Available online at www.sciencedirect.

com

ScienceDirect
International Journal of Pavement Research and Technology 9 (2016) 445449
www.elsevier.com/locate/IJPRT

A fuzzy set plasticity model for cyclic loading of granular soils

Cheng Chen a,, Lingwei Kong b, Xiaoqing Liu a, Zhonghui Han a
a
College of Civil Engineering and Mechanics, Xiangtan University, Xiangtan 411105, China
b
State Key Laboratory of Geomechanics and Geotechnical Engineering, Institute of Rock and Soil Mechanics, Chinese Academy of Sciences, Wuhan
430071, China

Received 8 March 2016; received in revised form 26 August 2016; accepted 16 September 2016
Available online 10 October 2016

Abstract

A constitutive model for describing the stressstrain behavior of granular soils subjected to cyclic loading is presented. The model is
formulated using fuzzy set plasticity theory within the classical incremental plasticity theory framework. A special membership function
is introduced to provide an analytical and simple geometrical interpretation to formulate hardening, hysteresis feature, material memory,
and kinematic mechanisms without resorting to complicated kinematic hardening formulations. The model can accurately describe cyclic
loading, dilatancy, material theory and critical state soil mechanics features eects. Two series of cyclic drained triaxial tests data are
considered. The characteristic features of behavior in granular soils subjected to cyclic loading are captured.
2016 Production and hosting by Elsevier B.V. on behalf of Chinese Society of Pavement Engineering. This is an open access article under the CC
BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords: Constitutive model; Fuzzy set; Plasticity; Cyclic loading; Hardening hysteresis feature

1. Introduction ties associated with topics such as earthquake engineering,

Cyclic response of granular materials is complex due to
the pressure and specic volume dependency of the stress
strain relationship and the highly nonlinear behavior of the
soil matrix. Until now, the mechanical behavior of granu-
lar soils has been mainly represented with constitutive
models which need dierent sets of constitutive parameters
for each density and eective conning pressure. In fact,
the study of loading and unloading response in granular
soils and development of relationships for its prediction
in natural formations and engineered materials has been
a major area of research in modern geomechanics. Con-
certed eort has been made to develop predictive capabili-
Corresponding author.
E-mail addresses: chencheng401@sina.com (C. Chen),
lwkong@whrsm.ac.cn (L. Kong), a8583272@qq.com (X. Liu),
1427687676@qq.com (Z. Han).
Peer review under responsibility of Chinese Society of Pavement
Engineering.
soil-structure interaction, soil liquefaction, o-shore engi-
neering, etc.
Development of constitutive models for a wide range of
engineering materials, including soils, has been found
extensively for recent decades [115]. A majority of the
models is based on the incremental plasticity theory.
Within the framework of classical plasticity theory, isotro-
pic hardening has been proved sucient to simulate the
stressstrain response of soil subjected to monotonic load-
ing while kinematic hardening and mixed hardening has
been typically used to mimic hysteretic phenomena of soil
under cyclic loading.
Nowadays, the cyclic behavior of unbound granular
materials under trac loading is another challenging task
for geotechnical engineers. A typical example is railroad
ballast. Thus, it is of special interest to determine the over
characteristics and constitutive properties of the ballast
and to ensure stable and long-lasting properties for such
a material that is not homogeneous. The research focuses
on the development of a cyclic constitutive model based

http://dx.doi.org/10.1016/j.ijprt.2016.09.004
1996-6814/ 2016 Production and hosting by Elsevier B.V. on behalf of Chinese Society of Pavement Engineering.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
446 C. Chen et al. / International Journal of Pavement Research and Technology 9 (2016) 445449

on fuzzy set concepts, its numerical integration, and nite
element implementation as well.
Unlike convention elasto-plastic hardening models, the
fuzzy set model is physically intuitive and easy to visualize.
It provides analytical and simple geometrical interpreta-
tions to formulate hardening, hysteresis features, material
memory, and kinematic mechanisms. In this model, based
on fuzzy set plasticity theory, the basic concept rests on the
assumption that there exists a fuzzy surface which in many
ways resemble a bounding surface. At each point within the
fuzzy surface, the value of plastic hardening moduli is
dened by the membership function. In this view, Bao
et al. presented a transparent and accurate kinematic-
cyclic constitutive model to capture the important features
of volume change and pore water pressure build-up related
to soil cyclic mobility [16].
In this study, a cyclic plasticity model based on fuzzy
plasticity theory is presented to model the cyclic behavior
of unbound granular materials under repeated loads. The
enhanced fuzzy-set model is built to adapt the simply for-
mat equations of plastic moduli and plastic potential to
simulated the pavement materials deformation problems
particularly related to cyclic mobility. Two series of cyclic
drained triaxial tests data are considered. The characteris-
tic features of behavior in granular soils subjected to cyclic
loading are captured.
2. Preliminaries
2.1. Notation
In the model presented, the material behavior is
assumed isotropic and rate independent in both elastic
nad elasticplastic response. Compression is considered
positive and tension is negative. For simplicity, triaxial
stress notation p0  q is adopted throughout;
0 0 0
p r1 2r3 =3 is the mean eective stress and
q r01  r03 is the deviator stress, where r01 and r03 are the
axial and radial stresses, respectively. The corresponding
work conjugates are volumetric strain ev e1 2e3 and
1 ep0 31  2m 1 ep0
K ; G 2
j 21 m j
where e is the void ratio and m is the Possions ratio.
2.3. Membership function
The membership function has been involved in the plas-
tic modulus equations. When c = 1, the material behaves
purely elastically and the corresponding value of the plastic
modulus is innite, while when c 0, the material reaches
a fully plastic state and the plastic modulus is equal to the
value on the fuzzy surface, i.e. H = H*.
With the assistance of the membership function c, we
can readily construct reversal plastic loading without
resorting to a kinematic hardening rule. The basic rules
of kinematic mechanism for the membership function are:
r Plastic loading: c_ < 0
r Plastic unloading: c_ < 0
r Elastic loading: c_ P 0
r Elastic unloading: c_ P 0
Although the value of the membership function is 1 at a
fully elastic state and 0 at the fully plastic state, the assign-
ment of the value in elastoplastic state is deterministic and
can be arbitrarily dened as needed. A linear variation with
respect to stress state was adopted in this study.
Fig. 1 displays an example of the deviatoric stressstrain
response and evolution of the membership function for a
material subjected to two varied amplitude cyclic loading
under a conventional triaxial stress path. The unloading
reloading points take place in two dierent stress levels
q 156 kPa and q 231 kPa, respectively. The two
graphs on the left highlight cycle 1 with the loading from
0 to 156 kPa and unloading from 156 to 0 kPa (in solid
line). The other two graphs highlight the cycle 2 with the
300 300
cycle 1 cycle 2

deviatoric strain ed 23 e1  e3 . The pairs of stresses and 200 200

q(kPa)
q(kPa)

strains are abbreviated in the vector form as r01 p0 ; q

T

T 100 100
and  ev ; ed  . The total strain rate is decomposed into
elastic and plastic parts according to 0 0
0 0.02 0.04 0.06 0 0.02 0.04 0.06
e p
_ _ _ 1 d
d

where a superimposed dot indicates an increment, and the 1 1

superscripts e and p denote the elastic and plastic compo- 0.8 0.8
nents, respectively. 0.6 0.6

0.4 0.4
2.2. Elastic behavior 0.2 0.2
cycle 1 cycle 2
0 0
The tangential elastic moduli are calculated assuming 0 50 100 150 200 250 0 50 100 150 200 250

that the slope of unloading/reloading occurs along a j line q q

in the e  lnp0 plane. The moduli are then dened as Fig. 1. Deviatoric stressstrain curve and evolution of the membership
function c for cycle 1 and cycle 2.
 C. Chen et al. / International Journal of Pavement Research and Technology 9 (2016) 445449 447

loading from 0 to 231 kPa and unloading from 231 to where A and B are constitutive parameters. If A B 1 in
0 kPa (in solid line). Eq. (6), the ow rule is equivalent to the ow rule of orig-
inal Cam Clay model. The terms a1 and b1 of Eq. (6) rep-
2.4. Cone fuzzy surface resent the slope of the fuzzy set surface line both on the
compression side and the extension side.
F ccone r  a0  a1 p  a2 p2 0 Plastic ow rules are expressed as
F econe r b0 b1 p b2 p2 0 3
_ p k_ cone m 8
where r and p are stress invariants, and r q  gh; v,
p I 1 =3. The coecient a1 in the cone fuzzy surface func-
In the fuzzy set plasticity theory, a forth-order tensor is
tion for triaxial compression incorporates the concept of
dened in such a way that m  T : n, where the fourth-
critical state soil mechanics, which is dened as:
order tensor T is dened as
a1 M c jhwi. For loose sand, a1  M c .
Since the critical state line in extension mean stress space 1
is not well dened and dicult to obtain experimentally, so T I  1  D1  1 9
3
it is reasonably assumed that the evolution of coecient b1
in the cone fuzzy surface for triaxial extension is attained
by keeping the ellipticity v xed, i.e. 1.5
b1 va1 4
1

2.5. Cone plastic moduli

0.5
It should be noted that cone loading surfaces are not
explicitly dened, and one can think that for the current
0

stress state, there exist cone loading surfaces such that

the cone plastic moduli is dened as follows.
-0.5
Mcd
H H 5
1  cd1
-1
d and M are model parameters that can be determined Laboratory Test
from test data. It is worth mentioning that cone member- Numerical Modeling
-1.5
ship functions have been involved in the cone plastic mod- -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04
d
ulus equations. And the value of the cone membership
function is determined by the second deviatoric stress Fig. 2. Deviatoric strain vs stress ratio for drained cyclic test on loose Fuji
invariant and the rst stress invariant respectively. river sand.

2.6. Plastic potential

The plastic potential denes the ratio between the incre- 0.04

mental plastic volumetric strain and the incremental plastic

0.035
shear strain. The most successful and widely used ow rule
in geotechnical engineering is based on Rowes stress- 0.03
dilatancy relationship [17]. Rowes original relationship is
modied here to take into account the dependence of dila- 0.025

tancy on the state parameter. This approach was also used

0.02
v

by Gajo and Muir Wood [9] and is given by

@epv 0.015
D AM c 1 k d w  g 6
@epd
0.01
For loose granular materials, we dened the dilatancy
0.005
ow rule as Laboratory Test
Numerical Modeling
@epv 0
Dc Aa1 B  g in triaxial compression -1.5 -1 -0.5 0 0.5 1 1.5
@epd

@epv Fig. 3. Volumetric strain vs stress ratio for drained cyclic test on loose
De Ab1 B  g in triaxial extension 7
@epd Fuji river sand.
448 C. Chen et al. / International Journal of Pavement Research and Technology 9 (2016) 445449

0.04 0.03
Laboratory Test
0.035 Numerical Modeling
0.025

0.03

0.02
0.025

0.02 0.015
v

v
0.015
0.01

0.01

0.005
0.005
Laboratory Test
Numerical Modeling
0 0
-0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 -1 -0.5 0 0.5 1 1.5
d

Fig. 4. Deviatoric strain vs volumetric strain for drained cyclic test on Fig. 6. Volumetric strain vs stress ratio for drained cyclic constant p0 test
loose Fuji river sand. on loose Toyoura sand.

0.03
1.5 Laboratory Test
Numerical Modeling
0.025

0.02

0.5
0.015
v

0
0.01

-0.5
0.005

Laboratory Test
-1 0
Numerical Modeling -0.025 -0.02 -0.015 -0.01 -0.005 0 0.005 0.01 0.015 0.02 0.025
-0.025 -0.02 -0.015 -0.01 -0.005 0 0.005 0.01 0.015 0.02 0.025 d
d
Fig. 7. Deviatoric strain vs volumetric strain for drained cyclic constant p0
0 test on loose Toyoura sand.
Fig. 5. Deviatoric strain vs stress ratio for drained cyclic constant p test
on loose Toyoura sand.

3. Model validation
Incremental stress-controlled formulation

Performance of the proposed model to simulate the
1 1 measured behavior of sands under cyclic loading was inves-
_ _ _ r_ D : r_
e p
n : r_ m De : r_ m  n :
e
H H tigated by comparing numerical simulations with experi-
  mental results from the literature. The results of cyclic
1
r_ De m  n : r_ Dep : r_ 10 loading are shown in Figs. 27. In all of these gures,
H model predictions are shown using continuous solid lines



_ v ep p_ and experimental data are shown by discrete symbols.
D 11
_ d q_
3.1. Drained cyclic test on Fuji river sand
where
  Figs. 24 show the simulation results for the cyclic
ep 1=K 0
D drained test conducted by Tatsuoka and Ishihara [18] on
0 1=3G loose Fuji river sand with increasing stress amplitude.
 2  12
1 1 a1 D a1 gD The initial conditions of the test were: p0 196 kPa and
q
 
H a2 g2 a2 D2 g2 a1 g g2 e 0:74. The basic material parameters were: j 0:001,
1 1 m 0:3, a1 M c 1:48, b1 va1 M e 1:08,
 C. Chen et al. / International Journal of Pavement Research and Technology 9 (2016) 445449
v 0:73, k 0:21, a 0:8. The plastic parameters were:
M 10; 450, d 2:48, a0 110, b0 100, A 0:17,
B 1:05. Once again, the model simulation matches the
experimentally observed trends. The model captures the
contractive responses both during loading and unloading,
and the successive stiening of the sample with the progress
of cyclic loading.
3.2. Dranied cyclic test on Toyoura sand
Pradhan et al. [19] executed a series of drained cyclic
tests on Toyoura sand which consisted of mainly quartz
with angular to sub-angular particle shape. The triaxial
tests were conducted on 75 mm diameter and 150 mm
high specimens prepared by pulviating air dried samples.
Figs. 57 show results of cyclic, constant p0 , drained tests
on loose samples of Toyoura sand. The initial conditions
of the test were: p0 98 kPa and e 0:845. The basic
material parameters were: j 0:001, m 0:3,
a1 M c 1:24, b1 va1 M e 1:08, v 0:87,
k 0:24, a 0:8. The plastic parameters were: M 8450,
d 2:8, a0 120, b0 60, A 0:28, B 1:05. These
tests are also well matched by the model simulation. The
model captures the stressstrain response and the succes-
sive stiening or softening of the sample with the progress
of the cyclic loading.
4. Conclusions
A cyclic constitutive model based on fuzzy set concepts
has been developed. The cyclic fuzzy set model is physically
intuitive and easy to visualize with the aid of membership
functions. The cyclic fuzzy set model provides analytical
and simple geometrical interpretation to formulate harden-
ing, hysteresis features, materials memory, and kinematic
mechanisms without invoking complex analytical formula-
tions. In addition the cyclic fuzzy set model accounts for:
realistic stressstrain behavior under repeated load cycles,
nonlinear dilatancy behavior, critical state soil mechanics
concepts, and non-proportional loading. The evolution
rule for the fuzzy surface can help simulate the post peak
soil behavior such as strain softening. The critical state soil
mechanics concept has been implemented into the fuzzy set
model by linking the fuzzy surface parameter a1 to the state
parameter w.
Acknowledgments
The support from the State Key Laboratory of Geome-
chanics and Geotechnical engineering is greatly appreci-
ated. The nancial support from the National Natural
449
Science Foundation of China (No: 41372314) and Major
Subject of The Chinese Academy of Sciences (No:
KZZD-EW-05-02) is also acknowledged.
References
[1] J.H. Prevost, Plasticity theory for soil stress behavior, J. Eng. Mech.,
ASCE 104 (5) (1978) 11771194.
[2] Z. Mroz, V.A. Norris, O.C. Zienkiewicz, An anisotropic hardening
model for soils and its application to cyclic loading, Int. J. Numer.
Anal. Meth. Geomech. 2 (1978) 203221.
[3] S. Sture, J.C. Mould, H.-Y. Ko, Elastic-plastic anisotropic hardening
constitutive model and prediction of behavior of dry quartz sand, in:
G. Gudenhus, F. Darve, I. Vardoulakis (Eds.), Constitutive Relations
for Soils, A. A. Balkema, Rotterdam, Rotterdam, Netherlands, 1982,
pp. 227248.
[4] J.P. Bardet, Bounding surface plasticity model for sands, J. Eng.
Mech. 112 (11) (1986) 11981217.
[5] P. Anandarajah, A granular material model based on associated ow
rule applied to monotonic loading behavior, Soils Found. 34 (3)
(1994) 8198.
[6] R.S. Crouch, J.P. Wolf, Y.F. Dafalias, Unied critical-state bound-
ing-surface plasticity model for soil, J. Eng. Mech. 120 (11) (1994)
22512270.
[7] M.T. Manzari, Y.F. Dafalias, A critical state two-surface plasticity
model for sands, Geotechnique 47 (2) (1997) 255272.
[8] A. Gajo, D.M. Wood, A kinematic hardening constitutive model for
sands: the multiaxial formulation, Int. J. Numer. Anal. Meth.
Geomech. 23 (1999) 925965.
[9] A. Gajo, D.M. Wood, Severn-trent sand: a kinematic-hardening
constitutive model: the p-q formulation, Geotechnique 49 (5) (1999)
595614.
[10] X.S. Li, Y.F. Dafalias, Constitutive modeling of inherently aniso-
tropic sand behavior, J. Geotech. Geoenviron. Eng. 128 (10) (2002)
868880.
[11] C. Chen, L. Ge, J.S. Zhang, Modeling permanent deformation of
unbound granular materials under repeated loads, Int. J. Geomech.
10 (6) (2010) 236241.
[12] C. Chen, J.S. Zhang, Constitutive modeling of loose sands under
various stress paths, Int. J. Geomech. 13 (1) (2013) 18.
[13] A. Elgamal, Z. Yang, E. Parra, A. Ragheb, Modeling of cyclic
mobility in saturated cohesionless soils, Int. J. Plast 19 (6) (2003) 883
905.
[14] X.S. Li, Y.F. Dafalias, A constitutive framework for anisotropic sand
including non-proportional loading, Geotechnique 54 (1) (2004) 41
55.
[15] N. Khalili, M.A. Habte, S.A. Valliappan, Bounding surface plasticity
model for cyclic loading of granular soils, Int. J. Numer. Meth. Eng.
63 (14) (2005) 19391960.
[16] Y. Bao, S. Sture, Numerical modeling of cyclic mobility based on
fuzzy-set concepts in plasticity theory, Comput. Geotech. 38 (3)
(2011) 375382.
[17] P.W. Rowe, The stress-dilatancy relations for static equilibrium of an
assembly of particles in contact, Proc. R. Soc. London, Ser. A 269
(1962) 500527.
[18] F. Tatsuoka, K. Ishihara, Drained deformation of sand under cyclic
stresses reversing direction, Soils Found. 14 (3) (1974) 5165.
[19] T.B.S. Pradhan, F. Tatsuoka, Y. Sato, Experimental stress-dilatancy
relations of sand subjected to cyclic loading 29 (1) (1989) 4564.
Copyright of International Journal of Pavement Research & Technology is the property of
Chinese Society of Pavement Engineering and its content may not be copied or emailed to
multiple sites or posted to a listserv without the copyright holder's express written permission.
However, users may print, download, or email articles for individual use.