See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245286785

Earthquake Safety Evaluation of Gravity Dams Considering Aftershocks and

Reduced Drainage Efficiency

Article  in  Journal of Engineering Mechanics · January 2008

DOI: 10.1061/(ASCE)0733-9399(2008)134:1(12)

CITATIONS READS
36 968

2 authors:

Pierre-Marie Alliard Pierre Léger

GDF SUEZ Polytechnique Montréal
9 PUBLICATIONS   39 CITATIONS    35 PUBLICATIONS   470 CITATIONS   

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Pierre-Marie Alliard on 17 April 2018.

The user has requested enhancement of the downloaded file.

 Earthquake Safety Evaluation of Gravity Dams Considering
Aftershocks and Reduced Drainage Efficiency
Pierre-Marie Alliard1 and Pierre Léger, M.ASCE2

Abstract: This paper develops a methodology to perform seismic response analyses of concrete gravity dams considering aftershocks,
and reduced drainage efficiency due to disruption of the drainage system. A database of earthquake records has been assembled to
characterize main shocks and aftershocks. A drain finite difference hydraulic model considering laminar or turbulent two-dimensional flow
in connecting cracks, geometrical distortions due to joint dilatancy while sliding, and misaligned drain segments is developed and
validated. Coupled-hydromechanical analyses on the seismic response of a drained 90 m gravity dam are used to show the importance of
cumulative displacements on the increase in uplift pressures, drain flow, and reduced sliding safety factors. Aftershock response is
especially sensitive to the drainage system dimensions and model parameters, such as the foundation stiffness, and the number of
potentially sliding joints. The proposed methodology forms the basis to develop displacement based performance criteria in stability
evaluation of existing dams reevaluated for much higher ground motions intensities that they have been designed for many years ago.
DOI: 10.1061/共ASCE兲0733-9399共2008兲134:1共12兲
CE Database subject headings: Dam safety; Uplift pressure; Sliding; Displacement; Seismic effects; Dilatancy; Hydromechanics.

Introduction
Earthquakes have affected several large concrete gravity and but-
tress dams in the past producing cracks in the upper region as
well as at the base of the structures and disturbing seepage paths
through foundation rock joints 共Fig. 1兲 共Hsinfenkiang, China;
Koyna, India; and Sefid Rud, Iran兲. Historical evidence has
shown that the base uplift pressures following earthquakes 共1兲
could increase 共Koyna Dam, 1967 earthquake兲 共Pant 1990兲 some-
times necessitating redrilling of drains 共Mingtan Dam, Taiwan,
1999 earthquake兲 共Charlwood et al. 2000兲; 共2兲 could decrease
共SefidRud Dam, 1990 earthquake兲 共ICOLD 2001兲; 共3兲 or could
increase and then return to the preseismic level 共Japanese dams,
1995兲 共Matsumoto et al. 1996兲. Base uplift pressure variations
could be explained from seismically induced displacements oc-
curring in a jointed rock mass foundation. These change the pre-
seismic steady-state seepage regime, and corresponding water
heads, to reach a new equilibrium condition. Joints could open or
close, new cracks could be formed during the earthquake, and
they may be in a compressed state afterward. Moreover, drains
could be disrupted by sliding of cracks and joints along the dam
1
Civil Engineer, Assystem France, Energy and Nuclear Division,
CEA Cadarache, Bat 784, 13108 St-Paul-Lez-Durance, Cedex, France.
E-mail: pierre-marie.alliard@centraliens.net
2
Professor, Dept. of Civil, Geological and Mining Engineering, Ecole
Polytechnique de Montreal, Montreal Univ. Campus, P.O. Box 6079,
Station CV, Montreal, P.Q, Canada H3C 3A7 共corresponding author兲.
E-mail: pierre.leger@polymtl.ca
Note. Associate Editor: Brett F. Sanders. Discussion open until June
1, 2008. Separate discussions must be submitted for individual papers. To
extend the closing date by one month, a written request must be filed with
the ASCE Managing Editor. The manuscript for this paper was submitted
for review and possible publication on November 29, 2006; approved on
May 25, 2007. This paper is part of the Journal of Engineering Mechan-
ics, Vol. 134, No. 1, January 1, 2008. ©ASCE, ISSN 0733-9399/2008/1-
12–22/$25.00.
height. Current dam safety guidelines require that adequate slid-
ing and overturning stability be maintained in the postearthquake
condition using uplift pressure appropriate to the displaced dam
position allowing for opening of joints and reduced drain effi-
ciency E, 共Fig. 2, ANCOLD 1998; FERC 1999, 2002; CDA
2006兲. Moreover, FERC 共1999兲 explicitly mentioned that the re-
sidual strength after a major earthquake should allow the dam to
resists aftershocks.
A coupled dam-foundation-reservoir hydromechanical finite-
element model for static loads with consideration of seeping foun-
dation joints has been presented by Erban and Gell 共1988兲.
Amadei et al. 共1989a兲 presented analytical solutions for uplift
pressures in drained cracks under hydrostatic loads. Deschamps et
al. 共1999兲 presented a foundation drain flow model for the Blue-
stone gravity dam using the distinct element method. Lemos et al.
共1997兲 presented the earthquake response analyses of an impervi-
ous dam on a jointed foundation. A coupled hydromechanical
analysis was conducted using distinct elements to update the up-
lift pressures and water flow in the foundation as a function of
foundation joint displacements. However, there is no rational
model to assess the postearthquake dams’ drainage system effi-
ciency as a function of seismic crack opening and sliding dis-
placements in the dam body.
This paper develops the methodology to perform seismic
response of concrete gravity dams considering aftershocks, and
reduced drainage efficiency due to disruption of the drainage sys-
tem. A database of earthquake records has been assembled to
characterize main shocks and aftershocks. A drain finite differ-
ence hydraulic model considering laminar or turbulent flow in
connecting cracks, geometrical distortions due to joint dilatancy
while sliding, and misaligned drain segments is developed and
validated. Coupled-hydromechanical analyses on the seismic re-
sponse of a drained 90 m gravity dam is used to show the impor-
tance of cumulative displacements on the increase in uplift
pressures, drain flow, and reduced sliding safety factors.

12 / JOURNAL OF ENGINEERING MECHANICS © ASCE / JANUARY 2008

 Fig. 1. Gravity dam subjected to earthquake: 共a兲 dam; 共b兲 contraction joints; and 共c兲 opening along lift joints 共cracks兲
Characteristics of Earthquakes Main Shocks
and Aftershocks
Aftershocks Definitions
Earthquakes are usually part of a sequence of ground motions
which can be defined in three steps: foreshocks, the larger events
called main shocks, and then aftershocks. Seismologists usually
describe foreshocks and aftershocks as the beginning and the end
of the energy relaxation process, and observed that the sequence
frequency pattern depends on the structure of material and the
distribution of stress applied underground 共Scholtz 2002兲. Thus,
for a homogeneous material and uniform external stress, several
foreshocks precede the main shock, and aftershocks typically
begin immediately after the main shock. For a partially heteroge-
neous material and nonuniform stress, there are very few fore-
shocks but many aftershocks. The Koyna 共1967兲 and Kariba
共1963兲 reservoir induced earthquakes are known as famous ex-
amples of such a sequence. However, for extremely heteroge-
neous material and a very concentrated stress, no dominant
sequences of ground motions are distinguishable.
Among many local phenomena, two main interpretations of
aftershocks are detailed by Gupta 共1992兲 and Scholtz 共2002兲. The
first one is the fluid diffusion theory. It is based on the observation
that frequency patterns of aftershock sequences usually decrease
with a law inversely proportional to the time after the main shock.
Fig. 2. Uplift pressures for structural stability evaluation
JOURNAL OF ENGINEERING MECHANICS © ASCE / JANUARY 2008 / 13
Hyperbolic decrease is a characteristic of fluid diffusion phenom-
ena leading to the following explanation. Due to the main shock,
changes in ground stress distribution cause water migration from
compressed areas to more dilated areas. Then, water can modify
pore pressure and decrease the shear resistance of faults creating
aftershocks. Later, a second theory was proposed under the name
of stress triggered hypothesis. The idea is that the main shock
rupture causes a stress transfer to relatively close zones to the
fault. Because the earth crust has thousands of small faults of
varying orientations, this theory assumes that stress changes will
create ruptures of faults properly oriented for triggering. In addi-
tion to static stress triggering, one should also consider the dy-
namic stress change associated with main shock waves. Indeed, if
the wave amplitude is large enough, it can trigger aftershocks on
the wave path even very far from the main shock hypocenter. The
most famous example is the Hector Mine earthquake 共M = 7.1兲
whose one aftershock was 40 km away.
Database of Main Shocks and Aftershocks
The writers present herein the evolution of seismic parameters of
accelerograms between the main shock and the largest magnitude
aftershock. A list of 20 significant earthquakes was selected
共Table 1兲. Most of them are natural earthquakes, whereas five are
reservoir induced earthquakes 共RIS兲. Ground motion digital
Table 1. Significant Main Shock Aftershock Sequences
Main shock Largest aftershock
PGA PGA
Name Country Year Magn. 共g兲 Delay Magn. 共g兲 Database
Reservoir induced earthquakes
Koyna India 1967 6.3 0.5 10 months 5.1 0.15 NGDC
Hsinfengkiang China 1962 6.1 0.6 36 months 5.3 — —
Kariba Zimbawa 1963 6 — 2 days 5.8 — —
Hoover United States 1939 5 — 2 hours 4 — —
Natural earthquakes
Tangshan China 1976 7.8 — 1 day 7.1 — NGDC
Chi Chi Taiwan 1999 7.6 0.8 1 week 6.8 — PEER
Izmit Turkey 1999 7.4 0.4 2 weeks 5.2 0.2 ESD
Manjil Iran 1990 7.3 0.7 2 months 5 0.15 ESD
Borah Peak United States 1983 7.3 0.08 1 day 5.8 0.07 PEER
Spitak Armenia 1988 7.1 0.2 4 min 5.9 0.15 COSMOS
Nahanni Canada 1985 6.9 0.2 2 days 5.7 0.1 STRONGMO
Landers United States 1992 6.9 0.3 10 hours 6.7 0.06 COSMOS
Northridge United States 1994 6.7 0.4 2 months 5.3 0.2 COSMOS
Valparaiso Chile 1985 7.8 0.3 1 month 7.2 0.15 COSMOS
Imperial Valley United States 1979 6.5 0.4 3 min 5 0.3 g COSMOS
Loma Prieta United States 1989 6.5 0.3 33 hours 5 — —
Friuli Italy 1976 6.5 0.35 4 months 6.1 0.2 PEER
Morgan Hill United States 1984 6.2 0.3 9 days 4.5 — COSMOS
Whittier Nar. United States 1987 6.1 0.3 3 days 5.3 0.15 COSMOS
Saguenay Canada 1988 5.9 0.16 1 day 4.1 — STRONGMO

records were obtained from public databases: the US National
Geophysical Data Center 共NGDC兲, the Columbia University
Strong Motion Database 共STRONGMO兲, the European Strong
Motion Database 共ESD兲, the Berkeley Strong Motion Database
共PEER兲, and the Consortium of Organizations for Strong Motion
Observations System 共COSMOS兲. The assembled database of
main shocks and aftershocks is available from Alliard 2006. It
includes records as well as parameters of amplitude, intensity,
frequency content, response spectra, and duration of strong
shaking.
Characteristics of Aftershocks
After the main shock, the delay before the largest aftershock is
very variable, from several minutes to months. To forecast this
delay is still very difficult even considering the particular tectonic
properties of each region.
On the contrary, one can often forecast aftershock Richter
magnitudes. Indeed, in most of the sequences studied, they seem
to respect empirical decreasing laws presented in Scholtz 共2002兲,
called Bath’s law for tectonic earthquakes, and Papazachos’ law
for RIS.
Regarding aftershock amplitude parameters 共example, peak
ground acceleration, sustained maximal acceleration, Arias inten-
sity,…兲 in most cases, they decrease as compared to main shock
values 共Figs. 3 and 4兲. Nevertheless, to generalize is hazardous
because one cannot ignore examples like the Valparaiso earth-
quake whose aftershock was more severe than the main shock in
the EW direction. Actually, considering both results and stress
triggered theory, there is apparently no possible rule concerning a
possible principal direction of motion such that aftershocks can
act in a completely different direction from the main shock.
Frequency analyses have indicated that aftershock Fourier

Fig. 3. Spitak earthquake, Gukasyan Station, Direction 0: 共a兲 main shock accelerogram; 共b兲 aftershock accelerogram; and 共c兲
acceleration response spectra 共5%兲

14 / JOURNAL OF ENGINEERING MECHANICS © ASCE / JANUARY 2008

Fig. 4. Landers earthquake, Joshua Fire St. Station, Direction 0: 共a兲 main shock accelerogram; 共b兲 aftershock accelerogram; and 共c兲
acceleration response spectra 共5%兲
spectra in acceleration have less major peaks. They are smoother,
and the bandwidth is often smaller. The same remarks can be
made for the acceleration response spectra. Besides, in almost
half of the sequences, the mean period Tm of the Fourier spectra
and the predominant period T p of the acceleration response spec-
tra increase as compared to that of the main shock 关Fig. 3共c兲兴.
However, Fig. 4共c兲 shows, on the contrary, a decrease. Finally, the
significant duration of aftershocks can be as long as the main
shocks and long pulses are not rare 共Fig. 3兲.
Aftershocks Damage Potential
According to FERC 共1999兲, the damage potential of aftershocks
has to be considered in earthquake safety evaluation of concrete
dams. An increase of T p shows a risk of additional cumulative
damage for a structure whose fundamental period has increased
because of cracks due to a main shock. The duration of accelero-
grams, and especially the long pulses, may favor further sliding
along existing cracks.
Hydromechanical Damage due to Movements
along Cracked Planes
Mechanical Strength and Crack Movements
In massive structures such as concrete dams, lift joints are usually
considered as planes of weakness. Thus, during an earthquake,
joints may open and cracks may form. Sliding displacement ␦t
and coupled dilatancy ␦n along cracked joints have to be consid-
ered 关Fig. 1共c兲兴. Indraratna and Haque 共2000兲 presented state-of-
the-art related to shear behavior of rock joints. In our study, the
joint constitutive model used is based on Coulomb gap—friction
elements 关Fig. 5共b兲兴. Parameters are the normal and tangential
stiffness kN and kS, tensile strength f t, fracture energy GF, coeffi-
cient of friction ␮, and cohesion c. When a crack occurs, the
tensile strength and the cohesion are progressively reduced from
the assigned values to zero over the fracture process zone. The
coefficient of friction is dependent on surface preparation after
concrete placement. Joints are untreated or very often water
blasted. Thus, the model is based on the concepts of a concrete
basic friction defined by the angle ␸b, and large scale surface
waviness friction defined by the dilation angle i. Both parameters
are calculated using empirical equations according to the applied
normal stress 共Fronteddu et al. 1998兲. The dynamic sliding behav-
ior of small concrete to concrete specimen was obtained from
experimental results. The friction coefficient is defined as
JOURNAL OF ENGINEERING MECHANICS © ASCE / JANUARY 2008 / 15
␮ = tan共␸b + i兲 共1兲
Fig. 5共a兲 illustrates typical complete forward-backward joint slid-
ing displacement cycles. The first cycle begins with an elastic
response segment reaching a peak, and then exhibits a residual
sliding plateau in the forward direction. In subsequent cycles,
there is no longer peak shear strength. The value of ␮ on the
plateau is given by Eq. 共1兲.
Besides, when sliding occurs, coupled relative normal dis-
placement between the interfaces of a cracked joint has to be
considered. Dilatancy may be computed in a time history analyses
using the dilatancy angle
⌬␦n = tan共i兲 . ⌬␦t 共2兲
The integration of normal displacement increments ⌬␦n leads to
the total dilatancy ␦n. When large sliding ␦t happens, a limit to the
dilatancy equal to one quarter of the maximum aggregate size is
used. This limit was derived by the experimental results of Front-
eddu et al. 共1998兲. Other dilatancy models could have been used
based on the theory of plasticity considering nonassociative flow
rules.
Drainage and Water-Stop Damage Model
To study the behavior of a dam during an aftershock, and to
compute uplift pressures after the main shock, considering dam-
age like crack opening and sliding is necessary. A numerical so-
lution programmed in MATLAB 共2004兲 is proposed to model uplift
pressures inside a drained monolith. Three kinds of joints are
studied: pervious 共nondamaged兲, partially cracked, and fully
cracked.
Fig. 5. Joint constitutive model: 共a兲 shear strength; 共b兲 gap friction
element
 Fig. 6. Equivalent hydraulic conductivity K of crack and exponent ␣ 共adapted from Louis 1969兲
Problem Geometry and Parameters
To calculate the cumulative drain flow, the head distribution h
along a joint, the efficiency E, the global uplift resultant, and the
overturning moment, the problem geometry, and parameters are
defined in Fig. 7共a兲. The joint is assumed to be horizontal, finite,
and rectangular: the length is L and the width is a. The parameter
a is given by the spacing between each drain line along the lon-
gitudinal direction. The joint is intersected by a single circular
drainage well centered at the distance y = y d from the upstream
face and at the middle of the width 共x = 0兲. The drain radius is Rd.
The discharge going to the drain is noted as Qd. Under normal
conditions, the joint is assumed pervious and discharge through
the drain is limited to seeping flow. The joint hydraulic conduc-
tivity is noted as K.
If the joint is open 共cracked兲, notations and hypothesis are as
follows. Water is assumed to enter into the crack, and the flow is
steady, incompressible, and two dimensional. The laminar-
turbulent flow transition usually affects significantly the uplift
pressure distribution. The kinematic viscosity ␯ is a decreasing
function of the water temperature at a particular joint entrance.
Crack geometrical parameters 关Fig. 1共b兲兴 are the aperture b, and
the absolute roughness k 共average height of crack wall asperities
in mm兲. For a partially cracked joint, the length of the crack is
Lc ⬍ L, and the hydraulic aperture is assumed constant. For a fully
cracked joint, Lc = L and the aperture is defined by a function b共y兲.
The water head 共h兲 boundaries conditions of the joint are
h = Hup at the boundary with the upstream face, h = Hdown at the
downstream face, and h = Hd in the drain. Boundaries with others
monoliths are assumed impervious, but if relative displacements
leading to water-stop damage are considered between monoliths,
the boundary condition is changed to a fixed head along a trans-
verse joint. For a partially cracked joint, the downstream crack
boundary is considered impervious for hydraulic calculation but a
triangular uplift pressure distribution is presumed to take place in
the ligament over a long period of time.
Head losses at the crack entrance and near the drain are ne-
glected. The validity of this approximation is discussed in detail
by Amadei et al. 共1989b兲: head losses slightly increase uplift pres-
sures but this effect is only significant for drain of small diam-
eters, less than 100 mm, and for smooth cracks with large
apertures.
16 / JOURNAL OF ENGINEERING MECHANICS © ASCE / JANUARY 2008
Flow Laws
For a nondamaged joint, flow is in an isotropic porous medium.
Darcy’s law 关Eq. 共3a兲兴 and the mass conservation Equation 关Eq.
共3b兲兴 lead to the classical Laplace’s equation 关Eq. 共3c兲兴, where
K = conductivity; v = velocity; and I = hydraulic gradient
v=K.I 共3a兲
ⵜ.v=0 共3b兲
ⵜ 2h = 0 共3c兲
For a damaged joint, flow laws in a crack with walls of varying
degrees of roughness were first established by Louis 共1969兲 and
could be also found in Wittke 共1990兲 and Amadei et al. 共1989a兲.
Let us define the crack hydraulic diameter Dh = 2b, the relative
roughness k / Dh 共between 0 in smooth cracks and 0.5 in rough
cracks兲, and the Reynold’s number R = Dhv / ␯ where
v = water velocity. Louis 共1969兲 classified five different flow con-
ditions according to the relative roughness and the Reynold’s
number. For each flow zone, the velocity v is obtained from Eq.
共4兲 in which the crack equivalent hydraulic conductivity K and
the constant exponent ␣ are defined in Fig. 6. In the particular
case of laminar condition with constant aperture, Eq. 共4兲 leads to
v = − KI␣ 共4兲
Hydraulic Zone I is for smooth crack and laminar flow. Isolines,
streamlines, and the Reynold’s number along the y axis are illus-
trated in Fig. 7共c兲 for a partially cracked joint in the hydraulic
Zone I 共Hup = 100 m兲. Zone II is for smooth crack and turbulent
flow. Zone III is for a hydraulically rough crack and turbulent
flow. For rougher cracks 共k / Dh ⬎ 0.033兲, Zone IV is a laminar
flow, and Zone V is a turbulent flow.
Finite Difference 2D Solution of Hydraulic Problem
The writers used a constant mesh size p 共m兲 in all sections. Let us
first assume a laminar flow 共Zone I or IV兲 and write at each node
i the flow mass conservation using Eq. 共3a兲 and Eq. 共4兲. Close
nodes are numbers i-j, i + j, i − 1, and i + 1 关Fig. 7共a兲兴. In the gov-
erning equation 关Eq. 共5兲兴, the hydraulic conductivity and the ap-
erture between two nodes i and j are noted Ki,j and bi,j. The head
at node i is hi
Fig. 7. Uplift pressure model: 共a兲 geometry; 共b兲 discharge in crack for different values of aperture b; 共c兲 isopressure lines, stream lines, and
Reynold’s number; and 共d兲 h / Hup along y axis at x = 0 for different values of a / Lc 共parameters similar to Amadei et al. 1989a兲
共Ki,i−jbi,i−j兲hi−j + 共Ki,i−1bi,i−j兲hi−1 − 共Ki,i−jbi,i−j + Ki,i+jbi,i−j
+ Ki,i−1bi,i+1 + bi,i+jKi,i+1兲hi + 共bi,i+1Ki,i+1兲hi+1
+ 共bi,i+jKi,i+j兲hi+j = 0
共n兲 共n兲
共Ki,i−jbi,i−j兲hi−j + 共Ki,i−1bi,i−j兲hi−1 − 共Ki,i−jbi,i−j + Ki,i+jbi,i−j
+ Ki,i−1bi,i+1 + bi,i+jKi,i+1兲h共n兲 共n兲
i + 共bi,i+1Ki,i+1兲hi+1
n−1
+ 共n兲
共bi,i+jKi,i+j兲hi+j =− 兺
s=1
⌬q共s兲
i 共6兲
Thus, adding boundary conditions, the laminar flow assumption
enables us to build a linear system of equations. But, when tur-
bulent flow conditions should actually be considered 共Zones II,
III, and V兲, the velocity is no longer proportional to the hydraulic
gradient and the formulation of the problem leads to a system of
nonlinear equations. To solve theses equations, the writers used
an iterative method described by Wittke 共1990兲. The laminar flow
condition is the starting point of the iterative process. Then, after
iteration n-1, for each node i where the Reynolds number exceeds
the laminar limit, a discharge correction ⌬q共n-1兲
i is calculated with
the difference between the discharge using the laminar formula-
tion and the discharge using the turbulent formulation. This dis-
charge correction is then included in Eq. 共5兲 to obtain Eq. 共6兲 for
the next iteration n, and a new system of linear equations is
solved. The actual conditions of turbulent flow are thus gradually
approached. This method is very efficient because the head dis-
tribution usually converges after only a few iterations.
Drain Boundary Condition
In the finite difference method, the drain is represented by only
one node. Yet, the drain boundary condition h = Hd must be along
a circle of radius Rd and centered on this node 关Fig. 7共a兲兴. To
achieve this condition, the hydraulic head has to be slightly dif-
ferent at the center of the circle. The program makes a linear
interpolation with close nodes to calculate the head at the distance
共5兲 Rd from the center of the drain and to adjust iteratively the correct
boundary condition. The adequacy of linear interpolation around
the drain depends on the finite difference mesh. With a large mesh
the singularity effect produces too small values of the pressure
head h near the drain. With a fine mesh, the representation of h
near the drain is correct 关see validation in Figs. 7共a and d兲兴. The
drain boundary condition also depends on the drain state. Under
normal operation, inside the drainage well, the pressure is atmo-
spheric above a gallery located at the dam-foundation interface.
The total discharge in the drainage well is obtained by summation
of the discharges supplied by all seeping joints or opened cracks.
The drain maximum discharge capacity Qmax 共m3 / s兲 at any level
is computed considering steady turbulent flow in a gravitational
field inside a pipe with smooth walls 关Blasius law in Eq. 共7兲兴
共Wittke 1990兲. When this value is exceeded, the drainage system
is actually drowned so that the head inside the drain equals the
head in the reservoir. In this case, drain efficiency is zero
冋 冉冊 册
1/4 4/7
g 2
Qmax = . 共Rd兲5/4 . ␲R2d 共7兲
0.079 ␯
Validation
For a pervious joint, the writers validated the numerical results
with analytical solutions 共Novak et al. 1996兲. For a cracked joint,
the writers compared parametric studies with solutions given by
Amadei et al. 共1989a兲. An example is illustrated in Fig. 7共d兲
where the ratio a / Lc is variable. Discharge calculations were also
compared to results presented by Wittke 共1990兲. Fig. 7共b兲 shows
JOURNAL OF ENGINEERING MECHANICS © ASCE / JANUARY 2008 / 17
Fig. 8. Influence of sliding displacement on discharges in cracked lift
joint
the evolution from laminar to turbulent flow for the discharge
in a crack without any drain for different values of constant
aperture b.
Influence of Sliding Displacement
As shown in Fig. 8, sliding displacement ␦t coupled with crack
opening b共␦t兲 at y = y d enables water coming from the drain above
to enter into the crack plane studied. For the uplift pressure cal-
culations of a lift joint, this is considered as a new boundary
condition in the vicinity of the drain. Coming from the drain flow
above the crack, the fraction of discharge f共␦t兲 going into the
crack is estimated in Eq. 共8兲. In this equation, the ratio of the area
A1共␦t兲 over the section ␲R2d controls the fraction of water dis-
turbed by the projection. This flow has two paths available. It can
go into the crack through the entrance cross section area
L1共␦t兲 . b共␦t兲 whose conductivity K1 is given by a cubic law 关Eq.
共9a兲兴. The second path is to fall into the lower drain, passing
through an assumed cross section 共2␲Rd−L1共␦t兲兲.共b共␦t兲2
+共2Rd−␦t兲2兲1/2 whose conductivity is K2 关Eq. 共9b兲兴. Such a
model may overestimate the quantity of water entering into the
crack because no energy loss is considered
冉 冊冉 冊
A1共␦t兲 K1
f= if兩␦t兩 ⬍ 2Rd, else f = 0
␲R2d K1 + K2
g . L1共␦t兲 . b共␦t兲3
K1 =
12␯
g . 共2␲Rd − L1共␦t兲兲 . 共b共␦t兲2 + 共2Rd − ␦t兲2兲3/2
K2 =
12␯
For the dam-foundation joint under the base gallery, the drain
well is assumed to be filled with water. Sliding displacement re-
duces locally the drain size so that it disturbs water flow. This
condition increases pressure and the hydraulic head considered in
the drain. The corresponding boundary condition is defined in Eq.
共10兲, where Hg = base gallery elevation, and assuming Hdown = 0
A1共␦t兲
Hd = Hg + 共Hup − Hg兲 if兩␦t兩 ⬍ 2Rd, else Hd = Hg
␲R2d
Parametric studies showed that sliding displacement influences
significantly uplift pressures only when an important discharge
exists in the drain above, and when the crack aperture is less than
18 / JOURNAL OF ENGINEERING MECHANICS © ASCE / JANUARY 2008
Fig. 9. Influence of sliding displacement on uplift pressure resultant
U 共Hup = 18 m, Hd = Hdown = 0 m, Rd = 75 mm, k = 0.3 mm, a = 4 m,
L = 17 m, Qd = 0.3 m3 / s兲
several mm. In such conditions, the resultant uplift pressure in-
creases when sliding displacement is greater than 90% of the
drain diameter 共Fig. 9兲. For large cracks, the hydraulic conductiv-
ity is so high that an additional flow due to sliding considerations
has no influence.
Water Stops
The drainage damage model implicitly considers that water stops
are not damaged and prevent water from going into the inter-
monoliths transverse joints. Therefore, one should keep in mind
that this assumption is questionable in some particular situations:
if there is a large differential motion between adjacent monoliths,
water-stops deformability can be exceeded losing water tightness.
Conventional design of a transverse joint is represented in Fig.
1共b兲. Materials used are usually copper, rubber, and PVC. Studies
of Jinsheng et al. 共2006兲 proposed several design recommenda-
tions to choose proper water-stop seismic resistant configuration
and materials. Such water stops could bear a water head of 290 m
and differential deformations of 20 mm.
共8兲 To illustrate the effect of one damaged water stop, Fig. 10共b兲
shows the increase in relative percentage to the initial value of the
uplift pressure resultant U and the discharge flow through drain
Qd for different values of the crack length. Equipotential lines and
共9a兲 stream lines are clearly affected 关Fig. 10共a兲兴, compared to the
same partially cracked joint with efficient water stops 关Fig. 7共c兲兴.
共9b兲
Seismic Analyses of Drained Dam Considering
Postseismic Conditions and Aftershocks
A seismic response analyses methodology for drained concrete
gravity dams is proposed in Fig. 11. First, the finite-element dam-
reservoir-foundation model is defined, the input ground motion is
selected, and the initial uplift pressure is computed. Then, seismic
analyses are performed with a main shock. Uplift pressures are
considered constant during the shaking in accordance with current
safety guidelines 共ANCOLD 1998; FERC 2002; CDA 2006兲. This
is a conservative assumption when the joints have an opening
共10兲
tendency. The joint constitutive model is implemented using gap-
friction elements 共Fronteddu et al. 1998兲. Time history analyses
enables us to compute sliding displacements and coupled dila-
tancy along cracked planes. The damaged drainage model is used
Fig. 10. Lift joint with one damaged waterstop 共at x / a = 0.5兲: 共a兲
isopressure lines and stream lines 共Y d / Lc = 0.25, a / Lc = 0.25, Fig. 11. Seismic safety evaluation considering severe aftershocks
L / Lc = 3, Hup = 100 m, Hd / Hup = 0.2, Hdown = 0, Rd / Lc = 0.00417,
k / 2b = 0.01, b = 0.1 mm兲; 共b兲 increase in percent of discharge to drain
Qd and of uplift resultant U for different values of L / Lc Seismic and Postseismic Response Analyses

Comparison of Seven Lift Joints Dam Model

and Single Lift Joint Dam Model
to update uplift pressures and drained water flow. At this stage, a
The Spitak event 共1988 earthquake兲 was chosen as input motions.
postseismic static analysis can be done to evaluate the dam sta-
A 8 s segment of the main shock, representing the strong shaking,
bility. The sliding safety factor 共SSF兲 can then be computed as the
was scaled to obtain a peak ground acceleration 共PGA兲 of 0.30 g.
ratio of available shear strength that could be mobilized over the
The aftershock was reduced to a 4 s significant segment, and the
driving shear force along a given plane. Aftershock response
PGA obtained with the same scaling was 0.17 g.
analyses are then performed in the same way as that for the main
The following finite-element dam models were considered:
shock. Uplift pressures are again updated considering the residual
• Model A: one joint at the base, and seven lift joints along the
displacements and dilatancy, and the final postseismic stability is
height 共Fig. 12兲; and
investigated. Drain seepage analyses are not performed during
• Model B: one joint at the base, and only one lift joint 共joint
transient applications of earthquake ground motions.
No. 6 in Fig. 12兲.

Dam-Foundation-Reservoir System Analyzed

To illustrate the proposed seismic analysis method, the sliding

response of a 90 m concrete gravity dam is investigated. The
concrete properties are the elastic modulus Ed = 27,690 MPa,
Poisson’s ratio ␯ = 0.2, and mass density ␳ = 2,400 kg/ m3. The up-
stream reservoir elevation is 86 m, and no downstream reservoir
is considered. In addition to self-weight and hydrostatic pressure,
Westergaard added masses are used to represent hydrodynamic
forces. Viscous damping is stiffness proportional with a value of
5% in the fundamental mode of the dam 共T1 = 0.27 s兲. The drain-
age system parameters are a = 4 m, Y d = 3 m, and Rd = 37.5 mm.
Viscosity is the function of the water temperature 共Wittke 1990;
from ␯ = 1.3e − 6 m2 / s for T = 10° C at the top of the reservoir, to
␯ = 1.5e − 6 m2 / s for T = 4 ° C at the bottom兲. The cracks absolute
roughness is k = 3 mm. Joints are water blasted: the residual aver-
age basic friction angle ␸b varies from 30.8° at the base to 32° at
the top; the average dilatancy angle is between 12° 共base兲 and 17°
共top兲 共Fronteddu et al. 1998兲. Fig. 12. 90-m gravity dam

JOURNAL OF ENGINEERING MECHANICS © ASCE / JANUARY 2008 / 19


Fig. 13. Sliding displacements of 90 m dam during Spitak ground
motions: 共a兲 seven lift joints Model A; 共b兲 one lift joint Model B
In both models, the maximum dilatancy 共5 mm兲 was obtained
in the upper joints after the main shock. The resulting discharge
drowned the drain. Uplift pressure was updated with this new
condition thus favoring additional sliding displacement during the
aftershock: 0.6 mm at the base and 10.7 mm at the top in Model
A; 1.6 mm at the base and 5 mm at the top in Model B. Com-
pared to a single lift joint dam 共Model B兲, a seven lift joints dam
共Model A兲 reduced the residual sliding displacement, including
the aftershock response, from 5 to 3 mm at the base, and from
163 to 37 mm in the upper joint of the dam 共Fig. 13兲. Neverthe-
less, historical evidence 共Koyna, India, Sefid Rud, Iran, and Hsin-
fengkiang, China兲 have showed that a single significant crack in
the upper part 共Model B兲 is more likely to occur. In this model,
the SSF has decreased between the preseismic and the postseis-
mic conditions, from 1.98 to 1.13 at the base, and from 3.31 to
1.90 at joint No. 6.
Each added joint or failure plane in the dam creates a new
zone for dissipation of seismic energy from tensile fracture and
sliding friction. Large displacement values computed from Model
B could be due to a smaller amount of seismic input energy dis-
sipation as compared to Model A. To follow historical evidence,
Model A could be used allowing cracking at all lift joints but
allowing friction sliding only at the base and at a top joint. More-
over comparison of computed sliding displacements from Models
A and B with those that actually occurred at Sefid Rud Dam, for
example, is difficult because sliding displacements are particu-
larly sensitive to the presence of long acceleration pulses in the
ground motion records.
This example also illustrates the importance of the choice of
drain size. Fig. 14 shows the evolution of uplift pressure distribu-
tion in joint No. 6 of Model A with a 75-mm-diameter drain. The
drain is drowned and the efficiency 共E兲 is zero. However, with a
150 mm drain diameter, E is only reduced from 72 to 61%, and
SSF decreases from 3.32 to 2.53.
20 / JOURNAL OF ENGINEERING MECHANICS © ASCE / JANUARY 2008
Fig. 14. Uplift pressure evolution in lift joint No. 6 共Model A兲: 共a兲
preseismic; 共b兲 postseismic 共Rd = 75 mm兲; and 共c兲 postseismic
共Rd = 37.5 mm兲
Influence of Foundation Stiffness on Sliding
Displacement
The single lift joint model 共Model B兲 was studied for different
values of the elastic modulus E f of the foundation using effective
springs at the base. A large 150 mm body and foundation drain
diameter was used to promote a nondrowned drain condition.
Responses to Spitak records 共1988 earthquake兲 were compared
to Nahanni records 共1985 earthquake兲. The Nahanni main shock
PGA was also scaled to 0.30 g, and the corresponding aftershock
PGA was 0.16 g. Frequency parameters of the sequences are
clearly different: the Nahanni predominant period T p of the re-
sponse spectrum is 0.06 s in the main shock and the aftershock;
whereas in Spitak, T p is equal to 0.26 s in the main shock, and
0.54 s in the aftershock.
The final displacements are illustrated in Fig. 15. The founda-
tion stiffness has a significant influence on sliding depending on
the frequency characteristics of the earthquake. In Spitak
共T p ⬎ T1兲, a soft foundation increases sliding displacements. On
the contrary, in Nahanni 共T p ⬍ T1兲, sliding is increased by a rigid
foundation.
Summary and Conclusion
Aftershocks were defined and almost 20 acceleration record pa-
rameters were compared to the main shocks. The great variety of
physical and mechanical phenomena associated with aftershocks
explains the difficulties in characterizing the evolution of the
strong motion parameters between a main shock and its largest
aftershock. Nevertheless, the damage potential of aftershocks is
clearly apparent on some records with a PGA possibly as high as
in the main shock, an increase of the predominant period in the
response spectrum, some long pulses, or a long duration of
shaking.
A hydromechanical method based on displacement calcula-
tions was proposed to evaluate the seismic safety of concrete
gravity dams considering severe aftershocks. A Coulomb joint
constitutive model was used to compute sliding displacement and
coupled dilatancy along cracked planes in finite-element time his-
tory analyses. A finite difference numerical solution was devel-
oped to model uplift pressures in cracks considering a damaged
drainage system due to crack opening and residual sliding dis-
placements. Water was assumed to enter into cracks, and flow was
steady, uncompressible, two dimensional, and could be laminar or
turbulent. From the proposed drainage model, several practical
Fig. 15. Sliding displacements for different stiffness of foundation: 共a兲 mesh and deformation; 共b兲 Spitak ground motions; and 共c兲
Nahanni ground motions
conclusions were obtained from parametric analyses:
1. The laminar-turbulent transition affects the drain efficiency
共E兲. In a fully cracked joint, when the aperture reaches sev-
eral mm, E is typically reduced from 70 to 60%;
2. In a partially cracked joint, where the crack extends past the
drain, when the drainage system is not drowned, a significant
uplift pressure reduction after the drainage well has to be
considered. The typical value of E = 50%, recommended in
ANCOLD 共1998兲, sometimes leads us to underestimate uplift
pressures when the ratio of the monolith width over the crack
length exceeds 0.5;
3. If the drainage system is drowned, E = 0% and uplift pres-
sures are quite larger than without effective drains so that
dam stability might be compromised. Drainage wells diam-
eter and spacing have to be carefully designed for discharges
under normal operations and postseismic conditions;
4. Damaged water stops strongly increase resultant uplift pres-
sure and discharge to the drain; and
5. When the value of sliding displacement along joints and
cracks exceeds 90% of the drain diameter, resultant uplift
pressure increases significantly, depending on the discharge
coming from the drain above the crack, and the hydraulic
conductivity of the crack.
Concerning the hydromechanical seismic safety evaluation
method proposed in this paper, the following conclusions were
obtained:
1. Aftershocks could create additional damage and sliding dis-
placements. Results are sensitive, especially regarding drain
diameter and spacing;
2. The adopted joint constitutive model strongly influences slid-
ing displacements and dilatancy computations, and conse-
quently affects uplift pressures computations;
3. For the 90 m dam model analyzed, the foundation stiffness
increases sliding when the predominant period T p of the
strong motion acceleration response spectrum is shorter than
the fundamental period T1 of the dam, and decreases sliding
when T p is longer than T1; and
4. The number of potentially sliding joints considered in the
dam’s finite-element model is a very important choice. Ac-
cording to historical observations, a model with only one or
two lift joints in the upper part of the dam is more realistic.
Consequences are larger sliding displacements than in a mul-
tijointed model.
The seismic safety evaluation methodology presented herein is
mainly limited by its sensitivity to the user’s adequate choices of
parameters to make a rational model of the dam-foundation-
JOURNAL OF ENGINEERING MECHANICS © ASCE / JANUARY 2008 / 21
reservoir system. Assumptions of rectangular and horizontal
confined aquifer cracks with uniform roughness may also be re-
strictive compared to reality. Regarding the influence of sliding
displacements on uplift pressures in a cracked segmented drain
system, the model should be confirmed by experimental results.
Nevertheless, the proposed methodology is an effective tool to
understand the seismically induced disruption of drainage
systems. Finally, it forms the basis of a displacement based per-
formance approach in stability evaluation of existing dams re-
evaluated for much higher ground motions intensities that were
designed many years ago.
Acknowledgments
Financial support provided by the Quebec Fund for Research on
Nature and Technology 共FQRNT兲, and the Natural Science
and Engineering Research Council of Canada 共NSERC兲, is
acknowledged.
References
Alliard, P.-M. 共2006兲. “Mainshocks and aftershocks sequences database.”
具http://www.polymtl.ca/structures/en/telecharg/index.php典.
Amadei, B., Illangasekare, T., Morris, D. I., and Boggs, H. 共1989a兲.
“Estimation of uplift in cracks in older concrete gravity dams.” J.
Energy Eng., 115共1兲, 19–38.
Amadei, B., Illangasekare, T., Morris, D. I., and Boggs, H. 共1989b兲.
“Estimation of uplift in cracks in older concrete gravity dams. 2:
Effect of head losses in drain pipes on uplift.” J. Energy Eng., 115共1兲,
39–46.
Australian National Committee On Large Dams 共ANCOLD兲. 共1998兲.
Guidelines for design of dams for earthquake, Australia.
Canadian Dam Association 共CDA兲. 共2006兲. Dam safety guidelines,
Edmonton, Alberta, Canada.
Charlwood, R., Little, T. E., and Lou, J. K. 共2000兲. “A review of the
performance of two large sub stations and eight large dams during the
Chi Chi Taiwan earthquake.” ICLR Research, Paper Series No. 6,
Institute for Catastrophic Loss Reduction, McMaster Univ., Hamilton,
Ont., Canada.
Deschamps, R., Yankey, G., and Bentler, D. J. 共1999兲. “Modeling uplift
and drain flow at Bluestone dam.” Proc., American Association of
State Dam Safety Officials (ASDSO) Conf. 共CD-ROM兲, St. Louis.
Erban, P. J., and Gell, K. 共1988兲. “Consideration of the interaction
between dam and bedrock in a coupled mechanic-hydraulic FE
program.” Rock Mech. Rock Eng., 21, 99–117.
Federal Energy Regulatory Commission 共FERC兲. 共1999兲. “Engineering
guidelines for the evaluation of hydropower projects. Chapter XI:
Arch dams.” Department of Energy, Washington, D.C., 具http://
www.ferc.gov/hydro/docsEngGuide/guidelines.htm典.
Federal Energy Regulatory Commission 共FERC兲. 共2002兲. “Engineering
guidelines for the evaluation of hydropower projects. Chapter III:
Gravity dams.” Department of Energy, Washington, D.C., 具http://
www.ferc.gov/hydro/docsEngGuide/guidelines.htm典.
Fronteddu, L., Léger, P., and Tinawi, R. 共1998兲. “Static and dynamic
behavior of concrete lift joints interfaces.” J. Struct. Eng., 124共12兲,
1418–1430.
Gupta, H. K. 共1992兲. Developments in geotechnical engineering. 64:
Reservoir-induced-earthquakes, Elsevier, Amsterdam, The Nether-
lands.
Indraratna, B., and Haque, A. 共2000兲. Shear behavior of rock joints,
Balkema, Rotterdam, The Netherlands.
International Commission on Large Dams 共ICOLD兲. 共2001兲. “Design fea-
tures of dams to resist seismic ground motion—Guidelines and case
studies.” Bulletin 120, Paris.
Jinsheng, J., Guiying, Z., Juato, H., and Yang, Q. 共2006兲. “Studies for the
transverse joint waterstops at Xiaowan dam.” Hydropower and Dams,
13共1兲, 50–52.
22 / JOURNAL OF ENGINEERING MECHANICS © ASCE / JANUARY 2008
View publication stats
Lemos, J. V., Cundall, P. A., and Dasgupta, B. 共1997兲. “Earthquake analy-
sis of concrete gravity dams on jointed rock foundations.” Proc., 2nd
Int. Conf., Dam Safety Evaluation, Balkema, Rotterdam, The Nether-
lands, 339–350.
Louis, C. 共1969兲. “A study of groundwater flow in jointed rock and its
influence on the stability of rock masses.” Rock Mech. Res. Rep. No.
10, Imperial College, London.
MATLAB. 共2004兲. The language of technical computing, version 7.0.1,
The Mathworks Inc., Natick, Mass.
Matsumoto, N., Nakamura, A., Sasaki, T., and Iwashita, T. 共1996兲. “Ef-
fects on dams of the great Hanshin earthquake 共the Hyogoken-Nanbu
earthquake兲.” Proc., Japan Society of Soils and Foundations, Japan.
Novak, P., Moffat, A. I. B., Nalluri, C., and Narayanan, R. 共1996兲.
Hydraulic structures, 2nd Ed., E & FN Spon, London.
Pant, B. 共1990兲. “Structural behavior of concrete and masonry gravity
dams.” Publication No. 215, Central Board of Irrigation and Power,
New Delhi, India.
Scholtz, C. H. 共2002兲. The mechanics of earthquakes and faulting, 2nd
Ed., Cambridge University Press, Cambridge, U.K.
Wittke, W. 共1990兲. Rock mechanics: Theory and applications with case
histories, Springer, Berlin.