j.1467-8667.2007.00538.x

7/30/2019 j.1467-8667.2007.00538.x

1/14

Computer-Aided Civil and Infrastructure Engineering 23 (2008) 309322

INDUSTRIAL APPLICATION

Response and Modeling of Cantilever RetainingWalls Subjected to Seismic Motions

Russell A. Green

Department of Civil and Environmental Engineering, University of Michigan, 2372 G.G. Brown,Ann Arbor, MI, 48109-2125, USA

C. Guney Olgun

Department of Civil and Environmental Engineering, Virginia Polytechnic Institute and State University,

Blacksburg, VA, 24061, USA

&

Wanda I. Cameron

Department of Civil and Environmental Engineering, University of Michigan, 2372 G.G. Brown,Ann Arbor, MI, 48109-2125, USA

Abstract: A series of nonlinear, explicit finite differ-ence analyses were performed to determine the dynamic

response of a cantilever retaining wall subjected to earth-

quake motions. This article outlines the calibration and

validation of the numerical model used in the analyses

and comparisons are presented between the results from

the finite difference analyses and results from simplified

techniques for computing dynamic earth pressures and

permanent wall displacement (i.e.,Mononobe-Okabe and

Newmarkslidingblockmethods).Itwasfoundthatatvery

low levels of acceleration, the induced pressures were in

general agreement with those predicted by the Mononobe-

Okabe method. However, as the accelerations increased

to those expected in regions of moderate seismicity, the

induced pressures are larger than those predicted by the

Mononobe-Okabe method. This deviation is attributed

to the flexibility of the retaining wall system and to the

observation that the driving soil wedge does not respond

monolithically, but rather responds as several wedges. Ad-

To whom correspondence should be addressed. E-mail: rugreen@

umich.edu.

ditionally, it was found that the critical load case for the

structural design of the wall differed from that for the

global stability of the wall, contrary to the common as-

sumption made in practice that the two load cases are the

same.

1 INTRODUCTION

Presented herein is an overview of a numerical modeland analyses performed to determine the trends in the

dynamic response of cantilever retaining walls subjectedto earthquake motions, where the walls are allowed toflex, rotate, and slide. Significant emphasis is placed onidentifying trends in the wall response and on calibrat-ing and validating the soil-wall system model used in thisstudy, as opposed to the development and implementa-tion of advanced component level models (e.g., soil andwall constitutive models). In this vein, the component-level models used in this study are available as optionsin many commercial finite element and finite differencesoftware packages. However, the validation procedure

C

2008 Computer-Aided Civil and Infrastructure Engineering. Publishedby Blackwell Publishing, 350 Main Street,Malden, MA 02148, USA,and 9600 Garsington Road, Oxford OX4 2DQ, UK.

7/30/2019 j.1467-8667.2007.00538.x

2/14

310 Green, Olgun & Cameron

used is readily extendable system models that incorpo-rate advanced, user-developed component-level models.

True validation of any system model/analysis ap-proach requires good agreement between predicted andactual field performance data, not only at the end ofshaking but throughout the entire duration of shaking.

Unfortunately, such detailed and well-documented fieldperformance data requisite for proper validation of seis-mic analyses of cantilever retaining walls (and for almostanytype of structure) do not exist. Consequently, the val-idation approach adopted by the authors was to performthe analyses using system models that have increasingcomplexity and to evaluate the trends in the response bycomparing them to results from simple analytical models(e.g., Okabe, 1924; Mononobe and Matsuo, 1929; New-mark, 1965). When the response results differed eitherin trend or significantly in amplitude, efforts were spentto determine whether the differences were due to inher-

ent limitations of thesimple models or whether themorecomplex system model was flawed. This assessment wasfacilitated by performing limited parametric studies us-ing both the complex and simple models and also by ex-amining trends identified in published results from thephysical model tests (i.e., shaking table and centrifugetests). The computer code, Fast Lagrangian Analysis ofContinua (FLAC) (Itasca, 2000) was used to perform thenumerical analyses (i.e., nonlinear, explicit finite differ-ence analyses). The analyses consisted of the incremen-tal construction of the wall and placement of the backfill,followed by dynamic response analyses, wherein the soilwas modeled as elasto-plasticwith a Mohr-Coulomb fail-ure criterion.

Brief mention needs to be made of the significant stud-ies that have been previously published for estimatingthe seismically induced forces on retaining walls and/orto estimate the seismically induced displacements of re-taining walls (e.g., Wood, 1973; Richards and Elms, 1979;Nadim and Whitman, 1983; Veletsos and Younan, 1994;Wu and Finn, 1999; Ostadan, 2005) and how the studypresented herein differs from them. Most previous stud-ies have focused on gravity retaining wall systems ornonyielding walls. Cantilever walls, which are the focusof this study, differ from gravity walls and nonyielding

walls, in that structural stability as well as global stabilityneed to be considered, with the critical load case for thetwo being different as concluded from this study. Also,the more complex numerical analyses are used to high-light the limitations of the simple analytical approaches.

Regarding the organization of the remaining parts ofthis article, first a description of the wall-soil system ispresented, followed by an overview of the finite differ-ence model used in the study. Next, a discussion of howthe ground motions were specified, how the wall andsoil model parameters were determined, and how the

wall-soil interface was modeled are presented. Compar-isons of the finite difference results are made with resultsfrom simplified analysis techniques for determining dy-namic earth pressures (i.e., Mononobe-Okabe approach(Okabe, 1924; Mononobe and Matsuo, 1929) and per-manent displacement of the wall (i.e., Newmark sliding

block approach (Newmark, 1965)). Subsequent to thesecomparisons, additional observations made from the fi-nite difference analyses are discussed.

2 NUMERICAL MODEL

2.1 Description of wall-soil system

The retaining wall analyzed was approximately 6.1 min height, retaining medium-dense, cohesionless, com-pacted fill (total unit weight: t = 19.6 kN/m

3; effec-tive angle of internal friction: = 35). Underlying the

wall/backfill was approximately 62.5 m of naturally de-posited dense cohesionless soil (t = 19.6 kN/m

3; =40). The small strain shear wave velocity profile of thesoil deposit is shown in Figure 1. The groundwater tablewas well below the base of the wall and was not consid-ered in the analyses. The geometry and the properties

100 200 300 400 500 600 700 80080

70

60

50

40

30

20

10

0

Shear Wave Velocity (m/sec)

Depth(m)

Rock Halfspace

Base of finite

difference model

Fig. 1. Small strain shear wave velocity profile analyzed.

7/30/2019 j.1467-8667.2007.00538.x

3/14

Response and modeling of cantilever retaining walls 311

0.6 m

0.2 m

0.9 m

HeelToe

Stem

Base

6.1 m

4 m

2.4 m0.5 m

Backfill

Fig. 2. Dimensions of the structural block of the wall-soilsystem analyzed, wherein the term structural block refers

to all that is shown above.

of the wall-soil system were specified by the U.S. ArmyCorps of Engineers (Ebeling and Filz, 2001) and are con-sidered to be representative of many medium-heightwalls in the United States.

The geometry and structural detailing of the wall weredetermined following the U.S. Army Corps of Engi-neers staticdesign procedures (Headquarters, U.S. Army

Corps of Headquarters, US Army Corps of Engineers,1989, 1992), with the dimensions of the structural block(i.e., wall and contained backfill) depicted in Figure 2.The properties of the concrete and reinforcing steel usedin the wall design are as follows: unit weight of concrete:c = 23.6 kN/m

3; compressive strength of concrete: fc =27.6 MPa; and yield strength of reinforcement:fy =413.4MPa. Additional details about the wall design and soilprofile are given in Green and Ebeling (2002).

2.2 Overview of finite difference model

The finite difference model consisted of the upper 9.1m of the wall-soil system, comprising the wall/backfilland approximately 3 m of the underlying natural de-posit (foundation soil). Laterally, the model was approx-imately 22.9 m wide, to include approximately 7.6 m ofthe foundation soil in front of the wall andapproximately15.3 m of the backfill/foundation soil behind the wall(Figure 3).

An elasto-plastic constitutive model, in conjunctionwith Mohr-Coulomb yield condition, was used to modelthe soil. The dilatancy angle for the soil was assumed

15.3 m

22.9 m

7.6 m

6.1m

3

m

9.1m

1

23

4

Fig. 3. Annotated finite difference mesh of the wall-soilsystem.

equal to zero (i.e., nonassociative flow rule). Elasticbeam elements were used to model the concrete retain-ing wall, with the wall/backfill being numerically con-structed similar to the way an actual wall would be con-structed. The backfill was placed in 0.61 m lifts, for a

total of 10 lifts, with the model being brought to staticequilibrium after the placement of each lift. Such place-ment allowed realistic earth pressures to develop as thewall deformed and moved in response to the placementof each lift. The constructed retaining wall-soil model isshown in Figure 3.

The model consists of four subgrids, labeled onethrough four in Figure 3. The separation of the foun-dation soil and backfill into subgrids one and two wasrequired because a portion of the base of the retainingwall was inserted into the soil/backfill. Subgrid three wasincluded so that free-field boundary conditions could bespecified along the lateral edges of the model (free-fieldboundary conditions cannot be specified across the in-terface of two subgrids). Subgrid four was included forsymmetry, but its inclusion was not necessary. The sub-grids were attached at the soil-to-soil interfaces, as de-picted by white lines in Figure 3, and interface elementswere used at the wall-soil interfaces.

The following subsections outline how the ground mo-tions were specified and the procedures used to deter-mine the various soil and wall model parameters.

2.3 Specification of input motions

In FLAC, the dynamic motions can be specified as ac-celeration, velocity, stress, or force time histories at theexterior boundaries or as interior excitations. A para-metric study was performed to determine the best wayto specify the ground motions for earthquake analyses.The parametric study involved performing a series ofone-dimensional (1-D) site response analyses using con-sistently generated acceleration, velocity, and stress timehistories. Generally, earthquake ground motions are notdefined in terms of force time histories and, therefore,were not considered in the parametric study. The use of

7/30/2019 j.1467-8667.2007.00538.x

4/14


stress time histories has the benefit of allowing the timehistory to be specified at energy absorbing (or quietboundaries), thus simulating radiation damping.

Using free-field acceleration time histories recordedat the surface of USGS site class B profiles, 1-D site re-sponse analyses were performed using a modified ver-

sion of SHAKE91 (Idriss and Sun, 1992). The analyseswere performed on a 68.6 m, 5% damped, nondegrad-ing profile, wherein the acceleration time histories werespecified as outcrop motions.Interlayer acceleration andstress time histories were computed at the profile surfaceand at depths of 7.6, 10.7, 15.2, and 68.6 m (i.e., bedrock).Interlayer velocity time histories were computed by in-tegrating the interlayer acceleration time histories usingthe trapezoidal rule.

The interlayer acceleration, velocity, and stress timehistories computed using SHAKE were used as basemo-tionsinaseriesoffinite difference site response analyses,

in which the acceleration time histories at the surface ofthe finite difference profiles were computed. The profilesused inthe finite difference analyses were comparable tothe SHAKE profiles down to the depths correspondingto the interlayer motions. An elastic constitutive rela-tion, with 5% Rayleigh damping, was used to model thesoil layers in the finite difference profiles. The centralfrequency of the damping relationship was setto thefun-damental frequencies of the respective finite differenceprofiles.

Fourier amplitude spectra (FAS) and 5% damped,pseudo-acceleration response spectra (PSA) were com-puted from the acceleration time histories of the surfacemotions of theSHAKE and finite difference profiles. Er-ror analyses were performed on the spectra correspond-ing to the different profiles and different types of speci-fied input motions. In the error analyses, the spectra forthe SHAKE motions were used as the correct mo-tions. The word correct does not imply that SHAKEprecisely models the behavior of an actual soil profilesubjected to earthquake motions. Rather, SHAKE givesthe analytically correct motion for a visco-elastic pro-file with constant damping applied to all frequencies ofmotion. On the other hand, the finite difference modelsused in this study give numerical approximations of the

correct analytical solution. The errors in the finite differ-ence spectral values were computed for a spectrum offrequencies using the following expressions:

ErrorPS A =PSAFLAC PSASHAKE

PSASHAKE(1a)

ErrorFAS =FASFLAC FASSHAKE

FASSHAKE(1b)

From the results of the parametric study, it was deter-mined thatthe specification of the input motion in FLAC

in terms of stress time histories gives the least accurateresults, wherein the stress time histories were applied ata quiet boundary along the base of the FLACmodel.The errors corresponding to specifying the motions interms of acceleration and velocity time histories wereessentially identical and considerably less than those as-

sociated with the stress time histories.

2.4 Development of input motions for wall analyses

As stated previously, the finite difference model of thesoil-wall system consisted of only the upper 9.1 m of a68.6 m profile. To account for the influence of the soilprofile below 9.1 m on the ground motions, the entire68.6 m profile, without the retaining wall, was modeledusing a modified version of SHAKE91. The interlayermotions at the depth corresponding to the base of thefinite difference model (i.e., 9.1 m) were computed. Theinput ground motions used in the SHAKEanalyses were

the same motions used in the parametric study discussedabove. The motions were specified as rock outcrop mo-tions at the base of the 68.6 m soil column.

The small strain fundamental frequency of the retain-ing wall-soil system in the finite difference model was es-timated to be approximately 6 Hz. At larger strains, thefundamental frequency of the system will be less thanthe small strain value. To ensure proper excitation of thewall, the cutoff frequency in the SHAKE analyses wereset at 15 Hz. This value was selected considering both thefundamental frequency of the wall-soil system and thefact that the input motions had little energy at higher fre-

quencies. The interlayer motions (at 9.1 m depth) com-puted using SHAKE were specified as acceleration timehistories along the base of the finite difference model.

2.5 Model parameters for soil

The soil was modeled using an elastic-perfectly plasticconstitutive relation, with the Mohr-Coulomb envelopedefining the yield criterion. The plastic flow rule was as-sumed to be nonassociative, as is appropriate for cohe-sionless soils. Five parameters arerequired forthis modeltobefullydescribed:effectiveinternalfrictionangle();dilatancy angle (); mass density (); small strain shear

modulus (Gmax); and bulk modulus (K). Both and are familiar to geotechnical engineers, where is the to-tal unit weight of the soil (t) divided by the coefficientof acceleration due to gravity (g), that is, = t/g. Asstated previously, for the backfill and foundation soilwere 35 and 40, respectively. These values are consis-tent with medium-dense compacted fill and dense natu-ral deposits.

is the angle of dilation that can vary from 0 to

depending on the form of the plastic potential functionused in the flow rule:

7/30/2019 j.1467-8667.2007.00538.x

5/14


ij= g(ij)

ij(2)

where ij and ij are the strain rate and stress tensors,respectively; g(ij) is the plastic potential function; and is a nonnegative multiplier for an elastic-perfectly plas-

tic material (not a material property). When =

, theplastic potential takes a form identical to the yield crite-rion (i.e., g(ij) = f(ij), where f(ij) = the yield crite-rion); hence, the flow rule becomes associative. For gran-ular soils fallingon the dense side of the critical state lineand sheared drained, as is the case for the backfill andfoundation soil, < . For the type and state of thebackfill and foundation soil being analyzed, dilation an-gles of 5 to 10 would be representative. However, it isthe authors experience that when soil is not significantlyrestrained, such as for slopes and retaining walls that canslide, the assumed dilatancy angle has little influence onthe resulting system response. Consequently, the soilwasassumed to be incompressible (i.e., = 0).

Several correlations exist that relate Gmax to othersoil parameters. However, the most direct relation is be-tween Gmax and small strain shear wave velocity (vs ):

Gmax = v2

s (3)

where vs maybe determined by various types of sitechar-acterization techniques, such as cross hole or spectralanalysis of surface waves (SASW) studies (e.g., Stokoeet al., 1994; Woods, 1994).

Values for K can be determined from Gmax and Pois-sons ratio () using the following relation:

K=2 Gmax (1 + )

3 (1 2 )(4)

in which may be estimated using the followingexpression:

=1 sin()

2 sin()(5)

which was derived from the theory of elasticity (e.g.,Terzaghi, 1943), in conjunctionwith the correlation relat-ing Ko and

proposed by Jaky (1944), that is, Ko = 1 sin(). Using the above expression, was determined

to be 0.26 and 0.3 for the foundation soil and backfill,respectively.

2.6 Model parameters for wall

The concrete wall was divided into five segments hav-ing constant parameters, as illustrated in Figure 4, witheach segment consisting of several 0.3 m elastic beamelements. Four parameters were required to define themechanical properties of the elastic beam elements:cross-sectional area (Ag); mass density (); elastic mod-

4 m

6.1m

2.4 m1.5 m

Beam

Elements

1

2

3

41.5 m

1.5 m

1.5 m

1.5 m

5

Fig. 4. Numerical model of retaining wall using elastic beamelements.

ulus (Ec); and second moment of area (I), commonlyreferred to as moment of inertia.

The basis for subdividing the wall into five segmentswas the variation of the mechanical properties along theheightof the wall. A wall havinga greater taper or largelyvarying steel reinforcement along the length of the stemor base would likely require more segments. For each ofthe segments, Ag and were readily determined from

the wall geometry and the unit weight of the concrete(i.e., 23.6 kN/m3). Ec was computed using the followingexpression (e.g., MacGregor, 1992):

Ec = 57,000

fc (6)

In this expression, fc is the compressive strength of theconcrete (e.g., 4,000 psi for the wall being modeled),and both Ec and f

c are in psi. Because the structure is

continuous in the direction perpendicular to the analysisplane, Ec that was computed using Equation (6) neededto be modified to account for plane-strain conditions.This modification was done using the following expres-

sion (Itasca, 2000):

Ec plane strain =Ec

(1 2)(7)

where 0.2 was assumed for Poissons ratio for concrete.I is a function of the geometry of the segments, the

amount and location of the reinforcing steel, and theamount of cracking in the concrete, where the latter inturn depends on the static and dynamic load imposedon the member. In dynamic analyses, it is difficult tostate a priori whether the use of sectional properties

7/30/2019 j.1467-8667.2007.00538.x

6/14


Fig. 5. Location of interface elements in the finite differencemodel.

corresponding to uncracked, fully cracked, or some in-termediate level of cracking will result in the largest de-mand on the structure. However, I= 0.4Iuncracked wasused as a reasonable estimate for the sectional properties(Paulay and Priestley, 1992).

2.7 Model parameters for wall-soil interface

Interface elements were used to model the interactionbetween the concrete retaining wall and the soil. How-ever, the finite difference code does not allow interfaceelements to be used at the intersection of branchingstructures (e.g., the intersection of the stem and base ofthe cantilever wall). Several approaches were attemptedby the authors to circumvent this limitation in finite dif-ference code, with the simplest and best approach, asfound by the authors, illustrated in Figure 5. As shownin this figure, three very short beam elements, orientedin the direction of the stem, toe side of the base, andheel side of the base, were used to model the base-stem

intersection. No interface elements were used on thesethree short beam elements. However, interface elementswere used along the other contact surfaces betweenthe soil and wall, as depicted by the hatched areas inFigure 5.

A schematic of the interface element is presented inFigure 6. As may be observed from this figure, the inter-face element hasfour parameters: S= sliderrepresentingshear strength; T= tensile strength; kn = normal stiff-ness; and ks = shear stiffness. The element allows per-manent separation and slip of the soil and the structure,

Fig. 6. Schematic of the finite difference interface element(adapted from Itasca, 2000).

as controlled by the parameters T and S, respectively.For the cohesionless soil being modeled, T= 0 (i.e., co-

hesionless soil), whereas S was specified as a function ofthe interface friction angle (). For medium-dense sandagainst concrete, = 31 (Gomez et al., 2000a).

As a rule-of-thumb, kn should be set to 10 timesthe equivalent stiffness of the stiffest neighboring zone(Itasca, 2000):

kn 10 max

K+

4

3 Gmax

zmin

(8)

In Equation (8), K and Gmax are the bulk and smallstrain shear moduli, respectively, and zmin is the small-estwidthofazoneinthenormaldirectionoftheinterfac-ing surface. The max[ ] notation indicates that the max-imum value over all zones adjacent to the interface beused. Arbitrarily large values for kn, as commonly usedin implicit finite element analyses, should not be used, asthis results in unnecessarily small time steps, and there-fore unnecessarily long computational times (Itasca,2000).

The determination of the ks required considerablymore effort than the determination of the other inter-face element parameters. In shear, the interface element

essentially is an elasto-plastic model, with an elastic stiff-ness of ks and yield strength S. The ks values wereselected such that the resulting elasto-plastic modelgave an approximate fit of the hyperbolic-type interfacemodel proposed by Gomez et al. (2000a, b). A compar-ison of the two models for initial loading (i.e., construc-tion of the wall) is shown in Figure 7.

The procedure used to determine ks values for ini-tial loading is outlined below. The reader is referred toGomez et al. (2000a, b) for more details concerning theirproposed hyperbolic-type model.

7/30/2019 j.1467-8667.2007.00538.x

7/14


0.000 0.002 0.004 0.006 0.008 0.0100

100

200

300

400

500

600

700

s (ft)

ult

f

(ps

f)

hyperbolicmodel

FLAC

model

r

ksKsi

Fig. 7. Calibration of the FLAC interface model to thehyperbolic-type model proposed by Gomez et al. (2000a,b).

1. Compute rusing the following expression.

r=f

Rf j Ksi(9a)

where,

f = n tan() (9b)

Ksi = KI w

n

Pa

nj(9c)

where Ksi is the dimensionless initial shear stiffnessof the interface; n is normal stress acting on the

interface (determined iteratively by first assuminga small value for ks and then constructing the wall); is interface friction angle = 31; Rfj is the failureratio = 0.84; KI is dimensionless interface stiffnessnumber for initial loading= 21,000; nj is dimension-less stiffness exponent = 0.8; w is unit weight ofwater in consistent units as r; and Pa is the atmo-spheric pressure in the same units as n. The valuesfor Rfj, KI, nj, and were obtained from Gomezet al. (2000b).

2. ks was computed using the following expression:

ks =

1

1

KI w

nPa

nj + Rf j rn tan() (10)The above computed ks values were used only forthe initial construction of the wall. The ks valueswere changed after the construction of the wall andprior to the application of the earthquake loadingto values consistent with the Gomez-Filz-EbelingVersion I load/unload/reload extended hyperbolicinterface model (Gomez et al., 2000a). The proce-dure used to compute ks for the cyclic loading is

outlined below. Again, the reader is referred to thecited report for more details concerning this model.

ks = Kur j w

n

Pa

nj(11a)

where,

Kur j = Ck KI (11b)

Ck = 0.5 (1 + Rf j)2 (11c)

where Kurjis the unload-reload stiffness number forinterfaces and Ck is interface stiffness ratio.

Using the above expressions, the interface stiff-nesses were computed for the interface elementsidentified in Figure 5. Although the ks for unload-reload were higher than the corresponding valuesfor initial loading (i.e., Equation (11a) vs. Equa-tion (10)), the values for kn were the same for both

initial loading and unload-reload.

2.8 Dimensions offinite difference zones

Proper dimensioning of the finite difference zones isrequired to avoid numerical distortion of propagatingground motions, in addition to accurate computationof model response. Itasca (2000) recommends that thelength of the element (l) be smaller than one-tenthto one-eighth of the wavelength () associated with thehighest frequency(fmax) component of the input motion.The basis for this recommendation is a study by Kuhle-meyer and Lysmer (1973). Interestingly, Lysmer et al.(1975) recommend lto be smaller than one-fifth the associated with fmax, and also refer to Kuhlemeyer andLysmer (1973) as the basis for the recommendation, thatis:

Itasca (2000): l

10(12a)

Lysmer et al. (1975): l

5(12b)

where is related to the shear wave velocity of the soil(vs ) and the frequency (f) of the propagating wave bythe following relation:

= vs

f(13)

Assuming that the response of the retaining wall will bedominated by shear waves, substituting Equation (12)into Equation (11a) gives:

lvs

10 fmax(14a)

or

fmax vs

10 l(14b)

7/30/2019 j.1467-8667.2007.00538.x

8/14


As may be observed from these expressions, the fi-nite difference zone with the lowest vs , for a given l,will limit the highest frequency that can pass through thezone without numerical distortion. For the finite differ-ence analyses performed in this investigation, 0.3 m by0.3 m zones were used in subgrids one and two (refer

to Figure 3). The top layer of the backfill has the low-est vs (i.e., 160 m/second). Using Equation (14b) andl= 0.3 m, the finite difference grid used in the analy-ses should adequately propagate shear waves having fre-quencies up to approximately 53 Hz. This value is wellabove the 15 Hz cutoff frequency used in the SHAKEanalysis to compute the input motion for the finite differ-ence analyses and well above the estimated fundamentalfrequency of the retaining wall-soil system being mod-eled (i.e., 6 Hz).

2.9 Damping

An elasto-plastic constitutive model, in conjunction withtheMohr-Coulombyieldcriterion,wasusedtomodelthesoil, which inherently introduces the following hystereticdamping:

D=2

1

G

Gmax

(15)

where D is the hysteretic damping; G is secant shearmodulus; and Gmax is small strain, or preyield, G. Asmay be observed from this expression, once the induceddynamic shear stresses exceed the shear strength of thesoil, the plastic deformation of the soil introduces hys-teretic damping, up to D= 2/ at large strains. However,for dynamic shear stresses less than the shear strength ofthe soil (i.e., G =Gmax), the soil behaves elastically (i.e.,D = 0).

As discussed in detail by Ni (2007), the sudden changein G at yielding and at the stress reversals causea numer-ical distortion that results in a migrating stress-strainresponse. To minimize this distortion, it is common toapply additional damping to the model, typically 2% to3% (e.g., Finn, 1988; Wang and Makdisi 1999). For theanalyses performed, 3% additional Rayleigh dampingwas applied, where Rayleigh damping provides a rela-

tively constant level of damping over a restricted rangeof frequencies. The central frequency corresponding tothe specified damping ratio is typically set to either thefundamental period (small strain) of the system beingmodeled (an inherent property of the wall-soil system)or predominant period of the system response (an in-herent property of the wall-soil system and the groundmotion). For the finite difference analyses performed,the central frequency was set equal to the small strainfundamental frequency of the retaining wall-soil system(i.e., 6 Hz).

3 COMPARISON OF FINITE DIFFERENCEAND SIMPLIFIED ANALYSES

A series of analyses were performed using the model ofthe wall-soil system described above, scaling the inputmotions to different peak ground acceleration values.

Each run required between 25 and 35 days to complete(long-run times is one of the shortcomings of explicitanalyses). To assess the applicability of simplified tech-niques that were developedfor estimatingdynamic earthpressures and permanent wall displacement of gravityretaining walls to cantilever walls, comparisons of theresults from the finite difference and simplified analysesare presented below. The reader is referred to Ebelingand Morrison (1992), Green et al. (2003), and Green andMichalowski (2006) for more detailed discussions aboutthe simplified techniques used.

3.1 Dynamic earth pressures

3.1.1 Simplified procedure for computing earth pres-

sures. The Mononobe-Okabe method for determiningseismically induced active and passive lateral earth pres-sures is based on limit equilibrium and is an exten-sion of the Coulomb theory for static stress conditions.Themethodentails three fundamental assumptions (e.g.,Seed and Whitman, 1970):

1. Wall movement is sufficient to ensure either activeor passive conditions, as the case may be.

2. The driving soil wedge inducing the lateral earth

pressures is formed by a planar failure surface start-ing at the heel of the wall and extending to the freesurface of the backfill. Along this failure plane themaximum shear strength of thebackfill is mobilized.

3. The driving soil wedge and the retaining struc-ture act as rigid bodies and, therefore, experienceuniform accelerations throughout the respectivebodies.

As demonstrated by Dr. Ignacio Arango (Seed andWhitman, 1970), the dynamic earth pressures may be de-termined from analogous static conditions. Accordingly,the Mononobe-Okabe expressions for dynamic earth

pressures can be derived from Coulombs expressionsfor static earth pressures. The analogous static condi-tions are achieved by rotating the wall-backfill systemby an angle , such that the vector sum of the horizontaland vertical inertial coefficients (kh and kv , respectively)is oriented vertically, where tan( ) = kh/(1 kv). Inregards to the mathematicalexpressions, the Mononobe-Okabe expressions can be derived from Coulombs ex-pressions by replacing the static values for the total unitweight of the soil (t), height of the wall (H), inclinationof the backfill (), and inclination of the wall face from

7/30/2019 j.1467-8667.2007.00538.x

9/14


the vertical (), with the corresponding dynamic values(i.e., td, Hd, d, and d). This substitution results in thefollowing set of equations for active conditions (i.e., khacting away from the backfill):

PAE =1

2

td H2

d KA(d, d)

=1

2 t H

2 (1 kv) KAE

(15a)

KAE =cos2( )

cos( ) cos2() cos( + +)

1

1 +

sin( + ) sin( )

cos( + + ) cos( )

2

(15b)

where, PAE is the resultant force acting on the wall.A plot of the Mononobe-Okabe active earth pressure

coefficients (i.e., KAE) as a function of the horizontal in-ertial coefficient (kh) is shown in Figure 8a for = == 0 and = 35. As shown in this figure, when kh =0 (i.e., static conditions), the value of KAE is equivalent

kh

0.0 0.2 0.4 0.6 0.8 1.00

1

2

3

4(a)

Later

alEarthPressure

Co

efficient(KAE

) M-O active

earth pressures

FD active

earth pressures

kh

0.0 0.2 0.4 0.6 0.8 1.00

1

2

LateralEarthPressure

Coefficient(KAE

)

KA

(b)

43

2

1

Fig. 8. Lateral earth pressure coefficients: (a)Mononobe-Okabe (M-O) and finite difference (FD) lateralearth pressure coefficients for = = = 0 and = 35;

and (b) Illustration of the process by which lateral stresses arelocked-in.

to Coulombs active coefficients (KA). However, as khincreases in value, KAE becomes greater than KA. Askh increases, the analogous static condition is achievedby tilting the wall forward, thus increasing the inclina-tion of the backfill and increasing the pressure inducedon the wall. The limiting pressure occurs when for the

analogous static condition, the inclination of the backfillequals the angle of internal friction (i.e., d =

). Atthis point, the failure wedge becomes infinite in size, orsynonymously, the angle of the failure plane (AE) fromthe horizontal equals the static inclination of the backfill(i.e., AE = , where AE decreases as kh increases). Forsuch a case, no sized wall could restrain the backfill frommovement.

3.1.2 Finite difference computed earth pressures on the

cantilever wall. The dynamically induced lateral earthpressures acting on the stem of the wall were computed

using finite difference. The corresponding lateral earthpressure coefficients (KFD) were computed from thesestresses using the following expression (Green et al.,2003):

KF D =2 PF D

t H2 (1 kv)(16)

where PFD is the resultant of the finite difference com-puted stresses acting on the stem of the wall; t is thetotal unit weight of the backfill; H is the height of thewall; and kv is the vertical inertial coefficient (assumedto be zero). Equation (16) inherently assumes that thelateral stress distribution is triangular, with the base ofthe triangle at the depth of the base of the wall. This as-sumption is consistent with that used in the derivation ofthe Mononobe-Okabe expressions (i.e., Equation (15a)).Equation (16) was used to compute KFD values at timescorresponding to the peaks in the time history of thehorizontal inertial coefficient (kh) acting away from thebackfill (i.e., active-type conditions), wherein spurioushigh-frequency spikes were filtered from the kh timehistory (Ni, 2007).

A plot of the computed KFD values versus kh is alsoshown in Figure 8a, labeled FD active earth pressures,for an analysis in which a motion recorded during the

1989 Mw6.9 Loma Prieta earthquake in California wasused. The Mononobe-Okabe lateral dynamic earth pres-sure coefficient (KAE) for the wall-soil system discussedaboveisalsoshowninthis figure.The reason for the devi-ationofthe finitedifference-computed stressesand thosecomputed by the Mononobe-Okabe expressions can beunderstood from examining Figure 9. Shown in this fig-ure is the deformed mesh from one of the finite differ-ence analyses, wherein the deformations are magnifiedby a factor of 3. At large values ofkh directed away fromthe backfill, the induced inertial forces on the structural

7/30/2019 j.1467-8667.2007.00538.x

10/14


Fig. 9. Annotated deformed mesh from one of the finitedifference analyses; deformations magnified by a factor of 3.

Note: Toe of wall not initially embedded. The deformed mesh shown

in this figure is only for illustration purposes. It is from an analysis

wherein the acceleration time history was scaled to a very large peak

acceleration. For this analysis, the right lateral boundary should have

been farther from the wall.

block cause it to simultaneously bend, rotate, and poten-

tially slide away from the backfill, at which time a smallwedge of soilor grabenmoves vertically downward. (Thestructural block consists of the cantilever wall and thebackfill contained within; see Figure 2.) As the direc-tion ofkh reverses (i.e., changes direction from away totoward the backfill), the graben prevents the structuralblock from returning to its undeformed shape, in effectlocking in the elastic stresses resulting from the bendingand rotation of the structural block.

This process is illustrated by the dashed arrows andcorresponding earth pressure coefficients in Figure 8b,wherein the initial stresses imposed on the stem of thewall correspond to active conditions. As kh increasesin the direction away from the backfill, the stresses onthe stem increase according to the Mononobe-Okabeexpression for active conditions (arrow 1). However,upon reversal of the direction of kh, the stresses im-posed on the stem do not decrease as predicted by theMononobe-Okabe expression, but rather remain rela-tively constant (arrow 2). As the direction ofkh reversesagain in the direction away from the backfill, the stressesacting on the stem remain relatively constant until khreaches the Mononobe-Okabe envelope (arrow 3), atwhich point the stresses increase again according to theMononobe-Okabe expression for active conditions (ar-

row 4). This stepwise increase in the locked-in stressescontinues until the residual stresses imposed on the stemcorrespond to at-rest (or Ko) conditions, while the dy-namically induced inertial stresses are superimposed onthe locked-in residual stresses. The increase in residualstresses is clearly shown in Figure 10, wherein plots areshown of both the time history ofkh and ofPFD.

The locked-in residual stresses on the wall are not re-leased by the slippage of the wall away from the backfill.This is because the driving soil wedge is not mono-lithic, but rather, in this case, consists of a graben and five

0 10 20 30 40

-0.4

-0.2

0.0

0.2

0.4

kh

PFD

(kN)

17.8

35.6

53.4

71.289.0

0.0

Stem

0 10 20 30 40

Koconditions

KA conditions

Time (sec)

Fig. 10. Time history of the horizontal inertial coefficient (kh)at approximately the center of the structural block, and the

time history of the resultant of the imposed stresses ( PFD) onthe stem of the cantilever retaining wall.

driving soil wedges (Figure 9), with the latter tending tomove downward and away from the backfill as the wallslides outward. As a result, the graben rides along withthe driving soil wedges maintaining its role of locking inthe residual stresses. The progressive increase in residualstresses was also observed in centrifuge model tests per-formed by Andersen et al. (1991), as shown in Figure 11.Further validating the multidriving soil wedge responseare the shaking table test results on a model retaining(Aitken, 1982). A post-test photo is shown in Figure 12a;for scale, the height of the retaining wall shown in this

photo is 0.32 m. The model was subjected to horizon-tal base excitations that were generated by a lever arm-springreleasemechanism.Theexcitationsweredecayingsine waves (Figure 12b), in which the firsttwo to threecy-cles exceeded acceleration required to cause sliding. Asmay be observed from Figure 12a, two driving wedges

0.00 0.04 0.08 0.12 0.16

Time (sec)

400

578

756

934

MagnitudeofE

arthForce(N)

Fig. 11. Lateral force induced on retaining wall by backfill incentrifuge model tests (adapted from Andersen et al., 1991).

7/30/2019 j.1467-8667.2007.00538.x

11/14


Fig. 12. Shaking table test results on a model retaining wall:(a) A posttest photograph showing two predominant driving

failure wedges; and (b) Excitation time history. (Adaptedfrom Aitken, 1982).

fully developed in the backfill, with the slip planes in-clined at 39.5 and 46 resulting from the first andsecond active pulses in the excitation, respectively.

3.2 Permanent wall displacement

Comparisons of the permanent relative displacements(dr) of the wall computed by finite difference and New-mark sliding block analyses (Newmark, 1965) of thestructural block (Figure 2) are shown in Figure 13. Fromthe finite difference analyses, dr was computed by sub-tracting the total displacement of the structural node at

the intersection of the stem and base of the wall fromthe total displacement of the grid point at the free-fieldboundary at the same depth. As may be observed fromFigure 13, dr computed from the finite difference analy-ses is about 0.33 m.

Newmark slidingblock analyses of thestructural blockwere performed using the acceleration time historyshown in Figure 14. This time history was computed inthe finite difference analyses at the free-field boundaryat a depth correspondingto approximately mid-height ofthestructuralblock. To performa Newmark slidingblock

0 5 10 15 20 25 30 35 40-0.10.00.1

0.20.30.4

0.50.60.7

Time (seconds)

Perm

anen

tre

lative

disp

lacemen

t(m)

Newmark

Finite

difference

(a)

Newmark

0 5 10 15 20 25 30 35 40-0.10.0

0.10.20.30.4

0.50.60.7

Time (seconds)

Pe

rmanentrelative

d

isplacement(m)

Finite

difference

(b)

Fig. 13. Comparison of the permanent relative displacementscomputed by finite difference and Newmark sliding block

analyses (Newmark, 1965): (a) Permanent relativedisplacements computed via the Newmark sliding blockprocedure assuming N g = 0.22 g; and (b) Permanent

relative displacements computed via the Newmark slidingblock procedure assuming N g = 0.27 g.

analysis, a maximum transmissible acceleration (N

g)has to be specified, which is the value of acceleration im-

parted to the block resulting in a factor of safety againstsliding equal to 1.0. Using the interface friction angle be-tween the concrete wall and foundation soil (i.e., = 31)in conjunction with the weight of the structural block,N g was determined to be approximately 0.22 g. Thesliding block analysis resulted in dr= 0.55m, as shown inFigure 13a, which is considerably larger than that fromthe finite difference analysis.

0 10 15 20 25 30 35 405

-1.0

-0.5

0.0

0.5

1.0

time (seconds)

N*g= 0.27 g N*g= 0.22 g

Acceleration(g)

Fig. 14. Acceleration time history used in the Newmarksliding block analysis of the structural block.

7/30/2019 j.1467-8667.2007.00538.x

12/14


A possible explanation for the difference in the dr val-ues is that the sliding block analysis did not account foradditional sliding resistance resulting from the plowingaction that occurs at the toe of the wall. Although thewall was not embedded in the foundation soil in its ini-tial, undeformed shape, the wall tended to rotate around

the toe as it translated away from the backfill. As a re-sult, the toe of the wall penetrated and plowed throughthe foundation soil. Such a mechanism was observed inthe deformed finite difference mesh (Figure 9). To ac-count for this additional resistance to sliding, N g wasrecomputed assuming a friction angle of 35, which isbetween the interface friction angle (i.e., = 31) andthe of the foundation soil (i.e., 40), with the revisedvalue ofN g= 0.27 g (Figure 14). A comparison of thepermanent relative displacements computed from the fi-nite difference and the sliding block analyses using therevised value ofN g is shown in Figure 13b. As may be

observed from this figure, the predicted displacementsare in very close agreement.

4 ADDITIONAL OBSERVATIONS

From an engineering perspective, the main concern for agravity retaining wall system is the global stability of thewall (i.e., sliding, overturning, and bearing capacity fail-ure). The structural design of such walls is a secondaryissue because the walls are so massive and have a rela-tively small aspect ratio in cross section. However, this isnot the case for cantilever walls, for which proper struc-tural design is just as important as global stability. Con-sequently, the critical load cases for structural design andglobal stability need to be determined in designing a can-tilever retaining wall system. A common assumption inpractice is that the critical load case for global stability,which is often determined using the Mononobe-Okabeprocedure, is also the critical load case for structural de-sign. However, the finite difference analyses showed thatthis assumption is not valid.

Figure 15 illustrates the critical load cases for bothstructural design and global stability from one of the fi-nite differenceanalyses. As shown in Figures 15aand 15c,

the critical load case for global stability occurs when thehorizontal ground acceleration is directed into the back-fill (or correspondingly, when kh is directed away fromthe backfill).For thiscase, in essence, the structural blockis treated as if it were a gravity retaining wall. Accord-ingly, the stresses and resultant force acting on the inter-face of the structural block and driving soil wedge (i.e.,the vertical section through the heel of the wall) are of in-terest. As may be observed from Figure 15a, the stressesacting on the structural block are relatively triangularlydistributed, similar to the stress distribution assumed for

-3.05 -1.52 0.00 1.52 3.05 4.57 6.10-1.52

0.00

1.52

3.05

4.57

6.10

7.62

Scaled force & stress, distance (m)

Heightfromb

ottomo

fwall(m)

(a)

PFD

stressdistribution

-3.05 -1.52 0.00 1.52 3.05 4.57 6.10

-1.52

0.00

1.52

3.05

4.57

6.10

7.62

Scaled force & stress, distance (m)

Heightfromb

ottomo

fwall(m)

(b)

PFD

stress distribution

0 5 10 15 20 25 30 35

-0.4

-0.2

0

0.2

0.4

Time (sec)

Accele

ration(g)

Towardsbackfill

Awayfrom

backfillCritical load case: Wallstructural design

Critical load case: Wallglobal stability

(c)

Fig. 15. Critical load cases: (a) for global stability of the wall;and (b) for structural design. (c) Free-field, free surface

acceleration time history computed at the elevation of thewall base. Denoted in this time history are the accelerations

and times corresponding to the critical load cases for both thestructural design and global stability.

static analyses of walls (Note: In Figure 15a, the reduc-tion in stress toward the base of the wall is due to the

arching of the soil in the failure wedge from soil belowthe failure plane to the structural block.). As a result, thepoint of action of the resultant force is around one-thirdthe height of the wall from the base.

In contrast to the critical load case for global stability,the critical case for structural design is shown in Fig-ures 15b and 15c. For this case, the stresses and resultantforce acting directly on the stem of the wall are of inter-est. As may be observed from Figure 15c, this case occurswhen the ground acceleration is directed away from thebackfill (or correspondingly, when kh is directed into the

7/30/2019 j.1467-8667.2007.00538.x

13/14


backfill). As a result, the stresses imposed on the wallby the soil are relatively uniformly distributed, and thepoint of action of the resultant force is about mid-heightof the wall.

Although the magnitudes of the respective resultantforces are approximately the same for the finite differ-

ence results shown in Figures 15a and 15b (Note: Thismay not be the case for all earthquake motions.), it isthe difference in the point of action of the resultantforces that distinguishes the critical load cases. Criticalfor structural design is the moment induced at the bot-tom of the stem. Hence, for a given resultant force, thehigher the point of action above the base of the wall,the larger the induced moment around the bottom ofthe stem. Consequently, in all the analyses performed bythe authors, the critical load cases for structural designoccurred when kh was directed toward the backfill (andhence, the soil imposed a fairly uniform stress along the

height of the stem) and the point of action of the resul-tant force was approximately mid-height of the wall.

5 SUMMARY AND CONCLUSIONS

The authors outline the details of a system model andits calibration and validation for use in computing thedynamic response of a cantilever retaining wall. The soilwas modeled using an elastic-perfectly plastic constitu-tive relation, with the Mohr-Coulomb envelope definingthe yield criterion. The plastic flow rule was assumedto be nonassociative. The wall was modeled with elasticbeam elements using a cracked second moment of area(Icracked) equal to 0.4 Iuncracked. Interface elements areused to model the wall-soil interface, wherein the inter-face element parameters are those that give a best fit ofthe Gomez et al. (2000a, b) hyperbolic interface model.The lateral stresses induced on the wall computed in thefinite difference analyses did not correspond with thosepredicted by the Mononobe-Okabe method. The reasonfor this deviation is attributed to the relative flexibility ofthe structural block and to the nonmonolithic motion ofthe driving soil wedge, bothof which violate assumptionsinherent in the Mononobe-Okabe method. The perma-nent relative displacement of the wall computed in the

finite difference analyses were in general accord withthose predicted using the Newmark sliding block proce-dure, once the plowing action at the toe was taken intoaccount. Finally, it was found that thecritical load case ofstructural design of the wall differed from that for globalstability, which is contrary to the common assumptionmade in practice that the two load cases are the same.Specifically, the critical load case for global stability oc-curs when kh is directed away from the backfill and thecritical case for structural design occurs when kh is di-rected toward the backfill.

ACKNOWLEDGMENTS

A portion of this study was funded by the Head-quarters, U.S. Army Corps of Engineers (HQUSACE)Civil Works Earthquake Engineering Research Pro-gram (EQEN). Dr. Robert M. Ebeling was the cog-

nizant Corps representative; his suggestions and com-ments throughout the project are greatly appreciated.The authors also acknowledge Itasca Consulting Group,Inc. for the academic loan of their FLACcode. This col-laboration is very much appreciated.

REFERENCES

Aitken, G. H. (1982), Seismic Response of Retaining Walls,MS Thesis, University of Canterbury, Christchurch, NewZealand.

Andersen, G. A., Whitman, R. V. & Germaine, J. T. (1991),Seismic response of rigid tilting walls, in Proceedings of Cen-

trifuge 91, Balkema, Rotterdam, The Netherlands, pp. 41724.Ebeling, R. M. & Filz, G. M. (2001), Soil-Structure Interaction

Analysis of Rock-founded Cantilever and Mass Concrete Re-taining Walls, Draft Report, US Army Corps of Engineers,Engineer Research and Development Center, Vicksburg,MS.

Ebeling, R. M. & Morrison, E. E. (1992), TheSeismic Design ofWaterfront Retaining Structures, US Army TechnicalReportITL-92-11, US Navy Technical Report NCEL TR-939, USArmy Engineer Waterways Experiment Station, Vicksburg,MS.

Finn, W. D. L. (1988), Dynamic analyses in geotechnical en-gineering, in J. L. Von Thun (ed.), Earthquake Engineeringand Soil Dynamics IIRecent Advances in Ground-Motion

Evaluation, Geotechnical Special Publication 20, ASCE, pp.52391.

Gomez, J. E., Filz, G. M. & Ebeling, R. M. (2000a), ExtendedLoad/Unload/Reload Hyperbolic Model for Interfaces: Pa-rameter Values and Model Performance for the Contact be-tween Concrete and Coarse Sand, ERDC/ITL TR-00-7, USArmy Corps of Engineers, Engineer Research and Devel-opment Center.

Gomez, J. E., Filz, G. M. & Ebeling, R. M. (2000b), Develop-ment of an Improved Numerical Model for Concrete-to-Soil

Interfaces in Soil-Structure Interaction Analyses, Report 2,Final Study, ERDC/ITL TR-99-1, US Army Corps of Engi-neers, Engineer Research and Development Center.

Green, R. A.& Ebeling,R. M.(2002), Seismic Analysis of Can-tilever Retaining Walls, Phase 1, ERDC/ITL TR-02-3, US

Army Corps of Engineers, Engineer Research and Devel-opment Center, http://www-personal.umich.edu/rugreen/papers/ITL-TR-02-3.pdf.

Green, R. A. & Michalowski, R. L. (2006), Shear band forma-tion behind retaining structures subjected to seismic excita-tion, Foundations of Civil and Environmental Engineering,7, 15769.

Green, R. A., Olgun, C. G., Ebeling, R. M. & Cameron, W.I. (2003), Seismically induced lateral earth pressures on acantilever retaining wall, in J. E. Beavers (ed.), AdvancingMitigation Technologies and Disaster Response for LifelineSystems, ASCE Technical Council on Lifeline EarthquakeEngineering Monograph No. 25, pp. 94655.

7/30/2019 j.1467-8667.2007.00538.x

14/14


Headquarters, US Army Corps of Engineers. (1989), Retainingand Flood Walls, EM 1110-2-2502, Washington DC.

Headquarters, US Army Corps of Engineers. (1992), StrengthDesign for Reinforced-Concrete Hydraulic Structures, EM1110-2-2104, Washington DC.

Idriss, I. M. & Sun, J. I. (1992), Users Manual for SHAKE91:A Computer Program for Conducting Equivalent Linear

Seismic Response Analyses of Horizontally Layered SoilDeposits, Center for Geotechnical Modeling, Departmentof Civil and Environmental Engineering, University ofCalifornia, Davis, CA.

Itasca. (2000), FLAC (Fast Lagrangian Analysis of Continua)Users Manuals, Itasca ConsultingGroup, Minneapolis, MN.

Jaky, J. (1944), The coefficient of earth pressure at rest, MagyarMenok es Epitesz Kozloi (Journal of the Society of Hungar-ian Architects and Engineers), October, 355358.

Kuhlemeyer, R. L. & Lysmer, J. (1973), Finite element methodaccuracy for wave propagation problems, Journal of theSoil Mechanics and Foundations Division, 99(SM5), 42127.

Lysmer, J., Udaka, T., Tsai, C.-F. & Seed, H. B. (1975), FLUSH:A Computer Program for Approximate 3-D Analysis of Soil-

Structure Interaction Problems, EERC Report No. EERC-75-30, Earthquake Engineering Research Center, Univer-sity of California, Berkeley, CA.

MacGregor, J. G. (1992), Reinforced Concrete Mechanics andDesign, Prentice Hall, Englewood Cliffs, NJ.

Mononobe, N. & Matsuo, H. (1929), On the determination ofearth pressures during earthquakes, in Proceedings: WorldEngineering Congress, 9, 17785.

Nadim, F. & Whitman, R. V. (1983), Seismically inducedmove-ments of retaining walls, Journal of Geotechnical Engineer-ing, ASCE, 109(7), 91531.

Newmark, N. M. (1965), Effects of earthquakes on dams andembankments, Geotechnique, 15(2), 13960.

Ni, B. (2007), Implementation of a bubble model in FLAC andits application in dynamic analysis, Ph.D. thesis, University

of Auckland, Auckland, New Zealand.

Okabe, S. (1924), General theory of earth pressures, Journalof Japan Society of Civil Engineering, 10(6), 1277323, plusfigures.

Ostadan, F. (2005), Seismic soil pressure for building walls:An updated approach, Soil Dynamics and Earthquake En-

gineering, 25, 78593.Paulay, T. & Priestley, M. J. N. (1992), Seismic Design of Re-

inforced Concrete and Masonry Buildings, John Wiley andSons, New York, NY.

Richards, R., Jr. & Elms, D. G. (1979), Seismic behavior ofgravity retaining walls,Journal of Geotechnical Engineering,

ASCE, 105(4), 44969.Seed, H. B. & Whitman, R. V. (1970), Design of Earth Retain-

ing Structures for Dynamic Loads, Lateral Stresses in theGround and Design of Earth-Retaining Structures, ASCE,10347.

Stokoe, K. H., II, Wright, S. G., Bay, J. A. & Roesset, J.M. (1994), Characterization of geotechnical sites by SASWmethod, in R. D. Woods (ed.), Geophysical Characterizationof Sites, Oxford and IBH Publishing, New Delhi, 1525.

Terzaghi, K. (1943), Theoretical Soil Mechanics, John Wileyand Sons, New York, NY.

Veletsos, A. & Younan, A. H. (1994), Dynamic soil pressure onrigid vertical walls, Earthquake Engineering and StructuralDynamics, 23, 275301.

Wang, Z. L. & Makdisi, F. I. (1999), Implementing a boundarysurface hypoplasticity model for sand into FLAC program,in Detoumay & Hart (eds.), FLAC and Numerical Modelingin Geomechanics, Balkema, Rotterdam, pp. 48390.

Wood, J. H. (1973), Earthquakeinduced soil pressures on struc-tures, Ph.D. thesis, EERL73-05, CaliforniaInstituteof Tech-nology, Pasadena, CA.

Woods, R. D. (1994), Borehole methods in shallow seismic ex-ploration, in R. D. Woods (ed.), Geophysical Characteriza-tion of Sites, Oxford and IBH Publishing, New Delhi, pp.91100.

Wu, G. & Finn,W.D. L. (1999), Seismic lateral pressuresfor de-

sign of rigid walls, Canadian Geotechnical Journal, 36, 50922.

Documents

j.1467-8667.2007.00538.x