Sie sind auf Seite 1von 8

The 12

th
International Conference of
International Association for Computer Methods and Advances in Geomechanics (IACMAG)
1-6 October, 2008
Goa, India

Numerical Analysis of an Embankment founded on Structured Clay
A. Grammatikopoulou
Geotechnical Consulting Group, London, UK

L. Zdravkovic, D.M. Potts
Dept. of Civil and Environmental Engineering, Imperial College, London, UK
Keywords: constitutive relations, numerical analysis, embankment, clay, structure
ABSTRACT: This paper presents the numerical analysis of an embankment founded on a soft structured clay
deposit. The paper investigates the effect of modelling the destructuration of the natural clay foundation. A
comparison of the embankment behaviour is presented when its soft clay foundation is modelled with a new
constitutive model, which can account for destructuration in natural clays and a pre-existing model applicable to
reconstituted clays. The paper demonstrates that taking account of the destructuration of the soft clay foundation
results in a lower embankment failure height and a different failure surface.
1 Introduction
Natural clays differ from reconstituted clays in a number of ways, due to the different structure that they posses
compared to reconstituted clays. The difference in structure between natural and reconstituted clays causes the
natural clays to exist at stress states that plot outside the state boundary surface of the reconstituted clays.
Shearing of many natural clays can cause a breakdown of this structure, resulting in many cases in the natural
soil moving towards the reconstituted state boundary surface. Several frameworks have been developed trying to
bring coherence to many aspects of the behaviour of natural clays (Burland, 1990; Leroueil and Vaughan, 1990;
Cotecchia and Chandler, 2000). Moreover, a number of constitutive models have been developed to model the
effects of destructuration in natural clays (e.g. Rouiania and Muir Wood, 2000; Kavvadas and Amorosi, 2000;
Koskinen et al., 2002, Baudet and Stallebrass, 2004), many of them based on the ideas put forward by Gens and
Nova (1993) of incorporating destructuration within the elastoplastic framework.

However, to date only a few models which account for destructuration have been fully implemented in finite
element codes. Hence, there are only limited applications of these models in boundary value problems
(Karstunen et al., 2005; Koskinen and Karstunen, 2006). In most cases, constitutive models for reconstituted
clays are still used to model natural clays, ignoring the effects of destructuration.

This paper presents the application of a constitutive model, which incorporates the effect of structure in natural
soils, in the numerical analysis of an embankment founded on a soft clay deposit. The model is an extension of
an existing kinematic hardening model for reconstituted clays. Both models have been implemented into the finite
element code ICFEP and this code has been used for the analyses presented in the paper. The paper presents a
comparison of the behaviour of the embankment when its soft clay foundation is modelled with the new
constitutive model for natural clays and the pre-existing model applicable to reconstituted clays. In this way the
paper aims to demonstrate the effect of taking account of the destructuration of the clay foundation on the
predicted behaviour of the embankment.
2 Problem analysed
The problem analysed in this paper aims to be typical of an embankment constructed on a soft structured clay
deposit. The chosen geometry of the embankment and the behaviour of the clay deposit are based on a test
embankment built on a soft structured Champlain clay deposit in Saint-Alban, Quebec (La Rochelle et al. (1994)).
Figure 1 shows the cross section of the embankment adopted in the analyses.
4041

1
.
5
m
2
:
1
1
.
5
:
1
23.7m 4.6m
A B
u
v

Figure 1. Cross section of embankment
3 Constitutive models adopted in analyses
3.1 Clay deposit
The new model for natural structured clays is an extension of an existing two-surface kinematic hardening model
for reconstituted clays (Grammatikopoulou, 2004; Grammatikopoulou et al. 2007). The latter is a version of the
two-surface bubble model (Al-Tabbaa and Wood, 1989) which employs a single kinematic yield surface within
the modified Cam Clay bounding surface. The extension of the two-surface model to incorporate the effects of
structure in natural clays follows the modifications made by Baudet and Stallebrass (2004) in a similar three-
surface model. Figure 2 shows the surfaces adopted in the extended model. The inner surface is a kinematic
yield surface which encloses the region of elastic response. The outer surface represents the natural bounding
surface, which is assumed to be larger than the reconstituted bounding surface. The size of the natural bounding
surface is represented by the term sp'
o
(see Figure 2), where s is the current sensitivity and p'
o
defines the size of
the reconstituted bounding surface. With plastic straining the sensitivity s reduces and the natural bounding
surface collapses towards the reconstituted one. Following Baudet and Stallebrass (2004) the model assumes
that the final value of sensitivity can be higher than unity. In this way it is possible to model soils which retain
some stable elements of structure. The destructuration law adopted in the model is as follows (Baudet and
Stallebrass, 2004):
( )
d f
s s
k
s

& &

=
* *
(1)
where

and

are the slopes of the isotropic normal compression line and the elastic part of the swelling line of
the reconstitured clay in lnv-lnp'

space respectively, k is a parameter which describes the rate of destructuration,
s
f
is the final value of sensitivity and &
d
is the destructuration strain rate. The destructuration strain rate is
assumed to be a function of the plastic volumetric and deviatoric strain rates. Baudet and Stallebrass (2004)
assumed equal contributions of the volumetric and deviatoric plastic strains. In this model an equation of the form
proposed by Rouainia and Muir Wood (2000) is adopted:
( )( ) ( )
2 2
1
p
d
p
v d
E B B
&
& & + = (2)
where
p
v
& is the plastic volumetric strain rate,
p
d
E
&
is the plastic deviatoric strain rate and B is a non-dimensional
scaling parameter, which controls the contribution of the volumetric and deviatoric plastic strains.

This extension of the two-surface model for natural clays requires the following extra parameters: s
f
, the final
sensitivity, k, the parameter controlling the rate of destructuration and B, the parameter controlling the
contribution of the plastic volumetric and deviatoric strains in the destructuration law. Moreover, it is necessary to
define the initial value of the sensitivity, s
i
. Appendix 1 gives the basic equations of the model, which is referred to
as the 2-SKH-S model.

Figure 2. 2-SKH-S model: Representation of surfaces in triaxial stress space
From the above equations it is obvious that if the values of the initial and final sensitivities are assumed equal to
unity then the model degenerates to the pre-existing model for reconstituted clays. The pre-existing two-surface
p' p'
a

q
a

q
sp'
o

g()
Natural bounding surface
Kinematic
yield surface
1
critical state line
4042

model for reconstituted clays is referred to as the 2-SKH model.
3.2 Embankment fill material
The embankment fill material was modelled as a linear elastic-perfectly plastic material with a Mohr-Coulomb
yield surface and a non-associated flow rule.
4 Model parameters
4.1 Clay deposit
The clay deposit was modelled with the 2-SKH and the 2-SKH-S models and the parameters are summarised in
Appendix 1. The parameters adopted for the 2-SKH model were those used in previous analyses of the test
embankment at Saint-Alban (Grammatikopoulou, 2004; Grammatikopoulou et al. 2002). The parameters common
to both models were taken to be the same for the 2-SKH-S and 2-SKH models (with only the parameter in the
hardening modulus slightly modified, refer to Appendix 1). For the extra parameters required in the 2-SKH-S
model the following values were adopted: the initial sensitivity was assumed equal to s
i
= 15, the final sensitivity
was assumed equal to s
f
= 1 (i.e. the clay was assumed to have no stable elements of structure), the parameter k
was taken equal to k = 1.8 and the parameter B was assumed equal to B = 0.8 (hence the destructuration was
assumed to be more distortional rather than volumetric). These parameters were chosen to represent the
destructuration of a generic soft clay deposit and do not specifically represent the soft Champlain Clay at St.
Alban.

In both models an effective stress ratio, K
o
, of 0.67 was assumed. For the 2-SKH model an overconsolidation
ratio, OCR, profile equal to that adopted in previous analyses was used (Grammatikopoulou (2004),
Grammatikopoulou et al. (2002)). As shown in Figure 3, this profile reduces from the ground surface to a depth of
2m and remains constant thereafter at a value of 2.0. This OCR profile results in the variation of the size of the
bounding surface size (i.e. the term p'
o
) with depth shown in Figure 4. This variation of p'
o
with depth results, in
turn, in the undrained strength profile in triaxial compression plotted in Figure 5. This undrained strength profile,
shows an increased strength in the clay curst, which is typical of a soft clay deposit. In the 2-SKH model the
undrained strength is the ultimate strength as no softening is predicted by the model.
0
2
4
6
8
10
12
14
0 5 10 15
OCR, YSR
D
e
p
t
h

(
m
)
OCR 2-SKH
YSR 2-SKH-S

Figure 3. OCR and YSR profiles for the clay deposit
The yield stress ratio, YSR, adopted for the 2-SKH-S model is plotted alongside the OCR profile in Figure 3. This
results in the size of the natural bounding surface, defined by the term s
i
p'
o
, varying with depth as shown in Figure
4. The YSR ratio was chosen so that the peak strength predicted by the 2-SKH-S was similar to the undrained
strength predicted by the 2-SKH model. Figure 5 plots the peak undrained strength predicted by the 2-SKH-S
model in triaxial compression alongside the undrained strength predicted by the 2-SKH model (in the figure the
two lines plot almost on top of each other).

Figure 6 illustrates the stress-strain curves obtained from single element finite element analyses modelling an
undrained triaxial compression test at a depth of 2m (in the figure J is the generalised deviatoric stress which in
the case of a triaxial test is equal to J = q/3 and E
d
is the corresponding generalised deviatoric strain which in
the case of an undrained triaxial test is equal to E
d
= 3
a
, where
a
is the axial strain). This figure shows that the
2-SKH-S model predicts a peak strength which is close to the ultimate strength predicted by the 2-SKH model.
The strength of the 2-SKH-S model then reduces significantly with straining. On the other hand, the 2-SKH
4043

model, which cannot model the process of destructuration, develops a stable ultimate strength.
0
2
4
6
8
10
12
14
0 50 100 150 200 250
p' o, si p' o (kPa)
D
e
p
t
h

(
m
)
p'o 2-SKH
sip'o 2-SKH-S

Figure 4. Profiles of p'
o
and s
i
p'
o
for the clay deposit
0
2
4
6
8
10
12
14
0 10 20 30 40 50 60
Undrained strength, Su (kPa)
D
e
p
t
h

(
m
)
2-SKH
2-SKH-S

Figure 5. Undrained strength profiles in triaxial compression for the clay deposit
0
2
4
6
8
10
12
14
0 5 10 15 20
Ed (%)
J

(
k
P
a
)
2-SKH
2-SKH-S

Figure 6. Stress-strain curves predicted by the 2-SKH and 2-SKH-S models for an undrained triaxial compression
test at a depth of 2m
It should be noted that, in these analyses the 2-SKH model does not actually model the behaviour of the truly
reconstituted clay. The behaviour of the truly reconstituted clay would be defined by a reconstituted bounding
surface which can be calculated by dividing the size of the natural bounding surface, i.e. the term s
i
p'
o
, with the
initial sensitivity s
i
. This truly reconstituted response would result in a stress-strain curve which would plot much
lower than the stress-strain curve plotted in Figure 6 and would also result in a much lower ultimate strength. It is
obvious that an analysis modelling the truly reconstituted clay would predict a much lower embankment height
4044

than the analysis modelling the natural clay. However, this is not what this paper aims to demonstrate. The
purpose of this work is to investigate how a model which does not account for destructuration but gives a good
estimate of the peak undrained strength can simulate the response of a natural clay and how the predictions of
this model compare with the predictions of a model which can simulate destructuration.
4.2 Embankment fill material
For the embankment fill material a Youngs modulus, E, of 10000 kPa and a Poissons ratio, , of 0.3 were
assumed. The angle of shearing resistance, , was taken equal to 44 and the angle of dilation, , was assumed
to be equal to 22.
5 Finite element analyses
5.1 Analyses details
Two analyses were performed, one in which the clay deposit was modelled with the 2-SKH model and one in
which it was modelled with the 2-SKH-S model. Both analyses were carried out using the finite element code
ICFEP.

Figure 7 shows the finite element mesh adopted in the analyses. Both analyses were plane strain and used eight
noded isoparametric quadrilateral elements with 2x2 integration. A modified Newton-Raphson scheme, with an
error controlled sub-stepping stress point algorithm, was used as the non-linear solver (see Potts and Zdravkovic,
1999).

The building of the embankment was simulated with the construction of successive layers, until failure occurred.
The clay was assumed to behave undrained whereas the fill was assumed to behave drained (the water table
was assumed to be at a depth of 0.7m below the ground surface). At the beginning of the analyses the kinematic
surface was assumed to be centred around the current stress state.



Figure 7. Finite element mesh
-0.25
-0.2
-0.15
-0.1
-0.05
0
0 1 2 3 4 5
Embankment Height (m)
H
o
r
i
z
o
n
t
a
l

d
i
s
p
l
a
c
e
m
e
n
t

(
m
)
2-SKH
2-SKH-S


Figure 8. Horizontal displacement at the toe of the embankment against embankment height
4045

-0.2
-0.15
-0.1
-0.05
0
0 1 2 3 4 5
Embankment Height (m)
V
e
r
t
i
c
a
l

d
i
s
p
l
a
c
e
m
e
n
t

(
m
)
2-SKH
2-SKH-S


Figure 9. Vertical displacement at the centerline of the embankment against embankment height



(a)



(b)

Figure 10. Vectors of increment displacement (a) 2-SKH model (b) 2-SKH-S model
5.2 Analyses results
Figure 8 shows the predicted horizontal displacement at the toe of the embankment (point B on Figure 1) plotted
against the embankment height for the two finite element analyses. The vertical settlement of the original ground
surface below the centreline of the embankment (point A on Figure 1) plotted against the embankment height can
be seen in Figure 9. Both figures show that the predicted deformations are very similar up to an embankment
height of approximately 2.5m. For higher embankment heights the 2-SKH-S model predicted a slightly stiffer
response than the analysis with the 2-SKH model, which agrees well with the picture depicted by the stress-strain
curves in Figure 6. However, the analysis which modelled destructuration (i.e. the 2-SKH-S model) showed a
more brittle response and the embankment failed at a lower height than the analysis in which no destructuration
was modelled. The analysis with the 2-SKH model predicted an embankment failure height equal to 3.95m, as
compared to the analysis using the 2-SKH-S model which predicted a failure height equal to 3.65m.

Figure 10 presents the vectors of incremental displacement close to failure for the two analyses. These vectors
show the directions of movement of the failing soil mass. Their absolute magnitude is not important as it is their
relative magnitudes which actually define the failure mechanism. Also plotted on Figure 10 are the failure
surfaces indicated by these vectors. Comparison of the failure surfaces shows that modelling destructuration
failure surface
failure surface
4046

results in a different failure surface which is narrower and shallower than the failure surface predicted when no
destructuration is accounted for.
6 Summary and conclusions
This paper presents the numerical analysis of an embankment founded on a soft structured clay deposit. The
behaviour of the embankment is examined and compared in two finite element analyses. In the first analysis the
clay foundation is modelled with a constitutive model for reconstituted clays which cannot take account of
destructuration, whereas, in the second analysis the clay foundation is modelled with a constitutive model for
natural clays, which can simulate destructuration.

The model for natural clays is an extension of a pre-existing two-surface kinematic hardening model for
reconstituted clays.

In the two analyses presented here the triaxial compression undrained strength predicted by the model for
reconstituted clays has been chosen to be the same as the peak strength predicted by the model for natural
clays. The main conclusions that can be drawn from the comparison of the two analyses are as follows:
Both analyses predicted similar pre-failure deformations, up to an embankment height of 2.5m.
For higher embankment heights the model which accounted for destructuration predicted slightly stiffer
response, but this reflects the stress-strain curves predicted by the models and hence the input
parameters to the models.
Modelling destructration resulted in a more brittle response close to failure and, more importantly, a
lower embankment failure height.
Taking account of destructuration also resulted in a narrower and shallower failure surface.
The above conclusions imply that if it is desirable to determine a failure height, in the case of an embankment, or
in general a failure load for a soft structured clay deposit, then it is not sufficient to use constitutive models which
do not take account of destructuration simply by modelling the undrained strength to be the same as the peak
strength of the natural clay. The work presented in this paper shows that such an analysis, which ignores
destructuration, can overpredict the failure load. In the case of an embankment this means overprediction of the
failure embankment height and the prediction of a different failure surface.
7 Acknowledgements
The authors would like to thank Dr Nesha Kovacevic of the Geotechnical Consulting Group for his useful
discussions on the finite element analyses presented in this paper.
8 References
Al-Tabbaa A. and Wood D.M. 1989. An experimentally based bubble model for clay. Proc. 3
rd
Int. Conf. Num. Models
Geomech., NUMOG III, 91-99.
Baudet B. and Stallebrass S. 2004. A constitutive model for structured clays, Geotechnique, 54(4), 269-278.
Burland J.B. 1990. 30
th
Rankine Lecture: On the compressibility and shear strength of natural clays. Geotechnique, 40(3), 329-
378.
Cotecchia F. and Chandler R.J. 1997. The influence of structure on the pre-failure behaviour of a natural clay. Geotechnique,
47(3), 523-544.
Gens A. and Nova R. 1993. Conceptual bases for a constitutive model for bonded soils and weak rocks. Proc. of Int. Symp. on
Hard Soils-Soft Rocks, Athens (Greece), 485-494.
Grammatikopoulou A., Zdravkovic L. and Potts D.M. 2002. The effect of a kinematic yield surface on the predicted behaviour of
an embankment. Proc. 8
th
Int. Conf. Num. Models Geomech., NUMOG VIII, Rome (Italy), 553-559.
Grammatikopoulou A. 2004. Development, implementation and application of kinematic hardening model for overconsolidated
clays, PhD thesis, Imperial College, London.
Grammatikopoulou A., Zdravkovic L. and Potts D.M. 2007. The effect of the yield and plastic potential deviatoric surfaces on
the failure height of an embankment. Geotechnique, 57(10), 795-805.
Karstunen M., Krenn H., Wheeler S.J., Koskinen M. and Zentar R. 2005. Effect of anisotropy and destructuration on the
behaviour of Murro test embankment. ASCE, International Journal of Geomechanics, 87-97.
Kavvadas M. and Amorosi A. (2000). A constitutive model for structured soils. Geotechnique, 50(3), 263-273.
Koskinen M., Karstunen M. and Wheeler S.J. 2002. Modelling destructuration and anisotropy of a natural soft clay. Proc. 5
th

Eur. Conf. on Num. Methods Geotech. Engineering, Paris (France), 11-19.
Koskinen M. and Karstunen M. 2006. Numerical modelling of Murro test embankment with S-CLAY1S. Proc. 6
th
Eur. Conf. on
Num. Methods Geotech. Engineering, Graz (Austria), 11-19.
La Rochelle P., Trak B., Tavenas F. and Roy M. 1974. Failure of a test embankment on a sensitive Champlain Clay deposit.
4047

Canadian Geotechnical Journal, 11(1), 142-164.
Leroueil S. and Vaughan P.R. 1990. The important and congruent effects of structure in natural soils and weak rocks.
Geotechnique, 40(3), 467-488
Rouainia M. and Muir Wood D. 2000. A kinematic hardening model for natural clays with loss of structure, Geotechnique, 50(2),
153-164.
9 Appendix 1
Equations for the 2-SKH-S model
Natural bounding surface:

( )
( )
0
4
:
2
1
2
2
2
2
=


=
o
b
o
b
p s
g
p s
p F

s s
(3)
Kinematic yield surface:
( )
( ) ( )
( )
( )
0
4
:
2
1
2
2
2
a a
2
a
=


+ =
o
y
y
p s
R
g
p p F

s s s s
(4)
where p is the mean effective stress, s is the deviatoric stress tensor (tensor quantities are represented by bold
type), p'
o
defines the size of the reconstituted bounding surface, s, is the current sensitivity, p'
a
and s
a
are the
mean effective stress and the deviatoric stress tensor at the centre of the kinematic surface, R is the ratio of the
size of the kinematic surface to that of the natural bounding surface. The functions g(
b
) and g(
y
) define the
shape of the natural bounding and kinematic surfaces in the deviatoric plane, where
b
and
y
are the values of
the Lodes angle for the natural bounding and kinematic yield surfaces respectively. In the analyses presented in
this paper a Mohr Coulomb hexagon was adopted as the deviatoric shape of the yield surface:
( )
3
sin sin
cos
sin


+

= g (5)
where ' is the angle of shearing resistance. The plastic potential surface was assumed to have a circular
deviatoric shape.
The parameters adopted in the 2-SKH-S model are:
*
= 0.215 and
*
= 0.005, the slopes of the isotropic normal
compression and the elastic part of the swelling line of the reconstituted clay in lnv-lnp'

space respectively, '
= 27, the angle of shearing resistance at the critical state, R = 0.06, the ratio of the size of the yield surface to
that of the natural bounding surface, G
e
, the elastic shear modulus (see Grammatikopoulou et al. (2002)),
= 1.7, parameter in the hardening modulus, s
i
=15, the initial sensitivity, s
f
=1, the final sensitivity, k = 1.8,
parameter which controls the rate of destructuration and B = 0.8, parameter which controls the contribution of the
plastic volumetric and deviatoric strains in the destructuration law.
If s
i
= s
f
= 1.0 then the model degenerates to the pre-existing 2-SKH model. The parameters
*
,
*
, ', R and G
e
for the 2-SKH model were assumed to be the same as for the 2-SKH-S model with only the parameter in the
hardening modulus being slightly different for the 2-SKH model (i.e. =1.5 was adopted for the 2-SKH model).

4048

Das könnte Ihnen auch gefallen