You are on page 1of 18

Color profile: Disabled

Composite Default screen

157

A new elastic viscoplastic model for time-


dependent behaviour of normally and
overconsolidated clays: theory and verification
Jian-Hua Yin, Jun-Gao Zhu, and James Graham

Abstract: This paper presents a new three-dimensional elastic viscoplastic (3D EVP) constitutive model for the time-
dependent stressstrain behaviour of both normally consolidated and overconsolidated clays. There are two major new
developments in the 3D EVP model: (i) incorporation of a nonlinear creep function with a limit for the creep volume
strain under isotropic stressing conditions; (ii) adoption of a new loading surface function with a smooth shape on the
plane. Based on the equivalent time concept of Yin and Graham, the formulation of the new 3D EVP model is
derived using the new nonlinear creep function and loading surface function. A new approach to determining all model
parameters is presented. The 3D EVP model is calibrated using data from one conventional isotropic stress creep test
and consolidatedundrained triaxial tests on soft Hong Kong marine deposits (HKMD). The model is verified by com-
paring model results with those from both normally and overconsolidated triaxial tests on HKMD and a mixture of
kaolin and bentonite.
Key words: creep, time-dependent, viscoplastic, equivalent time, stressstrain, constitutive model, verification, clay.
Rsum : Cet article prsente un nouveau modle de comportement tridimensionnel lastique visco-plastique (3D EVP)
pour tudier le comportement contraintedformation en fonction du temps des argiles tant normalement consolides
que surconsolides. Il y a deux dveloppements majeurs dans le modle 3D EVP. Lun est lincorporation dune fonction
de fluage non linaire avec une limite pour la dformation volumtrique en fluage dans la condition de mise en con-
traintes isotropes. Lautre est ladoption dune nouvelle fonction de chargement de surface avec une forme lisse sur le
plan . Sur la base du concept de temps quivalent de Yin et Graham, la formulation du nouveau modle 3D EVP
est drive au moyen des nouvelles fonctions, soit la fonction non linaire de fluage et la fonction de chargement de
surface. On prsente une nouvelle approche pour dterminer tous les paramtres du modle. Le modle 3D EVP est ca-
libr au moyen de donnes dun essai conventionnel de fluage sous contraintes isotropes et dessais triaxiaux consoli-
ds et non drains sur les dpts marins mous de Hong Kong (HKMD). Le modle est vrifi par comparaison des
rsultats du modle avec les rsultats des essais triaxiaux tant normalement que surconsolids sur les dpts de Hong
Kong et sur un mlange de kaolin et de bentonite.
Mots cls : fluage, dpendance au temps, visco-plastique, temps quivalent, contraintedformation, modle de
comportement, vrification, argile.

[Traduit par la Rdaction] Yin et al. 173

Introduction modelling of the time-dependent behaviour in one-


dimensional (1D) straining in odometer tests. In parallel, ef-
The time dependency of the stressstrain behaviour of soft forts have been made to study and develop models for time-
soils, especially highly plastic clay, is generally too signifi- dependent stressstrain behaviour of soils under both triaxial
cant to be ignored (Bjerrum 1967; Graham et al. 1983). The stress states and general stress states. A number of models
constitutive modelling of the time-dependent stressstrain have been proposed (Yin 1990).
behaviour of soils has been an active area of research for Most of the time-dependent models developed so far were
five decades and has attracted much attention from the inter- based on the elastic viscoplastic (EVP) modelling frame-
national geotechnical community in recent years. Bjerrum work of Perzyna (1963, 1966). For different elastic visco-
(1967), Leroueil et al. (1985), Yin and Graham (1989), and plastic models, the approaches used to determine the
others have done extensive work on the laboratory study and viscoplastic strain rates may be different. In developing elas-

Received 21 November 2000. Accepted 10 August 2001. Published on the NRC Research Press Web site at http://cgj.nrc.ca on
15 February 2002.
J.-H. Yin.1 The Department of Civil and Structural Engineering, The Hong Kong Polytechnic University, Hung Hom, Kowloon,
Hong Kong, China.
J.-G. Zhu. Research Institute of Geotechnical Engineering, Hohai University, Nanjing, China.
J. Graham. Department of Civil Engineering, The University of Manitoba, Winnipeg, MB R3T 2N2, Canada.
1
Corresponding author (e-mail: cejhyin@polyu.edu.hk).
Can. Geotech. J. 39: 157173 (2002) DOI: 10.1139/T01-074 2002 NRC Canada

I:\cgj\Cgj39\Cgj-01\T01-074.vp
Tuesday, February 12, 2002 9:31:16 AM
Color profile: Disabled
Composite Default screen

158 Can. Geotech. J. Vol. 39, 2002

tic viscoplastic models, the concepts of instant compression which is consistent with the physical implication; (ii) it is ap-
and delayed compression, proposed by Bjerrum (1967), have plicable to both normally consolidated and overconsolidated
been commonly adopted (Kaliakin and Dafalias 1990; Borja soils; and (iii) it is formulated in a general 3D stress space.
and Kavazanjian 1985; Kutter and Sathialingam 1992; Yin
and Graham 1999). The definition of instant and delayed Fundamental assumption in the new 3D
compression is shown in Fig. 1a. By separating the instant EVP model
and delayed compression, Bjerrum developed a timeline
model for 1D straining (odometer test conditions). However, According to Bjerrum (1967), in odometer compression, the
the full implications of the timeline concept in Bjerrums total volume strain of a soil subjected to a stress increment is
model and in subsequent models (Borja and Kavazanjian divided into instant volume strain and delayed (creep) vol-
1985; Kutter and Sathialingam 1992) were not properly de- ume strain, as shown in Fig. 1a. Since Bjerrum proposed
fined, and there was no explanation as to how to derive these concepts, the decomposition of total strains into instant
these time-dependent stressstrain models. Yin and Graham and delayed strains has been widely accepted in both the iso-
(1989, 1994) pointed out that the timeline proposed by Bjerrum tropic and general 3D stress states. There are mainly two
for considering creep compression under 1D straining types of assumptions adopting the separation of instant and
should be a line with a carefully defined equivalent time, delayed deformation. One type of assumption divides total
not the ordinary duration of loading. Otherwise, contradic- strain rates into elastic strain rate and viscoplastic strain rate
tions occur, especially when the clay is overconsolidated. components. A number of researchers used this assumption,
Yin and Graham (1989, 1994) defined equivalent time and for example, Zienkiewicz and Cormeau (1974), Adachi and
other concepts such as instant timeline and reference time- Oka (1982), Desai and Zhang (1987), and Kutter and
line for 1D straining condition. Using these concepts and log- Sathialingam (1992). On the other hand, some researchers,
arithmic functions, they derived a general 1D EVP constitutive such as Borja and Kavazanjian (1985), and Kaliakin and
model for time-dependent stressstrain behaviour of clays and Dafalias (1990), believed that total strain rates might be di-
calibrated and verified the model using test data from three dif- vided into elastic strain rates, plastic strain rates, and
ferent soils. The 1D EVP model has been successfully used in viscoplastic strain rates. In the present study, the former as-
1D consolidation analysis of clay (Yin and Graham 1994; Sun sumption is adopted, i.e., total strain rates & ij are the sum of
1999; Nash and Ryde 2000; Zhu and Yin 2000). elastic strain rates & eij and viscoplastic strain rates & vp
ij :
Yin (1990) and Yin and Graham (1999) have extended the [1] & ij = & eij + & vp
ij
concepts for 1D straining for triaxial and general stress
states and developed an elastic viscoplastic model that uses a where subindices i = 1, 2, 3 and j = 1, 2, 3; and the overdot de-
simple elliptical flow surface and Mises failure criterion. notes the rate of a variable. Here, the instant strains are as-
This original EVP model is considered to be a genuine ex- sumed to be elastic, time-independent, and recoverable, but the
tension of the modified Cam-Clay model (Roscoe and delayed strains are viscoplastic, time-dependent, and irrecover-
Burland 1968) for time-dependent stressstrain of soils (Yin able. A separate elastic relationship is suggested to relate the
1990; Yin and Graham 1999). This EVP model was cali- elastic strain rates & eij in eq. [1] to effective stress rates & kl :
brated and verified using data from triaxial tests on the
sandbentonite mixture proposed in a Canadian program for [2] & eij = C ijkl & kl
the safe disposal of nuclear fuel waste. Yin and Graham
demonstrated that the model predictions were in good agree- where Cijkl is a fourth-order compliance tensor with subindi-
ment with the measured results. ces k = 1, 2, 3 and l = 1, 2, 3. Summation is implied if two
Borja and Kavazanjian (1985), Kutter and Sathialingam subindices of two items are the same. If the elastic deforma-
(1992), Yin and Graham (1999), and Vermeer and Neher tion of soil is assumed to be isotropic, there are only two con-
(2000) have made considerable progress in the development stants (or moduli), for example, shear modulus G e and bulk
of suitable models for predicting the time-dependent stress modulus K e. It is noted that (i) compressive stresses and
strain behaviour of clayey soils. It is recognized by the au- strains are positive in soil mechanics, and (ii) the constitutive
thors, however, that these models have certain limitations. relationship is constructed in terms of effective stresses.
For example, the models by Borja and Kavazanjian and Yin Viscoplastic strain rates & vp
ij in eq. [1] are calculated from a
and Graham are only applicable to normally consolidated flow rule that was originally proposed by Perzyna (1963, 1966):
and slightly overconsolidated clays. The model by Kutter Q Q
[3] & vp
ij = ( F ) = S
and Sathialingam was formulated in the conventional triaxial ij ij
stress space instead of the general three-dimensional (3D)
space. Based on the framework proposed by Yin and Gra- where is referred to as the fluidity parameter (with units of
ham, this paper presents a new 3D EVP model for both nor- inverse time), is a function of F, Q is a viscoplastic potential
mally consolidated and overconsolidated clays in a general function, F represents the difference between the dynamic
stress state and the model verification. loading surface and the static yield surface, and ij is the
Compared with the original EVP model proposed by Yin stress tensor. In eq. [3], + (F), was originally used by
and Graham (1999) and other models, the new EVP model Perzyna. It is noted that + (F), = (F) for F > 0 or + (F), = 0
in this paper has the following features that eliminate limita- for F 0 in Perzynas model. In our EVP model, however,
tions in other models: (i) it uses a nonlinear creep function S is used instead of + (F),. The S is called a scaling function
to fit the relationship between creep strain and time and this and is assumed to be related to a loading (or flow) surface
nonlinear function assumes the creep strain to have a limit, function, which is denoted by F in this paper. The scaling
2002 NRC Canada

I:\cgj\Cgj39\Cgj-01\T01-074.vp
Tuesday, February 12, 2002 9:31:16 AM
Color profile: Disabled
Composite Default screen

Yin et al. 159

Fig. 1. (a) Definition of primary and secondary consolidation and instant and delayed compression. (b) Illustration of the instant
timeline, reference timeline, equivalent timelines, and limit timeline for 1D stressing (extended from Yin and Graham 1999).

function S was first proposed and used by Yin (1990) and Yin state below the limit timeline (to be introduced later), S
and Graham (1999), who suggested an approach to defining would be equal to 0; in any other state, S is greater than 0.
and determining S. This approach is based on the equivalent The elasticplastic behaviour can easily be dealt with using
time concept as introduced in the 1D EVP model by Yin and the new 3D EVP model in two ways: (i) the behaviour
Graham (1994), in which a logarithmic creep function was below the limit timeline is purely elasticplastic (time-
used to fit the creep data of volume strain against time under independent); and (ii) when the viscosity (or creep parame-
isotropic stressing for determination of the S function. ter ) is zero, the 3D EVP model becomes an elasticplastic
Though the logarithmic function has been used widely in en- model (Yin 1990).
gineering practice and research, this function has an errone-
ous limitation in that when creep time is infinite, the creep A new elastic viscoplastic relationship in
strain is infinite. To eliminate this limitation, a nonlinear
creep function (Yin 1999) with a limit is to be used in this pa- isotropic stressing
per to fit the creep data under isotropic stressing. The important concept of equivalent time is used in de-
It should be mentioned that the scaling function S is dif- veloping the elastic viscoplastic relationship. Bjerrum (1967)
ferent from + (F),. For the latter, when F 0, + (F), = 0. mentioned equivalent time for odometer creep behaviour, but
For the scaling function in this paper, in the case of a soil he did not give a clear definition of the concept. In deriving
2002 NRC Canada

I:\cgj\Cgj39\Cgj-01\T01-074.vp
Tuesday, February 12, 2002 9:31:17 AM
Color profile: Disabled
Composite Default screen

160 Can. Geotech. J. Vol. 39, 2002

a general EVP relationship for the time-dependent stress to (i + 1) along an instant timeline (the same slope as that
strain behaviour for a 1D straining condition, Yin and Gra- of the line), and then it will experience creep, i.e., delayed
ham (1989, 1994) developed the concept. They gave a spe- deformation from (i + 1) to (i + 1).
cific definition and an explanation for equivalent time. Yin
(1990) and Yin and Graham (1999) further extended the Reference timeline (or line)
equivalent time concept for the creep behaviour under iso- The reference timeline is written as
tropic stressing. Based on the above-mentioned work, this
pm
section adopts a nonlinear creep function to derive a general [5] rvm = rvmo+ ln
EVP relationship for isotropic stressing using the same V pmo
equivalent time concept.
Figure 1b shows four fundamental concepts used in the EVP where rvmo is the strain at mean effective stress pm = pmo ;
model under 1D (isotropic) stressing: (i) equivalent time te, are three parameters. The term /V
and rvmo, /V, and pmo
(ii) instant timeline (also called line), (iii) reference timeline is similar to that used in the modified Cam-Clay model
( line), and (iv) limit timeline (new in this paper). The term (Burland 1967) for defining the elasticplastic line for
vm in Fig. 1b is the volume strain corresponding to the mean isotropically consolidated specimens in a normally consoli-
effective stress, pm . The definitions and physical meanings of dated stress range. It has been shown by Yin and Graham
these concepts are similar to those in the earlier EVP models (1994) that if the viscous nature is ignored, the reference
(Yin 1990; Yin and Graham 1989, 1994, 1999), except for the timeline in eq. [5] expresses the elasticplastic compression
limit timeline which is newly introduced in this paper. A brief line as in the modified Cam-Clay model. The two parame-
discussion follows on how these concepts can be developed determine a point which the line passes
ters rvmo and pmo
by using a natural logarithmic function for the line and through, i.e., they are used to fix the line. The line cor-
line and a nonlinear function for creep volume strain. responds to zero equivalent time, and hence provides a refer-
ence for counting equivalent time te, which is discussed later
Instant timeline (or line) in this section. As shown in Fig. 1b, te is negative above the
The instant timeline is used to define the instantaneous line and positive below the line. Once the reference
volume strain at any point in Fig. 1b (above or below the timeline is determined, a unique relation of pm vm te is
limit timeline). The volume strain on the instant timeline can then determined and is presented later in this section.
be expressed as
Nonlinear creep compression
pm The logarithmic function has been commonly used to fit
[4] evm = evmo + ln
V pu the creep strain vp
vm versus time or secondary compression
versus time. One form of this kind of function can be written
where pu is a unitreference mean stress, pm is the mean ef- as vp
vm = (/V)ln[(t0 + te)/t0], where t0 is a curve-fitting pa-
fective stress under isotropic stressing, and evmo is the vol- rameter related to the choice of reference timeline (Yin and
ume strain at pm = pu . In eq. [4], V = 1 + e0 is the specific Graham 1989, 1994, 1999). One apparent limitation of this
volume, where e0 is the initial void ratio. The ratio /V is a function is that when time is infinite, creep strain is infinite.
material parameter used in the same way as in the modified An infinite creep volume strain implies that the soil particles
Cam-Clay model (Burland 1967). The instant timeline is ac- occupy zero volume. This is certainly incorrect. Yin (1999)
tually a set of lines with a slope of /V in the pm vm plane. proposed a nonlinear creep function with a limit and found
For example, if a soil in a state at point i in Fig. 1b is loaded that this function can fit creep test data well. This paper
+1, the state of the soil will first instantly move from i
to pm,i adopts this nonlinear creep function for the time-dependent
creep volume strain under isotropic stressing. The nonlinear
creep function is expressed as

o t0 + te o
ln
V t V t +t t +t
[6] vp
vm =
0 = ln 0 e = ln 0 e
o t0 + te o t0 + te t0 V t0
1+ ln 1+ ln
V vp
vml t0 V vp
vml t0

In eqs. [6] and [7], o /V, t0, and vp


vml are three constant param-
eters. If considering ln[(t0 + te)/t0] together as a variable, eq. [6]
where is in fact a hyperbolic function. (The authors have found that a
hyperbolic function using time t (or te) directly cannot provide
o a good fit to the creep data.) In eq. [7], if vp
vml = , the param-
eter /V = o /V. Consequently, eq. [6] would be reduced to
[7] = V the commonly used logarithmic function. From eq. [6], it is
V 1 + o ln t 0 + t e also found that when te = , vp vm = vml . This implies that
vp
V vp
vml t0 there is a limit for the viscoplastic (creep) volume strain.
2002 NRC Canada

I:\cgj\Cgj39\Cgj-01\T01-074.vp
Tuesday, February 12, 2002 9:31:17 AM
Color profile: Disabled
Composite Default screen

Yin et al. 161

With the time increase, the volume strain due to creep will earlier, for the heavily overconsolidated stress range, the state
never exceed (down) this volume strain limit of vp vml . The of soil may be below the limit timeline. In this case, the soil
timeline corresponding to this limit is called a limit is considered to be elasticplastic and time-independent.
timeline as shown in Fig. 1b. If vp vml is a constant, inde- Differentiating eq. [6], one obtains the creep rate at a
pendent of pressure, the limit timeline will be parallel to the point with an equivalent time te :
reference timeline, as shown in Fig. 1b. The vertical distance
o
between the limit timeline and the reference time lime is
d vp 1
equal to the value of vp vml . The state of heavily [8] vm = V
overconsolidated soil will be below the limit timeline. In this dt e t0 + te o t0 + te
2

case, the behaviour of soil is assumed to be elasticplastic 1 + ln


V vml
vp
t0
and time-independent.
It is noted that eq. [6] uses equivalent time as defined and Equation [8] indicates that constant equivalent time te lines
used by Yin and Graham (1989, 1994, 1999). The use of are also lines of constant viscoplastic strain rate, d vp
vm /dt e , for
equivalent time makes it possible to use a single equation, a loading history of creeping from the reference timeline to
for example eq. [6], to describe the time-dependent creep the current state, but this creep rate is derived the same way
above the limit timeline under any loading condition, includ- as that for any other loading history, as explained previously.
ing loading and unloading, and in a so-called normally con- In other words, this creep rate is derived using the equivalent
solidated or overconsolidated state. time and is independent of the loading path or history.
The meaning of the equivalent time te can be simply ex- In fact, it is well known that creep rate is independent of
plained here. There is a soil element in a state at point i in the loading path or history and is a function of stressstrain,
Fig. 1b. The load imposed on the element is increased from and this knowledge is used by almost all researchers. For ex-
i to pm,
pm, i +1. The state of the element will move instantly ample, the creep rates in eq. [3] in the model by Perzyna
from point i to point (i + 1) and then creep to point (i + 1) (1963, 1966) are in fact a function of the stressstrain state
under constant mean effective stress pm, i +1. Here we assume only, independent of the loading history. The simplest
that there is no water coupling, since the model is built for Maxwells rheological model also indicates that the creep
the soil skeleton using effective stresses. For the sake of con- rates are independent of loading history. Laboratory data
venience we have chosen point (i + 1) to be on the refer- support this assumption (Yang et al. 1999; Yin 1999). The
ence timeline (Yin and Graham 1994, 1999). If this is done, equivalent time concept is simply used to determine the
creeping from point (i + 1) to point (i + 1) takes place with creep rate independent of loading history. In other words, the
the equivalent time te equal to the real creep duration t. In creep rate determined using a particular loading history, for
this case, if the time in eq. [6] uses either equivalent time or example, the conventional creep tests, is the same and can be
real time, this equation can correctly express the volume used for any other loading history.
strain under this stress increment of pm, i to pm, i +1. The creep Thus, for any infinitesimal loading, the viscoplastic creep
rate can be calculated using eq. [6]. rate is given by & vp
vm = d vm /dt and, according to the equivalent
vp

Now consider what would happen to the creep rate time concept, there exists & vpvm = d vm /dt = d vm /dt e . This is
vp vp

(viscoplastic strain rate) if the state of point (i + 1) has very important and means that the equivalent time is not the
been reached by a path with a loading and unloading se- real time but is used to obtain lines of constant viscoplastic
quence [i (i + 2) (i + 2) (i + 1)], which is different creep rates (or field). Nash and Ryde (2000) call the lines of
from the path of [i (i + 1) (i + 1)]. The equivalent time constant viscoplastic creep rates a set of isotaches. Using the
concept indicates that the creep rate at point (i + 1) for any equivalent time (or the isotaches), the creep rate is uniquely de-
loading path is the same as that obtained by creeping from fined by the current state of the soil (Bjerrum 1967; Yin and
point (i + 1) on the reference timeline for an equivalent Graham 1989, 1994, 1999; Nash and Ryde 2000).
time te as in eq. [6]. In other words, the equivalent time of a
soil element experiencing a path of [i (i + 2) (i + 2) A new general constitutive relationship for 1D stressing
(i + 1)] is equal to that of an element experiencing a path of condition
[i (i + 1) (i + 1)]. This implies that the creep rate is in- Using the equivalent time concept, a unique general rela-
dependent of the loading path and loading history, and this tionship between pm , vm , and te in a condition of isotropic
can be easily described by the equivalent time concept (Yin stress for the time-dependent stressstrain behaviour can be
and Graham 1989, 1994, 1999). expressed as
According to Yin and Graham (1999), the line 21(i + 1)
p
(i.e., the reference timeline) in Fig. 1b can be very close to [9] vm = rvm + vp
vm = vmo +
r
ln m
the compression line obtained at the end of primary consoli- V pmo
dation (EOP), and thus it may be considered that the soil at
o t0 + te
point (i + 1) is in an overconsolidated state due to ageing ln
V t0
(Bjerrum 1967). According to the equivalent time concept, +
however, this overconsolidation can be reached by either age- o t +t
1+ ln 0 e
ing or loadingunloading, which do not have to be specified. V vp
vml t0
Therefore, the state of a soil element can be uniquely deter-
mined by (pm , vm , te), and the behaviour of volume creep in Equation [9] means that the volume strain at any point under
both the normally and overconsolidated (or unloadingreload- a certain mean effective stress is equal to the volume strain
ing) stress ranges can be described by eq. [6]. As mentioned at the reference timeline plus the creep strain under the same
2002 NRC Canada

I:\cgj\Cgj39\Cgj-01\T01-074.vp
Tuesday, February 12, 2002 9:31:17 AM
Color profile: Disabled
Composite Default screen

162 Can. Geotech. J. Vol. 39, 2002

Table 1. Isotropic consolidation test on Hong Kong marine deposits.


Loading Unloading Reloading
Stage: 1 2 3 4 5 6 7 8 9 10 11 12
Pressure (kPa) 25 50 100 200 400 800 400 200 100 200 400 800
Duration (min) 3260 2820 20 160 2850 2840 2850 600 1232 1500 900 1380 4300
One day vm (%) 4.38 9.28 14.93 20.39 26.30 31.34 30.92 29.76 28.21 28.97 30.20 31.92
End vm (%)a 5.07 9.55 15.54 20.52 26.51 31.58 30.92 29.76 28.21 28.97 30.20 32.17
a
Volume strain at the end of the corresponding loading stage.

Fig. 2. (a) Illustration of the reference timeline and instant timeline. (b) Changes of volume strain with time.

. This equation has an interesting feature, that is,


stress pm similar to the elasticplastic strain in the modified Cam-Clay
when the parameter o /V approaches zero, the total strain is model. From eq. [9], t e is then given by

2002 NRC Canada

I:\cgj\Cgj39\Cgj-01\T01-074.vp
Tuesday, February 12, 2002 9:31:17 AM
Color profile: Disabled
Composite Default screen

Yin et al. 163

Table 2. Values of parameters in the isotropic EVP relationship for HKMD and other parameters.

/V /V (kPa)
pmo o /V t0 (min) vmo
r
vp
vml M n
0.018 0.0793 15.2 0.0025 1440 0 0.601 1.265 6

Fig. 3. Failure surfaces or loading surfaces on the plane. Considering an infinitesimal time increment (dt) and loading
increment for a stress state above the limit timeline, say from
point i to point (i + 1) in Fig. 1b, the incremental volume strain
is expressed as d vm = d evm + d vpvm , where d vm is the elas-
e

tic strain increment from i to (i + 1) (taking no time for elastic


deformation) and d vp vm is the viscoplastic strain increment for
dt. From this, we have d vm /dt = d evm / dt + d vp vm /dt. Since
d vp
vm /dt = d vm /dte (the creep rate due to a different loading
vp

history is the same as that from the reference timeline to the


current point according to the equivalent time concept), we
have d vm /dt = d evm /dt + d vp
vm /dte. Consequently, a general
1D EVP relationship in an isotropic stress condition is derived
as
2
p& m
r
[12] & vm = + o 1 + vm vp vm
Vpm Vt 0 vml


rvm vm V
exp

1 + vm vm o
r

vp
vml
where d vp
vm /dte from eq. [11] is used.


p Determination of 1D EVP model parameters
vm rvmo ln m for Hong Kong marine deposits
V pmo
[10] t e = t 0 + t 0 exp
p There are seven parameters in eq. [12], i.e., /V, o /V, t0,
vm rvmo ln m vp
vml , /V, vmo , and pmo . This section uses the data from an
r
o 1 V pmo
isotropic consolidationcreep test on soft Hong Kong marine
V vml
vp
deposits (HKMD) (clayey silt) to illustrate the methodology
for determination of all parameters in eq. [12].

Substituting eq. [10] into eq. [8], the viscoplastic volume HKMD and isotropic consolidationcreep test
strain rate is written as The HKMD were taken from the seabed (12 m sediment
thickness) at a site in Hong Kong. From wet sieving and hy-
2
d vp
vm = o 1 + vm vm
r drometer tests, the HKMD tested in this study consisted
[11] mainly of clay and silt with some fine sand (27.5% clay with
dt e Vt o vml
vp
particle size < 0.002 mm, 46.5% silt with 0.002 mm < parti-
cle size < 0.063 mm, and 26.0% fine sand with 0.063 mm <
particle size < 0.15 mm). The basic physical properties of the
rvm vm V HKMD were measured (Zhu et al. 1999) and the values are as
exp
follows: specific gravity Gs = 2.66, liquid limit wl = 60%,
1 + vm vm o
r
plastic limit wp = 28%, and plastic index Ip = 32%.
vp One isotropic consolidation test on a remoulded HKMD
vml
specimen was performed in a triaxial cell and pore-water
where rvm is defined as in eq. [5]. pressure was monitored (Zhu 2000). The specimen was
According to eq. [1], the total volume strain rate & vm 50 mm in diameter and 100 mm in height. The initial void ra-
should be the sum of the elastic strain rate & evm and the tio of the sample was e0 = 1.504. Because of the length of this
viscoplastic strain rate & vp vp & &e & vp
vm = d vm /dt, i.e., vm = vm + vm . test, double membranes were used to prevent the possible
By differentiating eq. [4], the elastic strain rate is deter- leakage of water. The test results are summarized in Table 1,
mined as & evm = p& m
/(Vpm
). According to the equivalent time in which volume strains vm at 1 day and at the end of a load-
concept, the viscoplastic strain rate is d vp vm /dt = d vm /dte.
vp
ing step are presented.

2002 NRC Canada

I:\cgj\Cgj39\Cgj-01\T01-074.vp
Tuesday, February 12, 2002 9:31:18 AM
Color profile: Disabled
Composite Default screen

164 Can. Geotech. J. Vol. 39, 2002

Fig. 4. (a) Loading surfaces, total stress path (TSP), and effective stress path (ESP) in qp space. CS, critical state. (b) Corresponding
v (or vm ) versus p, instant timeline, reference timeline, and limit timeline for isotropically consolidated, undrained triaxial test at
constant axial strain rate (extended from Yin and Graham 1999).

Determination of /V, vp vml , o /V, and t0 vp


vml , o /V, and t0. The changes of volume strain with time
The relationship of vm ln pm in the loadingunloading under a pressure of 100 kPa are shown in Fig. 2b.
stages can be considered as linear (see Fig. 2a). Using the As mentioned earlier, vp vml is the limit of viscoplastic vol-
curve-fitting technique, the slope of the instant timeline fit- ume strain in eq. [6]. A special hyperbolic function in eq. [6]
ted to the unloadreload points in the vm ln pm plot is ob- has been demonstrated by Yin (1999) to provide a good fit to
tained as /V = 0.018. the data (or creep strain data) after primary consolidation. The
In all loading stages of the isotropic consolidation test, the determination of vp vml is not an easy task because it is impos-
longest duration of loading was about 1 week. In this stage sible to run a test of infinite duration. However, vp vml will
(stage 3) the consolidation pressure was 100 kPa. After the have an upper bound value. In an extreme case, assuming the
primary consolidation, the test data can be used to determine voids of a soil element (or specimen) all become zero under

2002 NRC Canada

I:\cgj\Cgj39\Cgj-01\T01-074.vp
Tuesday, February 12, 2002 9:31:18 AM
Color profile: Disabled
Composite Default screen

Yin et al. 165

creep at infinite time, the upper bound limit is vp vml,upper = Determination of /V, rvmo, and pmo
e0 /(1 + e0). The real limit will be vp
vml < e0 / (1 + e0). In case The parameters rvmo and pmo are used to fix the reference
no long-term creep data are available, the limit creep strain ) of the point on the reference
timeline. In fact, any ( vm , pm
can be estimated as vpvml e0 / (1 + e0). Since e0 = 1.504, the timeline can be taken as rvmo and pmo . It is convenient to
vml 0.601 for the HKMD in this paper.
limit vp choose them so that when pm = pmo , the strain vm is zero.
The data from stage 3 (vertical stress of 100 kPa) after Therefore, rvmo is generally taken to be zero, as shown in
1440 min are used to determine the parameters in eq. [6], since Fig. 2a. Since point A is on the reference timeline and
after 1440 min the excess pore-water pressure has been dissi-
pvmA = 100 kPa, rvmA = 14.93% (see Table 1); then, noting
pated and the strain increment is all creep strain. In Fig. 2b, vmo = 0, eq. [5] can be written as
r

point A is at time 1440 min. This point is considered here on


100
the reference timeline or the reference timeline passes point A [14] 0.1493 = ln
as shown in Fig. 2a. Thus, the equivalent time is counted start-
V pmo
ing at the real duration time of 1440 min. In other words, point
A in Fig. 2b is taken as a starting point of creep with te = 0 to The load increment from (i + 1) to (i + 2) is now considered
count for the equivalent time te. Equation [6] is then used to (see Fig. 2a). From eq. [4], vm,( i+ 2 ) = vm,( i+1) + ( /V)ln
best fit the creep data after point A as +2 /pm,
(pm,i i +1), since deformation is elastic from (i + 1) to
o t0 + te (i + 2) . According to eq. [9] (the equivalent time concept),
ln the volume strain vm,( i+ 2 ) at point (i + 2) can also be ex-
V t0
[13] vp
vm = ( vm ) total ( vm ) t =1440 min = pressed as
o t +t
1+ ln 0 e i+ 2
V vml
vp
t0 pm,
[15] vm, ( i + 2 ) = rvmo+ ln
V
pmo
where ( vm )t = 1440 min is the volume strain at the elapsed time
t = 1440 min. When carrying out curve fitting, it should be o t 0 + t e, ( i + 2 )
ln
noted that te = 0 corresponds to t = 1440 min. In this way, V t0
+
the two parameters in eq. [13] are determined as t0 = o t 0 + t e, ( i + 2 )
1440 min and o /V = 0.0025, assuming vp vml . The fitted line
1+ ln
is shown in Fig. 2b. The data from t = 0 to t = tA (point A) V vp
vml t0
are in a primary consolidation period and are not fitted by From eq. [15], te,(i+2) can be obtained as in eq. [10]. Noting
the creep function. i + 2 /pm,
that vm,(i +2) = vm,( i+1) + ( /V)ln(pm, i +1) and te,(i+2) =
t + te,(i+2) , the equivalent time te,(i+2) at point (i + 2) is then
obtained as


pm, i+ 2 p
vp
vml vm, i +1 + ln rvmo ln m, i + 2
V i +1
pm, V
pmo

[16] t e, ( i + 2 ) = t t 0 + t 0 exp
o i+ 2 i+ 2
vp pm, pm,

vml vm, i +1 + ln r
ln

vmo
V V i +1
pm, V pmo

where t is the test duration for this load increment. The method discussed above is rigorous. Yin and Graham
From eq. [9], the volume strain vm,( i+2 ) at point (i + 2) (1994, 1999) pointed out that /V is almost the same as the
can be given as slope of the line for normally consolidated specimens used
in the modified Cam-Clay model.
pm, i+ 2
[17] vm, ( i + 2 ) = rvmo+ ln
V
pmo Failure shape function for loading surface
o t 0 + t e, ( i + 2 ) in plane
ln
V t0
+ It is generally believed that the Mohr-Coulomb criterion
o t 0 + t e, ( i + 2 ) can properly express the failure of soil. However, the track
1+ ln
V vp
vml t0 of the failure surface in plane in the principal stress space
is an irregular hexagon. The surface exhibits corners or sin-
Using the load increment from pm, i +1 = 400 kPa to pm,
i+ 2 = gularities in a 3D generalization. To remove the singularity
800 kPa, vm,i+1 = 26.51%, vm,i+2 = 31.58%, and t = of the failure surface of the Mohr-Coulomb criterion, re-
2850 min, the parameters /V and pmo are obtained as 0.0793 searchers have proposed a number of functions (shape func-
and 15.2 kPa by solving eqs. [14], [16], and [17]. The pa- tion) for shaping the yield (failure) surface, such as the
rameters in eq. [12] are determined and listed in Table 2. Gudehus and Argyris function (Chen 1982), the Willam-

2002 NRC Canada

I:\cgj\Cgj39\Cgj-01\T01-074.vp
Tuesday, February 12, 2002 9:31:19 AM
Color profile: Disabled
Composite Default screen

166 Can. Geotech. J. Vol. 39, 2002

Warnke function (Chen 1982), and others. Xin (1988) sug- tests are different. Therefore, the authors suggest that is a
gested a smooth shape function, which is very close to the function of stress load angle and can be empirically deter-
Mohr-Coulomb surface. mined by = [6 + sin(3 /2)]/5.
To develop an EVP model for a general stress state, Xins As shown in Fig. 4, f2 is the part of the loading surface
(1988) shape function for shaping the loading surface and above the critical state line in the pq plane for the behav-
failure is adopted and is written as iour of heavily overconsolidated soil and is expressed by
[18] g ( ) = n
1 q
K [21] F = f2 = p 1 + po
n 1 Mg ( ) p

(2K 1) + (1 K) sin + 3 cos
6 6
where po is the mean effective stress value at which the
where K is a parameter and can be expressed by K = (3 loading surface in eq. [21] intercepts the mean effective
sin )/ (3 + sin ), in which is the effective friction angle of stress p axis in the pq plane (see Fig. 4). In Fig. 4, the
soil; and is the stress load angle defined as light-face broken line is not part of the loading surface but
shows the physical significance of po . The n in eq. [21] is a
1 3 3 J3
[19] = cos 1 parameter which in this study is determined to be 6; this
3 2J23/ 2 value is effective at least for the HKMD and a kaolin clay.
Figure 4a shows the total stress path (TSP) and effective
in which J2 and J3 are the second and third invariants of devi- stress path (ESP) from points A to B and to C of an iso-
ator stress tensors and are given as J2 = sijsij /2 and J3 = tropically consolidated, undrained triaxial compression test
sijsjk ski /3, where sij is the deviator stress tensor and is defined under a constant axial strain rate. Figure 4b shows the corre-
as sij = ij kk ij /3, where ij is the effective stress tensor. sponding state (pm , vm ) of points a, b, and c and the corre-
Figure 3 shows a comparison between the Mohr-Coulomb sponding state (p, v) of points A, B, and C.
surface, Xins (1988) surface, and the Gudehus-Argyris sur- Assuming that in a general stress state (p, q, ) or (1 , 2 ,
face (using the friction angle of the HKMD, i.e., = 31.5). 3, where 1 , 2 , and 3 are the major, intermediate, and mi-
Xins surface is closer to the Mohr-Coulomb surface than to nor principal effective stresses, respectively), there is a unique
the Gudehus-Argyris surface. Furthermore, the Gudehus- relationship between vm , p, q, , and te. The contribution of
Argyris surface would be concave when > 22.7 (see p, q, and to the volume strain vm is reflected by pm in
Chen 1982), but Xins surface is always convex. eq. [20]. Similar to the modified Cam-Clay model, in this
model the viscoplastic volume strain is taken as the hardening
Constitutive equation of the present 3D parameter. Thus, it can be assumed that the viscoplastic vol-
EVP model ume strain rate & vp v in a general stress state (p, q, ) is equal
to the viscoplastic volume strain rate vp vm under the corre-
To derive the constitutive relation, the scaling function S sponding isotropic stress condition (p = pm , q = 0), that is,
in eq. [3] must be determined. This section presents the ap- & vp & vp
v = vm , if (p, q) and (pm , 0) are on the same loading sur-
proach used to determine S and then formulate the general face as shown in Fig. 4. Thus, from eq. [11], we obtain
constitutive equation.
It has been recognized that the loading surface may not be 2

& vp o 1 + vm vm
r
an elliptic surface passing through the original point with [22] & vp
v = vm =
q = p = 0 in the pq plane (Desai and Zhang 1987; Kutter Vt 0 vml
vp

and Sathialingam 1992; Hibbitt, Karlsson and Sorensen, Inc.


1995). A new loading surface is proposed which consists of

two parts, f1 and f2, as shown in Fig. 4. As shown in Fig. 4, rvm vm V
exp

f1 is the part of the loading surface within the critical state

1 + vm vm o
r
line in the pq plane for normally or lightly over-
consolidated behaviour and is expressed by vp
vml
( p )2
2 p pm q2
[20] F = f1 = + 2 2 From eq. [3], the viscoplastic volume strain rate is expressed
2
(1 + ) 2
M g ( ) by & vp
v = S(F/p ). Considering eq. [22], the scaling func-
1 tion S can be determined by
+ )2 = 0
( pm
(1 + ) 2 2
o r
[23] S= 1 + vm vp vm
where M = 6 sin /(3 sin ) is the slope of failure line in Vt 0 vml
the pq plane; pm is the mean effective stress value at
which the loading surface in eq. [20] intercepts the mean ef-
fective stress p axis in the pq plane, as shown in Fig. 4;
and is a parameter controlling shear shape in the pq rvm vm V 1
exp

1 + vm vm o F
plane. After investigating the results of some undrained r
compression and extension triaxial tests, the authors have p
vp
found that the stress paths for compression and extension vml
2002 NRC Canada

I:\cgj\Cgj39\Cgj-01\T01-074.vp
Tuesday, February 12, 2002 9:31:19 AM
Color profile: Disabled
Composite Default screen

Yin et al. 167

where F/p takes an absolute value for consideration of Fig. 5. Comparison between the measured and predicted results
overconsolidated soil. This will not affect the behaviour of for constant strain rate shear tests on the HKMD under compres-
normally consolidated soil because F/p is always positive sion and extension tests: (a) axial strain a versus deviator stress
for normally consolidated soil. Substituting eq. [23] into q, normalized with respect to consolidation pressure 0 ; and
eq. [3] leads to (b) normalized effective stress paths.
2
o r
[24] & vp
ij = 1 + vm vp vm
Vt 0 vml


rvm vm V 1 F
exp

1 + vm vm o F ij
r

vml
vp p

Lastly, considering eq. [2], the elastic viscoplastic constitu-
tive relationship in a general stress state is written as
2
1 & p& r
[25] & ij = sij + ij + o 1 + vm vp vm
2G 3V p Vt 0 vml


rvm vm V 1 F
exp

1 + vm vm o F ij
r

vml
vp p

where G is the shear modulus.
The parameter M in eqs. [20] and [21] is the same as that
in the modified Cam-Clay model and is related to the effec-
tive friction angle by M = (6 sin )/(3 sin ). The effec-
tive friction angle can be obtained from either consolidated
undrained or drained triaxial compression tests. For the
HKMD, M was determined to be 1.265 using = 41.5 ob-
tained from the consolidatedundrained triaxial tests (Yin
and Zhu 1999). The proposed EVP model has the following
, t0 , vp
10 parameters: /V, /V, o /V, rvmo , pmo vml , M, , and
n. All 10 parameters can be determined by a multistage iso- effective stress on the reference timeline and the current
tropic consolidation test and undrained (or drained) triaxial (initial) mean effective stress (under isotropic stressing con-
tests (at least one in the normally consolidated range for can take the value of the mean
ditions). In a triaxial test, pmi
and one in the heavily overconsolidated range for n). The effective stress before shearing. If we assume that the soil,
values of these parameters (except for ) for the HKMD are which is under a normally consolidated state, is in a state on
listed in Table 2. the reference timeline (for example, point A in Fig. 2), the
Before using eq. [25], F/p , F/ ij , and pm included in volume strain of the soil can then be given by vmi = rvmo +
F should be determined. They are presented in Appendix A. /pmo
( /V)ln(pmi ). For an overconsolidated soil, if the OCR
and current (initial) mean effective stress pmi are known,
Verification of the proposed 3D EVP model then the initial volume strain can be estimated by
using data from the HKMD
If we use the EVP model to simulate the stressstrain be- pmc
pmc
haviour of a soil, for example, a consolidatedundrained [26] vmi = rvmo + ln ln
V pmi
V pmo
triaxial test on the HKMD during the undrained shearing,
the volume strain at the beginning of the shearing would not
be zero. This nonzero value is called the initial volume strain
pmi
= rvmo+ ln + ln(OCR)
of the soil or specimen. V V
V pmo
When the EVP model is used to simulate the stressstrain
behaviour of a soil, it is important to properly estimate the
initial volume strain. According to Zhu (2000), the com- In this paper, eq. [26] is used to estimate the initial volume
monly used overconsolidation ratio can be defined as OCR = strain of soil. If rvm vm in eq. [25] is replaced by
/pmi
pmc , where pmc
and pmi are the preconsolidation mean vm = vm
r
vm , eq. [25] is then written as
2002 NRC Canada

I:\cgj\Cgj39\Cgj-01\T01-074.vp
Tuesday, February 12, 2002 9:31:19 AM
Color profile: Disabled
Composite Default screen

168 Can. Geotech. J. Vol. 39, 2002

2
1 & p& Undrained triaxial shear tests combined with stress
[27] & ij = sij + ij + o 1 + vpvm relaxation
2G 3V p Vt 0 vml A step-changed, constant strain rate, undrained triaxial
compression test combined with stress relaxation is simu-
lated in this paper to verify the capability of the model for
describing the stress relaxation behaviour. In this test, the ef-
vm V 1 F fective cell pressure on the specimen was 300 kPa with a
exp

1 + vm o F ij
back-pressure of 200 kPa. The test procedure is presented in
p Table 3.
vml
vp
Figure 6 compares the simulated results with test results for
the step-changed, constant strain rate test on the HKMD in
terms of deviator stress q versus axial strain a (Fig. 6a),
If the initial volume strain is estimated by eq. [26], the pa-
pore-water pressure u versus axial strain (Fig. 6b), and effec-
rameters of the EVP model would be /V, /V, o /V, t0,
tive stress paths (Fig. 6c). Figure 6 shows that the predictions
vp
vml , M, , , n, and OCR. are close to the measured values, although some discrepan-
In this section, the proposed 3D EVP model in eq. [27] is cies are observed. Noting that the test loading history was
used to simulate a number of triaxial tests on different soils, complicated, the predictions are considered to be good.
such as the HKMD and a mixture of kaolin and bentonite.
The predictions from the EVP model are compared with the
Undrained triaxial tests on overconsolidated HKMD
test results. The 3D EVP model was calibrated in the previ-
A number of consolidatedundrained (CIU) triaxial com-
ous section using the test data from an isotropic consolida-
pression and extension tests under different OCRs of 1, 2, 4,
tion test and consolidated undrained tests on the HKMD.
and 8 were performed on the HKMD. In this section, eight
The model parameters have been determined for the HKMD,
tests with axial strain rates of 1.5%/h are used to verify the
and the parameters presented in Table 2 are used directly for
ability of the model to predict the stressstrain behaviour of
the following simulations for the HKMD.
overconsolidated soils. The preshear initial isotropic consoli-
In this paper, the shear modulus used for the elastic defor- dation pressure pmi = 400 kPa for the tests with OCR = 1, but
mation of the HKMD is calculated from Ge = 1.5K e(1 2v)/ = 100 kPa for the tests with OCR = 2, 4, and 8. Thus for
pmi
(1 + v), where the bulk modulus K e = p/( /V), and is tests with OCR = 1, 2, 4, and 8, the preconsolidation pres-
Poissons ratio, which can be estimated to be 0.3. The pa- sures pc = c were 400, 200, 400, and 800 kPa, respectively.
rameter is given by = [6 + sin(3 /2)]/5, where is Figure 7a compares the measured and predicted relation-
the stress load angle. The slope of the failure line (also ships of deviator stress versus axial strain for the HKMD
called the critical state (CS) line) in the q p plane is ex- with different OCRs. Figure 7b presents the effective stress
pressed as M = 6 sin (3 sin ), and n is a constant with paths normalized with respect to the preconsolidation pres-
a value of 6. sure c . Both the stressstrain relationship and the effective
When simulating a triaxial test, the triaxial specimen is as- stress paths show good agreement between the measured and
sumed to be an element of soil. Equation [27] can be integrated predicted results overall.
numerically using the fourth-order Runge-Kutta method.
Undrained triaxial creep tests
Undrained triaxial tests sheared at various strain rates Three consolidatedundrained triaxial compression creep
To examine the capability of the proposed EVP model in tests on remoulded HKMD were conducted by Zhu (2000).
reflecting the behaviour of strain-rate effects of soils, six For the three specimens of HKMD, the consolidation pres-
consolidatedundrained triaxial tests on the HKMD were sure of each specimen was 400 kPa, and the instantly ap-
simulated. Each specimen was first isotropically and nor- plied deviator stresses were 134, 189, and 243 kPa. Before
mally consolidated for 36 h in a compression test and 48 h the application of the deviator stresses, the specimens were
in an extension test under an all-round consolidation pres- normally and isotropically consolidated.
sure 0 = 400 kPa. To ensure the saturation of the specimen, Figure 8 compares the measured and predicted results for
a back-pressure of 200 kPa was imposed on each specimen. the undrained triaxial creep tests on the HKMD. Figure 8a
After isotropic consolidation, the specimens were sheared at shows the variation of axial strain with time, and Fig. 8b
different axial strain rates of 0.15, 1.5, and 15%/h under illustrates the development of pore-water pressure with time
compression and extension states. The plus sign denotes a during the undrained creep. The agreement between the mea-
compression test, and the minus sign an extension test. sured and predicted results is better at lower deviator stresses.
The feature of accelerating creep is well simulated, although
Figure 5 illustrates the comparison between the measured
not matching the data points when approaching failure.
and predicted results for constant strain rate shear tests on
the HKMD under both compression and extension states,
that is, deviator stress q = 1 3, normalized with respect Verification using data from tests on an
to consolidation pressure 0, versus axial strain a (Fig. 5a), overconsolidated mixture of kaolin and
and normalized effective stress paths (Fig. 5b). In general, bentonite
there was good agreement between the model predictions
and the measured results for both the compression tests and To further verify the proposed new 3D EVP model, four
the extension tests. Some discrepancies were noted for the consolidatedundrained (CIU) triaxial compression tests on
effective stress paths and stressstrain in extension. a mixture of kaolin and bentonite with an initial void ratio e0
2002 NRC Canada

I:\cgj\Cgj39\Cgj-01\T01-074.vp
Tuesday, February 12, 2002 9:31:20 AM
Color profile: Disabled
Composite Default screen

Yin et al. 169

Table 3. Step-changed constant strain rate compression test combined with relaxation.
Stage: 1 2 3 4 5 6 7 8
Test: Shear Unload Reload Relax Shear Relax Shear Relax
& a (%/h) 6.00 6.00 6.00 0 0.60 0 0.06 0
t (min) 29 7 20 2540 232 1320 830 705

Fig. 6. Comparison between the measured and predicted results Fig. 7. Comparison between measured and predicted results of
for step-changed constant rate undrained triaxial tests combined consolidatedundrained (CIU) tests for overconsolidated Hong
with stress relaxation on Hong Kong marine deposits: (a) axial Kong marine deposits: (a) axial strain a versus deviator stress
strain a versus deviator stress q, (b) axial strain a versus pore- q, and (b) effective stress paths normalized with respect to
water pressure u, and (c) effective stress paths. preconsolidation pressure c . OCR, overconsolidation ratio.

each specimen was 392 kPa, and the axial strain rate used
for simulation during shear was 0.6%/h.
According to Yin and Graham (1999), the EVP model
parameters can be estimated from conventional parameters.
Kutter and Sathialingam (1992) and Dafalias and Herrmann
(1986) presented values of conventional parameters as listed
in Table 4. Using these parameters, EVP model parameters
are estimated and listed in Table 4.
Figure 9 presents both the test results (after Kutter and
Sathialingam 1992; Dafalias and Herrmann 1986) and the
simulated results for the CIU tests on the mixture of kaolin
are simulated. The tests were carried out by Herrmann et al. and bentonite. The relationship of pore-water pressure ver-
(1981). The specimens tested were overconsolidated and had sus axial strain shows good agreement between the mea-
OCRs of 1, 1.3, 2, and 6. The preconsolidation pressure of sured and predicted results, as shown in Fig. 9b. For the
2002 NRC Canada

I:\cgj\Cgj39\Cgj-01\T01-074.vp
Tuesday, February 12, 2002 9:31:21 AM
Color profile: Disabled
Composite Default screen

170 Can. Geotech. J. Vol. 39, 2002

Fig. 8. Comparison between measured and predicted results of Fig. 9. Comparison between measured and predicted results of
undrained creep tests for Hong Kong marine deposits: (a) axial CIU tests on a mixture of kaolin and bentonite: (a) axial strain
strain a versus time, and (b) pore-water pressure u versus time. a versus deviator stress q, (b) axial strain versus pore-water
pressure u, and (c) effective stress paths (data from Dafalias and
Herrmann 1986).

Table 4. Parameters of the 3D EVP model for a mixture of ka-


olin and bentonite.
EVP model parameters Related conventional parameters
/V = 0.012 = 0.018
/V = 0.093 = 0.151
o /V = 0.0037 C = 0.0139
t0 = 1440 min
vp
vml = 0.383
= 31.1 = 31.1
n=6
= 0.3
Note: Values of the conventional parameters are taken from Kutter and
Sathialingam (1992) and Dafalias and Herrmann (1986); V = 1 + e0.
iour of both normally consolidated and overconsolidated soils.
This model is established for a general stress state and is
relationship between deviator stress versus axial strain shown more general than the original EVP model of Yin (1990) and
in Fig. 9a and effective stress paths in Fig. 9c, some discrep- Yin and Graham (1999).
ancy is observed. The proposed EVP model has a total of 10 parameters:
, t0, vp
/V, /V, o /V, rvmo, pmo vml , M, n, and . A method
Summary and conclusions for determining the parameters has been proposed and dem-
onstrated using the data from tests on the HKMD. All 10 pa-
A new 3D elastic viscoplastic model is presented in this rameters can be determined by a multistage isotropic
paper by incorporating a new loading surface and a nonlin- consolidation test and undrained (or drained) triaxial tests.
ear function for the creep (delayed) strain. The new EVP All these parameters will have a certain influence on the
model can consider the time-dependent stressstrain behav- time-dependent stressstrain behaviour. Yin (1990) and Yin

2002 NRC Canada

I:\cgj\Cgj39\Cgj-01\T01-074.vp
Tuesday, February 12, 2002 9:31:21 AM
Color profile: Disabled
Composite Default screen

Yin et al. 171

and Graham (1994) did a sensitivity analysis on the creep Kaliakin, V.N., and Dafalias, Y.F. 1990. Theoretical aspects of the
parameter /V and found that it has a strong influence on elastoplasticviscoplastic bounding surface model for cohesive
the strain-rate effects, relaxation, etc. In particular, when the soils. Soils and Foundations, 30(3): 1124.
creep parameter /V is zero, the 1D EVP model (Yin and Kutter, B.L., and Sathialingam, N. 1992. Elasticviscoplastic mod-
Graham 1994) becomes a 1D elasticplastic model and the elling of the rate-dependent behaviour of clays. Gotechnique,
3D EVP model (Yin and Graham 1999) becomes a 3D elas- 42(3): 427441.
ticplastic model (i.e., the modified Cam-Clay model). Leroueil, S., Kabbaj, M., Tavenas, F., and Bouchard, R. 1985.
The proposed EVP model is used to predict a number of Stress strain strain rate relation for the compressibility of
triaxial tests on two different soils, namely Hong Kong ma- sensitive natural clays. Gotechnique, 35(2): 159180.
rine deposits (clay) and a mixture of kaolin and bentonite. Nash, D.F.T., and Ryde, S.J. 2000. Modelling the effects of sur-
The tests have a wide range of types, including constant charge to reduce long term settlement of reclamations over soft
strain rate, CIU tests in compression and extension, relax- clays. In Proceedings of the Soft Soil Engineering Conference,
ation tests, and undrained constant-q creep tests. The CIU Japan, 2000.
tests include tests on specimens with different OCRs. The Perzyna, P. 1963. The constitutive equations for working hardening
and rate sensitive plastic materials. Proceedings of Vibration
proposed model has been used to simulate all these tests and
Problems, 4(3): 281290.
the simulated results are compared with the measured re-
Perzyna, P. 1966. Fundamental problems in viscoplasticity. Ad-
sults. The comparisons show that the model can properly de-
vances in Applied Mechanics, 9: 244368.
scribe the time-dependent stressstrain behaviour of soils in
Roscoe, K.H., and Burland, J.B. 1968. On the generalised stress
triaxial states under both normally and overconsolidated
strain behaviour of wet clay. In Engineering plasticity. Edited by
states. The model predictive capability under general loading
J. Heyman and F.P. Leckie. Cambridge University Press, New
conditions needs further validation. York, pp. 535609.
Sun, J. 1999. Rheology of geomaterials and applications. China
Acknowledgements Construction Publication House, Beijing, China.
Vermeer, P.A., and Neher, H.P. 2000. A soft soil model that ac-
Financial support from a RGC grant (PolyU 5065/97E) of counts for creep. In Proceedings of Beyond 2000 in Computa-
the University Grants Committee of the Hong Kong SAR tional Geotechnics 10 Years of PLAXIS International. Edited
Government of China and the Hong Kong Polytechnic Uni- by R.B.J. Brinkgreve. A.A. Balkema, Rotterdam, pp. 249261.
versity is acknowledged. Xin, Y.C. 1988. Experimental study on the elasto-plastic model of
loess. In Proceedings of the 3rd National Conference on Numer-
ical Analysis Methods in Geomechanics, Zuhai, China, Vol. 1,
References p. 193. (In Chinese.)
Yang, C.H., Yin, J.-H., and Daemen, J. 1999. Study on stress relax-
Adachi, T., and Oka, F. 1982. Constitutive equations for normally
ation of salt rock. Chinese Journal of Rock Mechanics and Engi-
consolidated clay based on elasto-viscoplasticity. Soils and Foun-
neering, Chinese Society of Rock Mechanics and Engineering,
dations, 22(4): 5770.
18(3): 262265. (In Chinese.)
Bjerrum, L. 1967. Engineering geology of Norwegian normally
Yin, J.-H. 1990. Constitutive modelling of time-dependent stress
consolidated marine clays as related to the settlements of build-
strain behaviour of soils. Ph.D. Thesis, The University of Mani-
ings. Gotechnique, 17(2): 83118.
toba, Winnipeg.
Borja, R.I., and Kavazanjian, E. 1985. A constitutive model for the
stressstraintime behaviour of wet clays. Gotechnique, 35(3): Yin, J.-H. 1999. Equivalent time and elastic visco-plastic modeling
283298. of geomaterials. Chinese Journal of Rock Mechanics and Engi-
Burland, J. 1967. Deformation of soft-clay. Ph.D. thesis, Cam- neering, Chinese Society of Rock Mechanics and Engineering,
bridge University, Cambridge, U.K. 18(2): 124128. (In Chinese.)
Chen, W.-F. 1982. Plasticity in reinforced concrete. McGraw-Hill Yin, J.-H., and Graham, J. 1989. Viscouselasticplastic modelling
Book Company, New York. of one-dimensional time-dependent behaviour of clays. Cana-
Dafalias, Y.F., and Herrmann, L.R. 1986. Bounding surface plastic- dian Geotechnical Journal, 26: 199209.
ity. II: application to isotropic cohesive soils. Journal of the En- Yin, J.-H., and Graham, J. 1994. Equivalent times and one-
gineering Mechanics Division, ASCE, 112(12): 12631291. dimensional elastic visco-plastic modelling of time-dependent
Desai, C.S., and Zhang, D. 1987. Viscoplastic model for geologic ma- stressstrain behaviour of clays. Canadian Geotechnical Journal,
terials with generalised flow rule. International Journal for Numeri- 31: 4252.
cal and Analytical Methods in Geomechanics, 11(6): 603620. Yin, J.-H., and Graham, J. 1999. Elastic viscoplastic modelling of
Graham, J., Crooks, J.H.A., and Bell, A.L. 1983. Time effects on time-dependent stressstrain behavior of soils. Canadian Geo-
stressstrain behaviour of natural soft clays. Gotechnique, 33(3): technical Journal, 36: 736745.
327340. Yin, J.-H., and Zhu, J.-G. 1999. Measured and predicted time-
Herrmann, L.R., Shen, C.K., Jafroudi, S., DeNatale, J.S., and dependent stressstrain behaviour of Hong Kong marine depos-
Dafalias, Y.F. 1981. A verification study for the bounding surface its. Canadian Geotechnical Journal, 36: 760766.
plasticity model for cohesive soils. Final Report to the Civil Engi- Zhu, J.G. 2000. Experimental study and elastic visco-plastic mod-
neering Laboratory, Naval Construction Battalion Center, Port elling of the time-dependent stressstrain behaviour of Hong
Hueneme, Calif. Kong marine deposits. Ph.D. thesis, The Hong Kong Polytech-
Hibbitt, Karlsson and Sorensen, Inc. 1995. ABAQUS program and nic University, Hong Kong.
manuals, version 5.5. Hibbitt, Karlsson and Sorensen, Inc., Zhu, G., and Yin, J.-H. 2000. Elastic visco-plastic finite element
Pawtucket, R.I. consolidation modelling of Berthierville test embankment. Inter-

2002 NRC Canada

I:\cgj\Cgj39\Cgj-01\T01-074.vp
Tuesday, February 12, 2002 9:31:22 AM
Color profile: Disabled
Composite Default screen

172 Can. Geotech. J. Vol. 39, 2002

national Journal for Numerical and Analytical Methods in Geo- vmi initial volume strain
mechanics, 24: 491508. evm volume strain on the instant timeline
Zhu, J.G., Yin, J.-H., and Luk, S.T. 1999. Time-dependent stress evmo volume strain at pm = pu
strain behavior of soft Hong Kong marine deposits. Geotechnical vmr
reference timeline
Testing Journal, ASTM, 22(2): 118126. vmo
r
EVP model parameter
Zienkiewicz, O.C., and Cormeau, I.C. 1974. Visco-plasticity, plastic- vp
vm viscoplastic creep strain
ity and creep in elastic solids: a unified numerical solution vp
vml EVP model parameter
approach. International Journal for Numerical and Analytical a function
Methods in Geomechanics, 8(2): 821845. effective friction angle
fluidity parameter
slope of swelling compression line in vm ln pm
List of symbols plane
Cijkl fourth-order compliance tensor plane
slope of virgin compression line in vm ln pm
e, e0 void ratio, initial void ratio Poissons ratio
f1, f2 function of loading surface c preconsolidation pressure
F function of loading surface ij stress tensor
g() shape function for shaping the loading surface in ij effective stress tensor
plane 0 consolidation pressure
G shear modulus 1 , 2 , major, intermediate, and minor principal effective
Ge elastic shear modulus 3 stresses, respectively
Gs specific gravity stress load angle
Ip plastic index
J2, J3 the second and third invariants of deviator stress ten- Appendix A
sors
K a parameter Determination of F/ p
Ke elastic bulk modulus The term F/p in eq. [25] can be determined from
m subscript for expressing the isotropic stress state eq. [20] or eq. [21] as
M slope of failure line in qp plane
F 2p
2 pm
n parameter [A1] = 2
p mean effective stress p (1 + ) 2
pc preconsolidation mean effective stress
pm mean effective stress under isotropic stressing when q/p < Mg(), or

pmc preconsolidation mean effective stress on the reference
F q
timeline [A2] = 1

pmi current (initial) mean effective stress under isotropic p Mg ( ) p
stressing

pmo model parameter when q/p > Mg().
po initial mean effective stress
pu reference mean effective stress Determination of pm
q generalized deviator stress is given by the following equations. When
In eq. [25], pm
Q viscoplastic potential function can be given directly from eq. [20] by
q/p < Mg(), pm
sij deviator stress tensor
= p+
q2 q
S scaling function for viscoplastic strain rates [A3] pm = 1 and < Mg ( )
t time M g (2) p
2 p
te equivalent time
t0 curve-fitting parameter related to the choice of refer- and
ence timeline
t test duration
=
p q2
[A4] pm ( p ) 2 (1 2 ) 2 2
u pore-water pressure 1 1 M g ( )
V specific volume, volume occupied by unit volume of
solids q
1 and < Mg ( )
wl liquid limit p
wp plastic limit
EVP model parameter For the case of q/p > Mg() for overconsolidated soil pm can
o creep parameter be derived as follows using eqs. [20] and [21].
ij Knonecker delta According to the critical state theory, the critical state line
a axial strain should pass through the point (for example, point D in
, v strain, volume strain Fig. 4a) on a loading surface in the p q plane. On this
ij strain tensor point, there should be F/p = 0 from the loading surface.
vc volume strain corresponding to pc Point D should lie on f1 and f2. Thus, point D should be the
vm volume strain corresponding to mean effective stress intersection of the critical state line with the loading surface

2002 NRC Canada

I:\cgj\Cgj39\Cgj-01\T01-074.vp
Tuesday, February 12, 2002 9:31:22 AM
Color profile: Disabled
Composite Default screen

Yin et al. 173

of f2, and it can be expressed as p = [(n 1)/n]po , and q = F 2q


[A9] = 2 2
[(n 1)/n]Mg() po . On the other hand, point D should also be q M g ( )
on the surface of f1. Substituting the p and q of point D into
eq. [20] gives F 2 ( p )2
[A10] =
n 1 2 n 1 3
[A5] =
pm po po
(1 )n 1 n
3 + 2 2 (1 + 2 ) 2
+ 2 p pm ( pm )
(1 + )
2 3
(1 + ) 2 3
where po can be determined from eq. [21]. Substituting po
in the case of q/p > Mg() for both = 1
into eq. [A5], pm 3
[A11] = cos 3
and 1 is given by 5 2

n F 2q 2
= 2 3
1 q
[A12]
=
(1 + )(n 1) g ( )
[A6] pm p 1 + M g ( )
n n 1 Mg ( ) p
For eq. [21], F/q, F/, and F/g() are written as follows:
q = 1 F n q n1
> Mg ( ) for [A13] =
p or 1 q n 1 M g (n) ( p ) n1
n

F
In this study, n = 6 is used, and then eq. [A6] can be written [A14] =0
as
F n qn

6 [A15] = n +1
(1 + )5 1 q g ( ) n 1 M g ( ) ( p ) n1
n
[A7] =
pm p 1 +
6 5 Mg ( ) p
From eq. [18],
q = 1 g ( )
> Mg ( ) for [A16]
p or 1

K(1 K) cos 3 sin
Determination of F/ ij 6 6
Using eqs. [20] and [21], F/ ij in eq. [25] can be de- = 2
rived. In eqs. [20] and [21], F is a function of p, q, and ,
(2K 1) + (1 K) sin + 3 cos
thus we have 6 6

F F p F q F p 1
[A8] = + + [A17] = ij
ij p ij q ij ij ij 3

F g ( ) q 3
+ [A18] = ( ij p ij )

ij 2q
g ( ) ij
3 3J3 2
where F/p has been determined in eq. [25]. For eq. [20], [A19] = sij sik skj + J2 ij
F/q, F/, and F/g() are expressed as follows: ij 4 J 32 27J 23 2 J2 3

2002 NRC Canada

I:\cgj\Cgj39\Cgj-01\T01-074.vp
Tuesday, February 12, 2002 9:31:22 AM

You might also like