Kikusawa, M. (1990). Geotechnique 40, No.

2, 201-210

Seismic stability analysis of rockfill dams

M. KIKUSAWA*

Seismic conditions considerably affect the stability Avec la mbthode pr&ente de construction les con-
of a rockfill dam in current design method. Thus in ditions sismiques influencent beaucoup la stabilitk
the present study pseudo-static analyses are con- d’un barrage en remblais rocheux. Des analyses
ducted for a typical dam with a central impervious pseudostatiques pour un barrage typique i noyau
wall by means of a limit equilibrium method and central impermbable furent effectu&s par la m&
an upper-bound limit analysis (Rigid-plastic finite thode d%quilibre limite et I’analyse des limites
element method is herein employed). Firstly are sup&rieures en employant la methode par &?ments
discussed the safety factor in limit equilibrium, sta- finis rigidesplastiques. L’article discute le coefti-
bility factor in limit analysis, and seismic coeffi- cient de &curit& le facteur de stabiliti dans
cient. It is confirmed that stability can be evaluated I’analyse des limites et le coefficient critique sis-
well by taking the stability factor NS as a param- mique. 11 est confirmC qu’on peut tr& bien bvaluer
eter and by the independent use of the safety factor la stabilite en prenant le facteur de stabiliti! N,
for the frictional strength. NS is defined as a func- comme parametre tout en employant ind(?pendam-
tion of the finite value of cohesion, which is nor- ment en m@me temps le coeflicient de s&urit(? pour
mally neglected for the rockfills. Secondly it is la r&stance P la friction. On d6finit NS comme
illustrated that the effects of the apparent cohesion une fonction de la valeur finie de cohbion qu’on
on the stability by employing a nonlinear strength nklige g&5ralement dans le cas des remblais
envelope. A useful chart is presented to address the rocheux. L’effet de la cohbion apparente sur la sta-
critical slope and the critical seismic coeficient for biliti! est par consi?quent illustrit zi I’aide d’une
the given NS. enveloppe de rupture non-IinCaire. Des abaques
sent p&en& pour la r&solution de la pente cri-
tique et le coefficient sismique critique pour une
valeur don& de NS . Ces abaques indiquent la lim-
itation de la methode d’&quilibre limite et
I’importance de la cohbion apparente pour
KEYWORDS: dams; earthquakes; finite element I’analyse de la stabiliti! des barrages i en remblais
method; safety factor; slope stability; rockfill. rocheux.

NOTATION slope angle

A, b material parameters unit weight of fill material
C, actual cohesion cotangent of slope angle b
ci cohesion mobilized critical value of rj
FS safety factor actual friction angle
FC safety factor for cohesion friction angle mobilized
positive coefficient that depends on
Fe safety factor for friction geometrical and mechanical proper-
gravity acceleration
IT slope height ties of the sliding mass
K seismic coeffkient RP-FEM rigid-plastic finite element method
KC critical value of seismic coefficient
K.3 design value of seismic coefficient
K equivalent seismic coefficient
NS stability factor INTRODUCTION
ci angle of the sliding direction to the Modern large dams constructed by the com-
horizontal pacted method have scarcely suffered from any
incurable damage. This is evidence for the effec-
Discussion on this Paper closes on 1 October 1990; for tiveness of the current design method established
further details see p. ii. empirically. However, complete confidence
* Department of Agricultural Engineering, Faculty of cannot be placed in the effectiveness of the
Agriculture, Kyoto University, Japan. method because of limited experiences of seismic
201
Downloaded by [ UC San Diego Libraries] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 202 KIKUSAWA

events. Enormous accelerations have been regis- slide that depends on the properties of soil, mag-
tered at dam sites, such as 1.2 g for the Morgen nitude of inertia force and slope angle. A similar
Hill (Shakal, 1984) and 0+0.7 g for the Mexican study was made by Prater (1985), in which lin-
earthquake (Marsal, 1987) which certainly have early varying cohesion with depth was newly
an intensity beyond the current design earth- incorporated. Most of these stability charts are
quake. Surprisingly, these severe motions brought calculated by analytical methods for homoge-
about only slight damage to both dams. From neous natural slopes.
these fortunate case histories it can be said that In the present study a numerical approach of
modern dams have slopes more gentle than is limit analysis in a finite element system is
necessary, but the current specification of the employed to assess the seismic stability of inho-
seismic coefficient does not seem to be conserva- mogeneous rockfill dams; an attempt is also
tive in relation to the tremendous intensity of made to examine the limitation of the use of the
earthquakes observed. De Mello (1977) pointed conventional limit equilibrium method. It is
out that steeper slopes were automatically con- intended to show the critical values of the slope
structed in old times when dumped rockfill was steepness and seismic coefficient that could be
used, and a trend for flatter slopes began with the used to calculate the permanent displacement
advent of compacted lifts accompanied by mentioned. Numerous pseudo-static stability cal-
laboratory-based theory. culations were done by means of the limit equi-
Nowadays intensive efforts are made to study librium analysis and the finite element-based
the seismic stability of embankment dams espe- upper-bound limit analysis. In order to compare
cially as concerns permanent deformation due to the results obtained from different methods it is
earthquakes. Kuwano Ishihara (1988) estimated necessary to review the relationships between the-
the permanent deformation of an earth dam safety factor F,, the stability factor N, and the
through strain potential in terms of a hybrid pro- seismic coefficient K. Since the cohesion cannot
cedure of static and dynamic FEM and triaxial be ignored so long as N, is employed, the effects
torsion tests. Daddazio, Ettouney & Sandler of the apparent cohesion in terms of the Mohr-
(1987) proposed a method that simulates progres- Coulomb type and the non-linear failure envelope
sive failure by incorporating geometric and are examined. A stability chart is presented for a
material non-linearities. However, it is usually typical rockfill dam with an impervious wall.
preferred to calculate the permanent displacement According to the chart, it is concluded that the
by way of a critical acceleration by the method limit equilibrium approach leads to a gentle slope
proposed by Newmark (1965). The key problem for a dam with N, greater than 100 due to the
in using Newmark’s method is of how to deter- underestimated strength at the low stress level
mine the critical acceleration K, . Makdisi & Seed and that the critical slope could be from 1 : 1.5 to
(1978) determined K, , which they called yield at most 1 : 1.7 for K = 0.15 and N, = 100.
acceleration, from the limit equilibrium-based sta-
bility analysis by employing the cyclic yield
strength estimated experimentally as a function of SAFETY FACTOR AND STABILITY FACTOR
embankment slope, material properties and loca- The limit equilibrium approach usually
tion of the potential sliding mass. Sarma (1975) employs the safety factor F, defined by the
derived f<! analytically from the limit equilibrium material strength which is specified empirically
of the rigid block mode on a plane surface. He for various types of structure. For instance, a
showed the effects of K, on slope stability for the rockfill dam can generally be designed by smaller
various parameters of embankment slope, F, than a clayey fill dam because of less dispersed
material strength and Skempton’s pore pressure friction angle according to the probabilistic works
coefficients. It is recognized from both these done by De Mello (1977). Besides, F, takes into
works that the earthquake-induced displacement account the facts that the peak strength cannot be
of the embankment is significantly affected by the simultaneously induced at all the points on a slip
value of K, used. Sarma & Barbosa (1985) assess- surface and that strain softening is expected
ed analytically the effects of K, on rockfill dams beyond the peak especially for the brittle rock
with central clay cores by means of a two-wedge materials. In this section is reviewed the relation
sliding mechanism, in which a great deal of infor- between the F, and the stability factor N,, the
mation was obtained from the dimensionless inverse of which is called stability number.
design charts about K, related to the embank- Two different factors of safety, F, and F, are
ment slope, material properties and safety factor. defined corresponding to the cohesion and to the
Instead of the limit equilibrium, Chang, Chen & friction of the Mohr-Coulomb criterion, respec-
Yao (1984) applied the upper-bound limit tively. That is
analysis to assess the seismic displacement. They
proposed seismic stability charts, showing the F, = (c&r), F, = tan $Jtan &Q (1)

Downloaded by [ UC San Diego Libraries] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 SEISMIC STABILITY OF ROCKFILL DAMS 203

in which c,, 4, are the strength parameters for N, is a parameter defined uniquely for a slope
the materials and cr, 4r are the working or the height H and material properties of y and c,. N,
mobilized strength parameters. Let the length of a is determined by using not F, but F, of equation
slip line be L, the weight of sliding mass be W, (5). In the limit analysis N, is determined by using
the average angle between sliding direction to the a load factor at collapse instead of F, or F, Thus
horizontal be a and the horizontal seismic coeffr- the stability will be evaluated in terms of N, that
cient be K, and the following equilibrium equa- can be given both by the limit equilibrium
tion is obtained. analysis and by the upper-bound limit analysis.
The results obtained from either analysis are
c,L+ W,tanr#+=T (2) almost the same as indicated by e.g. Chen (1975)
Leshchinsky, Baker & Silver (1985). In this study
T = W(sin a + K cos a),
N, is used as a parameter that represents the
W, = W(cos a - K sin a) characteristics of a dam. The use of F, and F,
instead of F, would be profitable in ordinary
Substituting equation (1) into equation (2) cases such as when the cohesion and the friction
angle are distributed stochastically in different
(cJF,)L + W,(tan 4JF,) = T (3) ways on the design space.
For F, = F, = F, The characteristics of critical seismic coefficient
K, , which is derived from the equations (4) or (5)
F =c,L+KtanA provided F, = 1 or F, = 1, are now examined.
(4)
T
5

K,=K,+cN;’ (11)
For F, = F, and F, = 1, then tan 4. - tan a
K, =
F, = c, L/(T - W, tan 4.) (5) l+tan4,tana’
(12)
F, is the conventional safety factor. F, varies with PC
5=
the slope size and the cohesion in a different way ~(cos a + sin a tan 4)
from F,, which may be explained as follows.
Suppose two geometrically similar slopes consist- Where < is a positive coefficient that depends on
ing of the similar material. One has the height of the geometrical and the mechanical properties of
H with F, and F,, and the other is n times higher a sliding mass. F,, a and 4, were mentioned
with F,, and F,, . F,, is calculated by substituting before and K is the ratio of the depth of the
nL, n’ W,, and n2T instead of L, W, and T of sliding mass to the slope height. From equation
equation (4). In the same way F,, is given from (12) it is found that if the sliding surface is steep
equation (5). or the material friction is small, such that a 2 4,)
K, will be less or equal to 0, which means the
slope fails by its own weight. However, in the case
F,,=F,+(c,L/T) i-1 of a < (6.) K, is positive as illustrated in Fig. 1, in
( > which K, approaches a constant positive value of
F,, = (FJn) (7) K, as N, infinitely increases. A rockfill dam is

Equation (6) implies that all the similar slopes

with different heights have the same F, when
c, = 0. Now, the stability factor N, is defined for
a slope of height H, the unit weight y and cohe-
sion cr.

N, = (YWJ (8)
Substituting equation (1) into equation (8)

N, = F,(~fUca) (9)
Again consider N,, for a slope with the height of
nH. Noting that F, = F, and F,, = F,,, N,, is
obtained, defined in the same way as in equation
(9), as follows, in which equation (7) is substituted

N = F ynH_ WV,vH = N
Fig. 1. Relation between critical seismic coefficient KC
sn 1” 5 (10) and stability factor N,
C, C,

Downloaded by [ UC San Diego Libraries] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 204 KIKUSAWA

mainly represented by its geometrical factors such

as the height H and the slope /I, and by the Xi A Wp(cos a, - K sin ai)“!: -b
F, = (14)
material factors such as y. c, and 4,. By consider- - K cos ai)
Xi K(sin CL~
ing a typical value for 4.) the seismic stability can
be evaluated by H, y, c, and b and the seismic where w and Ii are respectively weight and slip
coefficient K. Apparent cohesion c, is often length to a slice i of the sliding mass. mi is the
ignored in the design, but it will be taken into angle between the sliding direction and the hori-
account explicitly for the.reason that will be dis- zontal and K is the seismic coefficient. Equation
cussed in the next section. H, y and c,-which are (14) assumes no excess pore pressure developed.
considered as the key factors in addition to K Figure 2 shows the Mohr-Coulomb failure
and b-are linearly related to the stability of a envelopes and non-linear failure envelopes of
dam or a slope as indicated by the stability factor equation (13) for the data obtained by Matsu-
N, of equation (9) as well as F,, which is the moto & Watanabe (1987). Here limit equilibrium
safety factor applied to cohesion with full mobi- calculations are made assuming circular arc slip
lization of friction. As can be seen in equations (4) lines for two central core-type dams consisting of
and (5) (noting that F, = F,) the conventional rockfills shown in Fig. 2. The strength of the
safety factor F, is different from F,. However, impervious material is assumed to be ecu = 0.5
interest here is in the critical seismic coefficient kgf/cm* and $cu = 20”. Table 1 lists the critical
K,, where the safety factor, either F, or F,, seismic coefficient K, and the safety factor F, for
equals unity. Therefore, it is emphasized that the
use of N, together with p and K is reasonable to
evaluate the stability, and in doing so, the results
of the limit equilibrium analysis based on the
safety factor F, can be compared with those of
the limit analysis based on the stability factor N, .

APPARENT COHESION AND NON-LINEAR --- RockB

FAILURE ENVELOPE OF ROCKFILLS
Matsumoto & Watanabe (1987) conducted tri-
axial tests at stress levels of 2-8 kgf/cm’ for 17
types of rockfill. They performed 49 test cases (a) T, = 0.42 + ~1”tan43”
which were sampled out of the 100 m-high dam
(b) T, = 1.24 + on” 887
sites and whose parent material was mostly
igneous. The average values of cc- and &n are (C) T‘ = 1.88 + (1” tan420

1.02 kgf/cm’ and 41.2”. ccn is from 0.15 to 1.88 (d) T, = 1.91 + q,’ “’
kgf/cm’. Such apparent cohesions are usually
ignored in the stability calculations, which may
0 2 4
bring a critical slip surface coincident to the slope
on : kgf/cm’
surface, owing to an underestimation of the
strength at lower stress levels. Fig. 2. Linear and non-linear envelopes of shear strength
This section illustrates the effects of neglecting Q to normal stress on for rocks
the apparent cohesion and of expecting small
cohesion, say 0.1 kgf/cm*, by employing a non-
linear failure criterion through a simple stability Table 1. Effects of strength envelope on sta-
calculation. De Mello (1977) pointed out from the bility of rockfill dam
results of large-scale triaxial tests that the failure
envelope of rockfills is non-linear and expressed it Case Rock A Rock B
by the following equation.
I

Q = Aab” (13)
where zy and on are the shear strength and the
normal stress, respectively. A and b are material 3 0.160 1.024 0.145 0.991
parameters. When b = 1, the envelope becomes 4 0.249 1.203 0.232 1.168
linear as in the Mohr-Coulomb type, in which (1) Mohr-Coulomb, (2) non-linear envelope,
c = 0 and tan 4 = A. F, is expressed by the fol- (3) c = 0 in case (l), (4) c = 0.1 kgf/cm’ in case
lowing equation in the case of the Fellenius (1).
method adopted (Charles & Soares 1984). F, is calculated for K = 0.15.

Downloaded by [ UC San Diego Libraries] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 SEISMIC STABILITY OF ROCKFILL DAMS 205

K = 0.15, and the corresponding critical slip lines employed. In the upper-bound analysis the rigid-
are shown in Fig. 3. Case (1) employs the Mohr- plastic finite element method (Tamura, Kobay-
Coulomb criterion, case (2) employs non-linear ashi & Sumi 1984) is used.
failure criterion, and for cases (3) and (4) it is
assumed respectively c = 0 and c = 0.1 kgf/cm2
in the use of the Mohr-Coulomb criterion. Rigid-plastic$nite element method
Results by way of the non-linear failure criterion The upper-bound limit analysis is a method to
are akin to those of the Mohr-Coulomb criterion analyse plastic collapse of a rigid-plastic body
according to cases (1) and (2). It is emphasized and to find the optimum collapse mechanism in a
that the slight cohesion in case (4) makes the limit load system based on the theory of plasticity
results rather close to those of case (1) for less (Chen, 1975). The analysis is convenient and it is
cohesive rock A, while neglecting the cohesion easy to obtain a limit load for a typical boundary
results in F, and K, on the extremely safe side value problem of homogeneous media. Since the
and insignificant critical slip line on the slope collapse mechanism must be assumed beforehand
itself as seen in case (3). Through these simple cal- through insight resulting from knowledge of
culations the significant effects of the cohesive mechanics, the analysis becomes difficult for inho-
strength at lower stress levels are recognized. In mogeneous and arbitrary boundary conditions.
applying the conventional stability analysis, The rigid-plastic finite element method (RP-
careful attention should be paid to cohesion. For FEM) overcomes this difficulty. The optimum
this purpose, the use of F, and N, of equation (9) collapse mechanism is searched automatically in
may be helpful. RP-FEM, and can be applied to any inhomoge-
neous and arbitrary boundary conditions within
the practically permissible precision.
SEISMIC STABILITY CHART OF A TYPICAL In this study basically a pseudo-static stability
ROCKFILL DAM analysis has been conducted using the RP-FEM.
In this section is presented a chart to determine The difference between the conventional analysis
the critical seismic coefficient K, for a specified and the present RP-FEM is in the treatment of
slope 4 or critical slope rlc for a specified design the body force. The RP-FEM analysis is per-
seismic coefficient K, . The calculation is per- formed not for a particular sliding mass but for
formed for a typical central core-type rockfill dam the entire dam. Therefore, the seismic inertia
by means of pseudo-static limit equilibrium and forces are applied to all finite elements and they
upper-bound limit analyses. In the limit equi- can be either in phase or out of phase in such a
librium analysis only horizontal seismic force that way that the entire embankment vibrates. The in
acts on the slip line is assumed and the Fellenius phase inertia force is just that of the conventional
method that provides F, on the safe side is seismic coefficient method. However, the out of

(b)
Fig. 3. Effects of strength envelope on critical slip line (Table 1 details
cases (l)-(4)): (a) rock A;(b) rock B

Downloaded by [ UC San Diego Libraries] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 206 KIKUSAWA

phase inertia force incorporates the response material safety factor of 1.2. For example, let the
acceleration distribution at the most risky internal friction angle of 4, be 40” as in the case
instance as a space-variable seismic coefficient, so of a typical rockfill, a value of & = arctan (tan
that the response characteristics of the structure 40”/1.2) = 35” will be used. The same safety factor
can be taken into account (Kikusawa, 1987). It is is used for the ecu and &-u of the impervious
not quite obvious how the response acceleration materials. Sliding failure is confirmed to occur
intensity is related to the in phase seismic coefh- within the rockfill zone even as a result of severe
cient. Therefore, in the following calculations, the earthquake for the strength of the impervious
in phase inertia force will be applied to obtain a material specified in Table 2 (Kikusawa, 1987). In
seismic stability chart with an attempt to assess this case the cohesion strength of the rockfill
the conventional method of seismic stability, that affects only the limit load, not the failure mech-
is the pseudo-static limit equilibrium method. anism. The stability factor N, in equation (9) rep-
Okamoto (1971; pp. 403407) mentions the resents the dam characteristics, namely the dam
failure features of rockfill dams and refers to his height H, unit weight y and cohesion c, of the
model experiments by a shake table as follows. In rockfill.
the case where the stiffness of the impervious From the results of the stability analyses per-
material is almost the same as that of the rockfill, formed for the dams of various slopes and seismic
the upper part of the impervious wall slides coefficients a chart is obtained to resolve the criti-
together with the rockfills under seismic action. cal slope 4, or critical seismic coefficient K, for a
However, in the case of impervious material of given N, as shown in Fig. 4. K, is determined for
higher stiffness, the failure proceeds only within a given N, and an arbitrary v, and is the same for
the rockfill zone although some vertical cracks nc. Fig. 4(a) obtained from RP-FEM shows that
are induced in the impervious zone. And the the rl - K relation hardly varies as N, increases,
rockfill moves down along the surface of the finally being independent of N, when it is greater
impervious wall when no transient zone exists. than 100. This means that K, converges to a con-
RP-FEM can well simulate this sort of failure stant value for a given II, which can be under-
mechanism, and the proper mechanism in the stood from the discussion regarding Fig. 1 and
analysis may well lead to a proper limit load. The equation (11). Fig. 4(b), calculated by the limit
RP-FEM can be applied to an inhomogeneous equilibrium approach of the Fellenius method,
problem as well as a homogeneous one and it can shows similar characteristics to those of Fig. 4(a).
employ the non-associated flow rule (Tamura, However, the q - K relation varies with N, until
Kobayashi & Sumi, 1987). In this study, Drucker- it becomes 1000. In other words, the limit equi-
Prager criterion and associated flow rule are librium calculation produces different results
employed. from the upper-bound limit analysis especially for
N, greater than 100.
Figure 5 is a replot of Fig. 4 for N, = 100 and
Critical slope and critical seismic coefficient 1000. For N, = 100 the maximum difference
In the following calculation model rockfill among the analytical methods is only 12% which
dams will be used that consist of a central imper- might be caused by the safe side calculation of the
vious wall with slopes of 1 : 0.2 and rockfill slopes Fellenius method and by the numerical error
of 1 : 1 up to 1 : 2.5. For convenience, it is (danger side) of the FEM. For N, = 1000, on the
assumed that the model dam has the same exter- other hand, the difference between the limit equi-
nal slope steepness upstream and downstream. librium and the upper-bound analyses is remark-
Hereafter tl denotes the steepness and equals able. This difference, which indicates the
cot fi, where fi is the slope angle. The material conservatism inherent in the conventional limit
properties are summarized in Table 2. As can be equilibrium method, may be caused in the follow-
seen from the Table, the impervious material is ing way. The dam of N, greater than 1000 is
assumed to take CU-strength, while that of the either extremely high or has rather small cohe-
rockfill takes CD-strength because no excess pore sion, and it is for these rockfill slopes that the
pressure is induced. The strength parameters in limit equilibrium approach leads to an insignifi-
Table 2 were specified by taking into account the cant critical slip line having small depth and coin-

Table 2. Properties of model rockfill dam

Downloaded by [ UC San Diego Libraries] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 SEISMIC STABILITY OF ROCKFILL DAMS 201

Max.input
acceleration
0 0.39

0.01 I I I
(4

0.5 7
N,=20

O-01
I I I
1.5 2.0 2.5

Fig. 4. Relation between slope q and seismic coeffkient K for stability

factor NS : (a) from RP-FEM; (b) from Fellenius

cident with the slope surface itself as in the case of shear strength of lower stress level should be con-
Fig. 3(3). However, the RP-FEM as well as the sidered, otherwise, the upper-bound analysis is
upper-bound limit analysis cannot give such a recommended. K, N N, relations for typical
failure mechanism because the mechanism does slopes are shown in Fig. 6, which corresponds to
not depend on N,. It is often in the case of the Fig. 1 obtained from analytical considerations.
limit equilibrium approach that zero cohesion of From Fig. 6(a) any height of dam with the slope
rockfill is assumed, in which the optimum pro- 1 : 2 can be designed by either the Fellenius
cedure searching the critical slip line is not useful. method or the RP-FEM under the seismic coeffr-
Therefore, several risky slip lines with a finite cient less than 0.2 for the material properties
depth are empirically presumed but it is not easy mentioned in Table 2. However, Fig. 6(b) indi-
to assume them under the circumstances of cates that the RP-FEM allows the slope 1 : 1.5
seismic action. for any values of N, under the seismic coefficient
From the foregoing discussion, it is inferred K = 0.12 N 0.15 although the Fellenius method
that the critical slope can be determined if N, < does not. Therefore it is very important to notice
100 and may be from 1.5 up to 1.7 for a designed these significant effects of the seismic conditions
coefficient K, = 0.15 as indicated in Fig. 5(a). In and the method of analysis adopted in designing
the case of N, > 100, appropriate cohesion or the embankment economically by a steep slope

Downloaded by [ UC San Diego Libraries] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 208 KIKUSAWA

50 100

9
zc
?.
‘!, 25 50
2
E
Y
2

0
I I I I
(a)
,i
0.01 ’ ’ ’ ’ ’ ’ ’ ’ ’ ’ ’ ’ 1
(4
2
z?. - RP-FEM
“0 25 50 --- Fellenius
2

E
2

01 I I I I
0.1 0.2 0.3 0.4 0.5
- RP-FEM
KC
----. Fellenius
(b)

Fig. 6. Relations between critical seismic coefficient Kc

I and stability factor N,, dam height ff(r = 2tfld,
E = 0.1 kgf/cm*): (a) slope 1 : 2; (b) slope 1 : l-5
0.01 ’ ’ ’: * ’ f 1 ’ ’ ’ ’ ’ ’
1.5 2.0 2.5
rj=cotJ by means of the probabilistic work done by De
(b) Mello (1977).

Fig. 5. Comparison of q-K relationship for different

analysis methods: (a) N. = 100; (b) N. = 1000
Seismic acceleration and critical seismic coefficient
It is well known that the in phase inertia force
such as 1 : 1.5. In practice the slope of the zone expressed by the seismic coefficient leads to a
type dam is designed by taking into account the greater stability level for the same acceleration
permeability characteristics of the random zone, intensity. Here is examined how the seismic coef-
the location of the impervious wall and so forth. ficient is related to the input and the response
Therefore, no design criterion seems to be estab- acceleration of a dam. The symbols plotted by
lished unlike the case of a homogeneous dam. circles in Fig. 4(a) denote the results obtained
According to the available data, the average from RP-FEM, in which the response acceler-
values of representative rockfill dams of central ation distribution calculated by a dynamic
core-type (34 dams in Japan) are 2.71 in upstream response FEM of the computer program code
(US) and 2.18 in downstream (DS) while those of QUAD4 was used as an inertia force. In dynamic
the similar types of 21 dams in other countries FEM an actual earthquake record observed at
are 2.18 (US) and 1.80 (DS). It seems that gentle the bottom of a rockfill dam was used. The
slopes in the Japanese rocklill dams may be maximum amplitudes of the input horizontal
enforced by the severe aseismic code. Compared acceleration are adjusted to 0.1, 0.2, and 0.3 g (g
with these values the present calculations indicate being gravity acceleration) for each symbol in
much higher seismic resistance, which might be Fig. 4(a) and the vertical acceleration is about
true so long as the current stability analysis is 70% of the horizontal intensity (following the
applied provided no excess pore pressure is observed records). For the case of the input accel-
induced during the shearing. De Mello ramarked eration of 0.3 g and the vertical base acceleration
that the field engineer has felt sure of the ability of 0.21 g, the maximum horizontal response
to use compacted rockfills at slopes of 1 : 1 or acceleration is about 0.7 g on the crest. Three
more, while computational models in vogue have solid circles are plotted for the given slopes and
been requiring slopes of about 1 : 1.7. An even the calculated N, values. The response acceler-
steeper slope of 1 : 1.4 is calculated for F, = 1.5 ation characteristics are not dependent greatly on

Downloaded by [ UC San Diego Libraries] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 SEISMIC STABILITY OF ROCKFILL DAMS 209

(a) qc is 1 : 1.5 - 1 : 1.7 for the stability factor N,

of 100 and the seismic coefficient of 0.15 in
the case that the internal friction angle of the
rockfill is 35” for which the material safety
factor of 1.2 is taken into account.
(b) The limit equilibrium and upper-bound limit
analysis give almost the same results for N, <
100 within engineering precision.
(c) The limit equilibrium method leads the dam
of N, > 100 to have an extremely gentle slope,
I I I I I which indicates conservatism inherent in the
1.5 2.0 2.5 3.0
conventional method, while the critical slope
Slope 7 determined by the upper-bound analysis is the
Fig. 7. Relation between failure acceleration A, and same for N, > 100. The difference is due to
slope q of model dam by shake table test (after the fact that the failure mechanism depends
Okamoto, 1971) on N, in the former but not in the latter.

the dam slope and therefore the corresponding

seismic coefficients are almost the same for each ACKNOWLEDGEMENTS
slope, and here it is called the equivalent seismic Katsushi Yamazaki and Toshikazu Murayama
coefficient K, , K, = 0.4 for the case of 0.3 g are thankfully acknowledged for their great help
input. It should be noted in Fig. (4a) that there with the numerical calculations.
are off-plotted symbols (semi-solid and hollow
circles) for the cases of gentle slopes subjected to
small input accelerations of 0.1 g and 0.2 g. This
can be easily understood by recalling the fact that REFERENCES
no failure occurs due to the inertia force, which Chang, C.-J., Chen, W. F. & Yao, J. T. P. (1984).
Seismic displacements in slopes by limit analysis. J.
corresponds to the seismic coefficient, less than
Geotech. Engng. Div., Am. Sot. Ciu. Engrs 110, GT7,
K, of Fig. 1. S&874.
K, is 0.15, 0.26 and 0.4 for the maximum input Charles, J. A. & Soares, M. M. (1984). Stability of com-
acceleration of 0.1 g, 0.2 g and 0.3 g, respectively. pacted rockfill slopes. GCotechnique 34, No. 1,61-70.
It is about 1.3 times greater than the maximum Chen, W. F. (1975). Limit analysis and soil plasticity, pp.
input horizontal acceleration at the base divided 411413. Amsterdam: Elsevier.
by gravity acceleration in the present three cases. Daddazio, R. P., Ettouney, M. M. & Sandler, I. S.
This is because the response amplification is (1987). Nonlinear dynamic slope stability analysis. J.
Geotech. Engng. Div., Am. Sot. Ciu. Engrs 113, GT4,
almost the same in these cases. The factor of 1.3
285-298.
might differ according to the response amplifica-
De Mello, V. F. B. (1977). Reflections on design deci-
tion characteristics of the dam, and therefore the sions of practical significance to embankment dams.
seismic coefficient may well be considered to be GCotechnique 27, No. 3, 279-355.
the maximum horizontal acceleration averaged Kikusawa, M. (1987). Seismic stability analysis of a
over the cross-section divided by gravity acceler- rockfill dam subjected to a severe earthquake.
ation. From numerous shake table tests Okamoto Tsuchi-to-Kiso, JSSMFE, 35, No. 11, 15-20. (In
(1971) presented a relation (Fig. 7) between the Japanese).
failure acceleration and the slope of the homoge- Kuwano, J. & Ishihara, K. (1988). Analysis of per-
neous triangular-shape model dam made of river manent deformation of earth dams due to earth-
quakes. Soils Fdns 28, No. 1,41-55.
gravels. N, of the model dam might be signifi-
Leshchinsky, D., Baker, R. & Silver, M. L. (1985). Three
cantly great because no cohesion is expected. The dimensional analysis of slope stability. Int. J. Numer.
curve in Fig. 4 given numerically is very similar to Analyt. Meth. Geomech. 9, 199-223.
that in Fig. 7 obtained from model experiments. Makdisi, F. I. & Seed, H. B. (1978). Simplified pro-
cedure for estimating dam and embankment
earthquake-induced deformations. J. Geotech.
CONCLUSIONS Engng. Dia., Am. Sot. Ciu. Engrs 104, GT7,849-867.
Pseudo-static analyses have been conducted for Marsal, R. J. (1987). Lessons learned from measure-
a typical rockfill dam with a central impervious ments in earth and rockfill dams. Seminar note on
Earthquake engineering of rockjill dams, Tokyo.
wall by means of the limit equilibrium method Matsumoto, N. & Watanabe, K. (1987). The shear
and the upper-bound limit analysis (Rigid-plastic strength of rockfill materials: T$uchi-to-Kiso,
finite element method being employed). The fol- JSSMFE. 35. No. 12.49-54. (In Jananese)
lowing conclusions are made for the critical slope Newmark, N. w. (1965):Effects bf earihquaies on dams
‘I, and the critical seismic coefficient K, and embankments. GCotechnique 15, No. 2, 137-160.

Downloaded by [ UC San Diego Libraries] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.
 210 KIKUSAWA

Okamoto, S. (1971). Earthquake resistant engineering. Shakal, A. F., Sherburne, R. W. & Parke, D. L. (1984).
Tokyo: Ohmusha (in Japanese). Principal features of the strong motion data from
Prater, E. G. (1985). Stability of cuttings considering the 1984 Morgan Hill earthquake. In The IY84
variable cohesion and earthquake effects. Proc. 11th Morgan Hill, California earthquake, special pub-
Int. Con& Soil Mech. Fdn Engng 4, 7/B/5, pp. 1959- lication 68, 249-264.
1962, San Francisco. Tamura, T., Kobayashi, S. SC Sumi, T. (1984). Limit
Sarma, S. K. (1975). Seismic stability of earthdams and analysis of soil structure by rigid plastic finite
embankments. GCotechnique 25, No. 4, 743-761. element method. Soils Fdns 24, No. 1, 34-42.
Sarma, S. K. & Barbosa, M. R. (1985). Seismic stability Tamura, T., Kobayashi, S. & Sumi, T. (1987). Rigid-
analyses for rockfill dams with central clay cores. plastic finite element method for frictional materials.
Gtotechnique 35, No. 3, 319-328. Soils Fdns 27, No. 3, 1-12.

Downloaded by [ UC San Diego Libraries] on [15/09/16]. Copyright © ICE Publishing, all rights reserved.