1

Hyperplasticity Theory for Bangkok Clay Model

S. Likitlersuang
1
& G. T. Houlsby
2

1. Department of Civil Engineering, Chulalongkorn University, Phayathai Road, Pathumwan, Bangkok,
10330, Thailand, (fceslk@eng.chula.ac.th)
2. Department of Engineering Science, Oxford University, Parks Road, Oxford, OX1 3PJ, UK
Abstract: The main purpose of this paper is the development of a new constitutive soil model emphasising the use of thermodynamic
principles. This new approach to plasticity modelling, termed hyperplasticity, was first developed by Houlsby & Puzrin (2000). This
idea has been further extended to continuous hyperplasticity, in which smooth transitions between elastic and plastic behaviour can be
modelled (Puzrin & Houlsby, 2001b). Applying hyperplasticity to soil modelling, a kinematic hardening model specified by means of
two scalar functionals can accommodate the effect of stress history on stiffness. The rate-dependent calculation for the incremental
stress-strain response is introduced. The final model named kinematic hardening modified Cam-clay (KHMCC) model requires eight
parameters (plus an extra parameter for rate-dependent analysis). The model is developed in terms of triaxial stress-strain parameters
for the purpose of comparison them with experimental data from Bangkok clay. Triaxial test results from the Asian Institute of Tech-
nology (AIT), and cyclic undrained triaxial data from Chulalongkorn University are used for establishing the soil parameters.
1 HYPERPLASTICITY THEORY
1.1 Basic formulations
Houlsby & Puzrin (2000) describe an approach to plasticity
based on generalised thermodynamics, which is called hyperplas-
ticity theory. In that approach, the entire constitutive model is
specified by two scalar potential functions: an energy function
and a dissipation function (or yield function). Table 1 summa-
rises the formulations that define the constitutive law based on
Gibbs free energy (g) and Helmholtz free energy (f). Note that
thermal effects will be ignored in this research.

Table 1. Basic formulations for hyperplasticity theory
Gibbs free energy (g) Helmholtz free energy (f)
Energy
function
( )
ij ij
g g , = ( )
ij ij
f f , =
Dissipation
function
( ) 0 , , =
ij ij ij
g
d d & ( ) 0 , , =
ij ij ij
f
d d &
Generalised
stress
ij
ij
g

=
ij
ij
f

=
Formula for
stress and
strain
ij
ij
g

=
ij
ij
f

=
Dissipative
generalised
stress
ij
g
ij
d

&

=
ij
f
ij
d

&

=
Yield
function
( ) 0 , , = =
ij ij ij
g
y y ( ) 0 , , = =
ij ij ij
f
y y
Formula for
flow rule
ij
g
ij
y

=
&

ij
f
ij
y

=
&


1.2 Continuous hyperplasticity
Puzrin & Houlsby (2001a) describe the idea of extending the sin-
gle yield surface to multiple yield surface hyperplasticity. It is
still based on the concept that the entire constitutive response is
derived from two scalar potential functions. Finally, these are
further extended to be internal functions, which represent an infi-
nite number of internal variables. The principal advantage of this
development is that it provides realistic modelling of kinematic
hardening and smooth transitions between elastic and elastic-
plastic behaviour. Table 2 shows comparisons of the multiple in-
ternal variables and internal function formulations which define
the constitutive law based on Gibbs free energy (g).

Table 2. Examples of comparisons between multiple internal
variable and internal function formulations
Multiple internal
Variables
Internal function
Variables
) ( ) (
, , ,
n
ij
n
ij ij ij
( ) ( )
ij ij ij ij
, , ,
Typical
energy
function(al)
) , , , (
) ( ) 1 ( N
ij ij ij
g K
d g
ij ij
Y
) ), ( , (

Typical dis-
sipation
function(al)
) , ,
, , , , (
) ( ) 1 (
) ( ) 1 (
N
ij ij
N
ij ij ij
g
d


& K &
K



d
d d
ij
ij ij
Y
g g
) ), (
), ( , (

&

=

Typical
yield
function(al)
0 ) , ,
, , , , (
) ( ) 1 (
) ( ) 1 (
=
N
ij ij
N
ij ij ij
g
y


K
K

0 ) ), (
), (

, (
=
=

d
y y
ij
ij ij
Y
g g

Typical
derivatives
) (
) (
n
ij
n
ij
ij
ij
g
g

=

( )
( )

ij
ij
Y ij
ij
g
d
g

=


1.3 Kinematic hardening model
The kinematic hardening model for hyperplasticity was fully de-
veloped by Puzrin & Houlsby (2001a). The hardening term is
expressed in the energy function. An example of a one-
dimensional kinematic hardening is shown in Figure 1(a) and its
stress-strain response in Figure 1(b). The Gibbs free energy
function (g) and the yield function (y) are presented as in (1) for
this case:
2

+ =
2 2
2 2
H
E
g , 0 = = c y (1)
The kinematic hardening model is a key stage in developing a
model with multiple internal variables and its extension to a con-
tinuous internal functional. Only the continuous kinematic hard-
ening functional model can accommodate a smooth transition of
stiffness during unloading and reloading under the cyclic stress
history.


c
E
H

(a)
E

c
-c
2c
H E
H E
+

(b)
Figure 1. One-dimensional kinematic hardening model,
(a) Friction slip and spring model;
(b) Loading and unloading response
Based on the kinematic hardening function, the bilinear stress-
strain behaviour can be described by a single internal variable.
By using multiple internal variables, this idea can be extended to
explain a piecewise linear stress-strain behaviour. Finally, for
non-linear stress-strain behaviour, a continuous internal function
is required. The comparison of the bilinear, piecewise and non-
linear stress-strain behaviours is illustrated in Figure 2.

(a)

(b)

(c)
Figure 2. Stress-strain relationships
(a) Bilinear; (b) Piecewise linear; (c) Non-linear

2 DEVELOPMENT OF SOIL MODEL
2.1 Development of KHMCC Model
The following section describes the development of a new consti-
tutive model for soil mechanics, which begins with the one-
dimensional elasticity model. Then, the research is manipulated
into the elasto-plastic model respectively with a single internal
variable, multiple internal variables and a continuous internal
function. In addition, the model will extend to two-dimension
including triaxial stress-strain parameters. The description of the
development can be illustrated as a diagram shown in Figure 3.
Elasticity
Linear model
Elasto-plasticity with
multiple yield surface
Elasto-plasticity with
single yield surface
Elastic perfectly
plasticity
Elasto-plasticity with
infinite yield surface
Logarithmic model
1
5
4
3
2
1- D Triaxial 1- D Triaxial
Stage

Figure 3. Development of the kinematic hardening models
2.2 KHMCC model
For the triaxial stress-strain model, the energy formulations are
expressed by (2) and (3) for the linear and logarithmic models re-
spectively. Full details of this model are given by Likitlersuang
(2003).
( )

+ + + =
1
0
2 2
2 2
2

6 2


d
H H
q p
p g
q
K
p
g
q q p p
q p
x

0 /
2 2 2
= + = c M y
q p

(2)
( )

|
|
.
|

\
| +
+ +
|
|
.
|

\
|

|
|
.
|

\
|
=
1
0
2
0
2
0

2 / 3
exp

6
1 log

d
g
p q p
p g
q
p
p
p g
q p
q p
x
0 /
2 2 2
= + = c M y
q p

(3)
2.3 Continuous hardening function
The continuous hardening functions are required to define the
stress-strain response for the continuous hyperplasticity model.
In this paper, only the case of the linear stress-strain relation is
considered. A hyperbolic curve, which can control the asymptote
(c) and the rate of change of slope by adjusting two additional pa-
rameters (a, b), is shown in Figure 4(a). A hardening function
for the general hyperbolic curve is in the form:
( )
( ) 1 2
1

=
a
E
H
b

(4)
The modified version of the hyperbolic hardening function is
shown in (5).
( )
( ) 1 2
1

=
a
r E
H
b

(5)
The modification gives a sloping asymptote (when ap-
proaches to 1), which is controlled by the parameter r, see Figure
4(b). The value-r should be between 0 (asymptote approaches
vertical) and 1 (horizontal asymptote).
1
0.5
/c
/c
E
f(a)

(a)
1
0.5
/c
/c
E
f(a)
f(r)

(b)
Figure 4. Hyperbolic stress-strain backbone curve
(a) Horizontal asymptote; (b) Sloping asymptote
2.3 Evaluation of hardening parameter
Using the hardening function in (5), the linear consolidation
hardening function can be written as
( )
( ) 1 2
1

=
p
b
p
a
r K
H
p

, which
has four parameters (K, a
p
, b
p
and r) in total. The following are
four conditions defining the components of the consolidation
curve:
1. Initial bulk modulus
2. Tangential bulk modulus at particular stress
3. Plastic modulus (or asymptotic slope)
4. Secant modulus at particular stress
3
Thus, the above four conditions are adequate to estimate the
linear consolidated parameters.
On the shearing behaviour, the shear modulus (G) is assumed
to be linearly proportional to the mean stress that is p g G
x
= .
The shear hardening function is assumed by (4), which is
( )
( ) 1 2
1 3

=
q
b
x
q
a
p g
H
q

. The additional three parameters (g

x
, a
q
, b
q
)
are evaluated by using the shear stress-strain curve. Firstly, the
g
x
value can be obtained from the initial shear modulus (shearing
after the consolidation process), which is
c x ini
p g G = , where p
c

represents consolidation pressure. The a
q
and b
q
parameters are
calculated using the parametric study based on experimental re-
sults.
3 RATE DEPENDENT CALCULATION
3.1 Stress-strain incremental calculation
The incremental response for rate-dependence can be derived
from the differentiation of the energy function, and the internal
kinematic variable rate ( )
ij

&
is calculated from the flow potential.
For example, assuming that the model is defined based on the
Gibbs free energy function; the incremental response can be ex-
pressed in the form:
dt d
w g
d
g
d
kl kl ij
kl
kl ij
ij

|
|
.
|

\
|

=


2 2

(6)
dt
w g
d
g
d
kl kl ij
kl
kl ij
ij

2 2

=
(7)
In addition the boundary conditions during the calculation
(undrained condition, stress control etc.) are introduced by the
controlling statement:
dt T d E d S
ij kl ijkl kl ijkl
= +
(8)
where S and E are stress and strain control matrices, and T is
the strain or stress rate vector.
Substituting
ij
d into (8), the equation becomes:
( ) ( ) { }dt d

= + C B E T d A E S (9)
where
kl ij
g


=

2
A ,
kl ij
g

2


= B , and
ij
w

= C .
3.2 Error control
The accuracy of the rate-dependent calculation algorithm mainly
depends on the time step and the viscosity coefficient ( ) . A
rate-independent process can be approached by assuming a small
value of viscosity. A suitably small time step has to be used in
order to avoid instability in the solution.
dt
w
w
q
g
q
g
p
g
p
g
n
dq
dp
q
g
p q
g
q p
g
p
g
n
i
i
q
i
p
i
q
i
p
i
q
i
p

|
|
|
|
|
.
|

\
|

(
(
(
(
(

)
`

=
)
`

(
(
(
(

(
(
(
(

+
(

=1
2 2
2 2
2
2 2
2
2
2
1
0 1
1 0
0
0 1
1 0
0 0
0 0

&

(10)
An example of triaxial stress-strain parameters for the rate-
dependent calculation algorithm is presented in (10). Assume
that there are n yield surfaces; the governing matrix for strain
controlled triaxial undrained test as presented in (10).
The logic of the calculation is shown in the flow chart, Figure 5.
Define two potentials:
- Energy function (ex. g)
- Flow potential (w)
Set up rate-dependent constants
( , dt and & )
Solve for { } d
Update state parameter ( )
Update stress and strain
Set up initial condition
and Specify control statement
(S, E and T)
Calculate matrix components
(A, B, and C)

Figure 5. Calculation flow chart of rate-dependence
4 COMPARISON, RESULTS AND DISCUSSION
Table 3(a). Summary of KHMCC parameters for Bangkok clay
Parameter Value Physical meaning
K 5200 (kPa) Initial bulk modulus
ap 2.0
bp 2.5
r 0.8
Non-linear kinematic hardening pa-
rameter for p-direction
gx 60 Elastic shear modulus gradient
aq 3.5
bq 2.5
Non-linear kinematic hardening pa-
rameter for q-direction
M 0.9
Slope of critical state line in q-p
plane
-
Viscosity coefficient (for rate-
dependent calculation)
Table 3(b). Summary of MCC parameters for Bangkok clay
Parameter Value Physical meaning

0.045
Slope of swelling consolidation line
on log v-log p plot

0.38
Slope of virgin consolidation line on
log v-log p plot
gx 60 Elastic shear modulus gradient
M 0.9
Slope of critical state line in q-p
plane
-
Viscosity coefficient (for rate-
dependent calculation)
4
A hyperplasticity model called the kinematic hardening modified
Cam-clay model (KHMCC) in terms of triaxial stress-strain pa-
rameters has been developed as presented in the above section.
Some comparisons between experimental data and model predic-
tions based on a particular set of parameters will be presented
here. The experimental data for Bangkok clay have been ob-
tained from AIT (Likitlersuang, 2000) and Chulalongkorn Uni-
versity (Teachavorasinskun et. al, 2001).
The predictions of the KHMCC model are compared with all
experimental data. Both linear and logarithmic stress-strain rela-
tions are used. The linear model, however, seems to provide a
better result than the logarithmic model, especially on swelling
data from consolidation tests. Another advantage of using the
linear model is the simplicity of the calculation. Tables 3(a) and
(b) summary all parameters and their physical meaning based on
Bangkok clay experimental data for the KHMCC and the MCC
model respectively.
4.1 Undrained triaxial prediction
The KHMCC model is used to simulate a set of isotropic con-
solidation triaxial undrained compression tests for OCR-values
varying from 1.00 to 4.25. The normalised stress paths are
shown in Figure 6. Figures 7 and 8 present the KHMCC model
predictions of the isotropic undrained compression tests with
OCR values of 1.00 and 2.15. The figures also show a compari-
son between the KHMCC model and the MCC model. It shows
that the normally overconsolidated clay (Figure 7) is fully pre-
dicted by both the MCC and KHMCC model. The undrained
stress-strain response of normally consolidated clay exhibits only
the plastic behaviour because the stress path drags along the yield
surface boundary, which is only in the plastic region. Thus, the
MCC model can capture this special behaviour.
0.00
0.20
0.40
0.60
0.80
1.00
1.20
0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40
p/p
max
q/p
max

Figure 6. Prediction of isotropic consolidation undrianed shear
stress paths varying OCR from 1.00 to 4.25

On the other hand, the lightly and moderately overconsoli-
dated samples can still use the MCC model to predict their be-
haviour; however it does not fit very well for both stress path and
stress-strain responses. This is due to the limitation that the
MCC model only gives the elastic response for the stress state in-
side the yield surface. The multiple yield surface model can ac-
commodate the smooth transition from elastic stiffness to elastic-
plastic stiffness. Figure 8 shows that the KHMCC model can
capture the transition of shear stiffness. However, the undrained
stress path pattern (Figure 6) does not fit well with all tests, espe-
cially heavily overconsolidated samples. This is because the
KHMCC model is strictly based on the CSSM assumption, i.e.
every sheared sample must reach the same final critical state.
The experimental data however did not show a unique critical
state.
0.0
50.0
100.0
150.0
200.0
250.0
300.0
0.00 0.05 0.10 0.15 0.20
s
q (kPa)

Figure 7. Prediction of isotropic consolidation undrained test at
OCR = 1.00
0.0
50.0
100.0
150.0
200.0
250.0
300.0
350.0
400.0
0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.16
s
q (kPa)

Figure 8. Prediction of isotropic consolidation undrained test at
OCR = 2.15

The KHMCC model cannot accurately give the prediction of
the undrained shear strength, especially for the heavily overcon-
solidated sample. The most significant advantage over the MCC
model is that it can give a smooth transition of stiffness without
any additional expression.
4.2 Drained triaxial prediction
A set of data for isotropic consolidation triaxial drained tests has
been obtained in the testing program with variation of stress path
direction from 0
o
-360
o
. The data are compared with the predic-
tion for the models and some selected results are presented in
Figures 9 and 10. The figures also show a comparison between
the KHMCC model and the MCC model.
The samples are consolidated to 180kPa and then swelled
back to 67kPa. Thus, every sample is in a moderately overcon-
solidated state (OCR about 2.75) after the isotropic consolidation
history path. Then, the samples are sheared in the particular di-
rection with strain being controlled. The artificial rate-dependent
process is used for this prediction, using a & value that is equal
to 100Pa following the rate-effect in the consolidation behaviour
prediction section.

Model

Data

Model

MCC

Data

Model

MCC

Data
5
Figures 9(a) and (b) show the volumetric and shear stress-
strain curves of the CID-56 and CID-236 tests as illustrated by
the small diagram. There are three stress-strain curves, i.e. data,
MCC model prediction and KHMCC model prediction.
20.0
40.0
60.0
80.0
100.0
120.0
140.0
160.0
180.0
200.0
-0.100 -0.050 0.000 0.050 0.100 0.150
p(kPa)
v (-)

(a)
-80.0
-30.0
20.0
70.0
120.0
170.0
-0.10 0.00 0.10 0.20 0.30
s(-)
q(kPa)

(b)
Figure 9. Prediction of isotropic consolidation drained test at 56
o

and 236
o
direction of OCR = 2.75; (a) Volumetric stress-strain
curve; (b) Shear stress-strain curve
-120.0
-70.0
-20.0
30.0
80.0
-0.20 -0.10 0.00 0.10 0.20
s(-)
q(kPa)

Figure 10. Prediction of shear stress-strain curve of isotropic con-
solidation drained test at 90
o
and 270
o
direction of OCR = 2.75
Figure 10 presents the shear stress-strain curve of the CID-90
and CID-270 tests. Due to the fact that the MCC model does not
describe the transition of stiffness, the stress-strain response from
the MCC prediction gives a sudden change of stiffness from elas-
tic to plastic. The KHMCC model not only accommodates the
change of stiffness, but also gives a better prediction of shear
strength as shown in Figure 10. However, the prediction of
KHMCC model exhibits a symmetric stress-strain response about
the p-axis. This limitation of the KHMCC model will be tackled
in the prediction of the stress-strain behaviour with an anisotropic
stress history.
4.3 Cyclic loading prediction
The data on cyclic loading from Bangkok clay have been investi-
gated at Chulalongkorn University. Table 4 represents the defini-
tions of the cyclic stress-strain parameters used in this research.
For the simulation of cyclic tests, the shear modulus gradient
(g
x
) is taken at a slightly higher value than for static tests, i.e. g
x
=
150 in cyclic tests, while g
x
= 60 in static tests. The evidence
from the monotonic undrained compression test found that Chu-
lalongkorns clay samples are slightly stiffer than AITs samples.

Table 4. The definitions of cyclic stress-strain parameters
Definition Description
( )
2 2
3 1

=
SA
; in undrained condition
s SA
5 . 1 =
Single amplitude
shear strain
( )
=
1 2
E
G ; in undrained condition
s
q
G

=
Secant shear
modulus
E
L
c
A
A
h
4
= ; where
L
A and
E
A represent the
area of hysterisis loop and the area of elastic
zone (show in Figure 11)
Damping ratio

3
= -
1
/2

1

SA

(a)
A
E A
L
G
q

s

(b)
Figure 11. (a) Mohrs circle represents the strain parameters for
undrained condition; (b) Definition of damping ratio
Figure 12 present the simulation result of cyclic undrained
shear tests with varying confining pressure (p
c
) 100kPa. The re-
sults exhibit the variation of shear modulus and damping ratio
with single amplitude shear strain. The N-value represents the
number of testing cycles. Although, the testing program has not
been continued until failure (excess pore pressure develops due
to the cyclic load), the first ten cycles are sufficient to represent
the cyclic behaviour.
The prediction of the cyclic test at small strains is the most
important advantage of the multi yield surfaces model. In par-
ticular, the rate-dependent model shows the deformed elliptical
shape of hysteresis loops during the cyclic load. The area of the
hysteresis loop becomes larger when the number of cycles in-
creases. This evidence is exhibited on the damping ratio Figure
and 12(b).

Model

MCC

Data
56
o
236
o
90
o
270
o

Model

MCC

Data
E
L
c
A
A
h
4
=
6
0
1
2
3
4
5
6
7
8
9
10
0.0010 0.0100 0.1000 1.0000
()
G (MPa)
0.32 (data) N=23460
0.56 (data) N=5360
0.80 (data) N=653
0.32 (model) N=10
0.56 (model) N=10
0.80 (model) N=10

(a)
0
5
10
15
20
25
30
35
40
45
50
0.0010 0.0100 0.1000 1.0000
()
hc (%)

(b)
Figure 12. Prediction of cyclic undrained shear test varying shear
stress magnitude at confining pressure 100kPa, (a) Variation of
shear modulus; (b) Variation of damping ratio
5 CONCLUSIONS
The hyperplasticity model based on two potential functions can
define the constitutive law for soil. The advantage is that a con-
tinuous kinematic hardening function can exhibit a continuous
stress-strain response. The KHMCC model has been proposed
for implementation. A parametric study on Bangkok clay is
based on two sources of research. All predictions of the experi-
mental result have been processed by FORTRAN 90 with the as-
sumption that stress-strain is uniformly distributed under the tri-
axial condition (axi-symmetry condition). Every sample is
consolidated following its stress history and sheared in its spe-
cific condition such as undrained, drained or cyclic undrained
condition.
The prediction for consolidation behaviour is more accurate
than shear behaviour. Since, the consolidation prediction only
acts on the p term of the model, there is no p-q coupling term that
affects the stress-strain response.
The prediction for pure shear behaviour can be determined
from the CID-90 and CID-270 tests. These two tests are used to
evaluate the magnitude of the critical state cone both on the com-
pression and extension sides. The evidence on anisotropic tests,
however, shows that the extension strength is slightly higher than
compression. For simplicity, however the assumption that
E C
M M = has been proposed in this study.
The isotropic consolidated tests are accurately predicted in
terms of stress path and shear strength, but the anisotropic con-
solidated samples are slightly inaccurate, especially the stress
path. The variation of shear stiffness is relatively well fitted for
both isotropic and anisotropic consolidated tests. However, the
prediction of the shear strength of the heavily overconsolidated
sample should be improved.
The simulation of the cyclic undrained tests gives an accurate
result. The variation of the shear modulus is simulated by the
model accurately all shear strain amplitudes. The damping ratio
is also well predicted at all shear strain amplitudes. This leads to
the significant benefit of the model that it can simulate both the
monotonic loading and the cyclic loading behaviour by a unique
constitutive law. However, this version of the hyperplasticity
model still has to be improved. The suggested information is
summarised below:
1. Because the model is based strictly on the Masing rule; it
always gives a symmetric closed hysteresis loop when the stress
path returns to the same maximum past stress. Further research
has to introduce a ratchetting mechanism that will provide an un-
symmetrical hysteresis loop and a small drift after one load cycle.
2. The investigation shows that it is not necessary to strictly
enforce the CSSM assumption that every sheared sample must
reach the same critical state. Further research should consider ei-
ther the concept of localisation or the strain softening phenome-
non. The localisation idea assumes that material consists of two
zones i.e. elastic and plastic zones, where the ratio of these two
zones volume controls the material behaviour. In contrast, the
strain softening will result in a softening behaviour after material
yield. The yield surface will move inward or shrink when the
stress state touches the yield surface. These two concepts might
be used to explain the heavily overconsolidated behaviour.
3. The fact that the magnitude of the critical state cone (M)
depends on where the stress state is in stress space. For the gen-
eralised stress-strain model, the general expression to evaluate
the shape of critical state cone has to be considered.
4. For the better prediction of anisotropy, the yield surface
shape may be improved by, for example, deforming the yield sur-
face shape or rotating the yield surface. These ideas are believed
to be able to generate anisotropy.

ACKNOWLEDGEMENTS
The authors are grateful Prof. A.S. Balasubramaniam for provid-
ing triaxial experimental data from AIT, and to Ass. Prof. S.
Teachavorasinskun for offering a set of cyclic undrained triaxial
data from Chulalongkorn University.

REFERENCES
Houlsby, G.T. and Puzrin, A.M. (2000), A thermomechanical
framework for constitutive models for rate-independent dissi-
pative materials, International Journal of Plasticity, Vol. 16
No. 9, pp 1017-1047
Likitlersuang, S. (2000), Elasto-plastic analysis of soft clay be-
haviour, M. Eng. Thesis, AIT, Bangkok
Likitlersuang, S. (2003) A hyperplasticity model for clay behav-
iour: an application to Bangkok clay, D.Phil. thesis, Oxford
University, submitted
Puzrin, A.M. and Houlsby, G.T. (2001a), Fundamentals of ki-
nematic hardening hyperplasticity, International Journal of
Solids and Structures, Vol. 38, pp 3771-3794
Puzrin, A.M. and Houlsby, G.T. (2001b), A thermomechanical
framework for rate-independent dissipative materials with in-
ternal functions, International Journal of Plasticity, Vol. 17,
pp 1147-1165
Teachavorasinskun, S., Thongchim, P., and Lukkunaprasit, P.
(2001), Shear modulus and damping ratio of clay during
undrained cyclic loading, Geotechnique, Vol 51, No. 5, pp
467-470.