
20 UNS prediction 20
MMC prediction
10 10
Experiment
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
Displacement (m) Displacement (m)
(c) F-d plot for case 3-CB (d) f-d plot for test case 4-CB
Fig. 16. Load-displacement (f-d) plots for lateral pipeline tests in Chiba sand.
Pipe Load (kN/m length)

30 30
MC prediction
25 MC prediction 25
20 20
15 15 MMC prediction Experiment
10
MMC prediction
10 UNS prediction
UNS prediction Experiment
5 5
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
Displacement (m) Displacement (m)
(a) F-d plot for case 1-CN (b) f-d plot for test case 2-CN
Pipe Load (kN/m length)

70 30
60
25 MC prediction
50 MC prediction
20
40
15
30
Experiment MMC prediction Experiment
20 10
MMC prediction
UNS prediction
10 5
UNS prediction
0 0
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08

Displacement (m) Displacement (m)

(c) F-d plot for case 3-CN (d) f-d plot for test case 4-CN
Fig. 17. Load-displacement (f-d) plots for lateral pipeline tests in Cornell sand.

with the experimental results. Further, the predictions from the
proposed model have been capable of predicting similar responses
when compared with advanced UNS model predictions. Results
revealed that the accuracy of peak load prediction from MMC is
within 5% of the measure peak load based on physical model tests
(except 2 narrow outliers of 6% & 7.3%) as highlighted in Table 3.
However, the model predictions from MC model over-estimates
the loading (both peak and residual), especially for pipes buried
in Chiba sand, due to lack of depreciation in frictional and dilatancy
properties with changes in strains and saturations. For instance,
average accuracy of MC predictions for pipe loading in Chiba sand
is over-estimation of 12% and 28% for peak and residual pipe
loading respectively. On the other hand, the difference in pipe peak
loads between the models is less pronounced for pipes buried in
158 D.J. Robert / Computers and Geotechnics 91 (2017) 146160

Table 3
Peak and residual pipe loading extracted from the analysis.

Test case Pipe load peak (kN/m) Pipe load residual (kN/m)
a b c d e
MC MMC Exp MC-Accuracy MMC-Accuracy MCa MMCb Expc MC-Accuracyd MMC-Accuracye
1CB 39.0 34.2 36.4 7.1 6.0 36.0 30.5 30.8 16.9 1.0
2CB 44.9 42.4 41.2 9.0 2.9 43.6 38.4 29.9 45.8 28.4
3CB 49.3 45.5 45.2 9.1 0.7 49.2 38.2 38.4 28.1 0.5
4CB 14.9 12.7 12.2 22.1 4.1 14.9 12.7 12.2 22.1 4.1
1CN 22.5 19.9 19.6 14.8 1.5 22.5 18.2 16.4 37.2 11.0
2CN 25.4 23.0 24.8 2.4 7.3 24.3 20.0 20.9 16.3 4.3
3CN 43.5 41.4 39.9 9.0 3.8 43.5 40.0 39.9 9.0 0.3
4CN 21.0 20.2 20.0 5.0 1.0 21.0 18.4 17.2 22.1 7.0
a
Mohr-Coulomb model predictions.
b
Modified Mohr-Coulomb model predictions.
c
Pipe load obtained from large scale tests.
d
Over (+)/under () prediction (%) of MC model compared to large scale test data (>5% difference is highlighted).
e
Over (+)/under () prediction (%) of MMC model compared to large scale test data (>5% difference is highlighted).

Cornell sand due to lack of unsaturated cohesion (c0u ) for this sand. Appendix A. Functions and derivatives used in modified Mohr-
For example from Table 3, the average accuracy of peak load pre- Coulomb model
diction for Chiba sand cases from MC is an over-estimation of
12% in contrast to 8% over-estimation for the same in Cornell The main framework for the development of modified Mohr-
sand (Table 3). However, a substantial difference in the MC predic- Coulomb model has been demonstrated in the manuscript. The
tion of pipe residual loading has been encountered in Cornell sand appendix shows the functional forms and partial derivatives of
as highlighted in Table 3. These observations reveal that the linear- the yield and potential functions used in the formulation of the
elastic perfectly plastic Mohr-Coulomb may not be able to capture model. The standard functions [2,22] were extended to include
the complicated mechanisms of unsaturated soil behaviour inter- the saturation dependence on unsaturated apparent cohesion
acting with buried structures. The proposed modified Mohr- (c0u c00 in Eq. (10)) and unsaturated dilation (wunsat in Eq. (14)).
Coulomb model can capture the realistic loading on pipes buried
in unsaturated soils as it includes suction induced hardening in
A.1. Yield function and its derivatives
addition to classical hardening mechanisms.
Yield function for the model has been defined as;
r  
3 1  p 1  p e
6. Summary and Conclusion 0
q p 0
sin h cos h tan/ p tan/0  c0 0
2 3 cos/ 3 3 3 3
In this paper, a modified Mohr-Coulomb model is proposed to A:1
capture the mechanical behaviour of unsaturated soils. The model
has been developed within a generalized effective stress framework where
considering material softening and suction hardening effects. Strain
1
softening behaviour was captured by incorporating the reduction of e p r11 r22 r33 A:2
mobilized friction, dilation and apparent cohesion with an increase 3
in plastic deviatoric shear strain. Suction was incorporated into  1=2
yield and potential functions through the dependence of saturation 1 2 2 2
q r11  r22 r22  r33 r33  r11  2r212 2r223 2r231
on strength and dilatancy properties of soils. The model developed 3
in this study is capable of simulating the development of shear band A:3
with the use of elementary normalised FE computed shear strains 2 p 3
derived using actual shear strains obtained in soil element testing 1 1 43 3 J 3 5
of unsaturated soils. The model was implemented into commercial h cos where; A:4
3 2 J 32
FE program ABAQUS associated with user subroutine written in 2

FORTAN and was validated through a series of FE analyses in com-

parison with triaxial test data. The new model, which is called as J 3 r11  pr22  pr33  p  r11  pr223  r22  pr231
unsaturated modified Mohr-Coulomb model, only requires 3 addi-  r33  pr212 2r12 r23 r31 A:5
tional parameters compared to traditional linear-elastic perfectly
plastic Mohr-Coulomb model; 1.0 plastic deviatoric strain at soften- 1
ing completion (to characterize material softening), 2.0 relationship J2 r11  r22 2 r22  r33 2 r33  r11 2 
6
of unsaturated cohesion with saturation (cohesion enhancement), r212 r223 r231 A:6
and 3.0 relationship of unsaturated dilation with saturation (dila-
tancy enhancement). The study revealed that the proposed model Partial derivative of the yield function based on stress;
with appropriate model parameters was able to produce good out-
@F @F @ e @F @ q @F @h
comes when compared to large scale physical model experimental A:7
results. Hence, the developed model can be used to predict the @ r @ e @ r @ q @ r @h @ r
increased strength and stiffness associated with soil suction that where
increases lateral loads on pipelines, and thus has widespread rele- @F tan /0
vance for simulating the pipeline response under externally p A:8
@e 3
imposed ground movements.
 D.J. Robert / Computers and Geotechnics 91 (2017) 146160 159

@e 1 where
p dij ; dij Kroneckers delta dii 1; dij 0 A:9
@ rij 3 Rmw h; e
r  41  e2 cos2 h 2e  12 p 
@F 3 1  p 1  p p Rmc ; /0
0
p sin h cos h tan / A:10 21  e2 cos h 2e  1 41  e2 cos2 h 5e2  4e 3
@q 2 3 cos /0 3 3 3
A:25
@q 1 p  3  sin /0
sij A:11
@ rij q Rmc ; /0 A:26
3 6 cos /0
r  
@F 3 1  p 1  p
q p cos h  sin h tan /0 A:12 3  sin /
@h 2 3 cos /0 3 3 3 e A:27
3 sin /
 
@h @h @r @J 2 @r @J 3 wunsat is obtained on the basis of Eq. (14).
A:13
@ rij @r @J 2 @ rij @J 3 @ rij Partial derivative for the potential function;
@P @P @ e @P @ q @P @h
where A:28
@ r @ e @ r @ q @ r @h @ r
p
3 3 J3 where
r A:14
2 J 32 @P tan wunsat
2 p A:29
  @e 3
@h 1 1
 p with r 0 A:15
@r 3 1  r2 @P 3 qR2mw
r
q 2
A:30
0 1 @q 2
@r 9 p@J 3 A nc0 tan wunsat 2 3
2
R mw q
 3 5 with J 2 0 A:16
@J 2 4 J2 2
@P 3 q2 Rmw p  @R h; e
Rmc ; /0
r
q
mw

@J 2 @h 2   2 3 @h
2
sij A:17 nc0 tan wunsat 3
R q
2 mw
@ rij
A:31
0 1
@r p 1 where
1:5 3@ 3 A with J 2 0 A:18
@J 3 J2  
2 @Rmw h; e @ L M @L  L @M
@h 2 @h A:32
@h @h M M
@J 3 2
sik skj  J2 dij A:19
@ rij 3
L 41  e2 cos2 h 2e  12 A:33
Partial derivative of the yield function based on plastic deviatoric q
strain; M 21  e2 cos h 2e  1 41  e2 cos2 h 5e2  4e
A:34
@F @F @/0 @F @w @F @c0
p A:20
@ es @/0 @ eps @w0 @ eps @c0 @ eps
@L
81  e2 cos h sin h A:35
@F p  sin /0 1

@h
3=2 q sinh p=3 1=3 cosh p =3
@/0 cos2 /0 cos2 /0
e @M 21  e2 sin h  42e  11  e2 cos h sin h
p A:36
p @h 41  e2 cos2 h 5e2  4e
3 cos2 /0
A:21
References
@F
1:0 A:22
@c [1] Abolmaali A, Kararam A. Nonlinear finite-element modeling analysis of soil-
pipe interaction. Int J Geomech 2013:197204. http://dx.doi.org/10.1061/
Partial derivative of the yield function based on saturation (ASCE)GM.1943-5622.0000196.
[2] ABAQUS, Inc. ABAQUS V. 6.7 users manual, Providence, Rhode Island; 2011.
@F @co [3] Anastasopoulos M, Gazetas G, Bransby MF, Davies MCR, Nahas AE. Fault
1:0:l: A:23
@Sw @Sw rupture propagation through sand: finite-element analysis and validation
through centrifuge experiments. J Geotech Geoenviron Eng 2007;133
@co (8):94358.
where @Sw
is based on shape of the unsaturated cohesion depen-
[4] Attewell PB, Woodman JP. Predicting the dynamic of ground settlement and its
dence on saturation, and l as defined in Eq. (15). derivatives caused by tunneling in soil. Ground Eng 1982;15(8):1336.
[5] Bolzon G, Schrefler BA, Zienkiewicz OC. Elastoplastic soil constitutive laws
A.2. Potential function and its derivatives generalised to partially saturated states. Geotechnique 1996;46(2):27089.
[6] Bryden P, El Naggar H, Valsangkar A. Soil-structure interaction of very flexible
pipes: centrifuge and numerical investigations. Int J Geomech 2014;04014091.
Potential function for the model has been defined as; http://dx.doi.org/10.1061/(ASCE)GM.1943-5622.0000442.
[7] Chatterjee S. Numerical modelling of pipe-soil interactions [Ph.D.
q thesis]. University of Western Australia; 2012.
P ec0 tan wunsat 2 Rmw q2  p tan wunsat A:24 [8] Cheong TP. Numerical modelling of soil-pipeline interaction [Ph.D.
thesis]. University of Cambridge; 2006.
160 D.J. Robert / Computers and Geotechnics 91 (2017) 146160
[9] Cheong T, Soga K, Robert D. 3D FE analyses of buried pipeline with elbows
subjected to lateral loading. J Geotech Geoenviron Eng, 1090-0241 2011;137
(10):93948 [American Society of Civil Engineers, United States].
[10] Crofts JE, Menzies BK, Tarzi AI. Lateral displacement of shallow buried
pipelines due to adjacent deep trench excavations. Geotechnique 1977;27
(2):16179.
[11] EERI. The Izmit. EERI Special Earthquake Report. Kocaeli. Turkey Earthquake of
August 17, 1999; 1999.
[12] Jefferies MG. Nor-Sand: a simple critical state model for sand. Geotechnique
1993;43(1):91103.
[13] Jennings JEB, Burland JB. Limitations to the use of effective stresses in
unsaturated soils. Geotechnique 1962;12:12544.
[14] Jommi C. Remarks on the constitutive modelling of unsaturated soils. In:
Experimental evidence and theoretical approaches in unsaturated soils; proc
of an international workshop, Trento; 2000. p. 13953.
[15] Jung J, ORourke TD, Olson NA. Lateral soil-pipe interaction in dry and partially
saturated sand. J Geotech GeoEnviron Eng, ASCE 2013;139(12):202836.
[16] Karamitros DK, Bouckovalas GD, Kouretzis GP. Stress analysis of buried steel
pipelines at strike-slip fault crossings. Soil Dyn Earthq Eng 2007;27(3):20011.
[17] Koseki J, Matsuo O, Tanaka S. Uplift of sewer pipes caused by earthquake-
induced liquefaction of surrounding soil. Soil Found 1998;38(3):7587.
[18] Laloui L, Nuth M. On the use of the generalised effective stress in the
constitutive modelling of unsaturated soils. Comput Geotech 36:203.
[19] Leonards GA, Roy MB. Predicting performance of pipe culverts buried in soil.
Joint highway research project, report no. JHRP-76-15; 1976.
[20] Marshall TJ, Holmes JW. Soil physics. 2nd ed. Cambridge: Cambridge
University Press; 1988.
[21] Marston A, Anderson AO. The theory of loads on pipes in ditches and tests of
cement and clay drain tile and sewer pipe. Ames, Iowa: Bulletin 31, Iowa
Engineering Experiment Station; 1913.
[22] Menetrey P. Numerical analysis of punching failure in reinforced concrete
structures [Ph.D. thesis]. Switzerland: Ecole Polytechnique Federale de
Lausann; 1994.
[23] Menetrey P, William KJ. Triaxial failure criterion for concrete and its
generalization. ACI Struct J 1995;92:3118.
[24] Olson N. Soil performance for large-scale soil-pipeline tests [Ph.D.
thesis]. Cornell University; 2009.
[25] ORourke TD. Geohazards and large geographically distributed systems.
Rankine Lect, Geotech 2010;60(7):50343.
[26] ORourke TD, Jezerski NA, Olson T, Bonneau AL, Palmer MC, Stewart HE, et al.
Geotechnics of pipeline system response to earthquakes. In: Keynote paper,
proceedings, geotechnical earthquake engineering and soil dynamics IV
(GEESD) Sacramento, CA; May 2008.
[27] ORourke TD, Jung J, Argyrou C. Underground infrastructure response to
earthquake-induced ground deformation. In: Proceedings, 6 intl conf on
earthquake geotechnical engineering. Christchurch, NZ: Intl. Society for Soil
Mech. and Found. Engr.; 2015.
[28] Potts D. Numerical analysis: a virtual dream or practical reality? Gotechnique
2003;53(6):53573.
[29] Robert DJ. Soil-pipeline interaction in unsaturated soils [Ph.D.
thesis]. University of Cambridge; 2010.
[30] Robert D, Rajeev P. A modified Mohr-Coulomb model to simulate the response
of buried pipes subjected to large ground displacement. In: Geo-Chicago 2016:
sustainability and resiliency in geotechnical engineering, Chicago, United
States, 1418 August 2016; 2016.
[31] Robert DJ, Soga K. Simulations of soil-pipeline interactions in unsaturated
soils. Report to Tokyo Gas Co. Ltd., Cambridge University; 2009.
[32] Robert DJ, Soga K. Soil-pipeline interaction in unsaturated soils. In: Laloui L,
editor. Mechanics of unsaturated geomaterials. John Wiley and Sons; 2013. p.
30325 [chapter 13].
[33] Robert DJ, Soga K, ORourke TD. Pipelines subjected to fault movement in dry
and unsaturated soil. Int J Geomech 2016. http://dx.doi.org/10.1061/(ASCE)
GM.1943-5622.0000548.
[34] Robert D, Soga K, Britto A. Analysis of soil-pipeline interaction using ABAQUS/
Explicit. In: Raj Das, Sabu John, editor. Proceedings of the 8th Australasian
congress on applied mechanics 2014 (ACAM 8), Barton, Australia, 2428
November 2014; 2014. p. 42939.
[35] Robert D, Soga K, ORourke T, Sakanoue T. Lateral load-displacement behavior
of pipelines in unsaturated sands. J Geotech Geoenviron Eng 2016;04016060.
http://dx.doi.org/10.1061/(ASCE)GT.1943-5606.0001504.
[36] Roscoe KH, Burland JB. On the generalised stress-strain behaviour of wet
clay. Engineering Plasticity: Cambridge University Press; 1968. p. 535609.
[37] Roy K, Hawlader K, Kenny K, Moore I. Finite element modeling of lateral
pipelinesoil interactions in dense sand. Can Geotech J, 2016 2015;53
(3):490504.
[38] Saadeldin R, Hu Y, Henni A. Numerical analysis of buried pipes under field
geoenvironmental conditions. Int J Geo-Eng 2015;6(1). http://dx.doi.org/
10.1186/s40703-015-0005-4.
[39] Sheng D, Sloan SW, Gens A. A constitutive model for unsaturated soils:
thermomechanical and computational aspects. Comput Mech 2004;6
(33):45365.
[40] Spangler MG. The structural design of flexible pipe culverts. Ames,
Iowa: Bulletin 153, Iowa Engineering Experiment Station; 1941.
[41] Suleiman MT. The structural performance of flexible pipes [Ph.D.
thesis]. US: Iowa State University; 2002.
[42] Tamagnini R. An extended Cam-clay model for unsaturated soils with
hydraulic hysteresis. Geotechnique 2004;54(3):2238.
[43] Van Genuchten MT. A closed form equation for predicting the hydraulic
conductivity of unsaturated soils. Soil Sci Soc Am J 1980;44:8928.
[44] Vazouras P, Karamanos SA, Dakoulas P. Finite element analysis of buried steel
pipelines under strike-slip fault displacements. Soil Dyn Earthq Eng 2010;30
(11):136176.
[45] Wheeler SJ, Sharma RS, Buisson MSR. Coupling of hydraulic hysteresis and
stress-strain behaviour in unsaturated soils. Geotechnique 2003;53(1):4154.
[46] Yimsiri S, Soga K, Yoshizaki K, Dasari GR, ORourke TD. Lateral and upward soil-
pipeline interactions in sand for deep embedment conditions. J Geotech
Geoenviron Eng, ASCE 2004;130(8):83042.