Soil Dynamics and Earthquake Engineering 26 (2006) 231–252

www.elsevier.com/locate/soildyn

Practical applications of a nonlinear approach to analysis

of earthquake-induced liquefaction and deformation of earth structures
Zhi-Liang Wang*, Faiz I. Makdisi, John Egan
Geomatrix Consultants, Inc., 2101 Webster Street, 12th Floor, Oakland, CA 94612, USA

Accepted 11 November 2004

Abstract
Seismic stability, liquefaction, and deformation of earth structures are critical issues in geotechnical earthquake engineering practice. At
present, the equivalent linear approach is considered the ‘state of practice’ in common use. More recently, dynamic analyses incorporating
nonlinear, effective-stress-based soil models have been used more frequently in engineering applications. This paper describes a bounding
surface hypoplasticity model for sand [Wang ZL. Bounding surface hypoplasticity model for granular soils and its applications. PhD
Dissertation for the University of California at Davis, U.M.I. Dissertation Information Service, Order No. 9110679; 1990; Wang ZL, Dafalias
YF, Shen CK. Bounding surface hypoplasticity model for sand. ASCE, J Eng Mech 1990;116(5):983–1001; Wang ZL, Makdisi FI.
Implementing a bounding surface hypoplasticity model for sand into the FLAC program. In: Proceedings of the international symposium on
numerical modeling in geomechanics. Minnesota, USA; 1999. p. 483–90] incorporated into a two-dimensional finite difference analysis
program [Itasca Consulting Group, Inc. FLAC (Fast Lagrangian Analysis of Continua), Version 4. Minneapolis, MN; 2000] to perform
nonlinear, effective-stress analyses of soil structures. The soil properties needed to support such analyses are generally similar to those
currently used for equivalent linear and approximate effective-stress analyses. The advantages of using a nonlinear approach are illustrated
by comparison with results from the equivalent linear approach for a rockfill dam. The earthquake performance of a waterfront slope and an
earth dam were evaluated to demonstrate the model’s ability to simulate pore-pressure generation and liquefaction in cohesionless soils.
q 2005 Elsevier Ltd. All rights reserved.

Keywords: Stability of dams; Earthquake performance; Nonlinearity of soils; Plasticity model; Liquefaction

1. Introduction past earthquakes; (d) by comparing the dynamic-induced

The state-of-practice for evaluating the seismic stability
of earth dams was developed in the early and mid-1970s by
the late Professor Seed and co-workers at the University of
California at Berkeley [5]. The basic elements of this
analysis included the following steps: (a) earthquake ground
motions are estimated at bedrock underlying the dam and its
soil foundation; (b) the response of the embankment to the
base rock excitation is computed to estimate the dynamic
stresses induced in representative elements of the embank-
ment; (c) the cyclic strength of the embankment soils is
evaluated using in situ tests and liquefaction resistance
correlation curves based on observed performance during
* Corresponding author. Fax: C1 510 663 4141.
E-mail address: zlwang@geomatrix.com (Z.-L. Wang).
0267-7261/$ - see front matter q 2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.soildyn.2004.11.032
shear stresses to the cyclic strength, the potential for
liquefaction of the embankment and foundation soils is
estimated; (e) for the zones of the embankment that are
determined to have liquefied during the earthquake, a
residual strength is assigned based on the density of the soil;
(f) the stability of the embankment and its foundation is
evaluated using limit equilibrium stability analyses; if the
embankment is found to be stable, earthquake-induced
permanent displacements are estimated using Newmark-
type deformation analyses. In this procedure, the dynamic
response of an embankment is estimated using an iterative
equivalent linear approach [6] to model the nonlinear strain-
dependent modulus and damping properties of the soils.
With this approach, however, the computed seismic
response and shear stresses within embankment soils do
not reflect the effects of induced pore pressure during
earthquake shaking. The amount of induced pore pressure
and the potential for liquefaction are estimated at the end of
the specified duration of shaking.
232 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252

Nomenclature
a model parameter m state-dependent index
b model parameter M critical stress ratio in pKq space
A constant Mp phase transformation stress ratio in pKq space
c1,c2,c3,c4 constants (nD)ij, (nN)ij unit tensor indicating the direction of the
C(x) modulus degradation function deviatoric plastic strain increment, and devia-
d dilatancy toric unit loading, respectively
d3ij, d3eij , d3pij total, elastic, and plastic deviatoric strain p mean normal stress
increment pc mean normal stress at critical state
d3v, d3ev , d3pv total, elastic, and plastic volumetric strain pm maximum mean normal stress
increment pa atmospheric pressure
d3vd, d3pd shear-induced volumetric strain and plastic q deviatoric stress in triaxial space
strain increment rij, rij stress ratio and image stress ratio, respectively
dp increament of mean stress R, Rf, Rp, Rm, stress ratio invariants for loading, failure,
drkl increament of deviatoric stress ratio tensor phase transformation, and maximum,
dsij increament of deviatoric stress tensor respectively
e, ec void ratio, critical void ratio Rp stress ratio invariants for dilatancy
eij, eeij , epij deviatoric strain, elastic and plastic deviatoric sij deviatoric stress tensor
strains, respectively V(e) a function of void ratio
g gravity w a function controls shear-induced volume
g(q) shape function on pZconst plane changes [1,2]
Gmax elastic shear modulus a model parameter
Go model parameter aij projection center for rij
G secant modulus dij Kronecker delta
hr model parameter 3v, 3vd total and shear-induced volumetric strain
h(x) heavyside step function g engineering shear strain
Hr, Hp plastic shear modulus for drij and dp mechan- ga shear strain amplitude
isms, respectively k model parameter
Ip state pressure index r, r distances from projection center
J second deviatoric stress invariant svo overburden pressure
kr model parameter t shear stress
K elastic bulk modulus tm maximum shear stress
K0 coefficient of lateral earth pressure at rest h stress ratio in triaxial space
Kr, Kp plastic bulk modulus for drij and dp mechan- x accumulated plastic deviatoric strain
isms, respectively

In cases where an earth structure is subjected to severe
ground motions, and where liquefaction may occur soon
during shaking, an alternative approach to account for the
effects of buildup of pore pressure and the potential for
liquefaction during earthquake shaking is to use a nonlinear
effective-stress analysis. This paper describes a nonlinear,
fully coupled dynamic analysis procedure for estimating the
potential for buildup of pore pressure, the potential for
liquefaction, and the resulting permanent deformation of
earth structures. The basic elements for such analyses are:
(1) an estimated site ground motion; (2) a constitutive model
to simulate soil behavior under conditions of cyclic loading
and liquefaction; (3) a computer program capable of
performing dynamic analyses that are fully coupled (that
include mechanical aspects and groundwater flow); and (4)
relevant laboratory and in situ measurements of soil
properties.
The procedure uses a nonlinear, bounding surface
plasticity constitutive model for sand [1,2], incorporated
into the finite-difference computer program FLAC [4].
Because the basis for such an analysis is a cyclic plasticity
soil model, we will briefly introduce this model before
shifting our focus to applications. Recent improvements to
the model include: (1) a newly proposed state parameter, the
state pressure index, and its use in defining a dilatancy
curve; (2) simulation of the critical-state behavior of sands;
and (3) simulation of post-liquefaction deformation of
sands. These new formulations are verified by comparing
model simulations to laboratory test results.
This paper also introduces our current practice of using
this bounding surface plasticity model, as incorporated into
the computer program FLAC, to perform nonlinear, fully
coupled dynamic analyses. The procedures for using and
calibrating the model also are explained. The soil tests to
 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252 233

support such analyses are generally similar to those

J
currently used for equivalent linear and simplified effec-
tive-stress analyses. The advantages of using a nonlinear
approach are illustrated by comparing its results with those Line of Phase Failure Surface
from an equivalent linear analysis for a rockfill dam. The Transformation R-Rf=0
earthquake performance of a waterfront slope and an earth R-Rp=0
dam are also presented to demonstrate the model’s ability to
simulate pore-pressure generation and liquefaction in
cohesionless soils. Given the increasing availability of Maximum
Pre-stress Surface
high-speed personal computers, the newly formulated R-Rm=0
analytical methods are becoming more practical and
attractive despite their theoretical complexity. Loading
dσij

2. Bounding surface hypoplasticity model for sand Unloading

dσij

We will briefly introduce the basic features of the model

p
and its capability to simulate sand behavior during
monotonic and cyclic loading under drained and undrained Fig. 1. Surfaces and stress variables in JKp plane.
conditions, focusing on the prediction of liquefaction. Then,
we examine its advantages compared with using existing step function h(pKpm) and the Macauley brackets hdpi
models like the Mohr–Coulomb or hyperbolic stress–strain indicate that the plastic mechanism due to dp operates only
relations. when pZpm and dpO0. In a simplified two-dimensional
version of the model [3], the flat-cap is omitted, because
under undrained conditions mean pressure often decreases
2.1. Basic formulations
and the flat-cap is inactive. Be cautious to use this in a
settlement calculation, where consolidation may occur, in
The general analytical formulation of the bounding
such case pZpm and dpO0, so that the flat-cap is needed.
surface hypoplasticity model for sand [1,2] was written in
The elastic shear and bulk moduli are functions of p
terms of tensors of stress rate and strain rate. The elastic and
plastic strain rates are given as rffiffiffiffiffi
p
Gmax Z Go pa VðeÞ (3)
dsij dpdij pa
d3eij Z C (1)
2Gmax 3K in which


p ðnD Þij dij ð2:973 K eÞ2
d3ij Z C pðdrkl ðnN Þkl Þ VðeÞ Z
Hr 3Kr 1 Ce

 rffiffiffiffiffi
rij dij 1 Ce p
C C hðp K pm Þhdpi (2) K Z pa (4)
Hp 3Kp k pa
Gmax and K are the incremental elastic shear and bulk in which Go and k are model parameters that define the
moduli, respectively. Hr and Kr are the plastic shear and elastic moduli, e is void ratio, and pa is atmospheric
bulk moduli, respectively, associated with the tensor of the pressure.
stress ratio increment drij. Similarly, Hp and Kp are the
plastic shear and bulk moduli, respectively, associated with
the mean pressure increment dp. nD is a unit vector in stress
space along the direction of the deviatoric plastic strain
increment, and nN is the associated direction of deviatoric
unit loading. dij is a unit tensor along the p axis. Using J2Z
0.5sijsij and RZJ/p, the failure surface (RKRfg(q)Z0)
and the loading surface (RKRmg(q)Z0), as shown in Figs. 1
and 2, are defined in the stress space JKp with a shape
function g(q) on pZconstant plane (q is the Lode angle). A
flat-cap surface, pKpmZ0, is proposed for simulating
consolidation and rebound. The pm is the maximum value of
p experienced in the past loading history. The heavyside Fig. 2. Surfaces and stress variables in pZconstant plane.
234 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252
The plastic modulus Hr for monotonic and cyclic loading
is

m 
Rf r
Hr Z Gmax hr CðxÞ K1
Rm r
where RZRm is the maximum pre-stress bounding surface,
and RZRf is the failure bounding surface. r and r are
distances from the projection center (as shown in Fig. 2) to
the current stress and maximum pre-stress surfaces,
respectively. Index mZ ð1C gðqÞÞRm =r is state-dependent
for irregular cyclic loading. For stress-controlled cyclic
loading, the shear stress-shear strain relations derived based
on Eqs. (3) and (5) conform to the well-known Masing rules.
C(x) is a function of a hardening/softening measure; e.g. the
accumulated plastic strain, volumetric and/or deviatoric.
C(x) increases (during hardening) or decreases (during
softening) with x.
The plastic volumetric strain rate is controlled by the
plastic bulk modulus, Kr:
rffiffiffiffiffi
1 Ce p years, a PEER (Pacific Earthquake Engineering Research)
K r Z pa
wk pa
For plastic modulus Hp and Kp as well as other details of
the model formulation for cyclic loading, readers are
referred to the original publications [1,2].
Fig. 3. Model Simulation of Cyclic Torsional Test Results [8].
2.2. Comparison with laboratory test results
This cyclic nonlinear model has been used to simulate
sand and silt behavior under various loading and saturation
(5) conditions [1]. Yamada and Ishihara [7] conducted
extensive laboratory studies using a true triaxial device.
They studied Fuji River sand under complex stress paths
including rotational shear, and all these test results were
simulated using the model with one set of parameters [1,2].
Towhata and Ishihara [8] performed cyclic torsional shear
tests for Toyoura sand. The model simulation and test
results (for both the stress–strain relation and effective-
stress path) are compared in Fig. 3. In these comparisons,
the cyclic contractive and dilative behavior of the sand is
well predicted by the model.
2.3. Comparison with other nonlinear models
Although there is a number of plasticity constitutive
models for soils published in the literature, but, many of
these models have not been used in the industry. In recent
(6)
supported research project is aimed to shorten the gap
between the research tools and practical applications using a
multi-yield surface model [9]. But, in the industry,
currently, two nonlinear soil models are often used in the
dynamic analysis of soil structures—the Mohr–Coulomb
 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252 235

(or bilinear model), and the hyperbolic stress–strain 
K1

formulation [10]. Two parameters characterize these two
models in the simple shear mode: the soil strength, tm, and
initial shear modulus, Gmax. The hyperbolic stress–strain
relationship is defined by:
Gmax g
tZ
1 C Gtmax g
m
It has an equivalent functional form as
g Z t=G G=Gmax Z 1 K t=tm
in which G is the secant modulus and G/Gmax is the
‘modulus reduction curve.’
Modulus reduction curves represent basic soil properties
used for analyzing the dynamic response of soil structures.
Seed and Idriss [6] summarized laboratory tests and gave
upper- and lower-bound curves for sands, as shown by the
thin dashed lines in Fig. 4. The top first and second lines
show modulus reduction curves based on the Mohr–
Coulomb and hyperbolic models, respectively. For a given
soil strength, tm, and an initial shear modulus, Gmax,, these
modulus reduction curves are uniquely defined for each of
the two models, as shown in Fig. 6. These curves do not
adequately match the ranges for sand presented by Seed and
Idriss [6]. As commonly recognized, modulus reduction
curves differ for different soils. Therefore, it is desirable to
provide constitutive models that can better fit the curves
derived from the test data.
In the bounding surface hypoplasticity model, Eqs. (1)
and (2) can be integrated under monotonic simple shear
loading with pZconstant condition. The analytical solution
of the model for this loading condition is
Gmax = 2484 ksf, τm = 2.6 ksf
from Model of Wang (1990) with hr = 0.2
from Hyperbolic Curve
from Mohr-Colomb Model
Range for Sand after Seed & Idriss (1970)
1.0
Normalized Shear Modulus, G/Gmax
2t t t
G=Gmax Z 1 K m lnð1 K Þ C Þ
hr t tm tm (9)
g Z t=G
in which G is secant modulus and hr is a model parameter.
The modulus reduction curve, G/Gmax, for this model is the
first equation in Eq. (9).
(7) For a given soil strength, tm, and shear modulus, Gmax, the
modulus reduction curve (Eq. (9)) is a function of shear
stress, and the function varies with model parameter hr, as
shown in Fig. 6. This parameter can be calibrated against a
given soil modulus reduction curve to obtain the best fit to the
laboratory data. The capability of this model to fit specific
(8) modulus reduction curves enables a good comparison with
results obtained using an equivalent linear model, except at
large strain ranges (1% or higher), where the equivalent
linear approach may unreasonably compute stresses that
exceed the soil strength. Other ‘modified hyperbolic’ models
have been developed by adding more parameters to Eq. (7) to
provide a better fit to published soil modulus reduction
curves; however, these analytical equations are restricted to
relationships between shear stress and shear strain only. For
appropriate engineering applications, a soil model should
provide a general relationship between stress increment and
strain increment tensors (e.g. see Eqs. (1) and (2)).
The original hyperbolic stress–strain relation represents
nonlinear elasticity; i.e. volumetric strain only is affected by
mean stress changes, it does not predict the volumetric
changes induced by variation of shear stress or there is no
coupling between deviatoric stress (including shear stress)
and volumetric strain. Several empirical formulations have
been proposed to compute volumetric strains due to shear
strain changes, such as in Martin et al. [11]:
d3vd Z c1 ðga K c2 3vd Þ C c3 32vd =ðga C c4 3vd Þ (10)
Based on this empirical equation, the volumetric strain
increment per cycle is a function of the total accumulated
volumetric strain, 3vd, and the amplitude of the shear strain
cycle, ga. Although the relation considers the strain
hardening from accumulated volumetric strain, it does not
simulate the dilative behavior of sand within a cycle.
In the model of Wang [1,2], the plastic volumetric strain

0.9 increment (due to shear) is computed using an indepen-

0.8 dently defined plastic bulk modulus—Kr in Eq. (1). The
0.7 ratio between the plastic bulk modulus, Kr, and plastic shear
0.6 modulus, Hr, is identified [12] as:
0.5
rffiffiffi
0.4 d3pv 3 Hr
0.3 p Z
d3d 2 Kr
0.2
rffiffiffi

a  

0.1 3 Gmax hr p h b Mp K h
0.0 Z (11)
0.0001 0.001 0.01 0.1 1 10 2 Kkr pm M h
Effective Shear Strain Amplitude (%)
For a case in which model parameters aZ0 and bZ1, it
Fig. 4. Comparison of modulus reduction curves from nonlinear models. takes the following form
236 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252

coupled analysis is the one in which calculation of

Fig. 5. Model simulation for volumetric strain under cycle drained shear
loading.
d3pv
dZ Z AðMp K hÞ
d3pd
in which A is a constant, and hZq/p is the stress ratio. This
is similar to the dilatancy equation first proposed by Rowe
[13] using the energy concept. In plasticity theory, dilatancy
simply relates volumetric part of plastic strain rate to a scalar
measure of its deviatoric part. It can be seen from Eq. (11) or
(12), if stress ratio hZq/p is greater than Mp, we have
d3pv ! 0, indicating dilative behavior. Fig. 5 is an example of
computed volumetric strain under drained cyclic loading
showing the contractive and dilative behavior of sand [1].
3. Nonlinear, fully coupled dynamic analyses
The above is an effective-stress based, fully nonlinear
cyclic plasticity model. It has been implemented into the site
response analysis (with multi-directional shaking) program
SUMDES [14], and incorporated into the two-dimensional,
large-deformation code FLAC [4], to perform fully non-
linear, fully coupled dynamic response analyses. A fully
0.5
5%-Damped
10% input of M8: SHAKE
10% input of M8: Non-linear Results (SUMDES)
0.4
Spectral Acceleration (g)
displacements and pore pressure is carried out by solving
a coupled system of equations that includes the motion
equation and the diffusion equation. We applied this
approach to the following practical problems in geotechni-
cal earthquake engineering: (1) seismic response analysis
for soft soil sites subjected to very strong earthquake
motions; (2) seismic stability analysis for soil–structure–
foundation interaction involving liquefiable soils; (3)
earthquake-induced deformation analysis of embankment
and earth dams; and (4) estimates of periods for strength
recovery and settlement in post-earthquake and post-
liquefaction conditions.
(12)
3.1. Seismic site response using nonlinear approach
In current practice, the site response to a given
earthquake input motion is commonly obtained using an
equivalent linear approach. For low-level earthquake
excitations (e.g. input peak acceleration scaled down to
about 0.1g), the nonlinear and equivalent linear approaches
produce essentially the same results (see Fig. 6), when the
model was calibrated using the same dynamic soil proper-
ties, and similar modulus reduction and damping curves.
When an input motion with a peak acceleration of 0.56g was
applied at a depth of 6.4 m to a site having a soft clay layer
at depth between 2.4 and 6.4 m, the computed ground
motions at the ground surface were different. The equivalent
linear SHAKE analysis predicted a surface motion having
an amplified peak acceleration of 0.8g, as shown in Fig. 7.
5%-Damped
Conputed Surface Motion: Equivalent Linear Results (SHAKE)
Computed Surface Motion: Non-linear Results (SUMDES)
Spactral Matched Input Motion
4
3
Spectral Acceleration (g)

0.3
2

0.2

1
0.1

0 0
0.01 0.1 1 10 0.01 0.1 1 10
Period(sec) Period (sec)

Fig. 6. Comparison of response spectra at site surface using 10% of input Fig. 7. Comparison of response spectra at site surface and input motion
motion (M8 event). (M8 event).
 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252 237

300 this site during the earthquake [15]. The predicted

liquefaction in the top 20 m of fill is shown in Fig. 8 in
Effective Mean Pressure p', kPa

at 32m terms of time histories showing the reduction in mean

effective pressure within the sandy layers. The computed
200
surface acceleration time histories match the recordings
reasonably well for both pre- and post-liquefaction periods,
at 16m as shown in Fig. 9.
at 12m
100
at 6m 3.3. Predicting deformation using the nonlinear model

Given an infinite slope surface, the static stresses can be

0
10
Fig. 8. Computed time histories of mean effective pressure at Port of Island
during Kobe earthquake.
The nonlinear analysis [14] de-amplified the peak accelera-
tion at the surface. In the nonlinear analysis an effective
friction angle of 308 was assumed for the soft clay layer. The
equivalent linear approach, however, predicts shear stresses
in the soft clay layer that exceed the assumed shear strength
of the layer. For instance, the ratio between the computed
peak shear stress and vertical confining pressure from the
equivalent linear analysis reached 0.92 at a depth of 5.8 m.
3.2. Simulation of liquefaction and post-liquefaction
site response
We used records obtained from a downhole array at Port
Island, Japan during the 1991 Kobe earthquake, to compute
the site response and predict the observed liquefaction at
20 30 40 expressed by an element of soil having a non-zero initial
Times, Second shear stress (and normal stresses). Under a sinusoidal cyclic
shear stress loading, the initial shear stress renders the cyclic
stress–strain loops asymmetric, and shear deformation
accumulates during each successive cycle, as shown in
Fig. 10. If the soil element were saturated, pore water
pressure may be generated, resulting in a reduction in
effective stress. Because the elastic shear modulus of the soil
(Eq. (3)) is a function of the mean effective pressure, the shear
modulus also will decrease. Thus, the accumulated defor-
mation will significantly increase as shown by the dashed line
in Fig. 10. Incorporating the step-by-step degradation of
modulus due to pore water pressure generation is often called
‘fully-coupled’ effective-stress analysis.
3.4. Post-earthquake and post-liquefaction settlement
and strength recovery
The seismic stability of an offshore structure foundation
was analyzed for specified design earthquakes [16].

0.8 Recorded at Ground Surface

Acceleration, g

0.4 Computed Using SUMDES (3D)

0.0
–0.4
–0.8
10 15 20 25 30 35
0.6
0.4 at Depth 16m
Acceleration, g

0.2
0.0
–0.2
–0.4
10 15 20 25 30 35
0.4
0.2
Acceleration, g

at Depth 32m
0.0
–0.2
–0.4
–0.6
10 15 20 25 30 35
Time,seconds

Fig. 9. Comparison of recorded and computed acceleration time histories, N00E component at Port of Island.
238 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252

Initial Shear Stress 0.2 ksf 4. Practical applications

Initial Normal Stress 1.8 ksf
0.8
Drained Condition In this section, we describe practical applications of the
Undrained Condition
above nonlinear approach to evaluating the seismic stability
0.6
of a rockfill dam, the performance of a waterfront slope
(berth) during the 1989 Loma Prieta, California, earthquake,
0.4 and prediction of liquefaction of the foundation and
Shear Stress, ksf

estimates of permanent deformation of an earth dam.

0.2
0
–0.2
–0.4
–0.004 0 0.004 0.008
Shear Strain
Fig. 10. Model simulation for sand with initial shear stress under cyclic
loading.
The results of the analysis indicated that a silty sand layer
underlying the site (Layer No. 2 in Fig. 11) would liquefy
except for an area directly beneath the foundation. Effective
stresses at locations A–C beneath the foundation (Fig. 12)
were computed after earthquake shaking and were found
to increase with time, due to dissipation of excess pore
water pressure. This analysis was performed to resolve a
concern that the structure could become unstable if storm
loading occurred after an earthquake had softened the
silty sand layer, and before enough time had passed for
the earthquake-induced excess pore water pressures to
dissipate.
4.1. A rockfill dam
New Exchequer Dam, located on the Merced River east
of Merced, California, is a 150 m high, concrete-faced,
rockfill dam that impounds Lake McClure, providing a gross
storage capacity of one million acre-feet at normal
maximum reservoir level. The crest length is 426.7 m.
0.012 The volume of rockfill is four million cubic meters. Both
upstream and downstream slopes are 1.4H to 1.0V. The dam
sits in a deep, relatively narrow canyon eroded in highly
competent, meta-andesite rock, and all rockfill was placed
on a fully stripped foundation. The seismic stability of the
dam was evaluated for potential earthquake ground motions
from a maximum magnitude, Mw 6.5 earthquake on the Bear
Mountains fault, about 1/2 km from the site. Estimated
median peak horizontal accelerations at the site exceeded
0.5g. The dam was analyzed using both equivalent linear
and nonlinear approaches.
Two-dimensional dynamic finite element analyses were
performed using the computer program QUAD4M [17] for
the maximum cross-section of the dam (Fig. 13). The
variation of shear modulus and damping ratio with shear
strain is shown in Fig. 14. Details of this analyses are
presented in [18]. Dynamic finite element analyses provide

Fig. 11. Grid and boundary conditions for axis-symmetric analysis using FLAC.
 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252 239

During Earthquake After Earthquake 28 Hours

2.4 2.4
k=0.0001cm/sec for Silt layers

A
2.0 2.0
A

B
1.6 1.6
B
Mean Pressure, ksf

1.2 C 1.2

C
0.8 0.8

0.4 0.4

0.0 0.0
0 4 8 12 16 20 24 28 20000 40000 60000 80000 100000
Time, second Time, second

Fig. 12. Computed times histories of mean pressure during and after earthquake.

the distribution of peak accelerations within the embank-
ment and seismic coefficient time histories within potential
sliding masses for estimating permanent slope deformation.
Slope stability analyses were performed to estimate factors
of safety and yield accelerations, ky, for potential sliding
surfaces within the upstream and downstream slopes of the
embankment. Permanent deformations were estimated
using the concept of yield acceleration proposed by
Newmark [19] and modified by Makdisi and Seed [20].
The computed peak horizontal acceleration at the crest of
the dam was about 0.85g, and the greatest amplification
occurred in the upper 12.2 m of the embankment.
The greatest estimated earthquake-induced permanent
deformation (about 18.3 cm) was for a potential ‘shallow’
sliding surface within the upper one-fourth of the down-
stream slope.
The response of New Exchequer Dam also was evaluated
using nonlinear analyses employing the bounding surface
hypoplasticity model incorporated into the finite difference
program FLAC. Six parameters are required in the nonlinear
dynamic response analyses. Three of these parameters (f,
Go, and hr) can be estimated from parameters used in the
equivalent linear analysis. f is the Mohr–Coulomb friction
angle; Go is related to K2max; and hr is a model parameter
that characterizes the relationship between shear modulus
and shear strain. The model parameter hr for each material

Fig. 13. Maximum section of New Exchequer dam.

240 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252

1.0
0.9

Normalized Shear Modulus, G/Gmax

0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0001 0.001 0.01 0.1 1

Upper Bound Sand Curve (Seed & Idriss, 1970) p=8ksf

Non-linear Model at Typical Depths (p= 8 to 39 ksf)

30 18
23
39
Damping Ratio (%)

ksf
20

10
Lower Bound
(Seed & Idriss, 1970)
0
0.0001 0.001 0.01 0.1 1

Effective Shear Strain Amplitude (%)

Fig. 14. Relationships of normalized modules and damping ratio with shear strain.

zone was calibrated to provide modulus and damping
relationships with shear strain similar to those used in the
equivalent linear analysis (Fig. 14).
The response of the maximum section of the dam was
computed using the same input acceleration time history as
used for the equivalent linear analysis. For simplicity the
rock base was assumed rigid. This assumption may have
resulted in higher computed accelerations compared to
those from the equivalent linear analysis, which treated the
bedrock foundation as an elastic half-space. The nonlinear
analysis provides the time history of acceleration and
displacement at any location within the embankment, as
well as contours of horizontal and vertical displacement at
the end of earthquake shaking. Fig. 15 shows the time
history of computed acceleration at the crest of the dam,
along with the acceleration time history computed from the
equivalent linear analysis. The computed maximum crest
acceleration from the nonlinear analysis is about 1.1g,
compared to 0.85g from the equivalent linear approach. The
phases of the two time histories are in reasonable agreement
(Fig. 15).
Time histories of horizontal and vertical displacements at
the crest of the embankment are presented in Fig. 16.
Computed peak horizontal displacements at the crest of the
embankment were less than 61 cm in the downstream
direction; computed vertical settlements were about
36.6 cm. Contours of computed horizontal and vertical
displacements throughout the embankment at the end of
earthquake shaking are shown in Fig. 17. The maximum
horizontal displacements at the end of earthquake shaking
(within the upper one-fourth of the downstream slope) were
about 46 cm. The corresponding maximum vertical settle-
ments at the crest were about 37 cm. The estimated
permanent displacements from the equivalent linear
analyses were less than 20 cm.
4.2. A water front slope
Following the 1989 Mw7 Loma Prieta earthquake in
California, significant damage was observed at the site of
the Port of Oakland’s Seventh Street Marine Terminal [21].
Damage included a slope failure along part of a dike near the
southwest corner of the terminal; damage to battered, pre-
stressed concrete piles in the wharf structure; settlement as
great as 0.3 m of the landside crane rail; numerous cracks in
the pavement in the yard area; and movement of some of
 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252 241

1.0
Equivalent Linear Model

Acceleration (g)
0.5

0.0

–0.5

–1.0
0 4 8 12 16 20 24 28 32 36
1.0
Non-linear Bounding Surface Hypo-plasticity Model
Acceleration (g)

0.5

0.0

–0.5

–1.0
0 4 8 12 16 20 24 28 32 36
Time (seconds)

Fig. 15. Crest acceleration time histories computed from equivalent linear and nonlinear approaches.

the armor stone slope protection on the face of the dike.
Other earthquake effects at the site included liquefaction-
induced sand boils, cracks at the ground surface in unpaved
areas near the dike, and lateral and vertical displacement as
great as 0.3 m between the ground surface and adjacent pile-
supported structures along the dike. The level of ground
shaking recorded at a ground motion instrument within
2 km of the site was about 0.28g.
The observed behavior of a typical profile at Berth 40
without the wharf structure (shown in Fig. 18) during the
Loma Prieta earthquake was simulated using the fully
coupled, nonlinear effective-stress analysis described in this
paper.
Construction records of the berth at this site indicate that
after soft mud was dredged from beneath the dike area,
dredged sand fill was hydraulically placed by bottom-dump
barges up to the elevation of the original mud line. A small
rock toe dike was constructed on top of the first sand lift.
After construction of the rock toe dike, a second sand lift
was placed to the elevation of the top of the toe dike. Rock
spoil material was then placed by end-dumping from trucks.
Compacted sand fill was later placed on top of the rock
spoil. Materials used to create the terminal yard area behind
Displacement at Crest ( feet)
the perimeter dike are believed to consist primarily of
hydraulically placed dredged sand.
A finite difference model of the wharf embankment
profile is presented in Fig. 19. The two lateral side boundary
condition is so called ‘transmitting boundary’ allow wave
propagation through the vertical interface. In FLAC, this is
named as ‘Free Field boundary’.
Dynamic model parameters (maximum shear modulus
and the variation of modulus and damping ratio with shear
strain) were developed based on soil properties used in
equivalent linear analyses. Cyclic strength model par-
ameters were calibrated based on in situ standard
penetration test (SPT) data (converted to (N1)60) and the
observed occurrence of liquefaction during past earth-
quakes. Such cyclic strength curves, shown in Fig. 20, are
based on the liquefaction chart of Seed et al. [22], the
magnitude scaling factors reported in Youd and Idriss [23],
and the magnitude vs. equivalent number of cycles
relationship of Seed and Idriss [24]. The dashed lines in
Fig. 20 show the cyclic strength curves from the field data;
the solid curves were predicted based on calibrated model
parameters. Details of the model parameters used in the
analysis are described in [25]. The input ground motion used

2.0
1.5
1.0 Horizontal Displacement Relative to Rock Base
0.5
0.0
–0.5
–1.0
–1.5 Vertical Displacement
–2.0
0 4 8 12 16 20 24 28 32 36
Time (seconds)

Fig. 16. Computed time histories of crest displacements using FLAC and bounding surface hypoplasticity model[1].
242 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252

Fig. 17. Computed displacement contours using FLAC and bounding surface hypoplasticity model [1].

in the analysis was based on the recording from the strong
motion instrument near the site. The time history recorded at
the ground surface was deconvolved to compute an interface
motion at the top of the stiff old clay (bay mud) layer
underlying the site at a depth of about 20 m.
The analysis was performed first for the free-field zone
(i.e. a one-dimensional column) representing areas away
from the crest of the slope. Three layers (A–C in Fig. 19)
located at the bottom, middle, and top of the submerged
sand fill were selected to show the computed decrease in
mean effective stress, the effective-stress path, and the shear
stress vs. shear strain variation during earthquake shaking.
Fig. 21 shows that the computed effective mean stress in
layers A and B dropped to its lowest level after 15 s of
shaking, indicating the occurrence of liquefaction. In
contrast, the top layer, C, did not reach full liquefaction
because the underlying layers A and B had already liquefied
and fully softened, and thus could not transmit the strong
shaking to the soil above. However, this layer also was
softened by the significant reduction in effective stress.
Fig. 22(a) and (b) show the relationships between the
shear stress and effective mean stress, as well as shear
strain computed for layer B, using the bounding surface
model.
 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252 243

–100 –90 –80 –70 –60 –50 –40 –30 –20 –10 0 10 20 30 40 50 60 70 80 90 100
40 40

30 30

20 20

10 Compacted sand fill 10

Rip rap
0 Rock spoil dike: 0

–10 Sand fill Dike fill Rock toe dike: –10

Elevation, feet

–20 –20

–30 Medium dense to dense clayey/silty sand –30

–40 –40
Dense to very dense sand
–50 Medium dense to dense clayey sand –50

–60 –60
Dense to very dense sand
–70 –70
Stiff to very stiff silty clays
–80 –80
–100 –90 –80 –70 –60 –50 –40 –30 –20 –10 0 10 20 30 40 50 60 70 80 90 100
Horizontal Distance, feet

Fig. 18. Typical cross section at Berth 40, Port of Oakland.

Similar responses were computed for zones near the
slope (locations D–F in Fig. 19). The effective-stress time
histories (Fig. 23) show that the sand fill near the rock toe
dike (zone F) had liquefied, while the sand fill beneath the
crest (zone D) experienced dilative behavior after initial
liquefaction. The rock spoil portion (zone E) did not liquefy.
The computed time histories of displacement, both
horizontal and vertical, at the crest are shown in Fig. 24.
The base displacement is shown as a reference for the crest
movements. The computed deformed shape of the slope at
the end of earthquake shaking is shown in Fig. 25. The
greatest displacement values (about 0.5 m) were those
developed at the rock toe below the water surface. The
horizontal displacement at the crest was computed to be
about 37 cm. The computed settlement at the crests was
about 24 cm. These computed deformations are consistent
with those observed during the Loma Prieta earthquake.
SPT blow counts obtained from field investigations by the
US Army Corps of Engineers (Corps), Sacramento District,
are shown in Fig. 26. These values and laboratory test
results were used to calibrate the model parameters.
(1) Parameter f is the effective friction angle. For the
alluvium, the pervious and transition zones it was estimated
based on the laboratory data as shown in Fig. 27.
(2) The Corps of Engineers conducted a geophysical
investigation where shear wave velocities were measured
in borings drilled through the upstream and downstream
slopes of the embankment. The measured shear wave
velocities were presented in a ‘normalized’ form in terms of
Vs1 (see Fig. 26):

1=4
pa
Vs1 Z Vs (13)
svo

Based on Eq. (3), it can be shown that the model parameter

4.3. An earth dam Go is related to Vs1 by the following equation:

Success Dam serves primarily as a flood control and

water conservation project that was constructed between
1958 and 1961. The dam is a rolled-earthfill embankment
44 m high and 1038 m long. It consists of a central
impervious core along with upstream and downstream
pervious shells. The pervious shells of the embankment
were constructed on in situ recent alluvium, alluvial fan
deposits, terrace deposits, and overburden (slope wash)
deposits that may be susceptible to liquefaction during a
seismic event [26]. A typical cross-section of the dam
showing values of normalized shear wave velocities and Fig. 19. Finite difference grid and typical material zones.
244 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252

0.6

0.5

Cyclic Stress Ratio, τ/ mo' d=2 d=2.5

0.4
(N1 )60 =20

0.3

d=1

0.2 (N1 )60 =15

(N1 )60 =10

0.1
(N1)60=5

0.0
1 2 3 4 5 6 7 8 9 10 20 30 40 50 60 70 80 90 100
Number of Cycle to Liquefaction, N

Fig. 20. Cyclic strength of sand with 5% (or less) fines based on liquefaction chart [24].
rffiffiffiffiffiffiffi
g svo calibrated using the results of monotonic laboratory tests.
Go Z V2 (14)
gpa VðeÞ s1 p This calculation was made only for the recent alluvium that is
susceptible to liquefaction. Fig. 29 shows the triaxial shear
test results and model simulation using krZ0.55 and an
(3) Parameter hr characterizes the nonlinear relationship estimated phase transformation ratio, Rp, as 0.85 of the
between shear modulus and shear-strain amplitude. It can be failure ratio, Rf, that corresponds to a friction angle of 368.
calibrated against laboratory test results or published (5) Model parameter d controls the development of excess
relationships for different soils. For Success Dam, there are
insufficient cyclic test results to calibrate this parameter. (a) 0.8
Instead, we selected published modulus reduction relation- Computed Effective Stress Path in Layer B
Shear Stress (ksf)

ships. For example, for the cohesionless transition, pervious, 0.4

and alluvium zones, the upper-bound relationship for sand
developed by Seed and Idriss [24] was selected to calibrate 0.0
the model parameters, the calibration procedure involved
searching for a value of parameter hr that provides a best fit to –0.4
the selected modulus reduction curves as shown in Fig. 28.
(4) Parameter kr, which affects the changes in effective stress –0.8
under monotonic (and virgin) loading conditions, was 0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8 3.2
Effective Mean Pressure (ksf)
2.0 (b) 0.8
Effective Mean Pressure (ksf)

A At Inland free field Computed Shear Stess Vs. Shear Strain in Layer B
1.6
Shear Stress (ksf)

0.4
B
1.2
C 0.0
0.8
C –0.4
0.4

–0.8
0.0 –0.002 –0.001 0 0.001 0.002 0.003 0.004
0 4 8 12 16 20 24 28 32 36 40
Time (sec) Shear Strain

Fig. 21. Computed mean effective pressure time histories in sand fill layers Fig. 22. Computed mean effective pressure, shear stress and shear and shear
at inland free field. strain in sand fill layer B at inland free field.
 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252 245

2.0 pore water pressure due to cyclic loading under undrained

Effective Mean Pressure (ksf)

conditions (or shear-induced volumetric change under

1.6 Under Crest and Slope
drained conditions). The number of cycles required to
D cause liquefaction of a soil element was estimated based on
1.2
E the computed effective-stress path (or shear stress vs.
0.8 effective mean pressure), as shown in Fig. 30 for a given
cyclic stress. From the above example, for the selected
0.4 F parameters (krZ0.55 and dZ3) at the specified stress ratio of
0.3, seven cycles are needed to reduce the mean effective
0.0
0 4 8 12 16 20 24 28 32 36 40 pressure to the point of initiating or ‘triggering’ liquefaction.
Time (sec)
The procedure for simulating cyclic loading was repeated for
Fig. 23. Computed mean pressure time histories in sand fill zones under selected locations in the recent alluvium (represented by
crest and slope. mean effective stresses of 4 and 8 kips per square foot (ksf))
using different cyclic stress ratios. The number of cycles to
2.0 trigger liquefaction was determined for each stress ratio.
1.6
Each combination of stress ratio and the corresponding
Computed Displacement, feet

X-displacement at crest number of cycles to cause liquefaction in the simulated test

1.2
was used to construct the model simulation curves shown by
0.8
dashed lines in Fig. 31. For the recent alluvium layer under
0.4
X-displacement at base the upstream slope (represented by a SPT blow count of 10), a
0.0 model parameter dZ1 provided a reasonable fit to the
–0.4 strength curve based on the field data. A similar curve using a
–0.8 model parameter dZ3 was developed for the recent alluvium
Y-displacement at crest
–1.2 layer under the downstream slope, represented by a blow
–1.6 count of 15. Model parameters are presented in Table 1.
–2.0
0 4 8 12 16 20 24 28 32 36 40 A rock input motion provided by the Corps was used in
Time (sec)
the dynamic analysis. The estimated peak acceleration of
Fig. 24. Computed horizontal and vertical displacement time histories at the record is about 0.22g, and the time history represents
crest of slope.

JOB TITLE : (*10^ 2)

FLAC (Version 3.40)

0.750

LEGEND

18-Sep-99 4:14
step 209630
Cons. Time 3.7710E+10
0.250
–1.341E+02 <x< 1.335E+02
–1.626E+02 <y< 1.049E+02

Exaggerated Grid Distortion

Magnification = 5.000E+00
Max Disp = 1.761E+00
–0.250

–0.750

–1.250

Geomatrix Consultants, Inc.

2101 Webster Street, 12 Floor, Oa
–1.000 –0.500 0.000 0.500 1.000
(*10^ 2)

Fig. 25. Computed deformed slope at Berth 40 (displacements shown are magnified five times).
246 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252

Fig. 26. Typical cross section of success dam (1 footZ0.3m).

a relatively long duration of shaking (an 80 s record),
simulating a magnitude 8.0 event on the Owens Valley fault.
The input motion is specified at the base of the old dense
alluvium layer underlying the dam foundation. Fig. 32
shows both the input acceleration time history and the
computed acceleration time history at the crest. Fig. 32
indicates that the input motion was amplified at the crest of
the dam for at least the first 25–30 s. As discussed below, the
recent alluvium layer liquefied after about 20–30 s of
shaking.
Dynamic shear stresses result in buildup of excess pore
pressures in the zones within the embankment and
foundation that have relatively low blow counts, such as
the recent alluvium layer in the foundation. The increase in
pore water pressure results in a corresponding decrease in
the mean effective stress. Fig. 33 shows the time histories of
the mean effective stresses computed for four selected
locations in the recent alluvium underlying the upstream
slope of the dam. The initial (pre-earthquake) values were
those computed from static stress and seepage analysis. The
lowest allowable mean effective stress was specified at
a gradual softening and resulting in accumulation of shear
strains (maximum strains are not shown).
The predicted deformed shape of the embankment at the
end of earthquake shaking is shown in Fig. 35. This figure
shows that at the end of earthquake shaking, the computed
maximum settlement at the crest of the dam is about 5.5 m.
Maximum computed horizontal displacements of points on
the embankment slopes were 1.5 and 15 m at the upstream
and downstream toes, respectively.
It may be useful to compare the predicted large
deformation for this embankment with the actual perform-
ance of other similar dams observed in past earthquakes.
Sheffield Dam was constructed on a silty sand layer with an
estimated (N1)60 for clean sand of about 8 blows per foot.
Sheffield Dam failed near the end of earthquake shaking
during the 1925 Santa Barbara earthquake [27] as a result of
sliding of the entire embankment on the liquefied layer.
40
Pervious Zone
B-4B @ elev. 596.4'
35 B-5B @ elev. 597.2'
Principal Stress Difference, (σ1'–σ3')/2 (ksf)

B-6B @ elev. 608.5'

14.4 kPa (0.3 ksf) to simulate a state of initial liquefaction. B-7B @ elev. 613.0'
Liquefaction occurred first in the recent alluvium near the 30
B-8B @ elev. 618.6'
B-11B @ elev. 655.6'
upstream and downstream toes of the slope (at about 12– B-12B @ elev. 668.0'
16 s) and then extended toward the centerline of the dame sinφ'=tanψ, c'=a/cosφ B-15B @ elev. 595.5'
25 φ'=37 degrees, c'=0 B-18B @ elev. 651.5'
(at about 28–30 s).
Fig. 34 shows computed shear stresses and shear strains 20
for four zones in the recent alluvium underlying the
upstream shell of the embankment (similar, but much larger ψ=31 degrees
15
shear strains were computed in the alluvium beneath the
downstream side). The left columns on this figure show 10
relations of shear stress vs. mean effective stress. The failure
envelope, corresponding to a friction angle of 368, is also 5
shown in this figure using dashed lines.
The right column in Fig. 34 shows the shear stress-shear 0
a=0
strain relations for zones in the recent alluvium under the 0 5 10 15 20 25 30 35 40
upstream shell. As the effective-stress paths move from the Average Principal Effective Stress, (σ1'+σ3')/2 (ksf)
initial mean stress to minimum mean stress (triggering of
liquefaction), the shear stress and shear strain loops show Fig. 27. Effective Stress Stength Envelope of Previous Zone.
 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252 247

1.0

Normalized Shear Modulus (G/Gmax) and Damping (β)

0.9

0.8

0.7

0.6

0.5
from Seed and Idriss (1970)
Transition Zone
0.4
Upstream Pervious Shell
0.3 Downstream Pervious Shell
Upstream Recent Alluvium
0.2 Downstream Recent Alluvium
Old Alluvium
0.1

0.0
0.0001 0.001 0.01 0.1 1
Shear Strain, (%)

Fig. 28. Normalized Shear Modulus and Damping Ratios vs. Shear Strain for Sand.

Our estimated deformation for the Success Dam, using the represented by the concept of critical state. Applications
analyses described above, are consistent with the observed of critical-state soil mechanics to the constitutive modeling
performance of Sheffield Dam during and after the 1925 of sand have gained wide acceptance in geotechnical
Santa Barbara earthquake. engineering. To simulate the behavior of sand at the critical
state, the phase transformation stress ratio line originally
5. Recent developments in the constitutive model proposed by Ishihara et al. [29], was revised as a curved
for sand dilatancy line by Wang et al. [12] that passes through the
critical state. This line is defined using a ‘state pressure
Laboratory tests on Toyoura sand [28] show that sand index,’ IpZP/Pc. Here, pc is the mean pressure at the critical
behavior under large shear deformations can be well state corresponding to the current void ratio.

25 25

20 20
sin φ'=3M/(6+M) Line of Phase
φ'=36 degrees Transformation
Deviator Stress, q (ksf)

M=1.46, Failure Line

15 15
1

10 10

5 5

0 0
0 0.05 0.1 0.15 0.2 0.25 0.3 0 5 10 15 20 25
Axial Strain Mean Effective Stress, p', (σ1'+2σ3')/3' (ksf)

Model Simulations TH-15 specimen 2

TH-14 specimen 1 TH-15 specimen 3
TH-14 specimen 2 TH-15b specimen 1
TH-14 specimen 3 TH-15b specimen 2
TH-15, specimen 1 TH-15b specimen 3

Fig. 29. Comparison of Model Simulations and Triaxial Test Results.

248 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252

(a)

Shear Stress, τ (ksf)

0

–1
Effective Stress Path

0 2 4 6
Effective Confining Pressure, P' (ksf)

(b)

1
Shear Stress, τ (ksf)

–1
Shear Stess Vs. Shear Strain

–0.20 –0.15 –0.10 –0.05 0.00 0.05 0.10 0.15 0.20

Shear Strain

Fig. 30. Computed Cyclic Stress-Strain Relations for Stress RatioZ0.30 of Recent Alluvium.

0.6
Cyclic Strength Relations Based on Seed et al. (1985)
(N1)60=20 Model simulation for (N ) =10 using d=1, kr=0.55, from Po'=4ksf
1 60
Model simulation for (N ) =15 using d=3, kr=0.55, from Po'=4ksf
1 60
(N1)60=15 Model simulation for (N ) =10 using d=1, kr=0.55, from Po'=8 ksf
1 60
0.5 Model simulation for (N ) =15 using d=3, kr=0.55, from Po'=8ksf
1 60
Cyclic Stress Ratio, τ/σmo'

0.4

(N1)60=10

0.3

(N1)60=5
0.2

0.1

0.0
1 2 3 4 5 6 7 8 9 10 20 30 40 50
Number of Cycles to Liquefaction, N

Fig. 31. Cyclic Strength Relations and Model Simulations.

 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252 249

Table 1
Material zones and model parameters Success Dam

Zone no. Zone description Void ratio f0 Go k hr kr d Rp/Rf b

a
I Impervious core 0.773 18–41 200 0.02 1.663 100 100 1.0 2.0
II Transition 0.685 37 354 0.002 0.241 100 100 1.0 2.0
III Pervious shell (US) 0.392 37 335 0.002 0.308 100 100 1.0 2.0
(DS) 0.392 37 336 0.002 0.443 100 100 1.0 2.0
IV Recent alluvium 0.685 36 260 0.002 0.189 0.55 1.0 0.85 2.0
(US)
(DS) 0.685 36 354 0.002 0.227 0.55 3.0 0.85 2.0
V Old Alluvium 0.620 41 1230 0.002 1.200 100 100 1.0 2.0
a
For clay core at a depth that is equivalent to a strength of CZ0.8 ksf and fZ118.

0.4
Horizontal Acc. (g)

computed motion at crest

0.2

–0.2

–0.4
0.4
Horizontal Acc. (g)

input motion at base

0.2

0.0

–0.2

–0.4
0 10 20 30 40 50 60 70 80 90
Time(sec)

Fig. 32. Horizontal acceleration time histories at base and computed at crest of Success dam.

10
Upstream Recent Alluvium
at zone (20, 2)
8
Effective Mean Pressure (ksf)

6 at zone (16, 2)

4
at zone (12, 2)

at zone (7, 2)

0
0 10 20 30 40 50 60 70 80
Time(sec)

Fig. 33. Computed mean effective pressure time histories beneath upstream shell of Success Dam (1 ksfZ47.9 kpa).
250 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252

3 3
at zone (7, 2) at zone (7, 2)
2 2

Shear Stress, τ 1 1

0 0

–1 –1

–2 –2
0 1 2 3 4 5 6 7 8 9 10 –0.04 –0.03 –0.02 –0.01 0
3 3
at zone (12, 2) at zone (12, 2)
2 2
Shear Stress, τ

1 1

0 0

–1 –1

–2 –2
0 1 2 3 4 5 6 7 8 9 10 –0.04 –0.03 –0.02 –0.01 0
3 3
at zone (16, 2) at zone (16, 2)
2 2
Shear Stress, τ

1 1

0 0

–1 –1

–2 –2
0 1 2 3 4 5 6 7 8 9 10 –0.04 –0.03 –0.02 –0.01 0
3 3
at zone (20, 2) at zone (20, 2)
2 2
Shear Stress, τ

1 1

0 0

–1 –1

–2 –2
0 1 2 3 4 5 6 7 8 9 10 –0.04 –0.03 –0.02 –0.01 0
Mean Effective Stress, po Shear Strain, εxy

Fig. 34. Computed stress and strain relations in recent alluvium beneath upstream shell of Success Dam (stress ub ksf, 1 ksfZ47.9 kPa).

The dilatancy line is defined as: consolidated at different confining pressures. The model

simulation using this approach is compared with test results
p
q Z Md p Z Mo C ðM K Mo Þ p (15) on Toyoura sand. Readers are referred to [12] for details.
pc Another desirable improvement to the model of Wang
Using this state-dependent dilatancy line to replace the [1] is the ability to simulate post-liquefaction deformation.
phase transformation line (a straight line in Fig. 1), the model Recent laboratory tests (Kamerrer, 2001, personal com-
can simulate the behavior of Toyoura sand at critical state munication) demonstrate that shear deformation develops
using a unique set of parameters for different void ratios continuously following initial liquefaction, in sands
 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252
–5.000 –3.000 -1.000
Fig. 35. Computed deformed grids of the Success Dam at end of earthquake shaking (1 footZ0.3 m)
subjected to cyclic loading. The effective-stress paths after
liquefaction repeat dilative and contractive behaviors that
are similar in each cycle, but the corresponding shear strain
amplitude increases continuously. Wang and Dafalias [30]
proposed a strain-history dependence of plastic shear
modulus and used an intrinsic variable to capture the
reduction in the plastic shear modulus. The degredation
factor from Eq. (5) is modified as follows
1 Cx
CðxÞ Z
1 C ax
and the intrinsic variable is the accumulated plastic
deviatoric strain:
Ð qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
xZ 2depij depij (17)
Results of simple shear tests for post-liquefaction
deformation of Nevada sand (Kammemer, 2001) were
simulated using the model and Eq. (16).
The new development of the model is still limited
because there are new laboratory test results on the
uniqueness of critical-state line, so that we have not apply
the modification to our practical applications.
6. Conclusions
A nonlinear effective-stress approach is developed and
applied to the seismic stability and deformation analyses of
earth structures that may experience strong earthquake
shaking or induced liquefaction. The key element to
performing such analysis is a constitutive model capable
of simulating soil behavior under cyclic loading conditions.
The model’s capability is demonstrated by comparing its
simulation with laboratory test results.
The advantages of using the soil model [1,2] to simulate
nonlinear behavior and pore-pressure generation and
liquefaction in sandy materials are illustrated using results
for the seismic deformation and liquefaction analysis of
practical applications that include a rockfill dam, a water
front slope and an earth dam.
Because the soil properties needed to support such
analyses are generally similar to those currently used in
performing equivalent linear and simplified effective-stress
analyses, and considering the increasing availability of
251
Boundary of Undeformed Grid (*10^3)
Deformed Grid 0.700
0.500
1.000 3.000 5.000
(*10^2)
high-speed personal computers, the above analyses are
becoming more practical and useful in current practice.
Acknowledgements
The waterfront slope example was part of a study
performed for the Port of Oakland; the authors acknowledge
(16) their support. The authors also wish to thank Messrs. Ronn
Rose and Matt Allen, of the US Army Coprs of Engineers,
Sacramento District, for their permission to present the
results of analyses of Success Dam. L. Scheibel and C.C.
Chin of Geomatrix Consultants participated in the projects
described in this paper.
References
[1] Wang ZL. Bounding surface hypoplasticity model for granular soils
and its applications. PhD Dissertation for the University of California
at Davis, U.M.I. Dissertation Information Service, Order No.
9110679; 1990.
[2] Wang ZL, Dafalias YF, Shen CK. Bounding surface hypoplasticity
model for sand. ASCE, J Eng Mech 1990;116(5):983–1001.
[3] Wang ZL, Makdisi FI. Implementing a bounding surface hypoplas-
ticity model for sand into the FLAC program. In: Proceedings of the
international symposium on numerical modeling in geomechanics.
Minnesota, USA; 1999. p. 483–90.
[4] Itasca Consulting Group, Inc. FLAC (Fast Lagrangian Analysis of
Continua), Version 4. Minneapolis, MN; 2000.
[5] Seed HB, Lee KL, Idriss IM, Makdisi FI. The slides in the San
Fernando dams during the earthquake of February 9, 1971. J Geotech
Eng Div, ASCE 1975;101(GT7):651–88.
[6] Seed HB, Idriss IM. Soil moduli and damping factors for dynamic
response analyses. Earthquake Engineering Research Center, Univer-
sity of California, Berkeley, EERC79-10 1970.
[7] Yamada Y, Ishihara K. Undrained deformation characteristics of sand
in multi-directional shear. Soils Found 1983;23(1):61–79.
[8] Towhata I, Ishihara K. Undrained strength of sand undergoing cyclic
rotation of principal stress axes. Soils Found 1985;25(2):135–47.
[9] Yang Z, Elgamal A, Parra E. Computational model for cyclic mobility
and associated shear deformation. J Geotech Geoenviron Eng 2003;
129(12):1119–27.
[10] Konder RL, Zelasko JS. A hyperbolic stress–strain formulation for
sands. Proceedings, second Pan American conference on soil
mechanics and foundation engineering; 1963. p. 289–324.
[11] Martin GR, Finn WD, Seed HB. Fundamentals of liquefaction under
cyclic loading. J Geotech Div ASCE 1975;101(GT5):423–38.
252 Z.-L. Wang et al. / Soil Dynamics and Earthquake Engineering 26 (2006) 231–252
[12] Wang ZL, Dafalias YF, Li XS, Makdisi FI. State pressure index for
modeling sand behavior. J Geotech Geoenviron Eng 2002;128(6):
511–9.
[13] Rowe PW. The stress–dilatancy relation for static equilibrium of an
assembly of particles in contact. Proc R Soc Lond; Ser A 1962;269:
500–27.
[14] Li XS, Wang ZL, Shen CK. SUMDES: a nonlinear procedure for
response analysis of horizontally layered sites subjected to multi-
directional earthquake loading. Davis: Department of civil engineer-
ing, University of California; 1992.
[15] Wang ZL, Chang CY, Mok CM. Evaluation of a site response using
downhole array data from a liquefied site. Third international
conference on recent advances in geotechnical earthquake engineer-
ing and soil dynamics, San Diego, CA, No. 430 2001.
[16] Geomatrix Consultants, Inc. Seismic response, liquefaction, and
deformation analysis for platform site WHP-1. Bohai Bay, Offshore
China: Phillips China, Inc.; 2001.
[17] Hudson M, Idriss IM, Beikae M. User’s manual for QUAD4M: a
computer program to evaluate the seismic response of soil structures
using finite element procedures and incorporating a compliant base.
Davis: Center for geotechnical modeling, University of California;
1994.
[18] Makdisi FI, Wang ZL, Edwards WD. Seismic stability of new
exchequer dam and gated spillway. Twentieth annual USCOLD
lecture series. Seattle, Washington; 2000. p. 437–58.
[19] Newmark NM. Effects of earthquakes on dams and embankments.
Geotechnique 1965;15(2):139–60.
[20] Makdisi FI, Seed HB. Simplified procedure for estimating dam and
embankment earthquake-induced deformation. J Geotech Eng Div
ASCE 1978;104(GT7):849–67.
[21] Egan JA, Hayden RF, Scheibel LL, Otus M, Serventi GM. Seismic
repair at Seventh Street Marine Terminal. In: Borden RH, Holtz RD,
Juran I, editors. Proceedings, grouting, soil improvement and
geosynthetics, vol. 2. ASCE, Geotechnical Special Publication No.
30; 1992. p. 867–78.
[22] Seed HB, Tokimatsu K, Harder LF, Chung R. Influence of SPT
procedures in soil liquefaction resistance evaluations. J Geotech Eng
Div ASCE 1985;111(12):1425–45.
[23] Youd TL, Idriss IM. Summary reports. In: Proceedings of NCEER
workshop on evaluation of liquefaction resistance of soils. NCEER-
0022; 1997.
[24] Seed HB, Idriss IM. Ground motion and soil liquefaction during
earthquakes. Monograph series, Earthquake Engineering Research
Institute. Berkeley, CA: University of California; 1982.
[25] Wang ZL, Egan J, Scheibel L, Makdisi FI. Simulation of earthquake
performance of a waterfront slope using fully coupled effective stress
approach. FLAC and numerical modeling in geomechanics. Proceed-
ings of the second international FLAC symposium. Lyon, France;
2001.
[26] Allen MG, Nickell JS, Sherer SG. Geotechnical characterization of
success dam for a dam safety earthquake engineering evaluation.
Fourth international conference on recent advances in geotechnical
earthquake engineering and soil dynamics and symposium in honor of
Professor W.D. Liam Finn. San Diego; CA 2001.
[27] US Army Corps of Engineers. Report on investigation of failure of
Sheffield Dam. Santa Barbara, CA; 1949.
[28] Verdugo R, Ishiharam K. The steady state of sandy soils. Soils Found
1996;36(2):81–91.
[29] Ishihara K, Tatsuoka F, Yasuda S. Undrained deformation and
liquefaction of sand under cyclic stress. Soils Found 1975;15(1):
29–44.
[30] Wang ZL, Dafalias YF. Simulation of post-liquefaction deformation
of sand. Paper No. 183. ASCE 15th engineering mechanics
conference. New York: Columbia University; 2002.