Modeling Time-Dependent Behavior of Soft Sensitive Clay

Zhen-Yu Yin1; Minna Karstunen2; Ching S. Chang, M.ASCE3; Mirva Koskinen4; and Matti Lojander5

Abstract: The paper focuses on investigating the destructuration process during time-dependent stress-strain evolution. For this purpose,
various oedometer tests and triaxial tests on intact and reconstituted samples of soft sensitive Vanttila clay were carried out. Based on
experimental observations, a new elastic viscoplastic model, extended from the overstress theory of Perzyna, is developed. The proposed
model accounts for inherent and induced anisotropy, interparticle bonds and bond degradation, and viscosity. The determination of model
parameters is discussed, demonstrating how all model parameters can be determined in a straightforward way and no additional test is needed
for the proposed model compared to the modified Cam clay model. The model is implemented into a finite-element code, which enables
coupled consolidation analyses. The model is used to simulate various strain-rate and creep tests under one-dimensional and triaxial con-
ditions on the intact samples of Vanttila clay. The comparisons between experimental results and simulations show that the model has good
predictive ability on the time-dependent behavior of a soft sensitive clay. DOI: 10.1061/(ASCE)GT.1943-5606.0000527. © 2011 American
Society of Civil Engineers.
CE Database subject headings: Anisotropy; Creep; Clays; Time dependence; Viscoplasticity.
Author keywords: Anisotropy; Creep; Destructuration; Clay; Time dependence; Viscoplasticity.

Introduction (i.e., Leroueil and Vaughan 1990; Burland 1990); and (3) time-
dependent stress-strain behavior that has a significant influence
The extremely soft and sensitive clay deposits formed after the on the shear strength and the preconsolidation pressure (i.e., Suklje
Pleistocene period cover most of the densely populated low-lying 1957; Graham et al. 1983; Leroueil et al. 1985; Watabe et al. 2008).
coastal areas in Finland. Similar deposits can also be found in other Since all these features cannot be neglected, a new constitutive
parts of Europe, North America, and Asia. The construction on soft model (AniCreep) is proposed in this paper, which differs from the
soil deposits has become increasingly important in past decades, existing models for clay:
as more and more development occurs on areas that were consid- • The phenomenon of rate-dependent preconsolidation pressure is
ered unsuitable for construction just a few decades ago. The current incorporated into the proposed model. This is different from the
design methods are very crude, in most cases overconservative, but rate-independent elastoplastic models (e.g., Roscoe and Burland
in some cases unsafe. Therefore, safe and economical design and 1968; Asaoka et al. 2002; Wheeler et al. 2003; Karstunen et al.
construction on soft soils has become an important issue. 2005). Because of this hypothesis, the proposed model is able to
The behavior of soft sensitive clays is very complicated. They predict the behavior of natural soft clay in a more realistic man-
exhibit several features: (1) a significant degree of anisotropy ner than the rate-independent models.
developed during their deposition, sedimentation, consolidation • The proposed model is extended from overstress theory of
history, and any subsequent straining (see, i.e., Burland 1990; Perzyna (1963, 1966), by assuming that there is no purely
Wheeler et al. 2003; Karstunen and Koskinen 2008); (2) some elastic region. In the following, this is referred to as an extended
apparent bonding that will be progressively lost during straining overstress approach. This is different from the conventional
overstress models (Rocchi et al. 2003; Hinchberger and Rowe
1
Professor, Center for Marine Geotechnics Research, Dept. of Civil 2005; Yin and Hicher 2008; Karstunen and Yin 2010). This
Engineering, Shanghai Jiao Tong Univ., Shanghai 200240, PR China hypothesis eases the effort in parameter determination.
(corresponding author). E-mail: [email protected] • The proposed model considers initial anisotropy and its evolu-
2
Reader, Dept. of Civil Engineering, Univ. of Strathclyde, John tion attributable to irrecoverable straining by adopting an in-
Anderson Building, 107 Rottenrow, Glasgow G4 0NG, UK; and Aalto clined initial reference surface that rotates with viscoplastic
Univ., School of Science and Technology, Finland. E-mail: minna strains. It is different from the isotropic extended overstress
[email protected]
3 models (e.g., Kutter and Sathialingam 1992; Vermeer and Neher
Professor, Dept. of Civil and Environmental Engineering, Univ. of
Massachusetts, Amherst, MA 01002. E-mail: [email protected] 1999; Yin et al. 2002; Hinchberger and Qu 2009) and the mod-
4
Project Manager, City of Helsinki, Real Estate Department, Geotech- els that consider only initial anisotropy, neglecting the evolution
nical Division, P.O. Box 2202, 00099 City of Helsinki, Finland. E-mail: of anisotropy (e.g., Adachi and Oka 1982; Kimoto and Oka
[email protected] 2005). The consideration of anisotropy and its evolution allows
5
Dept. of Civil and Environmental Engineering, Helsinki University of the applicability of the proposed model to soils that exhibit gen-
Technology, P.O. Box 2100, 02015 TKK, Finland. E-mail: matti eral inherent and induced anisotropy.
[email protected] • The proposed model considers the effects of destructuration
Note. This manuscript was submitted on March 6, 2010; approved on
combined with anisotropy and viscosity, providing a straightfor-
February 24, 2011; published online on March 2, 2011. Discussion period
open until April 1, 2012; separate discussions must be submitted for indi- ward way to determine all parameters. It is different from gen-
vidual papers. This paper is part of the Journal of Geotechnical and eral anisotropic models that do not consider destructuration
Geoenvironmental Engineering, Vol. 137, No. 11, November 1, 2011. (e.g., Oka et al. 2004; Leoni et al. 2008) and the models that
©ASCE, ISSN 1090-0241/2011/11-1103–1113/$25.00. consider destructuration but require curve fitting to determine

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2011 / 1103

Downloaded 23 Dec 2011 to 203.197.118.86. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
 parameters (e.g., Kimoto and Oka 2005; Hinchberger and Qu Destructuration Effect on One-Dimensional
2009). Because of the consideration of destructuration, the Stress-Strain Behavior
proposed model is applicable to soft sensitive clays. The appli-
Fig. 1(a) shows the comparison of e
log σ0v curves (void ratio
cation of the model requires the same experimental information
versus vertical effective stress in log scale) for intact and reconsti-
as needed for the modified Cam clay model, which makes the
tuted samples of Vanttila clay. For intact samples, the postyield
proposed model attractive for geotechnical practice.
compression curves gradually converge to the compression curves
In a previous study by Karstunen and Koskinen (2008), drained
of reconstituted samples, as bonding is progressively destroyed
triaxial tests on reconstituted samples of Vanttila clay were con-
during inelastic straining, as shown by Leroueil and Vaughan
ducted to investigate the induced anisotropy and how this can be
(1990), Burland (1990), and others.
represented by a kinematic hardening law proposed by Wheeler
Fig. 1(b) shows a schematic plot for compression behavior
et al. (2003). This study focuses on investigating the destructuration
of a 1-day oedometer test for sensitive clays. In Fig. 1(b), the
process during time-dependent stress-strain evolution. For this pur-
initial amount of bonding χ0 ¼ St
1 is defined relating to the
pose, oedometer tests have been carried out on intact and reconsti-
soil sensitivity from oedometer tests as St ¼ σ0p0 =σ0pi0 . The amount
tuted samples of Vanttila clay. In the following sections, the results
of bonding χ decreases from its initial value ultimately toward
of oedometer tests on Vanttila clay are first discussed. Then a new
zero when all bonds are destroyed during straining [Fig. 1(c)].
model is proposed using a strain-rate-based scaling function, which
For the postyield compression, the current stress can be described
incorporates the effects of anisotropy and destructuration. The de-
by
termination of model parameters is discussed, and then the model
is used to simulate oedometer and triaxial tests on Vanttila clay.
σ0v ¼ σ0pi ð1 þ χÞ ð1Þ

Destructuration Effect on One-Dimensional with

Compression Behavior

evp
To investigate the destructuration process during time-dependent σ0pi ¼ σ0pi0 exp
λi
κ
stress-strain evolution, oedometer tests have been carried out on ð2Þ
e_ vp
intact and reconstituted soil samples: conventional tests with or in an incremental form σ_ 0pi ¼ σ0pi
load-step durations of 1 day, 10 days, and 100 days and tests at λi
κ
constant strain rates (see Table 1). A relative homogenous layer
of Vanttila clay from 2 to 3.5 m was selected for sampling. The where λi = slope of the intrinsic normal compression curve in the
Vanttila clay is a highly sensitive clay with sensitivity St > 50. e
ln σ0v plane; κ = slope of the swelling line; evp = viscoplastic
The unit weight varies from 12.9 to 14:3 kN=m3 . The plastic limit void ratio corresponding to σ0v [see Fig. 1(b)].
varies from 30 to 34%. The clay deposit has a high natural water Fig. 1(c) shows that the degradation of the amount of bonding is
content (w ¼ 115–120%), which is greater than the liquid limit associated with the viscoplastic strain level for Vanttila clay. The
(wL ¼ 98–105%). The void ratio varies from 2.79 to 3.64. The amount of bonding is plotted as a function of the void ratio during
percentage of clay-size particles is 53%, and the organic content one-dimensional compression. The bond degradation can be fitted
is 1.9%. by Eq. (3)

Table 1. List of Oedometer Tests on Intact and Reconstituted Samples of Vanttila Clay
Test Sample condition Initial thickness (mm) Depth (m) Loading condition
Conventional oedometer tests
Oed4642R, 4655R, 4661R, 5351R Reconstituted 20 2.10–3.00 1 day
Oed4721 Intact 15 2.93–2.95 1 day
Oed4723 Intact 15 2.91–2.93 1 day
Oed4725 Intact 15 2.87–2.89 1 day
Oed4872 Intact 20 3.39–3.41 1 day
Oed4873 Intact 20 3.27–3.30 1 day
Oed4788 Intact 15 2.85–2.87 10 days
Oed4789 Intact 15 2.83–2.85 10 days
Oed4814 Intact 15 2.77–2.79 100 days
Oed4861 Intact 15 2.89–2.91 100 days
Oedometer tests at constant strain rate
CRS2021 Intact 15 3.07–3.09 1:11 × 10
6 s
1
CRS2022 Intact 15 3.09–3.11 1:67 × 10
6 s
1
CRS2023 Intact 15 2.91–2.93 2:78 × 10
6 s
1
CRS2024 Intact 15 2.93–2.95 5:56 × 10
6 s
1
CRS2025 Intact 15 3.05–3.07 1:11 × 10
5 s
1
CRS2026 Intact 15 3.07–3.09 3:89 × 10
6 s
1
CRS2027 Intact 15 3.24–3.26 2:22 × 10
6 s
1

1104 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2011

Downloaded 23 Dec 2011 to 203.197.118.86. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
 ln( ’v)
4 80
'pi0 = 0.37 kPa 'p0 = 29 kPa Vanttila clay
’pi0 ’p0
’pi0 Curve fitting
3.2 60

i 40 vp
2.4 evp = 0 exp[- e /(1+e0 )]
e

’pi ’v
’pi = 16
1.6 Vanttila clay Intact samples 20
Intact 1 day Reconstituted
Reconstituted 1 day samples
0.8 0
0.1 1 10 100 1000 e 3.5 2.5 1.5
(a) 'v (kPa) (b) (c) e

Fig. 1. One-dimensional compression behavior based on 1-day oedometer tests: (a) void ratio versus vertical effective stress for intact and
reconstituted samples of Vanttila clay, (b) schematic plot for one-dimensional compression behavior, and (c) evolution of the amount of bonding
with void ratio



ξevp χ_
ξ e_ vp for σ0p
ε_ vK0 at any strain level with a constant β for a given sen-
χ ¼ χ0 exp or in an incremental form ¼
1 þ e0 χ 1 þ e0 sitive clay.
ð3Þ
Proposed Constitutive Model
where ξ = material constant controlling the rate of destructuration;
e0 = initial void ratio. According to Perzyna’s overstress theory (Perzyna 1963, 1966), the
total strain rate is additively composed of the elastic strain rates and
Destructuration Effect on One-Dimensional viscoplastic strain rates
Stress-Strain-Strain Rate Behavior
Fig. 2(a) presents the initial preconsolidation pressure normalized ε_ ij ¼ ε_ eij þ ε_ vp
ij ð5Þ
by the in situ effective vertical stress (σ0p0 =σ0v0 ) varying with the
applied strain rate in double log scale for different clays. The where ε_ ij = (i; j) component of the total strain-rate tensor; the
strain-rate effect on the initial preconsolidation pressure for Vanttila superscripts e and vp = elastic and viscoplastic components, respec-
clay is similar to that reported by Leroueil et al. (1985) for soft tively. The elastic behavior in the proposed model is assumed to be
sensitive clays. According to this figure, the relationship between isotropic similar to the modified Cam clay model (Roscoe and
the initial preconsolidation pressure and the strain rate can be Burland 1968).
expressed as follows: Test results by Leroueil et al. (1985) in Fig. 2(a) demonstrate
that the yield stress is highly rate-dependent and decreases with

0 β the strain rate. Since there are no results available at volumetric
ε_ vK0 σp0
¼ ð4Þ strain rate ε_ vK0 < 5 × 10
9 s
1 (a test at ε_ vK0 ¼ 1 × 10
9 s
1 up
ε_ rvK0 σ0rp0 to εvK0 ¼ 10% would need 1,157 days), it is difficult to explore
whether the elastic region exists. Therefore, for conventional over-
where the preconsolidation pressures σ0p0 and σ0rp0 correspond to stress models, the size of static yield surface (elastic region) needs
strain rates ε_ vK0 and ε_ rvK0 , respectively; σ0rp0 and ε_ rvK0 = selected to be assumed, and as a direct consequence the viscosity parameters
initial reference preconsolidation pressure and the corresponding are highly dependent on the assumed size. If the assumption of no
reference strain rate, respectively; subscript “K0” refers to the one- elastic region is made, a reference yield surface corresponding to
dimensional condition; β = material constant representing the slope a reference test (e.g., conventional oedometer test; see “Determina-
[β varying from 14 to 35 in Fig. 2(a), for Vanttila clay β ¼ 22]. tion of Viscosity Parameters”) can be easily adopted together with
Comparing to conventional oedometer tests with different load-step unique values for the viscosity parameters.
durations Δt [see Fig. 2(b)], the strain-rate slope β ¼ 22 obtained Therefore, given the strain-rate effect on the preconsolidation
by constant rate of strain (CRS) tests is the same as that by conven- pressure (see Fig. 2), it is possible to extend the model to general
tional tests. It is noted that initial preconsolidation pressure σ0p0 is stress space by adopting an elliptic-shaped surface. Perzyna’s over-
always positive even when the strain rate approaches zero. stress formulation (Perzyna 1963, 1966) for the viscoplastic strain
By normalizing the vertical stresses with the initial preconsoli- rate ε_ vp
ij is then extended to
dation pressure of each test, the normalized stress-strain curves are
d β
plotted in the εvK0
logðσ0v =σ0p0 Þ plane for all conventional and pm ∂f d
ε_ vp ¼ μ ð6Þ
constant strain-rate oedometer tests for Vanttila clay [see Fig. 2(c)]. ij
prm ∂σ0ij
The normalized stress-strain curves can be considered unique.
Therefore, by assuming that the initial preconsolidation pressure where μ = fluidity parameter; β = strain-rate coefficient; pdm = size
is strain-rate dependent, a unique stress-strain–strain rate relation- of dynamic loading surface; prm = size of reference yield surface.
ship is obtained for Vanttila clay, as proposed by Leroueil et al. The dynamic loading surface f d is treated as a viscoplastic potential
(1985) for Canadian sensitive clays. This uniqueness of stress- function. The scaling function is similar to those derived from the
strain–strain rate behavior implies that Eq. (4) can be applied exponential form of Perzyna’s formulation by Adachi and Oka

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2011 / 1105

Downloaded 23 Dec 2011 to 203.197.118.86. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
 0.4 Critical state line M=1.35
Batiscan 20
= 22 Corresponding
St Cesaire
0.3
Log( 'p0 / 'v0) Bäckebol to in-situ 'v0
Berthierville
0.2 10
Vanttila clay 1

q (kPa)
Vanttila clay
0.1
1 Experiment
0
0
0 10 20 30 40

-0.1
1E-9 1E-8 1E-7 1E-6 1E-5 1E-4 -10
-1 p' (kPa)
(a) d v /dt (s )
Fig. 3. Yield points and yield curves for Vanttila clay
100
Intact 1 day
Intact 10 days
q
Intact 100 days 0.0457
y = 29.035x Mc
2 Dynamic loading surface 1 fd
R = 0.9996
'p0

B ij

= 22 A Knc-line

Reference surface
1
Vanttila clay Intrinsic surface
10
0.001 0.01 0.1 1 10 pmi pmr pmd
-1 O p’
(b) 1/ t (day ) pmi
0

1
Vanttila clay Me

20 Intact 1 day
Intact 10 days Fig. 4. Definitions for the model in p0 -q space
(%)

Intact 100 days

CRS2021
v

CRS2022
40 CRS2023 (1986) and lately confirmed by Wheeler et al. (2003) and Karstunen
CRS2024
CRS2025 and Koskinen (2008) based on experimental investigations on vari-
CRS2026 ous natural soft clays. The yield surface equation can be expressed
CRS2027 in a general stress space as
60
0.1 1 10 100 3
ðσ0rd
p0r αd Þ : ðσ0rd
p0r αd Þ
(c) 'v / 'p0 fr ¼ 2 þ p0r
prm ¼ 0 ð7Þ
ðM 2
32 αd : αd Þp0r
Fig. 2. Strain-rate effect on the apparent preconsolidation pressure
from oedometer tests: (a) strain-rate effect on the initial preconsolida- where σ0d r and p0r = reference deviatoric stress tensor and reference
tion pressure from tests at constant strain rate, (b) initial precon- mean effective stress, respectively (see point A in Fig. 4 in p0
q
solidation pressure versus loading rate from conventional tests, and space); αd = deviatoric fabric tensor (see Wheeler et al. 2003). The
(c) unique stress-strain-strain rate behavior for oedometer tests on yield surface for the intact samples of Vanttila clay are shown in
intact samples of Vanttila clay Fig. 3. Corresponding to the current stress state (see point B in
Fig. 4), a dynamic loading surface f d is defined, which has an
elliptical shape identical to the reference yield surface, but a differ-
(1982) and Kimoto and Oka (2005). According to this equation, the ent size pdm . Fig. 4 also shows that the reference stress state “A” is
rate of viscoplastic volumetric strain always exists. the projected point of the current stress state “B” from the origin
To define the reference yield surface, drained triaxial compres- (point O).
sion and extension tests under different stress ratios η (η ¼ q=p0 To interpolate M between its values M c (for compression) and
M e (for extension), a modification of M by means of the Lode angle
where q = deviatoric stress and p0 = mean effective stress) on intact
θ is assumed (see Sheng et al. 2000)
samples of Vanttila clay (from a depth of 2:38–2:44 m with
e0 ¼ 3:15–3:32) were conducted. Yield points corresponding to  1=4
2c4
loading at different stress ratios η were determined from linear M ¼ Mc ð8Þ
1 þ c4 þ ð1
c4 Þ sin 3θ
stress-strain plots and are shown in Fig. 3. The value of M (slope
of the critical state line) for Vanttila clay, determined from un- where c ¼ ð3
sin ϕc Þ=ð3 þ sin ϕc Þ according to Mohr-Coulomb
drained triaxial tests, was used to define the shape of the yield yieldpffifficriterion
ffi (ϕc = friction angle);
π=6 ≤ θ ¼ ð1=3Þsin
1
3=2
curves in Fig. 3, which confirms the suitability of the assumed ð
3 3J 3 =2J 2 Þ ≤ π=6 with
J 2 ¼ ð1=2Þ
sij :
sij and
J 3 ¼ ð1=3Þ



shape for reference yield surface proposed initially by Dafalias
sij
sjk
ski ;
sij ¼ σd
p0 αd .

1106 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2011

Downloaded 23 Dec 2011 to 203.197.118.86. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
 Based on work by Wheeler et al. (2003) and Karstunen and The initial reference preconsolidation pressure σ0rp0 obtained
Koskinen (2008), the rotational hardening law that describes the from an oedometer test is used as an input to calculate the initial
development of anisotropy caused by viscoplastic strain, is size pm0 by the following equation [derived from Eq. (7)]:
expressed as follows:  


  ½3
3K nc
0
αK0 ð1 þ 2K 0 Þ
nc 2
ð1 þ 2K nc
0 Þ
3σd σd pm0 ¼ þ σ0rp0 ð14Þ
dαd ¼ ω vp

αd hdεv i þ ωd
αd dεdvp
ð9Þ 3ðM 2c
α2K0 Þð1 þ 2K nc
0 Þ 3
4p0 3p0
where K nc 0 = coefficient of earth pressure at rest for normally
where hdεvp vp vp vp
v i ¼ 0 for dεv ≤ 0 and hdεv i ¼ dεv for dεv > 0.
vp
consolidated states, which can be calculated from M c [M c ¼
The soil constants ω and ωd control the effect of viscoplastic strains 6 sin ϕc =ð3
sin ϕc Þ] by the simplified formula K nc
0 ≈ 1
sin ϕc
on the rotation of the ellipsoidal surface. of Jaky (1948) if no measurements of K nc 0 are available; αK0 =
The amount of bonding is described with a scalar state variable initial anisotropy of a natural undisturbed sample. This can be cal-
χ. Extended from one-dimensional behavior [see Eq. (1)], the size culated from M c by assuming that the initial anisotropy has been
of the reference yield surface prm can be expressed by the size of created by one-dimensional compression (see Wheeler et al. 2003),
intrinsic surface pmi and the amount of bonding χ as follows: which implies in the normally consolidated region that
prm ¼ ð1 þ χÞpmi ð10Þ 3M c
ηK0 ¼ ð15Þ
6
Mc
The scalar χ is similar to the R ðχ ¼ 1=R
1Þ by Asaoka et al.
(2002). The expansion of the intrinsic surface is expressed as

 M 2c
η2K0
1 þ e0 αK0 ¼ ηK0
ð16Þ
dpmi ¼ pmi dεvp
v ð11Þ 3
λi
κ
The value for the soil constant ωd can be determined from M c as
and the amount of bonding χ is changing owing to bond degrada- proposed by Wheeler et al. (2003); thus the ωd is not an indepen-
tion ultimately to zero, as proposed by Gens and Nova (1993) dent soil constant
dχ ¼
χξðjdεvp vp
v j þ ξ d dεd Þ ð12Þ 3ð4M 2c
4η2K0
3ηK0 Þ
ωd ¼ ð17Þ
8ðη2K0 þ 2ηK0
M 2c Þ
where the soil constants ξ and ξ d control the effect of viscoplastic
volumetric and deviatoric strains in destroying the bonds. Note that By assuming that the constant ω is the same for intact or reconsti-
Eq. (12) can be reduced to Eq. (3) when dεvp d ¼ 0. tuted clay, the simplified formulation based on the kaolin clay
proposed by Leoni et al. (2008) can be used to estimate a value
Model Implementation and Consolidation Coupled for ω
Analysis
1 þ e0 10M 2c
2αK0 ωd
The proposed model was implemented as a user-defined model in ω¼ ln ð18Þ
the 2D Version 9 of PLAXIS using the numerical solution proposed ðλi
κÞ M 2c
2αK0 ωd
by Katona (1984). The basic finite-element scheme for the pro- If results of an undrained triaxial test in extension are available, the
posed model is similar to that presented by Oka et al. (1986). value for the soil constant ω can be determined more accurately by
For a coupled consolidation analysis based on Biot’s theory, the curve fitting.
relationship of the load increment is given by applying the principle
of virtual work to the equilibrium equation as shown by Oka Parameters Relating to Destructuration
et al. (1986). The initial amount of bonding χ0 (state parameter) can be deter-
During consolidation coupled analyses for simulating conven- mined from oedometer tests, as shown in Fig. 1. The values for
tional oedometer tests, the permeability k varies with void ratio e as the other two destructuration constants ξ and ξ d could be deter-
suggested by Berry and Poskitt (1972) mined from two drained compression tests with different constant
stress ratio η by curve fitting. In this paper, a straightforward way
k ¼ k 0 10ðe
e0 Þ=ck ð13Þ
to determine the values for ξ and ξ d based on isotropic and one-
The initial value of permeability k 0 with constant ck can easily be dimensional compression tests is proposed as follows.
obtained from oedometer test. The incremental forms of Eq. (11) and Eq. (12) can be written in
full forms
Model Parameters 
 
1 þ e0 vp
pmi ¼ pmi0 exp εv ð19Þ
The proposed model involves a number of soil constants and state λi
κ
variables that can be divided into four main groups.

Parameters of Modified Cam Clay Model and Parameters χ ¼ χ0 exp½
ξðεvp vp

v þ ξ d εd Þ ð20Þ
Relating to the Anisotropy
By substituting Eqs. (19) and (20) into Eq. (10), the size of the yield
The modified Cam clay analogous parameters of the proposed
surface can be derived as
model include state parameters (e0 ; σ0rp0 ) and soil constants (λi , κ, 
 
υ0 , M c , where υ0 is Poisson’s ratio), and the values for these can be vp vp 1 þ e0 vp
determined in a standard way from triaxial and oedometer tests. pm ¼ pmi0 f1 þ χ0 exp½
ξðεv þ ξ d εd Þg exp εv ð21Þ
λi
κ
The value for λi can be measured from test results of intact samples
at a high level of vertical loading corresponding to samples with Because of the elliptic-shaped yield surface formulation adopted in
all bonds destroyed [see Fig. 1(a)]. Thus, no tests on reconstituted this model, for an arbitrary constant stress ratio η, the relationship
samples are needed. pm =pmi0 ¼ σ0v =σ0pi0 can be obtained. Eq. (21) is then written as

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2011 / 1107

Downloaded 23 Dec 2011 to 203.197.118.86. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
 
 
1 þ e0 vp values. If the isotropic consolidation stage of an undrained triaxial
σ0v ¼ σ0pi0 f1 þ χ0 exp½
ξðεvp vp
v þ ξ d εd Þg exp εv ð22Þ
λi
κ test is used, no additional tests are needed compared to the modified
Cam clay model.
From the equation of flow rule [Eq. (6)], the relationship dεvp vp
d =dεv
can be derived as Parameters Relating to Viscosity
The values for the parameters relating to viscosity (μ and β) can be
dεvp 2ðη
αÞ determined either from oedometer tests at a constant strain rate or
d
¼ 2 ð23Þ
dεvp
v M
η2 from a conventional oedometer test.
Viscosity Parameters Determined from Strain-Rate
For the case of isotropic compression, experimental evidence sug- Oedometer Tests. For reconstituted clay, from the definition of
gests that the initial anisotropy is erased by isotropic loading up to elastic and viscoplastic strains, the ratio between the elastic strain
a pressure that is two or three times larger than the preconsoli- rate and the viscoplastic strain rate under a one-dimensional con-
dation pressure (see Leoni et al. 2008). Based on this, the accumu- dition can be derived as
lated deviatoric viscoplastic strain is zero (εvp d ¼ 0) at the stress
σ0v ¼ 3σ0p0 . In this case, χ0 and σ0 0 ε_ evK0 κ
pi0 in the assumed e
log σ v vp ¼ ð26Þ
0
curve are defined using 3σp0 as the initial preconsolidation pressure ε_ vK0 λi
κ
[see the dashed line in Fig. 5(a)]; for a selected stress σ0v greater than
3σ0 p0, the value for ξ can be derived from Eq. (22) using χ0 and The total strain rate can then be written as
σ0
pi0 , expressed as
λi
ε_ vK0 ¼ ε_ vp ð27Þ
  λi
κ vK0

ð1 þ e0 Þ σ0v 1
ξ¼ ln 
 ð24Þ
Δevp χ0 expðΔe 0
λ
κ Þσpi0
χ0
vp
Based on the uniqueness of stress-strain–strain rate behavior
i
[see Fig. 2(c)], on the normalized lines for the strain rate ε_ vK0
where Δevp [the accumulated viscoplastic void ratio Δevp ¼ and reference strain rate ε_ rvK0 at a given strain level εvK0 , the points
ð1 þ e0 Þεvp 0  0
v ] corresponds to the selected stress σv ; χ0 , σpi0 , and
(σ0v =σ0p0 ; εvK0 ) and (σ0rp =σ0rp0 ; εvK0 ) are the same, since the normal-
ized lines are concurrent. Thus, σ0v =σ0p0 ¼ σ0rp =σ0rp0 is obtained.
Δe can be measured from the assumed curve in Fig. 5(a).
vp
Then, extending Eq. (4) and substituting Eq. (27) into this extended
For the case of one-dimensional compression, α ¼ αK0 . Thus,
equation, the viscoplastic volumetric strain rate can be written as
starting from the stress σ0v ¼ σ0p0 , the ratio εvp vp
v =εd ¼ 2ðη K0
αK0 Þ=
ðM c
ηK0 Þ is obtained using Eq. (23). On the compression curve
2 2

λi
κ σ0v β
[see Fig. 5(b)], a vertical stress σ0v (greater than σ0p0 ) is selected with _εvp
vK0 ¼ εvK0 _ r
ð28Þ
λi σ0rp
its corresponding Δevp before the end of the destructuration. For
this case, using Eq. (22) Adopting the elliptic-shaped reference surface, for the case of K 0
  consolidation, the relationship σ0v =σ0rp ¼ pdm =prm is obtained. There-
2ðηK0
αK0 Þ
ð1 þ e0 Þ σ0v 1
ξ þ ξ · ξd ¼ ln
fore, the ε_ vp
vK0 can be derived from Eq. (6) using the flow rule under
ðM 2c
η2K0 Þ Δevp χ0 expðλΔe
κÞσ0pi0 χ0
vp

i the condition of K 0 consolidation

ð25Þ
d β


vp pm ∂f d M 2c
η2K0 σ0v β
ε_ vK0 ¼ μ r ¼μ 2 ð29Þ
where Δevp corresponds to the selected stress σ0v ; χ0 , σ0pi0 , and Δevp pm ∂p0 K 0 M c
α2K0 σ0rp
can be measured from the curve in Fig. 5(b). Using the value of ξ
obtained by Eq. (24), the value for ξ d can then be obtained Regarding Eq. (28), the value for fluidity μ can be determined by
by Eq. (25). the reference strain rate ε_ rvK0
To minimize the error between the simulations and experiments,
λi
κ M 2c
α2K0
selecting three postyield points with different strain levels for each μ ¼ ε_ rvK0 ð30Þ
curve is proposed to calculate ξ and ξ d , then taking the average λi M 2c
η2K0

* ’pi0* ’pi0
’pi0* 0 ln( ’v) ’pi0 0 ’p0 ln( ’v)
3 ’p0
1

evp
i i
evp*
1 1

Intact sample Intact sample

Reconstituted sample Reconstituted sample

(a) e (b) e

Fig. 5. Schematic plot for definitions in e
lnðσ0 Þv space under (a) isotropic compression condition and (b) one-dimensional compression condition

1108 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2011

Downloaded 23 Dec 2011 to 203.197.118.86. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
The uniqueness of stress-strain–strain rate behavior [see Fig. 2(c)] stresses (see Fig. 6). Before that, the variation in the apparent value
suggests that the viscosity parameters are independent of the de- of C αe is attributable to the destructuration process, associated
structuration process, even when all bonds are destroyed and the with the variation of the apparent value of λ. Since the viscosity
soil tends to the reconstituted state [see Fig. 1(a)]. Thus Eq. (30) parameters are independent of the destructuration process, as sug-
can be used for soft sensitive clays. The reference strain rate is gested by the uniqueness of stress-strain–strain rate behavior [see
selected with its corresponding initial reference preconsolidation Fig. 2(c)], for convenience intrinsic values of λ and C αe measured
pressure σ0rp0 . The soil constant β can be measured from the from the oedometer test on the reconstituted sample (noted as λi
logð_εvK0 Þ
logðσ0p0 Þ curve [see Fig. 2(a)]. Therefore, the determi- and C αei ) are used for Eq. (31). By combining Eq. (31) with the
nation of μ and β is straightforward. linear equation obtained from constant strain-rate tests [Eq. (28)],
Viscosity Parameters Determined from Conventional it follows that
Oedometer Tests. Another attractive way is to determine viscosity
parameters from a conventional oedometer test, which is more λi C αei λi
κ
ε_ rvK0 ¼ and β¼ ð32Þ
common in practice than the test at the constant strain rate. Assum- ðλi
κÞ ð1 þ e0 Þτ C αei
ing that the conventional oedometer test is performed with a fixed
duration for each load increment, and the corresponding reference Therefore, the conventional oedometer test with a fixed load-step
preconsolidation pressure σ0rp0 is estimated from the test results, duration τ and the oedometer test at constant strain rate ε_ rvK0 can be
Kutter and Sathialingam (1992) and Vermeer and Neher (1999) equivalently conversed with each other. It is noted that Eq. (32)
suggested the following relationship under a one-dimensional con- implies a constant C αei =λi for reconstituted or insensitive clay.
dition for insensitive or reconstituted clays: Since the apparent values for λ and C αe for sensitive clay are in-

0 ðλ
κÞ=C fluenced by destructuration, it is not necessary to assume a constant
C αe σv αe
ratio of C αe =λ. Consequently, a variation of C αe =λ in experimental
ε_ vp
vK0 ¼ 0r ð31Þ
ð1 þ e0 Þτ σp results, as shown, e.g., by Mesri and Godlewski (1977) is still
possible.
where C αe ¼ Δe=Δ ln t = secondary compression coefficient; τ = In connection with the proposed model, the values for the vis-
reference time taken equal to the duration for each load increment cosity parameters can be obtained from a conventional oedometer
(Kutter and Sathialingam 1992; Leoni et al. 2008). test as follows:
For soft sensitive clays, experimental results show that the value
of C αe for an intact sample converges to the value for a reconsti- C αei ðM 2c
α2K0 Þ λi
κ
μ¼ and β¼ ð33Þ
tuted sample when all bonds are destroyed under high vertical τ ð1 þ e0 ÞðM 2c
η2K0 Þ C αei

0.03 The values for λi and C αei can also be measured from results of
Reconstituted 1 day
intact samples corresponding to a high level of vertical loading with
Intact 1 day all bonds destroyed [see Figs. 1(a) and 6].
Intact 10 days Overall, no additional tests are needed when using the proposed
0.02 Intact 100 days model, compared to the modified Cam clay model.
Vanttila clay
C

0.01 Experimental Validation

To evaluate the model’s predictive ability, various oedometer tests

and triaxial tests on Vanttila clay (Table 2) under different loading
0 conditions were simulated. Most attention is devoted to the coupled
10 100 1000 10000 effect of destructuration on the stress-strain–time behavior, rather
'v (kPa) than the anisotropy effect, since the importance of considering in-
herent and induced anisotropy on describing mechanical behavior
Fig. 6. Secondary compression coefficient versus applied stress from
of soft soil has been reported by Wheeler et al. (2003), Karstunen
conventional oedometer tests on intact and reconstituted samples of
and Koskinen (2008), and Leoni et al. (2008) using the same func-
Vanttila clay
tion for yield surface and rotational hardening law.

Table 2. List of Triaxial Tests on Intact Samples of Vanttila Clay

Test Depth (m) w (%) γ (kN=m3 ) ei σ01 (kPa) σ03 (kPa) Loading condition
Undrained triaxial tests at constant strain rate (axial strain rate)
CAUC1 2.77–2.88 118 13.9 3.294 47 35 1%=h
CAUC2 3.14–3.24 118.4 13.98 3.197 78 57 1 to 10%=h
CAUC3 2.75–2.85 118.4 13.78 3.256 33 23 1%=h
Undrained triaxial creep tests (applied stresses)
CAUCR1 3.16–3.27 124.2 13.61 3.442 26.9 16.5 σ01 ¼ 31 kPa, σ01 ¼ 16:5 kPa;
CAUCR2 3.31–3.41 111.7 13.77 3.145 26.9 16.5 σ01 ¼ 31:5 kPa, σ03 ¼ 14:2 kPa
CAUCR3 3.05–3.15 116.2 13.54 3.308 26.9 16.5 σ01 ¼ 36:5 kPa, σ03 ¼ 16:5 kPa

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2011 / 1109

Downloaded 23 Dec 2011 to 203.197.118.86. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
Table 3. Values of Model Parameters for Vanttila Clay
λi κ e0 σ0rp0 [kPa] υ0 Mc αK0 ω χ0 ξ ξd μ ½h
1  β
0.31 0.057 3.3 29 0.2 1.35 0.52 12 77 13.5 0.3 1:57 × 10
4 22.7

0 'v = 9.4 kPa

Calibration of Model Parameters
18.75 kPa
The calibration of model parameters is based on one conventional 10
37.5 kPa
oedometer test combined with one undrained triaxial test. As there 20
is natural variation between clay samples, a number of tests of the
30 75 kPa

(%)
same type were used to estimate the average values of parameters.
The procedure for calibrating model parameters is presented as 150 kPa

v
40
follows: 300 kPa
1. λi and κ were measured from 1-day conventional oedometer 50 600 kPa
tests on reconstituted samples (same as from intact samples at 1200 kPa
60
σ0v > 300 kPa; see Fig. 1). σ0rp0 was obtained with τ ¼ 1 day. Exp. Oed4725
Simulation
Exp. Oed4723
M c was measured from undrained triaxial test results. υ0 ¼ 0:2 70
was assumed. 0.001 0.01 0.1 1 10 100 1000
2. αK0 and ω were calculated by Eqs. (16) and (18), respectively. (a) Time (hr)
3. χ0 was determined from oedometer tests by χ0 ¼ σ0p0 =σ0pi0
1.
0 'v = 12.5 kPa
Note that as data on isotropic consolidation are not available on
25 kPa
Vanttila clay, a value ξ d ¼ 0:3 (ξ d ¼ 0:2–0:3 for Finnish clays 10
by Karstunen et al. 2005) was assumed. ξ was determined from
the oedometer tests by using Eq. (25). 20
50 kPa
4. C αei was measured from 1-day conventional oedometer tests

(%)
30
on reconstituted samples (same as from intact samples at σ0v > 100 kPa

v
300 kPa; see Fig. 6). β and μ were then calculated by Eq. (33). 40 200 kPa
5. k 0 ¼ 1:1 × 10
5 m=h with ck ¼ 1:65 was obtained from 50
400 kPa
1-day conventional oedometer tests on intact vertical samples 800 kPa
(horizontal permeability is not needed during oedometer test 60 Exp. Oed4788
Simulation
simulation). 70
The selected values of parameters for Vanttila clay, used for all 0.001 0.1 10 1000
test simulations, are summarized in Table 3. (b) Time (hr)
Simulations for Oedometer Tests 0 'v = 9.4 kPa
Fig. 7(a) shows good agreement between experiments and simula- 18.75 kPa
10
tions for the settlement-time curves of 1-day conventional oedom-
eter tests by using the average values of parameters in Table 3. 20 37.5 kPa
Good agreement between experiments and simulations was also 30
(%)

achieved for 10 days [Fig. 2(b)] and 100 days [Fig. 2(c)] conven- 75 kPa
v

tional oedometer tests by using parameters from 1-day oedometer 40 150 kPa
tests, demonstrating the model’s predictive ability. The proposed 300 kPa
50
model captures well the trend of the creep behavior, in particular 600 kPa
at stress levels that exceed the apparent vertical preconsolidation 60 Exp. Oed4814
stress. Simulation
70
To further study the effect of destructuration on creep, the
0.001 0.1 10 1000 100000
experimental results on the evolution of the apparent coefficient
(c) Time (hr)
of secondary compression [C 0αe ¼ Δe=Δ ln t for all time represent-
ing apparent strain rate of the soil; see Fig. 8(a)] and relative de- Fig. 7. Comparison between experiments and simulations for
structuration rate [defined by Δχ=ðχΔ ln tÞ; see Fig. 8(b) where conventional oedometer tests with loading increment duration:
the amount of bonding is obtained for the void ratio from Fig. 1 (a) 1 day, (b) 10 days, and (c) 100 days
as Eq. (3)] with time for the test Oed4814 have been plotted. Fig. 8
shows the model captured well the trend of the experiments.
Since the effective stresses change during consolidation, this
also influences the apparent value of C 0αe . The dissipation of rel- for 100 days test), the peak value of C 0αe occurs during the
ative excess pore pressure (i.e., excess pore pressure normalized primary consolidation. This is attributable to the increase of
by the load increment Δu=Δσ0v for each loading step) in the middle effective stress and bond degradation. Since the rate of increase
of the sample is plotted in Fig. 8(c). The time corresponding to full in effective stress during these stages is high enough to cause
dissipation of excess pore pressure from this figure is marked in significant bond degradation, the rate of degradation decreases
Figs. 8(a) and 8(b), in order to help in understanding the effect at a certain time, at which point C 0αe starts decreasing.
of destructuration on creep. Destructuration influences the creep 2. For the next two load increments (300 and 600 kPa), the
behavior under one-dimensional consolidation as follows: increase and decrease of C 0αe during the primary consolidation
1. For the first three load increments beyond the preconsolidation is mainly attributable to the change of effective stress. After
pressure (from 18.25 to 37.5, 75, 150 kPa, given σ0p0 ¼ 26 kPa that, a slight decrease of C0αe with time occurs, since most

1110 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2011

Downloaded 23 Dec 2011 to 203.197.118.86. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
 0.2 Oed4814 37.5 kPa bonds have already been destroyed (χ ¼ 0:16 at the end of the
Full dissipation of u from 75 kPa load increment of 150 kPa).
simulation 150 kPa
0.15 300 kPa Fig. 9 shows the model’s predictive ability for the strain-rate
600 kPa effect on the preconsolidation pressure, as expected by the principle
Simulation of the proposed scaling function. The predictive ability of the
u=0
e

0.1 model on the strain-rate test by using parameters from the conven-
C'

tional test demonstrates that, from a practical viewpoint, there is

no difference as to whether the parameters are determined from
0.05 oedometer tests at constant strain-rate or conventional oedometer
tests.
0 Simulations for Undrained Triaxial Tests at Constant
0.01 0.1 1 10 100 1000 10000
Strain Rate
(a) Time (hr)
Fig. 10 shows the comparison between the predicted and measured
0.8
lnt )

Oed4814 37.5 kPa

results for the shearing stage of undrained triaxial tests at constant
Full dissipation of u from 75 kPa strain rates. In test CAUC2, the sample was sheared at ε_ a (axial
0.6 simulation 150 kPa strain rate) of 1%/h up to 12.8% of deviator strain, followed by an
300 kPa unloading at ε_ a of
1%=h up to 12.4% deviator strain, and then
600 kPa reloaded at ε_ a of 10%=h up to failure. Even though the elastic
Destructuration rate =

Simulation
0.4 u=0 model is very simple, the predictions for the unloading-reloading
loop are reasonably good. Overall, the model predictions are
reasonably good and capture well the strain-rate effect on the de-
0.2 viatoric stress coupled with destructuration, although some discrep-
ancies were found between predicted and measured results. One
possible reason is the variation of natural samples (as indicated
0
by the variation in the physical properties shown in previous
0.01 0.1 1 10 100 1000 10000
sections) and different disturbances for different samples during
(b) Time (hr)
sample preparation for such a highly sensitive clay.
1 During an undrained triaxial test at constant strain rate, minor
Oed4814 viscoplastic volumetric strains occur and cause the initial intrinsic
0.8 reference yield surface to expand in the stress space. The viscoplas-
tic volumetric and deviatoric strains also cause changes in the
inclination of the yield surfaces and degradation in bonding. The
0.6
'v

destructuration may reduce the size of the reference yield surface.

'v from 75 to 600 kPa
The viscoplastic strains cause debonding of soil particles until the
u/

0.4 intrinsic and reference yield surfaces coincide (where χ ¼ 0). Since
the evolution of anisotropy and destructuration relates to (rate-
0.2 dependent) viscoplastic strains, viscosity is coupled with destruc-
'v = 37.5 kPa turation and the changes in anisotropy.
0
0.01 0.1 1 10 100 1000 10000 Simulations for Undrained Triaxial Creep Tests
(c) Time (hr)
Fig. 11 shows the predicted and measured results for undrained
Fig. 8. Comparison between experiments and simulations for test triaxial creep tests under three different stress levels (see Table 3).
Oed4814. The evolution of (a) C αe , (b) destructuration rate versus time, Generally, for higher applied deviator stresses the predicted excess
and (c) relative excess pore pressures versus time pore pressures are higher than those for low deviator stresses. Here
for the test CAUCR2 the excess pore pressure is lower than that for
test CAUCR1. This is attributable to the applied stress under con-
10 stant p0 condition for the test CAUCR2 after the consolidation stage
Experiment (i.e., Δσ01 ¼ 4:6 kPa, Δσ03 ¼
2:3 kPa). The model predictions
Simulation are reasonably good and capture the trends well. The difference
between predictions and experiments is possibly attributable to
the variation of natural samples and disturbance during sample
'p0/ 'v0

preparation.
The most interesting results are shown in Fig. 11(c), where the
axial strain rate has been plotted versus time. The model captures
well the evolution of strain rates for all stress levels, although
the strain rates are underpredicted for the first 10 min. It is worth
mentioning that at the highest deviator stress level, the model cap-
1 tures the creep rupture or tertiary creep, i.e., accelerating strain rates
1E-7 1E-6 1E-5 1E-4 with time.
-1
d v /dt (s ) Fig. 11(d) shows the evolution of the amount of bonding during
the creep process for the simulations by the proposed model. For an
Fig. 9. Comparison between experiments and simulations for strain-
undrained creep test, the size of the reference yield surface changes
rate effect on the initial preconsolidation pressure
slightly with viscoplastic volumetric strains, relating to the amount

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2011 / 1111

Downloaded 23 Dec 2011 to 203.197.118.86. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
 80 60
CAUC1
CAUC2
CAUC3
60 Simulation
40

∆ u (kPa)
1 %/h
q (kPa) 40
10 %/h

1 %/h 20 CAUC1
-1 %/h
20 CAUC2
1 %/h
CAUC3
10 %/h
Simulation
0 0
0 5 10 15 20 0 5 10 15 20
(a) ε d (%) (b) ε d (%)

Fig. 10. Comparison between experiments and simulations for undrained triaxial tests at constant strain rates: (a) stress-strain behavior, (b) excess
pore pressures

4 20
CAUCR1 ∆σ'1 =9.6 kPa CAUCR1
CAUCR2 ∆σ'3 =0 kPa CAUCR2
3 CAUCR3 15 CAUCR3
Simulation Simulation

∆ u (kPa)
∆σ'1=4.6 kPa
ε d (%)

2 ∆σ'3=-2.3 kPa 10

1 ∆σ'1=4.1 kPa 5
∆σ'3 =0 kPa
0 0
0.01 0.1 1 10 100 1000 10000 0.01 0.1 1 10 100 1000 10000
(a) Time (hr) (b) Time (hr)

80
Simulation with three features
1E+0 CAUCR1 ∆σ '1=4.1 kPa
∆σ '3=0 kPa
70
dε d/dt (%/h)

CAUCR2
1E-2 ∆σ '1=4.6 kPa
60
χ

∆σ '3=-2.3
CAUCR1 CAUCR3
1E-4 CAUCR2 50 ∆σ '1 =9.6 kPa
CAUCR3
Simulation ∆σ '3 =0 kPa
1E-6 40
0.01 0.1 1 10 100 1000 10000 0.01 0.1 1 10 100 1000 10000
(c) Time (hr) (d) Time (hr)

Fig. 11. Comparison between experiments and simulations for undrained triaxial creep tests under different deviator stress levels:
(a) deviatoric strains versus time, (b) pore pressures versus time, (c) deviatoric strain rates versus time and (d) the evolution of the amount of
bonding versus time

of overstress. The size of the reference yield surface is, however, decreasing. For high applied deviator stress levels, if the effective
also influenced by destructuration [prm ¼ pmi ð1 þ χÞ] influenced stresses reach the critical state line, there will naturally be no
also by the deviatoric creep strains. Therefore, viscosity and de- changes in the size of the intrinsic surface (_εvpv ¼ 0). For the case
structuration have strong coupling. In the case of a highly sensitive that the intrinsic surface and the reference yield surface coincide
clay, destructuration dominates, while for insensitive soils viscosity (χ ¼ 0), the overstress becomes constant, resulting in a constant
dominates. strain rate. Otherwise, the intrinsic surface is still inside of the refer-
For low applied deviator stress levels, destructuration results in a ence yield surface (and bonds exist). The viscoplastic deviatoric
slow decrease in the size of the reference yield surface [see the slow strains continue to increase, which causes a reduction in the
decrease in χ in Fig. 11(d)]. As a result, the amount of overstress size of the reference yield surface, which in turn increases the
(pdm =prm ) is decreasing during creep, and thus the strain rate is also amount of overstress. Through this way, the tertiary creep phase,

1112 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2011

Downloaded 23 Dec 2011 to 203.197.118.86. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org
i.e., deformation with an increasing strain rate, is reproduced. Karstunen, M., and Koskinen, M. (2008). “Plastic anisotropy of soft recon-
Therefore, the model explains one possible reason for the tertiary stituted clays.” Can. Geotech. J., 45(3), 314–328.
creep: the destructuration of soft sensitive clay. This is different Karstunen, M., Krenn, H., Wheeler, S. J., Koskinen, M., and Zentar, R.
from the theory of undrained creep rupture proposed by Sekiguchi (2005). “The effect of anisotropy and destructuration on the behaviour
of Murro test embankment.” Int. J. Geomech., 5(2), 87–97.
(1984) for insensitive clays.
Karstunen, M., and Yin, Z.-Y. (2010). “Modelling time-dependent behav-
iour of Murro test embankment.” Géotechnique, 60(10), 735–749.
Conclusions Katona, M. G. (1984). “Evaluation of viscoplastic cap model.” J. Geotech.
Eng., 110(8), 1106–1125.
Various oedometer and triaxial tests have been carried out on Kimoto, S., and Oka, F. (2005). “An elasto-viscoplastic model for clay
Vanttila clay to study the coupling between destructuration and considering destructuralization and consolidation analysis of unstable
behaviour.” Soils Found., 45(2), 29–42.
viscosity of soft sensitive clays. Based on that, a new constitutive
Kutter, B. L., and Sathialingam, N. (1992). “Elastic-viscoplastic modelling
model (AniCreep) accounting for soil anisotropy, destructuration,
of the rate-dependent behaviour of clays.” Géotechnique, 42(3),
and viscosity has been developed. The determination of model 427–441.
parameters is straightforward, requiring the same basic tests as Leoni, M., Karstunen, M., and Vermeer, P. A. (2008). “Anisotropic creep
for the modified Cam clay model. This makes the model attractive model for soft soils.” Géotechnique, 58(3), 215–226.
for geotechnical practice. Leroueil, S., Kabbaj, M., Tavenas, F., and Bouchard, R. (1985). “Stress-
The model was implemented into finite-element code, which strain-strain-rate relation for the compressibility of sensitive natural
enables coupled consolidation analyses. The experimental valida- clays.” Géotechnique, 35(2), 159–180.
tion was carried out with reference to the tests on intact samples of Leroueil, S., and Vaughan, P. R. (1990). “The general and congruent effects
Vanttila clay: oedometer tests with different durations (1, 10, and of structure in natural soils and weak rocks.” Géotechnique, 40(3),
100 days) of load increment, oedometer tests at constant strain 467–488.
rates, undrained triaxial tests at constant strain rates, and undrained Mesri, G., and Godlewski, P. M. (1977). “Time and stress-compressibility
triaxial creep tests. The comparisons between experimental results interrelationship.” J. Geotech. Engrg. Div., 103(5), 417–430.
and simulations show that the model has good predictive ability on Oka, F., Adachi, T., and Okano, Y. (1986). “Two-dimensional consolidation
analysis using an elasto-viscoplastic constitutive equation.” Int. J.
the time-dependent behavior of soft sensitive clay.
Numer. Anal. Methods Geomech., 10(1), 1–16.
Oka, F., Kodaka, T., and Kim, Y.-S. (2004). “A cyclic viscoelastic-
viscoplastic constitutive model for clay and liquefaction analysis of
Acknowledgments
multi-layered ground.” Int. J. Numer. Anal. Methods Geomech., 28(2),
131–179.
The work presented was sponsored by the Academy of Finland
Perzyna, P. (1963). “The constitutive equations for work-hardening and
(Grants 210744 and 1284594) and the European Community
rate sensitive plastic materials.” Proc. Vibration Problems, Warsaw,
through the program “People” as part of the Industry-Academia 4(3), 281–290.
Pathways and Partnerships project GEO-INSTALL (PIAP-GA- Perzyna, P. (1966). “Fundamental problems in viscoplasticity.” Adv. Appl.
2009-230638). Mech., 9, 243–377.
Rocchi, G., Fontana, M., and Da Prat, M. (2003). “Modelling of natural soft
clay destruction processes using viscoplasticity theory.” Géotechnique,
References 53(8), 729–745.
Roscoe, K. H., and Burland, J. B. (1968). “On the generalized stress-strain
Adachi, T., and Oka, F. (1982). “Constitutive equations for normally
behaviour of ‘wet’ clay.” Engineering plasticity, Cambridge University
consolidated clay based on elasto-viscoplasticity.” Soils Found., 22(4),
Press, Cambridge, UK, 553–609.
57–70.
Sekiguchi, H. (1984). “Theory of undrained creep rupture of normally con-
Asaoka, A., Noda, T., Yamada, E., Kaneda, K., and Nakano, M. (2002).
solidated clay based on elasto-viscoplasticity.” Soils Found., 24(1),
“An elasto-plastic description of two distinct volume change mecha-
129–147.
nisms of soils.” Soils Found., 42(5), 47–57.
Berry, P. L., and Poskitt, T. J. (1972). “The consolidation of peat.” Sheng, D., Sloan, S. W., and Yu, H. S. (2000). “Aspects of finite element
Géotechnique, 22(1), 27–52. implementation of critical state models.” Comput. Mech., 26, 185–196.
Burland, J. B. (1990). “On the compressibility and shear strength of natural Suklje, L. (1957). “The analysis of the consolidation process by the
clays.” Géotechnique, 40(3), 329–378. isotaches method.” Proc., 4th Int. Conf. on Soil Mechanics and
Dafalias, Y. F. (1986). “An anisotropic critical state soil plasticity model.” Foundation Engineering (ICSMFE), Butterworths Scientific, London,
Mech. Res. Commun., 13(6), 341–347. 200–206.
Gens, A., and Nova, R. (1993). “Conceptual bases for a constitutive model Vermeer, P. A., and Neher, H. P. (1999). “A soft soil model that accounts
for bonded soils and weak rocks.” Proc., Int. Symp. on Hard Soils—Soft for creep.” Proc., Plaxis Symposium “Beyond 2000 in Computational
Rocks, Athens, A. A. Balkema, Rotterdam, 485–494. Geotechnics”, Balkema, Rotterdam, 249–262.
Graham, J., Crooks, J. H. A., and Bell, A. L. (1983). “Time effects on the Watabe, Y., Udaka, K., and Morikawa, Y. (2008). “Strain rate effect on
stress–strain behaviour of natural soft clays.” Géotechnique, 33(3), long-term consolidation of Osaka Bay clay.” Soils Found., 48(4),
327–340. 495–509.
Hinchberger, S. D., and Qu, G. (2009). “Viscoplastic constitutive ap- Wheeler, S. J., Näätänen, A., Karstunen, M., and Lojander, M. (2003). “An
proach for rate-sensitive structured clays.” Can. Geotech. J., 46(6), anisotropic elasto-plastic model for soft clays.” Can. Geotech. J., 40(2),
609–626. 403–418.
Hinchberger, S. D., and Rowe, R. K. (2005). “Evaluation of the predictive Yin, J. H., Zhu, J. G., and Graham, J. (2002). “A new elastic viscoplastic
ability of two elastic-viscoplastic constitutive models.” Can. Geotech. model for time-dependent behaviour of normally and overconsolidated
J., 42(6), 1675–1694. clays: Theory and verification.” Can. Geotech. J., 39(1), 157–173.
Jaky, J. (1948). “Earth pressure in soils.” Proc., 2nd Int. Conf. on Soil Yin, Z.-Y., and Hicher, P. Y. (2008). “Identifying parameters controlling
Mechanics and Foundation Engineering, Rotterdam, The Netherlands, soil delayed behaviour from laboratory and in situ pressuremeter test-
pp. 103–107. ing.” Int. J. Numer. Anal. Methods Geomech., 32(12), 1515–1535.

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / NOVEMBER 2011 / 1113

Downloaded 23 Dec 2011 to 203.197.118.86. Redistribution subject to ASCE license or copyright. Visit http://www.ascelibrary.org