EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2008; 37:809–823

Published online 11 February 2008 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/eqe.791

Experimental seismic investigation of Sefid-rud concrete buttress

dam model on shaking table

A. R. Ghaemmaghami § and M. Ghaemian ∗, †, ‡

Sharif University of Technology, Tehran, Iran

SUMMARY
Owing to the devastating M7.6 earthquake of 20 June 1990 that occurred in the northern province of
Iran, Sefid-rud concrete buttress dam located near the epicenter was severely shaken. The crack penetrated
throughout the dam thickness near slope discontinuity, causing severe leakage, but with no general failure.
In this study, nonlinear seismic response of the highest monolith with empty reservoir is investigated
experimentally through model testing. A geometric-scaled model of 1:30 was tested on a shaking table with
high-frequency capability to study dynamic cracking of the model and serve as data for nonlinear computer
model calibration. Three construction joints are set up in the model to simulate effects of construction
aspects. The experimental results are then compared with smeared crack and damage mechanics finite-
element simulations using nonlinear concrete constitutive models based on fracture mechanics. The crack
patterns obtained from numerical models are in good agreement with those obtained from shaking table
tests for the case of including construction joint effects and rigid foundation. Copyright q 2008 John
Wiley & Sons, Ltd.

Received 6 July 2007; Revised 5 November 2007; Accepted 30 November 2007

KEY WORDS: Sefid-rud dam; buttress dam; seismic properties; shaking table; model test; nonlinear
fracture mechanics

INTRODUCTION

Sefid-rud concrete buttress dam was built during the period of 1958–1962 in the northern Iranian
province of Gilan. It is 106 m high in the middle section and 425 m long at the crest. It consists
of 26 monoliths, each 14 m long. The slope of the dam on the downstream face is 1 in 0.6 and

∗ Correspondence to: M. Ghaemian, Sharif University of Technology, Tehran, Iran.

†
E-mail: ghaemian@sharif.edu
‡
Associate Professor.
§ Ph.D. Student.

Copyright q 2008 John Wiley & Sons, Ltd.

810 A. R. GHAEMMAGHAMI AND M. GHAEMIAN

on the upstream side 1 in 0.4. This dam is constructed of plain concrete on a foundation material
of andesite and basalt. Its reservoir is more than 1700 million cubic meters in volume and is vital
for irrigating the vast rice fields in the downstream.
The dam was originally designed for static seismic coefficient of 0.25. An exceptional strong
ground motion in the seismotectonic province of Gilan in northern Iran occurred on 20 June
1990. Its magnitude is estimated to be 7.6 in surface wave scale. The epicenter of the catas-
trophic earthquake, which destroyed the two large cities of Manjil and Rudbar, was only about
5 km far from the dam. Following the earthquake, the most serious observed damage to the
dam was horizontal cracks that appeared in the upper parts of the monoliths, especially in the
highest monolith. A major crack ran almost the whole length of the dam at about 14 m below
the crest. Site investigations following the earthquake indicated that the dam monoliths moved
independently with permanent differential displacements of up to 50 mm. These large relative
displacements damaged the seals in the contraction joints and caused water leakage from the
joints.
Sefid-rud dam is one of the three large concrete dams that were severely damaged by earthquakes.
The others include Hsinfengkiang buttress dam in China during the 1962 Heyuan earthquake and
Koyna gravity dam during the 1967 Koyna earthquake in India. However, only the earthquake
damage of the Koyna dam has been studied extensively.
Most of the research conducted in the area of dam engineering is theoretical in nature. Although
substantial progress has been achieved in mathematical modeling, a few of various models have
been verified. Dam model testing has the potential of being used for analysis verification purpose.
Owing to the complex nature of the problem, the large size of the structure and difficulties in
physical modeling, little experimental work has been pursued in small-scale modeling of concrete
dams.
A significant number of early experimental research programs were concentrated on the nonlinear
seismic behavior of dam models. Donlon and Hall [1] tested a small-scale model of a dam
monolith using a shaking table. Difficulties were encountered due to shrinkage cracking of the
model in the process of drying. Ghobara and Ghaemian [2] conducted an experimental study
of a small-scale model of Pine Flat gravity dam to simulate the hydrostatic, hydrodynamic and
seismic loads. Mir and Taylor [3, 4] conducted an experimental investigation of the nonlinear
seismic response of a 30 m high concrete gravity dam to justify the validity of Westergaard’s
added mass theory. Harris et al. [5] used a two-dimensional model of Koyna dam on shaking
table to serve as data for nonlinear computer model calibration. Zhou et al. [6] performed a
series of dynamic experiments for 20 models of high arch dams on shaking table, with the
water ignored. Li et al. [7] performed a seismic analysis of the powerhouse monolith of the
Three Georges through model testing on a shaking table and numerical simulation using a three-
dimensional finite-element model of the structure. Tinawi et al. [8] conducted shake table experi-
ments on four 3.4 m high plain concrete gravity dam models to study their dynamic cracking and
sliding response. The experimental results were then compared with a smeared cracked numerical
simulation.
The difficulties in experimental modeling of concrete dams are mostly in material modeling,
simulating of reservoir and foundation effects and the availability of suitable equipment for
testing. The objective of this study is to investigate nonlinear seismic behavior of Sefid-rud dam
through the use of a small-scale model of the highest monolith at a geometric scale of 1:30 on
shaking table. Combined experimental and numerical analysis is performed to calibrate nonlinear
finite-element-based models. Of primary interest is crack pattern, crack opening and closing

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2008; 37:809–823
DOI: 10.1002/eqe
 SEFID-RUD CONCRETE BUTTRESS DAM MODEL 811

and sliding along crack planes. It should be noted that the reservoir and foundation effects are
neglected in this research modeling due to the experimental limitations in appropriate materials
and equipments. As a matter of fact, interpreting of the results should be done with regard to these
discrepancies.

MODEL DEVELOPMENT

Similitude requirements
For the behavior of a small-scale model to accurately represent the corresponding behavior of its
prototype, or full-scale structure, the model must follow certain laws of similitude. These laws,
which are determined by a dimensional analysis of the problem under investigation, are relationship
among the dimensionless ratios formed by corresponding parameters of the prototype and model
structure.
They establish requirements for the materials used to construct the model and the loading used
to excite it. For the case of neglecting reservoir effects, three basic requirements that relate model
and prototype are obtained. Equal gravity force (g) in the model and prototype gives the following
three formulas:


Tr = L r (1)

Sr =
r L r (2)

Ar = 1 (3)

where T, L , S, A and
are time, length, stress, acceleration and mass density, respectively. Subscript
r represents the ratio of these parameters in the prototype and model. The selection of a length
scale L r is typically based on the size of the prototype structure and the capabilities of the testing
facility.
The scale chosen for this model is a 1:30 geometric scale and a concrete-based material with the
same density is used for the material of the dam. According to similitude requirements, estimated
properties of Sefid-rud dam small-scale model are summarized in Table I.

Table I. Similitude requirements and model material properties.

Physical parameter Scale factor Ratio Prototype estimate Target value
Length Lr 30 — —
Density
r 1 2250 kg/m3 2250 kg/m3
Elastic modulus Er =

√r L r 30 20 GPa 0.667 Gpa
Time Tr = L r 5.47 — —
Acceleration Ar = 1 1 — —
Strain εr = 1 1 — —
Stress
f c 16.5 MPa 0.55 MPa
r =
r L r 30
f t 3.20 Mpa 0.107 MPa

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2008; 37:809–823
DOI: 10.1002/eqe
812 A. R. GHAEMMAGHAMI AND M. GHAEMIAN

Material properties
To satisfy the requirements for spatial, physical, boundary and moving condition similarities of the
scaled model with those of prototype, bentonite was used in the cement mixtures as a component
to reduce strength. Considerable work had been accomplished in Reference [5] to produce an
appropriate bentonite-based concrete mix. The use of bentonite in concrete mixes eliminates the
problems in curing and associate shrinkage cracking when using concrete mixes having highly
reduced properties. Ten trial mixes were made in the laboratory with the bentonite saturation
prior to mixing. Standard 30 cm×15 cm cylinders were made from each batch. Compressive and
tensile strengths of each mix were determined to select the best mix that meets the similitude
requirements. Tensile strength was obtained from splitting tensile test. The physical properties of
the best mix are shown in Table II. This mix has the minimum variation from target values for both
tensile strength and modulus of elasticity. It is obvious that not all parameters meet the similitude
requirements simultaneously.
An estimation of nonlinear properties of material, such as specific fracture energy, is a matter
of importance in this research. The specific fracture energy of dam concrete is a basic material
characteristic needed for a rational prediction of concrete dams fracture behavior. According to
the work done by Bazant and Giraudon [9], the fracture energy of concrete, as well as other
fracture parameters, can be approximately predicted from the standard compression strength f c ,
maximum aggregate size da , water–cement ratio w/c and aggregate type (crushed or river). A
very large database, consisting of 238 test series, was extracted from the literature and tabulated.
Optimization of the fits of this data set led to new approximate prediction formula:
 0.46  0.22  
f c da w −0.30
G f = 0 1+ , G f = 17.8% (4)
0.051 11.27 c

where G f is the specific fracture energy obtained in peak load tests such as size-effect method,
G f is the coefficient of variation for specific fracture energy parameter and 0 = 1 for rounded
aggregates, while 0 = 1.44 for crushed or angular aggregates. This formula leads to a value of
3.1 N m/m2 for specific fracture energy of the materials used in this model.
In fact, this equation is valid for normal and low-strength concretes with a maximum aggregate
size less than 50 mm. In this research, bentonite-based concrete is considered as a low-strength
concrete. For dam concrete, the validity of this equation is questionable. Ghaemmaghami and
Ghaemian [10] have conducted a series of tests on specific fracture energy determination of dam
concrete. According to their results, specific fracture energy of prototype concrete can be estimated
between 200 and 250 N m/m2 . Based on the similitude laws, none of the G f and da parameters
complies with the target values. For example, the maximum size of aggregate in this research is

Table II. Model mix physical properties.

Compressive Tensile Modulus of
strength (MPa) strength (MPa) elasticity (GPa)
w/c da (mm) B/C Slump (cm) 28 days 56 days 28 days 56 days 28 days 56 days
3 10 0.25 5.5 0.74 0.83 0.056 0.077 0.490 0.542

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2008; 37:809–823
DOI: 10.1002/eqe
 SEFID-RUD CONCRETE BUTTRESS DAM MODEL 813

limited to 10 mm, whereas its value for Sefid-rud dam is around 150 mm. A scale factor of 1:30 is
required for this parameter to completely simulate its effect on roughness of the crack, the fiction
coefficient and shear strength in the small-scaled model, but the target values for strength and
elasticity properties of the model concrete require a concrete mix with a higher maximum aggregate
size. As a result, a scale factor of 1:10 is used for this parameter in this research. It is obvious
that such distortion in similitude law is a source of error in simulation of the prototype behavior
through a small-scaled model. On the other hand, it is impossible to satisfy all the similitude
requirements at the same time.

Model design, construction and instrumentation

According to the structural drawings of Sefid-rud dam and also considering the actual conditions
of the shaking table, the model scale of 1:30 is adopted in this research. The highest, non-overflow
monolith was chosen because it was the most highly damaged monolith during Manjil earthquake.
The cross section shown in Figure 1 gives the dimensions of the monolith to be modeled.
After preparing bentonite-based concrete, a single model was constructed on a shaking table
in three stages to simulate horizontal construction joints effects as shown in Figure 2. In this
modeling, the joints are considered as cold joints similar to concrete gravity dams. The model
foundation was built from a high-strength concrete reinforced by steel bars to securely fasten the
model to the shaking table and treat the foundation as rigid. Steel bars were attached to a steel plate
that was bolted to the shaking table. Because the dam foundation is mainly composed of rocks
with higher strength and elastic modulus than those of the concrete dam, and also considering that
the base of the gravity dam was built on solid bedrock, the assumption of a rigid foundation is
relatively close to reality.
A detailed plan was developed to specify instrumentation types and locations. These were
selected specifically to ensure that recorded data could be compared directly with numerical

Figure 1. Model dimensions.

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2008; 37:809–823
DOI: 10.1002/eqe
814 A. R. GHAEMMAGHAMI AND M. GHAEMIAN

Figure 2. Layout of the construction joints.

model results. Instrumentation included 16 accelerometers, 17 strain gauges and 11 displace-

ment transducers. Accelerometers were mainly installed at the downstream, upstream and front
face of the model as shown in Figure 3(a). Two accelerometers were located inside the shaking
table to measure actual shaking table dynamic loads in longitudinal and transverse directions
(accelerometer nos. 15 and 16). Strain gauges were mainly concentrated on the top of the
model where the most potential of cracking is expected (Figure 3(c)). Displacement transducers
were placed on a steel frame with a rigid foundation adjacent to the shaking table in such
a way as to measure the total displacements of 11 installation points of the accelerometers
as shown in Figure 3(b). Relative displacements were calculated by subtracting shaking table
displacement.

Dynamic characteristics of the model

Using experimental sine-sweep tests on the model, the fundamental frequencies were obtained.
Vibration frequencies for first five modes of the model are 4.1, 11.6, 17.4, 23.2 and 27.1 Hz,
respectively. Mode no. 4 represents the motions along the stream. In this mode with the fundamental
frequency of 23.2 Hz, damping coefficient was measured around 2.9% of critical.

Test plan on the shaking table

For the scaled model test on the shaking table, Manjil earthquake records in stream and cross-
section directions were used as input. According to the scale factor, each seismic accelerogram
duration is reduced to 1:5.477 of the original duration. Peak accelerations of longitudinal and
transverse components are 0.51g and 0.49g, respectively. A biaxial shake table is used to excite the

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2008; 37:809–823
DOI: 10.1002/eqe
 SEFID-RUD CONCRETE BUTTRESS DAM MODEL 815

Figure 3. (a) Accelerometers, (b) displacement measure points and (c) strain gauges.

model in longitudinal and transverse directions. In actual dam structure, each monolith is laterally
supported by the adjacent monolith. Because of the experimental limitations, in this research
modeling no support was provided laterally. Therefore, to lessen lateral vibration effects, it was
decided to reduce the final transverse acceleration to 10% of the main scaled record, which is
the minimum level of earthquake loading as shown in Figure 4. It should be noted that applying
transverse earthquake component is only for obtaining a more detailed model response to serve as
data in numerical model calibration and has no major effect on the behavior of the model without
lateral support.
Earthquake loading was planned to begin at 10% peak ground acceleration (PGA) of Manjil
earthquake. After testing for each level of earthquake, within a limited time to check the obtained
results, another set of testing was performed for the next level of PGA. This was done to a
level in which damage can occur. The last input record on shake table was used in numerical
model.
Other factors such as hydrodynamic effects, foundation flexibility and pre-seismic cracking can
play a major role in simulating the seismic behavior of the prototype. Because of the experimental

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2008; 37:809–823
DOI: 10.1002/eqe
816 A. R. GHAEMMAGHAMI AND M. GHAEMIAN

Figure 4. Model-scaled base accelerations: (a) longitudinal component and (b) transverse component.

limitations, these factors are neglected in this research model and such experimental errors are
expected in the process of simulating the prototype behavior.

TEST RESULTS

The model displayed developed failure characteristics when the longitudinal peak acceleration
reached a value of 0.85g that is almost 1.67 times greater than the peak acceleration in Manjil
record. It is obvious that much larger inertia forces were required from the shake table experiments
than what occurred in the field. This phenomenon can be a result of neglecting other governing
factors such as reservoir effects and foundation flexibility in the experimental model. The main
structural damage in the model consists of an integrated crack in the upper part of the model. This
crack runs the whole length of the model about 40 and 61 cm below the crest at upstream and
downstream, respectively (Figure 5(a)). The second structural damage was formed at the upstream
face of construction joint no. 2 almost 122 cm below the crest as shown in Figure 5(b).
A comparison between crack patterns in model and Sefid-rud dam shows that the experimental
cracks are in a satisfactory agreement with those occurred after Manji earthquake (Figures 5(c)
and (d)).
The results obtained from shaking table test include the responses of accelerometers, displace-
ments transducers and strain gauges in both linear and nonlinear levels.

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2008; 37:809–823
DOI: 10.1002/eqe
 SEFID-RUD CONCRETE BUTTRESS DAM MODEL 817

Figure 5. Model cracking: (a) crack pattern at the top of the model; (b) crack pattern at the upstream face
of the construction joint no. 2; (c) model cracking locations; and (d) scaled Sefid-rud cracking locations.

In linear behavior, the average amplifications of acceleration along stream and cross-section
directions are factors of 1.66 and 5.25 from the base to the top of the model, respectively. There
is a great difference between magnification factors for stream and cross-section directions. This is
mainly due to the different model stiffness in these directions.
In nonlinear behavior, the average amplifications of acceleration along stream and cross-section
directions are factors of 4.57 and 3.1 from the base to the top of the model, respectively. A
high increase in magnification factor of the longitudinal accelerometers located above the cracks
line is clearly observed. However, this increase is accompanied by a decrease in magnification
factor of the top of the model in cross-section direction. This behavior can be a result of cracking
attitude.
The longitudinal displacement at the top of the model in nonlinear level is shown in Figure 6(a).
Owing to the excessive damage, the test was stopped at t = 4.55 s.
Figure 6(a) displays an abrupt change in the longitudinal displacement which would correspond
with cracking initiation in the model. Also, a permanent displacement about 2 mm is observed in
measuring instrument after stopping the test. This behavior is a result of a sliding towards down-
stream which corresponds with failure characteristics of Sefid-rud dam. Cross-stream displacements
are shown in Figure 6(b). Similar to longitudinal displacement, an abrupt change is clearly observed

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2008; 37:809–823
DOI: 10.1002/eqe
818 A. R. GHAEMMAGHAMI AND M. GHAEMIAN

Figure 6. Crest displacement in nonlinear level: (a) longitudinal and (b) out of plane.

Table III. Maximum dynamic tensile strains.

Strain gauge no. Location Maximum dynamic tensile strain

1 Construction joint no. 3 25 000×10−6

2 Construction joint no. 3 80×10−6
3 Construction joint no. 3 Out of order
4 Top of the transition zone 80×10−6
5 Top of the buttress at slope change zone 1000×10−6
6 Top of the buttress Out of order
7 Top of the buttress 290×10−6
8 Top of the buttress 290×10−6
9 Top of the buttress 110×10−6
10 Top of the buttress 90×10−6
11 Top of the buttress 130×10−6
12 Top of the buttress 225×10−6
13 Downstream of construction joint no. 2 75×10−6
14 Downstream of construction joint no. 1 90×10−6
15 Slope change above the heel Out of order
16 Slope change above the heel 100×10−6
17 Slope change above the heel 90×10−6

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2008; 37:809–823
DOI: 10.1002/eqe
 SEFID-RUD CONCRETE BUTTRESS DAM MODEL 819

in displacement pattern almost after t = 1.5 s which corresponds with crack development at the top
of the model.
The maximum tensile dynamic strains at various locations of model are shown in Table III.
Tensile strain threshold for cracking is almost 100×10−6 . It can be seen that the maximum dynamic
strain has occurred at construction no. 3 where the crack line has passed through. This large strain
with a value of 25 000×10−6 indicates that a complete opening has occurred in cracked zone, as
shown in Figure 5(a).
The strains at the top of the buttress are relatively large and show a wedge-shaped cracked zone
as that had occurred in reality. This is mainly due to the stress concentration at the slope. At slope
discontinuity above the model heel, despite the large strains, no cracked zone is observed. This
can be a result of material consolidation at the lower parts of the model which accompanies with
an increase in tensile stress and modulus of elasticity.

NUMERICAL MODEL CALIBRATION

Smeared crack and damage mechanics constitutive models

Smeared crack and damage mechanics models based on nonlinear fracture mechanics principles
and elasto-brittle damping model are used to represent crack initiation and propagation. The
development of the constitutive models relies on an energy-based softening initiation criterion,
fracture energy conservation during cyclic excitation, shear deformations in the fracture process
zone and the subsequence rotation of the crack planes. The basic material parameters are the
fracture energy G f , the tensile strength f t and the slope of the post peak response, as shown in
Figure 7. The maximum finite-element size, denoted by l0 , that can be modeled with strain-softening
behavior is determined from

2E G f
l0
(5)
2t

The three-dimensional constitutive laws and numerical procedures are presented in References
[11–16] in detail and are beyond the scope of this paper.

Figure 7. Concrete constitutive model.

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2008; 37:809–823
DOI: 10.1002/eqe
820 A. R. GHAEMMAGHAMI AND M. GHAEMIAN

Correlation with numerical analyses

Model finite-element meshing includes 780 20-node isoparametric solid elements as shown in
Figure 8. The modulus of elasticity, Poisson’s ratio, unit weight, the true tensile strength and the
ratio of the apparent to the true tensile strength, the specific fracture energy and the dynamic
magnification factor applied on both of the tensile strength and the specific fracture energy are
0.542 GPa, 0.2, 22072.5 N/m3 , 0.077 MPa, 1.25, 3.1 N m/m2 and 1.30, respectively.
Figures 9(a) and (b) represent the crack patterns in a monolithic model without considering
construction joint effects using damage mechanics and smeared crack models. In these scenarios,
cracks are concentrated at the lower part of the model and no cracking is observed at the top of
the model where the main cracks have occurred. First crack profile initiates from slope change in
downstream and expands into upper layers. In reality, no cracking was observed in this area. This
can be due an increase in strength properties of the material in lower layers of the model as its
effect was observed in strain gauges responses.
In second scenario, effect of construction joint is included in the numerical model and cracking
is prevented in layers below construction joint no. 1 through a slight increase in strength properties.
Figures 9(c) and (d) show the cracking patterns using damage mechanics and smeared crack models
in this case.
It is obvious that the resulted crack profiles including a weak layer upon construction joints are
in a satisfactory agreement with those reported experimentally. However, it is observed that cracks
are localized in downstream face of construction joint no. 2, which is contrary to the experimental
results.

Figure 8. Finite-element model including 780 elements.

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2008; 37:809–823
DOI: 10.1002/eqe
 SEFID-RUD CONCRETE BUTTRESS DAM MODEL 821

The crack profiles resulting from the damage mechanics approach (shown in Figure 9(c)) are
more diffused than those resulted from the proposed smeared crack method at the upper part of
the dam body. The more diffusion in the crack profile leads to excessive damage and therefore
analysis termination. In smeared crack model, the analysis is terminated at the end of the record
without any termination due to instability or excessive error in the energy balance. However, it
does not produce a through crack as shown in Figure 9(d). The reason is in the nature of the
nonlinear models as these differences have been already observed in the previous works of the
second author [15, 16].

Figure 9. Numerical crack patterns: (a) monolithic model without construction joint effects:
damage mechanics; (b) monolithic model without construction joint effects: smeared crack; (c)
heterogeneous model with construction joint effects: damage mechanics; and (d) heterogeneous
model with construction joint effects: smeared crack.

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2008; 37:809–823
DOI: 10.1002/eqe
822 A. R. GHAEMMAGHAMI AND M. GHAEMIAN

CONCLUSIONS

1. An experimental study on the dynamic characteristics and seismic responses of the 1:30
scaled model of the highest monolith of Sefid-rud dam was conducted on shaking table. It
was found that the nonlinear behavior of the model test was in satisfactory agreement with
the prototype behavior during Manjil earthquake.
2. In nonlinear behavior, a high increase in magnification factors of the longitudinal accelerom-
eters located above the cracks line is clearly observed. However, this increase is accompanied
by a decrease in magnification factor of the top of the model in cross-section direction. This
behavior can be attributed to the cracking.
3. Sliding displacements about 2 mm (equal to 6 cm in prototype) due to the longitudinal record
are approximately the same response in prototype.
4. From strain gauges responses, it can be found that an increase in strength properties of lower
layers due to the consolidation has prevented this area form cracking. As a result, the model
should be considered as heterogeneous in material properties.
5. Two approaches based on damage mechanics and smeared crack concept were proposed.
The accuracy and the numerical stability of the proposed models and the developed software
were established using the available experimental results. The crack patterns obtained from
numerical models are in good agreement with those obtained from shaking table tests for
the case of including construction joint effects and rigid foundation. However, smeared crack
model shows more numerical stability than fracture mechanics model.
6. As a result of similarity between model and prototype cracking pattern, the model can be
retrofitted according to prototype retrofitting plan after Manjil earthquake and re-tested on
shaking table to determine the final earthquake capacity.

ACKNOWLEDGEMENTS
The authors are grateful to the reviewers for their very useful comments and suggestions. The support
provided by Iranian ministry of energy for this study is gratefully acknowledged.

REFERENCES
1. Donlon WP, Hall JF. Shaking table study of concrete gravity dam monoliths. Earthquake Engineering and
Structural Dynamics 1991; 20:769–786.
2. Ghobarah A, Ghaemian M. Experimental study of small scale dam models. Journal of Engineering Mechanics
(ASCE) 1998; 124:1241–1248.
3. Mir RZ, Taylor CA. An experimental investigation into earthquake-induced failure of medium to low height
concrete gravity dams. Earthquake Engineering and Structural Dynamics 1995; 24:373–393.
4. Mir RZ, Taylor CA. An experimental investigation into the base sliding response of rigid concrete gravity dams
to dynamic loading. Earthquake Engineering and Structural Dynamics 1996; 25:79–98.
5. Harris DW, Snorteland N, Dolen T, Travers F. Shaking table 2-D models of a concrete gravity dam. Earthquake
Engineering and Structural Dynamics 2000; 29:769–787.
6. Zhou J, Lin G, Zhu T, Jefferson AD, Williams FW. Experimental investigation into seismic fracture of high arch
dams. Journal of Structural Engineering (ASCE) 2000; 126:926–935.
7. Li QS, Li ZN, Li GQ, Meng JF. Experimental and numerical seismic investigation of Three George dam.
Engineering Structures 2004; 27:501–513.
8. Tinawi R, Leger P, Leclerc M, Cipolla G. Seismic safety of gravity dams: from shake table experiments to
numerical analysis. Journal of Structural Engineering (ASCE) 2000; 126:518–529.

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2008; 37:809–823
DOI: 10.1002/eqe
 SEFID-RUD CONCRETE BUTTRESS DAM MODEL 823

9. Bazant ZP, Giraudon EB. Statistical prediction of fracture parameters of concrete and implications for choice of
testing standard. Cement and Concrete Research 2002; 32:529–556.
10. Ghaemmaghami AR, Ghaemian M. Large-scale testing on specific fracture energy determination of dam concrete.
International Journal of Fracture 2006; 141(1):247–254.
11. Bhattacharjee SS, Leger P. Application of NLFM models to predict cracking in concrete gravity dams. Journal
of Structural Engineering (ASCE) 1994; 120(4):1255–1271.
12. de Borst R, Nauta P. Non-orthogonal cracks in a smeared finite element model. Engineering Computations 1985;
2:35–46.
13. Gunn RM. Non-linear design and safety analysis of arch dams using damage mechanics, part 1: formulation.
Hydropowers and Dams 2001; 8(2):67–74.
14. Gunn RM. Non-linear design and safety analysis of arch dams using damage mechanics, part 2: applications.
Hydropowers and Dams 2001; 8(3):72–80.
15. Mirzabozorg H, Ghaemian M. Non-linear behavior of mass concrete in three-dimensional problems using a
smeared crack approach. Earthquake Engineering and Structural Dynamics 2004; 34:247–269.
16. Mirzabozorg H, Ghaemian M, Kianoush MR. Damage mechanics approach in seismic analysis of concrete gravity
dams including dam-reservoir interaction. European Earthquake Engineering 2004; 3:17–24.

Copyright q 2008 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2008; 37:809–823
DOI: 10.1002/eqe