GEOTECHNICAL SPECIAL PUBLICATION NO.

80
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

ANAlYSIS AND DESIGN

OF RETAINING
STRUCTURES AGAINST
EARTHQUAK
y
r-, t
- f.
I
! \ I
X~
/ ' He c-••, [x..,+nHa]nH

H
/ \c x, ~ -2
~ ... a
) k.fx,+nHe]nH
1

TI I
I
.4;.
j_.\'c.h cat [x,+% HB J
I tlljll •

· f ~i.\1 k. [ x,+% He J nH
e

I
I
j ', e
I
l
L--:---_-~- l x l
-c.te c.,x, •
"k.e 6
EDITBJ BY SHAMSHER PRAKASH

@seismicisolation
@seismicisolation
 GEOTECHNICAL SPECIAL PUBLICATION NO. 80

ANAlYSIS AND DESIGN

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

OF RETAINING
STRUCTURES AGAINST
EARTHQUAKES
Proceedngs of seal• sponsored by the
Soli Dynamics Committee of The Geo-lnslltute
of the American SOciety or Civil Engineers
In conJunction With the ASCE Nallonal Convenllon
. In W8shlngton, DC

November 1D-14, 1998

EDITED BY SHAMSHER PRAKASH

Published by

~
•sc•r S
AmertCIIn soel•ty
of Civil EnglnHrs

345 East 47th Street

New York, New York 10017-2398
@seismicisolation
@seismicisolation
 Abstract:
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

This proceedings, Analysis and Design ofRetaining Structures against Earthquakes, contains
both invited and contributed papers, which focus on the questions of 1) dynamic earth pressures
on fixed and movable rigid and flexible walls; 2) displacements in translation and rotation of
walls under earthquakes; 3) behavior offi!Js an.d abutments during earthquake; and 4) centrifuge
tests on walls. Both analytical and experimental data have been presented on possible behavior
of retaining structures under seismic loading. A study of this volume and other published
literature shows considerable effort is being devoted to determination of realistic dynamic
pressures, displacement in translation and rotation of retaining structures and behavior of fills for
abutments. Since a synthesis of these studies is not currently available, there are no unified and
generally acceptable solutions to the above questions. However, the discussions and
presentations of the papers during the session does highlight the need for such solutions and
more definite descriptions of unsolved problems.

Library of Congress Cataloging-in-Publication Data

· Analysis and design of retaining structures against earthquakes : proceedings of sessions

sponsored by the Soil Dynamics Committee of the ASCE Geotechnical Engineering Division in
conjunction with the ASCE Convention in Washington, D.C., November 10-14, 1996 I edited by
Shamsher Prakash.
p. em. -- (Geotechnical special publication ; no. 60)
Includes bibliographical references and indexes.
ISBN 0-7844-0206-X
I. Retaining walls--Design and construction--Congresses. 2. Earthquake resistant design--
Congresses. 3. Structural analysis (Engineering)--Congresses. I. Prakash, Shamsher. II.
American Society of Civil Engineers. Geotechnical Engineering Division. Soil Dynamics
Committee. IH. ASCE National Convention (1996 : Washington, D.C.) IV. Series.
TA77.A53 1996 96-44715
624.1'64--dc20 CIP

The Society is not responsible for any statements made or opinions expressed in its publications.

Photocopies. Authorization to photocopy material for internal or personal use under

circumstances not falling within the fair use provisions of the Copyright Act is granted by ASCE
to libraries and other users registered with the Copyright Clearance Center (CCC) Transactional
Reporting Service, provided that the base fee of $4.00 per article plus $.25 per page is paid
directly to CCC, 222 Rosewood, Drive, Danvers, MA 01923. The identification for ASCE
Books is 0-7844-0206-X/96/$4.00 + $.25 per copy. Requests for special permission or bulk
copying should be addressed to Permissions & Copyright Dept., ASCE.

Copyright © 1996 by the American Society of Ci vii Engineers,

All Rights Reserved.
Library of Congress Catalog Card No: 96-44715
ISBN 0-7844-0206-X
Manufactured in the United States of America.

@seismicisolation
@seismicisolation
 GEOTECHNICAL SPECIAL PUBLICATIONS
I) TERZAGID LECTURES
2) GEOTECHNICAL ASPECTS OF STIFF AND HARD CLAYS
3) LANDSLIDE DAMS: PROCESSES, RISK, AND MITIGATION
4) TIEBACKS FOR BULKHEADS
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

5) SETTLEMENT OF SHALLOW FOUNDATION ON COHESIONLESS

SOILS: DESIGN AND PERFORMANCE
6) USE OF IN SITU TESTS IN GEOTECHNICAL ENGINEERING
7) TIMBER BULKHEADS
8) FOUNDATIONS FOR TRANSMISSION LINE TOWERS
9) FOUNDATIONS AND EXCAVATIONS IN DECOMPOSED ROCK OF
THE PIEDMONT PROVINCE
10) ENGINEERING ASPECTS OF SOIL EROSION, DISPERSIVE CLAYS
AND LOESS
II) DYNAMIC RESPONSE OF PILE FOUNDATIONS- EXPERIMENT,
ANALYSIS AND OBSERVATION
12) SOIL IMPROVEMENT- A TEN YEAR UPDATE
13) GEOTECHNICAL PRACTICE FOR SOLID WASTE DISPOSAL '87
14) GEOTECHNICAL ASPECTS OF KARST TERRAINS
15) MEASURED PERFORMANCE SHALLOW FOUNDATIONS
16) SPECIAL TOPICS IN FOUNDATIONS
17) SOIL PROPERTIES EVALUATION FROM CENTRIFUGAL MODELS
18) GEOSYNTHETICS FOR SOIL IMPROVEMENT
19) MINE INDUCED SUBSIDENCE: EFFECTS ON ENGINEERED
STRUCTURES
20) EARTHQUAKE ENGINEERING & SOIL DYNAMICS (II)
21) HYDRAULIC FILL STRUCTURES
22) FOUNDATION ENGINEERING
23) PREDICTED AND OBSERVED AXIAL BEHAVIOR OF PILES
24) RESILIENT MODULI OF SOILS: LABORATORY CONDITIONS
25) DESIGN AND PERFORMANCE OF EARTH RETAINING STRUCTURES
26) WASTE CONTAINMENT SYSTEMS: CONSTRUCTION, REGULATION,
AND PERFORMANCE
27) GEOTECHNICAL ENGINEERING CONGRESS
28) DETECTION OF AND CONSTRUCTION AT THE SOIUROCK
INTERFACE
29) RECENT ADVANCES IN INSTRUMENTATION, DATA ACQUISITION
AND TESTING IN SOIL DYNAMICS
30) GROUTING, SOIL IMPROVEMENT AND GEOSYNTHETICS
31) STABILITY AND PERFORMANCE OF SLOPES AND EMBANKMENTS II
(A 25-YEAR PERSPECTIVE)
32) EMBANKMENT DAMS-JAMES L. SHERARD CONTRIBUTIONS
33) EXCAVATION AND SUPPORT FOR THE URBAN INFRASTRUCTURE
34) PILES UNDER DYNAMIC LOADS
35) GEOTECHNICAL PRACTICE IN DAM REHABILITATION
36) FLY ASH FOR SOIL IMPROVEMENT
37) ADVANCES IN SITE CHARACTERIZATION: DATA ACQUISITION,
DATA MANAGEMENT AND DATA INTERPRETATION
38) DESIGN AND PERFORMANCE OF DEEP FOUNDATIONS: PILES AND
PIERS IN SOIL AND SOFT ROCK
39) UNSATURATED SOILS
40) VERTICAL AND HORIZONTAL DEFORMATIONS OF FOUNDATIONS
AND EMBANKMENTS

@seismicisolation
@seismicisolation
 41) PREDICTED AND MEASURED BEHAVIOR OF FIVE SPREAD
FOOTINGS ON SAND
42) SERVICEABILITY OF EARTH RETAINING STRUCTURES
43) FRACTURE MECHANICS APPLIED TO GEOTECHNICAL ENGINEER-
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

ING
44) GROUND FAILURES UNDER SEISMIC CONDITIONS
45) IN-SITU DEEP SOIL IMPROVEMENT
46) GEOENVIRONMENT 2000
47) GEO-ENVIRONMENTAL ISSUES FACING THE AMERICAS
48) SOIL SUCTION APPLICATIONS IN GEOTECHNICAL ENGINEERING
49) SOIL IMPROVEMENT FOR EARTHQUAKE HAZARD MITIGATION
50) FOUNDATION UPGRADING AND REPAIR FOR INFRASTRUCTURE
IMPROVEMENT
5 L) PERFORMANCE OF DEEP FOUNDATIONS UNDER SEISMIC LOADING
52) LANDSLIDES UNDER STATIC AND DYNAMIC CONDITIONS -
ANALYSIS, MONITORING, AND MITIGATION
53) LANDFILL CLOSURES- ENVIRONMENTAL PROTECTION AND
LAND RECOVERY
54) EARTHQUAKE DESIGN AND PERFORMANCE OF SOLID WASTE
LANDFILLS
55) EARTHQUAKE-INDUCED MOVEMENTS AND SEISMIC REMEDIATION
OF EXISTING FOUNDATIONS AND ABUTMENTS
56) STATIC AND DYNAMIC PROPERTIES OF GRAVELLY SOILS
57) VERIFICATION OF GEOTECHNICAL GROUTING
58) UNCERTAINTY IN THE GEOLOGIC ENVORONMENT
59) ENGINEERED CONTAMINATED SOILS AND INTERACTION OF SOIL
GEOMEMBRANES
60) ANALYSIS AND DESIGN OF RETAINING STRUCTURES AGAINST
EARTHQUAKES
61) MEASURING AND MODELING TIME DEPENDENT SOIL BEHAVIOR
62) CASE HISTORIES OF GEOPHYSICS APPLIED TO CIVIL ENGINEERING
AND PUBLIC POLICY
63) DESIGN WITH RESIDUAL MATERIALS; GEOTECHNICAL AND
CONSTRUCTION CONSIDERATIONS

@seismicisolation
@seismicisolation
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

PREFACE

Retaining structure may be subjected to seismi c loads and experience either deformations
and/or increased earth pressures. These structures may be either rigid-masonry free -s tanding
retaining wall s for highways, sections of wing walls for hydraulic structures, and abutments
of bridges. Flexible retaining structures are most often reinforced earth sections.
Rigid structures have been designed for dynamic earth pressures most of the time. A sol u-
tion for displacements of rigid retaining walls was obtained by our Research Group at the
University of Roorkee (India) in I 974 for the first time. Subsequently for the last two de-
cades, their displacements were analyzed and designs based on permissibl e displacements
were attempted. We have not reached a stage where acceptable analytical too ls and design
procedures for such structures are available. Studies on such structures during earthquakes is
difficult because earthquakes cannot be made to order! Therefore, recourse is m ade to alter-
nate studies, e.g. on centrifuge and, shake table models, and analytical studies and their com-
parison with performance records.
Thus this session is organized with the objective to identify the state of practice in a nalysis
and design of retaining structures under dynamic loads and address the unsolved issues . The
papers were, therefore, invited from authors both within and outside the USA. There are 5
contributions from the US and 3 from overseas. This session was he ld at the ASCE Fall
Convention in Washington DC on November 12, 1996 and was sponsored by the Soil Dy-
nanJics Committee of the Geotechnical Engineeri ng Division of ASCE.
It is the current practice of the Geotechni cal Engineeri ng Division that each paper pub-
lished in a Geotechni cal Special Publication (GSP) be reviewed for its content and quality.
These special technical publications are intended to reinforce the programs presented at con-
vention sessions or specialty conferences and to contain papers that are timely and may be
controversial to some extent. Because of the need to have the GSP avail able at the conven-
tion, time available for reviews is generally not as long and reviews may not be as compre-
hensive as those given to papers submitted to the Journ al of the Division. Still we had orga-
nized our review process in a timely manner. At least 2-positive rev iews were obtained for
each paper accepted for publication and discussion. In fact one paper will be published in the
Journal based on our review process but was presented at this session. Thus there is hardly
any difference in the purpose and technical status of contributions to this geotechnica l special
publication as compared to those in the Journal.
In accordance with ASCE policy, all papers published in this vol ume are e lig ible for dis-
cussion in the Journal of the Geotechnical Engineering Division and are eligib le for ASCE
awards. Reviews of papers published in this volume were conducted by the Soil Dynamics
Committee of the Geotechnical Engineering Division. The following committee members or
cooperating persons from the general membership reviewed these papers:

T. Crespell ani Panos Dakoulas

Ahmed W.M. Elgamal W.D. Liam Finn
Steve Kramer Sanjeev Kumar
Jeen-Shang Lin Farrokh Nadim
Rowland Richards, Jr. Glenn J. Rix
Raj Siddharthan Cetin Soydemir
A.S. Veletsos Mishac Yegian

@seismicisolation
@seismicisolation
 Personal thanks go to Panos Dakoulas, Chairman of the Soil Dynamics Committee, for his
help and support in organizing this session. I also want to thank all the experts who gave both
the time and effort in reviewing the papers. Last but not least, thanks are due to all the authors
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

who kindly accepted the invitation to contribute to this volume and to the session in Washing-
ton D.C.

Shamsher Prakash, F. ASCE

Professor of Civil Engineering
University of Missouri-Rolla
Rolla, Missouri
August 28, 1996
Session Organizer and Editor

vi

@seismicisolation
@seismicisolation
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

CONTENTS

Seismic Pressures against Rigid Walls

G. Wu and W.D. Liam Finn .... ......... ...... .......................... ............... ... ............. ......... ............ I
Dynamic Response of Cantilever Retaining Wa:lls
A.S. Veletsos and A.H. Younan ............ .... ........ ...... .... ............... ..................... :.... ...... .. .... .. 19
On Seismic Displacements of Rigid Retaining Walls
Y. Wu and Sharnshet Prakash ............................ ............ .. .... ........ ...... .... ...... ...................... 21
Rotation of Large Gravity Walls on Rigid Foundations Under Seismic Loading
R.S. Steedman and X. Zeng ...... .. .. .. .... ...... .................. .. .... ............ ..... ...... ......................... 38
In-Situ Dynamic Response of Cantilever Walls
Sreenivas Alampalli and Ahmed W.M. Elgarnal ............................ ...... .. ............ .......... .. ... 57
Seismic Analysis and Model Studies of Bridge Abutments
K.L. Fishman and R. Richards, Jr....................... ...................................... ................ .. ...... 77
Earthquake-Induced Ground Settlements of Bridge Abutment Fills
Raj V. Siddharthan and Mahmoud El-Gamal ........ .............. .. ..................................... .. ... 100
Earthquake Destructiveness Potential Factor and Permanent Displacements
of Gravity Retaining Walls
T. Crespellani, C. Madiai and G. Vannucchi ............................ ..... .... ............................ .. 124

Subject I ndex ...................................... ......... .............. ...................................... ................. 135

Author Index ....... .. .. ... ......... ,....... .......... ..... ............ ............... ... ... ... .......................... ......... 136

vii

@seismicisolation
@seismicisolation
 Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

@seismicisolation
@seismicisolation
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

SEISMIC PRESSURES AGAINST RIGID WALLS

Guoxi Wul and W.D. Liam Finn2

ABSTRACT

Simplified linear elastic analytical solutions are presented for the seismic
pressures against rigid walls which agree closely with the exact solutions presented
by Wood (1973). The finite element method is used to extend the analyses to
nonhomogeneous elastic materials and to nonlinear soils. The finite element
analyses give almost exact solutions for the elastic cases. Some practical guidelines
are given for estimating the dynamic pressures for use in practice.

INTRODUCTION

Seismic pressures against retaining walls are usually determined using the
Mononobe-Okabe method (Mononobe and Matsuo, 1929; Okabe, 1924). This
method is based on the assumption that a wedge of soil bounded by the wall and the
shear failure plane in the backfill moves as a rigid body under the peak vertical and
horizontal ground accelerations or designated fractions thereof. Most commonly
only horizontal inertia forces are included.

The Mononobe-Okabe method is based on the assumption that the wall can
displace enough to permit a failure plane. Rigid walls, such as deep basement walls,
do not satisfy the displacement criteria for shear plane development, and therefore
the Mononobe-Okabe approach cannot be used. Matsuo and Ohara (1960)
formulated an elastic solution to the seismic pressures against a rigid wall, but did
not present any numerical values. Wood (1973) developed an exact analytical plane
strain solution assuming elastic response of a uniform backfill. The solution is quite
complicated and is usually applied approximately. Significant errors in estimating

1 AGRA Earth & Environmental, 2227 Douglas Road, Burnaby, B.C.

2 Professor, Department of Civil Engineering, University of British Columbia, Vancouver, B.C.

@seismicisolation
@seismicisolation
 2 RETAIN1NG STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

the dynamic thrust can result from using the response spectrum method proposed by
Wood (1973) for earthquake loading, as will be shown later. Arias et al. (1981)
developed a simple model of the elastic backfill and developed analytical expressions
for the response of the wall to both harmonic and seismic excitation. This model is
a modification of a shear beam model by including horizontal normal stresses. In
their analyses, they assumed that no seismic stresses would be developed in the
vertical direction. Veletsos and Younan (1994) developed a method based on the
equations of elasticity with the assumption of no vertical stresses and applicable to a
homogeneous semi-infinite backfill. All of these solutions are limited to . elastic
response and, therefore, are not directly applicable to backfills under very strong
shaking where the response is likely to be nonlinear.

A general model is presented here that includes all previous models as

special cases (Wu, 1994; Finn et a!., 1994). The model is applicable to both finite
and semi-infinite backfills. An analytical solution is presented for a uniform elastic
backfill and a finite element solution for nonhomogeneous elastic and nonlinear
backfills.

The elastic analytical solutions are validated by comparison with the exact
solutions by Wood (1973). Comparisons with the other approximate elastic
solutions show that the modified shear beam model of Arias (1981) gives results
closer to the Wood (1973) solution over all ranges of significant variables (Wu,
1994; Finn et al., 1994). Veletsos eta!. (1995) also adopted the simple Arias (1981)
shear model in their analyses of rigid walls with backfills of finite length rather than
the model with a more exact expression for shear stress used by Veletsos and
Younan (1994) .

FORMULATION OF ANALYSIS

Figure 1(a) shows the geometry of the plane strain problem and associated
boundary conditions. A uniform elastic soil layer is confined by two vertical rigid
walls and a rigid base. The soil layer has a total length of 2L and height of H.
When subjected to horizontal seismic body forces, the soil layer in the system
generates an antisymmetric field of horizontal normal stresses, cr,, with cr, = 0 at
x = L. Therefore, the original wall-soil problem can be reduced to the system in
Fig. 1(b) for all subsequent analyses. The ground acceleration is input at the base of
the wall-soil system.

Although the problem involves two displacement components, horizontal

and vertical components u and v, only the horizontal displacements u have been
taken into account in the analysis in order to simplify the solution of the problem. It
will be shown that this simplification does not induce a significant error in the
solution and that reliable estimates of the dynamic earth thrust can be obtained.

@seismicisolation
@seismicisolation
 SEISMIC PRESSURES AGAINST RIGID WALLS 3
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

·Y liu/liv-0

(a)
homogeneous elastic soil
(plane strain)
H u=O u=O

u=O
. ( X .X.X.&..&.~ . ~_...._. •

2L

ou/oy =0
(b) homogenous elastic soil
( plane strain )
liu/ox. = o

u=O X

Fig. I. Definition of rigid-wall problem: (a) original problem; and (b) equivalent
problem by using anti symmetric condition.

Initially, the soil is assumed to be homogeneous, isotropic and elastic, with a

shear modulus G, Poisson's ratio v, and mass density p. The equations of dyilamic
force equilibrium for the backfill in the horizontal and vertical directions are written
as

(1)

(2)

@seismicisolation
@seismicisolation
 4 RETAINING STRUCTURES ANALYSIS AND DESIGN

where ax and cry are the normal stresses in the X and Y directions and 'txy is the shear
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

stress in the X-Y plane. The displacements in the X and Y directions are u and v,
respectively.

For two-dimensional plane strain conditions, the stress components are

related to the displacements by

a =2G
•
au
- - [ (1-v)-+v-
1- 2v ax
av]
fly
(3)

a = 2G av
- - [ (1-v)-+v- au] (4)
Y 1-2v By ax

't =o(au + Bv) (5)

xy By ax

Since only the horizontal displacements u are taken into account in the analysis, only
the equation of dynamic force equilibrium in the horizontal direction, Eq. (1), is
used. Considering various forms of approximation to the problem, the governing
equation of the undamped free vibration of the backfill in the horizontal direction
can be written as

(6)

and the normal stress ax is given by

(j
x
=AG -au
,...ax

where pis the mass density of the soil backfill, tis time, and 8 and 13 are functions of
Poisson's ratio v (Wu, 1994).

e
The precise expressions for and p depend on ·the approximations used to
model the wall-soil system. Three cases are examined and their corresponding
expressions for 8 and 13 are given below.

@seismicisolation
@seismicisolation
 SEISMIC PRESSURES AGAJNST RIGID WALLS 5

Proposed Modified Shear Beam Model

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

In the p~oposed model, the shear stresses are modelled using the shear beam
analogy. Therefore, shear stress "txy is given by

au (8)
' xy = G -Oy

The normal stress cr., after assuming cry= 0 in the backfill, is found to be

cr =_2_ Gau (9)

X 1-V ax

Comparison with Eq. (7) gives

~ =-2- (10)
1- v

Substituting Eq. (8) and E q. (9) into Eq. (1) and comparing with Eq. (6),
one finds,

2
8 ::: - (11)
1- v

The values of 8 and ~ may be derived in a similar manner for other

assumptions. Values are given below for two other models; cry = 0 which was
adopted by Veletsos and Younan (1994) and v = 0.

v=O Model

In this case

8 =~ = 2(1 - v) (12)
1- 2v

Qy=OModel

In this case,

@seismicisolation
@seismicisolation
 6 RETAIN1NG STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

2- v
8=- (13)
1- v

2
13=-
1- v
(14)

The dynamic response of the wall-soil system for these three cases can be
represented by the saine equation, Eq. (6), using appropriate values for 8.
Therefore, the general derivation of dynamic solutions proceeds from Eq. (6).

SOLUTIONS FOR LlNEAR ELASTIC CASE: UNIFORM BACKFILL

The equation of motion for free vibration of the system is

(6 his)

This equation may be solved by the separation of variables technique.

Assume that the displacement solution has the form,

For the given boundary conditions, the solution reduces to

(16)

in which the mode shapes cl>.,.,(x,y) are given by

(17)

b = (2n - l)1t .
n = 1,2, ... (18)
n 2H '

(2m -1)7t
and am = L ; m = l,2, ... (19)
2

@seismicisolation
@seismicisolation
 SEISMIC PRESSURES AGAINST RIGID WALLS 7
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

The natural frequencies of the system are given by

(20)

The frequency of the first mode is given by

(21)

When the system is subjected to a ground acceleration ii 0 (t), the governing

equation for undamped vibration becomes

(22)

Modal solutions are given by

(23)

where the modal participation factor <Xmn is given by

(24)

or
16
a =------
2
(25)
mn (2m - 1)(2n - 1)7t

Damped solutions are obtained by introducing appropriate equivalent

viscous damping directly into the modal equations. IfY!IUI(t) = a.nnfmn(t), then

(26)

@seismicisolation
@seismicisolation
 8 RETAINING STRUCTURES ANALYSIS AND DESIGN

For a given ground excitation ii 0 (t) , a solution for the system is

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

u(x, y, t) = 2:2: ·am.fmn (t) sinam x ·sin bs (27)

m n

where fmn(t) is the time history solution of Eq. (26) corresponding to a particular
modal frequency <Onm· The dynamic earth pressure acting on the wall is the
compressive stress, cr., at x=O. Therefore, the dynamic pressure distribution along
the wall is

cr xlx=o = p(x, y, t) ••0 = l3G :1 x• O

(28)

The total dynamic thrust acting on the wall is

II
P(t) =f0 p(x, y, t),. 0 • dy (29)

(30)

or
(31)

The peak dynamic thrust on the wall for harmonic or earthquake motion can be
determined as accurately as desired using mode superposition. Studies were made
to determine the rate of convergence of the mode superposition method. Generally,
convergence to a solution acceptable for engineering purposes is very fast. For
walls with UH = 5.0, 75% accuracy is achieved using just the first mode; for L/H =
1.5, 80% accuracy was achieved. Generally, 95% accuracy was achieved by using
just the first ten modes. However, for the studies described later, 600 modes were
used in order to achieve an 'exact' solution for assessing the accuracy of the
approximate method compared to the exact method ofWood (1973).

Response Spectrum Method

For earthquake motion, the modal spectral thrust acting on the wall
associated with a frequency romn is determined using response spectral velocity S~ ,

@seismicisolation
@seismicisolation
 SEISMIC PRESSURES AGAINST RIGID WALLS 9
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

p
mn
= "'"""am.
J3G ".t.J.t.J
m n
a.mn
_ __
bn
-s~·.
-,
CO mn
m,n= 1,2, . . (32)

or

(33)

where

C0 2 mn = G[{(2n-l)7t}
p 2H
2
+-2
1- v
{(2m-1)7t}
2L
2
]
(34)

The peak dynamic thrust can be estimated by taking the appropriate

summation of the modal spectral thrusts for all modes considered. Two commonly
used summation methods are the square root of the sum of the squares of the peak
modal contributions and the sum of the absolute values of the contributions.

In the limited studies completed so far (Wu, 1994), the summation of the
absolute values of the peak modal contributions gave the best approximation to the
total thrust against the wall for input frequencies less than the fundamental
frequency of the backfill. In this range, the peak thrust was overestimated by about
20% for earthquake excitations using long backfills, L/H ~ 5.0. For higher
frequencies, the peak dynamic thrust may be overestimated by as much as SO%. The
square root of the sum of the squares method appears to underestimate the response
and the errors are somewhat larger. For harmonic input, the response spectrum
method gives exact results for input frequencies less than the fundamental frequency
of the backfill, but for greater frequencies the error can be greater than 50%.

EVALUATION OF MODELS

Static 1-g solutions for horizontal forces based on the above models are used
to test the various models against the exact solution by Wood (1973).

Horizontal Static 1-g Solutions

For a 1-g static horizontal gravity force, the function fmn(t) is found to be

(35)

The corresponding 1-g static thrust is obtained as

@seismicisolation
@seismicisolation
 10 RETAINING STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

(36)

The acting point of the static thrust, corresponding to a particular mode

number b., is located at (sinb~H)Ib. above the base of the wall. In particular, for
n = I it is found to be 0.637 H above the base of the wall.

For a given wall with known soil properties of the backfill, the total
horizontal thrust against the wall due to 1-g static horizontal gravity force is easily
determined by doing a double-summation for modes m and n in Eq. (16). To
generalize the solution of the problem, the concept of a normalized thrust ratio is
introduced. This ratio is defmed by

TOTAL THRUST
THRUST RATIO (37)
pH2Amu

where Amu is the peak ground acceleration in consistent units.

Comparisons with Wood (1973) Solution

Static 1-g solutions were conducted on walls with LIH = 5.0 and L1H = 1.5
using the models described above. The solutions are compared with the exact
Wood (1973) solutions (Wu, 1994; Finn et al., 1994).

The proposed modified shear beam model gives results that are in very good
agreement with the Wood (1973) solutions, as shown in Figs. 2(a) for LIH = 5.0 and
2(b) for L/H = 1.5 . The cry = 0 model yields results that in very good agreement
also when L1H = 5.0. Solutions for LIH ~ 5.0 are almost identical to the solution
for the semi-infinite backfill for which the Veletsos and Younan (1994) solution (cry
= 0) was obtained. For LIH :s;; 1.5, the cry = 0 model is less satisfactory.

The v = 0 model is quite accurate as long as the Poisson ratio v < 0.3. For
higher values of v, the solutions begin to deviate significantly from the exact
solution, with the accuracy depending the LIH ratio. For LIH = 5.0, the results from
the v = 0 model become unacceptable for v > 0.4.

These evaluation studies suggest that the proposed modified shear beam
model gives the best approximation to solutions for rigid wall systems with both
infinite and finite backfills. Therefore, this model is used for all further studies in
this paper.

@seismicisolation
@seismicisolation
 SEISMIC PRESSURES AGAINST RIGID WALLS 11
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

2.0
(o) L/H=S:O
- - Wood's exact
proposed model
(jy = 0
_v =0

O.<Q.o 0 .5
Poisson's ratio

2.0 ~----------------------------------------~
(b) L/H=1.5

0 1.5 >- Wood "s exact

~ proposed model
a:: (j' = 0 -----

~ 1.0 L~------------v•-~="_"~~-"-~-~-~-~-~-~-~~~~~-
-~-~--~-~~--~~~--~~-~~~--~~l­
I
1- 0.5 1-

I I I I
0 .1 0.2 0.3 0.4 0.5
Poisson's ratio

Fig. 2. Comparison of approximate solutions for rigid-wall systems with exact

Wood (1973) solution: (a) L/H=5.0; and (b) L/H= l.S .

DYNAMIC THRUST ON WALL

Uniform Backfill

The rigid wall system shown in Fig. 1 was subjected to both harmonic and
earthquake motions and the resulting dynamic thrusts against the wall were
evaluated for two ratios UH = 5.0 and LIH = 1.5.

@seismicisolation
@seismicisolation
 12 RETAINING STRUCTURES ANALYSIS AND DESIGN

The studies were conducted on a wall with H = 10m backfilled by an elastic

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

soil with a unit weighty= 19.6 kN/m3 and a Poisson's ratio v = 0.4. The dynamic
thrusts were evaluated as a function of frequency ratios fR1 and fR2, where fR 1 = fi/f.
and fR2 = fi/£11 , where t; =frequency of the input motion, f. = fundamental frequency
ofthe backfill, and f 11 = first natUral frequency of the soil-wall system.

The frequency of the harmonic input is easily controlled. The input

frequency of the earthquake motion is defined as the predominant frequency of the
record. To obtain a distribution of fu for the earthquake motions, the shear
modulus of the soil layer was varied. Two different earthquake motions were used;
El Centro (1940) NS component with A.r..x = 0.348 g and the Yerba Buena EW
component of the Lorna Prieta earthquake of 1989 with Anux = 0.067 g. The
corresponding predominant frequencies are 1.85 Hz and 1.61 Hz, respectively.

The dynamic thrusts are shown in Figs. 3(a) and 3(b) as a function of the
frequency ratio fRt · The data points are for both earthquakes. The peak dynamic
thrust ratios resulting from seismic input are much less than those from sinusoidal
input because there is less opportunity for resonant response to build up.

The dynamic thrust ratios from the earthquake motions are approximately
1.0 for frequency ratios less than 1.5. This suggests that the peak thrust, Pnux, can
be taken as, P.... = pH2 A...,, where p is the unit weight and A.r..x is the peak
acceleration. However, for frequency ratios greater than 1.5 to 2.0, the 1-g static
solution may be very conservative. More studies are needed to see if these findings
are of general applicability.

Non-Homogeneous Backfill: Parabolic Modulus Variation

Similar studies were conducted for non-homogeneous backfills in which the

shear modulus varied parabolically with depth. The shear modulus is usually
considered to vary as the square root of the vertical effective confining pressure
which is equivalent to a parabolic variation. The analyses were conducted using a
finite element representation of the equations of motion in Eq. (6). A six-node finite
element with six horizontal displacement variables (Fig. 4), was developed especially
for this study. The displacement field has a linear variation in the vertical direction
and a quadratic variation along the horizontal direction. The finite element
formulation was used to compute the dynamic thrust ratio as a function of time
using the El Centro record scaled to 0.07 g as input. The finite element response is
compared with the results from the closed form solution in Fig. 5. The results are
almost identical, confirming the accuracy of the finite element representation.

The dynamic thrust ratios against the wall were calculated as a function of
the frequency ratio, fR2, for (a) sinusoidal input, and (b) El Centro ground motions,
using a parabolic distribution of shear modulus with depth and elastic analysis. The
thrust ratios are shown in Fig. 6. For frequency ratios less than 1, the seismic input

@seismicisolation
@seismicisolation
 SEISMIC PRESSURES AGAINST RIGID WALLS 13

4
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Poisson's ratio : 0 .4
damping ratio 1015
o3
~
0:
- - sinusoidal motions
a.P.P.P.P earthquake motions
t;:;2
:::>
0:
I
1-1
... ----- ... _
0
0 2 3 4 5
FREQUENCY RATIO, fR1

4~--------------------------------------.
Poisson's ratio = 0 .4
(b) L/H= 1.5 damping ratio 1 0515

o3
~
sinusoidal motions
·~-·~~ earthquake motions

t;:;2
:::>
0:
I
~--,

0 2 3 4 5
FREQUENCY RATIO, fR1

Fig. 3. Normalized thrust ratios versus fR1 for sinusoidal and earthquake
motions: (a) L/H=5.0; and (b) L/H= l.5 .

gives thrust ratios of approximately 1, which suggests that in this range the seismic
thrust on the wall could be obtained using the 1-g static ~olution given by Eq. (36).

Nonlinear Response

The backfill with the parabolic variation in modulus is next analyzed

assuming a nonlinear dependence of damping arid modulus on shear strain. The

@seismicisolation
@seismicisolation
 14 RETAINING STRUCTURES ANALYSIS AND DESIGN

-
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

.....
u6 u5 u4
T b

ul
--
.!!.2
a
UJ, ~ X

Fig. 4. Finite element used in analyses.

E' 2oo from F.E. analysis

......_ (b) L/H• 1.5
z from cloaed-fonn aolutioo
c. 100
I-
(/)
::::>
~ o~~~~+4~~~~~~~~~~~~~~HY~~~
1-
u
~ -100
~ H•1 Om, Poisson •s ratio 0 .4;
a Shear Modulus J!l240 kPa : dampino 10..
El Centra input (0.079)
-200~~~~~~~~~~~~~~~~~~~
0 2 4 6 8 10
TIME ( SEC )

Fig. 5. Comparisons of dynamic thrusts between the finite element and closed-form
solutions for uniform soils for uniform backfill for LIH=l.5.

analysis is based on the equivalent linear approach of Seed and Idriss (1967). Two
analyses were carried out for stiff backfill material with a shear modulus G. =
132,000 kPa at the base of the dike fill, and for a more flexible backfill ·with a shear
modulus of 66,000 kPa at the base. The analyses were conducted for UH = 5.
Scaled El Centro acceleration records were used as input motions. The effects of
nonlinearity on the dynamic response of the system were explored by scaling the El
Centro motions so that records had peak accelerations ranging from 0.05 g to
0.35 g. The shear strain dependence of shear moduli and damping ratios were taken
from Seed and Idriss (1970). Maximum damping ratio D...,. was taken to be 30%.

@seismicisolation
@seismicisolation
 SEISMIC PRESSURES AGAINST RIGID WALLS 15
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

4
=
Poisson's ratio 0.4
=
LIH 5.0; damping 10%
0 3
~
a:: El Centro input
1-
C/)
;:J
2 ,, Sinusoidal motions
I
a:: I
I
1- 1

0
0 2 3 4 5
FREQUENCY RATIO, fR 2

Fig. 6. Thrust ratios as a function offrequency for: (a) sinusoidal input; and (b) El
Centro ground motions; parabolic distribution of shear modulus
with depth and elastic response.

The effect of the level of shaking on the seismic thrust on the wall depends
strongly on the degree of nonlinearity in the response. This is shown in Fig. 7 for
the stiff site in which the thrust ratios are plotted as a function of peak base
acceleration for both nonlinear and linear elastic analyses. At low levels of peak
base acceleration, when the response is effectively elastic, both equivalent linear and
linear elastic analyses give the same thrust ratio. Thereafter, this ratio remains
constant for the linear elastic analyses. The thrust ratios derived by nonlinear
analyses tend to increase with peak base acceleration, although the results are
affected somewhat by the relationship between the predominant period developed in
the nonlinear site and the predominant period of the earthquake motions. When the
peak ground acceleration is 0.3 5 g, the thrust ratio is about 25% higher than for the
linear elastic case. For levels of peak ground acceleration between 0.1 g and 0.3 g,
which is a very common range in practice, the thrust ratio is fairly well
approximated by 1.0, again confirming that the 1.0 g solution can be quite useful.
Note that for low levels of acceleration, when the response is sensibly elastic, that
the thrust ratio drops significantly below 1.0.

The peak dynamic thrust ratios for both the stiffer and more flexible backfllls
are compared in Fig. 8. The softer site shows the larger dynamic thrust ratios over
the entire range of peak accelerations.

@seismicisolation
@seismicisolation
 16 RETAINING STRUCTURES ANALYSIS AND DESIGN

1.5
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

~NON-LINEAR ANALYSES
· - - - LINEAR ELASTIC ANALYSES

L/H • 5 .0, H•1 Om

(/l Poisson 's ratio 0 .4
0 [I Centro input

~
I- 1.0
(/l
::::>
a::
::I:
I-

0.5
0.0 0., 0.2 0.3 0 .4 0.5
PEAK ACCELERATION (g)

Fig. 7. Effect of nonlinearity on dynamic thrust ratios for an initial parabolic

distribution of shear modulus with maximum modulus,
G. = 132,000 kPa.

The vanattons in the dynamic thrust ratio for different soil stiffnesses,
different L/H ratio and different peak ground accelerations, suggest that one should
be cautious in using simplified estimates of the peak dynamic thrust, such as P = yH2
A.w: or· the response spectrum approach. For important designs especially with
moderate factors of safety the finite element analysis developed herein is
recommended. The more simplified approaches should be used with a substantial
factor of safety.

CONCLUSIONS

A simplified approximate method based on a modified shear beam model has

been presented for the analysis of dynamic earth pressures against rigid walls. The
method gives pressures within 5% of the exact solutions proposed by Wood (1973)
for uniform elastic backfill.

Mode-superposition solutions much simpler than those proposed by Wood

(1973) are developed from the model. The modal solution converges very rapidly.
For harmonic input just one mode gives dynamic pressures within 75-80% of the
exact solution. If ten modes are used the accuracy is 95%.

@seismicisolation
@seismicisolation
 SEISMIC PRESSURES AGAINST RIGID WALLS 17

2
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

G0 = 132,000 kPa
0 G0 = 66,000 kPa

~
1-
en
~
J:
1-
t>
1 - - - - .,...:::-:::~----------
~
~
i:l

0 I T I
0.0 0.1 0.2 0.3 0.4
PEAK BASE ACCELERATION (g)

Fig. 8. A comparison ofthntst ratios resulting from nonlinear response

analyses of a stiff site and a more flexible site.

On the basis of a limited number of analyses of seismic response, the

accuracy of response spectrum analyses appears to be frequency dependent. For
input frequencies less than the frequency of the backfill, the thrust against the wall is
overestimated by about 20%; for greater frequencies, the error approaches 50%.
The modal contributions should be summed using the absolute values of the peak
modal contributions.

A finite element version of the proposed method was also developed for the
analysis of nonhomogeneous and nonlinear backfill soils. Analyses were conducted
to show the dependence of the peak dynamic wall pressure on the peak ground
acceleration for earthquake motions. The results show that the assumption of
dasticity may underestimate substantially the dynamic pressures under strong
shaking.

ACKNOWLEDGEMENTS

The basic research in tltis paper was supported by the National Science and
Engi neering Council of Canada under Grant No. 5-81498. The authors appreciated
the constntctive comments of the reviewers.

@seismicisolation
@seismicisolation
 18 RETAINING STRUCTURES ANALYSIS AND DESIGN

REFERENCES
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Arias, A. Sanchez-Sesma, F.J. and Ovando-Shelley, E . (1981). A Simplified Elastic

Model for Seismic Analysis of Earth-Retaining Structures with Limited
Displacements. Proc. Int. Conf. on R!!cent Advances in Geotech. Earthquake Eng.
and Soil Dynamics, St. Louis, MO, Vol. I, pp. 235-240.

Finn, W.D . Liam, Wu, G. and Ledbetter, R.H . (1994). Problems in seismic soil-
structure interaction. Proc. 8th Int. Con£ on Computer Methods and Advances in
Geomechanics, Morgantown, West Virginia, Vol. 1, pp. 139-151, May 22-28.

Matsuo, H. and Ohara, S. (1960). Lateral earth pressure and stability of quay walls
during earthquakes. Proc. 2nd World Conf. Earthquake Eng., Tokyo, Japan.

Mononobe, N. and Matsuo, M . (1929). On the Determination of Earth Pressures

During Earthquakes . Proc., World Eng., Congress, 9.

Okabe, S. (1926). General Theory of Earth Pressures. J. Japan Soc. of Civil Engrs.,
12:1.

Seed, H.B. and Idriss, I .M . (1967). Analysis of Soil Liquefaction: Niigata

Earthquake. J. Soil Mechanics and Foundation Division, ASCE, Vol. 93, No. SM3,
pp. 83-108.

Seed, H.B . and Idriss, I.M. (1970). Soil Moduli and Damping Factors for Dynamic
Response Analyses. Report No. EERC 70-10, Earthquake Engineering Research
Center, University of California, Berkeley, California, December.

Veletsos, A.S. and Younan, A.H. (1994). Dynamic Soil Pressures on Rigid Vertical
Walls. Earthquake Engineering and Structural Dynamics, Vol. 23, pp. 275-301.

Veletsos, A.S., Parikh, V.H. and Younan, A.H. (1995). Dynamic Response of a Pair
of \Valls Retaining a Visco-elastic Solid. Earthquake Engineering and Structural
Dynamics, Vol. 24, pp. 1567-1589.

Wood, J.H. (1973). Earthquake-Induced Soil Pressures on Structures. Ph.D. Thesis,

submitted to the California Institute of Technology, Pasadena, California.

Wu, G. (1994). Dynamic Soil Structure Interaction: Pile Foundations and Retaining
Structures. Ph.D. Thesis, Department of Civil Engineering, University of British
Columbia, Vancouver, B.C., Canada.

@seismicisolation
@seismicisolation
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

DYNAMIC REsPONSE OF CANTILEVER RETAINING WALLS

A. S. Veletsos, 1 Member ASCE, and A. H. Younan? Associate Member, ASCE

ABSTRACT
The complete paper will appear in the
Journal of Geotechnical Engineering

A critical evaluation is made of the response to horizontal ground shaking of flex-

ible cantilever retaining walls that are elastically constrained against rotation at their
base. The retained medium is idealized as a uniform, linear,, viscoelastic stratum of
constant thickness and semiinfi~te extent in the horizontal direction. The parameters
varied include the flexibilities of the wall and its base, the properties of the retained
medium, and the characteristics of the ground motion. In addition to long-period,
effectively static excitations, both harmonic base motions and an actual earthquake
record are considered. The response quantities examined include the displacements of
the wall relative to the moving base, the wall pressures, and the associated shears and
bending moments. The method of analysis employed is described only briefly, empha-
si being placed on the presentation and interpretation of the comprehensive numerical
solutions.

Following are some of the more important conclusions of this study:

1. For the soil-wall system examined, both the magnitudes and distributions of
the wall displacements, wall pressures and associated forces induced by horizontal
ground shaking are quite sensitive to the flexibilities of the wall and its base.
Increasing either flexibility reduces the horizontal extensional stiffness of the retained
medium relative to its shearing stiffness, and this reduction decreases the proportion of
the soil inertia forces that gets transferred to the wall and, hence, the forces developed
in it.
1 Brown & Root Prof., Dept. of Civil Engrg., Rice Univ., Houston, Texas 77005-1892
2 Post-Dor:toral Fellow, Dept. of Civil Engrg., Rice Univ., Houston, Texas

19

@seismicisolation
@seismicisolation
 20 RETAINING STRUCTURES ANALYSIS AND DESIGN

2. For realistic wall fl exibili ties, the total wall force or base shear is one-half or
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

less of that obtained for a fixed-based, rigid wall, and the corresponding reduction in
the overturni ng base moment is even larger. With the information presented, the
precise dependence of these critical forces on the flexibilities of the wall and its base
may be evaluated readily.
3. When the dynamic amplification effects of the retained medium are neglected,
the magnitude of the total wall force obtained for reali stic wall tlexibilities by the
present method of analysis is in reasonable agreement with that computed by the
limit-state, Mononobe-Okabe method whi ch also disregards the dynamic
amplincations. Additionally, the effective wall height, which is the height by which
the total wall force must be multiplied to obtain the overturning base moment, may
well be of the order of 40 percent or Jess of the actual wall height. These values are in
close agreement with the 1/3 value involved in the original M-0 method, and
substantiall y smaller than the 60 percent value recommended in the Seed-Whitman
modification of th e method.
4. For systems excited by earthquake ground motions of the type recorded
during the 1940 El Centro, California event, the dynamic amplification factor for total
wall force for the most unfavorable combination of system parameters is likely to vary
from 1.3 for fixed-based, rigid walls to 1.9 for walls of high fl exibility. The effective
wall height, on the other hand , is insensitive to the ground motion characteristics, and
may be taken equa l to that obtained for 'statically excited' systems.

5. Even for the 1940 E l Centro earthquake motion , the maximum wall
displacement relative to the moving base for realistic systems is found to be less than
the values of 0.1% to 0.4% of the wall height normally accepted as the minimum
required to develop a limit state in the retained material.
6. The comprehensive numerical soluti ons presented and their analysis provide
not only valuable insights into the effects and relative importance of the numerous
factors that influence the response of the systems considered, but also a sound
framework for assessing the behavior of even more complex soil-wall systems. It is
hoped that the information presented will also lead to a greater appreciation than
appears to exist at present of the va lue of elastic methods of analysi s for the problem
examined.
7. The effects of nonuniformity in the shear modulus of the retained medium and
of separation at the wall -medium interface were examined briefly from a static point
of view. The dynamic aspects of these issues require further study.

@seismicisolation
@seismicisolation
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

ON SEISMIC DISPLACEMENTS OF
RIGID RETAINING WALLS

By Yingwei Wu 1 and Shamsher Prakash\ Fellow, ASCE

ABSTRACT

Rigid retaining wa lls experience both sliding and rocking displacement during
earthquakes. Richard s and Elms' (1979) method incorporates only slidin g
displacemen ts in design. A realistic method of computi ng both sliding and rocking
displacements of the walls based on nonlinear soil properties both for the base soil and
backfill was developed by Rafnsson and Prakash (1994). A design proced ure based on
this method was then developed by Prakash. el a/ (1995, b).

Comparisons of the displacements of typical walls by both the meU1ods have been
made. Also, typical field data from recent earthquakes has beer1 analyzed to compare
the observed displacements with the computed displacements.

INTRODUCTION

Rigid retaining walls experience both sliding and rotational displacements during
an earthquake (Rafnsson 1991 , Prakash and Wu 1996). Several analytical procedures
to compute displacements of rigid walls have been proposed. Earlier methods
considered only sliding of the walls (Richard and Elms 1979, Prakash 198 1) . More
recently solutions have been developed to consider both sliding and rocking
di splacements (Nadim and Whitman 1983, 1984, Rafnsso n 1991 , Rafnsson and
Prakash 1994, Prakash et.al. 1995 a, b). In some stud ies, unrealistic asstunptions
regarding soil properties have been used, e.g., rigid plastic, or linear soil. Also artificial
restraints were placed on the wall so that it may either tilt first or rotate first. Prakash
and Wu (1996) compared the performance of two rigid walls during earthquakes with
their computations and found good agreement.

1
Graduate student, Civil Engrg. Dept., Univ. of Missouri-Roll a, Rolla, MO 65401
2
Prof. of Civil Engrg., Univ. of Missouri-Rolla, Rolla, MO 6540 1

21

@seismicisolation
@seismicisolation
 22 RETAINING STRUCTURES ANALYSIS AND DESIGN

For displacement-based design of rigid walls, only two of the above analyses have
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

been adapted.
1. Richards and Elms (1979) method; (RE method) .
2. Prakash, Rafusson and Wu (1995 a, b) method; (PRW method).

In this paper, computations of displacements of typical walls have been made by the
two methods. Also walls which experienced displacements in recent earthquakes have
been analyzed and the result compared with the field data.

At present (1996), no information is available on the permissible displacements of

rigid walls. Only European Prestandard, Eurocode (1994) has introduced the concept
of permissible displacement in aseismic design of retaining walls. A detailed discussion
of the permissible displacements has also been presented.

STATE OF THE ART

There are several procedures available in published literature to compute

displacements of rigid walls. Rafusson (1991) has prepared a comprehensive review of
these procedures. Only the one relevant to our work will be included here.

Nadim (1980) and Nadim and Whitman (1984) developed a method to evaluate
permanent rotation and translation of gravity retaining walls with dry, cohesionless
backfill. All elastic deformations were neglected. The work previously done by
Richards and Elms (1979) was extended to study the tilting effect on a wall. The
assumptions in this solution are:

1. The foundation soil has a constant moment capacity below which no rotational
movements take place. Once the moment capacity is reached, the foundation
soil deforms plastically in rotation. The soil thus behaves like a rigid-plastic
material.
2. The center of rocking is at a fixed point at the base of the wall.
3. When the active condition exists, a failure zone consisting of infinite number of
parallel planes develops in the backfill. This assumption allows to assume
continuity when the wall is tilting.

The resulting mathematical model led to a solution involving several coupled

equations which require an iterative procedure to obtain a solution. The horizontal
ground acceleration coefficient (N) initiating plastic rocking CNw,) and plastic sliding
(N..J is evaluated. The lower value of either Ntil, or N,ud determines whether sliding or
rocking motion will govern the displacement (D) of a wall that takes place during a
particular earthquake. This displacement is tht:n estimated from Eqn 1 (Wong, 1982).
The minimum value ofN,lidin& and Ntil, is used as N value in this equation.

@seismicisolation
@seismicisolation
 RIGID RETAINING WALLS SEISMIC DISPLACEMENTS 23

(¥)
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

_ 37V 2 - 9.4
D - -- e (1)
Ag
Wh~re : A is coefficient of peak horizontal ground acceleration.
Vis the peak earthquake velocity.
N is the coefficient of limiting wall acceleration

Nadim (1982) and Nadim and Whitman (1983) also developed a finite element
solution of the mathematical model of the soil system and concluded that:

1. Earthquake loading may result in a residual force on the wall, which may be
as much as 30% greater than the static active force.
2. If the ratio of dominant frequency of ground motion (f) to the fundamental
frequency ofbackfill (f1) is greater than 0.3, the amplification of motion in the
backfill plays an important role in the permanent displacement of the wall.

The fundamental frequency of the backfill is determined as:

v.
f1 = - Hz (2)
4H
Where: V, is shear velocity (m/sec)
His height of wall (m)

The finite element solution takes into account both sliding and tilting of wall.
However, because of the boundary conditions that are imposed in the idealization of
the problem, the results are conservative. The following design procedure was
recommended to determine displacements of retaining walls:

1. Determine f,.
2. Determine f.
3. If f/f1 is less than 0.25 neglect the amplification of ground motion.
4. Iff/f1 is approximately 0.5, increase the peak acceleration (A) and the peak
velocity (V) of the design earthquake by 25% to 30% respectively.
5. If f/f1 is between 0.7 and 1.0, increase A and V by 50%.
6. Use new values of A and V to determine the displacements (Eqn1 ).

Richards and Elms Method· (RE method)

A simplified method for dynamic design of rigid retaining walls had been proposed
by Richards and Elms (1979). This method is based on Newmark's sliding block
analysis (1965) and Franklin and Chang's (1977) solution for upper bound permanent
displacements for several natural and synthetic ground motions. This approach
determines the wall dimensions based on permissible displacement. A brief description
of this method is given below:

@seismicisolation
@seismicisolation
 24 RETAINING STRUCTURES ANALYSIS AND DESIGN.

1. Select permissible displacement (d),

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

2. Determine A, and A, from a given seismic zone (Applied Techn'ology Council,

1978),
3. Determine coefficient of cut off acceleration o:h (Eqn. 3).

[ o.2A:r· ~
2
(3)
o: , "A a -
A-d
a

Where d is permissible displacement.

A. and A, are acceleration-coefficient in Applied Technology Council
(ATC, 1978)

4. Compute dynamic active lateral earth pressure behind the wall by using Mononobe-
Okabe method for o:h computed in Eqn 3.
5. Compute weight of wall by using inertia force of the wall and considering force
equilibrium.
6. Apply a factor of safety to the calculated weight. A value ofl.S is recommended
and wall dimensions are then determined

Only sliding motion is considered in this method. For further details, see Richards
and E lms (1979).

Prakash, Rafnsson and Wu 's method: (PRW method).

A solution technique for simulating the response of rigid walls during an earthquake
has been proposed by Rafnsson (1991 ). Using this solution technique, a complete
design procedure has been developed by Prakash et a/.(1995 a, b and 1996). This
model consists of a rigid wall resting on the surface of the soil and subjected to
horizontal exciting ground motion. The soil behavior is non-linear for both backfill and
base soil. Both material and geometrical damping in sliding and rocking motions have
been considered, Figure 1, (Rafnsson 1991, Rafnsson and Prakash 1994). In Figure 1,
k represents the stiffness and c the damping of the soil. Mathematical model in Figure
2 represents the displacements in active case. Nonlinear behavior of soil is included in
defining the following properties, both at the base as well as the baclifill:

(1) Soil stiffness in sliding. (2) Soil stiffness in rocking.

(J) Geometrical damping in sliding. (4) Geometrical damping in rocking.
(5) Material damping in sliding. (6) Material damping in rocking.

The equations of motion for horizontal sliding and rotation in active case are written
as:
m~ + Cx ~ + kxXs + mea - ~lSEJ - kus8 = Px (t) (4)
~,)l + CR fJ + kR8 - ~lR ~ - k~m. Xs = Mx (t) (5)

@seismicisolation
@seismicisolation
 RIGID RETAINING WALLS SEISMIC DISPLACEMENTS 25

In the above equations, m represents the mass of the wall, Mmo - the mass moment
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

e
of inertia, x, the horizontal displacement, the angular rotatio n, and c the dynamic
damping. Subscripts "HS" and "HR" represent total damping for backfill in sliding and
rocking respectively, subscript "x" sliding, and "R" rotation. The stiffness (k) and
damping ( c) in several modes are presented elsewhere (Rafnsson and Prakash, 1991).

ex
e.G.
kx
mx

cxx (a)
kxx

Mmoe

(e.G.
) c<t>Ei
k.<t>e

~<l>e ./ (b)
k<J>8

~~fi cxx
k~B kxx./ (c)
Figure 1. System of forces in mathematical model of retaining wall: (a)
sliding only, (b) rocking only, (c) combined sliding and rocking
(Rafnsson 1991, Rafnsson and Prakash 1994).

@seismicisolation
@seismicisolation
 26 RETAINING STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

r_,

I
I \
I I
I I
I I

c-~
I
He
) k... [x, +nHEl]nH
I
I

~~X,+ 2

1
I
I . x.+nH8]nH
H I
I

I c~
[x,+Yl Hf) J .
T
e
I
I
I
I

k, [x,+Yl Hf) J nH
I

I
I
I

Figure 2. Mathematical model for stiffness and damping constants for the active
case (Rafnsson 1991, Rafnsson and Prakash 1994).

The method of computation is described here in brief only (Rafnsson 1991).

I. Wall dimensions are determined for given factors of safety under static
condition.
2. Cumulative displacements and rotation of wall are then computed for different
loading cycles (magnitude of earthquake) for a given ground motion.
3. The computed displacements are compared with the permissible displacement.
4. If the computed displacement is larger than permissible displacement, wall
section will be redesigned for the permissible displacement.

Nonlinear soil modulu s and material damping with strain used in this solution are
shown in Figure 3 and 4 respectively. The values of G/Gmax and damping ratio for silt
are obtained from the mean value of sand and clay (PI=30).

@seismicisolation
@seismicisolation
 RIGID RETAINING WALLS SEISMIC DISPLACEMENTS 27
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

o.a

~ 0 .6
0

'0 0.4

0 .2

Cyclic Shear Strain

Figure 3. Average values of G/Gmu versus shear strain (y) for different soils
(After Seed and Idriss 1970, for sand; Seed, Wong, ldriss and
Tokimatsu 1986, for gravel; Vucetic and Dobry 1991, for clay with
Pl=30).

30

~
2s
-~
.....
"'
0::
QD
20
0
·a. 15
8
"'
Cl
10
iii
·;:
.....Qj 5
"'
::;;
0
w-e w-• 10-~ 10-·
Cyclic Shear Strain

Figure 4. Average values of material damping ratio ( () and shear strain (y) for
different soils (After Seed and ldriss 1970, for sand; Seed, Wong,
ldriss and Tokimatsu 1986, for gravel; Vucetic and Dobry 1991, for .
clay with Pl=30).

@seismicisolation
@seismicisolation
 28 RETAINING STRUCTURES ANALYSIS AND DESIGN

COMPARISON OF RESULTS BY THE TWO METHODS

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Parameters Used for Comparison

For comparison of displacement by the two methods, a typical wall of 6m high was
studied. The basis. of comparison is as follow:

Fix the section by RE 's recommendations: then compute the displacements by PRW
method and compare with RE permissible displacement.

In RE method, permissible displacements must be known to determine the wall

section. Therefore, a permissible displacement of 2% of wall height, ie 0 .12m, was
adopted. The values of A, (0.4) and A, (0.4) for specific location, e.g . Orange County
(CA), have been adopted (ATC 1978). Other parameters used in this analysis are listed
in Table l. Cutoff acceleration for the parameters in Table l , computed frl'Jm Eqn 3 is
0.144g.

For a displacement of 0.12m, height of 6m and soil properties as in Figure 5, the

wall section was fixed according to RE recommendations for the above parameters.
Base widths used in this study are computed by RE method considering full base
friction angle (ie. & = <j> ) .

Table 1. Parameters used in the analys'is.

IParameter I Range of Valu es

Height of wall (m) 6

Location used for analysis Orange County, CA

A. 0.4

A. 0.4
Permissible displacement (m) 0 .12 (2% of height ofwalQ

Ground Motion (g) 0.4

Frequency (Hz) of ground motion 1

Magnitude of earthquake M6.75 and M7.5

Number of cycles 10 and 15
Soil properties for base soil and backfill See Table 2

@seismicisolation
@seismicisolation
 RIGID RETAINING WALLS SEISMIC DISPLACEMENTS 29
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Compacted poorly
graded gravel -silt (GM:)
<I> 33.0'
D

1i = 22.0'
c ~ 0 . 35
y,=2 1.60kNIM'
u = 0.3
w% = 10%
y,=23.58kNIM: '

r- 3.1037m - --1
Well graded sand (S W)
y,• l9.44 kNIM' w~. = 8%
<I> - 35.0' u - 0.3
1i = 23.3' ':':--
e - 0.46

Figure 5. Section of wall and soil combination used for analysis (BF1-BS3).

For typical properties of the base soil and backfill in Figure 5, a plot of cumulative
displacements with number of cycles is shown in Figure 6. In this figure, three
accelerations have been used.

Plot (a) in Figure 6 is displacement computed by RE 's cutoff acceleration of

0.144g. According toRE's procedure, there is no displacement of the wall for ground
motion smaller than cutoff acceleration. However, by PRW method, the displacements
in 10 cycles (M6.75) and 15 cycles (M7.5) are 4.39cm and 10.86cm respectively. In
RE's method, the soil is assumed rigid plastic. Therefore, these displacements cannot
be determined and are neglected.

Plot ( b ) in Figure 6 is for ground motion of 0.4g. The displacements for 10 cycles
(M6.75) and 15 cycles (M7.5) are 28 .1 5cm and 46.70cm respectively. This is larger
than the permissible displacement of RE's (12cm) which is the basis of fixing the wall
section.

Plot (c) in Figure 6 is for ground motion of 0.256g (the difference of 0.4g and
0.144g). The displacements for 10 cycles (M6.75) and 15 cycles (M7.5) are 14.56cm
and 27.12cm respectively. These are also larger than RE 's displacements of 12cm.

@seismicisolation
@seismicisolation
 30 RETAINING STRUCTURES ANALYSIS AND DESIGN .

I
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

....,J 0.4
~
Ql
E
Ql
CJ 0.3
ttl
0.
iJ"'
Ql 0.2
.~
~
;:;E o.1
;:l
u

15

Number of ~ycles

Figure 6. Cumulative displacements of proposed wall section by using three

ground motion (BF1-BS3).
Rotations of the wall for 0.4g ground motion and earthquake ofM6.75 and M7.5
are 1.29° and 2.15° respectively which results in 13.5lcm and 22.51cm displacement
at the top. Therefore, displacements in sliding only are 14.64 em and 24.19 em for
M6.75 and MJ-5 earthquakes respectively. These values are closer to the permissible
displacement of 12cm used in RE's design method. In fact sliding displacement (14.64
em) for M6.75 is extremely close to the permissible displacement of 12cm adopted
here.

It will thus be seen that:

I . There are significant displacements of this wall for the cutoff acceleration of
0.144g. In real soils, which may be not exhibit rigid-plastic behavior,
neglecting those displacements will be unsafe.
2. For design earthquake acceleration of0.4g, the displacements of the walls are
2.35 times and 3.89 times the pennissible displacement for real earthquakes of
M6.75 and M7.5 respectively as compared toRE's displacements.
3. RE method predicted closely only sliding displacements for this wall with
M6.75.
4. This wall experiences comparable displacements in rocking also. These cannot
be computed by RE method.

To continue this comparison further, seven types of base soils and three types of
backfills were selected (Table 2). Wall height and top width were 6m and 0.5m
respectively throughout. Cumulative displacements are computt!d for each of the cases

@seismicisolation
@seismicisolation
 RIGID RETAINING WALLS SEISMIC DISPLACEMENTS 31

and 3 grotmd motions, ie. 0.144g, 0.4g and 0.256g, and for M6.75 (10 cycles). In Table
3, for each of these accelerations, sliding displacements, rocking degrees and total
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

displacements at the top of the wall have been Listed. A critical study ofTable 3 shows
that:

1. The total displacement for cutoff acceleration of0.144g varies from 4.39cm to
20.84cm. This is neglected in RE method.
2. Displacements in sliding only for 0.4g ground motion is fro m 14.16cm to
43.93cm. Only in 12 of the 21 cases in this study, the sliding displacements
were within 150% of theRE's permissible displacements.
3. Total displacements for 0.4g ground motion are from·28.1 5cm to 45.6lcm,
neglecting cases with very large displacements which m ay be constitute
failure, as discussed later.

There are no guide lines on permissible displacements yet. A general discussion

on this question is presented later.

Table 2. Soil properties of base soil and backfill used for analys is.
BASE SOIL (BS)
soil Yd cj> 6 void c PI Gmax
v w%
type kN/m2 deg deg ratio kN/m 2 Mpa

BSI GW 21.07 37.5 25.0 0.25 0.3 - - 6 191

BS2 GP 19.18 36.0 24.0 0.36 OJ - - 6 162
BS3 sw 18.00 35 .0 23.3 0.46 0.3 - - 8 139
BS4 SP 16:82 34.0 22.7 0.56 0.3 - - 10 120
BS5 SM 16.5 I 33.0 22.0 0.68 0.3 - 4 11 164

BS6 sc 15.25 30.0 20.0 0.95 0.3 - 13 14 152

BS7 ML 14.15 32.0 2 1.3 0.85 0.3 9.57 4 14 102

BACKFILL (BF)
so il Yd cj> 6 void v c PI w% Gmox
BPI GM 19.6 33 .0 22.0 0.35 0.3 - - 10 164

BF2 GP 18.9 34 .0 22.7 0.40 0.3 - - 8 153

BF3 SP 15.6 34.0 22.7 0.50 0.3 - - 8 132
* All propert1es for backfill are determmed from 90% of the "Standard Proctor" test.

@seismicisolation
@seismicisolation
 32 RETAINING STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Table 3. Slidin g, and cumulative d isplacements and rockin g degrees computed

by various ground motions for M6.75 (10 cycles) by PRW method

Disp lacement(cm) and rocking(")' in 10 cycles

So il base'
0.144 (g) 0.256 (g) 0.4 (g)
comb. ¥idth (m)
Sliding Rock. Tota l Sliding Rock. Tota l S liding Rock. Total

BF I-BS I 2.5927 3.12 0.36 6.72 8.40 1.07 19.61 14.97 1.94 35.29

BF2-BS 1 2.2636 3.20 0.56 9.20 8.62 1.45 23.80 14.67 2.52 41.06

BF3-BS 1 1.7759 4.18 1.20 16.78 8.73 2.53 35.22 14. 16 4.10 57. 103

BFI-BS2 2.8891 3.83 0.40 8.03 10.77 1.1 1 22.3 9 18.10 2.10 40.09

BF2-BS2 2.5321 4.44 0.58 10.44 10.82 1.47 26.21 18.49 2.55 45 . 19

BF3-BS2 1.9970 5.26 1.16 17.26 10.83 2.49 36.9 1 17.55 4.05 59.963

BF I-BS3 3.1037 2.30 0.20 4.39 7.54 0.67 14.56 14.64 1.29 28.15

BF2-BS3 2.7266 2.46 0.28 5.39 7.79 0.89 17.11 14.33 1.65 31.61

BF3-BS3 2. 1572 3.07 0.56 8.93 7.88 1.48 23.38 13.68 2.59 40.80

BF1-BS4 3.3338 2.97 0.23 5.38 9.69 0.74 17.44 18.14 1.40 32.80

BF2-BS4 2.9350 3.29 0.32 6.64 9.88 0.98 20.14 17.74 1.77 36.28

BF3-BS4 2.3288 3.96 0.62 10.45 9.74 1.58 26.29 16.8 1 2.75 45.61

BF I- BS5 3.581 2 4.45 0.30 7.59 13.20 0.90 22.62 23 .99 1.64 41. 16

BF2-BS5 3.1591 4.88 0.42 9.28 13.34 1. 17 25 .59 23 .6 1 2.07 45.29

BF3-BS5 2.5 134 5.4 1 0.79 13 .68 12 .96 1.86 32.44 22.25 3.20 55.763

BF I-BS6 4.45 15 10.19 0.44 14.80 25.26 1.12 36.99 43 .93 1.95 64.35 3

BF2-BS6 3.9476 10.41 0.59 16.59 24.36 1.39 38.92 42.21 2.4 1 67.45 3

BF3-BS6 3.1627 10.32 0.97 20.84 22.8 1 2.12 45.01 38. 10 3.56 75.38 3

BFJ-BS7 3.848 1 3.99 0.2 1 6. 19 10.65 0.60 16.93 19.43 1.1 0 30.95

BF2-BS7 3.4010 4.06 0.29 6.99 10.59 0.77 18.66 18.73 1.36' 32.97

BF3-BS7 2.4 126 4.23 0.49 9.36 10.09 1.1 8 22.45 17 .18 2.02 38.33
1 Base w1dths are determmed based on RE method
2 Displacements are computed by PRW method.
3
Too large; represent fa ilure.

@seismicisolation
@seismicisolation
 RIGID RETAINING WALLS SEISMIC DISPLACEMENTS 33
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

COMPARISON WITH FIELD PERFORMANCE

A 4m high retaining wall had rotated by 1° - 2 o during Hokkaido-Nansi-Oki

earthquake in 1993 (Chung, 1995). The magnitude of this earthquake is 6.7 (9.7
cycles) and (a.h)max is 0.2g (see also Prakash and Wu 1996), For design of wall, a." is
taken as 2/3 x (a.h)max = 0.133g. Since no soil properties were described by Chung
(1995), this wall was analyzed by PRW method for several base soils with 4> of 27° -
35", but with one backfill Ccl> = 35"). The displacements are shown in Table 5:

Table 5. Displacements of a 4m high wall with varying base soil properties in

HNO earthquake (Prakash and Wu 1996).

35 34 33 32 31 30 29 28 27
0.35 0.4 0.45 0.50 0.55 0.6 0.65 0.7 0.75
y,,(kN/m') 22.16 21.72 2 1.3 1 20.93 20.5 7 20.23 19.92 19.62 19.34
Base width (m) 1.9590 2.0503 2. 1483 2.2523 2.3609 2.478 1 2.6032 2.7348 2.8779
Cumulative
0.0289 0.0323 0.0352 0.0384 0.042 1 0.0462 0.05 12 0.0571 0.0637
Displacement (m)
Cumulative
0.21 0.22 0.23 0.24 0.25 0.26 0.27 0.29 0.30
rocking degree
cumulative disp .
(') 0.4 1 0.46 0.50 0.55 0.60 0.66 0.73 0.82 0.9 1
height of wall

Since the soils were saturated, an allowance for this was made in this analysis. The
wall in HNO earthquake had rotation by I o to 2 °. Since the walls actually experiences
sliding and rotation, the rotation is computed based on cumulative di splacement (Table
5). The computed rotation is of the order 0.41° to 0.91 °.

This method is, therefore, capable of predicting the order of displacement of actual
walls during real earthquakes. Richards and Elms' (I 979) method cannot be extended
to perform analysis of wall with other than the parameters used in their analysis.

PERMISSIBLE AND FAILURE DISPLACEMENT

Permissi ble Displacement

There are no guide lines on the permissible displacements of wal l in the field. Free
standing retaining walls are used as wing walls along the spillway structures and as
retaining structures along highways. The magnitude of displacements of free standing
retaining walls that may be tolerated depends on a number of factors, eg. :

( i). Is it economical to design for a certain displacement?

@seismicisolation
@seismicisolation
 34 RETAINING STRUCTURES ANALYSIS AND DESIGN

(ii). How will the possible displacement affect nearby structures and buildings?
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

(iii). How much displacement can be tolerated before reconstruction of the

retaining wall?

Last but not the least is the factor of''psychologicaf' effect of a wall displacement
A wall that is sloping away from the fill after an earthquake may perhaps pose no
danger, but it may be perceived to have failed.

In the case of free standing retaining walls there are number of methods that may
be used to define the permissible displacement, such as:

1. A permissible displacement that is related to the wall height and level of

horizontal acceleration. The displacement could be defi ned as a certain
percentage of the wall height.
2. A requirement such that a wall is not allowed to slope forward. In that case the
horizontal distance from the toe of the wall to the top fro nt of the wall is the
permissible displacement. A solution would be to slant the wall on the backfill
in the first instance.
3. Stability and settlement requirements because of nearby structures. This has to
be analyzed and determined in each case.
4. Subjective judgement of the engineer.

Failure Displacement

A 4m high masonry retaining wall with base soil as gravel and sand experienced
complete failure during the Kobe earthquake (M=7.2, ahmax = 0.5) ofJanuary 17, 1995
on the JR Tohkaido line (Tateyama, eta/. 1995). The equivalent number of cycles for
M=7.2 at 0.65 a 1unax ( ah = 0.33) is 13 (Seed el al. 1983). No details of the section and
soil properties except a mention of alluvial fan deposit consisting of mainly gravel and
sand are listed. Therefore, several analyses were performed based on assumed data
(Prakash and Wu 1996) and it was concluded that a displacement exceeding I 0%
(40cm) of the wall height may be considered as failure condition. This is a first
recommendation in print on this subject. Eurocode 8 (1994) recommends permissible
displacement as 300x a (mm) ie ! Scm (for a =O.S , maximum design acceleration). This
wall is considered to have failed based on the Eurocode 8 recommendation and the
result computed by PRW method.

More analyses and performance studies on the walls are needed before unified
recommendations on ( i) acceptable displacement and (ii) fail ure displacements are
adopted.

@seismicisolation
@seismicisolation
 RIGID RETAINING WALLS SEISMIC DISPLACEMENTS 35

CONCWSIONS
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

I'. In this analysis, at M6.75 (10 cycles), the total displacements computed by PRW
method are generally larger than the permissible displacement used in RE's
method.

2. Only sliding displacements by the PRW method are close to the permissible
displacement of RE's method for M6.75.

3. The differences in displacements by the two methods are due to the fact that RE 's
method neglects rocking motion.

4. The displacements induced by rotation are comparable in most cases to the sliding
displacements. In several walls, eg. BF3-BS4, the displacement in rotation is in
fact larger than that in sliding.

5. More studies are needed to define: a) permissible and b) failure displacements.

REFERENCES

Applied Technology Council, (1978), "Tentative Provisions for the Development of

Seismic Regulations for Building" , Prepared by Applied Technology Council,
Associated with the structure Engineers Association of CA.
Chung, R.M. (1995) "Hokkaido Earthquake Reconnaissance Report" Earthquake
Spectra, Supplement to Vol. II April.
EUROCODE 8 (EUROPEAN PRESTANDARD, 1994) "Design Provisions for
Earthquake Resistance of Structures- Part 5: Foundations, Retaining Structures and
Geotechnical Aspects" The Commission of the European Communities.
Franklin, A. G., and Chang, F. K., (1977), "Earthquake Resistance of Earth and
Rockfill Dams; Report 5, Permanent Displacements of Earth Embankments by
Newmark Sliding Block Analysis", U.S. Army Engineer Waterways Experiment
Station, Soils and Pavements Laboratory, Vicksburg, Miss., pp. 1-38
Nadim, F. (1980), "Tilting and Sliding of Gravity Retaining Walls During
Earthquakes", Master's Thesis, Department of Civil Engineering, M. I. .T.,
Cambridge, Mass.
Nadim, F., (1982), "A Numerical Mode/for Evaluation ofSeismic Behavior of Gravity
Retaining Walls" , SeD Thesis, Department of Civil Engineering, M .I.T.,
Cambridge, Mass ..
Nadim, F. and Whitman, R. V., (1983), "Seismically Induced Movement of Retaining
Walls", Journal of the Geotechnical Engineering Division, ASCE, Vol. 109, No.
7, July, pp. 915-931
Nadim, F. and Whitman, R. V., (1984), "Coupled Sliding and Tilting of Gravity
Retaining Walls During Earthquakes", Proceedings of the Eighth World Conference
on Earthquake Engineering, July 21-28, San Francisco, CA, Vol. III, pp. 477-484.

@seismicisolation
@seismicisolation
 36 RETAINING STRUCTURES ANALYSIS AND DESIGN

Newmark, N. M. (1965), "Effects ofEarthquakes on Darns and Embankments", The

Institution of Civil Engineers, The Fifth Rankine Lecture, Geotechnique, Vol. 15,
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

No.2, January, 1965, pp.l37-16l.

Prakash, S. (1981), "Soil Dynamics" McGRAW-fiLL Book Co., New York, N.Y.,
Reprint, Sharnsher Prakash Foundation, Rolla MO.
Prakash, S.,Wu, Y. and Rafnsson, E. A., (1995, a), "On Seismic Design Displacements
of Rigid Retaining Walls", Proc. Third International Conference on Recent
Advances in Geotechnical Engineering and Soil Dynamics, ST. Louis Vol. III, pp.
1183-1192.
Prakash, S.,Wu, Y. and Rafnsson, E. A., (1995, b), "Displacement Based Aseismic
Design Charts For Rigid Walls", Sharnsher Prakash Foundation, Rolla MO.
Prakash, S. and Wu, Y., (1996), " Displacement of Rigid Retaining Walls During
Earthquakes", 11th World Conference on Earthquake Engineering, June, 23-28
(under print).
Prakash, S., Wu, Y. and Rafnsson, E. A., (1996), " DDRW-1- Soft Ware to Compute
Dynamic Displacements of Rigid Retaining Walls" Sharnsher Prakash Foundation,
August.
Rafnsson, E.A. (1991), "Displacement Based Design of Rigid Retaining Walls
Subje cted to Dynamic Loads Considering Soil Nonlinearity" , Ph.D. Thesis,
University of Missouri-Rolla, USA.
Rafnsson , E.A. and Prakash, S., (1991), "Stiffness and Damping Parameters for
Dynamic Analysis of Retaining Walls", Proc. Second International Conference on
Recent Advances in Geoteclmical Earthquake Engineering and Soil Dynamics, Vol.
III, pp. 1943-1952.
Rafnsson E. A. and Prakash, S., (1994), "Displacement Based Aseismic Design of
Retaining Walls" , Proc. XIII Inter. Conf. On SMFE, New Deihl, Vol3, pp. 1029-
1032.
Richards, R. and Elms, D. G., (1979), "Seismic Behavior of Gravity Retaining Walls
", Journal of the Geotechnical Engineering Division, ASCE, Vol. 105, No. GT4,
April, pp. 449-464.
Seed, H.B. and Idriss, I. M. (1970), "Soil Moduli and Damping Factors for Dynamic
Response Analysis" , Report No. EERC 70-10, Earthquake Engineering Research
Center, December, pp 1-40.
Seed, H.B. Idriss, I.M. and Arango, I., (1983), "Evaluation of Liquefaction Potential
Using Field Performance Data", Journal of Geotechnical Engineering, Vol. 109,
No . 3, March, pp. 458-482.
Seed, H.B., Wong, R.T., ldriss, I.M. and Tokimatsu, K ., (1986), "Moduli and Damping
Factors for Dynamic Analysis of Cohesionless Soils" , Journal of Geotechnical
Engineering, Vol. 112, No. 11, November, pp. 1016-1032.
Tateyama, M., Tatsuoka, F., Koseki, J. and Horii, K., (1995), "Damage to Soil
Retaining Walls for Railway Embankments During the Great Hanshln-Awaji
Earthquake, January 17, 1995", Proc. First International Conference on Earthquake
Geotechnical Engineering, Tokyo, 14-16 Nov, pp49-54.

@seismicisolation
@seismicisolation
 RIGID RETAINING WALLS SEISMIC DISPLACEMENTS 37

Vucetic, M. and Dobry, R. , (1991), "Effect of Soil Plasticity on Cyclic Response"

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Journal ofGeotechnical Engineering, ASCE, Vol. 117, No. 1, January, pp. 89-107.
Wong, C. P., (1982) "Seismic Analysis ans Improved Seismic Design Procedure for
Gravity Retaining Walls" , M.S. Thesis Dept. of Civil Engineering., M .I.T.,
Cambridge, Mass.

J::j"OTATIONS

A peak horizontal ground acceleration coefficient

A, seismic coefficient' representing the effective peak acceleration
Av seismic coefficient representing the effective peak velocity-related
acceleration
B base width
BF backfill
BS base soil
cohesion
computed displacement
permissible displacement
dominant frequency of ground motion
fundamental frequency of backfill
gravitational acceleration
maximum shear mudulus
Height of wall
coefficient of limiting wall acceleration
N slidlng horizontal ground acceleration coefficient initiating plastic sliding
N,;lt horizontal ground acceleration coefficient initiating plastic rocking
v, shear velocity
v peak earthquake velocity
a.h coefficient of cut off acceleration
Yd dry density
0 wall-soil friction angle
v poisson's ratio
<I> friction angle

@seismicisolation
@seismicisolation
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

ROTATION OF LARGE GRAVITY WALLS ON RIGID FOUNDATIONS

UNDER SEISMIC LOADING

R. S. Steedman 1 and X. Zeng2

A pseudo-static model is developed for the prediction of permanent residual

displacement of monolithic retaining walls in rocking. The model follows a Newmark
approach, and is validated by comparison with experimental data of the behaviour of a
large gravity wall subject to earthquake shaking. The experimental data show that
rocking of retaining walls comprises two components, an elastic recoverable element and
a permanent or residual component. The rocking model showed good agreement with
the observed data, matching the total displacement closely after a series of five
earthquakes. Comparison with conventional design approaches suggests that there is
opportunity to reduce conservatism in the design of gravity walls prone to rocking by
using a mechanism-specific model for the wall behaviour, particularly where the nature
of the earthquake input motion can be defined.

Introduction

The use of pseudo-static methods for the prediction of the sliding displacement
of gravity retaining walls is well established and widely used in the design of such
structures. In his Rankine Lecture Newmark (1965) noted that the slip displacement of
a block on a plane subjected to lateral accelerations could be calculated by integrating
the relative velocity of the block and the plane, having first defined a yield acceleration.
Newmark had in mind modelling the behaviour of earth dams or embankments but this
approach was subsequently extended to form the basis of a design approach for gravity

1
Director of Engineering, Sir Alexander Gibb & Partners Ltd, Earley House, London
Road, Reading RG6 lBL, UK
2
Assistant Professor, Department of Civil Engineering, University of Kentucky,
Lexington, Kentucky

38

@seismicisolation
@seismicisolation
 LARGE GRAVITY WALLS ROTATION 39

retaining walls by Richards and Elms (1979). The Richards and E lms approach followed
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

the work of Franklin and Chang (1977), who had systematically integrated a large
database of earth~uake records using the Newmark approach as a function of the ratio
of the yield (or threshold) lateral acceleration to the peak value. The earthquake records
were scaled to a peak acceleration of O.Sg and peak velocity of30 in/sec.

The new design approach was based on the view that it would be uneconomic
to design walls which were so massive that no movement took place under earthquake
loading, but that a realistic assessment could be made of the likely movement for walls
with threshold accelerations slightly less than the peak by using an envelope relationship
to the integrated earthquake records. Their proposed relationship for walls in sliding
was given as:

d =
v2 ( "A
o.o87 -:4 N) -4 ( in ) (I)

where Vis the design earthquake peak velocity in in/sec, N is the threshold acceleration
for sliding and A is the peak lateral ground acceleration (ftlsec 2) . At high values ofN/A,
this relationship provides a reasonable upper bound to the actual records. At low values
ofN/ A, however, Equation (1) overpredicts the likely sliding displacement significantly.

However, it is clear that such a calculation is sensitive to a number of key

parameters in the calculation of the threshold acceleration and in the nature of the
earthquake record. Whitman and Liao (1985) developed this approach further, building
on the work canied out at MIT, by analysing a selection offourt~en earthquake records
(twelve from earthquakes with magnitudes between 6.3 and 6. 7 and two larger) taking
into account the effects of orientation, vertical motion and site effects where possible.
A simple expression was first derived for the mean displacement as a function ofN/A,
analagous to Equation (1), to match the results of Franklin and Chang :

- 37
d =-- e
v2 -9. 4 !!.
(2)
A (in)
A

This expression then formed the basis for a statistical analysis of the likely sources of
error, leading to an alternative recommended design line based on a safety factor of 4,
equivalent to a probability ofnon-exceedance in excess of95%:

!!_ = 0.66 - - 1- In d A (3)

A 9.4 v2

suitable for most (larger) levels of ground acceleration. Consistent with the approach
used to derive this expression, N is used here as the expected or average threshold

@seismicisolation
@seismicisolation
 40 RETAINING STRUCTURES ANALYSIS AND DESIGN

acceleration based on average values for the various friction angles, and is denoted as N
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

in the original reference. The units are in/sec for V and ftlsec 2 for A, as for Equation l.

In this analysis, an allowance for rotation of the wall, contributing to outward

movement, was made based on earlier work by Nadim (1980) which suggested that
permanent outward movement of walls which could tilt may be of the order of 50%
greater than for walls which were constrained to slide only. Whitman and Liao
concluded that this aspect of wall behaviour, ie. tilting or rocking, is poorly understood
and should be the subject of further (centrifuge) model testing and analysis. (The
Whitman and Liao prediction is compared against data of rocking walls below.)

However, depending on the wall geometry and characteristics, rotation of a

gravity wall may be the critical mode of behaviour, 'triggered ' before a sliding threshold
is reached. In this respect, the use of the sliding mechanism is inappropriate to estimate
the rotational displacement. This paper describes the application of the principles of a
Newmark approach to the calculation of rocking displacement, ie. the permanent
element of a rotational mode of behaviour.

Fig. 1 Rocking of a monolithic wall

Rocking Displacement

Fig.1 shows the forces acting on a rigid monolithic gravity retaining wall founded
on a rigid base at the point of rocking. Considering the imposition of a lateral
acceleration field as pseudo-static, D' Alembert' s principle creates a body force opposed
to the direction of the lateral acceleration field, which affects the wall, the ground and
the retained soil. The total lateral force on the wall from the soil, P AE> is conventionally
calculated pseudo-statically using the well-known Mononobe-Okabe formulation

@seismicisolation
@seismicisolation
 LARGE GRAVITY WALLS ROTATION 41
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

following Coulomb, described in detail by Ebeling and Morrison (1992). The key
assumptions in this analysis are that:

a) the wall can be considered as rigid, rotating about its outer edge or corner;
b) the soil behind the wall behaves in a rigid-plastic manner, such that the soil
follows the wall movement outwards, but prevents inward movement;
c) stresses within the wall and the foundation do not lead to local failure;
d) vertical accelerations are ignored;
e) rotations are small, ie. errors associated with small changes of geometry can be
ignored.

A gravity retaining wall subject to lateral acceleration but prevented from sliding
can nevertheless deform in a variety of modes. The wall can rotate about its foundation,
if this is not rigid; or it can deform in bending or shear; or it can rock, which will be
initiated when the overturning moments exceed the righting moments and provided that
the stress state in the interior of the wall and in the foundation is acceptable. For the
purposes of this analysis, then, rocking can be defined as a permanent rotational
movement of the wall, analagous to sliding, which would represent permanent
translational movement.

As the lateral acceleration is increased, the horizontal inertia force on the wall
and the dynamic earth pressure force create an overturning moment sufficient to
overcome the restoring moment created by the wall self-weight. This acceleration is
defined as the threshold acceleration in rotation N, = ~ g, which can be found by
iteration from moment equilibrium :

Nr M Yc + PAE cos (o+l3) li =M g xc + PAE sin (o+P) ( B - h sin P ) (4)

The equation of motion for the wall is then (taking moments about Point 0) :

(5)

where Ic is the polar moment of inertia of the wall, 2c is the acceleration of the centroid
and i:l is the angular acceleration of rotation about Point 0 . At the centroid of the wall,
the acceleration has three components, one due to the lateral ground acceleration Rg, and
one due to the rotation of the structure and the third due to the centrifugal acceleration
of the wall. Expressed vectorially:

(6)

@seismicisolation
@seismicisolation
 42 RETAINING STRUCTURES ANALYSIS AND DESIGN

where ~ is the vector from Point 0 to the centroid and ro is the angular velocity of the
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

wall. Therefore :
PA£ cos (I'>+P) li - M g xc - PAE sin (o +P) ( B - li sin P)
(7)
= Mr c2 B - May
g c
+ I c El

from which 8 can be determined . The velocity of rotation is then given by :

e = f edt for t'l > 0

(8)
0

=0 for t'l s; 0

and the displacement can be calculated simply by integrating the velocity over time. On
this basis the model assumes that the rocking displacement starts to accumulate once the
threshold acceleration is exceeded and continues accumulating until the rotational
velocity becomes zero again, which will occur some time after the ground acceleration
has fallen below the threshold in rocking.

Model Behaviour

For clarity, the model is here demonstrated by application to a single pulse of

acceleration which rises instantly to a value of O.Sg, remains constant for 0.5 seconds
and then reduces instantly again to zero.

Consider the case of a rectangular mass concrete wall founded on a rigid

foundation (as Fig.l with~ = 0). Using sample data ofH = 10m, B = 5 m, y, = 24
kN/m3 for concrete, y, = 16 kN/m for soil, 8 = 20 ° and 4> = 40 ° and adopting the
Mononobe-Okabe solution, P AE is simply expressed as :

1
PAE -- Z KA£ Y, H 2 = 800 K AE '-'"
I<JYifll

and by iteration, assuming the dynamic earth force acts at H/3 above the base, the
threshold in rotation can be calculated as N, = 0.38g. Rearranging Equation (7), noting
~ = 0 and substituting for x. = B/2, y. = H/2 and ii = H/3 gives:

2
( M 1'c +Ic ) e= P AE cos 0 ~ - M g %- p AE sin 0 B + Mag ~ (9)

The polar moment of inertia about the centroid, I. = lx + ly = BH(B 2+H2)p/12.

@seismicisolation
@seismicisolation
 LARGE GRAVITY WALLS ROTATION 43
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Solving for the rotational acceleration i:) and integrating gives the velocity and
displacement as :

e" 0.21 rad/sec 2 '

e , e + e co) = o.21 1
(10}
e co) 1 + -e2t
2
e = e co) + o.1o5 1
2

Fort = 0.5 seconds, 8 = l. 5o . At this point, the lateral acceleration falls below
the threshold again, and gravity slows the rotational velocity to zero. Equation (9) is
reemployed to compute the new rotational acceleration e = -0.543 rad/sec2 (constant
in this simple example as a8 = 0 fort > 0.5 sec) which deccelerates the wall until the
rotational velocity reaches zero at time th when :

e = e co.5) - o.543 c t 1 - o.5 ) =o (11)

giving t 1 = 0.693 seconds, and a maximum rotation angle of :

8 =8 (0.5) + S (0.5) ( 11 - 0.5 ) + ~ ( 0.5 ) ( 11 - 0.5 f = 2.09 ° (12}

2

After t = t~> the model would predict that the wall should start to rotate
backwards into the soil. This moment is interpreted as the instant at which permanent
outward movement ceases and the wall and soil continue to follow the ground motion
elastically, until the lateral acceleration once again exceeds the outward threshold for
rocking.

Unlike the sliding mechanism postulated by Newmark, which can effectively

isolate the slipping block from the plane beneath, the rocking model continues to
transmit lateral accelerations through the ' pin' into the wall. Elastic deflections may then
be expected to be superimposed onto the permanent plastic movement predicted by the
model throughout the period of rocking, as will be shown below in an examination of
model data.

Validation of the model

Centrifuge tests of a monolithic concrete wall have been carried out following
well established principles for dynamic modelling, Schofield (1981}, Schofield and
Steedman (1988}, using the Equivalent Shear Beam model container, Zeng and
Schofield (1996) and the Bumpy Road earthquake actuator at the Cambridge
Geotechnical Centrifuge Centre. The model wall was constructed from three micro-
concrete blocks laid on top of each other to represent cracked construction joints. A

@seismicisolation
@seismicisolation
 44 RETAINING STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

thin ribbon of glue along the front (outer) edge ofthe blocks was used to provide a shear
connection, preventing sliding movement and therefore constraining the wall to rock
only.
44.8m
24.0m
...
end walls
and shear
sheet of
ESB
model

micro-concrete block wall

Fig. 2 Section through centrifuge model of

a concrete gravity retaining wall

Fig. 2 shows a cross section through the model, expressed in prototype

dimensions (17.6 m high with a base width of 7.4 m). The backfill material was a fine
sand (Leighton Buzzard 52/100, Specific Gravity 2.65) with a typical particle size of
around 0.25 mm and an initial void ratio of0 .763-(Relative Density 48%), reducing to
0. 698 (Relative Density 67%) after five earthquakes, based on the surface settlement.
The permeability of the sand was measured as 3.7 x 10-4 rn!s at a void ratio of0.76,
falling to 2 .6 x 10-4 rnls at a void ratio of 0.68, which using water in the model is
equivalent to a field permeability of around 0.03 rnls, ie. relatively free draining. The
critical state angle of shearing resistance for Leighton Buzzard sand can be taken as
33 o . A surcharge of28 kN/m2 was placed on top of the backfill in the form of a solid
aluminium plate, with a 2 mm (model scale) gap between the front edge of the plate and
the top of the wall. The strip of glue cementing the blocks together along their front
edge was about 3 mm wide (model scale) and, based on load tests which indicated a
shear strength of 10 MN/m2, provided an additional shear resistance of around 2400
kN/m. The backfill was water saturated with a water table at 4.2 m depth below the
ground surface. The test was run at 80 gravities.

The model wall was subjected to five earthquakes with peak lateral base
accelerations and velocities as indicated in Table 1. Although the walls were massive in
cross-section, their large height contributed to some amplification of the input motion,
with an average amplification 'of the peak acceleration of 1.36 at the wall centroid and

@seismicisolation
@seismicisolation
 LARGE GRAVITY WALLS ROTATION 45
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

this. was taken into account in the prediction of the threshold acceleration and the
outward movement as this has an important influence on the rocking model. A second
factor also accounted for in the model was the amplification in the soil, which reduced
from 1.82 in earthquake 1 (EQl) to 1.18 in EQS, probably related to densification
(noted above) .

E arthquake Peak +vc Peak -ve Peak +ve Peak -ve Threshold for
Acceleration Acceleration Velocity (m/s) Velocity (m/s) rotation (%g)
(%g) (%g)

EQ1 15.0 -1 5.2 0 . 110 -0.106 10.7

EQ2 19.8 -19. 2 0 .133 -0. 125 11.8

EQ3 16.4 - 17.8 0 .132 -0. 131 12.9

EQ4 22.5 -1 7.7 0 .147 -0. 157 13.9

EQS 21.9 -24.1 0 .174 -0.172 15.0

Table 1

The densification observed in the models had two effects which were considered
separately in the analysis of the data: firstly the effect on strength, which was accounted
for using correlations with relative density developed by Bolton (1986), and secondly
the effect on soil depth behind the wall, which had reduced by 0.8m (4.5%) total after
the five earthquakes.

The peak mobi lised angle of shearing resistance in the soil can be predicted as
a function of the effective confining pressure, the relative density and the critical state
angle of friction for plane strain or triaxial conditions using Bolton (1986) . In this
analysis plane strain conditions were adopted, and based on the effective confining
pressure at mid-depth behind the wall (as a typical mean value associated with a zone
of shearing) the peak value of~ was predicted as 42° , rising to 47.5° over the five
earthquakes in the series. The angle offriction on the back of the wall (li) was taken as
2/3 ¢.

A number of factors were ignored in the analysis. Firstly, although the angle of
fiiction on the base of the wall is clearly not significant in rocking, even a small tensile
capacity on this surface will have an important effect on the onset of outward
movement, raising the threshold for first yield. No tensile capacity was assumed in the
prediction of the model behaviour as the blocks were not cemented down. (In a design
case this would be a conservative assumption.)

Secondly, although the use of water as a pore fluid effectively increased the
permeability and reduced the development of residual excess pore pressure, cyclic excess
pore pressures were observed which reflect the change in effective confining pressure

@seismicisolation
@seismicisolation
 46 RETAINING STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

acceleration (%g)
30.-----------------------------------------------~

20

10 -

-10

soil top
-20
base input
EQ2

2 4 6 8 10 12
time (seconds)

Fig. 3 Amplification through the soil backfill

wall displacement (mm)

0.8.---------------------------------------------------.
EQ2

-0.2 '----~---'----'----'---~----'----~----'---~----'----~---'

0 2 4 6 8 10 12
time (seconds)

Fig. 4 Displacement at wall top, middle and base

@seismicisolation
@seismicisolation
 LARGE GRAVITY WALLS ROTATION 47
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

as the shear stress varied through the earthquake shaking. These did not appear to have
a significant influence on the prediction of outward movement. Similarly, vertical
accelerations were neglected as these were small and of a much higher frequency than
the base input motion. Table I shows the predicted threshold accelerations for each
earthquake.

The nature of the earthquakes generated using the ESB and the Bumpy Road
earthquake actuator are shown in Fig. 3 which compares the acceleration near the
ground surface with the base input motion for EQ2. Some amplification of the peak
motion is observed (1.44 in this exainple) and there is a small phase lag between ground
surface and the input motion, as expected. The earthquake motion comprises 10 major
cycles at a frequency of around 1.5Hz. Strong ground motion with a single dominant
frequency and multiple large cycles is typical of soft ground sites and is particularly
damaging to geotechnical structures. For this reason and also because of the ease with
which the data can be examined in the time domain, the 'tone burst' input is considered
to be highly appropriate to the validation of analytical and numerical models.

In each earthquake cyclic movement of the wall was observed, coupled with
incremental non-recoverable outward displacements. This outward movement was
clearly associated with a rocking mode of behaviour, as displacement transducers at mid-
height and at the base of the wall showed reduced and zero displacement respectively.
This can be seen clearly in Fig. 4 which shows the outward displacement of the top of
the wall, of the middle wall block and of the bottom wall block for EQ2.

The cyclical displacements which are superimposed onto the permanent or

residual rocking movement are caused by a dynamic elastic interaction between the wall
and the ground. The magnitude of typical recoverable displacements can be illustrated
by considering the· deflection of an elastic shear beam, representing a soil column, under
a pseudo-static lateral acceleration field . For a shear modulus G proportional to the
square root of depth, the surface deflection can be expressed as

X = (13)

where Gb is the shear modulus at depth H. For kh = 0.2, this predicts an elastic surface
ground displacement of around =F30 mm for values of shear modulus consistent with the
density and strain levels within the sand backfill, which compares favourably with the
order of magnitude of the observed cyclic displacements. ·

Figs. 5, 6, 7, 8 and 9 show the pattern of outward movement in each earthquake

compared with the predicted permanent movement using the model described above.
Dynamic effects on the displacement transducers will influence the magnitude and phase
of the dynamic response of these transducers to some extent, but the permanent

@seismicisolation
@seismicisolation
 48 RETAINING STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

acceleration (%)
20 ,-----------~--------------------------------~

15
threshold
10 ----·········-----------· •• --------------------·-···-·····-
5

-5
base input
-10

-15 EQ1
-20~--~---L--~----~--~---L----~--~--~--~~--~--~
0 2 4 6 8 10 12
time (seconds)

wall top displacement (mm)

20~----------------------------------------------------~

15 EQ1

- 10~--~--~--~---L--~--~L---~--J----L--~----~__J
0 2 4 6 8 10 12
time (seconds)

Fig. 5 Base input acceleration, and w,al l top

deflection for Earthquake 1

@seismicisolation
@seismicisolation
 LARGE GRAVITY WALLS ROTATION 49
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

acceleration (%g)
30

20
threshold
10

-10
base input

-20
EQ2
-3oL---~--L---~--L---~--L---~--L---~--L---~~
0 2 4 6 B 10 12
time (seconds)

wall top displacement (mm)

50

40 EQ2
30

20

10

-10 actual
-2oL---~--L---~--L---~--L---~--L---~--L---~~
0 2 4 6 B 10 12
time (seconds)

Fig. 6 Base input acceleration, and wall top

deflection for Earthquake 2

@seismicisolation
@seismicisolation
 50 RETAINING STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

acceleration (%)
20

............... ......... .. threshold

---········-················

EQ3
-2 0~--~--J---~---L--~----L---~--J_--~ ___ J_ _ _ _~__J
0 2 4 6 8 10 12
lime (seconds)

wall top displacement (mm)

50

40 EQ3
30

20

10

-10

-2 0~--~---L--~----L---~---L--~~--~--~--~L---~--~
0 2 4 6 8 10 12
time (seconds)

Fig. 7 Base input acceleration, and wall top

deflection for Earthquake 3

@seismicisolation
@seismicisolation
 LARGE GRAVITY WALLS ROfATION 51
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

acceleration (%g)
30

20 threshold
··············· .......•. . .. ... ····-· ·········-······-·

EQ4
-20L---~---L--~----L---~--~----~--~--~--~~--~--~
0 2 4 6 B 10 12
time (seconds)

wall top displacement (mm)

80

60
EQ4

40

20

actual

2 4 6 8 10 12
time (seconds)

Fig. 8 Base input acceleration, and wall top

deflection for Earthquake 4

@seismicisolation
@seismicisolation
 52 RETAINING STRUCTURES.ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

acceleration ('log)
3 0.-----------------------~--------------------~

20
threshold
·········-···-···
10

-10

-20
EQ5
-30~--~~L-~--~---J--~--~--~--~--~L-~~~
0 2 4 6 8 10 12
time (seconds)

wall top displacement (mm)

80,-----------------------------------------------~

60 EQ5
40

20

-20
actual

2 4 6 8 10 12
time (seconds)

Fig. 9 Base input acceleration, and wall top

deflection for Earthquake 5

@seismicisolation
@seismicisolation
 LARGE GRAVITY WALLS ROTATION 53
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

movement is accurately indicated. Good agreement is seen between the observed and
the predicted start of the permanent displacement, indicating that the calculation of the
threshold acceleration is close to the actual, and between the magnitude of the predicted
permanent wall displac!!ments and the observed. The cumulative residual rocking
displacement after five earthquakes was predicted as 68 mm, compared to an actual
displacement of 64 mm.

The nature of the wall acceleration after the onset of rocking can be seen in Fig.
10, which compares the base input motion with the wa,ll top acceleration. Large
negative .accelerations are seen to take place as the wall ' reconnects' with the ground
and peak positive accelerations climb above the threshold as the outward rocking
movement takes place. This amplification of motion has an important influence on the
outward movement as it increases the overturning moment. This was taken into account
in the analysis, which used the average amplification on the positive peak accelerations
ofl.36 evaluated at the wall centre of inertia, as noted above. The rounded nature of the
peak accelerations during rocking contrast with the flat ' caps' that have been observed
from experiments studying the sliding of large walls, Steedman (1984) and occur
because, unlike sliding, the rocking wall remains attached to the ground at the pin.

Comparison with design aP-proaches

In design, it is common to assume a critical state angle of shearing resistance for

the soil backfill and for the soil-wall interface. Fig. 11 presents a comparison between
the predicted and observed residual rocking displacements of the wall, normalised by the
height of the wall, and a 'design line' calculated using a value of¢ = 8 = 33 • . (The
observed displacements being plotted using predicted thresholds for simplicity.) The
earthquake record EQl was used for the .integration process as representing a typical
form of the earthquake to which the wall was subjected. Again, there is good agreement
with the data; such a comparison provides reassurance as to the general robustness of
the model for predicting permanent rocking displacement.

Richards and Elms, and Whitman and Liao, note the importance of addressing
uncertainty in the prediction of permanent displacement. Two major areas of uncertainty
are the nature of the earthquake and the prediction of the threshold acceleration. For
comparison purposes, the observed residual displacements from Eqs 1-5 were plotted
against Equations (1) and (3) in Fig. 12 where the data is seen to lie well below the
Whitman and Liao line. Taken together with the proposed rocking model, which
analysed the particular class of earthquake and the actual mechanism of behaviour, this
suggests that there is opportunity for reducing conservatisms in design by elimination
of uncertainties where these can be evaluated. Thus an improved estimate of the
permanent rocking displacement of the large wall in this example was achieved by use
of a similar form of earthquake to that actually experienced, and by use of a rocking
model for a wall known to be susceptible to rocking.

@seismicisolation
@seismicisolation
 54 RETAINING STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

acceleration (%g)
40.----------------------------------------------.

wall top
-40 - threshold
EQ2
base input
- 60L---~--L---~~L-~~~--~--~L---~--L---~~
0 2 4 6 8 10 12
time (seconds)

Fig. 10 Lateral acceleration ofdte wall in rocking

residual rocking displacement(%)

typical design envelope for residual

1,------ J--- - - l - - - - rocking displacement using scaled
earthquake record (EQ 1)

0.01

0.4 0.5 0.6 0.7 0.8 0.9

NIA

Fig. 11 Comparison with. design line for rocking

displacement

@seismicisolation
@seismicisolation
 LARGE GRAVITY WALLS ROTATION 55
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

residual displacement (%)

5

2 "l 1
design envelope fo~ residual
1 displacement for a typical -
- Richards ~ earthquake, based on EQ 1 -
0.5
and Elms•· ••• ~ ...f-.. ..
.... ~
......._,

-
0.2
··---~ ~
•••
0.1

0.05

0.01
0.4
actual ( • ) residual
0.02 - displacement for EQs
I
0.5
I
0.6
•
1-5
- 0.7
1-

0.8
~Whitman -
and Uao

0.9
N/A

Fig. 12 Comparison with conventional design approaches

Conclusions

A rocking model has been developed for the prediction of permanent

displacement of retaining walls analagous to sliding models currently used for design
purposes. The model uses a Newmark approach to predicting outward rocking
movement, which is triggered once a threshold acceleration in rotation is exceeded and
arrested once the wall velocity and the ground velocity are again identical.

The model has been validated by comparison with experimental data of a large
gravity wall constrained to rock and subjected to a series of damaging earthquakes. The
data showed clearly that the rocking motion is comprised of two elements: an elastic
recoverable movement and a permanent or residual rocking displacement. The rocking
model showed good agreement between the predicted residual displacement of the wall
top and the observed displacement, both using general design parameters and specific
experimental parameters.

Comparison with standard available design approaches illustrate the extent of

conservatism& inherent in a general approach and suggests that a mechanism-specific
rocking model such as that proposed here may be more appropriate to the prediction of
permanent rocking displacement, particularly where the nature of the design earthquake
can be defined.

@seismicisolation
@seismicisolation
 56 RETAINING STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Acknowledgements

The model was coded using Mathcad with the valuable assistance of Dr R E May
and his staff in the Geotechnical Division at Sir Alexander Gibb & Partners Ltd. Their
contribution is gratefully acknowledged.

References

Bolton MD (1986) The strength and dilatancy of sands, Geotechnique 36, No. 1, pp65-
78 .

Ebeling R M and Morrison E E (1992) The seismic design of waterfront retaining

structures, US Army Corps of Engineers, Technical Report ITL-92-11, Department of
the Army, Washington DC 20314-1000.

Franklin A G and Chang F K (1977) Earthquake resistance of earth and rockfill dams,
Report 5: Permanent displacements of earth dams by Newmark analysis, Misc. PaperS-
71 -17, USAE Waterways Experiment Station, November.

Nadim F (1980) Tilting and sliding of gravity retaining walls, SM Thesis, Dept Civil
Eng, MIT.

Newmark N M (1965) Effect of earthquakes on dams and embankments, 5th Rankine

Lecture, Geotechnique 15, No. 2, pp 139-160, June.

Richards Rand Elms D G (1979) Seismic behaviour of gravity retaining walls, Proc
ASCE, Journal Geotech Eng Div, 105, GT4, April, pp449-464.

Schofield AN (1981) Dynamic and earthquake geotechnical centrifuge modelling, Proc

Int Conf Recent Adv Geotech Earthquake Eng and Soil Dynamics, Rolla, Vol.3,
pp1081-1100.

Schofield AN and Steedman R S (1988) Recent development of dynamic model testing

. in geotechnical engineering, Proc 9th World ConfEarthquake Eng, Tokyo-Kyoto.

Steedman R S (1984) Modelling the behaviour of retaining walls in earthquakes, PhD

Thesis, Cambridge University.

Whitman R V and Liao S (1985) Seismic design of gravity retaining walls, Misc. Paper
GL-85-1, USAE Waterways Experiment Station, January.

Zeng X and Schofield AN (1996) Design and .performance of an equivalent shear beam
container for earthquake centrifuge modelling, Geotechnique 46, No.I, pp83-l 02,
March.

@seismicisolation
@seismicisolation
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

IN-SITU DYNAMIC RESPONSE OF CANTILEVER WALLS

Sreenivas Alampalli 1, Member, ASCE, and Aluned-W. ElgamaF, Member, ASCE

Full scale dynamic testing is undertaken to monitor the vibrational characteristics

of cantilever wall-soil systems. Impact hammer as well as controlled shaker excitation
at different force levels are employed. Two different reinforced concrete cantilever
wall-soil systems are studied. The employed testing techniques are described and the
recorded modal characteristics are presented. These modal characteristics may be used
for the purpose of seismic design; and for calibration of computational and/or small
scale laboratory centrifuge models.

Introduction

Damage to wall-backfill systems is often reported after earthquakes (Prakash

1981, Whitman 1991), mainly due to the phenomenon of soil liquefaction (Whitman and
Christian 1990); with significant financial expenses in repair or replacement. Recorded
quantitative seismic response of wall-backfill systems is virtually nonexistent.
However, the characteristics of this response are being increasingly rationalized through
experimental investigations. A large growing body of shake table and centrifuge test
results have been primarily focussed on wall-soil deformations under conditions of
strong dynamic excitation and/or soil liquefaction (Matsuo and Ohara 1960, Murphy
1960, Sherif et al. 1982, 1984, Ortiz 1982, Ortiz et al. 1983, Steedman 1981 , 1984,
Bolton and Steedman 1982, 1984, Andersen et al. 1987, Whitman and Ting 1993, Zeng
1993, Alampalli 1990).

1
Acting Head, Structures Research, Transportation R & D Bureau, New York
State Department of Transportation, 1220 Washington Avenue, Albany, NY
12232-0869.
2
Associate Professor, Department of Civil and Environmental Engineering,
Rensselaer Polytechnic Institute, Troy, NY 12180-3590.

57

@seismicisolation
@seismicisolation
 58 RETAINING STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

This paper addresses the in-situ dynamic response characteristics of reinforced

concrete cantilever retaining structures. Modal parameters (resonant frequencies,
damping ratios, and associated resonant mode shapes) of two retaining walls were
measured using different excitation techniques. Such modal parameters may be
employed for the purposes of seismic design. In addition, the recorded mode shapes
displayed a three dimensional response pattern, not accounted for by current analysis
procedures. In the following sections the employed testing procedures are described, and
salient results of these fu!J scale testing programs are presented within a unified
framework.

Test Structures

Two cantilever reinforced concrete retaining wall systems were tested. The first
was a small wall rigidly attached to the south side of the Jonsson Engineering Center
(Fig. I) at Rensselaer Polytechnic Institute (RPI). The wall (referred to hereafter as the
JEC wall) was 241 mm thick with a maximum height of 5 m. A soil backfill was
supported by this wall from the basement elevation to the ground surface (Fig. 1). The
second wall system, also on the RPI campus, was a long wall (42.7 m) of varying height
from 1.4 m to 3.05 m and was 406 mm thick (Fig. 2). This wall (referred to hereafter
as the CII wall) supported an elevated parking lot. A construction joint was visible at
about 22.7 m from the short end of the wall (Fig. 2).

Testing Procedure

In all tests, the structures were excited by an impact hammer or a shaking system
mounted on the wal l or on the backfill soil. Testing by impact hammer was less time
consuming, but imparted minimal energy at low frequencies (e.g., below 10 Hz).
Hence, for massive structures with resonant frequencies below 10 Hz, the impact
hammer technique may not be adequate as an excitation mechanism for the fundamental
mode. Tests were conducted on the JEC wall using two different impact hammers and
a small electrodynamic shaker, in order to compare the effectiveness of these methods
in testing wall-soil systems. For the CII wall, the impact hammer was used, and an in-
situ harmonic shaker was built at RPI in order to impart significant energy in the low
frequency range up to 17Hz. This section describes salient features of the test setup and
testing procedures.

Hammer and Shaker Tests

A typical test configuration is shown in Figure 3. It consists of a dynamic signal

analyzer to obtain real-time frequency domain representations of the recorded signals
with specified accuracy tolerances to assure measurements quality; a shaker (or an
impact hammer with a force transducer) for excitation; accelerometers for measurement
of response; amplifiers; signal conditioners; ~d a micro computer with a floppy and

@seismicisolation
@seismicisolation
loaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights res

"
I" 9 m----------------------------------~
~ ··-. . .-

Ia ()
)>
1 z
~
~ Soil rii:>:!
-~ ...~
Elevation View .....L. ~f;;

'
~
... "'
·~

t- "'E ·
<...>
w
....,
~

-, ~
)
....
0

B : Location of mounted shaker for

Random testing.
1eter
"' ~ "0
~ lSI
.....
)(

.....

Wall Cross-
Section
51
"'='
~Cll
ti1

Figure 1. JEC Wall with Grid of Measurment Points

Ul
\0

@seismicisolation
@seismicisolation
 60 RETAJNING STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

,...
...c
"(5
a.
~~ 'O
I •
,M / E
/
co
ci
'

...c
/ "(5
""')
c
0
-.::;
0
E
,.... ......
:I

Ill
C\i c
'<t 0
II 0
...
..c:
C'l
c
Ill
...J ~
>c
0
-.::;
c Ill
0
-.::; >
Ill
Ill jjj
,... >
.,...
...c Cll
jjj iii
3:
·a ~
a.
0
Ill
cc -
0
0
0
-.::;
Ill
E
Cll
..c:
0
(/)

co
,... C\i
....c ...
Cll
:I
·a C'l
iL
0.

@seismicisolation
@seismicisolation
 CANTILEVER WALLS DYNAMIC RESPONSE 61
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Accelerometer

Cables

Wall
Accelerometer
Backfill

Figure 3. Schematic of Shaker Test Setup

@seismicisolation
@seismicisolation
 62 RETAINING STRUCTURES ANALYSIS AND DESIGN

hard disk drive. Whenever possible, at the beginning of each test, reciprocity was
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

verified by comparing frequency response functions (FRF) obtained through

interchanging input and output locations. During a test, measurement locations were
chosen in such a way so as to reflect the behavior of the structure in the modes of
interest. Only the dynamic wall response perpendicular to its plane was measured.

In case of impact testing (Elgamal et al. 1990, Alampalli 1990) a 3 lb impulse

hammer (PCB model 086B20) or a 12 lb impulse sledge hammer (PCB model 086850)
were used. These harnmers were equipped with a load cell attached to the hammer tip
to measure input force imparted to the structure. An effort was made to keep the blows
as uniform as possible for repeatability. A PCB accelerometer (model 393C) was used
to measure the output acceleration. Excitation input Y;(t) was imparted at each sampling
point and the resulting acceleration output yoCt) was measured at a stationary point. An
input-output FRF was computed by the analyzer, which also performs signal digitW!tion
using appropriate anti-aliasing filters. This function may be defined as:

where f denotes frequency, uppercase denotes Fourier Transform, and(') denotes the
complex conjugate operator. The process was repeated for the same point and the
average of a number of FRFs was finally stored as the input-output transfer function for
this point. Inspection of the Coherence function (Y;)- also computed by the analyzer-
dictated the number ofFRFs to be averaged, where

in which O.Osy; 2(f)d.O. A coherence of zero denotes that yo(t) and y!Ct) are unrelated,
whereas a coherence of unity denotes that y.(t) is related to Y;(t) with no interference or
noise effects. In this study, a highly satisfactory coherence of 0.95 or greater was
always achieved.

In case of random excitation testing (Elgamal et al. 1990, Alampalli 1990) an

electrodynamic shaker (42lb maximum vector force, 76 em/sec velocity and 15.875 em
peak to peak stroke) manufactured by Acoustic Power S)'Stems Inc. (model No. 113)
was used. The HP3562A signal analyzer supplied a control signal to the shaker which
was amplified using a power amplifier (Encore Electroni~s Inc., model 505 with 700
Watts peak capacity and a current control feed-back system). Accelerometers (Kistler
model 8360AS with lV/g sensitivity) were used to measure shaker excitation as well as
acceleration response output at any particular sampling point on the structure. The
shaker along with an attached reaction mass was rigidly mounted at a convenient
location on the structure. The HP3562A was used to digitize incoming input and output
data from the two accelerometers, and to calculate input-output FRFs as well as the
associated coherence function (measurements at each point were repeated and averaged).

@seismicisolation
@seismicisolation
 CANTILEVER WALLS DYNAMIC RESPONSE 63

This coherence function as mentioned above was used as the data quality control
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

criterion.

A modal parameter estimation package (Modal 3.0 SE), developed by Structural

Measuring Systems (SMS) was employed for extraction of modal parameters (linearity
and reciprocity for the test structures is assumed). This package offers various options
for curve fitting. Three different options were employed in this investigation and are
discuss~d by Aliunpalli (1990) in detail. A different approach for testing and curve
fitting was adopted when harmonic;: excitation (swept sine mode) was employed. It was
found when performing such a test that a substantial amount oftime was expended in
trying to achieve a high accuracy variance convergence tolerance (in this case 1 %) in
the frequency bands away from a resonance. In these frequency bands, no significant
amplification occurs and the signal amplitude approaches that of ambient noise.
Attaining high convergence tolerances for such low amplitude signals was difficult and
consumes significant time with minimal impact on overall quality of the measured FRF.
In order to reduce the time required, while maintaining high data quality, the procedure
described by Alampalli and Elgamal (1991) was adopted when testing with swept sine
excitation.

Large Shaker Test

An eccentric-mass vibration generator (shaker) developed at RPI (Van Laak and

Elgan1al 1991) for resonance testing of full-scale structures based on the original
Hudson (1964) design, was used for these tests. Eccentric mass shakers are particularly
efficient in producing large dynamic forces at low frequencies. The RPI shaker was
designed to provide up to 22 kN of horizontal force within the frequency range 0.5 to
30 Hz. Rotational speed of this shaker was verified to remain stable at any specified
frequency within a 0.0625 Hz range. Operating speed was controlled by a calibrated
precision potentiometer. A 1.8 m square (0.45 m thick) reinforced concrete pad was
constructed flush with ground surface to provide a firm base for attachment of the
shaker. Terra Tek (model SSA102) force balanced accelerometers were used to measure
the horizontal wall and backfill accelerations. The HP3 562A signal analyzer was used
to digitize and record time histories; and to calculate transfer functions and coherence
functions between the shaker input and response accelerometer signals (Elgamal,
Alampalli, and Van Laak 1996). Testing was conducted over several nights in the
absence of traffic induced vibrations.

Test Results

JECWal l

Transfer functions of 85 measurement locations along the exposed side of the wall
(Fig. 1) were measured (Elgamal et al. 1990, Alampalli 1990). The entire wall was

@seismicisolation
@seismicisolation
 64 RETAINING STRUCTURES ANALYSIS AND DESIGN

tested three times, each time employing a different excitation technique (see Table 1).
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Natural frequencies and damping ratios obtained by curve-fitting the data (up to 100Hz)
of each test are shown in Table 2. Note that the small hanuner even with multiple
impacts (Tables 1 and 2) was unable to supply enough energy to excite the lowest
frequency mode. While estimates of natural frequencies were in agreement, a noticeable
difference in estimated damping may be observed (Table 2). The damping ratios shown
in Table 2 were found to depend on the employed curve-fitting algorithm/procedure and
consequently should be taken as rough estimates. Mode shapes obtained from all three
tests appeared to agree well (first mode was not obtained from the sm;~ll hammer test)
with maximum variation of 8% in natural frequencies. Modes obtained by the swept
sine shaker test (the third test) are shown in Fig. 4. Based on the above results, it may
be concluded that the results obtained by sledge hammer impact and those obtained by
harmonic shaker excitation are in good agreement.

It ]s noted that the modes shown in Fig. 4 were evaluated in the complex domain
and were fmmd to exhibit a gradual phase change (phase at resonance is not simply 0.0
or 180.0 degrees) as shown in Fig. 5. Such a phase relation between different points of
the structure may be due to the presence of non-proportional damping mechanism
(Ewins 1984), such as that due to radiation. It is also of interest to note that nonlinear
dynamic response of the wall was detected when a particular transfer function was
measured several times using a shaker force of varying amplitude (Fig. 6). The response
is similar in character to that of a nonlinear yielding system (Jennings 1964, Nayfeh and
Mook 1979), and should be a subject of further investigations.

High Frequency Tests: Initially, the same test setup as described for the JEC Wall
was used. Transfer functions of 178 measurement locations along the exposed side of
the wall were measured. Dynamic excitation was imparted using the large 12 lb sledge
hammer (Table 1). During the curve-fitting phase, it became evident that extremely
high modal coupling existed due to close resonances. In addition, the presence of
significant damping smoothed off most peaks in the measured transfer functions. This
complicated the curve-fitting process and no reliable resonant properties Clll\ be obtained
in the low frequency range. At higher frequencies, the natural frequencies and
associated mode shapes are given in Figure 7. A rough estimate of dari)ping ratios
suggested a value of about 8% for most of these modes.

Low Frequency Tests: The new large RPI shaker was employed in this case
(Elgamal et al. 1996). A plan view of the shaker and measurement locations on the wall
and backfill are shown in Figure 8. The shaker horizontal force and all measurements
were oriented in the direction perpendicular to the wall face. Data was recorded starting
at 6 Hz (before the fust resonance identified by analysis) and up to 17 Hz using a
frequency increment of0.125 Hz at points (Fig. 8) 2, 2A, 8, &A, 8B, 11, llA, and 0 (as

@seismicisolation
@seismicisolation
 CANTILEVER WALLS DYNAMIC RESPONSE 65

Table I. Summary of user defined settings for various excitation mechanisms

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

employed. .

Testing Excitation Frequency No. of Coherence Curve-fit

Device Employed (Hz) Avgs. Value Procedure
(Note 1) (Note 2) (Note 3) (Note 4)
Range 1.\f

Small Multiple 20-180 0.2 10 0.9-1.0 @

Hanuner Impact
(3 lb
weight)
Large Single 5-405 0.5 5-10 0.9-1.0 @,$,#
Hanuner Impact
(l2 lb
weight)
Electro Swept 10-100 0.11 2 0.98-1.0 *
Magnetic Sine (Note 5)
Shaker Constant
(with 8lb Force
force)

Notes:

I. A force-exponential window was employed with the large hanuner time domain data
(Alampalli 1990).
2. Number of averaged signals for each measurement.
3. Value ofthe coherence function in the frequency range of interest.
4. Different curve fitting techniques were used, which are briefly explained in this
paper. $ indicates Peak curve fitting method, @ indicates Rational fraction Polynomial
curve fitter with auto fitting method,# indicates Global curve fitting techniques, and *
indicates the use of the curve fitting algorithm developed by Alampalli &nd Elgamal
(1991).
5. Integration at each frequency point is done until I% variance limit on the value of
frequency response function is met or until the maximum integration time (120 sec. in
this case) is reached. This procedure is repeated for each average.

@seismicisolation
@seismicisolation
 66 RETAINING STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Table 2. Natural frequencies and damping ratios for JEC wall using three different input
excitation mechanisms

Mode Natural Frequency (Hz) Damping Ratio (%)

No.
Hammer Hammer Shaker Hammer Hammer Shaker
3lb 12 lb 18lb 3 lb 12 lb 18 !b
1 ----- 22.65 23.03 ----- 5.86 2.60
2 39.25 38.09 37.05 9.97 5.90 4.90
3 56.36 54.65 52.16 3.85 5.66 3.86
4 71.69 71.96 67.96 3.44 5.31 2.97
5 98.07 97.35 94.42 0.56 i.08 3.00

Table 3 Resonant frequencies of CII System

Resonant Frequencies (Hz)

Mode
Shaker Test FEM
(excitation to Soil) (Elgamal eta!. 1996)
l 6.7 6.67
2 8.2 7.75
3 9.5-9.7 8.86
4 10.4 10.27
5 11.3 I 1.75

Table 4. Observed viscous damping by half-power bandwidth method.

Mode Frequency (Hz) Damping(%)
1 6.7 8.7
2 8.2 15.9
3 9.5-9.7 12.2
4 10.4 8.7
5 11.3 . 6.6

@seismicisolation
@seismicisolation
 Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

@seismicisolation
@seismicisolation
CANTILEVER WALLS DYNAMIC RESPONSE

Figure 4. Mode Shapes of JEC Wall in Plan View

I~
67
 68 RETAINING STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

P1an View

~N2 .

N ~ Modal Nod e .
In animation, node appear s to
move in the r ange Nl- N3.

Figure 5. Various Stiii-Vi~ws During JEC Wall Mode 2 Animation

1.0 2 1
4f"-- 1\ 1=3 1-6 1b f rce
2=7 1-2 1b f rce
r\ >( ~
I % ~ x\ \\ \•18.-v 3=14.4 llb ior.
1,
[I U
r
I U •

~~
~
d.IQ)
N~ \ \
,.... ....
. ,... ::l

~· ..... / ~ i\\ 1\\ .

E c:
S...Ol
0 ~
pr ~ ~
Z::E
~~
~~
o .0

65.0 Fr~quency (Hz) 80.0

Figure 6. Transfer Functions Deoicting Nonlinear Yielding
Response

@seismicisolation
@seismicisolation
 Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

@seismicisolation
@seismicisolation
69
 70 RETAINING STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Elevation 75.6 m
0

5:1 Slope

Elevation 78.4 m o

8 Backfill
•

•
,..,E
N
.... • 36.6m
0 •
•
•
• 19.8 m
•
•
•
•
~~v.:) I4.9m
embedded
concrete pad
Stairway

0
Elevation 79.2 m

0 0

Elevation 75.9 m Elevation 75.6 m

• Accelerometer Loc~tlon
Figure a. Schematic Plan-View of Cll Wall-Soil System

@seismicisolation
@seismicisolation
 CANTU..EVER WALLS DYNAMIC RESPONSE 71
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

a reference). The frequency sweep was performed at each measurement point to

detennine resonant frequencies at which significant amplification of motion at or near
.the wall occurred relative to ground motion near the shaker. Once these approximate
frequencies were determined, a second series of measurements were performed in order
to obtain more precise values of the system resonant frequencies . Finally, a high
resolution frequency sweep (df = 0.0625 Hz) was performed around each resonance, as
identified by the first set of tests. Peak amplitude response within each frequency band
was taken to denote a system resonant frequency. Once, a resonant frequency was
identified with required precision, the system was excited at this frequency and the
corresponding amplitude ratios and 'phase relationships were measured at all
measurement points in order to obtain the corresponding resonant configuration.

A typical transfer function is shown in Figure 9. The associated mode shapes are
shown in Figure 10. Based on this test, the estimated resonant frequencies are shown
in Table 3 along with those obtained using a calibrated 3 dimensional finite element
model (Elgamal et al. 1996). The results indicated that the first wall-soil resonant
frequency was 6.72 Hz. The associated resonant wall configuration mimicked that of
a cantilever clamped plate, or a one-dimensional (lD) bending/shear bearp. model along
the height. Along the length, the amplitude gradually increased with the increase in free
cantilever wall height (Fig. 7). The higher resonant configurations were found to display
variation in response along the wall length. Viscous damping, using the half-power
method (Ewins 1984), was roughly estimated to be in the range of6.60 to 15.90 percent
(Table 4). All the obtained resonant configurations showed a gradual phase variation
along the wall length (phase at resonance is not simply 0.0 or 180 degrees). Such a
phase relation (Fig. 9) between different points (complex domain modes) may be
partially due to the presence of non-proportional damping mechanisms (Ewins 1984),
such as that due to radiation; and may also be influenced by the employed localized
shaking mechanism.

Discussion and Conclusions

Two cantilever reinforced concrete retaining wall systems were studied employing
forced vibration techniques. Modal parameters were evaluated in the complex domain.
Within the scope of this work, good agreement was observed between the data obtained
using impact hammer tests and harmonic shaker excitation tests (above 15Hz or so).
A harmonically forced transfer function repeatedly measured under an increasing force
level, suggested the presence of non-linear yielding behavior. The impact hammer was
unable to reliably excite the low frequency 6.7 Hz fundamental mode of the large CII
walL Hence, for massive heavily damped structure, an impact hammer may not be
suitable.

The tested retaining walls were observed to display spatial variability in motion
along the wall length as well as the "Yall height. Resonant configurations were found

@seismicisolation
@seismicisolation
 72 RETAINING STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

L2
0 Pt. 8/shaker
0 Pt. 8Nshaker
1.0
" Pt 8B/shaker
_g
0.8
'"
p::
<.)
"0
3 0.6
c.
E
-< 0.4

0.2

0.0
7 8 9 10
200

100

rl2
CJ)
<.)

8., -100
"'~
0:: -200

-300

-400
7 8 9 10

Frequency (Hertz)

Figure 9. Typical Acceleration Transfer Function

for Points 8, SA, and 88 (0-10 Hz range)

@seismicisolation
@seismicisolation
 CANTILEVER WALLS DYNAMIC RESPONSE 73
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Baseline WALL Wa ll (W)

PLAN VIEW
============================== Row A
BACKFILL
- - - - - - - - - - - - - - - Row B

.\fODE 1 : 6. 7 HZ ~B::;:a::::,se:;;l:;.:i;:;n:::,e_ _ _ _ _ _ _ _ _ __

w
A

w
A
B

Baseline
MODE 2: 8.2 Hz

MODE 3: 9.1 Hz ·~~~~~~~'-------~~w

A
B

Figure 10. Cll Wall-Backfill Mode Shapes during Shaker Test

@seismicisolation
@seismicisolation
 74 RETAINING STRUCTURES ANALYSIS AND DESIGN

to bear significant similarity to those of a clamped plate rather than those of a cantilever
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

beam (plane-strain analysis). Such a 3D response pattern is not accmmted for by current
analysis procedures, and may be of significance in the seismic response of highway
wing-walls of uneven height. ·

In general, it is often observed that most seismic energy during earthquakes is

imparted below 10Hz or thereabout. Hence, many typical highway wall-soil systems
(with h < l 0 m) will primm·ily respond to earthquake loading only in the first mode
along the height. Finally, this testing program suggested a modal viscous damping in
the range of 8% using the half-power bandwidth method . Thus, for typical situations
of seismic excitation with heights less than 10 m, a modal damping of 5% may be a
conservative value.

Acknowledgments

This research was supported by NCEER contract No. 90-1506 under NSF Master
contract ECE-86-07591. Dr. Paul Van Laak played a major role in conducting tills full-
scale testing program.

References

Andersen, G. R ., Whitman, R . V ., and Germaine, J. T. (1987). Tilting Response of

Centrifuge-Modeled Gravity Retaining Wall to Seismic Shaking. Research Report R87-
14, Department of Civil Engineering, Massachusetts Institute of Technology, MA.

Alampalli, S. (1990). Earthquake Response ofRetaining Walls: Full Scale Testing and
Computational Modeling. Ph.D. Thesis, Department of Civil Engineering, Rensselaer
Polytechnic Institute, Troy, NY.

Alampalli , S., and Elgamal, A-W.(l991). "An Efficient Procedure for Harmonic Testing
of Structirres." Proceedings ofIX International Modal Analysis Conference, Florence,
Italy, 1, 664-669.

Bolton, M.D., and Steedman, R . S. (1982). "Centrifug<ll Testing of Micro-Concrete

Retaining Walls Subjected to Base Shaking." Proceedings of Soil Dynamics and
Earthquake Engineering Conference, Southampton, 31i -329.

Bolton, M. D., cmd Steedman, R. S . (1984). "Modeling the Seismic resistance of

Retaining Stmctures." Proceedings of 11th Inti. Conierence on Soil Mechanics and
Foundation Engineering, San Francisco, 1-4.

@seismicisolation
@seismicisolation
 CANTILEVER WALLS DYNAMIC RESPONSE 75

Elgamal, A-W., Alampalli, S., and Van Laak, P. (1990). "Modal Response of Earth
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Retaining Structures." Proceedings ofIIX International Modal Analysis Conference,

Kissimmee, FL, 1, 479-486.

Elgamal, A-W., A1ampalli, S., and Van Laak, P. (1996). "Forced Vibration of Full-
Scale Wall-Backfill System." Journal of the Geotechnical Engineering Division,
ASCE, 122(10).

Ewins, D.J. (1984). Modal Testing: Theory and Practice. John Wiley, New York, 1984.

Hudson, D. E. (1964). "Resonance Testing of Full-Scale Structures." Journal of the

Engineering Mechanics Division, ASCE, 90(EM3), 1-19.

Jennings, P.C. (J 964). "Periodic Response of a General Yielding Structure," Journal of

Engineering Mechanics Division, ASCE.

Matsuo, H., and Ohara, S. (1960). "Lateral Earth Pressure and Stability of Quay Walls
During Earthquakes." Proceedings of Second World Conference on Earthquake
Engng., Tokyo, Japan, I , 165-181.

Murphy, V. A. (1960). "The Effect of Ground Characteristics on the Aseismic Design

of Structures." Proceedings ofSecond World Conference on Earthquake Engineering,
Tokyo, Japan, 1, 231 -247.

Nayfeh, A.H., and Mook, D.T. (1979). Nonlinear Oscillations. Wiley-Interscience.

Oritz, L. A. (1982). Dynamic Centrifuge Testing of Cantilever Retaining Walls. Ph.D.

Thesis, California Institute of Technology, Pasadena, California.

Oritz, L. A., Scott, R. F., and Lee, J. (1983). "Dynamic Centrifuge testing of a
Cantilever Retaining Wall." Journal of Earthquake Engineering and Structural
Dynamics, 11, 251-268.

Prakash, S. (1981). Soil Dynamics. McGraw-Hill Book Company, New York, NY.

Sherif, M.A., Ishibashi, 1., and Lee, C.D. (1982). "Earth Pressures Against Rigid
Retaining Walls." Journal of Geotechnical Engineering Division, ASCE, 108(5), 679-
695.

Sherif, M.A., Fang, Y. S., and Sherif, R. I. (1984). "K. and Ko Behind Rotating and
Non-Yielding Walls." Journal ofGeotechnical Engineering Division, ASCE, 110(1),
41-56.

@seismicisolation
@seismicisolation
 76 RETAINING STRUCTURES ANALYSIS AND DESIGN

Steedman, R. S. (I 981). The Effects ofEarthquakes on Model Retaining Walls . M. Phil.

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Thesis, Cambridge University, Cambridge, England.

Steedman, R .S. (1984). Modeling the Behavior of Retaining Walls il'! Earthquakes.
Ph.D. Thesis, Cambridge University, Cambridge, England .

. VanLaak, P., and Elganml, A-W. (1991). RPI Eccentric Mass In-Situ Shaker. Technical
Report, Department of Civil Engineering, Rensselaer Polytecluuc Institute, Troy, NY.

Whitman, R.V., and Christian, J.T. (1990). "Seismic Response ofRetaining Structures."
Symposium on Seismic Design for World Port 2020. San Pedro, CA.

Whitman, R.V. (1991). "Seismic Design of Earth Retaining Structures (state of the art
paper)." Proceedings ofSecond lntl. Conference on Recent Advances in Geotechnical
Earthquake Engineering and Soil Dynamics. St. Louis, Missouri.

Whitman, R.V., and Ting, N-H . (1993). "Experimental Results of Experiment No. 10."
Verifications ofNwnerical Procedures for the Analysis ofSoil Liquefaction Problems,
Arulanandan, K . and Scott, R.F. (eds.), Balkerna, Rotterdam, 1, 881-891.

Zeng, X. (1993). "Experimental Results of Model No. II." Verifications ofNumerical

Procedures for the Analysis ofSoil Liquefaction Problems, Arulanandan, K. and Scott,
R.F. (eds.), Balkema, Rotterdam, 1, 895-908.

@seismicisolation
@seismicisolation
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Seismic Analysis And Model Studies Of Bridge Abutments

K L. Fishman' and R. Richards, Jr.2 , Members, ASCE

An analytic method is described for determination of the most critical threshold

acceleration for bridge abutments. The method considers the posSibility of sliding,
bearing capacity, or mixed modes of failure. The authors have updated, and extended
the coupled equations of motion that appear in the literature for predicting seismic
induced permanent deformation. Comparisons between observed and computed
model respouses serve to verify the ability of the proposed algorithms to predict
sliding, tilting or mixed modes of defoiiiJJition. Threshold accelerations are
estimated for a representative sample of 40 bridges with free-standing abutments from
New York State. It appears that significant numbers are at risk for an earthquake with
peak accelerations above 0.15g. This would indicate that extending this easy screening
procedure to other areas in the East and Midwest in moderate earthquake zones would
be worthwhile.

Introduction

Free-standing bridge abutments on spread footings are susceptible to seismic

degradation of their horizontal and vertical capacity to sustain even moderate earthquake
loads. Moreover, free-standing designs make up the bulk of the inventory of bridge
abutments outside of areas in the West where the perceived risk is high.

For example, of the 40,000 bridge abutments in New York State almost all are free
standing and more than half are founded on spread footings. Abutments founded on
spread footings are not concentrated within one geographic locality but are distn'buted
evenly throughout the various regions of the State [NYSDOT (1991), Younkins
(1994)]. If the inventory of bridge abutments in New York is considered typical of the
Eastern United States and the Midwest, the seismic vulnerability of free standing bridge
1
Geotechnical Engineering ConsUltant, McMahon & Mann ConSUlting Engineers,
P.C., 2495 Main St., Suite 511, Buffalo, NY, 14214, (716) 834-8932.
2
Professor, Department of Civil Engineering, State University ofNew York @
Buffalo, 212Ketter Hall, Buffalo, NY 14260, (716) 645-2114.

77

@seismicisolation
@seismicisolation
 78 RETAINING STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

abutments founded on spread footings is a major cause for concern, even with the
moderate level of seismic risk associated with these regions.

Richards and Elms (1979) showed that a gravity wall designed for a static factor of
safety of 1.5 would slide in an earthquake a calculable amount whenever the lateral
acceleration exceeded O.llg. They therefore introduced the displacement based
approach for the seismic design of free standing, gravity wall type bridge abutments since
seismic movements· could not be totally eliminated at a reasonable cost even for
moderate earthquakes. Displacement based seismic analysis requires the determination
of a threshold level of acceleration beyond which equih'brium is lost (F.S = 1.0) and
relative displacement between the gravity wall and foundation soil will occul-. The
original work by Richards and Elms (1979) considered only horizontal equih'briurn and
therefore only the sliding mode of deformation. However, earthquake damage reports
and laboratory tests indicate that wall failure by rotation triggered by a loss of vertical
equih'brium is also significant.

These obseiVations are verified by recent analytic studies by Richards et a!.

(1990,1991,1993), Richards and Shi (1991) and Shi (1993) which show a dramatic
reduction in seismic bearing capacity is poSSL'ble. Thus, beyond a threshold acceleration
!eve~ a rotation or mixed mode of deformation can result. Seismic bearing capacity is
strongly dependent on the level of acceleration, the shear transfer between the wall
footing and foundation soil, and the shear strength of the foundation soil

This, in turn, leads to a general procedure for determining threshold acceleration levels
for free standing gravity wall bridge abutments for all modes of.fiillure. The procedure is
comprehensive in that the seismic rotation of retaining walls can be considered along
with the sliding mode so that sequential modes of movement can be seen to develop
when both horizontal and vertical equih'brium are violated.

Based on the work of Nadim and Whitman (1984), Siddharthan et al. (1992)
implemented coupled equations of motion to describe the relationships between wall
translation, rotation, and the forces and moments acting on the wall system. In this
paper a revised procedure for determination of permanent displacement of rigid
walls due to earthquake excitation will be presented. Coupled sliding and rotation
will be descn'bed, as before by Siddharthan et al. (1992), but the possibility of
seismic loss of bearing capacity will also be included in the analysis.

Although the ability to predict permanent seismic displacement with a sliding mode
of failure has already been verified in the laboratory (Lai(I979), Lai and Berrill
(1979), Aitken (1982), Steedman (1984), Elms and Woods (1986), Uwabe 11nd
Moriya (1988), Elms and Richards (1990), Whitman (1990)) there is a need to
extend this data base to include studies of coupled sliding and rotation. The tests
described herein are an improvement over previous studies (Aitken (1982), Musante
and Ortigosa (1984), Anderson et al (1987), and Whitman (1990)), in the sense that
the foundation soil beneath the abutment is included, and the model is not

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENTS SEISMIC ANALYSIS 79
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

constrained to a tilting mode of failure, but rather any possible mode of failure is
allowed.

In this paper first tlie analytic procedure for estimating threshold accelerations will
be reviewed, followed by the equations for estimating seismic induced permanent
deformation. The analysis has been verified based on observed performance of
laboratory models. Details of the bridge abutment models and the observed response
of models both with buried toes and forced to rotate about the top of the abutment
will be described. Finally, the results of a survey of bridge a_butments typical of those
constructed in New York State is presented.

Seismic Vulnerability Of Gravity Wall Bridge Abutments

A limit equilibrium approach is used to account for increased lateral earth pressure
(Mononobe (1929), Mononobe and Matsou (1929) and Okabe (1926)) and reduced
bearing capacity (Richards et al (1990) and Shi (1993)) due to seismic loading. The
application of the theory for analysis of seismic vuloerability of bridge abutments is
discussed in detail by Fishman et al (1995), Fishman and Richards (1995) and Richards
et al (1996). Brief details of the analytic procedure for determination of the most critical
threshold acceleration will be reviewed here.

The seismic vuloerability of gravity wall bridge abutments involves the determination of a
threshold acceleration beyond which permanent deformation of the gravity wall will
occur. A thorough seismic analysis must investigate the posSibility of both a sliding
mode of failure as well as a beariog capacity failure introducing rotation. The analysis
for the sliding failure mode is based on the theoretical and experimental work of
Richards and Elms (1979), and has been well documented in the AASHTO (1993) code
provisions and commentary. Seismic beariog capacity is a new development as applied
to gravity wall bridge abutments so details of the analysis follow.

Since seismic bearing capacity fuctors are dependant on ground acceleration,

determination of the threshold acceleration requires an iterative procedure. Refeoing to
Figure 1, it is assumed that there is no cohesion or depth of embedment and that kv= r
For walls free to move at the top F nn is zero or a known value and we can:

(1) Assume a trial value for~ and determine P AE from the

Mononab&-Okabe (M-0) analysis.

(2) Compute the vertical force resultant, P as:

(1)

@seismicisolation
@seismicisolation
 80 RETAINING STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

PAE

Ratalnad Soli Paramatal"8:

'Y • Yw • aoll unit weight

• • +.. = aoU lntemal frtction angle

F Foundat1on Soli Paramata~ :

p
yf • aoil untt ~ght
+, = aoU In tarnal tr1cUon angle

Figure 1. Gravity Retaining Wall Bridge Abutment Force Diagram.

(3) Compute the resultant of the shear traction to be transferred to the

foundation soil as

(4) Compute the filet or fusing the equation

f= S/~P (3)

(5) Sliding will occur if S =PtanSr and therefore

(4)

where Sr is the interfilce fiiction angle between the abutment footing and
the foundation soil

( 6) Given the fiiction angle of the foundation soil, 4>r, and the f filctor .from
step 4, find the seismic bearing capacity filctor from Figure 2.

(7) Compute the seismic bearing capacity PIE using the equation

(5)

(8) Bearing capacity failure will occur when P =PtE B and therefore:

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENTS SE ISMIC ANALYSIS 81

PmB
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

F.S.otc (6)
p

Iterate on kh to deteonine the threshold values given when F.S. = 1. That

is:

(9a) IfF.S.IIIC determined in step (8) is nearly equal to one, and F.s..... from
step ( 5) is greater than one, stop the ·iteration procedure since the
assumed value for ~ is the threshold value for bearing capacity failure,
i<J.b, which occurs first.

(9b) If F.s ..... deteonined in step (5) is nearly equal to one and F.S.JliC is
greater than one, stop the iteration procedure since the assumed value
for ~ is the threshold value for sliding failure, ~. In this case when
sliding occurs first there is still the potential for a bearing capacity failure
at a higher acceleration introducing a mixed mode. To estimate kJ.J, > k..
set, P and S at their constant values for sliding, compute NrE from
Equation (5) with PtE= P/ Band detemline kw.fromFigure 2.

(9c) If neither of the conditions in 9a or 9b is met, select higher trial value for
~and return to step (1).

II i .
t=2
I . :

.
II 8 I ft .5fi1
I J
~ I : ! : /
~
~
,, 6
I
I
I I
I
I
I
I

.'
..
•
I'

~
~ 4
I/ / I ,'
I
....0 I
I I
I
..
•

... /l' ~/
I :

~ I / .'
~:.~~-/
~
;g
~
C/) OL-~-~-~~-~---='
0 0.2 0.4 0.6
k,
Figure 2. Ratio of Nyu to Nys after Shi (1993).

Before going on to moclitY this analytic procedure for Fnn 0, it is important to *

reemphasize that, as in the M-0 analysis for lateral thrust, the bearing capacity analysis
involves only force resultants and equilibrium of forces. Just as we do not know the
position ofthe lateral thrust P,u:, we do not know the position of the force, P, driving the
foundation failure. Moment equilibrium is not considered since the analysis then

@seismicisolation
@seismicisolation
 82 RETAINING STRUCTURES ANALYSIS AND DESIGN

becomes highly indeterminate and nonlinear. The eccentricity of P and possible modes
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

of rotation including rocking and liftoff are not considered in the determination of
threshold accelerations nor are they likely for bridge abutments. Rather the analysis has
pwposely been kept simple at the same level of sophistication as the M-0 analysis \\hlch
is its major component. Eccentricities will be considered in the calculation of
deformations discussed later.

For abutments not free to move outward at the top due to the girder connection details
or other reasons, Fun will not be zero and the analysis procedure must involve the
moment equilibrium equation even iflines of action for PAE and P are assumed. For the
extreme case, the top can be considered pinned and the wall must rotate in the RT mode.

However, until the base moves, creating the active situation, it acts as a rigid wall. For
this case (Wood, 1975) the seismic lateral pressure increment is parabolic giving a thrust
PRE roughly twice the M-0 value and 0,. is close to zero.

Therefore, to modify the analytic procedure for walls restrained at the top (wltere Fun*
0) for determining threshold values:

a) In step 2 use PRE = 2PAE from step 1 and 0,. = 0. For a wall with
vertical interfilce

p Fov+Ww (7)

independent of kh .

b) Assume PRE = 2PAE acts 0.375H from the top and P acts at the
midpoint ofthe base, B/2.

c) Take moments about the top of the wall to determineS rather than usiug
horizontal equilibrium (Eqn. 2}. For a wall with a vertical iuterfilce with
its center of gravity at Zand X:
S = [2PAB(.375H}+kb WwZ+P(O.SB}-WwX] I H
(8)

d) If the value of Fun is desired it can now be computed fromEqn. 2.

Thus it is now possible with this analytic approach to evaluate the seismic vulnerability of
both existing gravity wall abutments and new designs. The procedure has purposely
been kept simple and many refinements are possible. Although we have assumed k. =()
this is not required.

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENTS SEISMIC ANALYSIS 83

Vertical acceleration can be included quite easily by using i<J,/(1-kv) instead of just kJ,
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

\Wen computing PAE and NyE/ NyS. W,. becomes (1-kv)W., in the equilibrium
equations. For the retaining wall problem where small incremental movements
accumulate, vertical accelerations are not as significant as they are for buildings. This is
because there iS little correlation between ~and kv. That is to say, because of asymmetric
resistance of walls (passive in one direction and active in the other), the pulses are as
likely to raise the threshold acceleration as to reduce it.

Although the backfill is usually granular the foundation soil may have some cohesive
strength. There may also be some depth to the foundation. Either, or both, may be
included by using Figures for N,EIN,. and N.v Ncs (Shi (1993)), and Eqn. 5 in Steps
6 and 7 of the procedure for determination of threshold accelerations.

Passive Restraint

In many instances the base of an abutment may be embedded to some depth within the
foundation soil Effects of foundation embedment include the development of passive
restraint against sliding, and a contnbution to bearing capacity from the surcharge. Both
of these effects may be incorporated into the newly developed analytic method.

Development of passive restraint under conditions of seismic loading has not received
the same level of attention as the active case. Inertia from horizontal accelerations,
driving the abutment to failure, tends to reduce the passive thrust that may develop
within the foundation soil. The M-0 analysis, and the seismic free field equations
developed by Richards et al (1990) descnbe the degradation of passive restraint with
increased horizontal acceleration. However, observations of the development of passive
restraint for seismic loading conditions are limited.

Richards and Elms (1987,1992) performed laboratory experiments to evaluate

passive soil resistance during seismic loading. They used a shaking table to study
the passive resistance behind walls which were forced to deform by: rotation about
the top (RT), rotation about the bottom (RB), and translation (T). The following is
a summary of several important conclusions which were reached:

• Within the passive region, compression of the soil is the first observed
mode of deformation. As horizontal accelerations increase, forcing the
wall to move further into the soil, a failure surface develops extending
from the base of the wall to the surface.
• The line of action of the dynamic passive thrust resultant moves
downward as horizontal accelerations increase, until the failure surface is
fully developed, at which point the line of action is located within the
bottom third of the wall.
• For dense cohesionless materials, the failure wedges move incrementally
when ground accelerations exceed the threshold acceleration. The active

@seismicisolation
@seismicisolation
 84 RETAINING STRUCTURES ANALYSIS AND DESIGN

wedge will settle and move in contact with the wall, while the passive
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

wedge will move ahead of the wall and be forced upward.

• Within the passive failure zone the shear strength of the soil is not a
constant, decreasing from the peak toward the residual strength during
deformation.

Equations of Motion

Nadim (1980), and Nadim and Whitman (1984) employed coupled equations of
motion to study the problem of seismically induced tilting of gravity retaining walls.
Based on the work ofNadim and Whitman (1984), Siddharthan et a! (1990, 1991,
and 1992), developed a method to predict the seismic performance of retaining
walls considering both rotation and translation deformation modes.
Deck Load

~
Pae ·

·xg(l)[

.. 1
Yg(l)
C.R.

N
Figure 3. Free Body Diagram of Bridge Abutment Model with Free Connection to
Bridge Deck.

Figure 3 is a free body diagram of a retaining wall subjected to seismic forces which
induce active earth pressures in the backfilL Inertial forces are applied according to
d'Alembert's principle. Much like the Richards and Elms (1979) approach to
translating walls Newton's fundamental laws of motion are applied to arrive at the
coupled eqautions of motion proposed by Siddharthan et al (1991). Equations 9 and
10 are the coupled equations of motion described by Siddharthan et al (1992).

(9)

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENTS SEISMIC ANALYSIS 85

[I- Y~t)]-..:~ +
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

l:rx)Ro,(") + [I• • [w;']e = [ ;}XJ.(t)ROn(") -

PAB(mH)cos(8)- PAEsin(o)[Rcos(ll) + a]-F0 v[a+Rcos(11)]- Mo

Pin Comection (10)

P.'AE

Figure 4. Free Body Diagram ofDridge Abutment Model with Fixed Connection to
Bridge Deck.

Siddharthan's equations of motion apply to a retaining wall, but not a pin-connected

bridge abutment shown in Figure 4. The resulting equation of motion is equation 11.

w 2] 8- = gXa(t)Rsin(ll)
[ll" + gR w- - WRcos(11) + PAllhcos(o) + Nb - SH

(11)

Values for the normal and shear forces at the abutment foundation-soil interface, N
and S, respectively, must be determined. The sliding threshold represents the
acceleration that the abutment can resist before sliding. Beyond the sliding threshold
acceleration the shear force, S, is:

S = P AEcos(ow) + k~uW =Ntan(or) (12)

where PAE is at the limit described by the threshold acceleration level for sliding, ki,' .

A newly developed theory describing seismic reduction of bearing capacity

(Richards et a!. (1990, 1991,1993), Richards and Shi (1991), and Shi (1993))
provides the limit to the normal force, N, at a given level of acceleration.

@seismicisolation
@seismicisolation
 86 RETAINING STRUCTURES ANALYSIS AND DESIGN

Model Studies
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Model Testing

Model bridge abutments were constructed within a seismic test chamber and subjected
to seismic loading via a shaking table. Complete details of the test chamber and model
construction have been descn'bed by Fishman et al (1995), Divito (1994) and Drotar
(1995). Bridge abutment models were unique in the sense that both the abutment
foundation and bridge deck were incolJlorated. A number of bridge abUtment models
were tested as descnbed by Fishman et al (1995), Fishman and Richards (1995) and
Richards et al (1996). Throughout the test program geometric, and physical model
parameters were varied including basal shear resistance, and details of the connection
between the bridge deck and the top of the bridge abutment. Comparison of observed
and predicted responses served to verify the newly developed analytic procedure for
predicting threshold accelerations. In this paper additional results from model testing will
be presented which demonstrate the effect of burying the toe of the abutment, and
deformations associated with a restrained connection to the bridge deck. The ability of
the seismic limit analysis and equations of motion to predict the model performance will
be demonstrated.

r :z.•.. ..,
lr"""'End .... 5tt••Oni·Rolrrlk4lfiG'1
Sorinho · -~~
Sfl'esn.n.fbwd Ortdg~Deck,2-W.. IO
<
o"'
:C 0.2lm :(
N
All ~~
f2 ~•
5
" ~ties 0nt . OUI • D. I4Gm
•M - -f- - M../1 - AI& Sttlu T"MJ · DIM • 0,1521'11
Set'"llwtt · Bne • O.lO)ft
l,r-Fir:lltii:EildW;al o.zm o.45m Al7
0AS - - [- - M• ~ -- n
Oll.l'WaSMKI
ASTWC-Iog .•, - -0.23m
1- - Al• I< AIJ
•AJ
RlgldEndWal - - -...

~~
o..,.. -- 1-- O.Oim
/ -- Rigid Fe~toummlllue

l!?ai:···
L-·- ----:.m '·""" - --1 I-_- _•.•
..J...- _s._ ___ ,_,., ______ J

Figure 5(a). Schematic of Seismic Test Chamber.

Figure 5(a) is a schematic showing a typical model with passive restraint. The weight of
the abutment was 1135 N (1.24 kN/m). The connection between the top of the abutment
and the bridge deck was free. Figure 5(b) shows the location of instrumentation.
Displacement transducers are designated as T1 through Tl4 and accelerometers are Al
throughA22.

In this study three models, referred to as Models lA, IIA, and IDA, were tested.
Variations between the models included deck load and depth of embedment as descnbed
in Table l Model ITA was configured with an increased deck load and decreased

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENTS SEISMIC ANALYSIS 87

embedment compared to Model lA. Model IliA was a model retaining wall since a
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

deck load was not applied.

TB T Tl 2 Ttl

~T5
111 Wt5T
H CAST
Tl
A9 AIO Al l ~AI2

"' A7

~T9
"'AI A2 "'

All.:~ Test Box Bose

Figure 5(b ). Instrumentation for Testing Abutments with Buried Toe.

TABLE l Models with Passive Restraint

MODEL DECKWAD(N) DEPTH OF

PASSIVE
EMBEDMENT
(em)
lA 356 7.62
ITA 592 3.81
lliA 0 7.62

The model response was evaluated with respect to dynamic sliding, bearing
capacity, and overturning threshold accelerations. Sliding threshold accelerations
were computed with and without passive restraint. The overturning mode of failme
exists when the line of action of the foundation soil pressme resultant is coincident
with the toe of the footing. Predicted and observed threshold accelerations are
shown in Table ll where,

= sliding threshold acceleration with no passive restraint

kt..t
k~u 1 =sliding threshold acceleration with passive restraint
kbb =bearing capacity threshold acceleration
khO.T. = threshold acceleration for overturning
N.O. =not observed.

Figmes 6, 7 and 8 present the history of deformation from pulse testing for Models·
IA, ITA, and IllA, respectively. Acceleration pulses were applied in increments of

@seismicisolation
@seismicisolation
 88 RETAINING STRUCTURES ANALYSIS AND DESIGN

0.05g to O.lg, through a range from 0.05g to 0.?g. At a given level of acceleration
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

pulses were repeated two or three times. The abscissa in Figures 6 and 8 show the
level of acceleration for a pulse that was repeated twice. In Figure 7 _pulses were
repeated three times at each level of acceleration.

TABLE ll Observed Response of Models with Passive Restraint

\ Failiiremode Mo~~IIA M;o~eiTIA · Model .

.
... :··
':: ···;~;-;,;_;.; ......, <·/> .· .. .< TIIA ·
.'·1 ~-- -:···:····· ~-.
'....
~·1

(nredicted/observed) 0.30/0.30 0.38/0.35 0.20/N.O

~·1

(nredicted/observed) 0.38/0.35 0.40/0.50 0.30/0.40

~·
(nredicted/observed) 0.60/N.O. · 0.40/N.O. 0.65/0.65
~O .T.

(nredicted/obscrved) 0.40/0.40 0.45/N.O. 0.30/0.45

~
0
"' a
N
ci ~ gci ~ ~ ~ ~
ci 0 ci 0 0
Base Ac:celeratlo n

Figure 6(a). Model IA Horizontal Displacemen (T3)

The accumulation of incremental permanent deformation · of the model walls

compared well with the predicted threshold accelerations· for sliding. Bearing
capacity failures were barely perceptible and occurred, as predicted, at relatively
high levels of acceleration. In what follows the initially oJ:>served sliding mode of
deformation will be discussed.

The onset of sliding occurred at acceleration levels corresponding to thresholds for

sliding which do not consider passive restraint. Displacement necessary to generate
the full active thrust is not compatible with that necessary to fully develop the
passive restraint. Therefore, at the onset of sliding, full passive restraint has not

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENTS SEISMIC ANALYSIS 89

developed. However, initial displacements are small and do not recur when pulses
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

are repeated at the same low levels of acceleration. 'Ibis nonrecurreuce is because
passive resistance is mobilized during deformation tending to increase the seismic
resistance. Once the restraint has fully mobilized deformations pulses at the same
5
'E4.s
E 4
:::3.5
..c 3
~ 2.5
u 2
Ill
~1.5
15 1
0.5
0
lll"'
0>
~ "' "' 8"<; ...."' ~ ~"'
0>
ci
"!
0
I()
N
ci
?l
ci ci 0
I()

ci 0
h1ci 0
~
0
Base Acceleration

Figure 6(b). Model IA Vertical Displacement (T8)

2.5

2
'E
g 1.5
~
..
Ill
:X:
1

0.5

0
"' ,g>
"' "' ?l"'
0>
~"' ~ ~
"" &l"' :R"' g"'
N I()
ci N
ci
ci ci 0 0 0 ci ci "'ci
Base Acceleration

Figure 6(c). Heave in Front of Model lA (TIJ)

level of acceleration reach a steady state. The level of acceleration at which this
happens compares well with the prediction of threshold acceleration including the
fully mobilized passive restraint.

Heaving of the passive wedge was measured at several locations on the soil surface
in front of the abutment. Figures 6(c) and 8(c) show the history of heaving during
pulse testing for Models lA and lliA, respectively. Accelerations observed at the
onset of heave correlate well with the apparent development of full passive restraint
based on observations of incremental displacements.

@seismicisolation
@seismicisolation
 90 RETAINING STRUCTURES ANALYSIS AND DESIGN

Models progressed from a sliding mode of failure, to tilting, followed by seismic loss
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

ofbearing capacity and incremental settlement. Sliding with a very small uplift of the
heel of the footing occurs at acceleration levels close to the threshold for
overturning as shown in figure 6(b) and 8(b ). As a result pressures concentrate at
the toe of the abutment footing until a bearing capacity failure is initiated. Loss of
bearing capacity is followed by settlement of the footing and movement of the center

25
e 20 . mPulse 1 · ~
.5.
"
~u
~
15
10
I
•Pulse 2
CPulso 3

c.
•
i5
5

0
0 .30g 0 .35g 0 .40g 0.5Dg 0.55g
Ba se Accellira tlon
Figure 7. ModeliiA Horizontal Displacement
(T 3)
of rotation toward the centerlioe of the footing. A bearing capacity failure is evident
from Figure 8(b) where it is seen that the toe of the footing is settling indicated by
displacement transducer T2.

~
0
~
0
~
0
Baae Acceleration

Figure 8(a) . Model IliA Horizontal

Displacement (T3)

Rotation Mode

A detailed analysis of seismic induced tilting of model bridge abutments was

performed by Kutschke (1995). The bridge abutment model was constrained at the
top by the bridge deck. The restraint forced the abutment to fail by rotation with
respect to the top of the abutment. Parameters and corresponding predicted
threshold accelerations for the model are presented in Table ill. Displacement
predictions are presented in Figure 9. Each point represents the accumulated
deformation subseq~ent to ail acceleration pulse. Acceleration pulses were repeated

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENTS SEISMIC ANALYSIS 91

5
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

e
.5.
4 BIPulse
•Pulse
1
1
TB
T2
3
'Cw lliiP u lse 2 TB
2
llu laPulse 2 T2

ii5 0
-1
0 d d d d
Baae Acceler•tlon

Figure 8(b ). Model IliA Vertical Displacement

2.5
2
e
.5. 1 .5
!t
~
~

"' 0 .5
0

~ ~ ~
0 ~ ~ ~
Baae Accelentlon

Figure 8(c). Heave in Front ofModel IliA

(T 13)
twice at 0.5g and three times at 0.55g and 0.6g. During the 0.65g accelerations the
wall failed in an observed deep-seated shear failure. Results obtained with Equation
(11} for rotation with respect to the top of the abutment rendered results in excellent
~greement with the observed response.

Table ill. RT Model Parameters and Estimated Threshold Accelerations.

Backfill Unit Weight, y 16.65 kNfm3

Backfill Friction Angle, cp 36°

Soil-Wall Friction Angle, Ow*Of ow= 22°, oc = 30°

Deck Load on Wall 0.39kN/m

Wall Weight L32kN/m

Bearing Capacity Threshold 0.46g

@seismicisolation
@seismicisolation
 92 RETAINING STRUCTURES ANALYSIS AND DESIGN

'E
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

10
.s 8
7
...d
"E ....d
Q) 6
8
E
4
.-:d ~
.!!! ~
g.
i:5 · 0
2

0.5 0.55
·------ 0.6
Kh

1--.octual --m-Predlcted I
Figure 9. Comparison of Predicted and Observed Displacements for RT Mode.

Survey OfExisting Bridge Abutments

As part of its bridge inspection program, New York State maintains a comprehensive
data base descnlling its bridge inventory (NYSDOT, 1991). The data base may be
obtained in compressed format on diskettes and is accesSJ"ble through a number of
commercially available data management programs. The data base contains over 170
fields of information for each bridge including the age of the structure, number of spans,
t;pe of bridge abutment, height ofbridge abutment, foundation type, and details of the
bearing between the bridge deck and the top of the bridge abutment. In the first part of
the survey, this data base was queried to study the demographics of the bridge abutment
population in the state and to identifY those for which the methodology under
investigation may apply.

Based on information from the initial survey, fifty bridge abutments were identified for
detailed analysis in part 2 of the survey. Construction drawings and subsurface soils data
were obtained for the selected bridge abutments. Using this information, wall geometry,
bridge deck loads, and shear strength parameters of the wall backfill and foundation soils
were determined. Each bridge abutment was analyzed to determine both static factors of
safety as well as the seismic resistance. The following is a summary of results from the
survey ofNew York State bridge abutments.

Inventory ofBridge Abutments in New York State

Of the 39,346 bridge abutments listed in the New York State inventory, 15,716 are
noted as being founded on spread footings placed on cut or fill material. Therefore, the
analysis addressed by this research may be applicable to 40% of all bridge abutments in
New York State. However, since the state bridge inventory does not ade(jiKltely reflect
abutments which could have piles under the footings this may be an overestimate.

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENTS SEISMIC ANALYSIS 93

There is no clear distinction between the construction practices of the New York State
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Department of Transportation (NYSDOT) and those of other agencies within the state.
NYSDOT is the owner of 43% of the bridge abutments on spread footings. Historically,
the practice of using spread footings has not changed when comparing those constructed
prior to 1960 to those after 1960. However, since 1960, there is a clear trend favoring
cantilever wall designs over gravity wall designs. Forty-five percent of all the bridge
abutments on spread footings are the gravity type design, and 39% are the cantilever
type. More than half (63%) of all the bridge abutments founded on spread footings are
for single span bpdges. Ofparticular interest is the fact that nearly half(44%) of the
bridge abutments on spread footings are taller than 6 meters. The significance of this
statement will become apparent from the results of the analysis.

The demographics of bridge abutments constructed in western New York were

compared to those of the state population and deviations were found to be minor.
Therefore, the inventory of bridge abutments in western New York was considered
representative of the state.

Analysis ofTypicalBridge Abutments

Fifty bridge abutments located in western New York were selected for detailed an~lysis.
As a point of reference, safety factors relative to a sliding failure, overturning and bearing
capacity were computed for static loading. The computed safety factors for sliding and
overturning were all above 1.5 and 2.0, respectively, but were much higher for shorter
walls. 1his was due to the filet that deck loads are not correlated to wall height. For
shorter walls, the deck load is large compared to the weight of the wall itself; thus, it has
a much larger impact on the computed safety factors when compared to taller, heavier
walls.

Seismic resistance of selected bridge abutments was computed using the newly
developed analytic procedure. Relatively speaking, for shorter walls the seismic
resistance is strongly affected by the colUlection detail between the top of the abutment
and the bridge deck If the bridge deck is fixed to the top of the abutment, an inertial
reaction is transferred to the abutment/soil system tending to drive the system to failure
during a seismic event. If a free colUlection exists between the bridge deck and the top of
the wall, this inertia is not transferred to the abutment and seismic resistance is higher.

Summary ofResults From Survey

Figure 10 presents results of the seismic vulnerability study applied to western New
York bridge abutments. Seismic vulnerability is compared to wall height. For shorter
walls a wide range of computed threshold accelerations is evident due to the influence of
the connection detail to the bridge deck. For taller, heavier, walls the contn'bution of the
bridge deck to the loading is less significant and correspondingly the range of computed
threshold accelerations narrows.

@seismicisolation
@seismicisolation
 94 RETAINJNG STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

z 0.35
0
I=
ii!w 0.~ ••
...r
• •• •
... . ..
w 0.25
(,)
(,)

"'...r 0.20
•l ••• .~
:!'7
z!!!.
0
0.15 ••• • ..,.:. •
N
a:0 • •
=
0.10
•
...r
"'I=
(,)
o.re •
a:
(,)
0.00
2.00 3 ,00 4.00 5.00 6.00 7.00 6.00 9.00
WALL HEIGHT (m)

Figure 10. Seismic Resistance ofTwical NYS Bridge Abutments.

All bridge abutments over 6 meters high had a computed threshold acceleration less than
0.2g with many less than 0.15g. These levels of acceleration are considered possible even
with the modest level of seismic hazard in parts ofNew York Considering that there are
7,000 bridge abutments over 6 meters high and founded on spread footings, a significant
portion of the bridge abutment inventory is vulnerable to seismic induced, permanent
deformation.

In addition, two of the cases studied, representing abutments over 7.5 meters high, had
computed seismic resistance ofless than O.lg with one as low as 0.06g. There are nearly
1,400 bridge abutments in New York state which could fit into this category. The
likelihood of these structures suffering excessive permanent seismic induced deformation
in the future is very high. A more thorough investigation of bridge abutments on spread
footings having heights in excess of 7.5 meters is warranted, and strongly
recommended. Remedial measures to protect existing abutments with high seismic risk
such as providing knock-off walls or drilled tiebacks can be straightforward and not
overly expensive.

Conclusions

Algorithms for predicting critical threshold accelerations and seismic induced

permanent deformations were applied to a model bridge abutment. The model was
constructed within a seismic testing chamber, and seismic loading was applied via a
shaking table. If the base of an abutment is embedded, passive restraint may develop
within the foundation soil Some deformation is required for the passive restraint to be
fully mobilized. Small lateral displacements may occur at sliding threshold accelerations
that do not reflect the development of any passive restraint. As deformations continue,
passive restraint is mobilized, and seismic resistance is increased. When full passive
restraint is mobilized a steady state seismic resistance prevails. Prediction of the sliding

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENTS SEISMIC ANALYSIS 95

threshold appears to be quite accurate and very good comparison between observed
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

and predicted thresholds was realized. Furthermore, the surcharge above the base of
the embedded foundation provides a significant increase in the seismic bearing capacity.
Predicted bearing capacity thresholds levels were relatively high and confirmed by
results from the model tests.

Whitman (1992) and Siddharthan, et al. (1990, 1991 , and 1992) proposed the use of
coupled equations of motion to predict seismic induced permanent deformation of
retaining walls. These · equations can be used to describe mixed modes of
deformation inciuding sliding and/or tilting. Equations of motion are cast in terms of
relative acceleration between the retaining wall and foundation soil. Relative
displacement and rotations are computed by double integration of the equations of
motion witlt respect to time, similar to Newmark (1965) and Richards and Elms
(1979). The coupled equations of motion as they appear in the literature were
modified to consider seismic reduction of bearing capacity, and rotation with respect
to the top of the abutment.

A survey of bridge abutments in New York State indicates that the analysis may be
applicable to a significant portion of the inventory. The methodology will be useful
from the standpoint of seismic risk assessment as applied to existing construction
and levels of seismic ltazard typical of the eastern United States.

Acknowledgement

This paper was prepared on the basis of research sponsored by the Federal Highway
Administration (FHWA) and administered through the National Center for
Earthquake Engineering Research (NCEER). The research reported herein is based
·on work conducted under tasks 106-E-4.5 ofFHWA Project DTFH61-92-00106,
Seismic Vulnerability of Existing Highway Construction, and 112-D-3.4 of FHWA
Project DTH61-92-00 112, Seismic Vulnerability ofNew Highway Construction.

Disclaimer
Any opinions, findings, conclusions, or recommendations expressed in tins paper are
strictly those of the authors and do not necessarily represent tlte views of the Federal
Highway Administration, the National Center for Earthquake Engineering Research,
or other participants in or sponsors of this work.

List Of Mathematical Symbols

f = shear transfer coefficient in seismic bearing capacity eq nation

Fnv = vertical force transferred to bridge abutment from deck
Fn11 = horizontal force transferred to bridge abutment from deck = khWn
kh = horizontal acceleration coefficient
k. = vertical acceleration coefficient
kbb = bearing capacity tlueshold acceleration

@seismicisolation
@seismicisolation
 96 RETAINING STRUCTURES ANALYSIS AND DESIGN

k~u = sliding threshold acceleration

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

M 0 = soil moment resistance

N or P = vertical force resultant acting on the foundation
N1 .~ NqE, NeE = seismic bearing capacity factors
N 1s Nqs, N eE= static bearing capacity factors
PAE = active thrust from M-0 analysis
PKJo; ==thrust on nonyielding wall during seismic loading
PtE = seismic bearing capacity
S = resultant of shear traction acting on the foWJdation
Ww or W = weight of the abutment
Wo . weight ofbridge deck
X, (t) = horizontal ground acceleration,
Y, (t) = vertical groWld acceleration,
x and S = relative horizontal and angular acceleration of the wall through and
about the center of gravity (CG), respectively,
e
x and = relative wall displacement and rotation through or about the center of
gravity {CG), respectively
cj> = internal angle of friction for backfill
cl>r = internal angle offiiction for foWldation soil
Or = wall base/fmmdation soil interface fiiction angle
Ow or 8 = abutment/backfill interface fiiction angle
Yr = 1mit weight of foundation soil

And the geometric variables are:

a = horizontal distance between the center of gravity and the wall backface,
n or b = width of abutment footing
H = abutment height
leg = mass moment of inertia about the CG
mH = height from the abutment base to the line of action ofPAE,
h = H(l-m)
R = radial distance from the center of rotation (CR) to tbe (CG),
Z and X = location of the abutment center of gravity
13 = angle oftbe abutment backface
lJ = angle that R makes with the horizontal.

List OfReferences

1. Anderson, G.A. , Whitman, R V., and ·Germaine, J.T. (1991). "Seismic Response
ofRigid Tilting Walls", Proc·eedings Centrifuge 1991, Balkema, pp. 417- 424.
2. Aitkeri, G.H. (1982). "Seismic Response of Retaining Walls", Research Report
82-5, Department of Civil Engineering, University of Canterbury, New Zealand.

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENTS SEISMIC ANALYSIS 97

3. American Association of State Highway and Transportation Officials (1993).

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Standard Specifications for Highway Bridges, Division 1-A, Seismic Design,

Washington, DC.
4. Divito, R. (1994). "An Investigation of the Seismic Reduction of Bearing
Capacity for Gravity Retaining Wall Bridge Abutments", Master Thesis, State
University ofNew York at Buffalo, Buffalo, New York.
5. Drotar, M.L. (1995). "Effects of Passive Restraint on the Seismic Bearing
Capacity ofGtavity Wall Bridge Abutments", Master Thesis, State University of
New York at Buffalo, Buffalo, New Yorlc.
6. Elms, D. and Richards, R. (1990). "Seismic Design of Retaining Walls",
Proceedings of the ASCE Specialty Conference On Design and Performance of
Earth Retaining Structures, ASCE, Geotechnical Special Publication No. 25,
pp. 854-871.
7. Elms, D.G. and Wood, J.H. (1986). "Seismic Design of Abutments and
Retaining Structures", Proceedings of the Second Joint United States-New
Zealand Workshop on Seismic Resistance of Highway Bridges, ATC 12-1, San
Diego, California.
8. Fislunan, KL., Mander, J.B., and Richards, R. (1995). "Laboratory Study of
Seismic Free Field Response of Sand", Soil dynamics and Earthquak£
Engineering, Elsevier, 14, pp. 33-43.
9. Fishman, KL., Richards, R., and Divito, R. (1995). "Critical Acceleration Levels
for Free Standing Bridge Abutments" Proceedings of the Third International
Conference On Recent Advances in Geotechnical Earthquake Engineering and
Soil Dynamics, Edited by S. Prakash, Univ. of Missouri, Rollo, Vol. 1, pp .163-
169.
W. Fislunan, K.L. and Richards, R. (1995). "Seismic Vulnerability of Existing
Bridge Abutments", NCEER Bulletin, NCEER, 9(3), pp.8-15 .
11. Kutschke, W. (1995). ''Bridge Abutment Rotations and Displacements Due to
Seismic Loads", Master Thesis, State University of New York at Buffalo,
Buffalo, New York.
12. Lai, C. S. (1979). "Behavior of Retaining Walls Under Seismic Loading", M E.
Report, Department of Civil Engineering, University of Canterbury, New
Zealand.
13. Lai, C. S., and Berrill, J.B. (1979). "Shaking Table Tests on a Model Retaining
Wall", Bulletin of the New Zealand National Society for Earthquak£
Engineering, Vol. 12, No.2, pp. 122-126.
l4. Mononobe, N. (1929). "Earthquake-Proof Construction of Masonry Dams",
Proceedings World Engineering Conference, Vol 9, pp. 275 .
15. Mononobe, N. and Matson, H.(I929). "On the Determination of Earth Pressures
During Earthquakes", Proceedings World Engineering Congress, 9, pp. 179-
187.
16. Musante, H. and Ortigosa, P. (1984). "Earthquake Induced Responses of Model
Retaining Walls", Proceedings of the Eighth World Conference on Earthquak£
Engineering, pp. 469-476.

@seismicisolation
@seismicisolation
 98 RETAINING STRUCTURES ANALYSIS AND DESIGN

17. Matsuo, H . (1941). "Experimental Study on the Distribution of Earth Pressure

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Acting on a Vertical Wall During Earth.quakes", Journal of the Japanese Society

of Civil Engineers, 27(2).
18. Nadim, F. (1980). "Tilting and Sliding of Gravity Retaining Walls", MS. Thesis,
Department of Civil Engineering, Massachusetts Institute of Technology,
Cambridge, Massachusetts.
19. Nadim, F. and Whitman, R. V. (1984). "Coupled Sliding and Tilting of Gravity
Retaining Walls D11ring Earthq11akes", Proceedings of the Eighth World
Conference on Earthquake Engineering, pp. 477-484.
20. Newmark, N . M . (1965). "Effects ofEarth.quakes on Dams and Embankments",
Geotechnique, 14(2), pp. 139-160.
21. NYSDOT, (1991). Bridge Inventory Manual, Bridge Inventory and Inspection
System, Albany, NY.
22. Okabe, S. (1926). "General Theory of Earth Pressllte", Journal of Japanese
Society of Civil Engineers, Tokyo, Japan, 12(1).
23. Richards, R. and Elms, D (1979). "Seisnlic Behavior of Gravity Retaining
Walls", Journal of the Geotechnical Engineering Division, ASCE, 105(4).
24. Richard, R. an.d Elms, D .G. (1987). "Seismic Behavior of Tied Back Walls",
Report 87-8, Dept. of Civil Engineering, Univ. of Canterbury, Christchurch,
N.Z.
25 . Richards, R., Elms, D. G., and Budhu, M. (1990). "Dynamic Fluidization of
Soils", Journal of Geotechnical Engineering, ASCE, 116(5).
26. Richards, R. and Shi, X. (1991). "Seismic Bearing Capacity of Shallow
FoWldations", Pacific Conference on Earthquake Engineering, November 20-
23, New Zealand.
27. Richards, R., B11dhu, M .,and Elms, D. G. , (1991). "Seismic Flu.idization and
FoWldation Behavior", Proceedings of the Second lniernational Conference On ·
Recent Advances in Geotechnical Earthquake Engineering and So;! Dynamics,
Edited by S. Prakash, Univ. ofMissouri, Rollo.
28. Richards, R. , Elms, D. G., and Budhu, M. (1993). "Seismic Bearing Capacity and
Settlements of Folllldations", Journal of Geotechnical Engineering, ASCE,
119(4).
29. Richards, R. and Elms, D .G. (1992). "Seismic Passive Resistance of Tied Back
Walls", Journal of Geotechnical Engineering, ASCE, 118(7), pp.996-1011.
30. Richards, R., Fishman, KL., and Divito, R.C. (1~96) . ' Threshold Accelerations
for Rotation or Sliding of Bridge Abutments" Journal of Geotechnical
Engineering, ASCE, scheduled to appear Sept. 1996.
31. Slli, X. (1993). "Plastic Analysis for Seismic Stress Fields", Ph.D. Dissertation,
State University ofNew York at Buffalo, New York.
32. Siddharthan, R. , Ara, S., and Anderson, J. {i990). "Seismic Displacement of
Rigid Retaiu.ing Walls", Proceedings of Fourth U.S. National Conference on
Earthquake Engineering, pp. 673-680.
33. Siddharthan, R., Gowda, P., and Norris, G. (1991). "Displacement Based
Design of Retaining Walls", Proceedings: Second International Conference on

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENTS SEISMIC ANALYSIS 99

Recent Advances in Geotechnical Earthquake Engineering and Soil Dynamics",

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

pp. 657-661.
34. Siddharthan, R , Ara, S., and Norris, G. (1992). "Simple Rigid Plastic Model
For Seismic Tilting of Rigid Walls", Journal of Structural Engineering, ASCE,
pp. 469-487.
35. Steedman, RS. (1984). "Modeling the Behavior of retaining Walls in
Earthquakes", Ph.D. Thesis, Engineering Department, Cambridge University,
Cambridge, England.
36. Uw'abe, T. and Moriya, M . (1988). "Shaking Table Tests of Sliding Gravity-
Type Retaining Walls During Earthquake", Proceedings of the Ninth World
Conference on Earthquake Engineering, Vol. Ill, Tokyo, Kyoto, Japan, pp.
685-690.
37. Whitman, R V. (1990). "Seismic Design and Behavior of Gravity Retaining
Walls", Design and Performance of Earth Retaining Structures, Proceedings of
a Conference, ASCE, pp. 817-841.
38. Whitman, R V. (1992). "Predicting Earthquake-Induced Tilt of Gravity
Retaining Walls", Retaining Stn1ctures, Proceedings of a Conference,
Institution of Civil Engineers, pp. 750-758.
39. Wood, J.M. (1975). "Earthquake Induced Pressures on a Rigid Wall Structure",
Bulletin of New Zealand Society for Earthquake Engineering, 8(3), pp. 175-
186.
40. Younkins, J. (1994). "Seismic Vulnerability of Bridge Abutments in New York
State", Master of Engineering Project, submitted to State University of New
York at Buffalo, Buffalo, New York.

@seismicisolation
@seismicisolation
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

EARTHQUAKE-INDUCED GROUND
SETTLEMENTS OF BRIDGE ABUTMENT FILLS

Raj v. Siddharthan, 1 MASCE, and

Mahmoud El-Gamal, 2 Student Member ASCE

Abstract

Simplified methods of analysis are herein proposed for

estimating earthquake-induced abutment fill settlements.
Primary factors such as excitation history, amplification
(or deamplification) within the fill, and nonlinear soil
properties and slope geometry which control the settlement
have been accounted for in the approach. Settlement caused
by the permanent slope rotation has been obtained from a
procedure similar to Newmark's sliding block model.
Additional settlement caused by grainslip-induced deforma-
tion is also incorporated. Readily usable simple relation-
ships have been developed based on the extensive seismic
excitation database generated from the 1994 Northridge
earthquake in order to evaluate settlements of typical
abutment fill heights of 4-10 m subjected to any earth-
quake excitation.

Introduction

One of the common damage features n ~ ar bridge

structures in earthquakes is the settlement in the bridge
abutment (approach) fill. A review of past earthquake
damage reports reveals that this phenomenon is widespread
even in moderate earthquakes (M - 6.0). For example, as
many as 80% of the bridge abutment fills near the
epicentral area in the 1994 Northridge earthquake (M =
6. 7) suffered measurable settlement. In many cases,

1. Professor of Civil Engineering, University of Nevada ,

Reno, NV 89557

2. Graduate student, Department of Civil Engineering,

University of Nevada, Reno, NV 89557

100

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENT FILLS SETTLEMENTS 101

clearly visible drops in excess of 20 to 30 em occurred

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

near the intersections of the deck and the approach . A

loss of access to the bridge structures can occur as a
direct consequence of excessive settlement in the abutment
fill. This means that abutment fill performance plays a
critical role and can affect the function of a bridge.
currently, no clear guidelines or methods exist to help
the designer predict which abutment fills will settle or
the severity of that deformation.
Estimating abutment fill settlement is not simple.
Many deformation mechanisms contribute to the settlement
in the abutment fill. These are: 1) settlement resulting
from slope movements or "failure" during shaking, 2)
settlement caused by grainslip induced by the cyclic
nature of the excitation, and 3 ) settlement in backfill
soil near the abutment due to permanent abutment movement.
All of these components of the settlement have to be
accounted for in a realistic settlement estimation
procedure. Many factors play a role in contributing to
each of the components listed above. A partial list of
these important factors includes the level of excitation,
amplification (or deamplification) of ground motion in the
abutment fill, variation of acceleration within the
approach fill, nonlinear soil properties, and height and
geometry of the approach fill. There are simplified
methods that are widely used in practice, e.g., Newmark's
and Tokimatsu-Seed procedures, to "estimate" the first two
components of the abutment fill settlement (Newmark, 1965;
Tokimatsu and Seed, 1987). However, as will be pointed
out later, many limitations exist to these approaches.

Existing finite element based methods can address

many of the factors listed in the preceding paragraph.
Some such methods use plasticity models and can,
therefore, represent the "true nonlinear" behavior of soil
(Prevost et al., 1985; Ueng et al., 1994 ) . However,
these sophisticated methods require a broad spectrum of
high quality data which, combined with their intensive
computing requirements, makes their general use
prohibitive in preliminary design estimates. ·
The major objective of this study is to present a
comprehensive approach to estimate approach fill
settlement. The study combines many of the well-accepted
and routinely us e d procedures and concepts. Effort was
expended to make certain that the procedure developed is
realistic, i.e., accounts for all the important factors
and in a form readily usable by the engineer.

@seismicisolation
@seismicisolation
 102 RETAINING STRUCTURES ANALYSIS AND DESIGN

Deck-Abutm e nt
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Gap
Approa ch
Deck
Fill

Approach Sla b
(Optional)
I
Bridge Abutment I

Piles
(Optional)
II
II
II
II
II
II
II
II
II
I

lA
+
Excitation Specified at
Foundation Level

Fig . 1: Seat Type Br i dge Abutmen t

Problem Statement and Past Studies

Figure 1 shows a seat-type bridge abutment and an

abutment fill. The bridge abutment may be on piles,
provided with an approach slab, either a seat type or a
monolithic type. The objective is to estimate the
settlement in the approach fill adjacent to the abutment.
When an approach slab is provided, the settlement at the
end of the approach slab is of interest. In seismic
response studies, it is customary to specify the base
excitation at a location where a "rigid or very stiff"
layer exists because the location of the "rigid" layer can
affect the ground response. The study undertaken here
relies on the excitation level being known at the base of
the abutment fill. Further, it is also assumed that the
water table does not lie within the region of interest
and, therefore, the effects of soil softening or lique-
faction are not addressed.

Fiqure 2 i l lus trates three de f ormation mechanisms

that contribute to the settlement of approach fill near a
bridge abutment. The first settlement component (p 1 ) is
from the deformation mechanism shown in Fig. 2(a). It
results from slope deformation that occurs in the approach
fill due to transverse shaking. Many researchers
(Franklin and Chang, 1977; Makdisi and Seed, 1978;
Ambraseys and Menu, 1988; Yegian et al., . 1991; Jibson,
1993; Ambraseys and Srbulov, 1994;. Ling and Leschinsky,
1995) have studied · this aspect of permanent slope
movement. These researchers, except Ling and Leschinsky,
used the "sliding block" on a plane model of Newmark
(1965) to study the permanent slope movement. On the
other hand, Ling and Leschinsky used a "rotating failure

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENT FILLS SETTLEMENTS 103
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Origlnnl Embankment

Sec llon A- A
(a) Selllemenl Due lo Sl ope )r.lovemenl

(b) Grains lip Induced S~lllemenl

Origlnnl Emba nkm ent

(c) Sc lllement from Abutment Movement

Fig . 2 : Components of Settlement in Abutment Fill

block". All these studies assumed that the slope movement
occurs after the minimum acceleration (referred to as
yield or critical acceleration, a y) required to bring
about incipie nt failure of the slope has been exceeded.
Many of these studies have numerous limitations.
First, they neglect the amplification (or deamplification)
of soil motions within the slope and nonlinear soil
behavior. The study by Makdisi and Seed (1978) shows the
importance of these factors. Second, though a circular
mode of failure is often used to arrive at the critical
acceleration, the computed movement is considered to be
the horizontal deformation of the slope in these studies.
An exception to this is the recent study of Ling and
Leschinsky (1995). The researchers consider the computed
deformation as horizontal because the deformation
computation is based on the horizontal acceleration
history. Since the objective of this study is to estimate
surface settlement, these approaches cannot be directly
used. The rotational failure mode gives rise to surface
settlement as shown in Fig. 2(a); it will be seen later
that surface settlement may be estimated using a "rotating
block" model. Third, these studies neglect the influence
of vertical acceleration. Since vertical acceleration
plays a major role in varying the normal stress on the
failure surface, it can substantially influence the
deformation calculations. Note that Richards and Elms
(1979) showed its importance relative to the permanent

@seismicisolation
@seismicisolation
 104 RETAINING STRUCTURES ANALYSIS AND DESIGN

deformation of bridge abutments. No study has

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

investigated the influence of vertical acceleration.

The recent study by Ling and Leschinsky (1 995 ) used

the rotating block model for the first time. It utilized
sinusoidal motions to illustrate the importance of the
frequency and amplitude of the excitation . Therefore,
their study cannot be applied directly in field situations
involving "earthquake-like" motions. Furthermore, as will
be seen later, their approach did not include a few terms
that are included in the proposed approach . . These terms
become important when slope rotation is high.

The second mechanism shown in Fig. 2(b) occurs due to

grainslip. This settlement component (p 2 ) commonly
develops in cohesionless soils because of the ability of
the grains to slip relative to each other . under cyclic
load. The amount of settlement depends on the excitation
level, thickness and relative density of the deposit, and
number of cycles of loading. Practicing engineers have
widely used a well-established procedure proposed by
Tokimatsu and Seed (1987) which accounts for all of the
above factors. Both longitudinal and transverse
excitations can contribute to grainslip. This method has
also been verified using many field observations.

The third settlement component ·(p 3 ) occurs from the

deformation mechanism shown in Fig. 2 (c) . This is a
result of the abutment movement away from the approach
fill during the excitation. This movement occurs due to
dynamic earth pressure induced on the abutment. This
deformation mechanism results from longitudinal shaking
and can occur mainly in seat type abutments, in which case
the abutment can undergo permanent deformation relative to
the bridge deck as shown in Fig. 2 (c). The manifestation
of surface settlement resulting from this deformation
mechanism will be limited to the abutment fill adjacent to
the abutment. When an approach slab is provided, the
influence of this mechanism is minimal. In addition, post-
earthquake field investigations also suggest that, with
monolithic abutments and abutments founded on piles, this
component of the settlement is small. This paper
addresses only the first two components of the
deformation.

Proposed Procedure

The foregoing description of deformation mechanisms

(Fig. 2) clearly shows · that an estimate of the total
settlement of the approach fill may be evaluated as
( 1)

@seismicisolation
@seismicisolation
 BR1DGE ABUTMENT Fll..LS SETTLEMENTS 105

Guidelines to estimate these components are presented

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

below. For a more complete treatment of the procedures,

see El-qamal (1996).

Settlement from Slope Deformation ( p 1l

Newmark (1965) made the earliest attempt to develop
a Jrocedure for evaluating permanent slope deformation due
t c earthquake shaking. The basis of Newmark's procedure

---

~
--
S-=::::/
c-
/

Fig . 3: Slope Defor mation Model (a fter Ambraseys

and Menu, 1988)

is that movements would begin to develop if the inertia

forces on a sliding mass were large enough to overcome the
yielding resistance. By using the steps involve~ in eval-
uating the downslope movement of a rigid sliding block on
an inclined plane, Newmark computed the permanent deforma-
tion of the sliding mass along a potential sliding surface
by integrating the "excess acceleration" (Newmark, 1965).
Makdisi and Seed (1978, 1979) proposed refinements to
the Newmark method in which they attempted to account for
nonlinear soil behavior and the variation of acceleration
within the embankment. For many circular sliding sur-
faces, they computed representative average acceleration
histc>ri~s within the sliding masses and subsequently used
llewmark's procedure to obtain conservative estimates ·of

@seismicisolation
@seismicisolation
 106 RETAINING STRUCTURES ANALYSIS AND DESIGN

permanent deformation. They developed designed charts

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

based on excitations from earthquakes of M = 6 ~ - 8~ and

assumed that the embankment was comprised of medium dense
sand with 2:1 side slopes. Many steps proposed by Makdisi
and Seed are used in the proposed method described below.

Figure 3 shows typical deformation patterns produced

by three successive pulses that cause yielding in a slope
(Amhraseys and Menu, 1988). The sliding mass on the left
side of the embankment slides downslope along the failure
surface as shown in Fig. 3 (a). This movement produces
certain surface settlement. When the pulse direction is
reversed (second pulse), the sliding mass on the right
side rotates as shown in Fig. 3(b); and, subsequent~
during the third pulse, the sliding mass on the left
rotates further. An estimation of the surface settlement
of any one of the sliding masses shown in the figure may
be evaluated using Newmark's block model concept by
monitoring the rotational movement of the failure mass.
Procedures to accomplish this are presented below.

rw, /gJn'e
JJlo ck

- ---g
·1 ' FaiLure
S u7'fo.cc

---~ jje x.
IV, k,,

Eit- 1
11'; '
\_..--:r
x,., - \ (
Ni = U; l;

F'o-rces in S Lice , ·i

F ig . 4 : Formul at i o n o f Circul a r Slip Mode l and

Slice Forc e s

Figure 4 shows a circular failure su;rface with a

radius R along with forces on a slice i. Here W1
represents the weight of the slice, 11 symbolizes the
length of failure surface, G and o denote the locations of
the cent~r of gravity and · the center of the failure
surface, y 1 depicts the· location of the center of gravity
from the failure surface, R* designates the distance
between G and o, and a 1 and ~~ illustrate the angles the
radial lines from o, shown in the figure, make with the

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENT FILLS SETTLEMENTS 107

vertical. The figure also presents the inertia forces on

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

the r;;lice due to the horizontal seismic coefficient, kh,

the vertical,. seismic coefL'icient, k. , and ,pngular
accelerat.~on, e. 1-!~re, k h = x. (t)/g and kv Y. (t)/g,
in which X, (t) and Y. (t) symbolize the average horizontal
and vertical acceleration, respectively, of the sliding
soil mass. As shown in the f iqure, X1 and E1 denote the
interelement forces and T1 and N1 represent the normal and
shear forces at the failure surface, respectively.
Following the steps used in the formulation of
Bishop's simplified method of slope stability analysis,
~e normal force on the failure surface during slip can be
deduced by resolving forces present on a slice (Fig. 4) in
the vertical direction as

WR
1
..
(W . (l -k ) - -s i n a1 6)
" v g (2)
cos a 1 ( 1 + tanq,tana1 }

in which a 1 represents the normal stress on the failure

surface and <1> symbolizes the frictional angle. Taking
moments for the entire sliding mass about the center, o,
for the slice after it rotates through a small angle, 8,
yields (Siddharthan, et al. 1992)

(3)

Assuming 6 is small and substituting for o 1 from Eq. 2

into Eq. 3 gives

0 (4)

@seismicisolation
@seismicisolation
 108 RETAINING STRUCTURES ANALYSIS AND DESIGN

Here, m1 = coscx 1 +tanq>sino: 1 and I, symbolize the mass moment

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

of the inertia of the sliding mass about the center, 0.

When the sliding mass is at the state of the incipent

failure, the acceleration in the body is the yield or
' critical acceleration. Thi§ state of equilibrium can be
obtained by neglecting the 8 and 8 terms in Eq. 4. This
gives

" W.Rtan4>
+ ~ , =0 (5)
mi
where the subscript y indicates the value at yielding.
This equation reveals that the ratio kh/ ( 1-kv) dictates
yielding. After defining

kh
k = --- (6a)
hv 1 - kv g - y.. ( t).

kY = - k--
h ) (6b)
hv ( 1 - kv y

and subtracting Eq. 5 from Eq . 4, the following

differential equation for 8 results.

(7)

The traditional way to obtain the failure mass is to

undertake an iterative approach to identify the critical
failure mass that yields a factor of safety of 1.0 when
seismic forces are also · included in the slope stability
calculations. The I o and the terms within the summation
in Eq. 7 need to be evaluated for this failure mass. This
has been accomplished by modifying a widely used slope
stability computer program PCSTBL5 (Achilleos, 1988) so
that it can compute these terms for the critical failure

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENT FILLS SETTLEMENTS 109

·mass that yielded a factor of safety of 1. 0. Bishop 1 s

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

simplified slope stability option is selected when using

the program. It is assumed that only a horizontal seismic
coefficient is present in the failure mass in yield
acceleration calculations. However, since kv = 0, it also
represents khv (Eq. 6b).

The formulation above requires that the acceleration

in the failure mass be uniform; in other words, "equiva-
lent uniform acceleration histories" X, (t) and Y. (t) need
to be estimated (see Eq. 6a) . The simplified procedure of
Makdisi and Seed (1978, 1979), which accounts for the
slope geometry, nonlinear soil properties, and
amplification in the slope, has been used to arrive at
~(t). This step requires a knowledge of the dynamic soil
properties such as the shear modulus and damping as a
function of strain plus the response spectrum of the
motion at the base of the fill. Dynamic soil property
variations appropriate for sandy soils, as shown in Fig.
5, have been u~ed (Seed and Idriss, 1970). However, in
the case of Y, (t), it has been assumed that it is
identical to the vertical ground acceleration at the base
of the fill.
The angle of rotation, e, of the failure mass can be
obtained by solving Eq. 7 using Newmark's numerical
integration scheme, widely known as the ~ .~ method
(Newmark, 1959; Clough and Penzien, 1975). The integra-
tion procedure gives 6 as a function of time. The input
acceleration history (horizontal and vertical) at the base
of the fill, I a and the summation terms in Eq. 7, and
yield acceleration k\; v are the input to the solution
procedure. The study adopted a linear acceleration
assumption within two time steps ( ~ = ~; ~ = 1/6).
Finally, the displacement of the rotating failure mass,
Pe1 is given by multiplying the permanent angular rotation
(6 p) by the radius of the failure surface. Estimates of
the surface settlement can be computed as (Fig. 6)

( 8)

in which ~o is the angle the radial line to the exit

point of the failure surface at the top of the slope makes
with vertical.
The methodology proposed thus far focused on one slip
surface, and only the positive acceleration pulses have
been used in the computations. As seen in Fig. 3, when
the acceleration is negative, the sliding mass located on
the right side of the figure slides. Though upslope
movement can develop in the sliding masses, the amount is
often quite small and may be neglected (Goodman and Seed,

@seismicisolation
@seismicisolation
 110 RETAINING STRUCTURES ANALYSIS AND DESIGN

,...._ 30
X 1.0
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

0
E .......
(.!) ~
........ 0.8 24
8. ""'....c:
ui Q)
::l
0.6 18 'o
:; 1;:::
'0 .....
Q)
0
::::;; 0
1... 0.4 1:2 u
0 0>
Q) c:
.c ·a.
(/)
6 E
....c: 0 .2
0
0 0
u
Q)
(/) 0.0
1.000E- 4 0.0 01 0 .01 0 0.100 1.000'
Shear Strain, -y (%)
Fig. 5: N~ ~linea~ a~~l Properties of Cohesionless

Soils (after Seed and Idriss , 1970)

-Original Position
-- Displaced Position

A.= Settlement Component

Fig. 6: Settlement Component, p 1 from Slope Rotation

1966; Ambraseys and Menu, 1988 ) . The proposed approach

to obtain the surface displacement is based on first
evaluating the r otational movement of one side ' of; t he
slope ( say , left side ) to a given acceleration history,
neglecting the upslope movement. So, only positive pulses
in the acceleration history initially are used in the
integration . Then , to estimate the rotational movement of
the right side of the slope , · the negative acceleration
pulses are subsequently used. The final estimate of the
surface settlement is the highest obtained from these two
computations .

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENT FILLS SETTLEMENTS 111

Estimation of· Settlement Component, o 2

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

As explained earlier, this component of settlement is

derived from grainslip. Tokimatsu and Seed (1987) pre-
sented a simplified procedure to compute settlements in
sandy soils on level ground . Their procedure takes into
account the nonlinea r behavior of soil. As a conservative
measure for mul tid i re.c tional shaking, Tokimatsu and Seed
recommended doubling the settlemen·t obtained with
unidirectional shaking.
The Tokimatsu and Seed (1987) procedure was applied
to evalua te p 2 using the following steps.
(1) Divide the approach fill into a number of horizontal
layers, and assign the initial shear modulus, G,.•., as

Gmax = 21B.s(x2 )ma x (a'}"

m
(9)

in which a~ mean normal stress of the layer in kPa,

and (K 2 ) m•• is a parameter that depends on the
relative density of the soil.

(2) Estimate the maximum shear stress induced in the

sublayer based on the maximum acceleration at the
top of the approach fill surface, a t~ · Makdisi and
Seed (1978, 1979) point out that the amplification
of motions in two-dimensional soil structures and
nonlinear soil behavior are important factors in the
estimation of a top • They provided simplified steps to
obtain a top • and this procedure has been adopted here
to obtain a top.
(3) Based on the shear stress in St e p 2, find the shear
strain in the sublayers using the G/G,. •• versus y
relationship appropriate for sandy soils (Fig. 5).

(4) Estimate the equivalent number of cycles, N0 q, of

excitation . Though some guidelines exist that give
N. q as a function of earthquake magnitude, a much
more realistic procedure has been employed here in
which a "weighting curve" shown in Fig. 7 has been
used (Seed, 1979). This procedure weighs the
excitation pulses with respect to the peak pulse.

(5) Utilize the shear strain (Step 3) and N8 q (Step 4) to

estimate the plastic volumetric strain for the sand
using the relative density (or SPT N1 value) of the
sand.

@seismicisolation
@seismicisolation
 11 2 RETAINING STRUCTURES ANALYSIS AND DESIGN

1 . 0+-----~~\--~r-----r-----+-----,_-----+
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

X
0
E
"\,
-saf 0.8

>
Q)
....J
C/l
C/l
~
.....
0.6

0.4
""- G......
~
-m_ --e- - - - <
(/)

.!d 0 .2
u
G 0.0+-----,_-----r-----+-----+----~r-----+
10.0 3.0 1.0 0.3 0.1 0.03 - 0.01
Equiva lent Number of Stress Cycles at
T = 0.65 T
max
Fi g . 7: We ighting Curve to Deduce Equival e nt St ress
History(after Seed, 1979)

(6) Double the volumetric strain to account for the

multidirectional shaking, and compute the sublayer
settlement by multiplying the volumetric strain by
the thickness of the sublayer. Finally, sum all of
the sublayer settlements to obtain the surface
settlement.

The steps above have been incorporated into a computer

program, and the surface settlement is computed as a
function of the base motion, relative density, and height
of the fill.

Development of Design Charts

Selection of Fill Properties and Earthquake Records

From the description of the procedures presented

above, it is clear that many parameters influence abutment
fill settlement. These parameters include the abutment
fill geometry (height and slope) , the dynamic properties
of the fill material, and the characteristics of input
motion. For example, the response spectrum of the input
motion is needed to undertake the estimation of the
"equivalent uniform" acceleration . history using the
simplified approach of Malcdisi. and Seed ( 1978, 1979).

Abutment fills are engineered fills and are typically

compacted to achieve above 90% relative compaction with
the side slopes often constructed at 2:1. This level of

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENT Fll..LS SETTLEMENTS 113

Table 1: soil P roperties Us e d in the Study

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Parameter Val ue
Relative De n sity , D, ( 'I; ) 65
1
Unit Weig ht of Soil, y (kN/m ) 18.5
Soil Parameter ( K,) mox 55
Po isso n' s Ratio, v 0.30
Angle of Internal Friction , <I> ( 0) 36

compaction leads to a medium dense sand of relative

density on the order of 65% (Byrne et al., 1987). The
recommended (K2l max value for this relative density by seed
and Idriss (1970) is 55. The unit weight and angle of
internal friction have been assumed to be 18.5 kN/m 3 and
36° respectively. Table 1 summarizes all of the assumed
soil properties. Three abutment fill heights of H = 4, 7,
and 10m have been considered .

An important step in the procedure is to arrive at

the "equivalent uniform" acceleration histories for the
failure mass. This step depends on the excitation
history, and the authors have utilized the extensive
database of excitation histories obtained from the 1994
Northridge earthquake (M = 6.7). These are high-quality
digital records obtained from 49 free-field stations by
the california Strong Motion Instrumentation Program
(CMSIP ). The Northridge earthquake is one of the most
extensively recorded urban earthquakes thereby providing
a large database of ground motions under a variety of soil
conditions and fault distances. Each station has records
of three components, two horizontal and one vertical. Out
of this database, records from 29 stations with
excitations above O.lg were singled out since often only
such excitations are important. Since each station record
consists of two horizontal components, altogether, there
are 58 horizontal acceleration histories. For practical
purposes, these histories may be considered independent
for the generation of the "equivalent uniform" time
histories. The CMSIP motions have been baseline corrected
and were used without scaling. These records represent a
maximum acceleration range of 0.1 to 1. 8g recorded at
stations within 55km of the epicenter.

Database or Settlement Component, p 1

The first step in evaluating the rotational movement

of the slope is to identify the critical failure circle
for the abutment fills selected in the study. Figure 8

@seismicisolation
@seismicisolation
 114 RETAINING STRUCTURES ANALYSIS AND DESIGN

1.6 -t-- - - - t - - - - - + - - - - - 1 - - - -----l

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

H = 4 .0, 7. 0, and 10.0 m

1.4 ~
:e
?';- """I' • After Leshchins ky e t ol. (1 994)

• PCSTBLS
~ 1.2 " ·
"'·---.,
b 1.0 +---------''-::o-~------FO
_ S_=_..:.1~.0-----I
..... ---~~
~ 0 .8 - · ---.......
~ k,W ~ TH -·----.
0.6 ~ _i kv = O.O

0.4+----~~----+-----1-----~
0.0 0.1 0.2 0 .3 0.4
Seismic Coefficient, kh
Fi g . 8 : Est ima tion of Cr i t i c al Sei s mic Coe f f ici e nt

shows the results obtained for factors of safety for

various values of kh using the computer program PCSTBL5
(Achilleos, 1988). Bishop's simplified method has been
used on a number of potential failure circles to arrive at
the critical failure circle. The plots for all H values
converge. For comparison, the results of a recent study
by Leschinsky and San (1994) are also reported. Even
though Leschinsky and San used a log spiral failure
mechanism that satisfies all of the conditions of global
equilibrium, their results are identical to those obtained
with Bishop 1 s simplified method. According to this
figure, the critical horizontal acceleration coefficient,
a y, for the abutment fills is 0.174. Since the
calculations using PCSTBL5 were carried out without any
vertical seismic coefficient, k v, this yield horizontal
acceleration will equal ki:v (see Eq. 6b) . Though khv is the
same for all of the abutment fill heights, the terms
defined in Eq. 7, for example, I 0 , W1 , y 1 , etc. differ for
each abutment fi l l height. The angle CX 0 obtained from
PCSTBL5 for the f ills H = 4, 7, and 10m are 38°, 37°, and
37° respectively.

The next step of the proposed procedure is to use the

1994 Northridge database to evaluate the "equivalent
uniform" acceleration histories for the failure masses.
These acceleration histories are to be used in the
differential equation of motion (Eq. 7) to compute the
permanent slope rotations. It is believed that the use of
the simplif ied procedure of Makdis i and seed (1978, 1979)
is consistent for abutment fill settlement computations
s~nce their procedure was developed for (1) embankments
w~th slopes at 2:1, (2) circular failure surfaces, and (3)

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENT FILLS SE'ITLEMENTS 115

medium dense sandy soils .

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

. 150~-----------------------------------,
.....
c(!) H =7.0 m
E
(!) 100 • with Vertical Ace.
~

B 0.. • without Vertical Ace.

(!) •
Cfl"CQ)
c 50 v
~ 0

--··· .....
.....c 0.
E
(!) 0
Eu
.....
::J
..0
0
~
- }
B ase
<(
-50
0.0 0 .4 0.8 1.2 1.6 2.0
Max. Base Acceleration (g)

Fig. 9: Settlement Component, p": 1994 Northridge

Database

Figure 9 shows as a typical case the permanent

displacement component, p 1 , computed for all of the
Northridge motions considered in the study for the
abutment fill with H 7m. Two cases have been
undertaken: excitation histories with and without vertical
accelerations . The figure shows that settlement in excess
of lOmm occurs only when the maximum base acceleration
exceeds 0.5g. The maximum p 1 value of 65mm is computed
when the maximum base excitation equals l.Bg. Another
important observation is that the influence of vertical
acceleration on p 1 is very minimal. On the average, the
increase in displacement when k v is included is only 5%
when p1 lies above the lOmm range. The vertical
acceleration, though it may be of significant magnitude,
is seldom in phase with the horizontal motion and is often
of a higher frequency. Therefore, although it can reduce
the normal stress on the failure plane, its net influence
on the settlement component, p 1 , has been insignificant.
The double integration of the acceleration above a
given yield acceleration value, widely know as Newmark's
approach, gives permanent slope displacement along a
planner slip surface. Many researchers (Ambraseys and
Menu, 1988; Makdisi and Seed, 1978; Ambraseys and Srbulov,
1994) have interpreted the displacement given by Newmark's
approach as horizontal slope displacement. Unlike the
circular slip displacement model used above, the planner
slip model requires only the yield acceleration and
excitation time histories to compute the slope movement.

@seismicisolation
@seismicisolation
 11 6 RETAINING STRUCTURES ANALYSIS AND DESIGN

<["" 200
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

~
,......, H = 7.0 m
cCl) E
c 0 2
0 ---: 150 r = 0.93
D.. ,......,
a:;
E '0
0
(.) 0
.....c ~ 100 Y = 0.82 X
Cl) .9-
E Ui
Cl) ...0
BCl) '3 50
(/)
2
D.. ~
Cl)

0
Vi
150 200
Slope Displacement (Planner Slip Model ), (em)

Fi g. 1 0: Correlat i on Between Sett l e me n t p 1 and

Di splacement from Planne r S li p Mode l

In studies by Ambraseys and coworkers, the measured

excitation histories are us e d directly without
modification for amplification or variation of the
acceleration within the failure ma s s . Such a simplified
interpretation has been used with the entire 58 Northridge
database records. Figure 10 shows th e se displacements
along with the settlement ( p 1 ) computed using the proposed
circular slip model for H = 7m . A linear correlation (y
2
= 0.82x) between these two models with an r value of 0.93
exists as shown. Similarly, for the other abutment fill
2
heights of H = 4 and 10m, the r values are 0.99 and 0.8'l.
The slopes of the corresponding correlation equations are
0.78 and 0.88 respectively. Since these correlation
coefficients do not differ greatly, an average value of
o. 8 3 may be used irrespective of abutment f i 11 height.
Because these levels of r 2 values are considered a good
fit in many geotechnical de s igns, it appears that the
displacements from one model can be obtained from the
other using the a p propriate correlation equation .
Ambraseys and Srbulov (1994) have reported on a
comprehensive study in which they computed permanent slope
displacement using the planner slip model . They used a
very extensive earthquake excitation database which
consisted of as many as 532 time histories. This database ·
came from 76 world-wide shallow earthquake events (focal
depth h < 25km) of magnitudes, M, r anging from 5.0 to 7.7.
The mean depth of all of the events in the database is 10
(±4 ) km and the mean magnitude 6.0 (± 0.7). The recordings
were made on soil ( 389 records ) and at rock sites ( 143 )
records . The researchers used this extensive database t o
develop a regression equation for permanent slope

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENT FILLS SETTLEMENTS 117

displacement, uP, in em for planner failure surface as

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

log (UP ) = C1 + C2 M +. C3 r + logl(l - q) "q mj (10)

in which r = focal distance in km; q = the ratio between

the yield acceleration and the maximum acceleration (ay/
a.,.,); and C1 - C3 , m, and n symbolize regression constants.
Ambraseys and Srbulov (199 4 ) give C2 = 0.473, C3 = -0.01,
m = -1. 02, and n = 2. 64. ·For an average value of uP, C1
= -2.41; and, for an 84% value, C1 = -1.83 . .
The regression equation (Eq. 10) above of Ambraseys
and Srbulov has very wide applicability since it is based
on an extensive database. Therefore, one can utilize this
regression equation to evaluate the abutment fill
settlement, p1 • It may be achieved by evaluating p 1 as

(11)

in which uP depicts the slope displacement computed using

Eq. 10 and F••• = 0.83 is the modification factor for
abutment fill heights of 4 to 10m. The 1994 Northridge
excitation database used here represents the most
extensively recorded earthquake to date, with records
representing a variety of soil conditions and epicentral
distances. These records also varied in excitation
strength ( a m•x = 0 .1 to 1. Bg) and, thus 1 covered a wide
range of soil strains (i.e., level of nonlinearity) and
amplifications (or deamplifications) in the slopes
considered in the study. Under such circumstances, the
equation above may also be used to estimate p1 for events
and field conditions other than the 1994 Northridge
earthquake. The Ambraseys-Srbulov equation (Eq. 10) is
simpler to use since it does not require a base
acceleration history, a study of amplification within the
slope, or the integration of Eq. 7. It may be argued
that, since integration of excitation records are used to
evaluate the slope deformation, the spectral
characteristics of the excitations from other earthquakes
of different magnitudes can influence the deformation, p1 •
This aspect is already accounted for in the evaluation of
uP since records from different earthquake magnitudes have
been used to arrive at the value of uP (Eq. 10).

Estimation of settlement Component, p2

The height of the fill and the excitation strength
govern the grainslip induced settlement component, p 2 , for
a given abutment fill. The abutment fill settlements have
been computed for the Northridge motions (Fig. 11) using
the approach based on the Tokimatsu and Seed (1987)

@seismicisolation
@seismicisolation
 liS RETAINING STRUCTURES ANALYSIS AND DESIGN
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

200~----------------------------------~
+'
cQ) H = 7.0 m 1994 Northridge Do\o
150
E
V N
:E<1l+-'': 100
(/J c
- (I)
~ 5 50
_,_,a.
c
<1)
E
0
Eu
+-'

.a
:J - so
<(

- 100+-----_,------~------+-----_,------~
0.0 0.4 0.8 1.2 1.6 2.0
Max. Bose Acceleration (g)

Fig. 11 : Settlement Component, p 2 : 1994 No rt h ridge

Database

procedure outlined earlier and the abutment fill

properties given in Table 1. As a typical plot, only the
case of H = 7m has been shown with an envelope of
settlement superimposed on the figure. The maximum base
acceleration in excess of 0.4g gives rise to a settlement
above 20mm. A maximum settlement o f as much as 75mm has
been computed. There appears to be a limit to the
settlement at around 70mm irrespective of the excitation
level. Volumetric measurements made by Youd ( 1972)
support this result. Since Tokimatsu and Seed (1987) used
Youd's data to develop their settlement procedure, such a
characteristic is expected in the results reported.
Deamplification, which occurs at higher excitation levels
is another reason for the limit. Figure 12 presents the
settlement envelopes for all of the abutment fill heights
as a function of the maximum base acceleration. As
expected, the p2 values increase for taller abutment fill
heights.

The plots given in Fig. 12 can be used directly for

earthquakes of a magnitude similar to the 1994 Northridge
earthquake (M = 6.7). A close look at the Tokimatsu-Seed
procedure reveals that, unlike in the case of p 1 , the
settlement component p 2 , is not a function of the
frequency components of the excitation but of the number
of cycles of motions induced by the excitation. For
excitations from other magnitudes · of earthquakes, the
component p 2 , can be readily scaled as described by
Tokimatsu and Seed (1987). They provide scaling factors
based on a reference earthquake magnitude M = 7.5. Figure
13 gives the scaling factor, F, , based on the 1994
Northridge earthquake magnitude (M 6.7). The

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENT FILLS SETTLEMENTS 119

200 ~-------------------------------------.
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

.... ···- H = 4.0 m 1994 Northridge Data

~ 150 ·· ··· H = 7.0 m
- H = 10.0 m
~ <"J
~ .:; 100
U1 ~ .... .... ... ......... ...... ... ....... ...
~ :5 50
~C1l E
0 0 _....·.•:.:.:·- ··-·· - ··-··-
_so
::J
~ - 50

-100+-------r------+------~------+-----~
0.0 0.4 0.8 1.2 1.6 2.0
Max. Base Acceleration (g)

F1g. 12: s~tt lement Compon e nt, p 2 for Va r ious Abutment

Heights
2.0..-- - -- - - - - - - -- ------ ---.,
Fv = 0.32 M - 1 . 18
2
I..J..> 1.5 r = 0.997
..:
0
+-'
u
~ 1.0
0>
c
0
u 0.5
en

0.0+----~~-----r------~--~~
5.0 6.0 7 .0 8.0 9 .0
Earthquake Magnitude, M
Fi g. 1 3: Mo dification Factor, Fv for Earthquake
Mag n itude

deformation component, p 2 , for any earthquake magnitude,

M, may be computed using

(12)

for 5~ < M < alo,

in which p 2 with subscript M 6.7 is given by Fig. 12.

@seismicisolation
@seismicisolation
 120 RETAINING STRUCTURES ANALYSIS AND DESIGN

'E 2oo.o.,---------------------,
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

'-'
tJ H = 7.0 m
....,
~ 150.0 1994 Northridge Doto
E
QJ

~ 100.0
(/)
...., 0
c
g 50.0
CD 0

E
~ 0 .0
0
+J
~ -5 0 . 0+-------+-------~-------r-------r------~
0.0 0.4 0.8 1.2 1.6 2.0
Max . Rose Acceleration (q)
F:~ . c4 : Total Settlempnr of Abutment Fills:
1994 Northridge Database
~lication to Field; 1994 Northridge

Thus far, procedures have been outlined to indiv~­

dua lly obtain p 1 and P2 • The total abutment fill settlement
is the sum of the components p 1 and p 2 • For the case of
the Northridge data, Fig. 14 shows the computed total
settlements for the case of H = 7rn. Equations 11 and 12
have been used to compute the total settlement utilizing
the available database of the Northridge acceleration
records. The figure also presents an envelope of total
settlement. Subsequently, Fig. 15 reports the correspon-
ding envelopes for all of the abutment fill heights
considered in the study. This plot can be directly used
to obtain estimates of the total settlement o f the
abutment fills for an earthquake similar in magnit11de to
E 200.0 .... . H = 4.0 m
-!!.,.
--H=7.0m
~

~ 150.0
- H = 10.0m -- ..--··I
.,E
/

"0
a. 100.0 _,
Ill
a / ... ····
c0 50.0
c:
tu
E
., 1994 Northridge Data
a.. 0.0
:s
~ -so.o.+:------::+----i----+----+- - --l
0.0 0.4 0.8 1.2 1.6 2.0
Max. Bose Acceleration (g)
Fig . 15: Total Settlement of Abutment Fills of Various
Heights

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENT FILLS SEITLEMENTS 121

that of the Northrid9e earthquake (M = 6.7).

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

conclusions

Two deformation mechanisms contribute to the surface

settlement of cohesionless abutemnt fills: slope movements
along the failure surface (component, p 1 ) , and grainslip
induced deformation due to the cyclic nature of the
seismic excitation (component, p 2 ) . These components of
deformation depend on the excitation history, amplifica-
tion (or deamplification) of the ground motion, slope
geometry, and nonlinear soil properties. A comprehensive,
yet simple, approach has been presented to estimate these
settlement components for medium dense cohesionless
abutment fills of height 4 to 10m with side slopes at 2:1.
Unlike other past studies, the proposed approach includes
the vertical acceleration in the formulation. The
procedure developed is realistic and is in a form readily
usable by designers.
The settlement component, p 1 , has been computed based
on the more realistic nonplaner (circular) failure
surface; while the procedure to evaluate p 2 is based on
the Tokimatsu-seed (1987) approach. simple relationships
have been developed to evaluate p 1 and p 2 (Eqs. 11 and 12)
as a function of earthquake magnitude, epicentral
distance, and the ratio between the yield acceleration and
the maximum acceleration at the base of the fill. These
relationships have been developed using the 58
acceleration time histories recorded during the 1994
Northridge earthquake. The study also reveals that the
influence of vertical acceleration on slope deformation is
i~siqnificant (only about 5%). Since the proposed
procedures are based on many well-established geotechnical
procedures, the relationships presented can be used, in
many cases, as a first approximation for evaluating slope
deformation in most earthquake situations.

Acknowledgments

This study was funded by National Science Foundation

grants (BCS-9221314 and CMS-9415955). The support is
gratefully acknowledged. Many thanks are due to Jeanie
Pratt for her excellent typing.

References

1. Achilleos, E. (1988). "User guide for PCSTABL5,"

Research Report, Department of Civil Engineering,
Purdue University, Indiana, USA.

@seismicisolation
@seismicisolation
 J22 RETAINING STRUCTURES ANALYSIS AND DESIGN

2. Alnbraseys, N.N., and Menu, J.M. (1988). "Earthquake-

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

induced ground displacements," J. Earthq. Engrg. and

struct. Dyn., Vol. 16(7), pp. 985-1006.
3. Arnbraseys, N.N., and Srbulov, M. (1994 ) . "Attenuation
of earthquake-induced ground displacements," J.
Earth. Engrg. and struct. Dyn., Vol. 23, pp. 467-
487.

4. Byrne, P.M., Cheung, H . , and Yan, L. (1987). "Soil

parameters for deformation analysis of soil masses,'
Canadian Geotech. J., Vol. 24, pp. 366-376.
5. Clough , R.W., and Penzien, J., (1975). Dynamics of
Structures, McGraw-Hill Book Co., New York.
6. El-Gamal, M., "Seismic design issues relating to
bridge abutments, Ph.D. Dissertation submitted to
the University of Nevada, Reno, May 1996.
7. Franklin, A.G., and Chang, F.K. (1977). "Permanent
displacements of earth embankments by Newmark
sliding blocks analysis," Miscellaneous Paper S-71-
17, Soil and Pavement Laboratory, u.s. Army
Waterways Experiment Station, Vicksburg,
Mississippi.

B. Goodman, R.E., and Seed, H. B. {1966). "Earthquake-

induced displacements in sand embankments," J. Soil
Mech. and Foundation Engrg., ASCE, Vol. 92(2), pp.
125-146.
9. Jibson, R.W. (1993). "Predicting earthquake-induced
landslide displacements using Newmark's sliding
block analysis," Transportation Research Record, No.
1411, National Research Council, Washington, D.C.,
pp. 9-17.
10. Le~ch~nsky, D., and San, K.C. (1994). "Pseudo-static
se1sm1c stability of slopes: design charts," J .
Geotech. Engrg., ASCE, Vol. 120(9), pp. 1514-1532.
11. Ling, H.I., and Leschinsky, D. (1995 ) . "Seismic
performance of simple slopes," Soils and Foundations,
Vol. 35(2), pp. 85-94.
12. Makdisi, F.I., and Seed, H.B. (1978 ) . "Simplified
procedure for estimating dam and embankment
earthquake-induced deformations," J. Geotech. Engrg.,
ASCE, Vol. 104(7), pp. 849-867.
13. Makdisi, F. I., and Seed, H. B. (1979). "Simplified
procedure for evaluating embankment response," J .
Geotech. Engrg., ASCE, Vol. 105(12), pp. 1427-1434 .

@seismicisolation
@seismicisolation
 BRIDGE ABUTMENT FIT.LS SETTLEMENTS 123

14. Newmark, N.M. (1959). "A method of computation for

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

structural dynamics, • J. Engrg. Mech. , ASCE, Vol.

85(3), pp. 67-94.

15. Newmark, N.M. (1965). "Effects of earthquakes on dams

and embankments," Geotechnique, Vol. 15(2), pp. 139-
160.

16. Prevost, J.H. (1985). "Nonlinear dynamic analyses of

an earth dam," J. Geotech. Engrg., ASCE, Vol. 111 (7) ,
pp. 882-897.

17. Richards, R., Jr., and Elms, D.G. (1979). "seismic

behavior of gravity retaining walls," J. Geotech.
Engrg., ASCE, Vol. 105(4), pp. 449-464.
18. Seed, H.B., and Idriss, I.M. (1970). "Soil moduli and
damping in soils; design equations and charts," Rep.
No. EERC 70-10, Earthquake Engineering Research
Center, University of California, Berkeley.

19. Seed, H. B. (1979). "Soil liquefaction and cyclic

mobility evaluations for level ground during
earthquakes," J. Geotech. Engrg., ASCE, Vol. 105(2),
pp. 201-255.
20. Siddharthan, R., Ara, s., and Norris, G.M. (1992). "A
simple rigid plastic model for seismic tilting of
rigid walls,• J. struct. Engrg., ASCE, Vol. 118(2),
pp. 469-487.
21. Tokimatsu, K., and Seed, H.B. (1987). "Evaluation of
settlements in sands due to earthquake shaking," J.
Geotech. Engrg., ASCE, Vol. 113(8), pp. 861-878.
22. Ueng, J., Hutchins, L., Jarpe, s., Kasameyer, Pl,
and Henze, F. (1994). "A new, combined seismic-
geotechnical approach for estimating site-specific
strong motion: application to the painter street
bridge site, Rio Dell, CA," 3rd Annual seism.
Research Workshop, Caltrans, Sacramento, California.
23. Yegian, M.K., Marciano, E.A., and Ghahraman, V .G.
(1991) . "Earthquake-induced permanent deformations:
probabilistic approach," J. Geotech. Engrg. , ASCE,
Vol. 117(1), pp. 35-50.
24. Youd, T.L. (1972). "Compaction of sands by repeated
shear straining," J. Soil Mech. and Foundation
Engrg., ASCE, Vol. 98, pp. 709-725.

@seismicisolation
@seismicisolation
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Earthquake Destructiveness Potential Factor and Permanent

Displacements of Gravity Retaining Walls

Teresa Crespellani1 , Claudia Madiai 2 and Giovanni Vannucchi 3

Abstract
An empirical equation for estimating the permanent displace-
ments of gravity retaining walls with dry backfill during earth-
quakes is proposed . It was obtained analysing the displacements
of Newmark's block produced by the horizontal components of 155
earthquakes which took place in various parts of the world.
The displacements were correlated with the earthquake de-
structiveness potential factor, Po, a parameter representative of
earthquake severity introduced by Araya and Saragoni (1984}. This
parameter, defined as the ratio between the Arias intensity and
the square value of zero crossings per second of an accelerogram,
synthetises the major characteristics of the seismic ground IrO-
tion which can affect the behaviour of the wall . For each record
of the earthquakes considered, the potential destructiveness fac-
tor and the final displacements of Newmark's block were de ter-
mined using different values for the resistance factor, N, and,
with the multiple regression technique, the equation that best
fits the calculated values was determined . This equation was com-
pared with the equation suggested by Wong (1982) and their uncer-
tainties were analysed. Finally an empirical equation for esti-
mating the displacements of a gravity retaining wall is obtained.
Introduction

Methods for estimating the permanent displacements of re-

taining walls away from waterfronts, during ·earthquakes are a

1
Assoc. Prof . , Dept. of Civil Engrg., Florence Univ.
2
Assist. Prof. , Dept. of Civil Engrg., Florence Univ.
3
Assoc. Prof. , Dept. of Civil Engrg., Florence Univ. ,
Via s . Marta 3, 50139 Florence, Italy

124

@seismicisolation
@seismicisolation
 GRAVITY RETAINING WALLS DISPLACEMENTS 125

useful design tool since they offer a fair compromise between the
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

need for simplicity in representing a complex physical phenomenon

and the need to obtain solutions which satisfy essential safety
requirements at a· minimum cost. Numerous efforts in this direc-
tion have been analysed in recent state of the art and papers
{e.g. Whitman, 1991; Nadirn & Whitman, 1993; Prakash et al.,
1995) .
The first approach to the seismic design of gravity retain-
ing walls based on the concept of Newmark's sliding block was de-
veloped by Richards and Elms (1979), who recommended the follow-
ing equation as conservative expression for displacements:

v
d $ 0.087.-.
2
(N)-•
- (1)
Aq A

where d is the displacement of the wall, Ag is the peak accelera-

tion, Vis the peak velocity and N is the resistance factor.
The limits, sources of errors and the uncertainties associated
with the use of such an equation are as follows:
A. Uncertainties linked to predicting the ground rrotion;
B. Uncertainties in the description of the ground motion derived
from few, synthetic earthquake severity parameters;
C. Uncertainties in the resistance factor N;
D. Model errors concerned with kinematic constraints, deformabil-
ity of backfill and tilting.
The uncertainties indicated in points A and B are generally
overlooked by the designer, given that they are specifically re-
lated to design factors in seismic codes , derived in turn from
studies of seismic hazard performed at the regional and local
level. Since the selection of the seismic parameters of the ex-
pected earthquake over a given period is the last stage in seis-
mic hazard analysis, the uncertainty associated with these pa-
rameters is very high due to the cumulative effect of the errors
and uncertainties in each stage of the analysis. In the present
paper the authors suggest replacing the equation (1) in Richards
and Elms design procedure with a different equation allowing es-
timation of the displacement of Newmark's sliding block (and not
only its upper limit) as a function of a single descriptive ~­
rameter of the design earthquake: Araya and Saragoni' s destruc-
tiveness potential factor, PD (1984). By using this parameter,
which synthetically summarizes the major characteristics of an
accelerogram, the uncertainties in obtaining an estimate are sig-
nificantly reduced.

@seismicisolation
@seismicisolation
 126 RETAINING STRUCTURES ANALYSIS AND DESIGN

Seismic destructiveness potential factor

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

The seismic destructiveness potential factor is defined by

the ratio between the Arias intensity, IAf and the square value
of zero crossing per second of accelerograms v/, i.e.

to
f a2(t) · dt
IA 7t 0 (2 ).
PD 2g
v2 v2
0 0

The horizontal destructiveness potential factor, Pw is the

vectorial sum of the Pv of the two horizontal corcp::>nents of the
earthquake. The destructiveness potential factor (expressed by
the unit 10-~ g·s 3 ) is a parameter which is independent of the
source mechanism of an earthquake and is correlated to the mac-
roseismic intensity and other significant seismic parameters. For
instance, using data from accelerograms recorded during North
.American and Chilean earthquakes having Pw ~ 20·10-4g·s3 , Saragoni
et al. (1989) established the following correlation with the Mer-
calli m::x:lified intensity Im:

log Pw = 0. 67 ·Im - 3. 04 (3)

For a given value of the resistance factor N, the permanent

cumulative displacement of Newmark's sliding block depends on
the characteristics of the design time history of acceleration,
that is the peak value, the duration of strong motion, the mode-
shape and the frequency content (in other words how much, how of-
ten and how long the earthquake acceleration of the base exceeds
the critical value); all this information is synthetically in-
cluded in the definition of the destructiveness potential factor.

Correlation between the destructiveness potential factor and the

displacements of Newmark's sliding block

The seismic data base was originally composed of 194 actual

acceleration records, each with two components of ootion, re-
corded in Europe (Italy, Greece, former Yugoslavia) and in the
USA. Nevertheless a few records were rejected either because tbey
represented very weak earthquakes or because one or more parame-
ters have anomalous values (outliers), probably due to recording
errors. Thus in defining the relationship between the displace-
ments of Newmark's block and the destructiveness potential fac-
tor, the data base is composed of 155 actual accel eration rec-
ords. In Table 1 the statistical distribution parameters of the

@seismicisolation
@seismicisolation
 GRAVITY RETAINING WALLS DISPLACEMENTS 127

rost significant seismic characteristics of the data base are

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

shown. The statistical distributions are all s trongly skewed

· right and the P0 and I A distributions exhibi t a very high kurto-
sis. It may be assumed that, in general, all the probability dis-
tributions are log-normal.

Table 1 - Parameters of the statistical distribution of some

charac t eristics of the data base
OOR. A V VI A llo P0 IA
[s] [an/s] [an/s] [s-1 ] [io-<~ g s 3 ] [an/s]
·ma:x·:. .....................a3·:·6............cl':'s2..........4'i':·ii-........23'eLT. .....43':·6.............3iL'66.............T9!L2 . ....
min. 5.1 0.01 0.5 11.2 4.7 0.002 0.1
mean 21.8 0.10 6. 3 61.0 12.2 1.996 14.4
std.dev. 13. 2 0.09 8.3 35.9 5.2 5.40 30.0
skewness 1.9 1.9 2 .5 1.6 1.8 4.5 3.6
kurtosis 4.3 3.6 6.4 3.3 5.6 22 .7 14.7

Figures la to 1d show the frequency histograms and cumulative

frequenc ies of the l ogarithms of the parameters: A, V, VI A and
Po.
histogram of in(A) histogram of inM
120 100% 60 100%
100 .. a)
80% .,"'c:
0

:>
50 80% .,"'c:
0

:>
i;' 80 . g- ~ 40 g-
c 60% ol= c: 60%
., ~ 30
ol=
""
60
">
"":;.. "":;.
>
:if
.1: 40
40%
! 20
40%

20 20% E 10 20% E
:> :>
0 0
0 0% 0 0%
.... ....
"'"!' 'I' 1 0 N
<)l
<D
~
9"" "'9 0
c;j
CD
c;j
"!
~ N
N
<'i
0

' ""'
histogram of In (VIA) histogram of in(P0 )
140 100% 60 . 100%
~ ~
120
1;- 100
c)
. 80% .,
c:
:>
50 d)
80% .,
c:

c:
80
g-
.. 60% ol=
ij' 40
60%
"g-
"" 60 .,> c:
~ 30 . .,
ol=
:if
.::
40
40%
~
:;
20% E
f 20
40% · ~
20% E
:;
..
20 10
:>
0 0%
0
0 0% "
0

.... 0 <D N ....

C'l <'i <'i
""'
"" ori
Ill
"' Ill

·~ "!' <)l 9
Ill If.!
"',..;
Figures 1a to 1d: ""'Frequency histograms and cumulative fre quencies
of ln (A) 1 ln (V), ln (VI A) and ln (P0 )

For each record the des tructiveness potential factor Po and

the final displacements d, of the sliding block (wi th one way
sliding) with both horizontal components and both directions of

@seismicisolation
@seismicisolation
 128 RETAINING STRUCTURES ANALYSIS AND DESIGN

acceleration, with six different values of the resistance factor

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

(N= 0.005; 0.01; 0.02; 0.05; 0.1; 0.2) were calculated (figs. 2a
to 2f).
N = 0.005 N = O.Ol N=O.D2
1~·r--r--r--r--r-~ 1000·r-~--~~--r-~

d[cm] b) d[cm] o)
d

0.1 0.1
0.001 0.01 0.1 1 10 100 0.001 0.01 0,1 1 10 100
Po [I IT' g IJ Po (liT' gl]

N= 0.05 N = O.l N=0.2

1~ 1~
1000
d) c)
d[cm) d[cm] d[cn •I I)

100 100 100

.I
10 10 0

~ i/_
~
a ""
0.11---- 1--""""""-'--+---l 0.1
0.001 0.01 0.1 1 10 100 0.001 0,01 0.1 1 10 100
Po [I IT' g II Po [liT' gl]

Figures 2a to 2f: Calculated final displacements of the sliding

block versus destructiveness potential factor for different val-
ues of the resistance factor and regression from eq. (4) (straight
line)
The scattergrams in a logarithmic scale indicate that the
variables are almost linearly correlated. The relationship link-
ing the displacement of Newmark's block and the destructiveness

@seismicisolation
@seismicisolation
 GRAVITY RETAINING WALLS DISPLACEMENTS 129

p:>tential factor obtained through a linear multiple regression

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

over a total of .l940 observations is as follows:

d = O.O ll 0 pgo977 0 N-lo33 o [an] (4)

The correlation coefficient is r = 0. 917 and the standard

error of the estimate of Y = ln (d) is Sy = 0. 792. In figures 2a
to 2f the traces of the plane of the above equation are also rep-
resented. It is possible to observe that the scatter increases
with an increase in N. Therefore the error of the estimate varies
with this parameter. It can also be observed that for a periodi-
cal motion the exponent of Pn is one.
Whitman and Liao (1984) indicated that a good estimate of
the mean value of the displacement of Newmark's block is given by
the following equation (Wong, 1982) :

d = 37 ° -
A ·g
v -
2
• exp-
( 9.4 · -
A
N) [an] (5)

This equation was obtained by a statistical analysis of the per-

manent displacements of a block sliding in one direction using a
data base consisting of the horizontal components of 14 acceler-
ometric records. The coefficient of variation of the estimated
displacement of Newmark ' s sliding block ranges from about 0. 6 at
smaller N/A to about 1.4 at larger N/ A.
In order to compare and to test the influence of the data
base, the same linear model by Wong was applied to the data base
adopted in the present paper .
The resulting relationship between the independent variable
X=~ and the dependent variable Y == ln( d · ;
0

g) is:
Aog
2
d = 34.08 · -v - · exp ( - 7.74 · -
A
N) [an] (6)

the correlation coefficient is r = 0.921, the standard error · of

the estimate of Y is Sy = 0. 582. Figure 3 shows the representa-
tive points of the ratio ( d · : 2
° g) obtained by integration ver-

sus the ratio (1f)in semi-logarithmic scale and the straight line
corresponding to the eq. (6) .
The coefficients of Eq. (6) are very close to the coeffi-
cients of Eq . (5), which, even though obtained with a smaller data
base , is thus confirmed to be an excellent means of estimation.

@seismicisolation
@seismicisolation
 130 RETAINING STRUCTURES ANALYSIS AND DESIGN

Eqs. (4) and (6) have very similar correlation coefficients,

Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

but the former has the advantage of requiring the knowledge of

only one seismic parameter, Po, instead of two parameters A and
V. The ratio VIA decreases with epicentral distance and varies
considerably with local lithology (Wang & Law, 1994), so that, as
can be noted in .Table 1, it presents a high scatter. Most na-
tional seismic codes give either no indication of the design peak
velocity or a constant value of the ratio VI A (i. e. the USA ATC
gives VIA = 76.2 an/s ; this value is quite similar to the aver-
age ratio of Table 1, equal to 61.0 cm/s). The correlation coef-
ficient between the A and V values of the data base is r = 0. 732.
However, the main advantage of equation (4) as regards equa-
tion ( 6) consists of the fact that the coefficient of variation
in the expected value of destructiveness potential factor,
c.o. v. (Po), resulting from seismic hazard analysis is decidedly
inferior to the corresponding coefficients of variation in ex-
pected values of peak acceleration, c.o . v . (A), and peak velocity,
c.o.v. (V). The typical ranges of the c .o.v. (A) and of the
c.o.v. (V) are 0.5-0.8, whereas the range of c . o.v. (Po) is 0.1-
0.2. This value was calculated starting from the value of the
c .o.v. of the macroseisrnic intensity, using Eq. (3). Consequently
the estimate of the block displacement is more precise.

1000

100

.... 10
::..
"b()
~
""<:!
0.1

0.01
+
0.001
0 0.2 0.4 0.6 0.8
NIA

Figure 3: Calcul ated d · A2 • g ratio. versus N and regression

. V A
from equation (6) (straight line)

@seismicisolation
@seismicisolation
 GRAVITY RETAINING WALLS DISPLACEMENTS 131
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

An example with realistic values of seismic and geotechnical

parameters is presented. A sliding block with resistance fact or N
= 0.01 is considered. The equations (4 ) and (6) are rrodified in
order to include the uncertainties of the regressions . The prob-
~ilistic analysis is carr ied out by means of numerical simula-
tion, assuming the log-normal distribution of the variables PD,
V, A, Cr and Cn.
a) Solution using eq. (4): d = 0.011 · p~· 917 • N - 1· 338 • Cx [em]
where Cx is a lognormal variate with unit mean and standard de-
viation equal to the standard error of the regression (0 .79 ).
design seismic data :

results of numerical simulation :

~ 4
97............ill(~)
·n;ea:n...........................cs· i5':·4s·?".........L.oos..........er:·2d43.....
std. dev. 0 . 238 0 . 148 0 . 802 6 . 758
c.o.v. 0.149 0.323 0 . 798 0.820
skewness 0. 464 0 . 016 2 . 969 3, 005
kurtosis 0 . 453 -0 . 054 15. 616 15.345

b) Solution using eq. (6):

d = 34.08 · ~
A·g
· exp(- 7.74 · A !!..) . c II
[em]
where Cu is a lognormal variate with unit mean and standard de-
viation equal to the standard error of the regression (0.58 ) . The
correlation coefficient between V and A is assumed to equal 0. 73 .
design seismic data:
v ln (V) A ln (A) Cn ln(Cu)
mean 7.2 1.798 0.12 -2.296 1.0 -0.145
std. dev. 4.68 0.594 0 . 08 0.594 0.58 0 . 538

results of numerical simulation:

V ln(V) A m (A) Cn d
"liieaii.......................,L.i36..........:r:·'7'sa.........cr:. n9. . . ::·:r:·3o'7..........a.:·9efs.. ... . ti":·6cfo. ..
std. dev. 4 . 589 0 . 597 0 . 079 0 . 601 0.559 12.156
c.o.v. 0.643 0.334 0.660 -0 . 260 0.567 1.412
skewness 1.891 -0.018 2 . 223 0.011 1.848 3.866
kurtosis 5.536 -0.025 9.184 - 0 . 004 6.093 22.08

@seismicisolation
@seismicisolation
 132 RETAINING STRUCTURES ANALYSIS AND DESIGN

The exampl e shows that even if the estimation error from the
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

regression of eq. {6) is lower than that of eq. {4) {respectively

0.58 and 0.79), and .the assumed correlation coefficient between A
and V is rather high {0. 73) , the c . o. v. {d) resulting from the re-
gression of eq. {4) is decidedly lower than the c .o.v. {d) ob-
tained from the regression of eq. {6) {respectively 0.820 and
1.412).

Wall displacements

The various sources of model errors, indicated above at

point D, were analysed by Whitman and Liao {1984) using statisti-
cal and probabilistic techniques. An important conclusion .of
their analysis is that the mean value of the permanent displace-
ments of a gravity retaining wall is equal to 3.51 times the mean
value of the permanent displacements of the corresponding New-
mark's sliding block.
Consequently the errpirical design relationship suggested in
this paper is:

d = 0.0 38 . p~·g'' . N-l.lle [em] {8)

Conclusions

In applying the methods of calculating the movements of

gravity retaining walls {and also of slopes) during an earth-
quake, errpirical correlations between the movements of Newmark's
block and parameters representing the seismic ground motion are
used. In this way the need to choose a single complete design ac-
celerogram is avoided. It is shown, for example, that there is a
very good correlation between the movement, d, and two seismic
parameters : the peak acceleration, Ag, and peak velocity, V.
In this study it is suggested that an errpirical correlation
between the displacement, d, and a single parameter representing
the seismic ground ootion, that is, the earthquake destructive-
ness potential factor , PD, may be used. The use of a single pa-
rameter instead of two considerably reduces the possibility of
uncertainty in estimating the displacement of the block.
Studies of regional and local seismic hazard could produce .a
zoning of the area based on the destructiveness potential factor ,
similar to that based on the maximum acceleration, maximum veloc-
ity and macroseisrnic intensity. Design values- of Po could be in-
cluded in seismic codes. Until such information is available, the
expected value of PD can be calculated using the existing regres-
sions between this parameter and the macroseismic intensity.

@seismicisolation
@seismicisolation
 GRAVITY RETAINING WALLS DISPLACEMENTS 133

References
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Araya, R., Saragoni, R. (1984) Earthquake accelerogram destruc-

tiveness potential factor. Proc .8th WCEE, 1984 , s. Francisco,
vol. 2, 835-841
ATC (1978 ) Tentative provisions for the developnent of seismic
regulation for buildings. Applied Technology Council 3-06 &
National Bureau of Standards, USA, 1978
Nadim, F., Whitman, R. (1993) Seismic analysis and design of re-
taining walls. Soil Dynamics and Geotechnical Engineering,
Ba1kema, Rotterdam, 387-411
Newmark, N. M. (1965) Effects of earthquakes on dams and embank-
ments. Geotechnique , vol . 15, No.2, 139-159
Prakash s., Wu Y., Rafnsson, E.A. (1 995 ) On seismic design dis-
placements of rigid retaining walls. Proc. 300 International
Conference on Recent Advances in Geotechnical Earthquake En-
gineering and Soil Dynamics, St. Louis, Missouri, 83-92
Richards, R., Elms, D. (1979) Seismic behavior of gravity retain-
ing walls. JGED, ASCE , vol . 105, No. GT4, April
Saragoni, R., Holmberg, A ., Saez, A. (1989) Potencial destructivo
y destructividad del terremoto del Chile de 1985, Proceed.
Sas. Jorn . Chilenas de Sismologia e Ing. Antisismica, Agosto
1989, vol. 1, 369-378
Whibnan, R.V. (1991) Seismic design of earth retaining struc-
tures . State of the art report . Proc. 200 International Con-
ference on ·Recent Advances in Geotechnical Earthquake Engi-
neering and Soil Dynamics, St. Louis, Missouri, vol. II,
1767-1778
Whibnan, R. V. , Liao, S. (1984) Seismic design of gravity re-
taining walls. Proc.sthWCEE, 1984, S. Francisco, vol. 3, 533-
540
Wang, J.G.Z.Q., Law, K.T. , (1994) Siting in earthquake zones,
Balkerna, Rotterdam
Wong, C.P., (1982) Seismic Analysis and an ~roved design proce-
dure for gravity retaining walls, S.M. Thesis , Research Re-:-
port R82-32, Dept . of Civ. Eng., Massachusetts Institute of
Technology

@seismicisolation
@seismicisolation
 Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

@seismicisolation
@seismicisolation
 SUBJECT INDEX
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Page number refers to the first page of paper

Bridge abutments, 77, 100 Models, 1, 124

Cantilevers, 19, 57 Passive control, 77

Centrifuge, 38
Concrete, 57 Retaining walls , 19, 21, 38, 77, 124
Rigid frames, 1, 21, 124
Design, 2 1 Rotation, 38
Displ acement, 38
Displacements, 19, 2 1, 57, 124 Seismic analysis, 77
Dynamic response, 19, 57 Seismic design, 57, 124
Seismic effects, 1, 2 1, 38, 100
Earth pressure, 19 Sliding, 77 , 124
Earthquake damage, 100, 124 Soil settlement, 100
Earthquake excitation, 19 Soil-structure interaction, 57
Earthqu ake loads, 38
Earthquakes, 1, 57, 77 Tiltmeters, 77
Elasti c media, 1
Validation, 38
Failures, 21
Fills, 100 Walls, 1, 57
Finite element method, 1

Gravity walls , 38, 124

Ground motion , 1, 19, 100

135
@seismicisolation
@seismicisolation
 AUTHOR INDEX
Page number refers to the first page of paper
Downloaded from ascelibrary.org by University of California, San Diego on 02/25/16. Copyright ASCE. For personal use only; all rights reserved.

Alampalli, Sreenivas, 57 Siddharthan, Raj V., 100

Steedman, R. S., 38
Crespellani, Teresa, 124
Vannucchi , Giovanni, 124
Elgamal, Ahmed-W., 57 Veletsos, A. S., 19
El-Gamal, Mahmoud , 100
Wu , Guoxi, 1
Finn, W. D . Liam, 1 Wu, Yingwei, 21
Fishman, K. L. , 77
Younan, A. H., 19
Madiai, Claudia, 124
Zeng, X., 38
Prakash, Shamsher, 21

Richards , R. , Jr. , 77

136
@seismicisolation
@seismicisolation