a

MISCELLANEOUS PAPER GL-85-1

SEISMIC DESIGN OF GRAVITY

of EngineerRETAINING
by

WALLS
Robert V. Whitman, Samson Liao Department of Civil Engineering Massachusetts Institute of Technology 77 Massachusetts Avenue Cambridge, Massachusetts 02139

Lf

(%J

.2g
- .5g

5g

N
..

DTIG ELECTE
~O
2 gAPR

12 985

0.5

15q

Al3
Approved for Public Release. Distribution Unlimited

DE

0.259

~2g
XO.39 Em.

US LLS
Prepared for

Anuay

1985s

AfEgnes.---

DEPARTMENT OF THE ARMY US Army Corps of Engineers

Washington, DC 20314-1000 _

Under

CWIS Work Units 31173 and 31589 39180-0631

.... ;2 .

Geotechnical Laboratory US Army Engineer Waterways Experiment Station

Monitored by

P0 Box 631, Vicksburg, Mississippi

. ... .._._ _'. ,. ._._.,,_,-

, . _._..:. _-, -: , . ,, -, , . .. . -, . .. _ ,, . -,_', - -.. ',.-,.,,,., ,_,._,._.,. '_,l: _

-..

Destroy this report when no longer needed. Do not return it to the originator.

o .

The findings in this report are not to be construed as an official

Department of the Army position unless so designated by other authorized documents.

The contents of this report are not to be used for advertising, publication, or promotional purposes. Citation of trade names does not constitute an official endorsement or approval of the use of
such commercial products.

...

..........

..

.,

Unclassified
SECURITY CLASSIFICATION OF THIS PAGE (*%hen

Data Entered)

.__

1.REPORT NUMBER
4.

REPORT

DOCUMENTATION PAGE

BEFORE AALOG NUMBER IEN 2.GOYJ ACCESSION No.3. REcIP COMPLETING FORM

READ INSTRUCTIONS

Miscellaneous Paper GL-85-1

TITLE (and Sublitt)

/lZ/
5. TYPE OF REPORT & PERIOD COVERED

SEISMIC DESIGN OF GRAVITY

Final report
6. PERFORMING ORG REPORT NUMBER
8. CONTRACT OR GRANT NUMBER(.)

RETAINING WALLS
7. AUTHOR(e)

Robert V. Whitman,
9.

Samson Liao
10. PROGRAM ELEMENT. PROJECT, TASK AREA & WORK UNIT NUMBERS

PERFORMING ORGANIZATION NAME AND ADDRESS

Department of Civil Engineering Massachusetts Institute of Technology

Cambridge,
11.

CWIS Work Units 31173 and 31589

12. REPORT DATE

Massachusetts

02139

CONTROLLING OFFICE NAME

AND ADDRESS

DEPARTMENT OF THE ARMY

January
13.
15.

1985

US Army Corps of Engineers Washington, DC 20314-1000

14. MONITORING AGENCY NAME & ADORESS(II different from Controlling Office)

NUMBER OF PAGES

156
SECURITY CLASS. (of this report)

US Army Engineer PO Box 631,

16.

Waterways Experiment Mississippi

Station

Geotechnical Laboratory
Vicksburg, 39180-0631

Unclassified
aIS. DECLASSIFICATION/DOWNGRADING SCHEDULE

DISTRIBUTION STATEMENT (of this Report)

Approved for public release; distribution unlimited.

1. DISTRIBUTION STATEMENT (of the abstract entered InBlock 20, Itdifferent from Report)
.

"'{
.1

IS. SUPPLEMENTARY NOTES

Available from National Technical Information Service, Springfield, Virginia 22161.

5285 Port Royal Road,

19.

KEY WORDS (Conlinue on reverse side if necesary

tnd Identify by block number)

Retaining

walls--design retaining

and construction wall design method

(LC)' (WES).

Richards-Elms

20.

ADSTIR ACT (

s ss

reverse as

If necessary and fdeitfy by block

bor

--- backfill

The report discusses the seismic design of gravity walls retaining granular
without pore water. The general features of behavior are illustrated

by field experiences, results from laboratory model tests and from theoretical analyses. Both the conventional method of design and the Richards-Elms method, A shortcoming of the n1c analogy to a sliding block, are reviewed. based upon sliding block analogy is discussed, and corrections obtained using a two-block model are presented. Several sources of uncertainty are examined in detail: -

(continued)

DD iX AtN 74T1473 IjR fM

EDITION

.vsSmorztnclojsified Tnlii OF I OV 6SIS OSSOLEE Ee

SECURITY

Umoo
CLASSIFICATION OF THIS PA ,F INWen Vats Fterrd)

...

--.

" i- . i

'

- . .

...

. ....

7..

Unclassified
SECURITY CLASSIFICATION OF TNIS PAGE(fWan Data Enterd)
I-

"

20.

ABSTRACT (Continued)

"he random nature of ground motions, uncertainty in resistance parameters, and model errors, including the important influence of deformability in the backfill. All of these results are then combined to develop a probabilistic method for predicting seismically induced displacements of walls and an improved The risk that walls designed version of the Richards-Elms method of design. by the conventional method might experience excessive displacements is .-. " - " .' ,,, analyzed.

Unlssf-

-0 L'

- .-

~.0
Unclassified"]SECURtITY CLASSIFICATION OF THIS PAGE(WI,.n 0.,. Elt.,.d) ""'

]"I

.'..-

. -.-.

---

--

,.-.

'.-.

. .

.'.-.".

."......

"

..

..-.--

..".-..";i'-',----"

, * -

..

PREFACE

This

report

was

prepared by

Professor

Robert

V.

Whitman

and

Mr. setts

Samson Liao of the Department of Civil Engineering, Institute of Technology, Cambridge, Massachusetts,

Massachuunder

Purchase Order No.

Waterways

DACW39-83-M-2088
Station (WES).

with the US Army Engineer

The work was conducted during

Experiment

the period April-December Office,

tion

1983 and was jointly (OCE),

sponsored by the InvestigaCWIS Work

Chief of Engineers

US Army Civil Works

Software,"

Studies,

"Computer-Aided

Engineering

Unit

31589,

and "Special 31173. Lucian G.

P. F.

Studies for Civil Work Soils Problems, respective OCE Technical Monitors and Richard F.
Chief,

CWIS Work Unit were Messrs.

tively. Dr.

The

Guthrie

Davidson,
Geotechnical

respec-

Hadala,

Assistant

Laboratory
under the

(GL)

was the WES Technical Monitor.

direction of Dr. W. F.

The work was

III, Chief, GL.

general

Marcuson

COL Tilford C.
the Commanders Mr. F. and R.

Creel,
Directors

CE,

and COL Robert

of WES during Director.

C.

Lee,

CE,

were

the period

of this

study.

Brown was

Technical

\AurcCZiofl
UTIS
.,' . ,:

For
""iTn-'in iorn
-.

GFA&T
"

.Vv .

,;t

', "

Codes --

o.r

-.

.10

.................................................

".....

-.........

..

ii
0

TABLE OF CONTENTS Page 1. 2. INTRODUCTION GENERAL FEATURES OF DYNAMIC BEHAVIOR 2.1 2.2 2.3 2.4 3. Complex Behavior and Simplified Models Field Observations Model Experiments Finite Element Results DESIGN 1 4 4
0

5 7 14 20 20 23 23 27 29 29 30 33 33 35 36
36
.

CONVENTIONAL 3.1 3.2

General Concepts Evaluating Dynamic Earth Pressure 3.2.1 3.2.2 Mononobe-Okabe Equation

"-

Validity of the Mononobe-Okabe Equation

3.3

Discussion of the Seismic Coefficient Method 3.3.1 3.3.2 3.3.3 3.3.4 3.3.5 Format of Typical Seismic Coefficients

Comparison of Two Seismic Coefficient Maps Judgement in Formulation and Use

Seismic Coefficients and Safety Factors Conclusion on Seismic Coefficients

4.

RICHARDS-ELMS METHOD 4.1 4.2 4.3 4.4 4.5

General

Newmark's sliding Block Model Evaluating Retaining Wall Displacements

36 43 46 47
. ."ii -.. "'. ... " i. . '. - i ,i. ":-

Richards-Elms Design Procedure Comments on the Richards-Elms Method

i . ....- :? l- i -: . . .i . . .. .' --. - .. -

i'. l' 2-'l~ --i

.. ii

i '

5.

KINEMATIC CONSTRAINTS UPON MOTION OF BACKFILL 5.1 5.2 5.3 5.4 5.5 The Two-Block Model Comparison with Single-Block Model Numerical Results Comparison with Experimental Results Summary

52 52 54 58 61 65 66 66 67 70 72 74 78 82 85 86 87
87
-

6.

RANDOM NATURE OF GROUND MOTIONS 6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 6.9 Introduction Scaling of Records Orientation Effects Scatter Among Different Sites and Events Effect of Vertical Accelerations Combined Uncertainty Predicting Residual Displacements Application to Retaining Walls Improvements to Predictions

7.

UNCERTAINTY IN RESISTANCE PARAMETERS

7.1 7.2 7.3 7.4 Introduction Uncertainty in Friction Angle

88 89 94

Block on Horizontal Retaining Walls

Plane

8.

MODEL ERRORS AND UNCERTAINTIES 8.1 8.2

8.3

100 100 105

i11
. . . .

Failure Plane Inclination Elastic Backfill Effects

Tilting

iv

Page 8.4 9. Approximations to 2-Block Analysis 115 118 118 118 125 130 131 Other Design 131 132 133 135 141 141 142 .
S

IMPROVED APPROACH TO DESIGN 9.1 9.2 9.3 Review of Objectives Equation for Predicting Motions Approach to Design Using Safety Factor Against Displacement 9.3.1 9.3.2 9.4 Choice of safety factor Examples

Reliability Implicit in Approaches 9.4.1 9.4.2

Conventional design Design following Richards-Elms

9.5

General Discussion

10. CONCLUSIONS AND OPPORTUNITIES 10.1 10.2 Conclusions Opportunities

REFERENCES LIST OF SYMBOLS APPENDIX A

144 147 151

0-;<.

_S

'9:
":-. i . .- .:i -. 2 - 2 " -2 . i-..-2 2 2- . ' - .- . .. .i - . -' - - . . ... .- - . ., - " .i " - .. . ' i

INTRODUCTION

The design of gravity retaining walls in environment is

an earthquake-prone

usually based upon static analysis using an This can be a suitable approach, is determined from a the use of

equivalent seismic coefficient.

provided that the seismic coefficient

rational analysis of actual dynamic behavior.,However, seismic coefficients in current practice is

largely empirical and

sometimes inconsistent,

leading to designs that may be either

excessively conservative or unsafe. In 1969, Richards and Elms presented a rational method for based upon the This approach is

the selection of a suitable seismic coefficient, concept of an allowable permanent displacement.

generally compatible both with the design philosophy used to design gravity retaining walls against static loads and with that used to design many other structures against earthquake loads. Richards and Elms utilized an analogy between the behavior of a gravity retaining wall and that of a block sliding on a plane,
which is an oversimplification of the actual behavior of a wall-

backfill system.

Consequently,

they suggested the use of a

liberal safety factor,

which to some extent takes into account the and other uncertainties in

effects of these oversimplifications the analysis. The work described in

this report improves upon and extends

the Richards-Elms approach to design by considering corrections to the simple sliding block analogy,
basis for the selection of a suitable

and by introducing a rational

safety factor for use in the

...

.-

approach. expression

The essence of the proposed method is

the following experienced

for prediction of the residual displacement wall during an earthquake:

by a gravity retaining

d Rw

Rv

R2/1

Q * R

(1.)

where d Rw

is

the predicted

residual displacement

Rv

is

the mean

(expected)

residual displacement

for a by

sliding block exposed to ground motion characterized a small number of parameters A and peak velocity V).

(such as peak acceleration

R2/1

is

a deterministic term accounting

for a specific

kinematic deficiency in

the single sliding block model.

is

a term accounting

for the unpredictable details in

shaking.

the random nature of future earthquake

is

a term accounting

for the uncertainty in

the backfill,

the

parameters characterizing foundation soil.

wall and

is

a term accounting for other,

and as yet,

poorly 0

understood deficiencies of the simple sliding block model.

-ot-9

The scope of this report is

restricted to gravity retaining where soil

walls with granular backfills subjected to earthquakes, liquefaction is not of importance. It

also primarily deals with treating 2 presents

the translational mode of retaining wall movements, rotational movements as a secondary concern.

Chapter

an overview of the complex nature of the dynamic retaining wall problem, design. and Chapter 3 discusses the conventional approach to Subsequent chapters terms in treat and discuss each of the detail.

individual It

Equation 1.1 in

should be noted that further research and development

remains to be done to render the basic Richards-Elms procedure completely satisfactory. knowledge summarized in is presented in Nevertheless, this report, based on the present

an improved design procedure

Chapter 9.

S< .i.

".0

2-

GENERAL

FEATURES OF

DYNAMIC

BEHAVIOR

2.1

COMPLEX Analysis

BEHAVIOR AND SIMPLIFIED MODELS of the behavior of is a gravity retaining walls interaction and such large as strains. the finite to branches and during problem

earthquake potentially Even with element simulate of

loading

complex soil-structure deformations procedures, feasible would occur.

involving the use of it is

plastic numerical not

method, all

presently

nor possible As in all

the phenomenon that simplified necessary

engineering,

models with to make

various approximations problems more tractable,

assumptions are

complex

particularly for purposes of design.

Various retaining walls, simplified will models, be presented useful in for engineering chapters design of of this

subsequent

report.

In this chapter,

the intent

is

to illustrate and examine and to highlight some

the complexities of retaining wall behavior,

of the major aspects of the problem that have been considered in the simplified models. However, more importantly, the phenomena

that have not been considered in identified, models.

the simplified models are also

to provide a basis for judging the limitations of the

A general overview of retaining wall behavior here, based on a review of field observations,

is

presented

laboratory model finite

experiments, element model.

and the results of a relatively sophisticated

Further aspects of some particular details of as necessary, in

these observations and data will be discussed, subsequent chapters.

18

the earth pressure connecting wedge. If

thrust

is

the force in

the spring

(labeled

'A')

the two masses representing

the wall

and the backfill

the ground is in Fig.

suddenly

accelerated

to the right and slip the spring 'A'

occurs as because

2.7(c), in in

a force would develop in the inertia the stiffness in forces of

of differences

between the two

masses and differences However, force slightly is is in there would the spring

the shear springs. 'B', and so the

be no force 'A'

the spring small,

would be fairly

and perhaps even

tensile.

On the other hand, left as in Fig.

when a sudden acceleration the force in spring 'B' As a force

applied to the activated slip

2.7(d),

to resist

the inertia

forces of the two masses. but a relatively large

result,

movements do not occur, in spring 'A',

would be present

the analog of the earth pressure. to explain in only a purely

The above arguments intuitive in

have tried

fashion the complex nature of forces and displacements retaining wall model. However, the major is that

a elastic-plastic

point of emphasis as it there is

applies to gravity wall design correlation of relative between

not a clear direct

the maximum earth that

pressure

force and the amount

displacement

occurs.

Focusing too much upon the forces exerted by the backfill and it is more essential think in

may lead to meaningless results, terms of displacements in design.

The location of the resultant

dynamic earth pressure

force

was observed to vary with time in a complex manner in element model. that shown in

the finite (than

Parametric studies using a finer mesh model Fig. 2.5),

indicated that the time variation of the

17

M-S

(a) Elastic-Plastic

Retaining Wall

BS

Shear Springs 71171n711771111ii17-Y ,

Axial Compressive Spring only (no tension)

(b) Idealized Lumped Mass System

A
Tr
Slip.',__ l . -

-Ground

Acceleration

(C) Occurrence of Slip

A B 0

--

Ground Acceleration

(d) No Slip, Elastic Deformations Only

.O

FIG.

2.7

AN INTUITIVE MODEL OF ELASTIC-PLASTIC WALL BEHAVIOR.

RETAINING

S. . .

. . . . .. .".. . ...

.".-".. .

.m '.:-'

'. ,.;.".".. ." m .r,,.l. '-' . -'.

-.

".".".. "

"

"

.-.

."."".".

"..

16

'i

-Slip

*Slip

-*Slip

~360

'520~280
5

240 200 0.5

-.-

1.0

1.5

2.0

2.5

3.0

i3 2 0 **
z280~240-

-Sloo ip
0 0.5 1.0

Slip 1.5

meSlip -s
2.0 2.5 3.0
*-

50 - -Slip.Z
wd

Slip

401-

301
-0J

U)

0.5

1.0 1.5 2.0 CYCLE NUMBER

2.5

3.0

FIG.

2.6

TYPICAL RESULTS FROM FINITE ELEMENT MODEL FOR RETAINING WALL SUB3JECTED TO 3 CYCLES OF SINUSOIDAL GROUND MOTION.

15

.8 m -

.9

/
/

Backfill properties: :30o c =0

H=8m

S..Pc=v .-" /.
2400 k

,otential
failure plane

= 0.3

2000 kg/mri 3
0.43

0= 54.67"

Ko:: 0

(a)

Contact elements with Cn=iE8kN/m/m, Slip elements with Cn=IE~kN/mlm,

Cs=O Cs lE5kNImlm 5

Slip element with Cn =1 E12 kN I/m/m. Cs =IE 5 kNI mlm4m

Rigid boundary Scale; 4m

Cn= Normal stiffness of slip elements CS: Shear stiffness of slip elements -0

(b

FIG.

2.5

FINITE ELEMENT MODEL OF A RETAINING WALL (AFTER NADIM, 1982).

0 .. .

14

2.4

FINITE

ELEMENT RESULTS finite element analyses are

Though even sophisticated simplified models of actual

behavior,

they can nevertheless offer

insight into physical processes that are difficult to observe or

measure experimentally. retaining wall used in 2.5. except The finite element idealization of a is shown in Fig.

a study by Nadim

(1982)

The material properties of this for the essentially at the wall-soil rigid-plastic interface,

model are

linearly elastic,

elements at the base of and along a preselected

the wall,

failure plane through the backfill.

for elastic development deformation

This model

is

able to account

of the backfill as well as for the wedge.

of a Coulomb-type failure

A typical set of results, obtzined using 3 cycles of sinusoidal motion at the base of the grid, There are three intervals is shown in Fig. 2.6.

(marked "slip" on the figure) during

In these intervals, the
S

which the wall slides upon the base.

shear force at the base of the wall is betwec , backfill and wall is
the maximum thrusts

constant and the thrust On the other hand,

times when no slip
S

relatively low.

from the backfill occur at

is

occurring,

and when the base shear resistance is

fairly low.
-

An explanation for the lack of direct correlation between the earth pressure
time) is illustrated

"

force and the amount of wall slippage

in Fig. 2.7. Here, the finite

(at any given

element model

is

further idealized as a lumped mass system consisting of two A unique feature 'B') is its ability

masses and axial and lateral-shear springs.

imagined for one of the axial springs (labeled to transmit compressive forces, but not tension.

The analog of

.. .. .. .. . .. ..

..

..

. ..

. ,= .

-72

. .

..

..

13

0.60 SlipCP

-Slipj
Base

a4

Acceleration
"A Wall"

z 0.20

0.2

Acceleration

w0 w0

's
0 0.08 0.16 0.24 0.32 0.40 0.48 0.56

I~ Ad
0.64

-0.20,

TIME in seconds

z46 w w
~i a. (0

2
00
p-I , p i..1 I p - I '

0.08

0.16

0.24

0.32

0.40

0.48

0.56

0.64

TIME in seconds

FIG.

2.4

TYPICAL MEASUREMENTS OF ACCELERATION AND RELATIVE DISPLACEMENT - IN EXPERIMENT PERFORMED BY LAI (1979). (Figure based on reduced data from Jacobsen, 1980).

.-.

..

12

Typical measurements of acceleration and relative displacement of the model wall obtained by Lai
2.4. model The first feature to note is

(1979)

are shown in
slip

Fig.

that the total

of the (i.e.,

retaining wall does not occur in but rather occurs in

a single movement

catastrophically), series of smaller

a stepwise fashion as a During the time

incremental displacements. is not occurring,

intervals when slip

the acceleration of the

model wall follows closely the input base acceleration. occurrence of slip is

The

associated with wall accelerations that are

and there appears to be a

less than the peak base motions,

critical acceleration at which slip starts to occur.

occurs in only the direction away from the backfill, to resist

Also,

slip

implying that wall

passive pressures are more than sufficient

movements into the backfill. A problem common to all model tests is in the lack of similitude In part, .

stresses and loads from using small scale models.

this scaling problem can be alleviated by conducting tests in a centrifuge, as has been reported by Ortiz, in et al. (1981) and by

Bolton and Steedman (1982) walls. Currently, there is

their experiments on model retaining

an extensive research program on

retaining walls being carried out at the Cambridge University

Centrifuge facility, but the results of those model tests are not

available to the writers at the time of publication of this report. It is almost certain that these series of tests will and may

provide further insight into retaining wall behavior, change some of the conclusions stated here.

"

"

'

..

. . . ... .

. .

FIG. 2.3

MODEL TEST SHOWING TRANSLATIONAL MODE OF FAILURE (FROM LAI, 1979). Note: Scale is marked in cm.

10

The formation of a single predominant backfill, along which slip occurs.

failure plane in

the

* *

A significant amount of backfill settlement. The occurrence of distortional backfill failure wedge. the area near the shear strains in the

The plastic deformation of the soil in

toe of the wall as a result of wall rotation.

Lai

(1979)

performed a series of experiments using L-shaped 12.6 inches (320 mm) high with a

model retaining walls approx.

base width of 8.7 inches (220 mm). aluminum,

The model walls were made of

and additional steel plates could be secured to the base The dynamic

of the wall to vary the total weight of the wall.

excitation was provided by a shaking table which could simulate both periodic and earthquake excitations. A photograph of one of Lai's model tests after failure is shown in Fig. 2.3. In contrast to the rotational failures, distortional strain in

translational movements produce very little the failing backfill soil wedge, rigid body motion. However,

and can be approximated as a

plastic deformations of the soil at the process of movement. failure, Lai noted the

the toe of the soil wedge must occur in

Similar to the result shown for rotational formation of a single predominant planes developed in Also, there is

failure plane,

though other

the failing wedge with larger movements.

clear evidence that increasing the weight of the inclinations of the failure plane.

wall leads to flatter

- .] , [. . [ [[ ..-[[ i .ii.. i .. ..

/ . .. . i

[ [.-iii.I i [

i.i...iI, ..

i " . -. .-.. 0

(a) Before startinq test

of gravity wall.

(b) Condition of gravity wall after l minute

of

vibration.

(c) Condition of gravity wall after 2- minutes 2 of vibration.

FIG. 2.2

MODEL TEST SHOWING ROTATIONAL MODE OF FAILURE (FROM MURPHY, 1960). Note: Scale is marked in inches.

2-

"(1970), and more recently by Nadim (1982).

Generally, the

experiments that have been performed can be classified into two groups: "* Experiments primarily concerned with the measurement of dynamic earth pressures and/or structural response. "* Experiments to measure the movements of retaining wall and observe general failure patterns during shaking.S The first group, involving experiments to measure dynamic earth pressure, has had limited success in the comparison of results with theoretical solutions, in particular the Mononobe-Okabe equation. Specific details of these comparisons It has generally been

with theory will be discussed in Chapter 3.

observed that the distribution of earth pressure does not increase linearly with depth as in the case for static pressures. Also,

the location of the resultant of the total force is usually located above the lower third point along the height of the wall. Several of these experiments involved model walls that were fixed or restrained, and subsequently did not correspond to true field
conditions. The second group of experiments are generally closer in their simulation of actual field conditions.
. . .-...-_

Murphy (1960) conducted

tests on a model gravity retaining wall made of solid rubber shaken with sinusoidal. base motion with a period of 1.48 seconds. ...-. .,... Figure 2.2 shows the sequence of failure of the retaining wall. Although the wall weight is improperly scaled and there are undoubtedly frictional effects (i.e. model against glass container), several significant behavioral features can be noted:
S
.-.

by Nadim (1980),

but the results are of a preliminary nature.

Settlements of the backfill behind a wall generally accompany outward movements of the wall. Evans (1971) reports fill height. settle0

ments of the order of 10% to 12% of the fill

Such orders

of magnitude of downwards movements of the backfill associated with outwards movement of the wall are consistent with the concept of the development of a wedge of soil failing along a plane behind the wall. It has been observed that movements are not always associated Evans (1971) noted that of the 39 bridges in New S

with damage or failure. examined in Zealand,

the vicinity of the 1968 Inangahua earthquake

23 showed measureable movement (without damage),

This is an important concept which is

and only
used by
S

15 were damaged.

Richards and Elms (1979)

in

the proposed design method. is a

Designing to limit the amount of outward movement

rational method to avoid failure not only for the translational mode, but perhaps also for rotational modes of failure. if the amount of translation is
by the bridge
S

In the

case of bridge abutments,

restricted so that contact is avoided,

and restraint

superstructure

then the likelihood of rotational

movements about the top of the retaining wall is

greatly reduced.

2.3

MODEL EXPERIMENTS

Tests have been performed by a number of investigators using

small scale models of earth retaining structures subjected to

dynamic base motion.

Reviews and summaries of the various results

from these experiments have been reported by Seed and Whitman

S..

.... ...

. ,.

....

n.

*-

..

..

,-/

i-

.-/
I I

I
/

(a) Outward Translation

;iI

(b)

Rotation about the Base

s-

lIliA\\-

"
-/

~//
/' I

:i::
,'. '

~(c) Rotation about the Top

.
--

FIG.

2.1

POSSIBLE MOVEMENTS OF GRAVITY

"p

RETAINING WALLS.

"j

o..i:

iL-. ... ;.' i.'-'. -.

i.1..-i'..

."i'.

/ '..

.i--, .:.ii

.i---<

-'-...'-'

..-

:...

. ..

< .. i- -

2.2

FIELD OBSERVATIONS Many reports of retaining wall movements during earthquakes

are available in

the literature.

Useful summaries of these data the Japan Society (1981).

have been presented by Seed and Whitman (1970), of Civil Enginee-s - JSCE (1977),

and by Mayes and Sharpe

Aside from the cases where liquefaction was a cause of failure, observed, * * * three types of retaining wall movements have been as schematically illustrated in Fig. 2.1. These are:

Outward translations of the wall Rotations about the base of the wall Rotations about the top of the wall

Most cases of movement involve a combination of translation and rotation. Rotations about the top of the wall appear to be

restricted to retaining walls forming part of bridge abutment structures. Mayes and Sharpe (1981) suggest that rotation about

the top occurs only after outward motion of the wall brings the *" top into contact with and restraint by the superstructure. However, the pattern of overall bridge movement in some cases

indicate that inertia forces from the superstructure may actually havepushed the top of the wall into the backfill In (Evans, 1971). chapters, in analysis.

the simplified methods presented in subsequent is considered

only the translational mode of movement This is

because translational movements are more analytically but it is is also an obvious

tractable than rotational movements, disadvantage . of the methods, since it

rare that purely Some

translational movements of walls have been observed.

"

analytical work on rotational modes of movement has been reported

...........................................

19

location of the earth pressure depended on factors such as the elastic modulus of the backfill and the frequency and amplitude of th( input ground motion. Another feature observed in the finite element model is the

amplification of ground motions due to the elastic properties of the wall and backfill. Since the system is elastic, a natural

frequency of vibration can be associated with the retaining wall and the soil. If the input ground motion due to an earthquake has effects similar to

a central frequency near the natural frequency,

resonance will tend to amplify the maximum ground acceleration, and cause larger displacements of the retaining wall.

**I

I4U

iIP

UL

LIE.

*I!*

U0

*-

-I

20

3-

CONVENTIONAL DESIGN

3.1

GENERAL CONCEPTS Gravity retaining walls are typically designed using a static

equivalent earthquake evaluate the static backfill,

coefficient.

This coefficient

is

used to

plus dynamic

force exerted on the wall by the the inertia shown in force Fig.

and should also be used to calculate A schematic of these forces is

due to the wall. 3.1,

along with definitions Having

and notation for the seismic forces, conventional static

coefficients.

found these

design procedures

are followed,

which nfeans ensuring that the

weight of the wall, passive

shear resistance on the base of the wall and (with appropriate and bearing

resistance at the toe are sufficient to resist sliding,

safety factors) capacity In failure. concept,

overturning

the use of a seismic coefficient is equivalent

(also called the to a static

pseudo-static method of analysis),

tilting

of the problem at an angle iA computed as:

NS
=tan-

NH 1NV

(3-1)

where NH = the horizontal seismic coefficient NV = the vertical seismic coefficient

with positive (+) *

inertia force directions as noted in

Fig.

3.1.

Figure 3.2 illustrates this concept for a simple case where Nv = 0.

.-..,- "?i.-'iil'i-. - ..l- .. . . . i .; ii i~- 1". .. ? ~ : -? ; . il . .~li i; < .; .ii i .' . -,i .i .i-Vi . .. -;1 .. [ . .. .;-i

21

iN WT

" H

F ailur'e
Plane

.il'

I -Nv

::

-FW
.W
=

NvW

Weight of Retaining Wall .0

Ws = Weight of Soil Backfill Wedge NH

*

Horizontal Seismic Coefficient

Nv = Vertical Seismic Coefficient

FIG.

3.1

SCHEMATIC OF STATIC EQUIVALENT SEISMIC COEFFICIENTS AND EARTHQUAKE INERTIA FORCES FOR CONVENTIONAL DESIGN.

_0

...-.

...-. .-. .

'.:

--.

'.

.-.

'.

-.

.. :.,-,.....'

-,

.......-..

.. -

..-.-... . .....

'.....

-....

22

90 A."

(a)

Retaining Wall with Gravitational and Horizontal Seimic Force Vectors

V.

-S

(b) Equivalent Static Problem with Gravitational Force Only

~D

FIG. 3.2

INTERPRETATION OF THE IP-ANGLE IN CONVENTIONAL SEISMIC COEFFICIENT ANALYSIS (NV 0).

23

As can be readily sec., interpretation of the angle,

logical conclusion from this that T can not exceed the angle

T is

of repose (0) for cohesionless flat backfill 1979).

(Richard and Elms, at the

Since the steepest slope that can be formed is T >

angle of repose,

would correspond to an impossible nonSimilarly, for a backfill inclined at

equilibrium condition.

angle i,

T would be restricted to have values less than 4k-i. as T increases, the critical angle of the failure 0-i, the failure plane becomes
S

Physically,

plane becomes flatter,

until at T =

parallel to the backfill slope.

3.2

EVALUATING DYNAMIC EARTH PRESSURE 3.2.1 Mononobe-Okabe Equation analyses can be

Although retaining wall seismic stability

performed by assuming several trial backfill (as in

slip planes of failure in the usually earth pressures equation the

.0

slope stability analysis),

are calculated using some version of the Mononobe-Okabe (Mononobe, equation is 1929 and Okabe, written as: 1926). In its complete form,

PAE

1/2

(1-Nv)

(3.2.a) _0

where

cos 2 (-M-8)
COS

A COS y cos 2 8 cos('+8+C)

[f

1 +

/sin(-+6) sin('_Y-i) cos(i-8) Cos(+8+i) (3.2.b)

.......................................... ...

..

...

..

....'

24

BACKFILL PROPERTIES

PAE

y*Unit Weight

4)uAngle

of internal

Friction

tail NH

FIG.

3. 3

DEFINITION OF PARAMETERS FOR MONONOBE-OKABE EQUATION.

25

AE

is

the combined active static in

and dynamic thrust,

and the

other quantities

the equation are:

y = unit weight of the backfill

H = height of the backfill
S=

angle of

internal friction

of

the backfill

6 = angle of friction between the backfill and the wall = angle of inclination of the back of the wall respect to vertical) i = angle of inclination of the backfill (with

The above variables are illustrated in

previously defined (Eqn. 3.1).

Fig.

3.3.

NV and T are as
-

Figure 3.4 provides various charts of the quantity K or AE KAE cos 6 plotted against the horizontal seismic coefficient NH. KAE cos 6 represents the horizontal component of the dynamic earth pressure. Fig. 3.4 illustrates the sensitivity of the Mononobethe various input parameters. Based 3.4
0

Okabe equation to changes in

on the observation that the inclination of the lines in Fig. are all approximately at the same slope (of about 3/4) relatively wide range of NH, t, and 6, Seed and Whitman
for KAE:
H

for a (1970)

proposed a useful approximate equation

AE A

KE

KA+

(3 / 4 )NH

(3.3)

where KA is

the static

earth pressure coefficient,

determined

using appropriate values of

*,

P, i,

and 6. "9

26

as*
0.4

-'
-

or-

/-

-000

-.

.2-d -lol

0.2-

3#

00.

-..

o0

0.

0.2

NH

0.3

0.4

0.5

0.30203

0.4

0.5

NH

o.i2

03

04

os%

oI

01

00J

;0.47._5

1-

"-" ".

O0
0 -4W.2 0

0.5
0. 0.1 0

NH 00
0

8.0

"-V

0.1

0.2

0.3

0.4

Q 5.

NH

FIG.

3.4

EFFECTS OF CHANGES IN VARIOUS PARAMETERS ON THE DYNAMIC EARTH PRESSURE COEFFICIENT KAE (AFTER SEED AND WHITMAN, 1970).

.-..

'.

27 The location of the total thrust PAE is

the Mononobe-Okabe analysis. Usually, it is

indeterminate
recommended that

from
the

resultant force be located above the lower third point of the wall. Seed and Whitman (1970) suggest that the dynamic component with the net result
thrust PAE would be

"-.

of PAE be placed at the upper third point,

being located that the combined dynamic of and static

at or near mid-height

the wall.

3.2.2

Validity

of Mononobe

Okobe

Equation

The Mononobe-Okobe equation is equation for active earth pressure,

nothing more than Coulomb's modified to incorporate a

horizontal inertia body force as well as a vertical gravitational body force. Indeed, as discussed previously, Equation 3.2 may be the

derived simply by starting from Coulomb's equation and tilting

wall and backfill until the resultant of all body forces is

vertical

(e.g.

see Antia,

1982).
.5

Equation 2.1 is

subject to all of the same limitations as the Failure lines through the backfill are which is an approximation but a good one.
that there be sufficient strain

static Coulomb equation. assumed to be straight,

Most important is the

requirement

along the assumed failure line to mobilize the full shearing resistance of the soil in
must of the then indeed be active is inertia plus

the active sense.

If throughout force stress is the the full

That is
shearing

to say,

there

conditions. realized body dynamic

resistance and if
.

the backfill horizontal the static

failure within

wedge, this

constant

wedge,

between backfill

and wall must

be distributed linearly with depth.

In many cases,

these may be

' . .. .''. ". .. "..' .. i-'.,': "- ".' ' " , - . " . -"-" ., ,

. - ". . -' , .

. ."

.. .

-, "-

" " - -

- -S

28 questionable assumptions, the deviation from which would lead to

quite different stress patterns. Other dynamic earth pressure equations have been suggested, usually derived on the assumption that the backfill is linearly

elastic with no limitation upon the shear stresses that can occur. A summary of these various solutions is Not surprisingly, thrusts, depth, presented by Nadim (1982).

such equations often predict much larger dynamic

and a different distribution of lateral stress with

than an analysis based upon Coulomb's assumptions. various experiments have been

As described in Chapter 2,

performed using shaking tables with the purpose of checking upon the validity of the Mononabe-Okabe equation. In general, the

conclusion has been that the observed total dynamic thrusts agree reasonably well with those predicted by the theory. However, many

of these tests have not satisfied conditions that permit sliding to occur along a failure plane through the backfill. the dynamic thrust varies during a cycle of loading, In addition, and it is not

clear which observed value should be compared to the MononobeOkabe value. Hence it is not surprising that there are experiments showing because of experimental conditions that On the "

disagreement with theory,

do not simulate the behavior of gravity retaining walls. other hand, at least some of the reported experimental Equation may be only

confirmation of the Mononobe-Okabe fortuitious.

-"' '-"'

":..

-: .

'-'

-"

:i

'

.".''.

..

"

..

. ""

"

I.

-.--.

I |I L

.,

..,

29
3.3 DISCUSSION OF THE SEISMIC COEFFICIENT METHOD

3.3.1

Format of Typical Seismic Coefficients earthquake coefficient is a

The use of a static equivalent

reasonable approach for the design of gravity retaining walls, provided that neither the backfill nor the foundation soils (i.e., in this

beneath the wall experiences a dramatic loss of strength liquefaction) during earthquake shaking. approach is in A key element

the proper selection of the seismic coefficient to use Equation. used for the design of most

the Mononobe-Okabe

The seismic coefficient method is civil engineering projects.

Building codes and other design which is

manuals provide recommended values for this coefficient, primarily dependent with respect most codes, dependent * *

on the geographical location of the project In

to regions as defined by seismic zoning maps. the coefficient is modified by factors that are

on:

The type of foundation soil profile at the project site The type of the structure (e.g. buildings vs. bridges)

The natural period of the structure * The importance of the structure warehouses) The last of these above factors, "importance factor", is often referred to as the (e.g. hospitals vs.

an attempt at incorporating a subjective

risk/benefit element

into the seismic coefficient. that have been developed Although -

The various maps and recommendations are strictly

for the horizontal seismic coefficients.

the general Mc-nonobe-Okabe equation can accommodate both vertical

-7 ..................................................................

30
and horizontal accelerations, for the vertical this factor in the present lack of recommendations
0

components of acceleration conventional is design. Also,

prevents considering the vertical component

of earthquake motion significance

generally not considered component.

to be of as much

as the horizontal

3.3.2

Comparison of Two Seismic Coefficient

Maps

Two examples of seismic coefficient maps of the United States are shown in Figures 3.5 and 3.6. Figure 3.5 is the seismic Army Corps of

coefficient Engineers -

zoning map currently USACE (1983),

used by the U.S. 3.6 is

and Fig.

from the tentative Council

building code proposed

by the Applied Technology

(ATC-3-06,

1978).

Similar maps for the United States are and in the ANSI

published in regulations.

the Uniform Building Code (UBC)

Comparison of the two maps in apparent differences in

Figures 3.5 and 3.6 indicate

the delimiting of seismic zones and the The ATC maps show coeffi-

magnitudes of seismic coefficients.

cients that are double the values shown on the USACE map. However, directly, the USACE coefficients are intended to be applied

while the ATC coefficients should be modified using For free-standing (Mayes and
-

various factors as previously described. gravity retaining walls, Sharpe, 1981) is

the current ATC recommendation where N is

to use NH = 1/2 No,

the value of the the final

seismic coefficient shown on Fig. comparison, first

3.6.

Thus in

the two maps do not conflict as significantly as at Nevertheless, differences do exist and it should be

glance.

S " '- -' -. '--.' .

- - - "

- ? .-'

.- ".-

'.- b 'i . .

'.' 'L - < - .

31

C4,)

~~44

E-c m

I:Z.

00 9U
E44

32

00

E-4

r:4 0

E-4
'14

ZrL

46

tan b 1 +

2Ww 2_

(4.5)

Once aT = Ng is approximate Eqn.

obtained by solving Eqn. 4.5, it is

4.4 or by using the a simple matter to use This 0

solution of Eqn.

4.1 to estimate the retaining wall displacement. illustrated in Example 4.2.

computation is

4.4

RICHARDS-ELMS

DESIGN PROCEDURE 0

The design of a gravity retaining wall essentially requires calculating the weight of the wall Ww, allowable displacements. This is given an imposed limit for

the inverse problem of solving The 0

for displacements discussed in

the previous section. as follows:

procedure proposed by Richards and Elms is 1. 2.

Decide upon an acceptable maximum displacement dR. Calculate N using Eqn. 4.1 in the form: .

2 0O 8 = 0.08 7 Ag

1/

A AR

(4.6)

3.

Use the Mononobe-Okabe equation

PAE" In doing so,

(Eqn.

3.2b)

to calculate
0, 6, and

the appropriate

values of 6,

i should be used.

4.

Calculate the required weight of the wall using Eqn. in the form:

4.4

-. . .. S

45 T FI + (PAE)H (4. 2a)

Making the appropriate substitutions,

we obtain

BN tan4 b
BNt n b-

WW

+ (PAE)H

aT
W

[w +AE-vI

+ (P AE)H

(4.2c)

Solving for aT:

aT

[Ww + (PAE)V 1

tan~b-

(PAE)H]

Ww

(4.3)

or

or or

N =tan b - (P aAE)H

SW
w

v tan4 b

(4.4)

Richards and Elms recommend using the Mononobe-Okabe equation (Eqn. cannot, 3.2) for evaluating PAE' and hence the above equation a non-linear

in general,

be solved explicitly since PAE is

function of aT (or N).

Iterative methods or an approximate LS Example 4.1 can be used to for PAE

graphical procedure as illustrated in solve the equation. (Eqn. 3.3) is If

the Seed-Whitman approximation 6 +

=

used and if

0,

then a simple explicit

expression for N can be obtained:

" .

"

.. . . . : , .. .

.. .-

.. .

..

-..

. .

- .

.:

..

VS

44

aS

0 T

7Note:

oT

Ng

-P.

Max a

=Ag

(a) Wall and Backfill Accelerations

(PAE)v

Rigid Block
~AE~H I%

calculated using T = Ng in MononobeOkabe Eqn.

(b) Free Body Diagram of Wall

FIG.

4.5

IDEALIZATION OF THE RETAINING WALL PROBLEM BY RICHARDS AND ELMS (1979).

43

corresponding displacements would be 0.35 in. Based on these results,

and 1.4 in. "

Richards and Elms proposed an

alternate and very convenient equation for calculating the block displacements d R in interest in design) the medium to low range of N/A (the range of as:

dR

0. 087 V2

41

where N and A are previously defined and V is velocity. This equation is also plotted in

the maximum ground 4.4 for

Fig.

comparison with the data and with Newmark's curves.

4.3

EVALUATING Although

RETAINING WALL DISPLACEMENTS 4.1 is based on the results of a sliding block

--

Eqn.

model originally and embankments, wall movements.

intended for use in it

predicting movements of dams retaining .

-"

can be easily applied to predict

The only difference

in application arises from a-.

slightly more complicated evaluation of the limiting acceleration

aT~ = Ng.
tancb.

For a block on a horizontal plane,

However,

N is simply equal to

additional vertical and horizontal earth pressure

=

forces,
(PAE)H
=

respectively denoted as (PAE)V cos(6 + 8) and shown in Fig.

PAE sin (6 + )

and

"i --

4.5, must be considered in

the equilibrium equations for the retaining wall. Summing forces body diagram in equilibrium: Fig. in the horizontal direction, using the free . e '-. ".

4.5(b)

and imposing the requirements of

'

............................................ ....................................................

42

-Various

Equations

Proposed by Newmark (1965) Richards- Elms Equation (1979)

>
-J
...................

AgA

...........

.....
................

(2pjaa)
..............

dipUcmet (10

.....................................

dslace~

N A A....

V.

0
zS

~2 Ag

0.11

0.01

0.10

1.0

10

N A
FIG. 4.4

TRANSMITTABLE BLOCK ACCELERATION MAXIMUM GROUND ACCELERATION

RANGE OF NORMALIZED DISPLACEMENTS USING NEWMARK SLIDING BLOCK MODEL, AND VARIOUS EQUATIONS APPROXIMATING THE UPPER ENVELOPE.

41

shown were obtained using several strong motion records from the San Fernando earthquake. in Note that there is considerable scatter as a result

the calculated displacement for each factor of N/A,

of differing characteristics The data in Fig.

of the various earthquake records.

4.3 were plotted using a "standardized" inputs to a

displacement scale,

obtained by scaling the earthquake

maximum acceleration Ag = 0.5g and a maximum velocity V = 30 in/sec. These same data can be replotted using a normalized

dimensionless displacement scale by dividing the calculated displacements by V2 /Ag. Figure 4.4 shows such a plot with the from all earthquakes used by Also shown are several 6

ranges of normalized displacements Franklin and Chang in

their analyses.

expressions suggested by Newmark giving conservative estimates for

the residual displacements, each most applicable for a different
.

"

range of N/A. upper bounds,

Note that while these expressions are not true they do form nearly an upper envelope for most of O-

the computed points. To illustrate the implications of these results, V2 /Ag typically ranges from 1 in. the quantity 0

(for moderate earthquake with a (for a major earthquake with a

peak acceleration of 0.2g) to 4 in. peak acceleration of 0.6g).

In many problems of interest the At N/A = 0.3, the normalized 6

value N/A ranges from 0.3 to 0.7. residual displacement falls in

the range from 0.4 to 10.0 so that

displacements during a moderate earthquake would be from 0.4 to 10 inches, 40 in. while those in a major shaking would range from 1.6 in. At N/A = 0.7, is 0.35, the upper envelope value of normalized so that for minor and major earthquakes, the
O
...... ..........-. .-. o".............." .... .J .-. ".". . ".............. .. . .

to

displacement

40

I00I

I S

I 1000 I I I

Ii

~I I I I II

--

SAN FERNANDO, CALIFORNIA EARTHQUAKE. 2/9/71

500 7_

SS

M=6.5 EPICENTRAL DISTANCE 22.4 2 TO 18SKm -

ISCALED
.0.5&
mEAN VALUE

34SOIL SITES TO A V 30 INJSEC

I-o
w
U

,,,

IL

0I

S --

100

NONSYMMETRICAL RESISTANCE
01b

0.01

0.05

0.10

%5

--

N .. A

.TransmittableI . Block Acceleration . . . . . . . . ..

Maximum Ground Acceleration

,_...

FIG. 4.3

EXAMPLE OF RESULTS USING NEWMARK SLIDING BLOCK

MODEL (AFTER FRANKLIN AND CHANG, 1977).

39

transmitted to the block through friction forces is the subscript acceleration. block is

aT = Ng where

'TI denotes the transmittable horizontal or limiting Then the consequent acceleration experienced by the Fig. 4.2(a).

shown by the dashed lines in

The resulting velocity profile as a function of time can be deduced as shown in Fig. 4.2(b). The plane's velocity increases

linearly at a slope Ag and levels off at time to, the end of the rectangular input pulse. However, the block continues to

accelerate until its plane (at time t acceleration

velocity catches up to the velocity of the

) and this limits the time interval of the The resulting simply in the

impulse experienced by the block.

relative displacement

between the block and the plane is Fig. 4.2(b), i.e. the difference

the shaded area shown in

integrals of plane and block velocities over time. The basic concepts described above can be applied to more complex earthquake acceleration time histories, simple computer program. included is using a relatively

An additional feature that must be i.e.,

the non-symmetric resistance of friction forces, This is

slip occurs only in one direction.

consistent with the

physical behavior of retaining walls in generally more than sufficient

that passive pressures are

to resist wall movements into the

backfill during earthquake shaking. An example of the type of results obtained by Newmark and later expanded by Franklin and Chang (1977) 4.3. This figure is is shown in (1965) Fig.

a plot of standardized residual block

displacements d

versus the ratio of transmittable block The data

acceleration to maximum ground acceleration aT/a = N/A.

.. .

E i

- -

38

ACCELER AT ION APlane Acceleration AgBlock Acceleration

NgL

/
to

~I

tm (a) Plane Velocity V Ag to

"Block Velocity

I iME

VELOCITY

Vb to tm (b)

Ng t TIME
0

FIG.

4.2

ACCELERATION AND VELOCITY PROFILES OF BLOCK AND PLANE SUBJECTED TO A RECTANGULAR PULSE EXCITATION.

S:: :!:

90

37

--i.OTr Ng
77117111111IIIIII

Max a a Ag
0

(a) Notation for Block and Plane Accelerations

W.

Fj, = Mo a

g-oT

T= BN ton
BN= W

b W ton~b

F, = Inertia Force W = Weight BN =Base Normal Force T = Shear Force (b) Free Body Diagram of Block 0

FIG.

4.1

NOTATION AND FORCES FOR SLIDING BLOCK ON A PLANE.

-0

. .

-F

. :

_. .

v. -

. -

. .

- . ,

"

,-

36

4-

RICHARDS-ELMS METHOD

4.1

GENERAL

Recognizing the shortcomings of the conventional approach for seismic design of gravity retaining walls, Richards and Elms (1979) developed a design philosophy based on the concept of an In the end, the design of a

allowable permanent displacement. wall is still

accomplished using an equivalent static seismic but with a more rational basis for the selection of
S

coefficient,

this coefficient. The key to the Richards-Elms approach is the method of The approach is to evaluate the

calculating the amount of residual wall movement. similar to the method suggested by Newmark (1965)

amount of slip occurring in dams and embankments during earthquakes.

section,

The Newmak sliding block model is

which is

discussed in the next

notation and to set
.

also intended to introduce

the stage for discussions of more complex models for evaluating retaining wall displacements.

4.2

NEWMARK'S SLIDING BLOCK MODEL Consider the rigid block shown in Fig. 4.1 with weight W and constant. It is

mass M = W/g,

where g is

the gravitational

assumed that the coefficient of friction between the block and the plane is -P tanOb Suppose that a rectangular earthquake 4.2(a) is applied to the equal to _

impulse (solid lines) shown in Fig. plane. Ag.

The magnitude of the plane's acceleration a is

Suppose also,

that the maximum acceleration which can be

35
3.3.5 CONCLUSION ON SEISMIC COEFFICIENTS O The conclusion that is is arrived at from the above discussion

that there are rational ways to select and use the conventional in design. However, the emphasis of design

seismic coefficient

should not concentrate

on the evaluation of equilibrium of forces, that In a recent @

but rather on the evaluation of the retaining wall slip should be allowed document to occur during a major earthquake. it is stated that:

issued by the USACE,

"... the seismic coefficient method, often referred to as the pseudo-static method, is no longer regarded as being appropriate for analysis of embankment or foundation response use for this Therefore its in seismic loading. purpose should be discontinued." 1983) (USACE,

The above statement should equally apply to gravity retaining walls.

-S ,

S.- -

-.S

-9.

..................................

~.**.**

...

.**-

.-.

*-*,.*..*"'""*

...

"..

-.

--

.-

.-.

--

tS

34 much higher than the seismic coefficient. Thus, it is recognized

that buildings designed using these recommended coefficients can be expected to yield, should a major earthquake occur. However,
-

these designs are such that the yielding should not cause unacceptable damage or danger of injuries and fatalities. The implication for gravity retaining walls designed using a seismic coefficient method is that slip of the of the wall will This is especially true

likely occur during major earthquakes. in

light of the fact that relatively low factors of safety are in conjunction with seismic design. The

usually recommended

design manual used by the Naval Facilities (NAVFAC, 1982)

Engineering Command

DM-7.2 currently allows a factor of safety between and for quay walls in Japan the 1977).

1.1 and 1.2 for seismic analysis, recommended

factor of safety against sliding is

1.0 (JSCE,

Although the USACE does not have specific factor of safety guidelines for retaining walls guidelines for dams (USACE, would be acceptable in (USACE, 1970) 1965), it is inferred from the

that a factor of safety of 1.0

earthquake design.

An alternative to designing retaining walls using the seismic coefficient would be to instead use the peak ground acceleration expected for a future earthquake. However, if It this practice is the inertia force of has also been suggested

considered to be generally uneconomical the wall is considered in the design.

that the horizontal earth pressure PAE be evaluated using the peak ground acceleration and that the inertia of the wall be ignored. However, this is an illogical procedure and cannot consistently

lead to sound designs.

..- .- ...... ..

,-.. . -,

./

.- i

.3 .,-

[ .- .

-i

..-

...---.-

. .<

. .-

,,

. ..

33 noted that for walls which are restrained from horizontal movement, NH = 1.5 N is recommended by the ATC code.

3.3.3

Judgement in

Formulation and Use in the various codes and

The recommended seismic coefficients

manuals are derived partly from theory and partly from experience data during actual earthquakes. Considerable judgement is

necessary to formulate the zoning maps and to determine suitable values of seismic coefficients. the differences in Thus it is not surprising that and

various codes and manuals should occur,

also that updating of the values of the seismic coefficients occur from time to time. It is also important to note that the intended use of the

various recommendations may significantly affect seismic coefficient values. * For example, the USACE maps were originally designing earth dams, which make Thus,

formulated primarily for use in

up a significant part of the USACE's constructed projects.

applying the USACE coefficients to other structures should be done * cautiously. In Japan, a similar situation exists with seismic

coefficients and maps differing for port and harbour structures, *roadways, buildings, etc. (JSCE, 1977). -I 3.3.4 Seismic Coefficients and Safety Factors typically have lower values than the

Seismic coefficients

peak ground accelerations that have occurred during earthquakes. In designing buildings, it is expected that the peak accelerations the structure) could be I

(due to amplification of ground motion in

47

(PAE)H

(PAE)V
-

tan b (4.7)

w w

tan~b

5.

Apply a factor of safety of 1.5 to the wall weight RW w.

A design problem, Example 4.3.

using the above procedure,

is

illustrated in

4.5

COMMENTS ON THE RICHARDS-ELMS

METHOD

The Richards-Elms procedure It is, in effect,

and Hall,

is

rational and simple to apply.

a counterpart of a procedure used for buildings

1982) where the ratio of design seismic

(Newmark

coefficient *.

is

chosen on the bas~s of the ductility ratio (of that a structure possesses before Its "

expected strain to yield strain) there is

extreme structural damage or danger of collapse. does not consider certain the

major disadvantages are that it *

kinematic restrictions upon retaining wall behavior, deformability of the backfill or possible tilting, statistical
fashion,

and the In a

variability of earthquake ground motions.

these factors have been taken into account in somewhat

the factor 4

of safety of conse-vative

1.5 on the wall weight,

which is

compared to usual values of recommended

safety

factors ranging from 1.0 to 1.2, However, it is

as discussed in

Section 3.3. -4

not clear that there is

a rational basis for the

suggested safety factor of 1.5 on wall weight.

"2 .. . . .. ..
-**.

,' .
.*. * .. .*

..

"

|0

48

The remainder of this report will consider some of the

deficiencies in the Richards-Elms procedure, and will suggest

improvements and corrections,

while retaining the essential

simplicity and soundness of the basic approach.

0 ,.

J'. 0

0-

..

".

,o

"

o '."

"

.J

..

..

49
EXAMPLE 4.1

Given: Find: Solution:

Retaining wall and backfill with properties shown in Figure E4.1. The maximum transmittable acceleration N, using the Richards-Elms method. The weight of wall Ww is calculated to be 32.81 K/ft. From Eq. 3.2a, PAE= (1/2)(0.120) (25) 2 KE = 37. 5 KAE K/ft. Assuming values of N, values for P and KAE are calculated from Eqs. 3.1 and 3.2b. A new value of N is then computed from Eq. 4.4. Results of these computations appear in Table E4.1 and are graphed in Fig. E4.2. The answer is given by the intersection of a curve through the computed points and a line
through the origin at 450.

.0

-*
4.

-2.5'
2.5
440

Bockfill Slope i =0

25'
'

,. q4

zo=

Bockfill Properties 300

'

tE120

C =0

PCF

. .

=CV 4 .% ,.
150

PCF-

15'

0-.0 '. FIG. E4.1

. .

50 EXAMPLE 4.1 (continued)

Table E4.1 AS SUMED

N

(KAE) 'P 2.860

5.710 8.570 H

COMPUTED
N

0.05
0.10 0.15

0.364
0.397 0.433

0.161
0.123 0.082

0.20

11.310

0.473

0.036

0.3-

SOLUTION
z0.2 N 0.I12

a.o0.280.1

Data from Calculations

Table E4.1 0 0.1 0.2 ASSUMED N FIGURE E4.2 0.3

For comparison,

N is

also computed using Eq.

N
=

4.5,

yielding

0.106

...

......... ....

51 EXAMPLE 4.2 Given: Find: The wall in Example 4.1.

A = 0.3 g's and V
=

The permanent displacement caused by an earthquake

characterized by 15 in/s,
-

using the Richards-Elms approach. Solution: From Eq. 4.1:

dR = 0.087 0.3'386

-0.112

= 0.087
= 8.7 in.

(1.94) (51.48)

EXAMPLE 4.3 Given: Find: The backfill and frictional resistance properties in Example 4.1. For a wall 25 feet high, the required weight of wall if an earthquake with A = 0.3 g's and V = 15 in/s is to cause a permanent displacement of 1 inch, according to the Richards-Elms approach.

Solution:
Step 1 d
=

1 inch 4.6, N
= =

Step 2 - From Eq. Step 3

-

0.192

Eq. Eq. Eq.

3.1 gives

10.890 1
= =

3.2b yields KAE 3.2a gives P 4.7, W AE

=

0.467 17.51 K/ft.

Step 4

From Eq.

45.44 K/ft.

Step 5

Applying a safety factor of 1.5 to computed Ww: Required weight of wall = 68.2 K/ft.

-0<
O

52

5-KINEMATIC CONSTRAINTS UPON MOTION OF BACKFILL

5.1

THE TWO BLOCK MODEL

-- '

"

In the Richards-Elms model,

the retaining wall is when in

modelled as

a single sliding block on a plane, behavior is much more complex.

fact the actual the

A more realistic model is (1979), which is

two-block model developed by Zarrabi schematically in Fig. 5.1.

shown represented (behind O

In this model the wall is

as a block on a horizontal plane, the wall)

and the wedge of soil

that "fails" during sliding is

represented by another

rigid block on an inclined plane. The kinematic constraints on the two-block model are that during sliding, contact force and acceleration continuity must be and between each of
.

maintained between the two blocks themselves,

the blocks and their respective sliding planes. This gives rise to three equations of acceleration continuity that must be satisfied simultaneously with the equations of equilibrium. The most significant constraint in the problem is terms of the mechanics of " .

that of maintaining contact between the sliding For outward movement of the

soil wedge and the inclined plane. wall to occur,

there must be a simultaneous outward and downward Thus, even when there is no vertical

movement of the soil wedge.

ground acceleration,

the backfill wedge would still

experience

vertical accelerations.

--

53

Kinematically constrained directions of relative slip

//
7
-.-

ilure Plane

Ground Acceleration

(a) Actual Physical Situation

Rigid Sliding

Plastic deformations Interaction of blocks through active force PAE (b) necessary for movement are ignored
C

Two Block Model Idealization by Zorrobi

FIG.

5.1

SCHEMATIC OF IDEALIZATION OF RETAINING WALL PROBLEM BY ZARRABI (1979).

S.

-. ..- ... .

..-,

..- ..

- ..

..

.-

....

. ..

..

: -- : .:

54

..:

..

5.2

COMPARISON WITH

SINGLE-BLOCK MODEL

Vertical accelerations

in

the backfill wedge affect the This is

'..

----.
. . -...

active earth pressure PAE between the wall and the soil reflected by the term and the factor (I-Nv)
Equation (Eqn. 3.2). It can be she::,,n that for

.-.- -. :,

in

the Mononobe
of

continuity

acceleration normal to the failure plane at any instant in the following equation must hold:

time,

.i

Nv(t)
or

= Av(t)

+ [AH(t)

- NH(t)I tan [@(t)]

(5.l-a)
.

NH(t)

= AN(t)

+ [Av(t)

-My(t)]

cot [@(t)]

(5

2-b)
/

" .
'" /

where AH(t) Av(t) NHIt)

is is is

the horizontal ground acceleration coefficient.

the vertical ground acceleration coefficient ""'"'"'" the transmittable horizontal acceleration.i2222211?'".'.'-.'",.:;..,:
w

coefficient of the wall and soil wedge.

Nvlt) is
81t)
The notation

the transmittable vertical acceleration coefficient of the soil wedge. is the angle of inclination of the failure plane with respect to horizontal (see Figure 5.1).
It) indicates the above quantities to be variable

[ ..:.

--?i-i':< i<.-'.

"-.'-"-"-.', "' "'"'""

" .... :-
.- ,

with time time is

This is

Thus,

the transmittable acceleration at any instant in upon the ground acceleration at the same time.
to the single-block model proposed by

S. ?.i-.-

dependent
in

sharp contrast

Richards and Elms where the transmittable acceleration is with time.

constant

'"'"--""
J ,- J ,0

.... ." 121?.i.i.."

is ?[''.i[[i[[

A schematic comparison of the sliding processes of the

Zarrabi (two-block) and Richards-Elms (single-block) models

shown in

Fig.

5.2

In

the two-block model there is

a threshold.

"--"-"7-

"?--"-'?-."?':. ,."-.'--A?--?--?-:..':--?--..,,.,,,,..,: ,.. ,.,,. ,.,-v.-.--..-.-.-.-.-..'...:. . ..-..-.. . v...v.-...:.-..-..-i.. i. ?.-i .---.>-i.i. . ..-.- . . :- ...-.. , .,_-. , '.,,...-- --. -. --.-" . -'-" ..... . . ... ... . . : .. .... ... . .. ,

55

-Ground

-.-

z
2
uJ NI

Richards - Elms (Single Block) Model Zorrabi (Two-Block)

model
N M

I.

TIME

oe'

TIME

I-

z
w~

>I00

> W

00

TIME

FIG.

5.2

SCHEMATIC COMPARISON OF SLIDING PROCESSES OF THE RICHARD-ELMS (R-E) AND ZARRABI MODELS.

--.

56

acceleration NT required to initiate

slip.

Provided

that

comparable assumptions are made concerning the properties of the backfill, the value of NT is exactly the same as the value of N However, after initiation
0

used for the Richards-Elms procedure.

of slip, the limiting acceleration NH at any time during a cycle of slip can be greater or less than NTg. As a result, sliding. in Fig. the active thrust P

AE

is

also changing during shown

An illustration of how this physically occurs is 5.3 for the case where there is (AV =0). no vertical ground

acceleration

When the ground acceleration AHg exceeds [Fig. in 5.3(a)], the vertical

the transmittable acceleration NHg backfill wedge acceleration Nvg is Hence, the inertia force is in

the downward direction. 70

the opposite upward direction, the weight of the soil and a

effectively causing a decrease in subsequent decrease in PAE"

The reverse occurs during the later Fig. 5.3(b)

0

stages of slip when AHg < NHg as shown in

The equations applicable to the evaluation of residual slip

in the two-block model was developed by Zarrabi (1979). The

solution procedure for these equations is that during slip, time-step. Also,

fairly complicated

in

the value of NH must be evaluated at every since e is a function of NH and NV, Wong (1982) the solution

for NH must be obtained iteratively. developed a more efficient scheme, to the governing equations is memory,

subsequently

in which part of the solution computer it might 9

precomputed and stored in Alternatively,

thus requiring fewer iterations.

be assumed that 0 remains fixed in which case the Mononabe-Okabe

equation no longer applies and the basic equations for dynamic

57

00
U)

1-4

(n
U)U

E-4

>w

z
0

0..Z
h.r

zz

\rU x

(0 24
o'x

-j E-4

A0

z
0

"wlo

+
-

in;in

CIO.

~44

58

equilibrium and continuity are solved simultaneously at each time step. There is one other feature of Zarrabi's two-block model that This is the implicit nonso that unlike the

deserves mention at this point.

symmetrical resistance of two-block model, Richards-Elms/Newmark model, non-symmetrical necessary.

no explicit assumptions regarding the

nature of the sliding block resistance are

5.3

NUMERICAL RESULTS The net result of the kinematic constraints in the two-block

model is

that the calculated residual displacements are smaller This is illustrated in

than those using the single-block model. Fig. 5.4, which shows the ratio R2 /
1

plotted against N/A (for the where R2 /

1

single-block model) is defined as:

or NT/A (for the two-block model),

2/1

dR dR

Residual displacement of two-block model Residual displacement of single-block model as used here, are not

Note also that the values of A, N and NT, functions of time,

but are constants depending on the earthquake

record or the wall/backfill properties. The results shown in Fig. 5.4 are based on limited results

using the average values of residual displacement calculated using four earthquakes (Antia, 1982). The unit weight of the soil, the

wall height and the height of the wall are properties that can be collectively described by the value of NT. S - ".-.-'- .. / . . l .' '. .i i . ] .. i i.

However,

the soil and

...5. - .

-, - . - - i- i

-- -

- -

59

0C

(0

fl
i

.0
01 0
Lii

cozZ

E-

2.

cn

0I

00

00

E 4E

z
0tt

.~

00

z
-7

LAo0

1- 40

I)

~~0

uc

.LN3VV43DV-dSIOI 1300Vd MOO19 3-19NIS

J.N3bVGOV-dSla -13GOVI A30-Oi9

P4..

.........

60

backfill properties cannot be as easily incorporated in a single parameter,and the results shown in
case which might be encountered
=

Fig.

5.4 are only for a typical

(
= =

in

practice

300;

0).

5
is seen from Fig. 5.4 that the differences between the two

It models

(smallest R2/ 1 )

are greatest for small values of N/A and/or An explanation for this trend is that as
S

for small values of A. either N or A increases,

the angle of the failure plane 0(t) (flatter). Hence, Nv (t) (Eqn. which is 5.1-a), so In O

becomes generally smaller directly related

to tan [G(t)]

becomes smaller

that the vertical acceleration and its the limit, as A or N becomes large

effects are reduced.

(roughly corresponding to T (horizontal),

becoming large),

the angle e(t) would be nearly zero

and hence no vertical backfill motions would result from purely horizontal ground motions. The reason for the ratio R2 / one is
1

being consistently less than It would be not

not completely clear at the present.

unreasonable to envision that although NH(t) and hence PAE vary with time during slip, that on the average, the results of the
S

two-block model should be same as the single-block model. Intuitively,

model

however,

the mere fact of adding "constraints" to a

of otherwise freer motions. Another

implies a restriction

intuitive notion,

from a work-energy viewpoint,

is

that the

two-block model has more energy-dissipating mechanisms than the single-block model. Whereas in the single-block model, the O

earthquake energy causing motion can only be dissipated through friction forces at the base of the wall, the two-block model has

-.

--.

..

.-.-

.-

--.

-. .

-.

-.

, .

--

...-. ,

..--

--

-.-.--.

--.

74

rhe goodness of fit 3.7,

is

shown in

Fig.

6.3; for N/A between 0.1 and within 10% of the

the value predicted by this equation is

-omputed C1 (within 5% for N/A > 0.4). Re as it it ideally should,

This expression does not,

go to zero as N/A approaches unity although that range. indicated by the which

predicts

insignificant values in

The scatter of the record means dRo is coefficients of variation in

are statistics mediate for

the second line of Table 6.1,

Fs Ro /Z
.

the random variable the uncertainty

For interis

values of N/A,

from record to record

about the same as for differently

oriented walls during any one

shaking.

At larger N/A,

the orientation effect has much greater

uncertainty.
Again, all these results were developed using only the

horizontal components of recorded ground motions.

6.5

EFFECT OF VERTICAL ACCELFRATIONS

Downward acceleration

of the plane supporting a block will thus decreasing the

to slip. In

decrease the normal force at the interface,

transmittable acceleration Conversely, and increasing

the tendency

upward acceleration

increases resistance

to slip.

a ground motion with many peaks of acceleration causing slip,

the

effects of the vertical component of ground motion may be expected to cancel. Hence the vertical component of ground motion has

generally been ignored when computing sliding block displacements. The actual effect of vertical ground accelerations has been
studied using the suite of 14 earthquake records described above.

For each computation,

the vertical component of acceleration was

S .............................................................

73

100

50 _

__

Mean Value Curve from Fig. 4.3

5 dRe

0.5

0.01

0.05

0.1 N/A

0.5

1.0

FIG.

6.2

MEAN DISPLACEMENTS FROM THIS STUDY COMPARED WITH THOSE FROM FRANKLIN AND CHANG (1977).

9ii.i

72

eynonential it

with a spike at the origin. similar to a log-normal

For N/A of 0.4 and 0.5, distribution.

ii somewhat

A'I of these results components

were computed using only the horizontal 0

of the recorded ground motions.

6.4

SCATTER AMONG DIFFERFNT SITES AND EVENTS The next step was to examine the record means d Ro For each normalized

selected N/A,

each of these fourteen values was first

to a common peak acceleration scaling.

and peak velocity using V2/Ag A is the largest absolute and V is the peak

(As previously discussed,

acceleration

from both components of a record,

absolute velocity from the component containing tion.) Then the 14 normalized values of

that accelerato

Ro were averaged that is:

obtain the overall mean displacement dre;

dRe = Ave

[aRe]

(6.1)

The scatter analyzed.

of the record means about

the overall means was also

The overall mean displacements, plotted as a function of N/A in Fig.

in

normalized

form,

are

6.2.

As a result

of the

scaling scheme used in Wong's analysis,

the average curve from Fig. 4.3.

these results fall below

A simple expression which

provides an approximate fit

to the mean slips is:

Re

37V Ag

-9.4N/A (6.2)

..

. .. ..-..

.-..

.- . . .

. ..,

. . ,

. .

. .

. . ... . .

. "

71

Table 6.1 COEFFICIENTS OF VARIATION ARISING FROM UNCERTAIN ASPECTS OF GROUND MOTION

COMPONENT OF UNCERTAINTY ORIENTATION OF WALL AT A SITE EARTHQUAKE TO EARTHQUAKE A 0.2 0.3 0.4 0.5 u 0.6 0.7

N/A
-

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.32

0.42

0.51

0.64

0.86

1.12

1.30

0.53

0.54

0.58

0.58

0.56

0.50

0.41

... <0.05 <0.05

<0.05 <0.05 0.05 0.06 0.07 0.08

0.05 0.07 0.10 0.12

0.15

0.07 0.13 0.18 0.25

0.33

0.15 0.22 0.30 0.37

0.44

0.12 0.27 0.37 0.57

0.66

0.19

0.42

0.51

0.73

. .

.-

70

6.3

ORIENTATION For a

EFFECTS analysis of this effect, components it of would each be desirable record, only the the

complete from

to compute, time histories two

the two observed in

of motions

many directions. used in this

However, study

recorded components

have been

to evaluate

four possible permanent displacements. Four values of permanent slip dRe were computed from each of the 14 records for values of N/A from 0.2 to 0.7, with no

normalization of the records and ignoring vertical accelerations. For each of the records, the analysis
d

an average aRo was determined

(used for

in

Section 6.4),
for each N/A,

plus four values of the ratio E0 =

56 values unity. of The line in at of F0 were obtained. of The

/a
of E
0

Thus is, are It

mean variation

by definition, listed on

coefficients Table 6.1. E

the first the scatter

may be seen that This

the ratio the larger

increases one or

as
-

N/A

increases.

occurs because,

N/A,

more component-directions example, the east-west at N/A

=

may not cause any 1940 El

permanent causes wall

slip. no slip

For in so

0.7

the

Centro record little for a

direction,

and very

oriented

that it
17

can slip to the south.

for
=

Overall,
are other zero

of the 56 computed slips,

for N/A = 0.6, and as 2 are

are zero for N/A

N/A 0.5.

0.7,

eight

zero

Numerous

values

are so

small

to be

essentially zero. Because of the tendency for an increasinq number of zero

values as N/A increases, the distribution of E0 changes as N/A

changes.

For N/A = 0.1,

the distribution was found to be

=

approximately normal.

For N/A

0.7,

it

is

more nearly

............

........

69

N
-0.-25g0.2g
--

Eg0

'V"

' --

-"vVYo-:669:

SW

0.3g

(a) Example N-S and E-W Components of an Earthquake

0.5 0/.2g

O0.2g

/E

0 0.3g

0.

15 g

"

b
"0,,.2g

ifrn o R rett i

W25ga

...... ....

.2
E,

0/.3g

Earthquake Components

FIG.

6.1

EFFECTS OF RETAINING WALL ORIENTATION.

68

tions are different in component.

the positive and negative sense of each

An important implication is

that the permanent slip S

experienced by a retaining wall during an earthquake will depend upon the orientation of the wall. walls shown in Fig. For example, the four retaining

6.1 would be expected

to experience differing 0

amounts of permanent displacement during any one earthquake. Indeed, some walls might have permanent displacement while the This is an important aspect of

others would not yield at all.

uncertainty in the prediction

of the motion which may be

experienced by any particular wall. In previous work, it has been usual practice to normalize a 0

component of a record to the maximum acceleration and velocity in

that component. However, this procedure tends to obscure the

orientation effect just described.

acceleration -

In the study by Wong,

one
O

the largest absolute acceleration was used

from either of This was always .

the components done because characterized

to normalize both components. at a site is

the earthquake motion recorded by this

largest absolute acceleration,

and this (or

number would always be used when any ether structure)

judging whether or not a wall

located at the site

lived up to expectations. the ground motions

The choice of a velocity to characterize at a site is less obvious. For this study,

use has been made of

the largest absolute velocity in largest peak acceleration.

the component containing the is consistent with past

This choice

practice and should lead to the least confusion concerning interpretation of results.
O

67

characterized resulting (Actually, 6.7). all

by having peak accelerations

greater

than 0.15 g 6.3 and 7.7.

from earthquakes with magnitudes between all but 2 records

are for magnitudes between 6.3 and imposed so that

The restriction

on peak acceleration was

records would be typical

of those which might cause signifiwalls, and distortion would

cant displacement

of actual retaining

not be introduced by scaling of weak motions.

The limitation upon

magnitude was used to narrow the range of durations of earthquake shaking, which roughly correlates with magnitude. Ideally a

similar study should be performed corresponding

using other sets of records but as yet this

to smaller and larger magnitudes,

has not been done. Analysis of scatter in sliding arising from differences First in
S

ground motions has been divided the differences in

into three parts.

there are components

sliding associated with the several

of motion at one location during one earthquake. are the differences quake. Finally from site to site

Second there to earthcomponent of

and earthquake of the vertical

there is In

the effect the

ground motion.

the end,

influences of these three effects considering them separately

will be lumped together.

However,

will provide an understanding several effects.

of the relative

importance of the O

6.2

SCALING OF RECORDS A typical ground motion record has two horizontal components. are

As a matter of course, different

the peak accelerations Moreover,

(and velocities) the peak accelera-

for the two components.

.*. -

66

6 - RANDOM NATUPE OF GROUND MOTIONS

6.1 INTRODUCTION This is in the first of three chapters dealing with uncertainty for a gravity

the prediction of the residual displacement

retaining wall.

As discussed in

the introduction, in

such uncer-

tainty arises because of differences motions,

the details of ground resistance of a wall

because of doubts as to the actual and because of errors in

to sliding,

the models used to predict

residual displacement resistance parameters.

for a given ground motion and given

This chapter deals with the consequences of the essentially random nature of ground motions. Fig. 4.3, As discussed in connection with

different ground motions each normalized to the same

and velocity can produce quite different amounts

peak acceleration

of sliding for the same N/A.

These differences are associated

with differing frequency contents,

differing distributions of Because the prediction

peaks and differing directions of shaking. procedure ultimately recommended sliding block model,
model. Thus the results

in

this report uses the simple

these various effects are studied using that

potentially apply to all problems for 0

which the sliding block model provides a reasonable prediction of permanent displacement or deformation including certain earth slope movements in earthquakes as well as retaining walls.

The analysis here uses results from a study of the mean and distribution of sliding caused by a suite of normalized ground motion records (Wong,
-

1982).

This study utilized a relatively Appendix A

S -, - .

small set of records

14 in number and listed in

65

5.5

SUMMARY The use of a single-block model analogy by Richards and Elms

S
-

(1979)

tends to overestimate the residual slip of a retaining wall Zarrabi's model, using the concept of

due to earthquake shaking. two interacting blocks,

provides a better estimate of the actual

slip, as confirmed by model tests on a shaking table. An important issue mentioned briefly in whether it is this chapter involves 0

valid to assume a failure plane inclination 0 which This and other

varies with the instantaneous ground acceleration. issues regarding how changes in other parameters 6) affect the results, are treated in Chapter 8.

(i, *b' i,

and

/- S-.

S" .o

.-

,S

-S

-. .

..

..

64 0.40

Measured (Ground)
U,

Measured

0.20-(al ~R-E 0-

100
0-0.20-040

0.10S

0.1

SE

E
Co

6-

/Zarrobi
M-`teasuredy

0
FIG. 5.6

0.08

0.16

0.24

0.32

0.40

0.48

0.56

0.64

TIME in seconds
COMPARISON OF RICHARDS-ELMS (R-E) AND ZARRABI SLIDING BLOCK MODELS WITH EXPERIMENTAL RESULTS (FROM JACOBSEN, 1980) (SEE CHAPTER 2 FOR DETAILS OF EXPERIMENTS, WHICH WERE PERFORMED BY LAI, 1979).

63

experimental

results

obtained by Lai in

(1979).

These experiments

were previously described A typical Fig. 5.6.

Chapter 2. shown in a better and is not 5 0

comparison of theory and experiment is the Zarrabi

As can be seen,

two-block model is

simulation of the time histories relative constant displacement. Clearly,

of acceleration,

velocity,

the limiting acceleration

as predicted by the single block model.

Also noted by

Jacobsen was the fact that the two-block model predictions are in better agreement during to the latter in the beginning of the shaking as compared It is conjectured that this is

part of the shaking.

part due to the physical constraint

that the toe of the soil in order to slide

block has to undergo some plastic with the wall.

deformation

The rigid two-block model

inherently assumed this

effect to be negligible. An important point with regard to the comparisons made by

Jacobsen, is that the calculations he performed should be

recognized as "Class C" fact. In particular,

predictions,

i.e.

predictions after the

Jacobsen used the angle of inclination 6 of

from the experimental results as input
S

the failure plane obtained to the computuer

simulation models.

As opposed to Zarrabi's

procedure where 8 varied with acceleration, to use a fixed 0 (measured in the model)

Jacobsen chose instead

and found that this gave

better agreement with the experimental

results.

_0

0 "-

62

E-4

z
00

u0

ci

00

d0

w0 0 0

E-E-

00 0z 0~ 4 0 05 04
0

0E

N ci 0

0 0

5:4

0D

IS

61

at

least an additional

frictional

energy dissipation the backfill. plausible,

surface

through the failure explanations justified

plane in

Though these

are intuitively

they would need to be

rigorously by further Another important

research. shown in Fig.

observation from the results

5.4 is

that it

is

not possible to normalize the residual as was then

displacements of the two-block model by dividing by V2 /Ag the case for the single-block model. If it were possible,

2/1

RR R

dR/(V2/AG) dR/(V /Ag)

should only be a function of N/A or NT/A. dR/(V 2 /Ag) is

Since it

is

known that but that is not

only a function of N/A (see Section 2.4), it

R2/1 depends on A,
*

can only be concluded that dR/(V 2 /Ag)

and hence cannot be normalized.

solely a function of N/A,

However,

as a practical matter in design considerations, 5.4 can be replotted as shown in Fig. 5.5,

the

"results of Fig.

where

the horizontal axis has values of N or NT instead of N/A or NT/A. This scheme condenses the values of R2 / data, minimizing the influence of A.
1

to a narrower band of has suggested

Wong (1982)

that R can be approximated for design purposes as solely a 2/1 function of N.

5.4

COMPARISON WITH

EXPERIMENTAL RESULTS

Jacobsen

(1980)

performed a detailed comparison of the

Richards-Elms single-block and Zarrabi two-block models with

"~~~~~~~~~~~.'-............ . -.-. "-.'''"- - . .... .i.. ."'............

,....:.....,,....,.........o... :........,...... ....

,M ,,, :.

.. 75

00

CD

ga~
0

le

0
2z a.W

00

z
0D

0z E40

E-40

aZ4

CY

H cf

0~~

00 B00 0ll

76

scaled in

the same ratio as the horizontal

component.

The

influence of the vertical accelerations is Ev = dR/dR

accelerations

indicated by the ratio

where d

is

the slip computed when vertical

The ratio E was found to depend (the coefficient A) as

-..

are considered.

upon the strength

of the input acceleration

well as upon N/A.

values of E

For each pair of values for A and N/A,

(less those cases where d

up to 56
0 and
S

were computed

Re

cases where dRe is of E in

V

so small that division by it error).

would give a value

which might be much in 6.4,

Average values are plotted the lower

S

Fig.

and coefficients of variation are listed in

portion of Table 6.1. Figure 6.4 indicates that, on the average, incorporating . .
-

vertical ground accelerations causes greater residual displacements. in This can be understood by considering a hypothetical case

which upward and downward

accelerations both reach peak values

of 1.0g.

When the peak downward acceleration occurs,

resistance

to slip disappears entirely,

whereas at the peak upward merely double that for zero the potential effect of downward than the influence of upward
S

acceleration the resistance is vertical acceleration. acceleration on slip is acceleration.

Clearly, greater

The coefficient of variation for the function E

except when both A and N/A are large.

is

small
with

As would be expected,

large N/A there are fewer intervals during which slip occurs, hence the sense of the vertical acceleration quite important. in

and
S

these moments is

The influence of vertical accelerations would

.-

~
.

..

..

". .

. ''.. . - . .

.."."-.-' . -", . . .

."' . "''" ..

' -' *.- .".',

. ''-

-.

. '.

-...

-. .

.. -. -"'",''.

''..'..

77

1.5

1.41.3
IL1.2
_ _

A0.0 0.5
0_

w
0.

w0

0-

0.i

0.2

0.3

0.4

0.5

0.66

0.7

N/A

FIG. 6.4

EFFECT OF VERTICAL GROUND ACCELERATION ON RESIDUAL DISPLACEMENT OF BLOCK.

78

appear even greater included. Wong suggested accelerations:

if

cases

in

which dRe

0 but dRy P

were

an equation

for the average effect of vertical

1.015

0.2N/A + 0.72(N/A) 2 vS

(6.3)

which is this

valid

for 0.2 < A < 0.7 and 0.1 < N/A< 0.7.

"Average"

in

sense

implies averaging

over a range of values of A as well

as over a set of computed

slips.

6.6

COMBINED UNCERTAINTY One way to estimate the overall uncertainty of the wall, arising from the and is

combined effect of orientation event-to-event differences,

site-to-site

and vertical

ground

accelerations

to combine the coefficients of variation in Table 6.1. estimate the slip of a block as:

We woulO-

dRv =o EosEv Re

(6.4)

where d and Eo, If

Re
Es,

is

a deterministic

function of N/A

(as given in

Fq.

6.2)

Ev are random variables which depend upon A and N/A. F and F is: are independent, then the

we further assume that Eo,

coefficient

of variation VR of dRv

)(1

+ V2

)(1

+ V)

(6.5)

79

where Vo, and E v

Vs and V

are the coefficients of variation of E ,

E-

respectively.

Table 6.2.

The resulting values of VR are tabulated in R As would be expected from comparison of the individual vertical ground accelerations contribute relatively 0

V0 , Vs and Vv, little

to the overall uncertainty except at large values of N/A.

Even here the predominant uncertainty comes from the unknown orientation of a wall relative to the principal motion. Alternatively, values of VR may be determined directly from These it 6 axes of the ground 0

the 56 computed values of residual slip for each A and N/A. results are given in is Table 6.3.

Comparing Tables 6.2 and 6.3, (Table 6.3)

seen that the directly-evaluated VR

are always less

than those (Table 6.2) discussed in

computed by assuming that the three effects 6.4, and 6.5 are independent. Clearly

Sections 6.3,

some degree of correlation actually exists among these effects. From the results at small N/A, orientation and site-to-site degree. it may be deduced that the O

effects are correlated to a slight

At large N/A the VR do not increase significantly as A RO implying that the effect of vertical ground strongly correlated to one or both of the other vertical 0

becomes larger, acceleration is two effects.

This latter conclusion seems reasonable:

accelerations are important only when there are a very few spikes of horizontal acceleration that cause slip, and having only a few

such spikes can also lead to a strong orientation effect. For design considerations only the results in Table 6.3 are of interest. However, having looked at the various effects
.

................... . ..
"-" ~~~.. .. '~~~~~~~~~~~~~~ ..........-,........
.'..'-..i i..i-L

..

.I'-..

-....-.

?.,...-.,..-.,

.-.--.

00 80 Table 6.2 OVERALL COEFFICIENT OF VARIATION ASSOCIATED WITH UNCERTAIN NATURE OF GROUND MOTION, BY COMBINING UNCERTAINTIES IN CONTRIBUTING EFFECTS

N/A 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.64 0.64 0.64 0.64 0.64 0.64 0.2 0.72 0.72 0.72 0.72 0.72 0.72 0.3 0.83 0.83 0.83 0.83 0.83 0.83 0.4 0.94 0.94 0.95 0.95 0.96 0.98 0.5 1.14 1.15 1.17 1.19 1.24 1.30 0.6 1.37 1.40 1.44 1.48 1.54 1.60 0.7 1.48 1.54 1.60 1.78 1.87 1.95
-

..

..

..

81

Table 6.3 OVERALL COEFFICIENT OF VARIATION ASSOCIATED WITH UNCERTAIN NATURE OF GROUND MOTIONS, FROM STATISTICS OF COMPUTED RESIDUAL DISPLACEMENTS

N/A A 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.63 0.63 0.63 0.63 0.63 0.63 0.2 0.68 0.68 0.68 0.68 0.68 0.68 0.3 0.78 0.78 0.78 0.78 0.78 0.78 0.4 0.88 0.88 0.88 0.88 0.88 0.88 0.5 1.02 1.02 1.02 1.01 1.01 1.01 0.6 1.17 1.17 1.17 1.16 1.16 1.16 0.7
S

1.39 1.40 1.41 1.42 1.43 1.43

Ao

.
.. .

..
..

..
..

.
..

..
. .

..
.

..
.

. .
.

..
. .

.
.

.
.

.
.

.
.

.
. .

.
.

.
.

.
.

.
. .

.
.

.
.

.
.

.
.

82

separately has been of considerable importance of the several

value in

understanding

the

contributions to overall

uncertainty.

6.7

PREDICTING RESIDUAL DISPLACEMENTS If the effects of vertical ground acceleration are ignored,

Eq.

6.2 provides a good prediction for a given A, is still

for the average residual There is a small "fitting in the 0

displacement

V and N/A.

error" and there determination

some statistical

uncertainty

of the actual mean displacement owing to the limited used in the calculation. to the scatter in 0

size of the suite of earthquakes However, these uncertainties

are small compared

residual displacement motions. When vertical

resulting from accelerations

the random nature of ground are introduced,Eq. 6.2 may

be modified to:

1Rv

vaRe

(6.6)

where

v is
V

given by Eq.

6.3.

As discussed but the error

in

Section 6.5,

E is 0

actually also a function of A, Eq. 6.3 is small compared

introduced with

by using

to the scatter effects.

associated

orientation and event-to-event

Table 6.3 provides estimates of the coefficient

of the residual displacements. is If the distribution Eq.

of variation
function for

the displacements

known or assumed,

6.6 plus Table 6.3 that various levels of

provide a basis for estimating probability ground motion might be exceeded. uncertainty in d arises

Since to some degree the the three

.

from a multiplication of

-. _,.._.,.?": , .'- --. -.i,:.:.L%.._ "..~. -.iLii..-.. ..

.. ...

.?i

.i...:,>...:...i-:-L:......... ...... .... . .. .

..

---

'.?....-*,,:?-

83

effects discussed in preceeding assume a log-normal distribution

sections, function.

it

seems reasonable to Figure 6.5 compares

cumulative distribution functions derived as suggested here with those developed from the actual computed residual displacements. In general appear tiat the agreement is reasonable. In particular, it would
S

a satisfactory estimate can be obtained for the

S

displacement which will not be exceeded with 95% probability. Assume that the actual residual displacement can be written as: d
=

Rv0

(6.7)

where Q is

a log-normally distributed random variable with mean of VR. The mean and standard S

unity and coefficient of variation V deviation for InQ are:

2S

ln (1+ VQ)

(6.8)

mlnQ= ln(E[Q])

Var [lno]

-1

Q2
-.

ln(l+v 2
0 lnQ

(6.9)
-..,n-

The value of lnQ which will not be exceeded with 95% probabality is then:
..............

_0

..

m.h ..... .

ilm

..

..

.....

. ,

77

84

0a

hI

00

0A 0

=0.2A

0.

0N/A=

100

~~

0.2
1_

N/A= 0. 5
Goo

200

300

400

Cv

OISPLCEMET

~DISPLACEMENT

(me)

LOSS

00

00 10 2

N/

DISPACEENTtom)DISLACMEN

0 4

o0
EUT

0.5NA

I9oI

OPRSNO FIG.~~~~~~~~~
WITH DERIVED DISRIUTINSFORDIPLAEMNT

FSIIGBOKAAYE 0.
TTSIA

CUMULAT~~IVELGNRA

85

lnQ9 5
or lnQ 5 9

mlnQ + 1.649 in1

(6.10)

I n 52

(+V

+ 1.645 Vln

(1+ VQ )

(6.11)

or 1n
95

(1+ V)
-

+ 1.645
+.Vo

in(l

(6.12)

in-(1+

Values of 095 are given in Table 6.4, The interpretation of these results is the allowable residual displacement is

for typical values of V0 . as follows. dL. Then, in Suppose that order for .

there to be 95% probability that dL will not be exceeded during an earthquake with given peak acceleration the expected (average)
9 5.

and peak velocity,

then I

residual displacement for such an earthThis assumes that other parameters of a

quake should be dL/Q problem,

such as the resistance of a block or wall to the are known with certainty. The effects of the next chapter.
.

" .

initiation of sliding, uncertainty in

these parameters as discussed in

6.8

APPLICATION TO RFTAINING WALLS As discussed in Chapter 5, for a given A, V and N the d., is not equal to the

S
-

residual displacement of a retaining wall, residual displacement of a sliding block.

Thus Equation 6.2 does

not apply exactly for the mean residual slip of retaining walls. However, calculations by Wong (1982) have shown that the

86

coefficients of variation for dRe are very nearly equal for dRe' for each A and N. Hence Equation

to those in

6.11 and the results

Table 6.4 apply to retaining

walls as well as sliding blocks.

6.9

IMPROVEMENTS TO PREDICTIONS The procedure outlined here for predicting is, of course, the average dis-

placement suite

of a sliding block

only as good as the

of earthquakes

used to develop the correlations and as the intensity of .

assumption that A and V are the best measures of shaking. It will certainly be desirable 6.2 by using a larger

to refine and improve

upon Equation Another limited

suite of earthquake motions. suites, each for a so

improvement will be to use several such range of magnitudes

(and perhaps epicentral distances) It may

as to reflect

the influence of the duration of motion.

also be that there are parameters better occur;

other than A and V which are displacement distance. than can Explanathe

indicators of the amount of residual for example, magnittle and epicentral

tory studies direction of Section 6.3,

into these questions are ongoing at MIT under Prof. Daniele Veneziano. Finally, as noted in

a more thorough treatment of the orientation effect

would be desirable.

S .

,.

...................................................

..

..

87

7-

UNCERTAINTY IN RESISTANCE PARAMETERS

INTRODUCTION The analysis in Chapter 6 has assumed that the mean displace0

nt dRe is ckfill.

a known function of the properties of the wall and Actually, dRe is itself uncertain - partly because of 0

certainty in

these properties and partly because of uncertainty

the model being used to compute displacement for given operties and given ground motion input. e effect of uncertainties in This chapter deals with 0

the wall/backfill properties.

These properties include the weight of the wall; the unit ight of the wall; the friction angles at the base of the wall, thin the backfill and at the wall-backfill ometry of the problem. certain to some degree, interface; and the S

While all of these properties are the most important uncertainties are The influence of insignificant, these other S

ose associated with the friction angles. certainties in

the other properties generally is this analysis. That is

d will be ignored in

to say,

operties will be considered as deterministic. Computation of the uncertainty in permanent displacement ivolves: id then (a) (b) evaluation of uncertainty in the friction angles, through the

propagation of these uncertainties

ialysis connecting the friction angles to permanent displacement.

ie latter step is described in sections 7.3 and 7.4, beginning
-"-. -.

th the case of a sliding block on a horizontal plane.

..............

88

7.2

UNCERTAINTY The friction

IN

FRICTION

ANGLES for analysis of a particular using based the actual the upon friction in most in as 6

angles

selected

retaining actual past soils

wall at

seldom are based that wall, but

upon values measured rather are estimated

experience

with similar as sliding with this as part

soils. progresses. latter of

Moreover,

angles may change movement associated properly Chapter resulting angles. When friction starting small. deviation It from a be treated 8, but for

The uncertainty consideration error it should discussed is

the model this report

purposes of in

considered friction

from uncertainty

the evaluation

of the

angles specified

of a initial

granular density,

soil

are scatter by a

measured, is relatively 0

might reasonably a of 1 or

be characterized However,

standard the backfill

2 degrees.

because

sometimes is

poorly compacted and its

actual density unknown,

the

possible variation about the actual mean value must be greater. Because backfill does tend to be loose, decrease in continues. there is relatively little 0

friction angle past the peak value as strzining Furthermore, and it is

for backfill is

almost always selected

conservatively,

rather unlikely that the minimum value S

can be more than several degrees less than the anqle typically
used for design. (It is the possibility that the actual t may be

less than the value assumed for design that is to us). All in all, it

of primary concern 2 degrees or

..-

seems reasonable to use a.

possibly 3 degrees.

-j -- -i >. . .' .' - i- ... " >>~i -> ii ' > >> :..' -i''> >

.: :

" ". "." i3. -"

i : . .. "

.-- '-'>S .

I-

.,.....

89

These same arguments also apply to the friction angle *b at the base of the wall. compacted, decrease in significant.

bi

Because foundation soils tend to be better greater than t. By the same token, likely to be more
0

*b usually is

*b with continued sliding is However,

once again the values nominally used for Again it seems

calculations are likely less than the actual mean.

reasonable to use a standard deviation of 2 or 3 degrees. Engineers tend to feel quite uncertain as to the choice of a suitable value of wall friction angle 6. actual peak interface friction angle is there is in It is not that the but rather is .

great doubt,

uncertainty as to how much frictional resistance

mobilized at a wall before static failure occurs. problem where slip is actually expected, Nonetheless,

In the seismic .-

this type of uncertainty it is considered that a the case of wall

shvuld not be so important.

larger standard deviation should be considered in friction; say a6= 5.

0

For the calculations in

the following sections,

it

is

necessary to pay careful attention to the units for the standard deviations of friction angles, re-expressing the foregoing values are summarized in and this is best handled by Appropriate
S

results in

radians.

table 7.1.

7.3

BLOCK ON HORIZONTAL PLANE When uncertainty in resistance is considered, the term d

in Rv Equation 6.7 becomes a function of the random variable N; that is, this term is found by: itself a random variable. The variance of d may be

...

-.....

...

..

.....

....

*..

..

..

..

..

-.

90

Table 7.

"ESTIMATED STANDARD DEVIATIONS "FOR FRICTION ANGLES

Location

a in

radians

Backfill Base Wall

0.035 to 0.052 0.035 to 0.052 0.035 to 0.087

.0

.o .S

. ,

-.

..

91

[2 8N1 Rvv Var~dRy I = [--gf2 Var[N] (7.1)

i~ 2

Alternatively,

but less precisely,

we may rewrite Equation 6.6 as:

dr

dRyQR,

(7.2)

where now

is

a deterministic function computed using average the problem at hand and ".

values for the friction angles involved in R is

a random variable with mean unity reflecting uncertainty in The variance of R is given by Equation 7.1. .

the friction angles.

Computations show that 83Rv/bN is

1/N.

closely approximated by

Hence the variance of R* may be expressed as:

[@]= Fv Va

a 2_8N

Var[R

37V2 -9 4/A) 2 A ] Var[N] (-- e9

(7.3)

[_14 (_-j

Rv 2 Var[N] 2]

Taking the square root to obtain the standard deviation of R

and

dividing by

0Rv'

of R the coefficient of variation

becomes:-

Vrj,

9..

(7.4)

.-..

..

. . . . .. . . . . . .. . . . . . . . . . .. . . . . . . . . . .

.-

-.-........

.-.

-. .~

--

92

Note that V R coefficient A.

is

inversely related to the ground acceleration This means that, since it is ratio N/A that

determines the amount of slip, a given uncertainty in N is much more important when A is small than when A is large.

For a sliding block of known weight on a horizontal plane, the transmittable acceleration N is determined entirely by the friction angle between the block and the supporting plane:

tano

(7.5)

The variance of N is

then found by:

Var[N]

[.]

DN2 Var[4]

4 sec4q

Var]

(7.6)

and the standard deviation aN is:

aN =a c/cos

(7.7).

Typical values for

range from 300 to 350.

Using the values for

a, from Table 7.1, the range for aN is 0.05 to 0.08. Combining Eqs. 7.4 and 7.7 leads to the results in Table 7.2. Note that some of the values of V - those for the smaller A are quite large, much larger than the coefficients of variation
-

..

discussed in

Chapter 6.

Thus it

is

evident that uncertainty in

resistance parameters can be a very important consideration.

. .. . .

93

Table 7.2 COEFFICIENT OF VARIATION OF RESISTANCE TERM

VR

A 0.2 0.3 0.4 0.5 0.6 0.7

aN -0.05 2.35 1.57 1.18 0.94 0.78 0.67

aN

0.08 3.76 2.51 1.88 1.50 1.25 1.07 S O

S,

:S:.,: .

.S

"S":

_O

94

7.4

RETAINING WALLS For retaining walls one proceeds in a generally similar

fashion.

In this case,

an exact solution would require

derivatives

R Values of the partial R 2/1'Vle f h ata derivative have not been evaluated, although approximate results by Wong (1982) indicate that the value obtained for a sliding That

for the product d

block are a good approximation for the retaining wall case. is to say, Equation (7.3) may still be used.

Hence interest then .

focuses upon evaluation of Var[N]

for retaining walls. the three

The variance of N depends upon uncertainty in angles 0, *b and 6. each other, If

these angles are assumed independent of

then Var[N] may be evaluated by:

Var[N]

N 2

Var[b] b

(-) D6

)2

Var[61

(7.8)

The derivatives are obtained from Equation 4.4, rewritten as:

which may be

N =tan~b -

[cos (p+6) -sin (p+6) tanOb] PAE W w

(7.9)

.-

"

Since PAE is

a function of N as well as of Wong (1982)

and 6,

implicit

differentiation must be used. necessary steps, Table 7.3.

carried out the

-

and the resulting equations are reproduced in

Numerical evaluations of these equations have been performed for various combinations of values of the three angles and of N.
>~ - i-.. i- ? . , li-lllll2 .llli'.-

~~

- ii....-

. -..

-.-

...

.ili-i. -i

-i.

i..i

. .:

i'

.-.

95 Table 7.3 EQUATIONS FOR PARTIAL DERIVATIONS OF , AND ~ b N WITH RESPECT TO

-N

Ge b F [ cot(a+6 ) 3 b G L~ot (+6) -tan~b)

-N

(tan~b-N)J

IN

1.

[2r
-~~~~~~

nOa

cot (0+6) +cot (o-i-i)1 G jtn(--)+1+ j

_1

CY

6 [cot(ct06)-Jtan(iP++6) tan(a-I )+tan b1 + 1 + J l-tan(+ 6 )tan bj

Ptan+N

+ N + cot

(--i) +Htan ('P+a+

+tan

-N

/cos (i-a)cos (t+a+6) sin 6sin(i

96

(All have is

involved

= =

0.) 0.

One special case of some interest Then equation 7.8 may be rewritten
---.

that when

*b

and 6

as:

Var[N]

+ +(N)

IN

Var[p]

(7.10)

The term in

brackets is

nearly constant for the several combinaThus:

tions investigated,

at a value of about 0.85 radians-2

aN=

/8

0.92a

(7.11)

That is

to say,

the uncertainty in

N is

actually somewhat less This reduction relatively unlikely (or

than that in

each of the two friction angles. assuming independence, it is

occurs because,

that both friction angles will be simultaneously larger smaller) than mean values. If there is

some actual dependence then the ratio aN l

between the values taken on by these angles, would be increased to some degree. When wall friction is situation is present,

as indeed it

must be,

the

more complicated.

Results from some typical calculafor the case

=

tions are presented in Table 7.4, 5

0

=b 2

a =

and

i = 0 0.

It

may be seen that a

varies with the

combination of values used for the three angles and N, and is greatest when N is smallest. Somewhat greater values of aN may be Further-

found when other combinations of parameters are used. more, depending among the three friction angles

(especially

97

Table 7.4 STANDARD DEVIATION FOR RESISTANCE FACTOR N FOR GRAVITY WALLS WITH WALL FRICTION, ASSUMING a = b =
AND

06

504

pb

N
_b_

N 0.05 0.1 0.2 0.3

6=0 0.032 0.031 0.032 0.033

6=200 0.043 0.041 0.038 0.036

6=300 0.049 0.047 0.043 0.039

300

30 0

300

350

0.05 0.1 0.2 0.3

0.034 0.033 0.033 0.033

0.051 0.048 0.044 0.041

0.061 0.058 0.052 0.047

350

350

0.05 0.1 0.2 0.3

0.033 0.033 0.033 0.033

0.049 0.047 0.044 0.042

0.059 0.057 0.052 0.048

S

S- . .

98

between

and 6)

will cause the aN to increase.

It

was judged "

that an average of the numbers in

would be most representative average values are plotted in of

the final column of Table 7.4

N, and such

the uncertainty in

Fig.

7.1 as a function of N.

Values of aN may be entered into Equation 7.4, is is is still at least approximately valid.

and Table 7.2

Once again the conclusion in the friction angles

that the overall effect of uncertainty very significant.

.. S

S :i i

-S

S"m"
""

" "" m

"Ia

"

"( ;

"

" "i

" "

99

0.08

0.06

0-N

0.04

-(
S

0.02

00.I

0.2

0.3

FIG.

7.1

TYPICAL VALUES FOR WALL FRICTION.

N FOR RETAINING N

WALL WITH

. .

..

. .

. .

. .

. .

100

8-

MODEL

ERRORS

AND UNCERTAINTIES

This chapter discusses several ways in which the two block model of Chapter 5 is still imperfect, but where knowledge is

inadequate to permit use of an improved model. include the proper choice of a failure plane,
deformations tilting choose considerable available. within the backfill In addition, before there results sliding

These aspects the effect of

begins, because lieu of and one may

of walls.

are errors in

to use simpler effort in

approximate using

expending that are

the best

methods of

analysis

8.1

FAILURE In Zarrabi's

PLANE

INCLINATION (1979) two-block model, it was assumed that the

inclination of the failure plane in instantaneous ground acceleration.

the backfill varies with the This is a natural assumption

that arises from the formulation and application of the MononobeOkabe Equation.
tests performed

However,
by Murphy

as described in Chapter 2,
(1960) and Lai (1979) indicate

the model
that it S

may be more reasonable

tion of remains Lai's constant model

to assume that the failure plane inclinaduring test slip. Jacobsen's with calculations in a that (1980) comparisons the

(1979) model

results this

using

two-block is better

concur with

reasoning,

the agreement angle of

between

theory and experiment, assumed. the calculation Thus, it of would residual

when be

fixed

inclination fixed angle

o is
0 in

reasonable

to use a

displacements.

114

depth more rapidly rotational

than linearly. of the wall horizontal

At other

times when the the opposite is true

-

acceleration

reverses, lies

and then the resultant

force

above the lower third 0

point.
Once revealed by this are quite obvious. simple conceptual model, these results that

The analysis demonstrates clearly

conventional wisdom concerning which was derived

the location of the resultant, that does not tilt, may is 0

for the case of a wall

be quite misleading as regards free to tilt. Nadim used sliding and tilting. this model

the response of a wall which

to study the relative

importance of

Moment resistance at the base of the wall in character, and it was further A
.

was assumed to be rigid-plastic asumed that the axis of tilting

was at a fixed location. NR' is

threshold transmittable moment required

acceleration,

reached when the base equals the maximum

for dynamic equilibrium just

moment.

(The dynamic stress from the backfill

is

assumed to Any tendency

is resisted from 0

increase linearly with depth in

for the thrust from the backfill inertia

this calculation.)
to increase further

by rotational

of the wall.

The principal conclusions

Nadim's study were: Tf N > N, then only sliding will occur. That is, having R resistance to sliding which is less than the resistance to
tilting protects the wall against tilting.

If

NR < 0.85N,

then only tilting

and no sliding

occurs.

In this case,

the permanent displacement at the

-

113

0 ,all may slide and/or rotate about is just its base. The horizontal

icceleration of each slice icceleration of the wall at pall. Each slice

equal

to the horizontal meets the to

the point where the slice vertically, With

also accelerates

as necessary these various

iaintain contact between the slices. issumptions, listribution it is

possible to develop an equation between

giving the in terms of

of stresses

the wall and backfill base

he horizontal liffer

acceleration

of the wall at its of the underlying

(which may and the then used

from the acceleration acceleration

ground) is

otational

of the wall.

This equation

.ogether with equations as in Chapter 5)

for the dynamic equilibrium

of the wall

to compute the motions of the wall. analysis - a rather surprising result at

One result

of this

irst

sight

is

that the resultant of the dynamic stresses

)etween backfill and wall sometimes lies below the lower third point. This happens whenever the tilt of the wall away from the acceleration of the
S

)ackfill is [round is

accelerating while the horizontal (which is

towards the backfill

just the situation of (see figure 8.5b), the

0

freatest concern to us). ibsolute acceleration ibsolute acceleration issumptions,

At such times

at the top of the wall is at the base of the wall.

less than the With our

this means that the uppermost slice through the than the slices below it. increases with

)ackfill has a smaller acceleration lence the horizontal

stresses between soil and wall

It turns out that the thickness of the individual slices can in the limit be set to zero, so that a continuous equation is obtained.

112

ol
4e.

0 IIA E- 0

U3

Ns
In

0l
00

S-

~~0

u 0H

O E-4O

NN
"P4
HD

N
C-0E-4Z
=

tipii-

. .-

111

depicted by the dashed "proposed design lines" shown in Fig. 8.4.6

The factor RE obtained from these lines can be justifiably factors to be applied used as to
S

relatively

conservative amplification in

simpler sliding block analysis, suggested. It backfills should be noted will typically

the same vein as previously

that the fundamental

frequency

fBF for
S

range from 5 Hz to 25 Hz.

Dominant from 2 Hz to 5 ratio fE-Q/fBF

frequencies Hz. is Thus,

of ground motions fE-Q range generally the most typical range of the frequency

0.2 to 0.6.

8.3

TILTING Observations in the field following earthquakes suggests that

permanent tilting. little

displacement of gravity retaining walls usually involves This important aspect of behavior has received very standpoint, tests and hence is in still is

attention

from a theoretical

poorly understood. tilted motions while it

There have been model

which a wall like

and the backfill et al. 1981).

are subjected to earthquake These tests have confirmed

(Sherif,

the

influence of tilting behind a wall. amount of tilting A preliminary conceptual is

upon the development

of active conditions

However,

they give no direct evidence as to the in a free-standing (1980) wall. used the

that might develop study of this in

problem by Nadim The failing

model shown

Figure 8.5.

wedge of soil S

subdivided

into thin slices

which may slide over each other

once frictional resistance on the interslice failure surface is reached. Except for such sliding, the backfill is rigid, while the

S.

110

idealization

of the problem.

In

all

of

these cases,

it

was

assumed that there is a rigid boundary beneath the backfill at the

same level as the base of the wall. motions reflect However, in This means from this that seismic wave boundary. not rigid, and a at

back into the backfill most real problems this

boundary is

portion of the waves radiate this boundary

away from the region of excitation Because of this effect, the is

(radiation damping).

actual

amplification

of ground motion through the backfill

probably not as large as one predicts from the finite analysis wall, with a rigid boundary at the level

element

of the base of the

especially Another

near resonance conditions. in Fig. 8.4 is that the amplification of

trend shown is

residual displacements If N/A

=

more significant

for high values of N/A. there would be

1 and the rigid block model

were correct,

no slip.

However, in

because of amplification through the backfill the upper portions of the backfill would
and hence slip would occur. Thus for N/A equal

the acceleration

actually exceed N,

to and slightly greater than unity,

previously N/A is noted, however,

the ratio R is F

infinite.

As

the usual range of practical

values of

from about 0.3 to 0.7. of preceeding observations into a practical (19P2) and

The incorporation design criteria

can be done in (1983)

several ways.

Nadim

Nadim and Whitman and V to reflect

have proposed modifying

the values of A and using

the amplification

of ground motion, input

these amplified values of A and V as

into a block model

analysis,
alternative

such as the Richards-Elms

method that is perhaps

Equation

(Eqn.

4.1).

An
is

simpler and more direct

109

0 d RE
=

R d

using elastic finite elements R using rigid 2-block model

f and T are respectively excitation, the dominant and f .

The variables frequencies T

and period of the earthquake fundamental

BF

and

BF

refer to the first

frequency and period of the

backfill.

From Fig.

8.4,

it

is

clear that much larger displacements the

element 10) is as S

than predicted by the rigid-block model can occur if

elasticity model. the ratio of the backfill The ratio R is considered, as in (on

the finite the order of 1.0, which

can become very large fQ/fBF

of the frequencies

approaches

analogous to a resonance condition for a SDOF vibrating system. Since the fundamental
approximated as

frequency of the backfill can be

BF

Cs

(8.4)

4H where Cs is the shear wave velocity of the soil and H is the

height of the wall.

This implies a direct correlation between the Thus the amplifica-

frequency ratio and the height of the wall.

tion problem is more pronounced

for higher walls.

It

should be noted the finite element results obtained

for

excitation frequencies near the resonance condition are conservative because of the boundary conditions that are imposed in the

...................................

..........-.--

,_-

108

E-

0 N

-E-

0
zz
E..-m...:

0EODo
0
0
E-4c

0S

I.
444

OH"

SZz 0 Uo
,---,C.-,. c

z
-4

000
Oi_ 0HOr. "I

ri

~ EE

E-6

~0 Cl)'
0 LO i 0 0

sl)

a07oie 13001N
SIN3fl313 3IJNI

-z a]li8

buisn Up

OI1LSV713 buusn Up =L"

...................................-..........-.... -.

'.--:.

....

........

. ."...

.'.......'.

..

".

...

'

107

increases the ground accceleration coefficient A.

Assuming that

the transmittable acceleration Ng is approximately the same

regardless of the flexibility assumption, the effective N/A ratio than for the rigid backfrom rigid block Also,
S S

would be lower for the flexible backfill fill case. Consequently, considering

the results

models, there is

larger

residual displacements would be expected. amplification

additional

of the transmittable accelera-

tion coefficient, expected

such that larger earth pressures would be if envisioned in the context of seismic
S

to develop,

coefficients. Obviously, the preceeding statements are only intuitive affects

arguments for explaining how ground motion amplification residual displacements. It should be recognized strict

that the concepts of N, A,

2 v/

',

of rigid block models, the ratio N/A,

including

quantifications

and notions of normalizing displacement in view of the non-uniform distribution finite element model.

using V2/Ag of However, the 0

are no longer valid these quantities in corresponding

an elastic

rigid block model quantities references

can be calculated and S

can serve as useful Numerical results results

for comparison of results. finite element model

from the elastic

as compared with those from the rigid two-block model are in Fig. 8.4. Computations are shown for values of N/A. The
0

(assuming constant e) three earthquake

records and two different is analogous

format of comparison

to the frequency response cure vibrating mass. factor, is The ratio defined as: .0

for a single degree-of-freedom RE, a residual displacement

(SDOF)

amplification

ES

~.
.

..

. . .

"- -

.--- :--.

.-.-

..

-.

", . .

. =. ..

....-.._

-..

-"-

-"

-"

."

.- ,

-"

-.-

-'-

-,

-".,.-

..

.-

106

Some ambient ground motion at both locations

"'

Acceleration constant ~within block

(a) Rigid Blocks and Backfill

Ground motion amplified by elastic backfill

2'/ -

Ambient ground motion

Acceleration variable and amplified in block

(b) Flexible (Elastic) Blocks and Backfill

FIG.

8.3

SCHEMATIC COMPARISON OF EFFECTS OF RIGIDITY AND FLEXIBILITY (ELASTIC) ASSUMPTIONS IN RETAINING WALL BEHAVIOR.

.)iii:
--7,.

105

:1
because of lack of data. cance, However, in terms of practical signifione

for the ranges of N and A commonly encountered,

whether

uses a fixed or variable 8 may not be important. especially true in shown in Fig. light of the uncertanties

This is

in evaluating 8 (as in the

8.1) and the even larger uncertainties

earthquake ground motion characteristics.

8.2

ELASTIC BACKFILL EFFECTS

The sliding block model developed by Zarrabi (1978) assume

the blocks representing the retaining wall and the soil backfill wedge to be perfectly rigid. assumption, This is actually a fairly good
-

particularly for walls with relatively low heights,

flexibility are essentially

where the effects of wall and backfill

negligible.

However,

for high walls,

the rigidity assumptions may The effects".

lead to severe underestimation

of the residual slip.

of assuming the wall and backfill to be elastic,

flexible, element has been studied by Nadim as described in in (1982)

and hence

using the finite 2. These result (1983).

idealizations

Chapter

have also been reported

a paper by Nadim and Whitman backfill similar is

The basic effect on an elastic amplification of ground motion,

the resulting which is

to the phenomenon

expected to occur in earth dams (see Seed and Martin, Makdisi and Seed, 1979). in The amplification phenomenon Fig.

1966 and is schema.

tically illustrated
case,

8.3 and compared to the rigid block

without amplification.

Firstly,

amplification of the ground motion occurs in the and effectively

backfill outside of the soil failure wedge,

104

0 O

z
o
US

w~
0

dw

Iz U)
--

"c

z
tun

OD
-

OW x
0I4

04X
bu)Z U

IOXl

80 X-I

"

0us
N.::..:
0

-..

.... . .
.. .-...... ... .., . .-.. . ... .-,. ... ..
. . -. . . . _ -. . . .

..
. " . " .

.. ..... i -.C,.,
. -O " .'

:rz.:.

.. . . . .. . : .- . . ., ,. , ,

2 "j -

- . o. .

, -

. -,

.. ; ,.,..

o. .

.. , ,.-

. . .. -

--. , . . ,, ,.-.

.,,

-U).-.-

'

103

(increasing

Fww),

a larger slip.

inertia

force from the backfill

is soil ..

required to initiate

Thus a larger mass of backfill in Fig.

""

must be mobilized

as reflected by a decrease results for 6 shown in

The experimental tently higher

8.1 are consiscalculated based (1979) and effects

than the theoretical of 0,

threshold eTV

on measurements Jacobsen between (1980).

*b and 6 evaluated by Lai

This may be attributed

to frictional

the model and the glass sides of the model container, increase *. There is also a range

which would tend to effectively of possible interpretation that the failure line of failure. failure zone, surface is

of the failure

angle 6 due to the fact

actually a zone rather than a thin there is a slight curvature to this 0

Furthermore,

deviating from ideal

linear conditions. in the calculated

The numerical differences residual

that arise

displacement between assuming a constant or variable are illustrated in Fig. 8.2. The results shown

plane of failure

are obtained from calculations (1982) and Wong (1982)

=

performed

separately by Nadim

using three earthquake records and a 0.112. The Ratio Re is defined as:

constant value of N

Residual Residual

displacement using fixed 0 displacement using variable

e
to one, of

(8.2)
S

The value of R8 those values greater numerical greater

is

generally

less than or equal

with

than one thought to be the result There is clearly a

round-off errors. for

trend toward given a fixed '

discrepancies but

increasing accelerations,

value of N,

further generalizations

of trends are difficult

"................

*i

"

t*'

.'."

....

...

' .

"....-...

..

"

-"...

...

"""'

102

0
0

"E-

z z

Cd

z.

0E-

3 0 4
0
co1-

~0
w 0

pO

" a-

0~

I--4

..-

0CJ

00 00

44

101

J
it

Physically,

once a

failure

plane develops weaker

in

the backfill,

the plane may become slightly

than the surrounding soil and at subsequent stages of that develops should failure

would be the preferred plane of failure Thus, the initial failure plane

movement.

logically be used in

the analysis.
0

Thus the critical

plane angle of inclination

which would occur during earth-

quake motions is

only dependent on the wall weight and the Nadim (1982)

properties of the foundation and backfill soils.

gave equations that determine the angle of inclination of this plane explicitly as:
OT = 1/2(a + a,) (8.1)

where

a = tan-1

(B/A)

a1 = cos -1 (-C.cosa/A) and

A
=

cos(4+6+i)sin(4b-)

sin(+6-i)cos(bcos(i)cos(
6

+$b)F

"B =

cos(i)sin(s

b-

)sin($+6)

C = sin(o b+6+i)

- cos(i)cos(6+sb)/Fww

W"

ww

2 i/2yH

w 2

wall

weight

factor

A comparison of the theoretical and the results * 8.1. 8 with from Lai's (1979)

equation

give by Nadim is shown in

(1982)

experiments

Fig. trend for trend

Both theory and experiment increasing wall weight

show a general

downward

is ww explained physically by the fact that as the wall becomes heavier

. . . .. - . . . .. . ..-.-- , ' ' ..- .,.- a .d '.:.= . . . . . . . . ,.' . . ," ". : .

factor Fw.

This downward

: . ..- " " % " "

115

top of

the wall was found to be about 1 1/2 times the using a sliding block with N set equal some reason to think that a sliding in estimating permanent is the predominant mode

movement predicted to NR. Thus there's

block model maybe useful displacement

even though tilting

of displacement. These conclusions must, of course, be treated with great caution. tilting may occur is

Unlike the case of sliding, before the maximum resistance reached. problem. Figure

significant

against overturning moment

There has been no adequate study of this Some results obtained using the finite that the resultant

aspect of the element model of from the

2.5 have confirmed

thrust

backfill can lie below the lower third point at various times
during a cycle of shaking. It does seem clear that any tendency

".-

for a wall to tilt

will relieve the overturning moment acting upon

the wall. All in all, there is reason to believe that the sliding block
and Elms is a reasonable model for

""

model proposed by Richards

predicting the permanent movement of actual gravity walls, provided they have been designed using a typically conservative ' safety factor against overturning by static loads.

8.4

APPROXIMATIONS

TO 2-BLOCK ANALYSIS

If

one knows the inclination of the failure plane through the or if one is willing to accept the assumption that this the effect of
as discussed in

backfill,

inclination varies continuously during shaking,

kinematic constraints can be taken into account

. .i ., . >..' - . .-- . .

.- ' -. - } -'-} .- -/

.. - .. } .. . . . .. ..- *.* *

*:.*. *.'

- '. '.-

'

.'*.*.

" .

.. :.*

116

Chapter complex,

5.

However, it

the required generally is

analysis is

at least

nmoderately

and hence

desirable to accept approxi-

mations in Wong

order to achieve simplicity. (1982) has suggested an equation for the factor R2/ 1 . '

[0.7
R2/1
=

+ 1.2N(l-N);

N < 0.5 (8.5) 0

N> 0.5

This equation describes an average curve through the several curves in Figure 5.5. Each of those curves is itself drawn ground

through a scattering

of points calculated using different

motions.

However,

the scatter in in Eq.

these points is

quite small.

The approximation

8.5 lies in
for Antia

ignoring the effect of A.

=

Wong computed his results specific valus of 0 and varying *8"

the case f (1982)

0 and for the effect of

S

examined

these parameters upon the ratio computed by the two-

R2/1 of permanent His

displacements

and one-block models.

results are illustrated by the following tabulation N = 0.4 and Taft earthquake record):

(for A

0.2,

i 0
7.50

6 0
0

Ratio of R22/1 1 / 0.731

0.767

0
0

0
0
7.50

150
0
150

0
150
150

0.828
0.764
0.917

117

In

all

cases

investigated,

the effect

is

to decrease the computed using the Note that each weight of

difference between two models. case in wall,

the residual

displacements

Wong reached similar conclusions.

the tabulation above corresponds to a different

so as to hold N constant. These results, plus those discussed "2-block" it in Section 8.1, emphasize no

the complexity of the so-called doubt that the effect actual permanent is

effect.

There is

real and that

acts to reduce the

sliding compared

to that predicted using the

Richard-Elms simple sliding block model.

The problem is

in

predicting this reduction accurately by any simple calculation. Having the inclination of the failure plane fixed tends to make
the reduction greater than suggested by Eq. 8.5. As the results

in

Figure 5.6 show,

this reduction can be very large indeed.

show that this

On
...-

the other hand,

Antia's results

equation may not

always be conservative. Taking all these factors into consideration, to use a mean value of 0.65 for the factor R2 /I,
the uncertainty in this

-0 it is reasonable

and to represent
S

factor by a standard deviation of 0.2i

IdS
'.

.~~~.

.-. .

..

............................................................................
........................................................
,...---._-.....

118

9-

IMPROVED APPROACHES TO DESIGN

9.1

REVIEW OF OBJECTIVES As discussed in Chapter 1, the general objective of this invovled in using a
0

study has been to quantify the uncertainties displacement-limiting approach

to the design of gravity The specific goal has a suitable safety if possible, the
S 0

retaining walls for seismic loadings. been to select,

with greater confidence,

factor for use with the Richards-Elms approach or, to develop an improved design methodology.

In so doing,

desire to maintain the essential simplicity of the Richards-Elms approach has been an overriding consideration. The various aspects of the problem have now been examined, to the extent that knowledge permits. synthesize them into a unified approach The next step is to design.
S

to

9.2

EQUATION FOR PREDICTING MOTIONS At the heart of any displacement-limiting approach to S

design is

an equation

for predicting displacements

in terms of

the specified ground motions and the physical parameters characterizing a gravity retaining wall and its backfill. achieve the objectives of this study, probabilistic in nature. That is To S

such an equation must be is necesary to

to say it

quantify the probability that various amounts of permanent displacement will be exceeded. Equation this purpose. 1.1 was suggested at the outset of this study for In it, the permanent displacement of a retaining
. . . . .. . . . . . . . . . . . . . . . . . . . . . . . . .

_
>--

119

wall is dR v

predicted by the product of five terms: : The permanent displacement of a one-way sliding block with maximum transmittable acceleration Ng, averaged over all earthquake ground motions characterized velocity V. by peak acceleration Ag and peak

R2/1 :

A deterministic factor accounting for the effect of a kinematic constraint ignored when using a sliding block to represent a retaining wall and its fill. backS

0 :

A random variable with mean of unity,

describing

the

variation of permanent displacement aroused by different earthquake motions all having the same Ag
and V.

RO :A

random variable with mean of unity, effect of random uncertainties parameters in

describing the

the resistance

for the wall and backfill. random in nature, which accounts

A model error term,

for as yet poorly understood p The first problem.

aspects of the

four of these factors have been studied in R2 /

1

considerable detail: and RO in Chapter 7.

in

Chapter 5,

dR and Q in

Chapter 6

Methods for accounting

for these several some detail.

aspects of the problem have been developed in However,

p,

several of these methods are rather complicated.

. .. .. :.

"---"''

-....-. *.. -.. .... . -... . . .-. -" '" - ''" "- : -:" -""-

-.. . .- . ... .- .. ... "" - :,,, * ,,

.--. .- ... ',, ;.

.. . . "". . .

. ".

............-.... -.. ..-....".".". "". '.". . .

... .' " "

*. .. .. " '.

. ."

. "

. . . . . . . : ": --' " "- ' ' "' :: :

120

The effect of vertical ground accelerations enters expression for dR, and this causes in

into the

the factors A and N to enter Complicated calcula-

into computations

a complex fashion.

tions are necessary

to arrive at exact values for the ratio influenced by A and N in a non-simple the effect

R 2 /1 , and this ratio is way. Similarly, it is

not an easy matter to evaluate

of deformable backfill, have been suggested In still in

even though guidelines for this purpose Chapter 7. and keeping Equation in mind the

order to retain simplicity,

imperfect state of our knowledge,

1.1 has been

simplified to:

dRw d 37V2 -9.4N/A -ge " Ag The first term is

Q RM

(9.1)(9.1 (9.1).

.. Equation 6.2 for the mean displacement of a ground accelerations. The errors

sliding block with no vertical caused by ignoring vertical by omitting the factor R2/1, term M. N is the expected

accelerations,

and those introduced

are lumped into the model error (in the average sense) threshold

transmittable acceleration

coefficient

for the wall/backfill,

evaluated using average values for the various friction angles. Q, R2/1 and M are all random variables, unity. Based in and in Q and R having means of

part upon available numerical part upon

results (see (i.e. it is

Chapter 6)

the form of Equation 9.1 it is

a product of random variables) lognormally distributed.

assumed that dR is

rr

-.

121

Using results

from the theory of probability and assuming

that Q, R and M are also independent,

it

is

now possible to

write expressions for the mean and stantard deviation of dRw: Mend Mean [dRw] where M is
2

37V e-9.4N/A Ag e

(9.2)

the mean value of M7 and *

94 2 (9) N

a2

nndRW

a2

lnQ deviation for

.M lndR,

Actually,

Equation

9.3 gives

the standard

rather than for dRw directly. deviations for InM and lnQ,

alnM and alnQ are the standard

respectively. in Equations 9.2 and

Information

concerning the parameters Table 9.1.

9.3 has been summarized in for Q are found in

Coefficients of variation unity, the

.

Table 6.37 since the mean of Q is

standard deviation equals the coefficient of variation. computed by: G

= ln(l+VQ2)

OlnQ is S .4

inQ

Estimates for aN were discussed in The model error term M is as indicated

*

Chapter 7. factors,

a composite of several

in

Table 9.1.

Figure 6.4 has been used to estimate Equation 9.1 should have been )) .4 + log Q + log m. The first term in Equation 9.3

In a more proper derivation, written as: in dRw n(

where N is a random variable. is then obtained by employing:

n RN L 3NVar[N]
=

(-

a
N

122

Table 9.1
SUMMARY OF PARAMETERS IN EQUATION 9.1

FOR PREDICTION OF PERMANENT DISPLACEMENTS

Factor Ground motion factor Q Resistance factor N

Mean 1 N

Standard Deviation 0.6 -1.4 0.04-0.065

Log normal Standard Deviation 0.58-1.05

-

MODEL ERROR Vertical acceleration Ignoring R Deform, Tilting backfill 1.2 0.65 3 1.5 0.2 0.2 2 0.75 0.2 0.3 0.6 0.5
.

Combined model error

3.5

3.6

0.84

...

..

..

. . .

. .

.".-

w II

..

,"

123

the mean and standard deviation of the error introduced by failing to take the effect of vertical accelerations account directly. The discussion in into

Section 8.4 has provided 0

the basis for estimating the effect of not accounting directly for the ratio R2 /1. correction Similarly, rather than introducing a

for the effect of deformable backfill as a function a mean correction and a measure of The possible In 5 0

of period ratio and N/A,

scatter have been estimated from Figure 8.4. effects of tilting have been discussel in

Section 8.3.

arriving at these various values,

attention has been focussed A = 0.3 to

upon the range of A and N/A of greatest interest: 0.4 and N/A = 0.3 to 0.7. Given the means and standard deviations effects that enter into the model error term,
this term may be calculated. The mean of M is

for the several the parameters for

-

just the product

of the means.
G2 inM

The standard deviation is

Z

computed using:
(9.5)

(V

+ i)

where Vi is M.

the coefficient

of variation

for the ith factor in Table 9.1

Applying these approaches to the values listed in

under the heading of model error leads to the results M = 3.5 and alnM = 0.84. 9.4 3.6.
Table 9.2 presents results for the standard deviation of S

VM,

and thence aM, can be found using Equation leading to VM = 1.03 and aM = MVM =

(with 0 replaced by M),

indR,

calculated using Equation 9.3 together with values of values for M and OlnM from Table 9.1

OlnQ based on Table 6.3,

S. ....................................................

" '" ............. .''".

'

."".- ..>.........."-"

... '"..." ..

.... .. "

".'....
..

J................

.. ':..

."..... ..

-".--

.'.."..:....'-...

124

Table 9.2 G lnd lnR FOR GRAVITY RETAINING WALLS

N/A 0.1 VQ 02 inQ

21nQ +

0 0.5 1.02 0.71 1.42 0.6 1.17 0.86 1.57 0.7 1.41 1.10 1.81 0

0.2 0.68 0.38 1.09

0.3 0.78 0.48 1.19

0.4 0.88 0.57 1.28

0.63 0.33 1.04

lnQ

2n

mM

0.2 0.3 0.4 A 0.5 0.6 0.7

2.9 2.1 1.7 1.5 1.3 1.3

2.9 2.0 1.6 1.4 1.3 1.3

2.8 2.0 1.6 1.5 1.3 1.3

2.8 2.0 1.6 1.5 1.3 1.3

2.8 2.0 1.7 1.5 1.4 1.3

2.8 2.0 1.7 1.5 1.4 1.4

2.8 2.0 1.7 1.6 1.5 1.4

So

": " <--"'"-" : " : "" :" '

: " > " "":i":"'"

"" "-""

i '""

.-- .:iiS

125

and values of aN interpolated values of A, the first that is to say, term in

from Fig.

7.1.

For the smaller quite dominant: factor N is the 9.3

Equation 9.3 is the resistance

uncertainty in

controlling factor. are more-or-less is substantially

At the larger A, all terms in importance. of N/A,

Equation

of equal

The result is

that alndR

independent

but depends strongly on A.

9.3

APPROACH TO DESIGN USING SAFETY FACTOR AGAINST DISPLACEMENT Taking logarithms on both sides of Equation 9.2 and

rearranging
N

the resulting equation gives:

1 37MV_

in

Agdw

(9.6) if

Equation 9.6 may be solved to find the value of N required the mean residual displacement given dR, in is

to be equal to or less than a According to the argument

when A and V are specified.

section 9.2, If

M should be taken as 3.5.

an engineer could accept having the average residual or less than, a

displacement of many walls be just equal to, specified permissible may be replaced by dL. would, together displacement dL,

then dR in

Equation 9.6

The value of N found from this equation

with average values for soil weights and be used to calculate 4.7). - sin(+6)tan] tan ,b-N
to be applied

friction angles, wall from (see Eq. w

No safety wall.

the required weight of

[cos(+)

(9.7) (9.7
to this weight of

"

factor would need

126

However,

the discussion in

this and previous chapters has

emphasized that often the residual displacement will be much larger than the mean value. The engineer may wish to use a

safety factor so as to reduce the likelihood that the actual displacement wil exceed dL. safety factor F to dL, dL dRw Then Equation 9.6 becomes N A in 137MFV2 _ AgdL 9.4 ( ( One way to do this is to apply a

such that: (9.8)

This equation would be used as before to calculate N and thence wall weight from Equation would be applied to Ww. In the absence of extensive tests or field observations theory of to 9.7. No additional safety factor

use as a basis for selecting the safety factor F,

probability may be used to guide the choice of a suitable value. The engineer must first having dL exceeded. of not exceeding dL. is (approximately) decide upon an acceptable risk for

Let us denote by P the desired probability Since we have already concluded that dRw lognormally distributed, probability theory 6.11): (9.10)

gives the following expression for F (see Eq. F =exp [-_i where Dp is
2

i nd

ain.

a factor dependent upon P, in

as listed in Table 9.3.

The information

Tables 9.2 and 9.3 may then be combined to (Note, in

produce a table giving F as a function of P and A. Table 9.2, that 01ndR is

virtually independent of N/A" an

..

127 Table 9.3 FACTOR D FOR VARIOUS PROBABILITIES OF NON-EXCEEDANCE 0

Probability of Non-Exceedence P-%

Factor D P in Eq. 9.12

50 75 85 90 95

0 0.675 0 1.037 1.286 1.645

..... . .

..

128

average value has been used for each A.) in Table 9.4. Several somewhat noted. First,

The results are given

surprising aspects of this table may be

for P = 50% all of the safety factors are less

than unity.

This result actually reflects a well-known fact the median (point for which is always less is

concerning lognormal distributions:

half of values are smaller and half are larger) than the mean. For the alndRw of interest,

this situation

almost always true for P = 75%. Second, for P = 95% the safety factor first Rw increases (A decreases)
The trend for all

increases

somewhat as alnd
again with for

and then decreases

for F to decrease values of P.

large values alnd

of (lndRwis evident

increasing

Thus,

once

alnd

Rw

Rw has increased past a certain point (which

other

depends upon P), non-exceedence

further increase causes the point for P percent to move back toward (and even past) the mean! for this behavior may be
that uncertainty in dR, in N

The physical situation

explained as contributes smaller A. enormous follows.

that accounts
It

has been noted

very strongly to uncertainty An N smaller than

especially with an

the mean will contribute

number of cases with small displacements,

while an N

larger than the mean implies that a few cases with very large displacement are possible.
decrease in the value

The net effect is

(100-P) percent

actually a
of the time.

exceeded

..-

142

predicting the displacement of a sliding block has been ieloped, 5. and suitable safety factors have been suggested. Based upon this analysis, it is possible to assess the

liability of previously proposed methods for designing gravity taining walls. Use of a seismic coefficient corresponding to one-half of the peak acceleration for the design earthquake, together with safety factors on wall weight in the range of 1.0 to 1.2, gives satisfactory designs for a moderate seismic environment (probability of excessive displacement less than 10% for peak accelerations less than 0.2 g). In a severe seismic environment (peak acceleration of 0.5 g of more), there is generally an unacceptable risk (probability greater than 20%) that walls designed by the seismic coefficient approach will experience permanent deformations in excess of one or two inches. For walls designed by the Richards-Elms approach, with a safety factor of 1.1 to 1.2 on wall weight, there is at least 95% probability that the limiting displacement will not be exceeded. .2 OPPORTUNITIES Even though design of gravity retaining walls is the really major problems in earthquake engineering, not one the

tire class of problems involving prediction of permanent splacements presents one of today's major challenges. ere is considerable value to be derived Overall 0

from further work upon

d understanding problems.

of one of the simpler cases within this class

The opportunities for further work upon the three categories. to

avity retaining wall problem fall in 1. There is

need for further theoretical analysis,

udy the effect of deformability of the backfill upon both

iding and tilting. del is now available A reasonably satisfactory it finite element

for study of sliding;

should be used

" " .

" "

- .

i" "

-,

."

."

"

'.

..

..

"0
141

10- CONCLUSIONS AND OPPORTUNITIES

0.1

CONCLUSIONS The conclusions from this study may be summarized by the 0

ollowing statements: 1. The design of gravity retaining walls against the 0

ffects of earthquakes logically should be based upon a isplacement-limiting 2. approach. is appropriate (but

Use of a sliding block analysis for a gravity wall.

arginally so!) 3.

The basic sliding block model must be modified to

ccount for:
*

The actual interaction between two sliding blocks, one representing the wall and the other the failing wedge of backfill. This effect is reasonably well understood. The effect of the deformability of the backfill prior to failure. This aspect of the problem is now partly understood. Tilting of the wall. understood. This effect is as yet poorly the design of 0 6

4.

The choice of a safety factor for use in

wall for the seismic loading case should be based upon the ,robability that a limiting permanent displacement will be
xceeded.
*

It

is

necessary

to consider:

Variability

of ground motions

* *

Uncertainty in

resistance

parameters in the model used as

S

Approximations ("errors") a basis for computation.

.n analysis has been made of these considerations, !urrently best available data and information.

using the (9.1)

A new equation

,.-.:... ....

..-..-. . .,.... -..... ..

... ...

,....... ..

. ...- :..

..-

.- _. . ..-

. .:

. ..

:::::.::.:

140 EXAMPLE 9.2 Given: A wall designed for A = 0.3 and safety factor of 1.1 on wall weight for seismic case, using = b = 300 and 5 = 200. The height is H and the unit weight is y. Probability that permanent displacement will exceed 1, 2 or 4 inches, during earthquake with A = 0.3 and
V = 15 inches/sec.

Find:

Solution:

The first
3.2b,

step is

to design the wall.

K

From Eq.

with N = 1 A = 0.15;

From Eq.

9.7:

2 Ww = 0.707

AE

= 0.407

0
2

1 yH

Applying the safety factor:

Ww = (1.1) (0.707)fYH
= 0.778 1 yH2

Using Eq.

4.4 iteratively the actual N for this wall is Eq.

0.171.

For the actual earthquake,

9.2 gives dRw =1.19 inches. "

The following table gives results for various d L From Table 9.2, the appropriate value of alndRwfor AL= 0.3 is 2.0.

d
dL F
=

=
0.84

in

=2 in
1.68

=4 in
3.36

Rw

Dp P[dRw<a P[dRw>d L

1.09 R L0.82 0.18

1.26 0.90 0.10

1.61

Eq.

9.12

0.95 0.05

Standard Probability Table Tal

.....

".-.

..........

139 EXAMPLE 9.1 Given: Find: Solution: The requirements in Example 4.3.

Appropriate weight of wall using approach of Chapter 9. W will be computed for two values of F. w

F = 2.5

F = 4

N/A N0.20

0.68

0.73 0.22

Eq.

9.13

'11.30
KAE 0.473

12.40

Eq.

3.1

0.491

Eq.

3.2b

PAE

17.7 k/ft 46.9 k/ft

18.4 k/ft 51.5 k/ft

Eq. 3.2a

Eq.

9.7

These weights may be used directly to proportion the wall: no additional safety factor is required.

.,'

..

..

..

..

..

.-

138

shearing exert

through

the backfill. force

In

addition, or less

pore water may independently of

a dynamic

on a wall more

the mineral

skeleton.

These are matters

for future

research.

-S.

Si
S ..

... S-.

o S

,---

'-.....-...............-.......--...........-...-..................-.....-

."."....'...-..-."""-..-.--..--.-'.....-...-".

137

displacement of such a wall actually results from the type of deformation pattern envisioned by a sliding block type of analysis. apparent Even here, however, there are problems, such as the

importance of "elastic" deformations Section 8.3. Indeed, when it

of the backfill, becomes necesary

as discussed in

to use a multiplicative account

correction

factor of 3.5 (primarily to one must think being pushed

for elastic backfill effects and tiltng),

that the usefulness of the sliding block method is

tu its limit.

Clearly more research is of tilting

required concerning

the importance It

and of the "elastic" deformability of backfill. to conduct experiments in which these

will be very important

effects are properly simulated. stress-strain behavior in

Having reasonably correct

.

the soil used for these experiments

will be critical.

Thus model tests should be carried out on a normal gravity. this report deserv-

centrifuge rather than in

Two other matters not dealt with in mention. a wall. One is

the effect of passive resistance at the toe of into is

In principle this resistance might be incorporated Eq. 4.2a. However, it

the analysis as an additional term in

well known that some displacement must occur before full passive resistance is developed, and such a situation cannot be modelled Great care should be used in if passive toe resisto sliding.

well by a rigid-plastic model.

applying the methodology of this report tance is

a significant part of the total resistance

The other effect is backfill.

the influence of pore water within the to

= " , -

Pore water of course can influence resistance

-'m % . -" = ' Imm m- ' ~ " " "" " , -

136

There is incomplete.

one aspect of the probabilistic analysis which is Eq. 6.2 for predicting permanent displacement of a

sliding block was developed using data from earthquakes with magnitudes from 6 to 7. Presumably larger earthquakes, causing 5

ground motions of greater duration, displacements for the same A and V.

would result in greater It would be desirable S

to repeat the analysis leading to Eq. from larger earthquakes. In

6.2 using ground motions

the years since Newmark's paper suggesting the use of there have been numerous efforts to
0

sliding block analysis,

apply the method to predicting permanent displacements of earth structures as the result of earthquakes. suggested use of the method in Newmark originally and the As a
-

connection with slopes, in fact, been to slopes.

bulk of the applications have, conceptual aid to understanding deformations and in

the evolution of permanent the the

the development of general guidelines, However, in

sliding block has indeed been very valuable. case of earth dams it

appears that a significant amount of 0

permanent displacement may result from distortions distributed throughout the dam, before a definite failure surface develops.

Thus the quantitative use of the sliding block analysis as a tool for predicting the permanent deformation of earth dams may be limited. Sliding block analysis has always seemed more suitable for the analysis of gravity retaining walls, very litle since it usually takes
.

outward motion to develop a failure condition within That is

.' ." . . .. "

the backfill.

to say,

the observable permanent

"" "" ' " '" " -'

. , - 2 - * " ," - '- " " " ' " ," ," ", " " " .- -

135

seismic environments, inches,

several different dL ranging from I to 4

and several values of safety factor applied to the The results of these calculations may
0

calculated wall weight. be summarized as follows:

Safety factor applied calculated wall weight 1.0 1.1

1.2

Probability that dL used for design will be exceeded if design earthquake occurs -10%
~ 5% < 5% 0

If

5% probability-of-exceedence

is

taken as a target,

these

results justify the use of a safety factor of 1.1 or 1.2 on wall

weight in conjunction with the Richards-Elms procedure.
-

9.5

GENERAL DISCUSSION The methodology developed in this report has focussed

-

primarily upon predicting the probability that a retaining wall will experience various amounts of permanent displacement Quantifying uncertainty requires good

during an earthquake.

knowledge of the several aspects of a problem and adequate statistics for the pertinent parameters. Obviously there is yet

not enough such information to do with great confidence. theless, the estimates developed in

None-

this report using the best

available information appear to provide good guidance for

purposes of design.

The methodology itself

should remain useful

as additional data are developed,

and may be applied to other is appli0

types of problems for which a sliding block analysis cable.

-... -...-. .......-... .,..,..: .. .. .-. ... ........? .,... . .,..- .., ?:

. ...-.-. -< < , - ........- ..,.- ,.-.-....-. .......... .-,-,. .,

134

Table 9.5 PROBABILITY THAT MOVEMENT OF WALLS WILL EXCEED VARIOUS LIMITING VALUES. WALLS ARE DESIGNED FOR STATIC SEISMIC COEFFICIENT OF 1/2 OF PEAK ACCELERATION = 300 and 6 = 200 WITH 0 =

FACTOR ON
WALL WGT. SEISMIC CASE

SAFETY

dL
L inches 1

CHARACTERISTICS

OF EARTHQUAKE
A = 0.3 V=15 in/s 28% 18% 10% 18% 10% 5% 12% 6% 3%

A = 0.2 V=10 in/s 10% 7% 4% 5% 3% 2% 2% 2% 1%

A 0.5 V=25 in/s 56% 38% 22% 47% 30% 16% 40% 23% 12% S S

1.0

2 4 1

1.1

2 4 1

1.2

2 4

S 1
. .. .

.-.

133

coefficient for this wall is

N = 0.71.

The actual earthquake

is

assumed to have the same peak acceleration as the "design earthquake in/s. (i.e. a = 0.3), together with a peak velocity V = 15 of the actual

Note that the results are independent

height of the wall and of the actual weight of the backfill. Table 9.5 summarizes results for walls designed in using several different design (and actual) this way

safety factors nad several different Similar results were obtained

earthquakes.

using other combinations for 4, parameters

*b

and 6.

(Changing

these

changes the wall weight,

but has little

effect upon For a design

the N for a wall just meeting moderate seismic environment will have reasonably ments. However,

the design criteria.) (A = 0.2),

a conventional

low probabilities of excessive displace-

in a severe seismic environment the probability much larger, at least for safety

of excessive movements is factors of 1.0 to 1.2. In

computing these results,

it

has been assumed that If

average values of

*,

*b and 6 are used for design of a wall. design,

conservatively low values are used in

the probability of

excessive movements will be less than suggested by these calculations.

9.4.2

Design Following Richards-Elms With the Richards-Elms method, the required weight

of wall depends upon the allowable displacement dL as well as upon the seismic.. environment. A number of calculations have b and 6various

been made, using various combinations of C

...............

..

...

..

,..

132

2.

The corresponding

value of N is

computed,

using one of

the techniques discussed in 3. The value of dR is parameters 4. computed,

Section 4.3. using Eq. 9.2 and the

appropriate for the earthquake of concern. calculated for the selected threshold 0 9.12.

The factor F is

permanent displacement. 5. The factor Dp is calculated by inverting Eq. Rw 9.2 6. The probability of exceedance corresponding to Dp is looked up in a standard normal cumulative probability table. . is

The appropriate value of Tlnd

selected from Table

9.4.1

Conventional Design While there are many different "conventional" ways

to select a seismic coefficient

for the design of gravity

retaining walls (se Chapter 3),for purposes of illustration this coefficient will be taken as one-half of the peak acceleration in the earthquake used to define the seismic threat. Example that 0

9.2 outlines a sample calculation of the probabilities various levels of displacement will be exceeded, case. Note that the wall of 0.3q, is

for a specific

"desiqned" for an earthquake with a _

peak acceleration (1/2)(0.3)

=

using a seismic coefficient N =

0.15.

Because of the safety factor applied to the the actual threshold acceleration

calculated wall weight,

-9

131

9.3.2

Examples Using M
=

0 3.5,
1 in

Equation 9.11 now becomes:

Agdi A F = 4

[0.66

(9.11a)

9.
0.61 + 1 in

AgdL(.1a
A F = 2.5 (9.11b)

The factor 37,

M and F have all been combined in

the first

terms

of these equations. To illustrate the use of Equation 9.13, Example 4.3, developed in with Nv = 0 which is reworked in let us return to
S

Example 9.1 using the results 3.1 and 3.2 are used

this chapter.

Note that Eqs.

the possible effects of Nv * 0 have been accounted this chapter. The wall weights computed
S

for by the analysis in

using both values of the safety factor are larger than that determined in However, in Example 4.3 before any safety factor was applied. _

the current example no additional safety factor need and hence the design is more

be applied to the computed weight, economical.

9.4

RELIABILITY IMPLICIT IN OTHER DESIGN APPROACHES The results developed in Section 9.2 may be used to

estimate the probability that walls designed by a conventional approach, or using the Richards-Elms method, will P-: .rience as

"more than a given value of movement.

follows: 1.

The methodology is

The wall weight required by the approach, safety factor, is calculated.

including any

130

9.3.1

Choice of Safety Factor There is no standard to guide the choice of a suitable In the case of buildings, the probability of not

value for P.

failing during a major earthquake, building codes, P = 99%). appropriate P = 90%. It apparently is

as implied by good modern

less than 1% (corresponding to

A somewhat lower probability of nonfailure seems for gravity retaining walls: say P = 95% or even about as far as

should also be noted that P = 95% is concerning the distribution This is

the information

of dR may

comfortably be pushed.
even beyond the limit)

certainly the upper limit (perhaps

in the statistical analysis

of confidence

of computed displacements Referring

in

Chapter 6. with P
=

to Table 9.4,

95% the safety factor F

should be about 3.6 to 3.9, except for the smallest ground accelerations where a smaller value is design of gravity walls is little In justified. However,

affected when expected ground the interests of simplicity

accelerations are this small.

a single factor of 3.8 might be used at all levels of ground acceleration. reasonable if Similarly, use of a safety factor of 2.4 seems 0

a somewhat less conservative design corresponding

desired.

to P = 90% is

Recognizing that one characterization of dR is

of the distribution one

uncertain at such higher levels of non-exceedence, the choice of safety or F

-

should not attempt to be too precise in factors. Hence values of F are recommended.

" _

4 (conservative)

2.5 (less

conservative)

-i- i,

--- "-

_-'''-

L~-+L JL t.

'-'

JL .L

A'J

'-X

-- ]--'

_--_

.'-

.","

"

"

"

'"

-+.

129

Table 9.4 SAFETY FACTORS REQUIRED FOR VARIOUS PROBABILITIES OF NON-EXCEEDANCE, AS A FUNCTION OF PEAK GROUND ACCELERATION

Probability of non-exceedance - % A 0.2 0.3 0.4 0.5 0.6 0.7 lndpw 2.83 2.01 1.66 1.50 1.36 1.33 50 0.018 0.13 0.25 0.32 0.40 0.41 75 0.12 0.52 0.77 0.89 0.99 1.01 85 0.34 1.07 1.41 1.54 1.63 1.64 90 0.69 1.44 2.13 2.23 2.28 2.28 95 1.92 3.62 3.87 3.83 3.72 3.68

%S

I.S

- --

_ .-.

-..

-.

I-

"143

to study the influence of varying the several parameters. Further improvement of the model likely will be necesary for study of tilting: for this case it seems essential at least to 0

have multiple parallel

failure planes through the backfill.

Modelling the resistance of the soil beneath the wall to tilting is a very poorly understood problem. 2. Valuable results can come from model tests of gravity It is essential that the wall

walls carried out on a centrifuge. be free-standing; will. i.e.

free to move relative to the soil as it 0

A program of tests involving sliding has been underway at in England, but results are not yet tilting

Cambridge University available.

Additional tests to explore the complex

problem should be even more valuable. 3. The analysis of uncertainty should be extended. A

primary need is earthquake

to bring the influence of the magnitude of the

into the equation for predicting the displacement of

a sliding block.

. . .

.-.-

.... ........................ ..... . .

r -r

-"

v .

ri k

k-

-.

144 REFERENCES Antia, H. 1982. "Predicting Relative Displacement for LimitedSlip Seismic Design of Gravity Retaining Walls with Non-Liquefying Backfills," S.M. Thesis, Research Report R82-34, Department of Civil Engineering, Massachusetts Institute of Technology. ATC-3-06. Jun 1978. "Tentative Provisions for the Development of Seismic Regulations for Buidings," Applied Technology Council, Report No. ATC-3-06, Palo Alto, Calif. Bolton, M. D., and Steedman, R. S. July 1982. "Centrifugal Testing of Microconcrete Walls Subjected to Base Shaking," Proceedings, Conference on Soil Dynamics and Earthquake Engineering, Southampton, Vol 1, pp 311-329. Evans, G. L. 1971. "The Behavior of Bridges Under Earthquakes," Proceedings, New Zealand Roading Symposium, Victoria University of Wellington, Wellington, New Zealand, Vol 2, pp 664-684. Franklin, A. G., and Chang, F. K. Nov 1977. "Earthquake Resistance of Earth and Rock-Fill Dams; Report 5, Permanent Displacement of Earth Embankments by Newmark Sliding Block Analysis," MP S-7117, Soils and Pavements Laboratory, US Army Waterways Experiment Station, Vicksburg, Miss. Jacobsen, P. N. 1980. "Behavior of Retaining Walls Under Seismic Loading," M.E. Report 79/9, Department of Civil Engineering, University of Canterbury, Christchurch, New Zealand. JSCE. 1977. "Earthquake Resistant Design for Civil Engineering Structures, Earth Structures and Foundations in Japan," Japan Society of Civil Engineers. Lai, C. S. 1979. "Behavior of Retaining Walls Under Seismic Loading," M.E. Report 79/9, Department of Civil Engineering, University of Canterbury, Christchurch, New Zealand. Makdisi, F. I., and Seed, H. B. Dec 1979. "Simplified Procedure for Evaluating Embankment Response," Journal of the Geotechnical
Engineering Division, ASCE, Vol 105, No. GTI2, pp 1427-1434.
-

0.

Mayes, R. L., and Sharpe, R. L. Oct 1981. Guidelines for Highway Bridges," Report No. Federal Highway Administration, Washington, Mononobe, N., and Matsuo, H. Pressure During Earthquakes," Conference, Vol 9, p 177.

"Seismic Design FHWA/RD-081-081, DC.

1929. "On the Deformation of Earth Proceedings, World Engineering .0

Murphy, V. A. 1960. "The Effect of Ground Characteristics on the Aseismic Design of Structures," Proceedings, Second World Conference on Earthquake Engineering, Japan, Vol 1, pp 231-247.
. . ,. .*

145 Nadim, F. 1980. "Tilting and Sliding of Gravity Retaining Walls," S.M. Thesis, Department of Civil Engineering, Massachusetts Institute of Technology, Cambridge, Mass.
1982. "A Numerical Model for Evaluation of Seismic

. 0

Behavior of Gravity Retaining Walls," Sc.D. Thesis, Research Report R82-33, Department of Civil Engineering, Massachusetts Institute of Technology, Cambridge, Mass. Nadim, F., and Whitman, R. V. July 1983. "Seismically Induced Movement of Retaining Walls," Journal of Geotechnical Engineering, ASCE, Vol 109, No. 7, pp 915-931. NAVFAC. May 1982. "Foundations and Earth Structures, DM-7.1 and DM-7.2," Department of the Navy, Naval Facilities Engineering Command, Alexandria, Va. Newmark, N. M. 1965. "Effects of Earthquakes on Dams and Embankments," Geotechnique, Vol 15, No. 2, pp 139-160. Newmark, N. M., and Hall, W. J. 1982. "Earthquake Spectra and Design," Earthquake Engineering Research Institute, Berkeley,
Calif.

Okabe, S. 1926. "General Theory of Earth Pressure," Japanese Society of Civil Engineers, Vol 12, No. 1.

Journal, .

Ortiz, L. A., Scott, R. F., and Lee, J. 1981. "Dynamic Centrifuge Testing of a Cantilever Retaining Wall," Soil Mechanics Laboratory, Division of Engineering and Applied Sciences, California Institute of Technology, Pasadena, Calif. Richards, R. J., and Elms, D. Apr 1979. "Seismic Behavior of Gravity Retaining Walls," Journal of the Geotechnical Engineering Division, ASCE, Vol 105, No. GT4, pp 449-464. Seed, H. B., and Martin, G. R. May 1966. "The Seismic Coefficient in Earth Dam Design," Journal of the Soil Mechanics and Foundations Division, ASCE, Vol 92, No. SM3, pp 25-59. Seed, H. B., and Whitman, R. V. 1970. "Design of Earth Retaining Structures for Dynamic Loads," ASCE Specialty Conference,

"

Lateral Stresses in the Ground and Design of Earth Retaining Structures, pp 103-147. Sherif, M. A., Ishibashi, I., and Lee, C. D. Jan 1981. "Dynamic Earth Pressures Against Retaining Structures," Soil Engineering Research Report No. 21, University of Washington, Seattle, Wash. * USACE. Jan 1965. "Retaining Walls, EM 1110-1-1401 Change 3," Department of the Army, US Army Corps of Engineers, Washington, DC.
-

AD

146 USACE. Apr 1970. "Stability of Earth and Rock-Fill Dams, EM 1110-1-1902," Department of the Army, US Army Corps of Engineers, Washington, DC.
Feb 1982. "Stability of Earth and Rock-Fill Dams,

EM 1110-1-1902 Change 1," Department of the Army, of Engineers, Washington, DC.

May 1983.

US Army Corps

"Earthquake Design and Analysis for Corps

of Engineers Projects," ER 1110-2-1806, US Army Corps of Engineers, Washington,

Department of the Army, DC.

S

Wong, C. P. 1982. "Seismic Analysis and an Improved Design Procedure for Gravity Retaining Walls," S.M. Thesis, Research Report R82-32, Department of Civil Engineering, Massachusetts Institute of Technology. Zarrabi-Kashani, K. 1979. "Sliding of Gravity Retaining Wall During Earthquakes Considering Vertical Acceleration and Changing Inclination of Failure Surface," S.M. Thesis, Department of Civil Engineering, Massachusetts Institute of Technology, Cambridge, Mass.

. .

. . .

..

. .

. .

..

..--

.-

. . . ...

...

..

147

LIST OF SYMBOLS

A AH Av
a

Maximum horizontal ground acceleration coefficient of an earthquake record (max. acceleration is Ag). Horizontal ground acceleration time. coefficient: varies with

=
=

Vertical ground acceleration coefficient.

Ground acceleration.

aT Cs Dp

= = =

Transmittable

(or threshold,

or limiting)

accelerations.

Shear wave velocity. Standard normal deviate (i.e., number of standard devitions from the mean) corresponding to the probability P of non-exceedance of a given parameter (e.g. residual wall displacement). Allowable residual displacement Residual displacement calculated using the rigid single block model. Also, the residual displacement calculated using the Richards-Elms equation or one of
.-

dL dR

Newmark's equations.
AdR =

Predicted residual

displacement

for 2-block

model.

dRe

Residual displacement of a rigid single-block model; when no vertical component of an earthquake record is taken into account. Average of dRe over a number of records. Average residual displacement of a rigid single-block model; averaged over 4 orientation directions as shown in Fig. 6.1 for a single earthquake record. Residual displacement of a sliding rigid single-block model, when vertical component of an earthquake record is taken into account. Average of dRy over a number of records. Residual displacement of a retaining wall. Expected value of dRw.
S ..........................................................

URe dRo

= =

dRy

dRy dRw dRw

= = =

148

E[]

The expected value operator, with the quantity inside the brackets being the operand: equivalent to calculating the mean of the operand. Correction
account

factor to block model slip calculations,

for wall orientations.

to

independently

A random

variable. Es
=

Correction factor to block model slip calculations, account independently for earthquake record variability. A random variable.

to

Ev

Correction factor to block model slip calculations, to account independently for vertical earthquake motions. A random variable. Average value of Ev at a given N/A. Wall weight factor,
weight.

Ev Fww fE-Q fBF g i KA KAE Ko H M

= =

a non-dimensional

indicator of wall
.

= = = =

Central frequency of earthquake. Natural frequency of backfill.

Constant of gravitational acceleration. Angle of inclination of the backfill with respect to horizontal. Coefficient of active earth pressure (static).

= =

Coefficient of active earth pressure due to earthquake (includes static and dynamic effects). Coefficient of earth pressure at-rest. S

= = =

Height of the wall. Correction factor, accounting for various modelling Also, mass of a errors of the single block model. sliding block. Expected value of correction factor M.
Mass of the backfill soil.

M
m Mw m

=
= = =

_S

Mass of the wall. Mean of a quantity the particular quantity would be

_

indicated by a subscript to m, e.g.

mQ,

mN,

etc.

. . . .. . . . . . .. . ..

149 N Maximum transmittable horizontal ground acceleration coefficient; also seismic coefficient depending on the context.
Expected value of N.

NH

Transmittable horizontal ground acceleration coefficient, variable with time. Also, the seismic coefficient depending on the context. Seismic coefficient number from Fig. 3.6. for

No NR Nv NT P PAE

= =

Threshold transmittable acceleration coefficient rotational mode of failure. Transmittable vertical acceleration variable with time. Threshold transmittable acceleration coefficient coefficient.

=Probabilityi

probability of non-exceedance.

Total active thrust due to dynamic plus static earth pressure.

(PAE)H= (PAE)V= Q RE
R2/
=

Horizontal component of PAE Vertical component of PAE. for random nature of factor, accounting

Correction factor, accounting earthquake shaking.

Residual displacement amplification for effects of elastic backfill.

Correction factor, ratio of 2-block

to 1-block model

displacement. R
=

Correction factor, accounting for uncertainties parameters characterizing backfill, wall, and foundation soil.
Ratio of residual displacements

in

Re

of 2-block models using

fixed e vs. TBF

=Natural

variable 0 in

analysis.

-9

period of backfill.

TEQ=
t
=

Central period of earthquake.

Time.

"V
Vo

=
=

Maximum ground velocity of an earthquake record.

Coefficient of variation of Eo 0

,=.: .-

'...-

. -.-........ -.. -........ ..................... . ... -..

...'-.

' :..-..

*.'

..

.'.%-.-.-."

.'.

150 VR
=

Combined coefficient assuming Eo, Es,

of variation of Eo,

Ea, and Ev

and Ev independent.

VR* Vs Vv

= = =

Coefficient of variation of RO. Coefficient of variation Coefficient of variation The variance operator, of EB. of Ev. .
S

Var[] =

with the quantity inside the

brackets being the operand; equivalent to W

= =

Weight of a sliding block. Angle of inclination of the back of a retaining wall with respect to vertical.
Unit weight Unit weight of soil. of concrete.

= = =

Y 6
S=

Friction angle between the back of the wall and the backfill soil.
Friction angle of soil backfill.

Friction angle between the base of the wall and the foundation soil. Coefficient of friction.

= = =

Unit mass of concrete.

Ps
a

Unit mass of the backfill soil. Standard deviation of a quantity; the particular quantity would be indicated by a subscript to a, e.g. aQ, aN, etc. Angle of inclination of failure plane with respect to horizontal. Threshold angle of inclination of failure plane withS respect to horizontal.
S

e
OT

SEquivalent angle of tilt

analysis

to transform a dynamic into a pseudo-static analysis.

.-.

151

APPENDIX A CATALOGUE OF STRONG-MOTION EARTHQUAKES CHOSEN FOR

-

STATISTICAL SLIDING BLOCK ANALYSES

0-i i

...

...........................

152

1 %0rl%

%Q

E-4Lnr* E-4 1 0 1

%iC O oo

1-

'

'C4'

zI

I
e

I 4-I I4

r-4.
-P

9-4O O
1-1
-

zJ
I4 I 0I I I I I: Hd I

O I.~41

01

00-.

Jj
di cc VI wW 0 co 0O, co 00 0x .

0A

dQ

I i4

c .

& d414 (a 41 coL

- )41
w $a 00 Aj a

CtoS

-4
00j, m 4
.0

zS

IAU S
0.,a.) I 4to fe 4)

lU)01 I.4~A I I.44

0 4 i Ci H 4 U) 00J- = go Ai o

C >) l) 0 &J:1-AI

p0

l .

c
k

0 0 1: vaa 0

w 0- PW
C P-4 0

P..
0.

10 I4 AJ r.

c'r. a* to4~U ov $4 w U)

-4
-ImL

OZ

I) C U 0.,v4 .i q4 9id. aO,U I.t0JvA c 0&J to~. 0

40 4 U 0d

0 . o4

I4 I

001. 0

J4

S:

OH~c E-4

I I)

~
P-

Id000)~ A)00 0 00

~
U

0 * 0.

W CA cc

I I

aM 00

I 1

O4

o.4

t4c