Professional Documents
Culture Documents
SUMMARY
A new concept to determine state of the damage in concrete gravity dams is introduced. The Pine Flat
concrete gravity dam has been selected for the purpose of the analysis and its structural capacity, assuming
no sliding plane and rigid foundation, has been estimated using the two well-known methods: nonlinear
static pushover (SPO) and incremental dynamic analysis (IDA). With the use of these two methods, perfor-
mance and various limit states of the dam have been determined, and three damage indexes have been
proposed on the basis of the comparison of seismic demands and the dam’s capacity. It is concluded that
the SPO and IDA can be effectively used to develop indexes for seismic performance evaluation and damage
assessment of concrete gravity dams. Copyright © 2012 John Wiley & Sons, Ltd.
KEY WORDS: concrete gravity dams; structural capacity; damage index; nonlinear static pushover;
incremental dynamic analysis; seismic performance
1. INTRODUCTION
Damage index (DI) measures the amount of damage and degradation that is imposed to a structure. It is
a physical value that is well correlated to a critical state in a structure [1]. The idea of describing the
state of damage of the structure by one number on a defined scale in the form of DI is attractive
because of its simplicity [2]. The development of general DIs for application in various structural
systems is more complex than the generation of DIs for a particular structure. Empirical and
theoretical approaches have been applied to obtain various estimates of structural damage. In the
case of concrete dams, because there is limited database of field and experimental observations of
damage, the numerical methods for estimation of DI are preferred. Usually, the DI uses a set of
structural response parameters that can evaluate the amount of damage suffered by the structure.
These parameters can be as deformations, reactions, stiffness, energy dissipations, or combination of
them. A reasonable method for computing the DI is to compare the response parameters demanded
by the earthquake with the structural capacities or ultimate values of the response parameters [3].
The capacities of a structure can be computed from its maximum response values under increasing
applied loads. Two well-known methods for computing the structural capacities are the static
pushover (SPO) analysis and the incremental dynamic analysis (IDA). In an SPO analysis, a
monotonically increasing lateral load is statically applied to the structure, and the response
parameters are monitored until yielding and failure states of the structure are captured [4]. In an
IDA, which is in fact a dynamic pushover analysis, the intensity of ground shaking applied to the
base of the structure is incrementally increased until limit states of the structure are captured and
probable collapse mechanisms are determined [5].
*Correspondence to: Mohsen Ghaemian, Department of Civil Engineering, Sharif University of Technology, Tehran, Iran.
†
E-mail: ghaemian@sharif.edu
Damage indexes are usually normalized so that they have zero value when the structure is in no-
damage state and unit value when failure or total collapse of the structure occurs. The damage state
of a structure can be defined in several ways: a binary damage state (failure/no failure) and a
discrete valued damage state by using qualitative indicators such as none, minor, reparable, severe,
and failure [2]. Because the actual database for earthquake-induced damage of gravity dams is
restricted and inadequate, suggestion of such qualitative indicators of damage for gravity dams is
not easily possible.
The DIs are divided into two general categories: local and global DIs. A local DI is an indicator of
damage for a part of a structure, whereas a global DI gives an estimate of overall damage imposed to
the structure. There are some relations for developing the global DIs from local ones. In addition, the
DIs are separated into cumulative and non-cumulative ones. Those indexes that can consider the
accumulation effect of seismic excitation to structural damages are called cumulative indexes [6].
The cumulative DIs can better capture the damage progress in the structure.
There are some available DIs that can be used for concrete gravity dams. These DIs, which are
generally utilized in damage assessment of buildings, use the maximum deformations [7], dissipated
energy through plastic deformations and damage [3], general structural stiffness [2], or combination
of them [8]. As far as the authors know, there are limited works in the field of DIs for gravity dams.
One of them is ligament ratio, which is the ratio of un-cracked ligament length, during the
earthquake excitation, to the whole crack path obtained from larger maximum ground acceleration
when the crack passes through the dam body [9]. However, these indexes are crude and require
much further investigation. In this research, it is attempted to reasonably extend the concepts of DIs
to the field of concrete gravity dams so that they will be appropriate for damage evaluation and
safety assessment of gravity dams.
2. NUMERICAL MODELING
As it is noted, in this research, the numerical approach has been selected for damage assessment of
gravity dams. For this purpose, the tallest non-overflow section of the Pine Flat gravity dam has
been chosen and assessed using finite element technique. The Pine Flat dam is on the Kings River
in the San Joaquin Valley, California, and forms Pine Flat Lake. It is a 122-m-high concrete gravity
dam. Its primary function is flood control, with irrigation and recreation secondary in importance. Its
reservoir is large, with a capacity of 1.2e + 9 m3 [10]. The tallest monolith of the dam used for the
purpose of the analysis. Figure 1 shows a finite element mesh of the dam and truncated reservoir.
There are two major failure modes in concrete gravity dams: tensile overstressing and sliding along
cracked surfaces in the dam or at the dam–foundation interface or planes of weakness within the
foundation [11]. The sliding and overturning failures are very rare to occur in concrete gravity dams
[12]; therefore, this paper focuses on the overstressing failure mode of the Pine Flat dam, where
(a) (b)
Figure 1. (a) Configuration of the tallest monolith of the Pine Flat dam; (b) finite elements mesh of the dam
and the truncated reservoir.
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
DAMAGE ASSESSMENT OF GRAVITY DAMS
there is no sliding plane within the dam body or at the dam–foundation interface. This assessment is
carried out using the IDA and SPO analysis of the dam section.
The dam and the reservoir are modeled in the 2D plane stress analysis. The depth of the reservoir is
considered 116.31 m for seismic analyses. The length of the reservoir is 5 times the dam height, and the
non-reflective boundary condition has been assigned to its far end to model the radiation of the pressure
waves into infinity. The rigid foundation has been considered for the model in all of analyses. The base
of the dam has been tied to the foundation, and there is no sliding at the dam–foundation interface. The
dam–water interaction and water compressibility are considered in the analyses.
Material nonlinearity has been considered in the static and dynamic analyses. The concrete material
is modeled using damage-plastic model [13], and its features have been listed in Table I.
In this model, the concrete behavior in uniaxial tension is controlled by tension stiffening and tensile
damage (dt) that consider the degradation of the material stiffness due to damage propagation in terms
of cracking normal displacement. The constitutive behavior and input parameters of the model have
been shown in Figure 2. The specific fracture energy that is a common parameter in the fracture
mechanics and is the area under the tension softening branch of the stress-normal cracking
displacement curve is taken to be 400 N/m. The computational model used does not consider sliding
displacements and does not address the post-earthquake condition in a damage state where uplift
pressure may built up in cracks that have lost cohesive characteristics.
The material damping with the mass proportional damping coefficient of 1.64 and the stiffness
proportional damping coefficient of 0.0012 has been considered for the concrete that produces 5%
critical damping in the first and third modes of the coupled dam–reservoir model. The explicit
approach has been used for time integration of the equation of motion. This approach is based upon
the implementation of an explicit integration rule together with the use of diagonal or lumped
element mass matrix. The equations of motion for the dam structure are integrated using an explicit
central difference integration rule.
The explicit procedure integrates through time by using many small time increments. The central
difference operator is conditionally stable, and the stability limit for the operator (with damping) is
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
M. ALEMBAGHERI AND M. GHAEMIAN
given in terms of the highest eigenvalue in the system. A conservative estimate of the stable time
increment is given by the minimum taken over all the elements. In this way, the stability of the
procedure is maintained and the analysis can be progressed even in severe damage conditions.
The obtained results are based on the assumptions that the dam body is homogenous without any
joint at the start of the analysis. Only the horizontal component of the earthquakes was considered
for the purpose of the dynamic analysis. One of the major limitations of the current study is the
uncertainty in the crack profile prediction.
Despite criticism on its performance-estimation abilities, the pushover is a simple method that suggests
useful insight on the expected behavior of a structural model [14]. Corresponding to the concepts of the
nonlinear SPO analysis, after exerting gravitational and hydrostatic loads, a lateral incremental load is
applied on the dam body with same pattern as the imposed seismic lateral loads in earthquakes. In
buildings, this pattern is often dominated by the first mode of vibration of the structure if there is
uniform distribution of mass along the building height; otherwise, the effects of non-uniform
distribution of mass should also be considered [4].
In concrete gravity dams, there are two types of lateral loads that are generated in earthquakes:
inertial loads and hydrodynamic loads. Thus, for the SPO analysis of the gravity dams, these two
loads should be considered for generation of the lateral increasing load distribution. There is clearly
non-uniform distribution of the mass along the height of a gravity dam. Because of the 2D
homogenous analysis of concrete gravity dams, the mass distribution is proportional to the width of
the dam along its height. It is also important to assign reasonable relative weight factors for these
two load distributions. By using only the first vibration mode of the coupled dam–reservoir system,
the lateral load distribution is considered as follows:
in which, b is the dam width, ’1 is the normalized first-mode shape of the dam, rc is the concrete
density, a is the weight factor, H(y) is the hydrodynamic load distribution, F(y) is the inertial load
distribution, P(y) is the total lateral load distribution, and y is the height measured from the base of
the dam. From primary modal analysis of the coupled dam–reservoir system, it is found that the first
vibration mode of the system has an effective mass over 70% of the mass of the dam and the
reservoir. Thus, only consideration of the first mode is roughly adequate. For the hydrodynamic load
distribution, the well-known Westergaard’s added mass distribution [15] can be used as follows:
7 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
H ðyÞ ¼ rw hw ðhw yÞ for y⩽hw (3)
8
in which, rw is the water density and hw is the reservoir height. In the added mass approach, the inertial
and hydrodynamic loads have the same direction in earthquakes. The reservoir height for SPO analysis
is considered at the crest level, and hence, the total hydrodynamic load on the upstream (U/S) face of
the dam is about 8660 tons. Comparison of the results of preliminary linear seismic analyses of the Pine
Flat dam, with and without reservoir, shows that generally, the base shear with the presence of the
reservoir, excluding the hydrostatic load, is about 1.5 to 2 times the base shear without the presence
of the water. Therefore, it was decided that the total inertial load being 2 to 3 times the total
hydrodynamic load. Thus, the a factor in Equation (2) is taken 10, and the total inertial load is about
19 000 tons. Because of incremental increasing of the lateral load, this relative measure seems to be
reasonable. The distribution of the inertial and hydrodynamic loads along the dam height is shown
in Figure 3.
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
DAMAGE ASSESSMENT OF GRAVITY DAMS
120
100
60
40
Inertial Load 20
Hydrodynamic Load
Applied Lateral Pushover Load
0
400 300 200 100 0
Load (ton/m)
The fracture of the dam structure in tension is only the considered dominant potential failure of the
model. The compression damage is not of concern in the dam model, and essentially, the compressive
yielding does not occur in concrete gravity dams. The limit states of the dam model are, therefore, tension
softening initiation (yielding state), cracking initiation, stable crack growth, and breakthrough or unstable
crack growth that are considered as ultimate (collapse) state.
Because of un-biased nature of the concrete gravity dams for loading into the U/S and downstream
(D/S) directions, the incremental lateral loads should be separately applied in both directions. It is also
possible to cyclically apply the lateral increasing load for better investigation of the dam’s behavior.
EI þ EK þ ED ¼ EE (4a)
ES þ EF ¼ EI (4b)
in which EE is the work done by externally applied loads, EI is the internal energy, ED is the viscous
energy dissipated, EK is the kinetic energy, ES is the recoverable elastic strain energy, and EF is the
energy dissipated by inelastic processes and damage. Variation of model’s energies versus total
lateral load in the SPO analysis has been shown in Figure 4(c). Because the SPO analysis is
essentially a static analysis, all of the external work is transformed to the internal energy and almost
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
M. ALEMBAGHERI AND M. GHAEMIAN
70 70
Stable
60 Elastic Cracking 60
40 40
30 30
Softening
20 Initiation 20
Unstable
First Crack
10 Crack 10
Appearance Growth
0 0
0 1 2 3 4 5 6 7 8 0 10 20 30 40 50
Crest Displacement (cm) Base Crack Length (m)
(a) Maximum crest displacement (b) Base crack length
1.2
Elastic Stable Cracking Collapse
Region Region Region
1
EE , EI
0.8
Energy (MJ)
0.6
ES
0.4 Softening
Initiation
Unstable
Crack
0.2 First Crack EF Growth
Appearance
0
0 20 40 60
Total Lateral Load (MN)
(c) Energy variations
all of the internal energy is composed of recoverable strain energy; small fraction of the internal energy
is dissipated through the plastic deformations and damage.
Table II. Model responses at three limit states from SPO analysis into the D/S direction.
Yielding state Cracking initiation Ultimate state
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
DAMAGE ASSESSMENT OF GRAVITY DAMS
50
45
30
25
20
15
10 Unstable
Crack
5 Growth
0
0 1 2 3 4
Crest Displacement (cm)
Because the hydrostatic load is present in opposite direction against the pushover load, higher amount of
load is needed to start cracking compared with that in the previous section.
Table III. Model responses at one limit state from SPO analysis into the U/S direction.
Ultimate state(Cracking Initiation)
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
M. ALEMBAGHERI AND M. GHAEMIAN
3
into the D/S direction
-1
-2
into the U/S direction
-3
stiffness of the dam structure in the two directions is the same, but by increasing the lateral load, the
first crack appears in the D/S direction at the heel of the dam and causes the nonlinear continuation
of the D/S SPO curve and generation of residual displacement after loading removal. After that,
crack growth at the base of the dam participates in the dissipation of input energy and causes stable
rising of the D/S SPO curve. This behavior does not exist in the U/S direction because there is no
stable cracking region. The stable cracking region for monotonic loading into the D/S direction acts
like a fuse and causes the higher ultimate state compared to the U/S direction. But for the cyclic
loading, there is approximately the same ultimate lateral load for both directions. The ultimate
damage state of the dam in the cyclic SPO analysis is shown in Figure 8.
After all, on the basis of the results of the monotonic and cyclic SPO analyses, the yielding and
ultimate response values for the dam’s motion into the D/S and U/S directions have been listed in
Table IV. In this table, u is the crest displacement, and subscripts y and u indicate the yielding and
ultimate states, respectively.
Incremental dynamic analysis is a parametric analysis that estimates more thoroughly structural
performance under seismic loads. It involves subjecting a structural model to some ground motion
records, each scaled to multiple levels of intensity, thus producing some curves of response
parameterized versus intensity level [5].
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
DAMAGE ASSESSMENT OF GRAVITY DAMS
For scaling ground motions, a non-negative scale factor (SF) is used that is multiplicatively applied
to the un-scaled (natural) acceleration time histories. A value of SF = 1 indicates the natural record,
SF < 1 is a scaled-down record, and SF > 1 is a scaled-up record. Plotting an IDA curve requires
monotonic scalable ground motion intensity measure (IM) and structural state variable (damage
measure (DM) or engineering demand parameter). An IM is a non-negative scalar that characterizes
the intensity of a ground motion record and is monotonically increasing with the SF. Common
examples of scalable IMs are the peak ground acceleration (PGA), peak ground velocity (PGV), and
the 5% damped first-mode spectral acceleration Sa(T1,5%). A DM is a non-negative observable
scalar that characterizes the state of the structural model to a prescribed seismic loading that can be
deduced from the output of the corresponding nonlinear dynamic analysis [5]. Possible choices for
concrete gravity dams could be maximum crest displacement, or energy dissipations through plastic
deformations and damage, to assess limit states or modes of failure of such structures in a
performance-based earthquake engineering assessment. As it is explained in the previous section, the
structural response of gravity dams is a signed scalar that depends on the motion of the dam into the
D/S or U/S direction, so either the absolute values are used or the magnitudes of the positive (D/S)
and negative (U/S) parts are separately considered [5].
To perform IDA of the Pine Flat gravity dam, a set of 12 ground motions are used (Table V) to
roughly provide accurate estimates of the complete range of the model’s responses, from elastic to
yielding, then to nonlinear inelastic and finally to collapse state. These records belong to a group of
relatively large magnitudes of 6.5 to 7.3 and various distances of 8.2 to 77 km, all recommended by
FEMA440 [16], and extracted from PEER strong motion database and recorded on stiff soil (class C
of FEMA440; type B or C of USGS) [17]. The pseudo acceleration spectra of the selected records
that have been scaled so that have spectral acceleration of 0.5 g at period of 0.5 s have been shown
in Figure 9.
Table IV. Summary of the dam model responses from SPO analyses.
D/S direction (+) [monotonic] U/S direction () [cyclic]
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
Table V. Ground motions selected for IDA of the Pine Flat dam.
No. Earthquake name Record name Component (deg) Moment magnitude PGA (g) PGV (cm/s) R (km) USGS soil type Sa(5%) at T1 = 0.349 s (g)
10.0
IVPTS
LPDMH
SFPAS
IVELC
SFPPP
LADSP
LPAND
LPGIL
LPLOB
Sa (g)
LPSTG
1.0 MHG06
NRORR
0.1
0.1 1.0
Period (sec)
Each of the records has been scaled to multiple levels of Sa(T1,5%) from zero to 1 g that have been
arranged in 0.1 g (sometimes 0.05 g) steps. From the primary modal analysis of the coupled Pine Flat
dam–reservoir model, it is found that the period of the first mode of vibration of the system is
T1 = 0.349 s. After extraction of DMs, a set of discrete points are left for each record that resides in
the IM–DM plane and lies on its IDA curve. By interpolating them, the entire IDA curve can be
approximated without performing additional analyses. To do so, the spline interpolation is utilized.
There are some methods for summarization of an IDA curve set that is a collection of IDA curves of
the same structural model under different accelerograms, which are all parameterized on the same IM
and DM. One of these methods, which is a non-parametric method, is using of scatter-plot smoothers
such as the running mean, running median, or the smoothing spline [18]. In the running mean with a
zero-length window (or cross-sectional mean), values of the DM at each level of IM are obtained,
and then, the average and standard deviation (s) of them are calculated. Then, the mean, mean plus
one s (84% fractile), and mean minus one s (16% fractile) representative curves can be plotted. In
the cross-sectional median, the median of DM values at each level of IM are calculated [5]. By
using these methods, an IDA curve set is summarized into one representative (characteristic) curve.
In this section, tension overstressing and fracture failure of the Pine Flat dam is investigated using
IDA. The nonlinear structural model of the dam after exerting the weight and hydrostatic loads is
analyzed under the set of earthquake ground motions listed in Table V, in multiple levels of
intensity; then on the basis of the obtained results, the capacity, performance, limit states, and failure
modes of the dam are recognized.
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
M. ALEMBAGHERI AND M. GHAEMIAN
0.8
1
0.7
0.8 0.6
Sa (T1,5%) (g)
0.5
PGA (g)
0.6
0.4
0.4 0.3
Median 0.2
0.2 Mean
84% fractile
16% fractile 0.1
0 0
0 2 4 6 8 0 2 4 6 8
Maximum D/S Crest Displacement (cm) Maximum D/S Crest Displacement (cm)
(a) Sa (T1,5%) as IM (b) PGA as IM
70
60
50
PGV (cm/s)
40
30
20
10
0
0 2 4 6 8
Maximum D/S Crest Displacement (cm)
(c) PGV as IM
The median and mean IDA curves show good agreement (Figure 10(a)). It is also seen that selection
of the Sa(T1,5%) as IM, instead of PGA or PGV, produces a lower dispersion over the full range of D/S
crest displacements. This holds for other DMs. Because the IDA can be used for estimation of collapse
capacity of structures, the Sa(T1,5%) seems preferable to PGA and PGV for this structure as the
dispersion of IDA curves is less at high values of IM. Thus, the remaining IDA curves are plotted
only for Sa(T1,5%) as IM. These curves have been shown in Figures 11 and 12.
All IDA curves have been plotted until Sa(T1,5%) = 1.0 g, which is associated with scaling factors of
0.74 to 5.18 for different earthquakes. The first-mode spectral acceleration of 1.0 g generates extensive
damages to the dam body in all of analyses that is indeed unrealistic. This unrealistic damage is due to
the ‘diffused cracking pattern’ that is imposed to the dam section in high-intensity shaking, and most
probably, if sliding is allowed, the top block will separate from the rest of the dam body along a single
major crack. Generation of these responses is because of the model properties especially that the
sliding planes along cracked surfaces are not considered; however, they play a significant role in the
behavior of the cracked dam.
From Figure 11(b), the D/S curve is generally hardened, whereas the U/S one is softened. There is the
same initial stiffness for both motions as observed from the SPO analyses. It is worth noting that in almost
all records, at high IM values, there is permanent deformation for the dam crest into the U/S direction.
Because each increment of damage energy dissipation occurs in the peaks of the crest displacement
and because positive (D/S) and negative (U/S) peaks are associated with the heel and neck cracking,
respectively, the energy dissipation is assessed either in two opposite directions or totally (Figure 12).
Until the mean first-mode spectral acceleration of about 0.18 g, there is no damage in both directions
where the first crack is appeared at the heel of the dam. The first crack at the neck of the dam, which is
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
DAMAGE ASSESSMENT OF GRAVITY DAMS
1.2
1
1
0.8
Sa (T1,5%) (g)
0.8
Sa (T1,5%) (g) 0.6
0.6
0.4
0.4 Median
Mean D/S Crest Displacement
84% fractile U/S Crest Displacement
0.2 16% fractile 0.2
0 0
0 4 8 12 0 2 4 6 8 10
Maximum U/S Crest Displacement (cm) Maximum Crest Displacement (cm)
(a) (b)
Figure 11. IDA curves: (a) maximum U/S crest displacement, (b) mean IDA curves.
caused by the dam motion into the U/S direction, appears at various Sa(T1,5%) among the records,
varies from 0.4 g to 0.6 g. The growth of the intensity causes increasing in the rate of the damage
energy dissipation, which shows vastness of damage that is imposed to the model. The energy
dissipation IDA curves show no tangible signs of hardening that is reasonable specifically for the
motion of the dam into the U/S direction, which generates essentially unstable cracking at the neck
of the dam. However, there is minor hardening in the IDA curves of the D/S direction that is
because of rather fracturing at the neck of the dam that dissipates the input energy.
1 1
0.8 0.8
Sa (T1,5%) (g)
Sa (T1,5%) (g)
0.6 0.6
0 0
0 100 200 300 400 500 0 100 200 300 400 500 600 700
Maximum Damage Dissipation in D/S Direction (kJ) Maximum Damage Dissipation in U/S Direction (kJ)
(a) Dam motion into the D/S direction (b) Dam motion into the U/S direction
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
M. ALEMBAGHERI AND M. GHAEMIAN
appurtenances can be a possible choice. The DM-based rules have the advantage of simplicity and ease
of implementation, especially for performance levels other than collapse; however, even a unique CDM
value may imply multiple limit-state points on an IDA curve (Figure 10) [5]. Nevertheless, comparison
of the dam responses at various states makes possible the definition of such CDM values. For example,
from IDA analysis, the initiation of cracking at the heel of the dam occurs at the maximum D/S crest
displacements of 1.187 to 1.709 cm with the average of 1.429 cm. This value is comparable with the
SPO result, that is, u+y = 1.367 cm (Table IV), with the difference of 4.5%. Thus CDM = 1.429 cm can
be a good choice for elastic limit state of the dam performance with corresponding IM of Sa
(T1,5%) = 0.18 g (Figure 11(b)).
Initiation of cracking at the neck of the dam (from U/S face) occurs at the maximum U/S crest
displacement of 2.931 to 3.979 cm with the average of 3.378 cm. The corresponding extracted value
from the SPO analysis is u y = 2.654 cm (Table IV), which differs 27% with the IDA result. Because
the dynamic analysis is more realistic, the CDM = 3.378 cm can be a critical value for the limit state
of neck’s cracking initiation with corresponding IM of Sa(T1,5%) = 0.44 g (Figure 11(b)). Until the
CDM = 3.378 cm, there is approximately the same maximum crest displacement into the D/S and U/S
directions; therefore, the two aforementioned CDMs can be used in the absolute maximum crest
displacement IDA curve. The dispersion (standard deviation divided by mean) of the response
results at each level of IM for various IDA curves are shown in Figure 13. The low dispersion of
the results at corresponding selected IMs, that is, 0.18 and 0.44 g, (Figure 13(a)) assures the quality
of the selected CDMs.
The alternative IM-based method is primarily intended for better assessment of collapse capacity
because it generates a single point on the IDA curve that divides it to two regions, non-collapse and
collapse. Definition of a critical value on IM (CIM) that signals collapse for all IDA curves is more
difficult than on DM [5]. For example, the FEMA 20% tangent slope approach [19] is in effect an
IM-based method; the last point on the curve with a tangent slope equal to 20% of initial elastic
slope is assumed to be the capacity point. Such a flattening that is a typical final state in building’s
1 1
0.8 0.8
Sa (T1,5%) (g)
Sa (T1,5%) (g)
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 0.05 0.1 0.15 0.2 0.25 0 0.2 0.4 0.6 0.8 1 1.2
Dispersion Dispersion
(a) Maximum absolute crest displacement (b) Damage energy dissipation in the D/S direction
0.8
Sa (T1,5%) (g)
0.6
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3
Dispersion
(c) Damage energy dissipation in the U/S direction
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
DAMAGE ASSESSMENT OF GRAVITY DAMS
IDA curves is not observed in the Pine Flat dam’s IDA curves. However, there is extensive damage
imposed on the dam body at high intensities; the maximum crest displacement IDA curves follow
closely the equal displacement rule, that is, the inelastic section generally approximately follows the
elastic section. In addition, there is an infinite slope in the elastic section of the dissipated energy
IDA curves. Thus, the FEMA method is not applicable here. In the following section, estimation of
the model’s capacity is investigated by comparison of the IDA and SPO results.
The common incremental loading nature of the IDA study and the SPO suggests an investigation of the
connection between their results [5]. For reasonable comparison of the IDA and SPO curves, it is
required to establish one-to-one mapping between the first-mode spectral acceleration in an IDA
curve and the lateral load (base shear) in a SPO curve. This purpose is achieved by dividing the
base shear by the mass of the model and then adjusting it so that the elastic segments of the SPO
curve and the mean IDA curve (when DM is maximum crest displacement) lie on each other. The
obtained adjusting factors are also applied to the other SPO curves. Now, the two curves are
traceable in one plot (Figure 14).
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 2 4 6 0 2 4 6 8
Maximum D/S Crest Displacement (cm) Maximum U/S Crest Displacement (cm)
(a) Maximum D/S crest displacement (b) Maximum U/S crest displacement
0.6
0.4
0.2
0
0 100 200 300
Maximum Dissipated Energy in D/S Direction (kJ)
(c) Damage energy dissipation in the D/S direction
Figure 14. Comparison of some mean IDA and nonlinear monotonic SPO curves.
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
M. ALEMBAGHERI AND M. GHAEMIAN
The IM values of the IDA curves are almost higher than the monotonic SPO curves at the same DM. It
is because of monotonic application of loads in the SPO analyses that causes concentration of damage at
specific regions (heel or neck). But in IDA analyses, the damages are scattered in both regions.
The ultimate value of D/S direction crest displacement in the monotonic SPO analysis is 6.205 cm,
which is not reached in the corresponding IDA curve (Figure 14(a)). From the cyclic SPO analysis, the
ultimate value of U/S direction crest displacement is 4.138 cm (Table IV) that corresponds to Sa
(T1,5%) 0.6 g in the IDA curve (Figure 14(b)). Because the unstable cracking at the dam’s neck
(from the place of slope changes on the U/S face of the dam) is the dominant behavior at the
ultimate state and the average crest displacement in the U/S direction is higher than the D/S for
Sa = 0.6 to 1.0 g, it can be said that CIM = 0.6 g can be used as ultimate state of the dam performance.
It is worth noting that the damage energy dissipations of the ultimate state of the SPO analyses, that
is, (EF)+u = 103339.9 J and (EF) u = 20276.2 J (Table IV), are comparable with the corresponding
values from mean IDA curves at CIM = 0.6 g, that is, 96481.7 and 23772.5 J, respectively
(Figure 12). In addition, the selected CIM has low dispersion from Figure 13.
The recognized limit states of the dam based on the explanations of this and previous sections have
been shown in Figure 15.
The IDA curves and correspondingly the limit-state capacities display large record-to-record
variability as evident in the previous sections. This observed dispersion is closely connected to the
selected IM; for example, it is proven that PGA is deficient relative to Sa(T1,5%) in expressing limit-
state capacities as it increases their dispersion (Figure 10) [20]. Using spectral accelerations at other
periods, instead of the first-mode period, has shown that it often does not decrease the dispersion
especially for first-mode-dominated short-period structures such as gravity dams [21]. This issue is
shown for the dispersion of IDA curves of the Pine Flat dam in Figure 16, for three limit states of
the dam model. The lowest dispersion belongs to the first mode with the same value for different
limit states. The second and third vibration periods (T2 and T3) approximately lie on the valleys of
1 1
Elastic Stable
0.8 Region Heel 0.8
Cracking
Median IDA Curve
Sa (T1,5%) (g)
Sa (T1,5%) (g)
0.8
Sa (T1,5%) (g)
0.6
Collapse
Region
0.4
Median IDA Curve
Mean IDA Curve
0.2
0
0 100 200 300 400
Maximum Dissipated Energy in U/S Direction (kJ)
(c)
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
DAMAGE ASSESSMENT OF GRAVITY DAMS
0.9
0.6
Dispersion
0.5
0.4
0.3
0.2
0.1
T3 T2 T1
0
0.01 0.1 1 10
Period, (s)
Figure 16. Dispersion of IDA curves for various limit states versus the period.
the dispersion curves. The dispersion curves varies with the same pattern and show that Sa(T1,5%) is
the best choice among other Sa(t,5%) or even combination of them and that the dispersion is mostly
affected by the value of Sa(T1,5%), not by elastic spectral shape.
Based on the results of the SPO and IDA studies, the damage state of the Pine Flat dam under selected
earthquake ground motions at various intensity levels (Table V) is evaluated using following proposed DIs.
a. Based on the maximum deformations, proposed DIs are
DId ¼ max DIdþ ; DId
(
0 if uþmax < uþ
y
uþmax
DId þ ¼ if uþ þ þ
y ⩽u max ⩽ uu
uþu
( 1 if uþmax > uþ
u (5)
0 if umax < u
y
umax
DId ¼ if u
y ⩽u max ⩽ uu
uu
1 if umax > u
u
(
DIu ¼ max DIuþ ; DIu
0 if uþmax < uþ
y
þ þ
u u
DIuþ ¼ max y
if uþ þ þ
y ⩽u max ⩽ uu
uþ þ
u uy
1( if uþmax > uþ
u (6)
0 if umax < u
y
umax u
DIu ¼ u
y
if y ⩽u max ⩽ uu
uu u
y
1 if umax > u
u
in which umax is the maximum crest displacement in an earthquake analysis, uy is the crest
displacement at the yielding state, uu is the crest displacement at the ultimate state, + and
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
M. ALEMBAGHERI AND M. GHAEMIAN
superscripts are indicators of the dam’s motion into the D/S and U/S directions, respectively.
These indexes are global, non-cumulative DIs.
b. Based on the damage energy dissipation, proposed DI is
þ
(DIE ¼ max DIE ; DIE
ðEF Þþmax
if ðEF Þþmax < ðEF Þþ
DIEþ ¼ ðE F Þþ
u
u
in which, the u subscript is indicator of the ultimate state. This index is a global, cumulative DI.
Between the above DIs, the DIE seems more reasonable because it is based on the damage
propagation within the dam body and is a cumulative DI.
The results of nonlinear dynamic analyses of the Pine Flat dam–reservoir system under different
earthquakes at various intensity levels, in IDA analysis, are used for calculation of DI of the dam
model. Before processing the results and calculating the DIs, it is better to investigate the bias factor
of the results. It has been shown that if low to medium SFs are used to scale the records, based on
the Sa(T1,5%) as IM, there is no concern from generation of biased results [22]. The bias factor of a
response result is calculated as
structural response to scaled records
Bias Factor ¼ (8)
mean structural responses to unscaled records naturally at target Sa
If a record, considering a target Sa(T1,5%), has 0.8 ⩽ SF ⩽ 1.2, it means that it is naturally at that
target Sa. This approach is taken because of low number of the selected records. The bias factor of
the results versus their SF for maximum crest displacement and damage dissipation are shown in
Figure 17. It is decided to leave the results with 50% bias factor, that is, maintaining the results with
bias factor of 0.5 to 1.5. Thus, all results of the maximum crest displacement in both directions are
maintained, and many results of damage dissipation energy are kept. The off results (results beyond
the mentioned bias factor range) are thrown away in calculation of DIs.
The dam model’s responses at the yielding and ultimate states have been extracted from Table IV for
D/S and U/S directions. Figure 18 shows the variation of DIs with the Sa(T1,5%). In this figure,
the + and signs indicate the dam’s motion into the D/S and U/S directions, respectively.
Bias, maximum damage dissipation
4
Bias, maximum crest displacement
3.5 3.5
D/S Direction
3 U/S Direction
3 D/S Direction
U/S Direction
2.5 2.5
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0.1 1.0 10.0 0.1 1.0 10.0
Scale Factor Scale Factor
(a) (b)
Figure 17. Bias versus scale factor: (a) maximum crest displacement, (b) maximum damage dissipation.
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
DAMAGE ASSESSMENT OF GRAVITY DAMS
1.1 1.1
1 1
DIδ + DIu +
0.9 DIδ -
0.9 DIu -
0.8 0.8
Sa (T1,5%) (g)
Sa (T1,5%) (g)
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
DIδ DIu
(a) (b)
1.2
1 DIE+
DIE-
Sa (T1,5%) (g)
0.8
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
DIE
(c)
The damage of the dam structure in its motion into the U/S direction (neck cracking) is mostly
dominant at high values of Sa(T1,5%), but in lower intensities, the damage of the D/S motion (heel
cracking) is dominant. Lowest scatter belongs to the DId values. For DIE, at low and high Sa(T1,5%)
values, there is lower dispersion compared with the medium Sa(T1,5%) values, and generally,
dispersion of DIs (DIs for the crest motion into the U/S direction) are more scattered. For example,
at Sa(T1,5%) = 0.5 g, DI u varies between 0.06 and 0.96, which shows great variability of the dam
state under different records scaled to the same Sa(T1,5%). Crack profile and damage state of the
Pine Flat dam under some selected records along with the DIs are shown in Figure 19.
As it is seen, values of the DIs increase with damage progressing. In some records, there are
uncompleted cracks at the neck of the dam that show non-unit DI in the U/S direction; however, the
force-control monotonic SPO analysis estimates unit DI for these situations. Thus, the cyclic SPO
analysis better captures the behavior of the dam in the U/S direction. In other earthquake analyses, there
are complete or no failure in the U/S direction, that is, breakthrough or no cracking at the neck of the dam.
In the situations where there is no crack at the neck of the dam, DI E always is equal to zero, but the
other two indexes sometimes show non-zero values. Based on the analysis results, the ratio of the crack
length at the base of the dam to the crack length at the ultimate state of the D/S SPO analysis is
comparable to DI+E. This relation shows that the cracking behaviors of the dam in the two directions
are approximately independent of each other. It is also observed that the cracking regions in motion of
the dam into the D/S and U/S directions in the earthquake analyses are the same as the SPO analyses.
It shows that the considered lateral load distribution in the SPO analyses is rational and adequate.
There is difference between the obtained crack profiles and the resulted crack profile of the SPO
analysis (Figures 8 and 19). It is generally because of the distinct nature of the static and dynamic
loadings. However, one vertical crack is observed in Figure 19, and the place of the seismic neck
cracking has been truly estimated from the SPO analysis.
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
M. ALEMBAGHERI AND M. GHAEMIAN
Figure 19. Crack profiles of the dam model under selected six records.
The average of DIs at each level of Sa(T1,5%), after omitting the off results, are plotted in Figure 20
along with the various limit states of the dam performance. This curve can be effectively used in
performance estimation and evaluation of the Pine Flat concrete gravity dam.
It is observed that all of DIs have unit value for Sa(T1,5%) more than 0.65 g that approximately
corresponds to the value of CIM that was used for ultimate limit state. DIE and DIu have
approximately equal value for various levels of Sa(T1,5%). Their rate of changes increases from heel
cracking initiation limit state, but partially decreases by initiation of neck cracking.
The values of DIE versus DId and DIu have been plotted in Figure 21 along with their quadratic trend
lines. This figure contains off results as well.
1.1
1
DI
0.9 DIu
DIE
0.8 Collapse Region
Sa (T1,5%) (g)
0.7
0.6
0.5 Neck Cracking Initiation
0.4
0.3 Heel's Stable
Cracking Region
0.2
0.1 Elastic Region
Heel Cracking Initiation
0
0 0.2 0.4 0.6 0.8 1
Damage Index
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
DAMAGE ASSESSMENT OF GRAVITY DAMS
0.8
0.6
DIE
0.4
0.2 DI
DIu
0
0 0.2 0.4 0.6 0.8 1
DI , DIu
The two trend lines finally approach to 1, which shows that the three proposed DIs equally estimate
the ultimate limit state of the dam model. The trend line of DIE DIu approximately lies near the Y = X
line that shows these DIs estimate roughly the same damage for the dam model in various levels of
shaking intensity. Therefore, it can be said that DIE and DIu along with the results of the SPO and
IDA can be efficiently used for damage assessment of concrete gravity dams.
7. CONCLUDING REMARKS
By extending the concepts of the SPO, IDA, and DI into the field of concrete gravity dams, structural
performance, capacity, and limit states of such structures were investigated using the tallest
non-overflow 2D section of the Pine Flat gravity dam. The dam–water interaction was considered,
but the dam–foundation was not considered. There is no sliding plane in the model. The damage-
plastic model was considered for the nonlinear concrete behavior.
For SPO analysis a lateral load distribution based on the earthquake-induced lateral loads on the
gravity dams, was developed and incrementally, statically, monotonically, and cyclically applied to
the dam. It was observed that the dam behaves differently in the U/S and D/S directions. These
behaviors are roughly independent. The dam’s motion into the D/S direction contains a stable
cracking region at the base of the dam that is not observed in the U/S motion of the dam. The
cracking of the dam in its motion into the U/S direction is essentially unstable and occurs at the
neck of the dam. It was shown that the cyclic application of the lateral load can better capture
the behavior of the dam in the U/S direction.
For IDA of the dam model a set of 12 records with various un-scaled intensities was used. It was
observed that the spectral acceleration at the first-mode period of the coupled dam–reservoir is the
best choice for IM among the PGA, PGV, and spectral acceleration at other periods. Two limit
states (heel and neck cracking initiation) were determined based on the DM, and collapse limit state
was determined base on the IM. Investigating the dispersion of the IDA results showed the quality
of the defined limit states.
It was shown that the SPO and IDA studies can be used in performance evaluation and capacity
estimation of concrete gravity dams. The results of these methods well correlate with each other
especially for determination of the limit states of the dam model. Simple DIs based on the maximum
crest displacement or energy dissipation through plastic deformations and damage were proposed
that effectively estimate the damage level of the dam. The DIs increase reasonably with damage
propagation. Qualitative indicators of the damage, such as reparable or severe, were not introduced
in this research, but it can be said that the DI of unit is an indicator of collapse.
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe
M. ALEMBAGHERI AND M. GHAEMIAN
REFERENCES
1. Mita A, Takahira S. Damage index sensor for smart structures. Structural Engineering and Mechanics 2004;
17(3–4):1–23.
2. Ghobarah A, Abou-Elfath H, Biddah A. Response-based damage assessment of structures. Earthquake Engineering
and Structural Dynamics 1999; 28:79–104.
3. Bozorgnia Y, Bertero VV. Improved shaking and damage parameters for post-earthquake applications. Proceedings
of the SMIP01 Seminar on Utilization of Strong-Motion Data, Los Angeles, California, 2001.
4. Chopra AK, Goel RK. A modal pushover analysis procedure for estimating seismic demands for buildings. Earthquake
Engineering and Structural Dynamics 2002; 31:561–582.
5. Vamvatsikos D, Cornel CA. Incremental dynamic analysis. Earthquake Engineering and Structural Dynamics 2002;
31:491–514.
6. Arjomandi K, Estekanchi H, Vafai A. Correlation between structural performance levels and damage indexes in steel
frames subjected to earthquakes. Scientia Iranica 2009; 16:147–155.
7. Powell GH, Allahabadi R. Seismic damage prediction by deterministic methods: concepts and procedures. Earthquake
Engineering and Structural Dynamics 1988; 16:719–734.
8. Park YJ, Ang AHS. Seismic damage analysis of RC buildings. Journal of Structural Engineering ASCE 1985;
111:740–757.
9. Horri H, Chen SC. Computational fracture analysis of concrete gravity dams by crack-embedded elements—toward
an engineering evaluation of seismic safety. Engineering Fracture Mechanics 2003; 70:1029–1045.
10. http://en.wikipedia.org/wiki/Pine_Flat_Dam [17 August 2001]
11. Ghanaat Y. Failure modes approach to safety evaluation of dams. Proceedings of 13th World Conference on Earthquake
Engineering, Vancouver, Canada, 2004.
12. Fishman YA. Stability of concrete retaining structures and their interface with rock foundations. International Journal of
Rock Mechanics & Mining Sciences 2009; 46:957–966.
13. Lee J, Fenves GL. A plastic-damage concrete model for earthquake analysis of dams. Earthquake Engineering and
Structural Dynamics 1998; 27:937–956.
14. Vamvatsikos D, Fragiadakis M. Incremental dynamic analysis for estimating seismic performance sensitivity and
uncertainty. Earthquake Engineering and Structural Dynamics 2010; 39:141–163.
15. Westergaard HM. Water pressure on dams during earthquakes. Transaction ASCE 1933; 98:418–472.
16. FEMA440. Improvement of Nonlinear Static Seismic Analysis Procedures. Federal Emergency Management
Agency: Washington, DC, 2005.
17. Vamvatsikos D, Cornell CA. Applied incremental dynamic analysis. Earthquake Spectra 2004; 20:523–553.
18. Hastie TJ, Tibshirani RJ. Generalized Additive Models. Chapman & Hall: New York, 1990.
19. FEMA. Recommended seismic design criteria for new steel moment-frame buildings. Report No. FEMA-350. SAC
Joint Venture, Federal Emergency Management Agency: Washington, DC, 2000.
20. Vamvatsikos D. Seismic performance, capacity and reliability of structures as seen through incremental dynamic
analysis. PhD Thesis, Stanford University, USA, 2002.
21. Shome N, Cornell CA. Probabilistic seismic demand analysis of nonlinear structures. Report No. RMS-35. RMS
Program, Stanford University, 2004.
22. Luco N, Bazzarru P. Does amplitude scaling of ground motion records result in biased nonlinear structural drift
responses? Earthquake Engineering and Structural Dynamics 2007; 36:1813–1835.
Copyright © 2012 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. (2012)
DOI: 10.1002/eqe