Академический Документы
Профессиональный Документы
Культура Документы
Research Paper
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.
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 147
provided geotechnical engineers with an extremely powerful anal- interaction problems [8,44,1,37,15]. These models are often chosen
ysis tool [28]. They facilitate to assess the response of pipes consid- considering its simplicity, ease of use, reasonable computational
ering wide range of variables in timely and efficient manner. The time and the high level of understanding among the engineers.
most recent research work conducted in the area of numerical Such models are readily available in commercial finite element
modelling of soil-pipeline interaction problems has been able to programs so that researchers/practitioners can use them at less
highlight the development of proper numerical tools to capture cost. However, they have their own limitations which can signifi-
the realistic behaviour of pipelines when subjected to ground cantly influence the prediction of buried pipeline response unless
movements. For example, Yimsiri et al. [46] have adopted state appropriate model changes are incorporated.
dependent hardening/softening model (Nor-Sand [12]) to study The linear elastic-perfectly plastic Mohr-Coulomb model is
the behaviour of pipelines subjected to lateral ground deformation. unable to predict the realistic loading that can apply on buried
Cheong et al. [9] also used the Nor-Sand model to investigate the structures during large ground deformation. Especially, the original
response of elbow pipelines subjected to lateral ground deforma- Mohr-Coulomb model is not applicable to simulate the unsatu-
tion. Chatterjee [7] has adopted modified Cam-Clay model [36] to rated soil loading which can result larger than anticipated loading
quantify the lateral resistance of pipes in soft soil. Bryden et al. due to suction induced additional normal force between soil parti-
[6] have used hyperbolic hardening soil model, which can capture cles. Recent studies showed that soil suction under unsaturated
the stress-dependent variation in stiffness to study the effect of condition can change the stiffness and strength, affecting the beha-
surface loading on buried pipelines. The uses of such advanced soil viour of the pipeline systems subjected to ground deformation
constitutive models were effective for predicting the behaviour of [25,15,38,33,35,31]. Therefore, utilization of traditional Mohr-
pipelines due to their superior capability of capturing the load Coulomb failure criterion could lead to under-estimation of soil
dependent material state changes. However, in most occasions, loads on pipes, thus could result in unexpected pipeline failures
they demand comprehensive calibration of soil parameters using in the events of construction or earthquake induced ground defor-
detailed and advanced testing programs prior to use in the desired mations. This study showed that Mohr-Coulomb model could well
application. On the other hand, it is very common in practice to uti- be used to predict realistic pipeline loading if it can be modified to
lize standard material models such as Mohr-Coulomb model to include the suction into the constitutive modelling framework. A
simulate the soil behaviour in the application of soil-structure user defined unsaturated modified Mohr-Coulomb model is
148 D.J. Robert / Computers and Geotechnics 91 (2017) 146160
q
developed within a generalized effective stress framework includ-
P ec0 tan w2 Rmw q2 p tan w 3
ing suction hardening effect. Firstly, the model has been developed
considering microscopic and macroscopic suction hardening where
mechanisms, and was implemented into a commercial finite ele-
ment program associated with user subroutine written in FORTAN. Rmw H; e
Then the model was validated through a series of unsaturated tri-
41 e2 cos2 H 2e 12
axial compression tests conducted on the basis of different sand p Rmc p=3; /0
types having various initial conditions. Finally, the model has been 21 e2 cos H 2e 1 41 e2 cos H 5e2 4e
applied to simulate the behaviour of pipelines subjected lateral soil
loading in unsaturated soils. The results revealed that the modified 3 sin /0
Rmc p=3; /0
Mohr-Coulomb model has reasonable predictions when compared 6 cos /0
to the load-displacement response of pipes obtained from two
large scale testing programs. The developed model can be used where /0 and w are the friction angle and the dilation angle mea-
to predict the increased strength and stiffness associated with soil sured in the p Rmw q plane, respectively. c0 is the initial cohesion
suction that increases lateral loads on pipelines, and thus has wide- yield stress; H remains as defined in yield criterion; e is a parameter
spread relevance for simulating the pipeline response under exter- which is referred to as meridional eccentricity which defines the
nally imposed ground movement. 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
2. Mohr-Coulomb model
parameter referred to as the deviatoric eccentricity [2] that
describes the out-of-roundedness of the deviatoric section in terms
Mohr-Coulomb model is a simple linear elastic-perfectly plastic
of the ratio between shear stress along the extension meridian
model which is widely used for the design applications in geotech-
H 0 and shear stress along the compression meridian
nical engineering to simulate material response under monotonic
H p=3.
loading. The model is widely popular in the community for mod-
The original Mohr-Coulomb model can be modified to capture
elling the behaviour of soils due to its relative simplicity and the
the strain softening behaviour of the material at large deforma-
requirement of the basic soil properties (such as friction and dila-
tions [32]. The softening behaviour is captured by reducing the
tion of soils).
mobilized friction and dilation angle with an increase in plastic
Within the Mohr-Coulomb (MC) framework, the soil is mod-
deviatoric shear strains [3] as shown in Fig. 2, where /0max and
elled as an isotropic dilatant elastic-perfectly plastic material. Elas-
tic behaviour is modelled assuming linear isotropic elasticity of /crit are the peak friction angle and critical state friction angle,
which the stress is defined from the elastic strain as; respectively, and wmax is the ultimate dilation angle, wres residual
dilation angle, cpF is the plastic deviatoric shear strain when the
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 required on the basis of direct shear tests in order to obtain a
is the elastic strain. mesh-independent solution. The procedure introduced by Anasta-
The model behaves elastically and obeys Hookes law until the sopoulos et al. [3] can be adopted when deriving plastic shear
onset of yielding which is determined by the Mohr-Coulomb yield strain at the completion of softening cpF based on direct shear test
criterion. For general states of stress, the yield criterion of the data. This approach can be effective for obtaining mesh size inde-
model in terms of three stress invariants is defined as (p: compres- pendent solution for soil-pipeline interaction analysis involving
sion negative) [2]; large ground deformations [32,30].
1 1
Rmc H; /0 p sinH p=3 cosH p=3 tan /0
3 cos /0 3
1=3 1=3
9 9
r S:S : S Sij :Sjk : Ski
2 2
fp dev
p
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
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
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).
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 153
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 iments which were conducted to investigate the behaviour of
against axial strain under three net-confining stress levels of pipelines subjected to lateral ground movements. Experiments
10 kPa, 25 kPa and 55 kPa. The constitutive model predicts realistic were conducted at different initial states of sands (dry density
stiffness, peak strength, residual strength and dilation for unsatu- and moisture content) as well as different depths on the basis of
rated tests at all the confining stress levels to an axial strain of 0.04. pipelines buried in Chiba sand and Cornell sand. Some results of
Even though the partial saturation increases the dilatancy at the physical model tests are presented here, and further details
low moisture contents and low stress levels (which is predicted (including results) of the large scale tests are reported by Robert
well by the model), such behaviour is diminished by larger confin- et al. [35], Jung et al. [15], ORourke [25] and Robert [29].
ing stresses due to excess densification (confining) effect over
unsaturation. The proposed model is unable to predict such over-
5.2. Numerical simulations of the full-scale soil-pipe tests
shadowed dilatancy at low moisture and high confining stress
levels (see Fig. 11b). This is because the incapability of the model
FE simulations of the laterally-loaded pipe experiments were
to predict the behaviour of unsaturated sands close to the region
carried out using ABAQUS [2] under plane strain conditions. The
of residual saturation at larger stress levels. In other words, it
modified Mohr-Coulomb model was implemented as VUMAT in
shows the limitation of the generalised effective stress formulation
ABAQUS. To simulate the pipe behaviour under box movement,
applied to the pendular regime of water saturation. However, the
the pipe was pulled laterally by imposing equal lateral displace-
model is capable of predicting the strength and dilatancy of tested
ment on all pipe nodes while setting it to move freely in the verti-
sands at other initial conditions. The calibrated model parameters
cal direction. Considering the fast pipeline loading in low
for modified Mohr-Coulomb model are showed in Table 1 for Chiba
permeable unsaturated soils, the analyses were conducted under
and Cornell sand under unsaturated conditions.
undrained condition (i.e. volume of water remains constant). The
wall boundaries of the FE model (length = 10 m and height = 2.7 m)
5. Model application to simulate buried pipe response in were assumed to be smooth and supported in the normal direction.
unsaturated soils Adaptive meshing has been incorporated in the analyses to control
the mesh distortions that result from large deformations of the soil
5.1. Large-scale physical model experiments caused by lateral pipe displacements. The soil and pipe elements
were 4-node bilinear, plane strain, reduced integration with hour-
The developed and calibrated modified Mohr-Coulomb model glass control (CPE4R) elements. The pipe was assumed to be a lin-
has been applied to simulate the large scale physical model exper- 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- experiment are compared with the model predictions obtained
ulate the laterally loaded pipeline experiments. from the proposed modified Mohr-Coulomb model (MMC) for the
The interaction between the soil and pipe surface was modelled lateral pipe test based on Chiba sand which was having initial soil
by assuming that the allowable frictional stress, scrit , is given by moisture and dry density of 15.2% and 1.61 g/cm3 respectively. The
lr0n where l tan /l and /l is the interface friction coefficient. model responses using linear elastic-perfectly plastic Mohr-
Slip along the soil-pipe interface occurs when s scrit and the sep- Coulomb model (MC) are also showed in comparison.
aration between the soil and pipe occurs when r0n < 0. This beha- The deformation mechanism of the soil in the FE models was
viour was modelled using the finite movement solution available derived from the deviator strain contours at a dimensionless pipe
in ABAQUS [2]. In the current study, /l was set to equal to half displacement, d/D, of 0.7, where d is displacement and D is the pipe
of the saturated peak frictional angle of soil [8,46]. 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
5.3. Model prediction of pipe response MC and MMC models, respectively. Similar to what was observed
in large scale tests, the lateral pipe movement results forming
The response of the pipe is discussed on the basis of soil defor- shear bands 1 and 2 in the FE models. Even though the inclinations
mation mechanisms and pipe loading. The results of large scale 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.
Depends on H/D
5.0m
2.0m
10.0m
Fig. 12. Geometry and mesh discretisation of the model used for the FE analyses.
Region 2
MC model results thicker shear bands with larger amount of dis- suddenly collapsed (see Fig. 13c for the Chiba sand). This behaviour
placed soil than MMC model. This is shown by the up heave ground could not be observed in the FE model as the pipe displacements in
displacement plotted in Fig. 15 which shows more ground surface FE simulations were restrained upto 150 mm as of numerical dif-
heave in MC model than in MMC model. Further, model predictions ficulties encountered by high mesh distortion resulted at larger
show no active soil depression at the ground response as was seen ground displacements.
in large scale experiments at small pipe displacements. However, The proposed model was applied to predict load-displacement
at large pipe displacements (i.e. 400 mm), the soil behind the pipe relationships for pipes buried in Chiba sand and Cornell sand.
156 D.J. Robert / Computers and Geotechnics 91 (2017) 146160
SB-2
SB-2
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.
Table 2
Large scale test cases (extracted from [29]).
Test case Sand type Pipe diameter (mm) H/D cd (g/cm3) w (Sw) (%)
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
with the experimental results. Further, the predictions from the the loading (both peak and residual), especially for pipes buried
proposed model have been capable of predicting similar responses in Chiba sand, due to lack of depreciation in frictional and dilatancy
when compared with advanced UNS model predictions. Results properties with changes in strains and saturations. For instance,
revealed that the accuracy of peak load prediction from MMC is average accuracy of MC predictions for pipe loading in Chiba sand
within 5% of the measure peak load based on physical model tests is over-estimation of 12% and 28% for peak and residual pipe
(except 2 narrow outliers of 6% & 7.3%) as highlighted in Table 3. loading respectively. On the other hand, the difference in pipe peak
However, the model predictions from MC model over-estimates 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
@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 [29] Robert DJ. Soil-pipeline interaction in unsaturated soils [Ph.D.
subjected to lateral loading. J Geotech Geoenviron Eng, 1090-0241 2011;137 thesis]. University of Cambridge; 2010.
(10):93948 [American Society of Civil Engineers, United States]. [30] Robert D, Rajeev P. A modified Mohr-Coulomb model to simulate the response
[10] Crofts JE, Menzies BK, Tarzi AI. Lateral displacement of shallow buried of buried pipes subjected to large ground displacement. In: Geo-Chicago 2016:
pipelines due to adjacent deep trench excavations. Geotechnique 1977;27 sustainability and resiliency in geotechnical engineering, Chicago, United
(2):16179. States, 1418 August 2016; 2016.
[11] EERI. The Izmit. EERI Special Earthquake Report. Kocaeli. Turkey Earthquake of [31] Robert DJ, Soga K. Simulations of soil-pipeline interactions in unsaturated
August 17, 1999; 1999. soils. Report to Tokyo Gas Co. Ltd., Cambridge University; 2009.
[12] Jefferies MG. Nor-Sand: a simple critical state model for sand. Geotechnique [32] Robert DJ, Soga K. Soil-pipeline interaction in unsaturated soils. In: Laloui L,
1993;43(1):91103. editor. Mechanics of unsaturated geomaterials. John Wiley and Sons; 2013. p.
[13] Jennings JEB, Burland JB. Limitations to the use of effective stresses in 30325 [chapter 13].
unsaturated soils. Geotechnique 1962;12:12544. [33] Robert DJ, Soga K, ORourke TD. Pipelines subjected to fault movement in dry
[14] Jommi C. Remarks on the constitutive modelling of unsaturated soils. In: and unsaturated soil. Int J Geomech 2016. http://dx.doi.org/10.1061/(ASCE)
Experimental evidence and theoretical approaches in unsaturated soils; proc GM.1943-5622.0000548.
of an international workshop, Trento; 2000. p. 13953. [34] Robert D, Soga K, Britto A. Analysis of soil-pipeline interaction using ABAQUS/
[15] Jung J, ORourke TD, Olson NA. Lateral soil-pipe interaction in dry and partially Explicit. In: Raj Das, Sabu John, editor. Proceedings of the 8th Australasian
saturated sand. J Geotech GeoEnviron Eng, ASCE 2013;139(12):202836. congress on applied mechanics 2014 (ACAM 8), Barton, Australia, 2428
[16] Karamitros DK, Bouckovalas GD, Kouretzis GP. Stress analysis of buried steel November 2014; 2014. p. 42939.
pipelines at strike-slip fault crossings. Soil Dyn Earthq Eng 2007;27(3):20011. [35] Robert D, Soga K, ORourke T, Sakanoue T. Lateral load-displacement behavior
[17] Koseki J, Matsuo O, Tanaka S. Uplift of sewer pipes caused by earthquake- of pipelines in unsaturated sands. J Geotech Geoenviron Eng 2016;04016060.
induced liquefaction of surrounding soil. Soil Found 1998;38(3):7587. http://dx.doi.org/10.1061/(ASCE)GT.1943-5606.0001504.
[18] Laloui L, Nuth M. On the use of the generalised effective stress in the [36] Roscoe KH, Burland JB. On the generalised stress-strain behaviour of wet
constitutive modelling of unsaturated soils. Comput Geotech 36:203. clay. Engineering Plasticity: Cambridge University Press; 1968. p. 535609.
[19] Leonards GA, Roy MB. Predicting performance of pipe culverts buried in soil. [37] Roy K, Hawlader K, Kenny K, Moore I. Finite element modeling of lateral
Joint highway research project, report no. JHRP-76-15; 1976. pipelinesoil interactions in dense sand. Can Geotech J, 2016 2015;53
[20] Marshall TJ, Holmes JW. Soil physics. 2nd ed. Cambridge: Cambridge (3):490504.
University Press; 1988. [38] Saadeldin R, Hu Y, Henni A. Numerical analysis of buried pipes under field
[21] Marston A, Anderson AO. The theory of loads on pipes in ditches and tests of geoenvironmental conditions. Int J Geo-Eng 2015;6(1). http://dx.doi.org/
cement and clay drain tile and sewer pipe. Ames, Iowa: Bulletin 31, Iowa 10.1186/s40703-015-0005-4.
Engineering Experiment Station; 1913. [39] Sheng D, Sloan SW, Gens A. A constitutive model for unsaturated soils:
[22] Menetrey P. Numerical analysis of punching failure in reinforced concrete thermomechanical and computational aspects. Comput Mech 2004;6
structures [Ph.D. thesis]. Switzerland: Ecole Polytechnique Federale de (33):45365.
Lausann; 1994. [40] Spangler MG. The structural design of flexible pipe culverts. Ames,
[23] Menetrey P, William KJ. Triaxial failure criterion for concrete and its Iowa: Bulletin 153, Iowa Engineering Experiment Station; 1941.
generalization. ACI Struct J 1995;92:3118. [41] Suleiman MT. The structural performance of flexible pipes [Ph.D.
[24] Olson N. Soil performance for large-scale soil-pipeline tests [Ph.D. thesis]. US: Iowa State University; 2002.
thesis]. Cornell University; 2009. [42] Tamagnini R. An extended Cam-clay model for unsaturated soils with
[25] ORourke TD. Geohazards and large geographically distributed systems. hydraulic hysteresis. Geotechnique 2004;54(3):2238.
Rankine Lect, Geotech 2010;60(7):50343. [43] Van Genuchten MT. A closed form equation for predicting the hydraulic
[26] ORourke TD, Jezerski NA, Olson T, Bonneau AL, Palmer MC, Stewart HE, et al. conductivity of unsaturated soils. Soil Sci Soc Am J 1980;44:8928.
Geotechnics of pipeline system response to earthquakes. In: Keynote paper, [44] Vazouras P, Karamanos SA, Dakoulas P. Finite element analysis of buried steel
proceedings, geotechnical earthquake engineering and soil dynamics IV pipelines under strike-slip fault displacements. Soil Dyn Earthq Eng 2010;30
(GEESD) Sacramento, CA; May 2008. (11):136176.
[27] ORourke TD, Jung J, Argyrou C. Underground infrastructure response to [45] Wheeler SJ, Sharma RS, Buisson MSR. Coupling of hydraulic hysteresis and
earthquake-induced ground deformation. In: Proceedings, 6 intl conf on stress-strain behaviour in unsaturated soils. Geotechnique 2003;53(1):4154.
earthquake geotechnical engineering. Christchurch, NZ: Intl. Society for Soil [46] Yimsiri S, Soga K, Yoshizaki K, Dasari GR, ORourke TD. Lateral and upward soil-
Mech. and Found. Engr.; 2015. pipeline interactions in sand for deep embedment conditions. J Geotech
[28] Potts D. Numerical analysis: a virtual dream or practical reality? Gotechnique Geoenviron Eng, ASCE 2004;130(8):83042.
2003;53(6):53573.