Computers and Geotechnics 75 (2016) 1–11

Contents lists available at ScienceDirect

Computers and Geotechnics

journal homepage: www.elsevier.com/locate/compgeo

Research Paper

A state-dependent critical state model for methane hydrate-bearing

sand
J. Shen a, C.F. Chiu a,⇑, C.W.W. Ng b, G.H. Lei a, J. Xu a
a
Key Laboratory of Geomechanics and Embankment Engineering of the Ministry of Education, Geotechnical Research Institute, Hohai University, 1 Xikang Road, Nanjing 210098, China
b
Department of Civil and Environmental Engineering, The Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong Special Administrative Region

a r t i c l e i n f o a b s t r a c t

Article history: Methane hydrate exists in the pores of methane hydrate-bearing sand (MHBS) and is considered to be a
Received 2 August 2015 potentially significant source of methane and thus energy for mankind. However, before conducting a
Received in revised form 11 December 2015 large-scale extraction of methane from MHBS, it is crucial to simulate the mechanical behaviour of
Accepted 15 January 2016
MHBS and evaluate its stability during drilling and methane production. In this paper, a state-
dependent critical state model for MHBS is presented. The critical state of MHBS is discussed, and critical
state line formulations are introduced as functions of hydrate saturation. A simple nonlinear bonding and
Keywords:
linear debonding law is incorporated considering the cementing mechanism of hydrate. A modified state-
Methane hydrate
Sand
dependent dilatancy is proposed to account for the effects of stress level, internal state (density), bonding
Critical state strength and hydrate saturation. Determination of the model parameters is described in detail. The pro-
Dilatancy posed model is employed to predict results of drained triaxial compression tests on MHBS. Satisfactory
Constitutive modelling performance is demonstrated, i.e., the model can adequately capture the stress–strain and volume change
behaviours of MHBS over a wide range of hydrate saturations, confining pressures and densities using a
unified set of parameters.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction
Gas hydrate is a crystalline compound formed from a combina-
tion of water and low molecular weight gas, e.g., methane and car-
bon dioxide. Methane hydrate is the most commonly occurring
natural hydrate and is considered as one of the potential sources
of energy for the future [1–4]. Natural methane hydrate occurs in
abundance in deepwater marine sediments and permafrost regions
where appropriate temperature and pressure conditions exist [5,6].
Coarse-grained sediments tend to contain methane hydrates in
pore spaces, while in fine-grained sediments, methane hydrates
often exist in the forms of discrete nodule, vein, veinlet and layer.
According to field investigation, more methane hydrates tend to
occur in coarse-grained sediments (i.e., Nankai Trough and
Mallik-Mackenzie Delta) than in fine-grained sediments (i.e.,
Hydrate Ridge and Blake Ridge) [7]. Thus, more attention has been
given to extracting natural methane hydrates from coarse-grained
reservoirs due to economic benefits. Understanding the mechani-
cal properties of methane hydrate-bearing sand (MHBS), a natural
⇑ Corresponding author. Tel.: +86 13770554507; fax: +86 25 83713073.
E-mail addresses: shenjiehhu@163.com (J. Shen), acf.chiu@gmail.com (C.F. Chiu),
cecwwng@ust.hk (C.W.W. Ng), leiguohui@hhu.edu.cn (G.H. Lei), cexujie@gmail.com
(J. Xu).
soil deposit, is crucial to evaluate its stability during drilling and
methane production [8–10].
Past studies have shown that the mechanical properties of
MHBS are significantly influenced by the presence of hydrates
(hydrate saturation and hydrate accumulation habit), confining
pressure and density [11–18]. It should be noted that the hydrate
saturation (Sh), which is widely used in the literature to describe
the hydrate content in sediments, is defined as follows [11]:
Vh
Sh ¼ ð1Þ
Vh þ VV
where Vh and VV are the volumes of the hydrate and the pore space,
respectively. For a given hydrate saturation, the hydrate accumula-
tion habit within the soil pore space is one of the biggest factors
influencing the response of MHBS under different loading condi-
tions. Waite et al. [19] presented the following three possible
hydrate accumulation habits in coarse-grained sediments:
(1) Pore filling – hydrates exist only within the pore space and
do not contribute to the load bearing of the soil skeleton.
(2) Load bearing – hydrates form part of the soil skeleton that
contributes to the stability of the soil structure [20].
(3) Grain cementing – hydrates act as bonding agents to cement
the soil grains at intergranular contacts.

http://dx.doi.org/10.1016/j.compgeo.2016.01.013
0266-352X/Ó 2016 Elsevier Ltd. All rights reserved.
2 J. Shen et al. / Computers and Geotechnics 75 (2016) 1–11
Different hydrate accumulation habits lead to different mecha-
nisms of deformation for MHBS. Fig. 1(a) illustrates the pore-filling
mechanism of hydrates. This figure shows that the hydrate exists
only in the pore space and does not come into contact with soil
grains. Under this condition, the hydrate primarily affects the pore
fluid stiffness and conduction properties [21]. The load-bearing
mechanism is depicted in Fig. 1(b), where hydrates bridge neigh-
bouring soil grains and become part of the load-bearing framework
mainly derived from dilation enhancement and frictional resis-
tance at the hydrate-grain contacts. Fig. 1(c) shows the cementing
mechanism of hydrates. In this case, the hydrate cements the inter-
granular contacts, and a small amount of hydrate can dramatically
increase the stiffness and shear strength of the host sand by bond-
ing adjacent grains together [22].
In the last decade, experimental studies on synthesised
MHBS have been widely reported in the literature
[11,13,14,16,17,23,24]. However, only a few attempts have been
made towards modelling the behaviour of MHBS. Freij-Ayoub
et al. [25], Rutqvist and Moridis [26] and Klar et al. [27] developed
extended Mohr–Coulomb models for MHBS, which consider the
stiffness, strength parameters and dilation angle of MHBS as
functions of hydrate saturation. However, the strain-softening
behaviour and debonding of hydrate were not modelled, and
the volume change of the specimens could not be satisfactorily
predicted. Uchida et al. [28] and Lin et al. [29] extended the
Cam-Clay model for MHBS by taking into account the degradation
of hydrate bonding based on the concept of critical state soil
mechanics. However, the critical state of MHBS was assumed to
be independent of the presence of hydrates in the above critical
state-based model. This assumption is unlikely to be true and will
be discussed in the following section. Moreover, the associated
flow rule was adopted, which had been proved inappropriate in
sand modelling, thus resulting in an inaccurate volume change
prediction of MHBS. In addition to hydrate saturation, density is
also an important factor that can significantly influence the
mechanical behaviour of MHBS. So far, the mechanical behaviour
of MHBS has not been captured over a wide range of densities
using a unified set of parameters.
In this study, a state-dependent critical state constitutive
model for MHBS is developed. The critical state behaviour of
MHBS is introduced based on the experimental data from large-
strain triaxial compression tests conducted by Hyodo et al. [13].
A modified state-dependent dilatancy is presented to account
for the effects of stress level, internal state, bonding strength
and hydrate saturation. A simple bonding and debonding law is
used to describe the evolution of the hydrate-induced bonding.
Details of the model formulation and parameter calibration are
described. The accuracy of the model is demonstrated by compar-
ing its predicted values with results from the drained triaxial
tests on MHBS under various hydrate saturations, densities and
confining pressures.
(a) (b) (c)
Fig. 1. Three possible hydrate accumulation habits in coarse-grained sediment: (a)
pore filling, (b) load bearing, and (c) grain cementing.
2. Model formulation
The proposed model is developed within an elasto-plastic
framework and takes into account of the possible occurrences of
irrecoverable plastic strains. The total strain increment (de) is the
sum of the elastic strain increment (dee) and the plastic strain
increment (dep). The model assumes an elastic behaviour if the soil
remains inside a yield surface and a plastic behaviour commences
once the yield surface is reached. The model is formulated in a p0 –q
space. Mean effective stress and deviator stress are defined as
p0 ¼ ðr01 þ 2r03 Þ=3 and q ¼ r01  r03 , respectively, where r01 and r03
are the major and minor principal stresses, respectively. The work
conjugate strain rates for p0 and q are dev and deq, respectively
(volumetric strain increment dev ¼ de1 þ 2de2 and shear strain
increment deq ¼ 2ðde1  de3 Þ=3, where de1 and de3 are the principal
strain increments). Under the above elasto-plastic framework, dev
and deq are decomposed into
dev ¼ deev þ depv ð2Þ
deq ¼ deeq þ depq ð3Þ
where deev and depv are elastic and plastic volumetric strain incre-
ments, respectively, and deeq and depq are elastic and plastic shear
strain increments, respectively.
2.1. Elastic strain
According to the theory of elasticity, deev and deeq are calculated
as
0
dp
deev ¼ ð4Þ
K
dq
deeq ¼ ð5Þ
3G
where K and G are the elastic bulk and shear moduli, respectively.
For clean sand, G can be expressed by the following empirical
equation [30]:
ð2:97  eÞ2 pffiffiffiffiffiffiffiffi
ffi
G ¼ G0 p0 pa ð6Þ
1þe
where G0 denotes a model parameter, and pa is the atmospheric
pressure, taken as 100 kPa.
For MHBS, Uchida et al. [28] suggested that G is the summation
of the shear modulus of the sand skeleton and the shear stiffness
increase due to the presence of hydrates. The shear stiffness
increase was assumed to be a linear function of hydrate saturation.
As expressed by Eq. (6), G of the sand skeleton is a function of the
void ratio and the effective stress. However, the dependence of the
shear stiffness increase on the void ratio and the effective stress is
unknown for MHBS. In the proposed model, it is assumed that
Eq. (6) is still applicable for MHBS, and G0 is affected by hydrate
saturation. Fig. 2 shows the resonant column test results in terms
of G0 for MHBS reported by Clayton et al. [24]. The MHBS speci-
mens were prepared using the ‘partial water saturation’ method,
in which methane hydrate was formed by flushing pressurised
methane gas through unsaturated sand specimens and cooling into
the hydrate stability zone. An approximately linear relation
between G0 and hydrate saturation is observed. Based on the above
observation, it is assumed that G0 increases linearly with increas-
ing hydrate saturation for MHBS:
G0 ðSh Þ ¼ G0 þ nSh ð7Þ
where G0(Sh) and G0 are model parameters for MHBS and clean
sand, respectively, and n is a positive model parameter denoting
 J. Shen et al. / Computers and Geotechnics 75 (2016) 1–11 3

Fig. 2. Variation in G0(Sh) with hydrate saturation (test data from Clayton et al.
Fig. 3. Isotropic compression of MHBS under a confining pressure ranging from 1 to
[24]).
5 MPa.

the gradient of the best-fit line as shown in Fig. 2. In Eq. (7), Sh

represents the combined effects of pore-filling and cementing
mechanisms of hydrates on the stiffness of MHBS. However, the
contribution to an increase in stiffness for each mechanism cannot
be identified separately. As MHBS degrades from a cemented state
to a cohesionless state, the contribution of hydrate cementation to
stiffness is expected to decrease. On the other hand, the contribu-
tion of pore-filling of hydrates to stiffness remains unchanged.
Therefore, as a first approximation, the decrease in stiffness due
to hydrate debonding is not considered in the proposed model.
Based on elasticity theory, the elastic bulk modulus K is equal to
2ð1 þ mÞ
K¼G ð8Þ
3ð1  2mÞ
where m is the Poisson’s ratio, which is assumed to be independent
of density, pressure and hydrate saturation in the model.
Fig. 4. Yield surface in the p0 –q plane.
2.2. Yield function

For a frictional material such as sand, under confining pressures where Kp is a plastic hardening modulus. The plastic shear and
of interest, a constant stress-ratio loading induces only relatively volumetric strain increments can then be written as follows [32]:
small plastic volumetric strains before grain crushing [31]. Li
1
depq ¼ L ¼
0
et al. [32,33] assumed that a stress path of a constant stress ratio, ðdq  g dp  g dpb Þ ð11Þ
which is not equal to the ultimate stress ratio, induces no plastic Kp
deformation in the modelling of clean sand. Past studies on MHBS
D
depv ¼ LD ¼
0
[13] have shown that negligible plastic strains are produced under ðdq  g dp  g dpb Þ ð12Þ
isotropic compression (i.e., a type of constant stress ratio loading) Kp
at a confining pressure ranging from 1 to 5 MPa, as shown in Fig. 3.
where D is the dilatancy defined as depv =depq . Combining with Eqs.
Therefore, in the proposed model for MHBS, plastic deformation is
(2)–(5), the incremental stress–strain relationship in the p0 –q space
assumed to occur whenever the stress ratio exceeds its historic
can be expressed as follows:
maximum. However, when a higher confining pressure or plastic
deformations under a constant stress ratio are to be considered, a 0
dp D
dev ¼ deev þ depv ¼
0
cap representing the compression mechanism can be added [34]. þ ðdq  g dp  g dpb Þ ð13Þ
K Kp
With the above assumption, the yield criterion for MHBS in the
p0 –q space (illustrated in Fig. 4) can be written as
dq 1
deq ¼ deeq þ depq ¼
0
þ ðdq  g dp  g dpb Þ ð14Þ
f ¼ q  g ðp0 þ pb Þ ð9Þ 3G K p
where pb is the bonding strength, and g⁄ is the stress ratio at yield-
ing, which is defined as g ¼ q=ðp0 þ pb Þ. 2.3. Critical state
By the theory of plasticity [35], a loading index L can be
obtained through the plastic consistency condition, which is The hydrate accumulation habit significantly influences the
expressed as follows: mechanical behaviour of MHBS. The following discussion is limited
  to MHBS exhibiting behaviours characterised by the aforemen-
1 @f 0 @f @f 1 0
L¼ dp þ dq þ dpb ¼ ðdq  g dp  g dpb Þ ð10Þ tioned three possible mechanisms. Hyodo et al. [12,13] conducted
K p @p0 @q @pb Kp
a series of triaxial compression tests on MHBS to study the effects
4 J. Shen et al. / Computers and Geotechnics 75 (2016) 1–11
of density and hydrate saturation on its mechanical behaviour.
They used specimens that were formed by the ‘partial water satu-
ration’ method. The test specimens were all sheared to large
strains, and it was assumed that the critical state was approxi-
mately reached at the end of each drained triaxial test. Fig. 5(a)
shows the effect of hydrate saturation on the critical state lines
(CSLs) in the e  ln p0 plane of MHBS. The CSL of the host sand is
also shown in the figure for comparison. The void ratio is defined
as follows:
VV þ Vh
e¼
VS
where VV, Vh and VS are the volumes of pore space, hydrate and sand
grains, respectively. Despite some scattering of the data, an approx-
imately linear relationship between the void ratio and the loga-
rithm of the mean effective stress (ranging from 2 MPa to 10 MPa)
is observed for each hydrate saturation tested. In this range of high
confining pressures, a similar linear relationship is also found in
clean sand [36]. The CSL shifts towards the upper side with increas-
ing hydrate saturation. In other words, MHBS exhibits a higher ten-
dency to dilate during shear with an increase in hydrate saturation
for a given confining pressure and void ratio.
Fig. 5(b) shows the effect of hydrate saturation on the CSL of
MHBS in the p0 –q space. The test data can be well fitted by several
straight lines, where the slope increases with increasing hydrate
saturation. In other words, the internal angle of friction at the crit-
ical state increases with increasing hydrate saturation. Fig. 6
depicts the variation in the stress ratio at the critical state (critical
Fig. 5. Critical state lines of MHBS in the (a) e  ln p0 plane and (b) p0 –q plane (test
data from Hyodo et al. [13]).
ð15Þ
Fig. 6. Variation in the critical stress ratio M(Sh) with hydrate saturation.
stress ratio), i.e., M(Sh) (M(Sh) = q/p0 ), with hydrate saturation. For
Sh below 60%, it is shown that M(Sh) increases nonlinearly with
increasing hydrate saturation. The change in the critical stress ratio
is negligible when the hydrate saturation is relatively low because
the pore-filling mechanism is dominant in the deformation of
MHBS (shown in Fig. 6), and the hydrate only has a small effect
on the mechanical behaviour of MHBS. As hydrate saturation fur-
ther increases, the critical stress ratio increases more rapidly
because of greater contacts between hydrates and grains and
because more hydrates will contribute to the sand-hydrate fric-
tional resistance and cause the load-bearing mechanism to become
dominant (shown in Fig. 6).
Based on the previous analysis, it is proposed that the critical
stress ratio and CSLs for MHBS alter with the hydrate saturation
as follows
MðSh Þ ¼ M þ ASBh ð16Þ
ec ðSh Þ ¼ eC ðSh Þ  kc lnðp0 =pa Þ ð17Þ
eC ðSh Þ ¼ eC þ aSbh ð18Þ
where M(Sh) is the critical stress ratio for MHBS for a given hydrate
saturation, M is the critical stress ratio for clean sand, and A and B
are two model parameters. pa is the atmospheric pressure, and eC,
kc , a and b are model parameters. Both ASBh and aSbh represent the
influence of hydrates in terms of the load-bearing habit on the crit-
ical state behaviour of MHBS.
2.4. State-dependent dilatancy
Various stress-dilatancy relationships have been proposed in
the literature for clean sand. Rowe [37] postulated that dilatancy
could be expressed as a function of the stress ratio and the internal
angel of friction between the mineral surfaces of the particles.
However, experiments found that the dilatancy of clean sand
depends on both the stress ratio and the internal state (e.g.,
density) [38]. Li et al. [32,33] introduced a framework of
state-dependent dilatancy by relating it to the stress ratio and
the density (or void ratio). They developed a constitutive model
for sand that can capture the stress–strain and volume change
behaviours of clean sand over a wide range of densities and pres-
sures. To date, state-dependent dilatancy has been widely used
in modelling the behaviours of clean sand [33], unsaturated soils
[34] and rockfill materials [39].
 J. Shen et al. / Computers and Geotechnics 75 (2016) 1–11
For MHBS, experiments have also revealed that the dilatancy of
MHBS depends on the stress ratio (or stress level), internal state,
bonding strength and hydrate saturation [13,14,17]. Fig. 7 illus-
trates the volume evolution of MHBS in the e  ln p0 plane during
drained triaxial shear tests. Three MHBS specimens with the same
hydrate saturation were prepared. Two specimens with low densi-
ties (high void ratios) were isotropically consolidated at effective
confining pressures of 1 MPa and 3 MPa before shear (see states
1 and 2 in Fig. 7), and a third specimen with a high density was
isotropically consolidated at an effective confining pressure of
3 MPa (see state 3). For MHBS specimens represented by states 1
and 2, completely inverse volume changes are produced during
the triaxial shear process. MHBS is dilative at a low confining pres-
sure but compressive at a high pressure. When subjected to the
same confining pressure (e.g., states 2 and 3), the two specimens
with low densities display volume compression, while the one
with a high density exhibits volume expansion. Thus, the dilatancy
of MHBS is significantly dependent on the internal state (mainly
density).
For clean sand, Li and Dafalias [33] proposed a general expres-
sion for the state-dependent dilatancy, D = D(g, e, Q, C), where g
is the stress ratio, defined as g = q/p, and Q is an internal state vari-
able distinct from the void ratio e, and C denotes the intrinsic mate-
rial constant. Two requirements must be satisfied for the general
expression. First, the dilatancy must be zero at a critical state,
i.e., D(g = M, e = ec, Q, C) = 0. Second, a ‘phase transformation state’
could be reached, at which D = 0, g – M and e – ec, i.e., D(g – M,
e – ec, Q, C) = 0.
To incorporate the accumulation habit of methane hydrate into
the dilatancy of clean sand, a modified general dilatancy expres-
sion is proposed:
D ¼ Dðg ; e; pb ; f h ; Q ; CÞ
where pb and fh denote the effects of the cementing and load-
bearing mechanisms of hydrates on dilatancy, respectively. The
above basic requirements for formulating the dilatancy of clean
sand are also applicable for MHBS.
At the critical state,
Dðg ¼ MðSh Þ; e ¼ ec ðSh Þ; pb ¼ 0; Q; CÞ ¼ 0
and at the phase transformation state,
Dðg –MðSh Þ; e–ec ðSh Þ; pb ; Q ; CÞ ¼ 0
To describe the state of clean sand, Been and Jefferies [40] pro-
posed a state parameter w ¼ e  ec to measure the difference
between the current state and the critical state in terms of the
Fig. 7. Volumetric behaviour of MHBS in the e  ln p0 plane during drained triaxial
shear tests (test data from Hyodo et al. [13]).
5
density (void ratio), as shown in Fig. 8. Based on the above require-
ments, Li and Dafalias [33] suggested an expression of state-
dependent dilatancy for clean sand by incorporating the state
parameter w. The expression is
d0
D¼ ðMemw  gÞ ð22Þ
M
where d0 and m are two model parameters. Similar to Eq. (22), a
modified state-dependent dilatancy equation satisfying the require-
ments of Eqs. (20) and (21) is proposed for MHBS:
d0
D¼ ½MðSh ÞemwðSh Þ db  g  ð23Þ
MðSh Þ
where w(Sh) is the state parameter for MHBS, defined as w(Sh)
= e  ec(Sh). db is a scaling factor, which is a function of the bonding
strength pb, to characterise the cementing mechanism of hydrates
on dilatancy. One type of formulation for db is suggested as follows:
pffiffiffiffiffiffiffiffiffi
db ¼ emb pb =pa
ð24Þ
where mb is a positive model parameter. Applying Eqs. (9) and (21)
to Eq. (23), the stress ratio at the phase transformation state for
MHBS can be obtained: g ¼ MðSh ÞemwðSh Þ db ð1 þ pb =pÞ. It can be
observed that the addition of a scaling factor, db, which is always
lower than 1, reduces the phase transformation stress ratio. On
the other hand, the last term ð1 þ pb =pÞ on the right-hand side of
the equation slightly increases the phase transformation stress
ratio. Thus, such a formulation provides the possibility of increasing
or decreasing the phase transformation stress ratio for MHBS with
hydrate bonding, which depends on the value of the parameter mb.
ð19Þ 2.5. Bonding and debonding laws
Similar to cemented sand, interparticle bonding in artificial
MHBS due to hydrate cementing is generally measured by the tri-
axial tensile strength of MHBS [41], called the bonding strength
(pb) here, which is employed as a macroscopic counterpart to the
particle-scale bonding to account for the cementing habit of the
ð20Þ hydrate. Abdulla and Kiousis [42] found that the bonding strength
increases nonlinearly with cementing agent content. Therefore, the
initial bonding strength pb0 can be defined as a power function of
ð21Þ hydrate saturation:
pb0 ¼ aSbh ð25Þ
where a and b are two positive model parameters.
Fig. 8. State parameter in the e  ln p0 plane.
6 J. Shen et al. / Computers and Geotechnics 75 (2016) 1–11

Experimental observations showed that the bonding of soils
weakens gradually because of the development of plastic deforma-
tion and is almost completely damaged when the critical state is
reached [39,42–44]. According to the test data from Hyodo et al.
for MHBS (shown in Fig. 3) [13], the volume change is very limited
under isotropic compression under a confining pressure ranging
from 1 to 5 MPa, which covers most in-situ stress states for MHBS.
It is therefore assumed that isotropic compression under the con-
fining pressures of interest results in a negligible debonding of
hydrates. Thus, debonding of hydrates due to isotropic compres-
sion is not considered in this paper. Under the above assumption,
a simple linear relation between the degradation rate of the bond-
ing strength on a natural logarithmic scale and the increment in
the plastic shear strain is assumed as follows:
3.2. Calibration of parameters for MHBS
Apart from the 10 model parameters for the host sand, 9 addi-
tional parameters are required for MHBS (see Table 2), i.e., an elas-
tic modulus parameter (n), four critical state parameters (A, B, a
and b), two bonding parameters (a and b), one debonding param-
eter (kd) and one dilatancy parameter (mb). The following sections
provide detailed information on the calibration of these 9 addi-
tional model parameters using data from the drained triaxial tests
for MHBS.
Table 2
Additional model parameters for MHBS.

Parameter Symbol Value

d ln pb ¼ kd depq ð26Þ
Elastic modulus n 1200
Critical state A 0.35
where kd is a positive model parameter. B 1.4
a 0.6
b 2.0
2.6. Hardening rule Bonding a (MPa) 0.7
b 1.6
For the plastic modulus Kp, the following modified constitutive Debonding kd 8
relation is proposed:
Dilatancy mb 0.04
nwðSh Þ
hGe
Kp ¼ ½MðSh ÞenwðSh Þ  g  ð27Þ
g

where n is a model parameter, and the plastic modulus parameter h

is used to scale the effect of density on the plastic modulus, defined
as a linear function of the initial void ratio e0:

h ¼ h1  h2 e0 ð28Þ

where h1 and h2 are model parameters. Kp takes a similar form as

the plasticity modulus proposed by Li and Dafalias [33] for clean
sand. Eq. (27) relates the plastic modulus to the stress ratio g⁄
and the state parameter w(Sh). This type of formulation can well
capture the strain-hardening and strain-softening behaviours of
sand over a wide range of densities and confining pressures
[33,34,39].

3. Calibration of model parameters

3.1. Calibration of parameters for the host sand

The model consists of 19 model parameters. 10 of these param-

eters describe the mechanical behaviour of the host sand (see
Table 1) and can be calibrated using the method proposed by Li
and Dafalias [33] and Xiao et al. [39].

Table 1
Model parameters for the host sand.

Parameter Symbol Value

Elastic moduli G0 300
m 0.05
Critical state M 1.18
eC 1.116
kc 0.129
Hardening rule h1 1.98
h2 2.19
n 0.8
Fig. 9. Calibration of the bonding parameters: (a) Peak strength envelopes of MHBS
Dilatancy d0 1.6
at different hydrate saturations and (b) variation in the initial bonding strength
m 1.3
with hydrate saturation (test data from Hyodo et al. [13]).
 J. Shen et al. / Computers and Geotechnics 75 (2016) 1–11
(a) Calibration of the elastic modulus parameter
As G0 and m have already been determined in the process of the
parameter calibration for the host sand, only n remains to be deter-
mined for MHBS. Combining Eqs. (6) and (7) yields
Gð1þeÞ
pffiffiffiffiffiffi  G0
ð2:97eÞ2 0
p pa
n¼ ð29Þ
Sh
Strictly speaking, independent small strain tests are needed to
determine the value of shear modulus G for MHBS. A relevant inves-
tigation of the stiffness for MHBS has been reported by Clayton et al.
[23,24]. However, owing to the different types of host sands used by
Hyodo et al. [13] and Clayton et al. [23], the test data provided by
Clayton are not suitable for the stiffness determination in this
paper. Thus, a simple method is adopted in which the initial shear
modulus G of MHBS at different effective confining pressures and
hydrate saturations is obtained in terms of the relationship between
the deviator stress q and the deviatoric strain eq with the axial strain
limited to 0.2%. Based on Eq. (29), an average value of n is adopted
for MHBS (as also shown in Fig. 2).
(b) Calibration of the critical state parameters
According to the previous analysis, the four critical state param-
eters (A, B, a, and b) can be determined by fitting the test data
obtained at the critical state under drained triaxial compression
conditions. As shown in Fig. 6, the parameters A and B can be
Fig. 10. Calibration of the dilatancy parameter mb: (a) ea–ev curve and (b) stress–
strain curve (test data from Hyodo et al. [13]).
7
obtained by fitting the trend of the variation in the critical stress
ratio with hydrate saturation. The other two critical state parame-
ters a and b can be determined similarly.
(c) Calibration of the bonding parameters
As we have clarified in the above section, triaxial tensile
strength is employed as the bonding strength to account for the
cementing habit of hydrates. Neglecting the small bonding break-
age before the peak strength state, the bonding strength can be
determined indirectly by back calculating from experimental data
on the peak strength of a soil [45,46]. As illustrated in Fig. 9(a),
the mean effective stress corresponding to zero deviator stress
in the peak strength envelope is the initial bonding strength. a
and b can be determined by fitting the trend of the variation in
the initial bonding strength with hydrate saturation (as shown
in Fig. 9(b)).
(d) Calibration of the debonding parameter
Under drained triaxial compression conditions, the following
relation holds neglecting the small elastic deformation:
dq dq
 ð30Þ
depq deq
Fig. 11. Comparison between the drained triaxial shear test results and model
simulations for the host sand with varying effective confining pressures (test data
from Hyodo et al. [13]).
8 J. Shen et al. / Computers and Geotechnics 75 (2016) 1–11

Substituting Eqs. (26), (27) and (30) into Eq. (11) yields: strain levels below 4%. According to Eq. (23), the dilatancy parame-
nwðSh Þ g
ter of the host sand ‘‘m” plays a dominant role in controlling the
hGe
g ½MðSh ÞenwðSh Þ  g   ddqeq ð1  3 Þ amount of dilatancy in the small strain regime. In this study, ‘‘m”
kd ¼ ð31Þ
pb g is assumed to be independent of the presence of hydrates, resulting
in an overestimation of dilatancy in the small strain regime. Fur-
The parameter kd can then be determined by curve-fitting the mea- thermore, the predicted volumetric strains underestimate the mea-
sured eq–q data for MHBS. sured values in the large strain regime. This may be because the
(e) Calibration of the dilatancy parameter calculated shear strength at the critical state under-predicts the
Ignoring the small elastic deformation, the dilatancy is given by experimental measurement as shown in Fig. 10(a), resulting in an
depv dev overestimation of the critical void ratio or an underestimation of
D¼  ð32Þ
depq deq the volumetric strain.

Then, by substituting Eqs. (24) and (32) into Eq. (23), mb can be
expressed as 4. Model validation
h i
dev
ln½d0 MðSh Þ þ mwðSh Þ  ln deq
MðSh Þ þ d0 g The experimental results used here for the model validation
mb ¼ pffiffiffiffiffiffiffiffiffiffiffiffi ð33Þ
pb =pa came from drained triaxial compression tests conducted by Hyodo
et al. [13], which are currently the most complete set of triaxial
Once all of the above parameters are determined, mb can be deter- tests published on MHBS. All the MHBS test samples were made
mined by curve-fitting the measured ea–ev data for MHBS obtained out of the host sand (Toyoura sand in this study) with methane
from drained triaxial compression tests. As shown in Fig. 10(a), hydrate formed by the ‘partial water saturation’ method. The initial
mb = 0.04 gives the best-fitted result for the measured axial–volu- porosity of the host sand was controlled to be approximately 40%
metric strain relationship. However, it can be found that the pre- and 45%, and the hydrate saturation was varied within the range
dicted volumetric strains overestimate the experimental data at of 0–60%. The specimens were then isotropically compressed at
effective confining pressures of 1 MPa, 3 MPa and 5 MPa under a
controlled temperature of 5 °C.

Fig. 12. Comparison between the drained triaxial shear test results and model Fig. 13. Comparison between the drained triaxial shear test results and model
simulations for the host sand with varying void ratios after isotropic compression simulations for MHBS with varying hydrate saturations (test data from Hyodo et al.
(test data from Hyodo et al. [13]). [13]).
 J. Shen et al. / Computers and Geotechnics 75 (2016) 1–11 9

As mentioned before, for MHBS with zero hydrate saturation

(i.e., the host sand), a series of drained triaxial compression tests
is also carried out by Hyodo et al. [13] for comparison with MHBS.
The experimental results are also used here to calibrate the 10
model parameters for the host sand. Based on the detailed calibra-
tion results (summarised in Table 1), model simulations of clean
host sand with different effective confining pressures and densities
are presented in Figs. 11 and 12, respectively. The eq–q and eq–ev
relations predicted by the model can well describe the overall
trend of the deformation process observed in the tests for the host
sand. However, the simulations tend to slightly underestimate the
peak strength of the host sand. The critical stress ratio M of Toy-
oura sand, obtained by fitting the test data provided by Hyodo
et al. [13] (as shown in Fig. 6), is equal to 1.18. The generally used
M value for Toyoura sand is 1.25 [33,47], which is slightly larger
than that used in this model. Nevertheless, the 10 calibrated
parameters for the host sand are reasonably acceptable and will
be used in the model simulation for MHBS.
Fig. 13 presents the simulation results for MHBS with varying
hydrate saturations at an effective confining pressure of 5 MPa.
The void ratios of the specimens after isotropic consolidation are
approximately 0.65. The increases in stiffness and peak strength,
as well as the enhanced dilatancy and strain softening, with
increasing hydrate saturation are more or less reproduced by our
model. The trend of the whole deformation process can be

Fig. 15. Comparison between the drained triaxial shear test results and model
simulations for MHBS with varying void ratios after isotropic compression (test
data from Hyodo et al. [13]).

captured with acceptable accuracy by the eq–q relations predicted

Fig. 14. Comparison between the drained triaxial shear test results and model
simulations for MHBS with varying effective confining pressures (test data from
Hyodo et al. [13]).
by our model. Meanwhile, the simulated eq–ev relations are in per-
fect agreement with the test results, indicating the effectiveness of
the modified state-dependent dilatancy in modelling the deforma-
tion of MHBS. Likewise in Fig. 14, the eq–q and eq–ev relations pre-
dicted by our model for MHBS with varying effective confining
pressures at a mean hydrate saturation of 53.7% and a void ratio
of 0.65 match well with the experimental results. Increasing stiff-
ness and peak strength, as well as weakened dilatancy and strain
softening, with increasing effective confining pressure are also
clearly shown. However, a major discrepancy exists between the
predicted volumetric strain and the experimental results at an
effective confining pressure of 3 MPa. At effective confining pres-
sures of 3 MPa and 5 MPa, the tested volumetric strains are nearly
identical. Hence, it is doubtful whether the volumetric strain at an
effective confining pressure of 3 MPa was accurately measured.
Fig. 15 shows the ability of our model to predict the mechanical
behaviour of MHBS with various void ratios after isotropic com-
pression using a single set of model parameters, which cannot be
achieved by existing models. It is observed that the dense MHBS
specimen with a low void ratio before shear displays slight
strain-softening and volume dilation behaviours, while the loose
specimen exhibits strain-hardening and compressive behaviours.
Our model is able to capture explicitly the trends in the changes
of the stress–strain and volume change behaviours with varying
densities. In general, the results predicted by our model are
10 J. Shen et al. / Computers and Geotechnics 75 (2016) 1–11
sufficiently close to the data from the drained triaxial tests, consid-
ering the complex mechanical behaviour of MHBS.
5. Conclusions
An elasto-plastic model based on the framework of state-
dependent dilatancy has been proposed in this paper to model
the effects of hydrate saturation, confining pressure and density
on the mechanical behaviour of MHBS. The influence of hydrate
accumulation habit on MHBS was thoroughly investigated and
modelled in terms of the three mechanisms of deformation for
MHBS based on the analysis of the experimental data from
Hyodo et al. [13]. The cementing mechanism of hydrate was
quantitatively described by a bonding strength parameter (pb).
The degradation of bonding was modelled as a function of plastic
deformation. The critical state behaviour, including CSLs in the
e  ln p0 plane and the critical stress ratio, was modified for MHBS
to account for the load-bearing mechanism of hydrates. By further
introducing the state-dependent dilatancy, the mechanical beha-
viour of MHBS with varying densities can be reproduced using a
unified set of model parameters. The increase in the peak strength
of MHBS over that of the host sand can be explained by the
increase in the critical stress ratio, existence of the bonding
strength and the increase in dilatancy.
All model parameters can be conveniently calibrated based on
conventional laboratory tests for clean host sand and MHBS. The
model simulations and test results from drained triaxial tests on
artificial MHBS were compared. It was demonstrated that the pro-
posed model can satisfactorily capture the stress–strain and vol-
ume change behaviours of MHBS over a wide range of hydrate
saturations, confining pressures and densities of interest.
Acknowledgements
The authors would like to acknowledge the financial supports
from GRF 617213 and M-HKUST603/13 provided by the Research
Grants Council (RGC) of HKSAR and FP204 by HKUST. Specialized
Research Fund (No. 20130094140001) for the Doctoral Program
of Higher Education (SRFDP) by the RGC, grant 2012CB719805
from the National Basic Research Program (973 Program) project
(No. 2012CB719800) provided by the Ministry of Science and Tech-
nology of the People’s Republic of China, grants 51378178 and
51578213 by the National Natural Science Foundation of China,
grant B13024 by the 111 Project of China and grant 2015B06014
by the Fundamental Research Funds for the Central Universities
of China are also acknowledged.
References
[1] Dallimore SR. Scientific results from the Mallik 2002 gas hydrate production
research well program, Mackenzie Delta, Northwest Territories,
Canada. Ottawa (Ontario, Canada): Geological Survey of Canada; 2005.
[2] Grauls D. Gas hydrates: importance and applications in petroleum exploration.
Mar Petrol Geol 2001;18(4):519–23.
[3] Holder GD, Kamath VA. Hydrates of (methane+ cis-2-butene) and (methane+
trans-2-butene). J Chem Thermodyn 1984;16(4):399–400.
[4] Ruppel C. Tapping methane hydrates for unconventional natural gas. Elements
2007;3(3):193–9.
[5] Collett TS. Energy resource potential of natural gas hydrates. AAPG Bull
2002;86(11):1971–92.
[6] Kvenvolden KA. Potential effects of gas hydrate on human welfare. Proc Natl
Acad Sci 1999;96(7):3420–6.
[7] Soga K, Lee SL, Ng MYA, Klar A. Characterisation and engineering properties of
methane hydrate soils. In: Tan TS, editor. Characterisation and engineering
properties of natural soils. London: Taylor and Francis; 2006. p. 2591–642.
[8] Jaysinghe A. Triaxial compression strength of methane hydrate-bearing
sediments. PhD thesis. Calgary (Canada): UC; 2013.
[9] Mienert J, Vanneste M, Bünz S, Andreassen K, Haflidason H, Sejrup HP. Ocean
warming and gas hydrate stability on the mid-Norwegian margin at the
Storegga Slide. Mar Petrol Geol 2005;22(1):233–44.
[10] Nixon MF, Grozic JLH. Submarine slope failure due to gas hydrate dissociation:
a preliminary quantification. Can Geotech J 2007;44(3):314–25.
[11] Hyodo M, Nakata Y, Yoshimoto N, Ebinuma T. Basic research on the
mechanical behavior of methane hydrate-sediments mixture. Soils Found
2005;45(1):75–85.
[12] Hyodo M, Li YH, Yoneda J, Nakata Y, Yoshimoto N, Nishimura A, et al.
Mechanical behavior of gas-saturated methane hydrate-bearing sediments. J
Geophys Res: Solid Earth 2013;118(10):5185–94.
[13] Hyodo M, Yoneda J, Yoshimoto N, Nakata Y. Mechanical and dissociation
properties of methane hydrate-bearing sand in deep seabed. Soils Found
2013;53(2):299–314.
[14] Masui A, Haneda H, Ogata Y, Aoki K. Effects of methane hydrate formation on
shear strength of synthetic methane hydrate sediments. In: Proceedings of the
15th international offshore and polar engineering conference; 2005. p. 364–
69.
[15] Masui A, Haneda H, Ogata Y, Aoki K. Triaxial compression test on submarine
sediment containing methane hydrate in deep sea off the coast off Japan. In:
Proceedings of the 41st annual conference. Japanese Geotechnical Society;
2006.
[16] Miyazaki K, Yamaguchi T, Sakamoto Y, Tenma N, Ogata Y, Aoki K, et al. Effect of
confining pressure on mechanical properties of artificial methane hydrate
bearing sediments. In: Proceedings of the 20th international offshore and polar
engineering conference; 2010. p. 132–7.
[17] Miyazaki K, Masui A, Sakamoto Y, Aoki K, Tenma N, Yamaguchi T. Triaxial
compressive properties of artificial methane-hydrate-bearing sediment. J
Geophys Res: Solid Earth 2011;116(B6). 1978–2012.
[18] Winters WJ, Waite WF, Mason DH, Dillion WP, Pecher IA. Sediment properties
associated with gas hydrate formation. In: Proceedings of the 4th international
conference on gas hydrates; 2002. p. 19–23.
[19] Waite WF, Santamarina JC, Cortes DD, Dugan B, Espinoza DN, Germaine J, et al.
Physical properties of hydrate-bearing sediments. Rev Geophys 2009;47(4).
[20] Yun TS, Santamarina JC, Ruppel C. Mechanical properties of sand, silt, and clay
containing tetrahydrofuran hydrate. J Geophys Res: Solid Earth 2007;112(B4).
1978–2012.
[21] Helgerud MB, Dvorkin J, Nur A, Sakai A, Collett T. Elastic-wave velocity in
marine sediments with gas hydrates: effective medium modeling. Geophys
Res Lett 1999;26(13):2021–4.
[22] Dvorkin J, Prasad M, Sakai A, Lavoie D. Elasticity of marine sediments: rock
physics modeling. Geophys Res Lett 1999;26(12):1781–4.
[23] Clayton CRI, Priest JA, Best AI. The effects of disseminated methane hydrate on
the dynamic stiffness and damping of a sand. Geotechnique 2005;55
(6):423–34.
[24] Clayton CRI, Priest JA, Rees EVL. The effects of hydrate cement on the stiffness
of some sands. Geotechnique 2010;60(6):435–45.
[25] Freij-Ayoub R, Tan C, Clennell B, Tohidi B, Yang JH. A wellbore stability model
for hydrate bearing sediments. J Petrol Sci Eng 2007;57(1):209–20.
[26] Rutqvist J, Moridis GJ. Numerical studies on the geomechanical stability of
hydrate-bearing sediments. Berkeley (CA, US): Ernest Orlando Lawrence
Berkeley National Laboratory; 2007.
[27] Klar A, Soga K, Ng MYA. Coupled deformation–flow analysis for methane
hydrate extraction. Geotechnique 2010;60(10):765–76.
[28] Uchida S, Soga K, Yamamoto K. Critical state soil constitutive model for
methane hydrate soil. J Geophys Res: Solid Earth 2012;117(B3). 1978–2012.
[29] Lin JS, Seol Y, Choi JH. An SMP critical state model for methane hydrate-bearing
sands. Int J Numer Anal Meth Geomech 2015;39(9):969–87.
[30] Richart FE, Hall JR, Woods RD. Vibrations of soils and foundations. Englewood
Cliffs (New Jersey): Prentice-Hall Inc.; 1970.
[31] Manzari MT, Dafalias YF. A critical state two-surface plasticity model for sands.
Geotechnique 1997;47(2):255–72.
[32] Li XS, Dafalias YF, Wang ZL. State-dependent dilatancy in critical-state
constitutive modelling of sand. Can Geotech J 1999;36(4):599–611.
[33] Li XS, Dafalias YF. Dilatancy for cohesionless soils. Geotechnique 2000;50
(4):449–60.
[34] Chiu CF, Ng CWW. A state-dependent elasto-plastic model for saturated and
unsaturated soils. Geotechnique 2003;53(9):809–29.
[35] Dafalias YF. An anisotropic critical state soil plasticity model. Mech Res
Commun 1986;13(6):341–7.
[36] Verdugo R, Ishihara K. The steady state of sandy soils. Soils Found 1996;36
(2):81–91.
[37] Rowe PW. The stress–dilatancy relation for static equilibrium of an assembly
of particles in contact. Proceedings of the royal society of London a:
mathematical, physical and engineering sciences, vol. 269(1339). The Royal
Society; 1962. p. 500–27.
[38] Cubrinovski M, Ishihara K. Modelling of sand behaviour based on state
concept. Soils Found 1998;38(3):115–27.
[39] Xiao Y, Liu H, Chen Y, Jiang J, Zhang W. Testing and modeling of the state-
dependent behaviors of rockfill material. Comput Geotech 2014;61:153–65.
[40] Been K, Jefferies MG. A state parameter for sands. Geotechnique 1985;35
(2):99–112.
[41] Gao Z, Zhao J. Constitutive modeling of artificially cemented sand by
considering fabric anisotropy. Comput Geotech 2012;41:57–69.
[42] Abdulla AA, Kiousis PD. Behavior of cemented sands–I. Testing. Int J Numer
Anal Meth Geomech 1997;21(8):533–47.
[43] Consoli NC, Viana da Fonseca A, Cruz RC, Heineck K. Fundamental parameters
for the stiffness and strength control of artificially cemented sand. J Geotech
Geoenviron Eng 2009;135(9):1347–53.
 J. Shen et al. / Computers and Geotechnics 75 (2016) 1–11 11

[44] Consoli NC, Cruz RC, da Fonseca A, Coop M. Influence of cement-voids ratio on
stress–dilatancy behavior of artificially cemented sand. J Geotech Geoenviron
Eng 2011;138(1):100–9.
[45] Gao Z, Zhao J, Yao Y. A generalized anisotropic failure criterion for
geomaterials. Int J Solids Struct 2010;47(22):3166–85.
[46] Lade PV. Three-parameter failure criterion for concrete. J Eng Mech Div
1982;108(5):850–63.
[47] Ling HI, Yang S. Unified sand model based on the critical state and generalized
plasticity. J Eng Mech 2006;132(12):1380–91.