SOILS AND FOUNDATIONS Vol.31, No.1, 47-59, Mar.

1991
Japanese Society of Soil Mechanics and Foundation Engineering

A PROBABILISTIC ANALYSIS OF SEISMICALLY

INDUCED PERMANENT MOVEMENTS
IN EARTH DAMS

S.K.PAL 1>, M.S.RAHMAN 11 ' and C. C. TuNG 111 >

ABSTRACT
The earth dams built with compacted cohesive soils or dense sandy materials are vulnerable
to seismically-induced permanent displacements which in some cases may be excessive.
Analysis of seismic response involves an uncertainty inherent in the specification of ground
motion. In this study, a probabilistic analysis is developed for evaluating seismically-
induced permanent displacements in earth dams. The ground motion is modelled as a sta-
tionary random process which is defined by the Kanai-Tajimi spectral density function. The
dynamic response of the dam to vertically propagating shear waves is formulated in frequency
domain. The nonlinearity in shear modulus and damping is incorporated by using an equiv-
alent linear model for the stress strain behavior of soils. The permanent displacement in
dams is assumed to be in the form of the cumulative movements of a block of soil mass
<fetached from the body of the dam. Numerical results of an example analysis and a param-
etric study are presented.

Key words : dam, deformation, earthfill, earthquake, earthquake resistance, probability, slip
surface (IGC : E 8/E 13)

and stability of dams and embankments

INTRODUCTION during earthquakes. The permanent displace-
The study of the performance of earth ment caused during shaking is now consid-
dams built with compacted cohesive soils and ered a more reasonable criterion of perform-
dense sands during earthquakes shows that ance than factor of safety based on limit
the accumulation of permanent deformations equilibirium principles. Various methods of
due to cyclic loading may be excessive evaluating. the permanent displacement/defor-
leading to their failure. mation have been developed (1, 2, 3, 4, 5, 6).
In past decades significant advances have A probabilistic method of evaluation of the
been made towards the analysis of response permanent displacement has also been con-

1> Graduate Research Assistant, Dept. of Civil Engineering, Box 7908

N.C. State University, Raleigh, NC 27695-7908, USA.
li) Associate Professor, ditto.
m> Professor, dirto.
Manuscript was received 'for review on May 8, 1989.
Written discussion on this paper should be submitted before October 1, 1991, to the Japanese Society
of Soil Mechanics and Foundation Engineering, Sugayama Bldg. 4 F, Kanda Awaji-cho 2-23,
Chiyoda-ku, Tokyo 101, Japan. Upon request the closing date may be extended one month.

This is an Open Access article under the CC-BY-NC-ND license.

NII-Electronic Library Service

 48 PAL ET AL.

sidered (7). where w is component frequency. The three

In this study, a simple fully probablistic parameters s 0 , w~ and '~' defining the power
model is developed for the evaluation of the spectrum are estimated from chosen sample
expected value of the random permanent time history. Two estimation procedures
displacement and its distribution. The darn have been used, i.e., the least square fitting
is modelled as a non-homogeneous trapezoidal technique and the spectral moment estimation
soil mass. The earthquake induced ground method (12).
motion is considered to be a random process. The Kanai-Tajirni power spectrum thus
The uncertainties associated with soil proper- obtained represents a random motion with
ties and the failure mechanism are certainly the chosen time history as one sample. This
important but are not included in this study. is stationary however and therefore only a
segment of this motion (having the same
RANDOM SEISMIC ENVIRONMENT central frequency and the same intensity as
in the original nonstationary motion) is used.
Details of input motions have significant
In summary, the random ground motion is
effects on the response of soil masses. This
defined by (a) Kanai-Tajirni spectrum with
is illustrated by the wide range of variability
in the response reported by different research- parameters s0 , w~, and c~ and (b) the equiv-
alent stationary duration, S. Lai (13) has
ers (8, 9). The results of some observed
estimated above parameters of the model by
behavior of an earth darn (10) also suggest
processing 120 recorded motions each treated
that the response of such systems may vary
as a sample time history.
greatly due to variation in frequency content
and other details of earthquake excitation.
Therefore, the use of a single time history
DYNAMIC RESPONSE
of motion is not adequate and may yield
erroneous results. The earthdarn is idealized as a trapezoidal soil
The ground motion is a random process of mass of unit thickness with shear modulus
finite duration with time varying frequency varying with vertical condinator z (see Fig.1)
content. For simplicity and ease of analysis, and its response is assumed to be caused by
earthquake motion is considered to be station- the vertical propagation of shear waves from
ary random process (with constant frequen- the underlying formation. The equivalent
cy content) truncated for a finite duration. linear model is used for the stress-strain
In this study an approach suggested by relationship of soils.
Vanrnarcke and Lai (11) is used in which a
chosen time history of recorded ground motion Equation of Motion:
is considered as one sample of the entire A trapezoidal mass of soil (representing an
ensemble representing the random process. earth darn) supported by a half space is shown
The varying function is a box car function in Fig.l. Vertical propagation of shear wave
described by an equivalent stationary dura- through the darn will cause only horizontal
tion. The Kanai-Tajirni spectral density displacement, u(z, t) which must satisfy the
function is used to characterize the frequency following equation of motion
content of the strong earthquake ground
motion. This function for the horizontal
ground surface acceleration SuaCw) has the
following f orrn:

[1-( =~ YJ+4c/( :g y
( 1) Fig. 1. Seismic response of Earth dams

NII-Electronic Library Service

 PROBABILISTIC ANALYSIS OF EARTH DAMS 49

p f)2u =G(z) [ f)2u_+l_ ~] The following boundary conditions are used

fJz 2 fJt 2 z fJz to evaluate the constants C 11 C 2, Eb, andFb in
_J_[ 2
J
+ r; fJt f)fJzu2 +1_z f)u
f)z
h<z<H ( 2)
the above equations.
At the crest of the dam:
where p, r; and G(z) are respectively density, fJu
G(z) f)z (z=h, t) =0
viscosity and shear modulus of soil, and H
is height of dam (see Fig. 1) At the base of the dam:
On the basis of field measurements, Oner u(z=H, t) =ub(z=O, t)
(17) recommended the following variation of
-r:(z=H, t) =7:'b (z=O, t). (7)
shear modulus with depth
where -r: is used to denote shear stress
G(z) =GsmCz/H) 112 (3)
At the base outcrop:
where Gsm is the maximum (spatial) shear ilb (z = 0, t) = 1. etwt
modulus at the base of the dam. Harmonic
-r:b(z=O, t) =0.
displacement u(z, t) with frequency (J) satisfy-
ing the above equations can readily be found
as: Transfer Functions for Response:
The transfer functions for various response
u(z, t) =Z- 114 {CJ1;3[ARGT]
variables such as displacement u, acceleration
+C2Y1;3[ARGT]}etwt, (4) il, shear stress 7:', and shear strain r, due to
where unit ground acceleration at the flexible rock
ARGT= (4/3)Ksm*H 114 z 314, base can be readily obtained. That is

where, c1 and c2 are constants to be evaluated Hu(z, i(J)) =u(z, t)/ilb(O, t) =fz/((J) 2hzH)
from the boundary conditions, J 113 and Y1; 3 Hu (z, i(J)) =il(z, t) /ilb (0, t) =- fz/ (hzH)
are Bessel Functions of the first and second Hr(z, i(J)) =r(z, t)/ilb(O, t) =Yz/((J) 2hzH)
kind of one-third order and Ksm * is the com- H~ (z, i(J)) = -r:(z, t) /ilb (0, t)
plex wave numbers defined as:
=Gsm*(z/H) 112 Hr(z,i(J)), (8)
Ksm *P(J) 2/ (Gsm +i(J)r;) = P(J) 2/Gsm * = ((J)jv*) 2
where
(5)
In the above, Gsm * is the complex shear fz=z- 114 { -KBESJ1/3[ARGT]
modulus, and v* is the complex shear wave + Y1;3[ARGT]}
velocity. Since the modulus, G(z) and the Yz= -KBEs {- (1/4)z- 514 J 1; 3[ARGT]
damping ratio, /3, are independent of frequen- + z-1/2 H114 Ksm *J1;a' [ARGT]}
cy over the range of interest in the anal-
+ {- (l/4)z- 514Y1; 3 [ARGT]
ysis, the complex shear modulus is taken as
GsmC1+2i{3). +z- 112 H11 4Ksm * Y1t/[ARGT]}
The solution to the wave equation governing h::H=H- 114 {KBEsJ1;3 [ARG]- Y1;3[ARG]}
the horizontal displacement in the horizontal -i {Gsm *I (Kb *Gb *)}
base formation (Fig. I) is given by X {KBEs {- (l/4)H- 514J 1; 3[ARG]
ub (z, t) =Ebexp {i (Kb *z+(J)t)} +H- 114 Ksm *J ;/[ARG]}
1

+Fbexp {-i(Kb *z-(J)t)}, H<z<oo - {- (1/4)H- 514 Y1; 3[ARG]

(6) +H- 114Ksm*Y1;/[ARG]}
where the first term represents the incident {- (l/4)h- 514 Y1; 3 [ARGh]
wave travelling upward and the second term +h-112 H114Ksm * Yt;a[ARGh]}
represents the wave reflected downward; and KBEs= {- (l/4)h- 514J 1; 3 [ARGh]
+h- 112H 114 Ksm *J1; 3[ARGh]}
Kb * is the complex wave number in the half
space. where J 1; / and Y1; / =derivatives of J 1; 3 and

NII-Electronic Library Service

 50 PAL ET AL.

¥ 1; 3 , Kb* and Gb*=the complex wave number the quantity enclosed in the brackets. The
and the complex shear modulus in the flexible expected total number of peaks per unit time,
baserock.
For reasons to be made clear later, consider E[M J=-1 ~ ioo (1)4S:c((l))d(l)] (13)
a simple geometrical shape (hatched zone in
Fig. 1). The transfer function of the accel-
:& 2;r lJ:oo (J)2Sx((l))d(l)
eration of the block in the dam can be evalu- The probability density function of peaks, Xp,
ated from the following: for a stationary Gaussian zero-mean random
izs H(z, i(l)) (b 0 /H)zp dz process x is
fxp ()..) = (2 ;r) -I/20'r-1 (1- ~-2) 112
H;;B=H;; •.(z., iw) .[' B(b,/H)zp dz
exp {-A 2 [2 0' :c 2 (1-~- 2 ) J- 1}
(9) + c2c1/)- ~x 1

112
2~- 2 -2)-
where b0 is as shown in Fig.l.
X {1+erf( ::c }exp( ;;: )
Similarly, the relationship between spatial
average shear modulus Gav and spatial maxi- (14)
mum shear modulus, Gsm can be evaluated where, 0' :c is the standard deria tion of x,
as:

in Gsm(z/H) 112 (b 0 /H)zdz

is bandwidth parameter and
iH (b 0/H)z dz 2 rx
erf(x) =-/ Tt Jo exp( -z 2 )dz
[H5/2_h5!2]
=4/5H-ti2Gsm [H2-h2] (10)
Small values of e correspond to a situation
It should be noted that the above response is where the expected number of peaks is much
not influenced by 'b 0'. larger than the expected number of zero cross-
ings in which case Eq. (14) maybe approxi-
Spectral Density Functions for Response mated by a Gaussian probability density
Considering linearity of the system, the function.

(2;r~1/20'z exp[-; ( :x )]
spectral density functions for any of the above
response variables can readily be obtained: fxpCA)
S:c(z, (!)) = IH:c(z, i(l)) I2Sii/(l)) (11) -co<J.<co (15)
where Sug((l)) is the spectral density function On the other hand, the upper limit e = 1
of the input ground acceleration (see Eq. (1)), corresponds to a narrow band random process.
and x represents any of the response variables, For this case the number of peaks is equal to
e. g., U, il, UB, y and 7:'. that of zero crossings and the probability
density function for the random variable Xp
Statistics of the Maxima reduces to a Rayleigh distribution:
Following Lin (18), the following relations
for a stationary random process can be writ- fxpO.) = 0'~ 2 exp[-; (0'~ ) ] O<J.<co
ten: (16)
The expected number of zero crossings per
unit time, The Extreme Value Problem
Many solutions have been proposed for the
extreme value· problem (19); each solution
being based on different assumptions and
simplifications. The solutions are usually
where E[.] denotes the expected value of presented in the form:

NII-Electronic Library Service

 PROBABILISTIC ANALYSIS OF EARTH DAMS 51

(17) Newmark (1) and later extended by Sarma

-where Ex is the extreme value of x which ( 4), Makdisi and Seed (9) and Lin et al. (10).
·has the probability p of not being exceeded The elements of the analysis and associated
-within the duration S. The function rs.p is assumptions are discussed below:
·called the peak factor and is the basic
parameter which has to be determined. Slip Surface, Sliding Block and Critical Ac-
Vanmarcke (20) presented an approximate celeration
:.solution for the peak factor given by : The earth dam is considered flexible with
its response depending on its stiffness and
r 2 s. p =2ln[2 n-exp(o)e1. 2 .J 7"t In (2 n)]
damping characteristics and the intensity of
QS
-where, n= 7"C (-In p )- 1 ground shaking. There exists a level of
2
shaking which induces large inertial forces
1 2
A2 ) 1 to form a well defined slip surface [Fig. 2]
Q= ( l:;
delineating a block of soil mass from the main
_ (1 A
oe- - AoA2
12)112 (18) body of the dam. Thereafter, whenever the
inertial force associ a ted with this block over-
Ai = JmiS x (m)dm i= 0, 1, 2
comes the horizontal components of the total
resistance along the slip surface, the block is
·In above Q is the central frequency and Oe assumed to move as a rigid body. This re-
:represents a bandwidth parameter. sistance 'R' is defined in terms of a critical
acceleration of the block for which the soil
The Equivalent Linear Model strength along the slip surface is completely
The shear modulus and the damping ratio mobilized developing a factor of safety of
·of soils depend on the level of strain induced. unity. In this study we assume that by a
'This nonlinearity of soil behavior is incorpo- conventional analysis, choosing various trial
·rated in the following manner. First, initial ship surfaces, the value of the critical accel-
values of Gsm and {3 are assumed. The dynam- eration has already been evaluated. It should
·ic response is then evaluated. From the be noted that there are uncertainties associ-
·power spectra of the strain, the maximum ated with the soil properties. The effect of
·value of the shearing strains r max (z) are eval- these uncertainties on dam response can be
·uated by using eqn. (17), with P=60%, (11). taken into account in this step but is not
The average (spatial) effective strain at any considered in this study.
·point is calculated as
1
rerr= H-h)o
rH-h 0.65rmaxCz)dz
- (19)
The Block Acceleration
Earthquake-induced accelerations m the
'The values of Gav and {3 compatible with the embankment are determined by dynamic
:.strain r eff are then read from the laboratory response analysis. The average block accel-
obtained curves. These new values of Gsm eration uB(t), is then evaluated. At the stage
.and {3 are then used in the next iteration of of evaluating ilB(t), the earth dam is treated
. all the above calculation steps. Usually 3---4 as a continuous system. After the formation
iterations are required for convergence; there- of the yield surface (i.e., development of a
. after the changes in G and {3 are not discontinuous system) the block of soil mass
.significant. is assumed to behave as a rigid body subjected
to the same uB(t) [Fig. 2(b) ]. It should be
noted that a trapezoidal shape of the block
ELEMENTS OF ANALYSIS FOR PERMA- is assumed for convenience of evaluating the
-_NENT DISPLACEMENTS block acceleration. The associated bilinear
The basic formulation here follows the slip surface should approximate the corre-
:simplified method originally proposed by sponding slip surface used in evaluat ingthe

NII-Electronic Library Service

 52 PAL ET AL.

Sliding Block placements caused by individual slips.

The block of soil mass sliding on the failure~
plane and forces acting on it during slip are
shown in Fig. 2(b). Based on the free body
diagrams, we have
i.i 9(t) m(x+ilB) +R=O
(a)
or
x= -ilB-Rfm.
Reference Axis
From the above:
+Ve it /l .
~---mue
x= (KB-Kc)g, when KB>Kc' (Kc>O)
(20}

---Lie otherwise
x=O,
(b)
Ke=·iie
where
g
Ke m =mass of soil block
Kc ilB=KJig earthquake induced block accel-
eration
X=relative acceleration of the block
R= (Kc mg), the resistance along the·
failure plane
KB=earthquake induced average block
(c)
Kc = = acceleration
Fig. 2. Earthquake induced permanent Kc=critical acceleration of the block
displacement in darns For the sake of simplicity, KB is assumed to·
have an idealized form:
critical acceleration. (21)
where W 0 is predominant frequency of ground:
Permanent Displacement of the Block
motion obtained from s;;g(w), ¢ is random
Whenever the earthquake induced block
phase angle uniformly distributed and Kp is.
acceleration KB(t) [ =ilB(t) fg] exceeds the
Rayleigh distributed amplitude of KB.
critical acceleration Kc (which is assumed to
From very basic considerations the following
have already been evaluated) [Fig. 2(c) ], the
relationships are readily obtained [Fig. 3]:
block is assumed to slip along the failure
plane. The slip occurs only in the positive KBCto) =0
x-direction since the critical acceleration corre- where t 0 is time of occurence of peak, and,.
sponding to negative block acceleration is from KB=O,
considered to have infinitely large value [Fig.
2(c) ]. The displacement of the block starts From
when KB(t) exceeds Kc and it stops when the
relative velocity between the block and the
horizontal slip plane vanishes. we have
In this study it is further assumed that the sin(w 0 J..-¢) = ±KdKP
effective pulses (shown with hatched lines in Similarly, from
Fig. 2(c) are so far apart that displacement
KB(J..+o) =Kc
during any two consecutive slips do not over-
lap. This assumption is valid for large we have
values of Kc. The total permanent displace- sin[wo(J..+o) -¢] = ±KdKP
ment, therefore, is equal to the sum of dis- In the above,

NII-Electronic Library Service

 PROBABILISTIC ANALYSIS OF EARTH DAMS 53

fKe K2= [2(Kp-Kc)/Kp] 112 (22 b)

from which
a+ J.. =K2fwo+ cpfwo
/Idealized Pulse Shape
By integrating the expression for relative
(a) acceleration x in eqn. (20), the relative veloc-.
t
ity of the sliding block is obtained (see Fig.
3):
(a) For t<J..+a:

x(t) = Kpg [sin(w 0 t+¢) -sin(w 0 J..-¢)]

Wo
-Kc(t-J..)
(b) For t=J..+a:
Kc (b)
x(J..+a) = (2gfwo) (Kl-KcK2) (23)
(c) For t>J..+a:
x(t) =x(J..+a) -Kc[t- CJ..+a)]
·j
After the sliding of the block is initiated, it
continues to move till the time, te, at which
{c) its relative velocity reduces to zero. This
te gives:
I
I
I
I
te=2 gKd (woKe) -2 gK2/w 0 +cpfw0 +K2jw 0
I
I
I (24)
I
(d)
_Jt -r------ ,__ - --· -----=-=-=-=-'--;-
I I
After integrating the expression for the rela-
1' I
I
l
I
tive velocity of the block the amount of
I
I
I
relative displacement of the block during one
slip, d, is obtained as follows:
d =X(te) =2gKt 2/W 0 2Kc-2K K gjw 2
1 2 0

Fig. 3. Displacement of block during an (25)

acceleration pulse

(22 b) PROBABILISTIC EVALUATION OF PER-

The value of a
can be obtained using MANENT DISPLACEMENTS
jnverse trigonometric functions but this will For a narrow band random process KB(t),
make subsequent computation cumbercome. the statistical properties of the permanent
Instead, we derive the following approximate displacement d may be found in the following
but simple relation for a.
The method utilizes manner:
the Taylor series expansion of KB about t 0 •
'That is, Expected Value of the Displacement
In Eg. (25), K 1 and K 2 in Eqs. (22 a) and
Kc=KBCto+a/2) =KBCto)
(22 b) are random due to the presence of Kp
+KB(to)l_+ KB(to)
2 2 2
(.§_)2+··· which is Raleigh distributed with probability
density function denoted by fKp and given by
·To the order of a , the following approximate
2

·expression for a is obtained: fKp= lfP exp( -Kp/2<1 2Ks) (26)

<J Ks
a=2 K2/Wo The expected value of the block displacement
where during one slip can be written as:

NII-Electronic Library Service

 54 PAL ET AL.

= Var[N(S)]·E 2 [d]+E[N(S)] Var[d]

(31)
2gioo K1K2fKpdKP (27)
--2 For large. value of the ratio, Kc, N(S)>
lVo Kc (JKB ;
Substituting (26) in (27) the expected value approaches a Poisson's process, giving
of the permanent displacement in one slip, Var[N(S)] =E[N(S)] (32)
d, can be obtained which involves the nu- Therefore,·
merical evaluation of the associated integrals.
Var[D(S)]=E[N(S)] {(E[d]) 2 + Var[d]}
The expected total displacement E[D(S)],
in duration S, will be the sum of expected (33)
independent displacement during each slip. where,
That is, Var[d]=E[d 2 ]-(E(d)) 2 (34}
E[D(S)] =E[N(S)]E[d] (28) in which,
in which E[N(S)] represents the expected 4 2
E[d 2]=____!LE[K 4]+ g E[K 2 K 2]
number of slips in time, S. Following Lin (J)o4Kc2 1 Wo4 1 2

(18) the expected number of crossings per

-~E[K
2 1
2 K]
2 (35)
unit time at level 'Kc' of a stationary random h\
owo Kc
process KB can be obtained. From E[d], Using the Rayleigh distribution for Kp ail.d1
the expected total number of slips of the performing the various integrations numeri-
block during an excitation of an equivalent cally E[d 2 ] can be evaluated.
stationary duration 'S' can be shown to be
given by: Probability of the Permanent Displacement·
112
2 Exceeding a Limiting Value: P(D>Dz)
S [ ioo w KK 8 (w)dw ]
If the permanent displacement 'D' is as-
E[N(S)]= 2;r loo sumed to have a log normal distribution th~m
SK 8 (w)dw
0 the probability density function of D can be:
exp( -Kc /2dK 8 2 )
2
(29) written as
Thus from Eqs. (26), (27), (28) and (29) the
expected value of total permanent displace-
p(D) DdnlV 2 7r exp[ -112( D*d~P*) J
ment of a potential sliding mass of an earth (36)
dam can be evaluated; where
D*=lnD
Coefficient of Variation for "D(S)"
The coefficient of variation of D(S) can be D*= In D-1_ (dn*) 2
2
written as:
dn*= [ln (V[D 2 ] + 1) ] 112
V[D(S)] = Var[D(S) ] 112/E[D(S)]
where, the variance of D(S) is V[D]= d_!J
D
2 2
Var[D(S)] =E[D (S)] -E [D(S)]
The probability of the event D>Dt can be
where, written as:
E[D 2 (S)] =E[N 2 (S)]E 2 [d] oo 1

Therefore,
+E[N(S)]Var[d] P(D>Dz)=1- _
J
D1 'V12 ;z'(J D
*
2
1 ( D*-D* ) (37)
Var[D(S)] exp [ - 2 dn* dD
=E[N 2 (S)]E 2 (d) After performing the integration in the above
+E[N(S)] Var[d] -E 2 [N(S) ]E 2 [d] equation, the following is obtained:
= {E[N 2 (S)]-E 2 [N(S)]}E 2 [d]
1 { 1-erf ( D*-D*
P(D>Dt) =2 V2 dn 2 )}
+E[N(S)]Var[d]

NII-Electronic Library Service

 PROBABILISTIC ANALYSIS OF EARTH DAMS 55

0.020
NUMERICAL RESULTS S ..
Ug 0.015
The formulations discussed in the previous
sections to evaluate the seismically induced 0.010
permanent displacement in an earth dam have (s::s)
0.005
been incorporated in a computer program,
0 50
PROSE (18). Using this program an example
study was conducted to illustrate the· evalua· 30
tion process. A series of analyses was then Inhomogeneous Model
made to study the effect of various parameters 20
on P(D>Dz), the expected value and the IH··Uav 12
coefficient of variation of seismically..:..induced
10
permanent displacements in an earth dam.

Permanent Displacement in an Earth Dam: 40 50

An Example study
This numerical example analyzes a 49 m
0.2
high homogeneous embankment for h=O, Suav Inhomogeneous Model
density fJ =2. 09 gmfce and the maximum shear +--,20 ZB/H
.40
modulus, G 8 m=175, OOON/m 2 • The seismically
induced permanent displacement in this dam
(s::s) 0.1 .60
.80
1.0
is evaluated. The seismic environment is 0 10 20 30 40 50
modelled as a random process. The N 21 E co (Rad/Sec.)
component of the Taft record of Kern County Fig. 4. Response of an earth dam : example
earthquake (1952) with a maximum accelera· problem
tion of 0. 2 g is used as a sample process.
The model parameters describing Kanai-
Tajimi spectrum are S 0 =. 007 m 2 /sec 2 , {J)g =
E[dJ (meter)
18. 45 radfsec. Cg = 0. 36, and S = 13. 7 sees.
The results of the analysis are presented in .005 .01 .015 .02 .025 .03

Figs. 4 and 5. The spectral density of the

input ground motion S;;u, the transfer func-
0.2
tion and the spectral density of the average
acceleration of five potential sliding blocks
0.4
are shown in Fig. 4. The values of O'uB'
E[N(S)] and E[d] for these blocks are § 0.6
shown in Fig. 5. These parameters decrease
with increasing ZBJH• ratio of the sliding block, 0.8
for the same level of critical acceleration.
For the purpose of illustration, it is assumed
that the most critical block of the dam is 2 3 4 5 6
the wedge extending to a depth of zB!H = 0. 8, E[N+)
with Kc=0.1g. For this, the results are
summarized below:
0 0.05 0.10 0.15 0.20 0.25
0' .. =0.068g
us
O'ue ('g' units)
E[N(S)] =7.526
Fig. 5. Expected values of displacement
E[d] =.226cm parameters and RMS block accelera-
E[D] =1. 706cm tion in example problem

NII-Electronic Library Service

 56 PAL ET AL.

V[D] =1. 672 block to the height of the dam;

P(D>lOcm) =. 017 defines the size of the block.
The results of the parametric studies are
Permanent Displacement in Earth Dams: A presented in Figs. 6 to 11, from which the
Parametric Study following observation can be made.
The same dam mentioned above is studied. From all the results it is evident that the
The mean of mixed site recorded motions
(15) with a peak acceleration 0. 3g and equiv· H/Vs =0.25 sec.
alent stationary duration of S = 10 seconds ii9max=0.3g
is used for the random ground motion model
with the following parameters: S 0 =. 020m2/
sec2 ; w11 =20.3radfsec. and <:' 11 =0.32
The following are the primary parameters
affecting the seismically induced permanent
displacement in earth dams:
H/Vs :The ratio of dam height to shear
wave velocity which defines the
flexibility characteristics of the
dam.
ilumax : The maximum ground accelera·
tion.
Kc/d"U 9 : The ratio of critical acceleration 1Q-2
to the standard deviation of
input ground acceleration, (which
characterizes the resistance of the Fig. 7. Influence of block depth on the
dam to permanent deformation.) coefficient of variation of permanent
zBfH : The ratio of the depth of sliding displacement

101

Ugmex" 0.3g
ZBtH" 0.5

10"1 100 101

Keto··uo

Fig. 6. Influence of block depth on expected Fig. 8. Influence of dam flexibility on

permanent displacement expected permanent displacement

NII-Electronic Library Service

 PROBABILISTIC ANALYSIS OF EARTH DAMS 57

101 jj9max= 0•39 101

ZB/H = 0.5 HIVs =0.25 sec.
ZB/H: 0.5

1Q-2 10•1
Kefa··Ug 10·2 10·1

Fig. 9. Influence of darn flexibility of the Kcta"Ug

coefficient of variation of permanent
displacernen t Fig. 11. Influence of the maximum ground
acceleration on the coefficient of varia-
tion of permanent displacement
HIVs = 0.25 sec.
ZB/H=0.5 of variation of permanent dieplacement tends
tOO
to increase for the deeper blocks (Fig.7).
Figs. 8 and 9 show very clearly that the
~
() more flexible dams will give smaller permanent
CD
~ displacement for the same Kc/cJ'Uu' while the
-€ coefficient of variation has the inverse tend-
..s
..e
lo(
ency. However it should be noted that the
til
::;,
more flexible dams will also have the smaller
~
Cll 10·1 value of Kc/cJ:Uu·
Q The values of expected displacement, and
w
the coefficient of variation normalized with
respect to the maximum ground acceleration
decreases with increasing level of maximum
ground acceleration. (Figs. 10 and 11).

1Q-1
CONCLUDING REMARKS
Kc/cr"Ug
A simple procedure and a convenient ana-
Fig. 10. Influence of the maximum ground
lytical tool is developed for the probabilistic
acceleration on expected permanent
displacement evaluation of seismic response of earth dams
with emphasis on permanent displacements
expected value of the permanent displacement in dams built with compacted cohesive or
decreases with increasing Kc/cJ'Uu, while the dense sandy soils.
coefficient of variation of permanent defor- The uncertainty inherent in specifying the
mation tends to increase. ground motion is handled by modelling them
The deeper sliding blocks will have smaller as a portion of a stationary random process.
permanent displacements with other variables The dam is modelled as series of shear slices.
remaining the same (Fig. 6). The coefficient The nonlinearity in shear modulus and damp-

NII-Electronic Library Service

 58 PAL ET AL.

ing are incorporated by using the Equivalent ]. engr. mech. div. ASCE 92[EM 2], Proceed-
Linear model for stress-strain behavior of ings Paper 4793, April, pp. 197-212 ..
soils. The seismic response is assumed to be 7) Lin, J. S., Whitman, R. V. and Vanmarcke, E. H.
only due to the vertically propagating shear (1982) : "Probabilistic Evaluation of the Seis-·
mically-Induced Permanent Displacements in.
waves and is formulated in frequency domain
Earth Dams," Research Report No. R 82~21,
providing the power spectra of various re-
Department of Civil Engineering, Massachuset·
sponse parameters. The permanent displace- ts Institute of Technology, Cambridge,
ment in the dams is assumed to result from Massachusetts.
accumulation of the slips of a block of soil 8) Franklin, A. G. and Chang, F. K. (1977) :
mass detached. from the main body of the "Earthquake resistance of earth and rockfilR
darh. The occurence of the slip events are dams : Permanent displacement of earth ·em-
assumed to be goverened by a Poisson process. bankments by Newmark sliding block analysis,,.
It should be noted that in this study many Misc. Paper S-71-17, Report 5, U.S. army·
simplifying assumptions have been made. It Engineer Waterways Experiment Station,
is hoped that some of these are compatible Vicksburg, Mississippi.
9) Romo-Organista, M.P. (1977) : "Soil-structure
with the limitations of the sliding block model
interaction in a random seismic environment,,.
for failure, which has been used in many
Ph. D. Dissertation, University of California,.
studies. The limitation associated with the Berkely.
assumption of one dimensional response may 10) Okamoto, S. et al. (1969) : "On the dynami-
be removed by using a finite element model. cal behavior of an earthdam during earth-
The uncertainties associated with the soil quakes," Proc. 4 th World Con£. on Earthqk.
properties are quite important and should be Engr., Santiago, Chile.
included in future analyses. This can best be 11) Vanmarcke, E. H. and Lai, S. P. (1977) :
done by modeling the soil properties as ran- "Strong motion duration of earthquakes," Re-
dom fields, in conjunction with a stochastic search Report, No. R 77-16, Department of Civ-
finite element analysis of the seismic response. il Engineering, Massachusetts Institute of
Technology, Cambridge, Massachusetts.
The assumptions that the acceleration pulses
12) Binder, R. (1978) : "Strong motion duration
are sinusoidal and that slip events are Poisson
and spectral density functions for a set of 39'
are rather servere and they should be removed earthquake ground motions," Seismic Safety
in future studies. of Buildings Internal Study Report, No.16, De-
partment of Civil Engineering, Massachusetts:
Institute of Technology, Cambridge, Massachu--
REFERENCES setts.
1) Newmark, N. M. (1965) : "Effect of earthquakes 13) Lai, S. P. (1980) : "Overall safety assessment
on dams and embankments," Geotechnique 15 of multi-story building subjected to earthquake·
[2], pp.139-160. loads," Ph. D. Dissertation, Department of Civil
2) Seed, H. B. and Martin, G. R. (1966) : "The Engineering, Massechusetts Institute of Tech ..
seismic coefficient in earth dam design, " J. nology, Cambridge, Massachusetts.
Soil Mech. Found. Div. ASCE 92[SM 3]. 14) Oner, M. (1984) : "Estimation of the Funda.
3) Ambrasseys, N. N. and Sarma, S. K. (1967) : mental Period of Large Earthfill and Rockfill
"The response of earth dams to strong earth- Dams," Soils and Foundations, Journal of Japa-
quakes," Geotechnique 17[3], pp. 181-213. nese Society of Soil Mechanics and Foundation
4) Sarma, S. K. (1975) : "Seismic stability of Engineering, 24[4], pp.1-10.
earth dams and embankments," Geotechnique 15) Lin, Y. K. (1967) : Probabilistic Theory of
25[4], pp.743-761. Structural Dynamics, McGraw Hill, New
5) Makdisi, F. I. and Seed, H. B. (1978) : "Simplified York.
procedure for estimating dam and embankment 16) Crandall, S. H. (1970) : "First-crossing Proba-
earthquake-induced deformations," ]. Geotech. hili ties of the linear oscillator, " J. Sound
Engr. Div. ASCE 104[GT 7], pp. 849-867. Vib. 12[3], pp.285-299.
6) Clough, R. W. and Chopra, A. K. (1966) : 17) Vanmarcke, E. H. (1976) : "Structural res-
"Earthquake stress analysis in earth dams," ponse to earthquakes," Chapter 8 of Seismic

NII-Electronic Library Service

 PROBABILISTIC ANALYSIS OF EARTH DAMS

Risk and Engineering Decisions, Lomitz, C. and sponse of soil masses in random seismic envi-
Rosenblueth, E. eds., Developments in Geotech- ronment," A Report to National Science Foun-
nical Engineering, 15, Elsevier Scientific Pub- dation, Grant No. CEE-8106301, Dept. of Civil
lishing Company. Engineering, North Carolina State University.
18) Rahman, M.S. and Pal, S. K. (1983) : "Re- Raleigh.

NII-Electronic Library Service