You are on page 1of 26

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)

RESPONSE-BASED DAMAGE ASSESSMENT OF


STRUCTURES

A. GHOBARAH *, H. ABOU-ELFATH, AND ASHRAF BIDDAH


 Department of Civil Engineering, McMaster University, Hamilton, Canada L8S 4L7
 Structural Engineering Department, Ain Shams University, Cairo, Egypt

SUMMARY
The structure’s ability to survive an earthquake may be measured in terms of the expected state of damage of
the structure after the earthquake. Damage may be quantified by using any of several damage indices defined
as functions whose values can be related to particular structural damage states. A number of available
response-based damage indices are discussed and critically evaluated for their applicability in seismic
damage evaluation. A new rational approach for damage assessment is presented which provides a measure
of the physical response characteristics of the structure and is better suited for non-linear structural analysis.
A practical method based on the static pushover analysis is proposed to estimate the expected damage to
structures when subjected to earthquakes of different intensities. Results of the analysis of ductile and
non-ductile reinforced concrete buildings show that the proposed procedure for damage assessment gives
a simple, consistent and rational damage indicator for structures. Copyright  1999 John Wiley & Sons, Ltd.

KEY WORDS: damage; assessment; structures; frames; index; pushover analysis

1. INTRODUCTION

There is renewed interest in damage analysis for application in the seismic assessment and
rehabilitation of existing buildings as well as performance-based engineering approaches. The
idea of describing the state of damage of the structure by one number on a defined scale in the
form of a damage index is attractive because of its simplicity. However, its development is
complex since the index should apply to various structural systems at advanced stages of inelastic
deformation and up to collapse. The damage state of a structure can be defined in several ways: (a)
a binary damage state (failure/no failure); and (b) a discrete valued damage state using qualitative
indicators such as none, minor, reparable, severe and failure. Empirical and theoretical ap-
proaches have been applied to yield various estimates of structural damage.
The empirical damage models are based on statistics of observed structural damage following
earthquakes. Although, these damage observations may be subjective, they provide useful
qualitative information on the overall seismic performance of structural systems. However, the

* Correspondence to: A. Ghobarah, Department of Civil Engineering, McMaster University, Hamilton, Canada L8S
4L7. E-mail: ghobara@mcmaster. ca

CCC 0098—8847/99/010079—26$17)50 Received 28 October 1997


Copyright  1999 John Wiley & Sons, Ltd. Revised 16 April 1998
80 A. GHOBARAH, H. ABOU-ELFATH AND A. BIDDAH

empirical evaluation does not lend itself well to rationally predicting the strength reserve and
response characteristics of a structure with a specified degree of damage because: (a) it completely
disregards the mechanics of materials that undergo large inelastic cyclic deformation; (b) future
earthquakes may have different intensities, duration, and frequency content; (c) buildings in other
locations and recently built structures that are designed to current codes may differ significantly
from the structures used to develop the damage statistics; and (d) the dynamic characteristics of
the population of structures included in the statistical analysis may have altered due to repairs
and damage accumulation from previous earthquakes.
The analytical damage models may involve various degrees of complexity as they account for
the characteristics of the structure and its seismic response. They can be broadly divided into two
classes which are: (a) strength-based damage indices; and (b) response-based damage indices.
Strength-based damage indices are simple and do not require response analysis. However, the
index must be calibrated against observed damage using a large database. If field observations of
damaged structures due to seismic loads are not available, the damage index may be calibrated
based on damage prediction using non-linear dynamic analysis. Strength-based damage indices
were first proposed in 1968 by Shiga et al. and later applied by Yang and Yang. These indices
depend on the geometry of structural elements such as the column and wall area and their general
material properties. A strength-based damage index is proposed by the Japan Building Disaster
Prevention Association Standard for the evaluation of the seismic capacity of existing buildings.
Damage evaluation based on the non-linear response analysis is relatively complex but may
require less data for calibration. Detailed information of structural and material models and
description of ground motion consistent with the site of structure are needed.
The seismic performance of structures is commonly related to the capacity to undergo inelastic
deformations, defined as the ratio of peak inelastic response to the corresponding yield response
or ductility. Experimental studies show that ductility as well as alternative measures of seismic
structural performance based solely on the low-cycle fatigue theory do not seem to provide
a satisfactory index for seismic damage. These test results are consistent with the notion that
failure of brittle systems is caused by excessive deformation while failure of ideal ductile systems is
initiated by repeated inelastic deformations. Damage indices for structures that are neither ideal
brittle nor ideal ductile, should account for the damage effects of both excessive and repeated
inelastic deformations. There is a need for more general and reliable indices to characterize the
performance of structures.
The objective of this study is to review and evaluate some of the available response-based
damage models and to develop a new rational approach for damage evaluation. The proposed
damage assessment is based on the determined response of the structure and the performance
characteristics of its members. The concept is general and applies to both steel and concrete
structures of different lateral load resisting systems. A ductile and non-ductile three-storey
reinforced concrete frame office buildings are used to illustrate the application of the damage
evaluation procedure and to compare the results to a number of response-based indices.

2. RESPONSE-BASED DAMAGE MODELS

The response-based damage indices can be divided into three groups according to what the index
accounts for: (a) maximum deformation; (b) cumulative damage; and( c) maximum deformation
and cumulative damage.

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
RESPONSE-BASED DAMAGE ASSESSMENT OF STRUCTURES 81

2.1. Damage indices based on maximum deformation

2.1.1. Ductility ratio (DR). The ductility ratio is defined as the ratio of the maximum deforma-
tion to the yield deformation. It has been used extensively in seismic analysis to evaluate the
capacity of structures undergoing inelastic deformation and develop inelastic response spectra.
As a damage index, the ductility ratio may be unsatisfactory, especially when shear distortion in
joints and beam bottom bars pullout are anticipated. As demonstrated by experimental studies,
the ductility ratio does not account for the effect of the duration and frequency content of the
ground motion. It is normally assumed that failure occurs when the ductility demand (response)
exceeds the structural ductility (capacity) that is equal to the ratio of the ultimate deformation
under monotonic static load to the yield deformation.

2.1.2. Interstorey drift (ID). The interstorey drift is the maximum relative displacement between
two storeys normalized to the storey height. According to Sozen the percentage of damage to
the structure is given by

% of damage"50 (maximum interstorey drift in percentage)!25 (1)

From the analysis of test data on components and small-scale structures, it was found that an ID
value smaller than 1 per cent corresponds to damage of non-structural components while values
of ID larger than 4 per cent may result in irreparable structural damage or collapse. Collapse is
considered to occur when ID exceeds 6 per cent. Similar to the damage index based on the
ductility ratio, the interstorey drift does not account for effects of cumulative damage due to
repeated inelastic deformation. In addition, the relationship between damage and interstorey drift
varies depending on the maximum deformation at collapse which depends on the ductility class of
the structure.

2.1.3. Slope ratio (SR). The slope ratio is a measure of damage due to stiffness degradation
during seismic loading. It is defined as the ratio of the slope of the loading branch of the
force—displacement diagram to the slope of the unloading branch. From tests on small-scale
structural systems, it has been determined that SR with values of 1)0 and 0)2 correspond to safe
structural behaviour and critically damaged structures, respectively.

2.1.4. Flexural damage ratio (FDR). Roufaiel and Meyer suggested that the ratio of initial
stiffness to the reduced secant stiffness at the maximum displacement can be used as a measure of
damage. Damage indices based on extreme inelastic deformations seem to be strongly correlated
so that their predictions are usually similar. For example, the correlation coefficient between
the two ratios DR and FDR have been found to be 0)95. The FDR index does not account for
effects of cumulative damage caused by repeated load reversals. Critical values of the DR, SR and
FDR damage indices are determined from laboratory tests and field observations. Therefore,
their use in the prediction of seismic damage for structures with characteristics significantly
different from those used in the calibration process requires caution. Additional difficulties in the
use of these damage indices relate to the differences between the characteristics of the expected
earthquake and the earthquakes used in the calibration such as intensity, duration and frequency
content.

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
82 A. GHOBARAH, H. ABOU-ELFATH AND A. BIDDAH

2.1.5. Maximum permanent drift. Permanent drift is closely related to the plastic deformations
in a structural system. Toussi and Yao and Stephens and Yao introduced a qualitative
classification of damage, which among other things included the permanent drift experienced by
the building. They defined four levels of structural damage as: (a) Safe with storey drift not
exceeding 1 per cent of the storey height and no permanent drift; (b) lightly damaged with
permanent displacement of approximately 0)5 per cent of the storey height; (c) damaged with
permanent displacement of 1 per cent of the storey height; and (d) critically damaged with top
storey displacement showing some aperiodicity at the end of the record with poor correlation
between base shear and top level displacement.
The shortcoming of the maximum permanent drift as a measure of damage is that light damage
implies a maximum permanent drift of 0)5 per cent or less. However, a permanent drift of 0)5 per
cent does not necessarily indicate light damage. For example, a damaged non-ductile frame
structure may exhibit no permanent drift, even after severe inelastic deformation.

2.2. Damage indices based on cumulative damage

2.2.1. Normalized cumulative rotation (NCR). A simple measure of structural deterioration


during a seismic event is the sum of all inelastic excursions experienced by the structure. The value
of this measure depends on the duration and intensity of the earthquake. The normalized
cumulative rotation is defined as the ratio of the sum of the inelastic rotations during half cycles
to the yield rotation. Statistical analysis of data on beam-column elements subjected to cyclic
loads shows that damage indices based only on cumulative inelastic deformation or dissipated
energy, may be inadequate to characterize the complex process of damage propagation and
subsequent failure in concrete members.

2.2.2. Low cycle fatigue (LCF). The theory of low-cycle fatigue has been applied to the seismic
analysis of structures subjected to strong ground motion to estimate the state of damage. The
determination of the damage index is somewhat complex and involves the entire response history.
However, the index does not account for the effect of maximum inelastic deformation.

2.3. Damage indices accounting for maximum deformation and cumulative damage

2.3.1. Park and Ang+s local damage index. The index is based on scaled values of ductility and
dissipated energy of the local element during the seismic ground shaking.  The ductility defined
as the ratio of the maximum experienced deformation d , to the yield deformation, is scaled by

the ratio of the ultimate deformation d , under a monotonic static load to the yield deformation.

The dissipated energy is scaled by b /(Q d ) where Q is the yield force and b is a constant
    
determined from experimental calibration. The ultimate deformation is determined when the
concrete reaches a specified ultimate strain. Although the value of the index may exceed unity,
failure of the structure is assumed to occur when the damage index equals 0.8—1.0. Under
monotonically increasing loads, the dissipated energy is zero and the value of the damage index is
d /d so that failure is predicted to occur, as expected when d "d .
 
Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
RESPONSE-BASED DAMAGE ASSESSMENT OF STRUCTURES 83

The assumptions used in the development of the damage index expression are: (a)
the contributions to damage of the extreme deformation and dissipated energy can be superim-
posed linearly, and (b) the related evolution in time of these components can be disregarded. The
results obtained by Banon and Veneziano do not support these assumptions. In addition, the
value of the constant b is not specified and has to be obtained by calibration using laboratory or

field data. The behaviour of this index is strongly dependent on the hysteretic model of the
elements.

2.3.2. Chung, Meyer and Shinozuka+s local damage index. Chung et al. proposed a damage
index which contains damage modifiers that reflect the effect of the loading history. This index
considers the difference in response of members to positive and negative moments. The effect of
the loading history is taken into account by a damage modifier which includes the change in
stiffness and the bending moment sustained up to the calculation cycle. The damage index
definition does not explicitly account for the damage caused by the maximum deformation
experienced by the element.

2.3.3. Maximum softening. DiPasquale and Cakmak developed a damage model based on the
evolution of the natural period of a time-varying linear system equivalent to the actual non-linear
system for a series of non-overlapping time windows. This global damage index depends on
a combined effect of stiffness degradation and plastic deformation. However, to compute the
maximum softening it is necessary to have the input ground acceleration and the acceleration at
another location such as at the top of the structure. The maximum softening, index does not
explicitly account for the dissipated hysteretic energy and strength deterioration and does not
provide information concerning the extent of local damage sustained by the members.

2.3.4. Final softening. DiPasquale and Cakmak used the change in the fundamental period of
the structure as a measure of the change in the stiffness caused by the earthquake. However, the
instantaneous fundamental period includes the effect of the inertia and damping forces. The
advantage of the final softening is that it can be evaluated from the initial natural period and the
final period determined from vibration field testing after the earthquake. In effect, it is not
necessary to know the actual structural response. However, the measured change in period could
be caused by cracking of infill walls while the structural system may remain intact. The
final period is affected by the changes in the fundamental mode of the structure due to
inelastic response. These changes will cause a corresponding change in the modal mass leading
to final softening index that is no longer representative of the global stiffness deterioration.
A shortcoming of damage measurements based on the final softening is that local element and
storey damage as well as the information contained in the response to the earthquake are not
available. A recognized difficulty in the calculation of the final period is due to the idealization
used in the analytical procedure. The period calculation at the final time step of the earthquake
loading may be affected by the randomness of the instantaneous tangent stiffness at the end of the
dynamic load. In the inelastic hysteretic response range of reinforced concrete, the stiffness of the
loading direction may be significantly different from the stiffness in the unloading direction. In
addition, the stiffness at the zero load position may differ from the stiffness at the loaded
positions.

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
84 A. GHOBARAH, H. ABOU-ELFATH AND A. BIDDAH

2.4. Global damage indices

Several of the discussed damage indices such as DR, NCR, LCF, Park and Ang and Chung
et al. represent the local damage sustained by the individual structural elements. The analysis of
local damage indices identifies the weak or vulnerable elements that should be rehabilitated.
However, it is difficult to get a clear idea of a structural system’s response to a given input ground
motion from a long list of element damage indices. Important decisions concerning the residual
strength and safety of a damaged structure are currently based on a single overall or global
damage index. Global indices are required for post-earthquake evaluation of structures, reliabil-
ity studies and applications in the performance-based engineering approach.
The global damage index is obtained from special combinations of local damage measures. The
simplest technique for combining local damage indices is to use a weighing scheme. The weighing
factor can reflect the replacement cost and/or the relative importance of the member or substruc-
ture in maintaining the integrity of the structure. For example, the lower storey of a building
might be assigned more importance than the upper storeys. The weighing factor for any storey
could also depend on the magnitude of the damage index for that storey, so that severely
damaged storeys are weighted more heavily. Due to the integration of detailed damage informa-
tion of an entire structure into a single global estimator, much information is lost thus allowing
only a crude estimate of structural performance during seismic events. Park and Ang presented
an approach to calculate the global damage index by using weight factors for local indices defined
as functions of the hysteretic energy dissipated by each element. Chung et al. used the damage
index from each storey to define the global damage index. The storey damage index is obtained as
a weighted average of the local damage indices of all elements in the storey, with the energy
dissipated in the member as the weighing function.
The use of weighted average procedure to calculate global damage index does not properly
account for the local concentration of damage, does not distinguish between a column and
a beam, and may lead to misleading results. It is possible for a few structural members of the
building to have undergone severe damage without being reflected in the global index. To
illustrate the point, consider two identical portal frames A and B. The first frame A is subjected to
a lateral force causing one plastic hinge on the right side of the beam with local damage index d ,
0
hence, the global damage index is the also d . The second frame B is subjected to a higher lateral
0
force causing two plastic hinges to form at both ends of the beam, one to the right and one to the
left. The local damage indices of the plastic hinges are d , d and the dissipated hysteretic energy
0 *
of each plastic hinge are E and E . The global damage index is (d E #d E )/(E #E ). If (d ,
0 * 0 0 * * 0 * 0
E ) of frame B are slightly larger than (d , E ) of frame A and (d , E ) of frame B are fractions of
0 0 0 * *
(d , E ) of frame B the calculation may yield a global damage index of frame B that is lower than
0 0
that of frame A. This is incorrect since frame A is at an earlier stage of loading and should have
less damage than frame B.

3. PROPOSED DAMAGE MODEL

From a series of earthquake simulation tests of concrete elements, Otani et al. observed that
when loading reversal was repeated at the same newly attained maximum displacement, the
loading stiffness in the second cycle was lower than that in the first cycle, although the resistances

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
RESPONSE-BASED DAMAGE ASSESSMENT OF STRUCTURES 85

at peak displacement were almost identical. Otani and Sozen observed from testing of
multi-storey frames that if a reinforced concrete structure, subjected to a base motion of sufficient
intensity to develop yield, was tested for a second time with a base motion of similar intensity, it
developed essentially the same maximum drift it did in the first test. The stiffness of the structure
at the beginning of the second test was lower than the initial stiffness of the structure. From this
observation, Sozen concluded that the maximum drift response appeared to be a function of the
initial properties of the structure and not related to the stiffness of the structure at the beginning
of the second test. However, when a reinforced concrete member is subjected to cyclic loading at
constant displacement level, the member stiffness will decrease with the loading cycles. In other
words, the damage to the structure after it was tested for the second time is more than the damage
to the structure after the first test. This is the reason why drift alone may not be the best measure
of damage. The stiffness before and after the loading is a more consistent indicator of damage.
This is the concept of the final softening damage index.
To overcome the analytical difficulties in the calculation of the change in period as a measure of
damage, a new approach for determining the change in stiffness of the structure is proposed. The
approach is to perform pushover analysis for the structure twice; once before subjecting the
structure to the earthquake and once after subjecting the structure to the ground motion, as
shown in Figure 1. Before performing the second pushover analysis, the structure is returned to
the unloaded static state. Relating the initial stiffness before and after subjecting the structure to
an earthquake, a new global damage index as well as damage indices for each storey level can be
determined. The damage indices of the storeys are useful in determining the deterioration in
stiffness and the distribution of damage to the various storeys. The stiffness damage index (DI) of
)
the whole frame can be calculated as follows:

(DI) "1!(K /K ) (2)


)    
where K is the initial slope of the base shear-top deflection relationship resulting from the
 
pushover analysis of the frame before subjecting it to the earthquake ground motion and K is
 
the initial slope of the same relationship but after subjecting the frame to the earthquake time

Figure 1. Pushover analysis. Change of stiffness for calculating the damage index of: (a) the whole frame;
(b) a storey

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
86 A. GHOBARAH, H. ABOU-ELFATH AND A. BIDDAH

history. The stiffness damage index of the ith storey is defined as:

(DI)G "1!(KG /KG ) (3)


)    
where KG and KG are the initial slopes of the base shear-storey drift relationships of the ith
   
storey resulting from the pushover analysis of the frame before and after subjecting it to the
earthquake, respectively.
The values of the damage indices range from zero to one depending on the amount of damage
experienced. A value of zero represents no damage while a damage index value of one corres-
ponds to collapse. However, in practical terms, collapse may be defined at a lower damage index
corresponding to a certain percentage loss of stiffness.
An inverted triangular loading representing the code lateral load distribution may be used in
the pushover analyses for structures whose response is dominated by the first mode. The loading
can be modified to represent the contribution of higher modes. The proposed index is based on
relative stiffness before and after the earthquake. For this reason, the index is not sensitive to the
distribution of the applied load in the pushover analysis. The effect of torsion and building
irregularities can be represented by the 3D pushover analysis. Research programs to widen the
applicability of the pushover analysis are underway by several researchers.

4. ADVANTAGES OF THE PROPOSED PUSHOVER DAMAGE INDEX

The advantages of the proposed damage evaluation approach are numerous. Some of these
advantages include: (a) The calculation of the damage index is based on rational response analysis
procedure with little need for calibration. (b) The stiffness is calculated after removing the inertia
and damping force effects and bringing the frame to a static state. (c) The damage can be
estimated at any stage of loading without the need to guess the maximum displacement or
deformation of the structure near collapse. (d) Two different final stiffness can be calculated
depending on the direction of the load in the pushover analysis (either from right to left or from
left to right). The smaller stiffness can be used in calculating the stiffness index; (e) A very
important advantage of the proposed procedure is that it provides information regarding the
element and storey damage and the sequence of element damage and failure. A stiffness damage
index is calculated for each storey level separately as well as for the whole frame which can be
used to identify which storey is contributing to the damage; (f ) The storey and global damage
indices for the whole frame are obtained without the need for an averaging or weighting
procedure to integrate the effect of the frame elements. (g) The proposed index is capable of
modelling damage due to mechanisms other than flexural yielding. In this case, the models used in
the analysis should include all possible failure modes. For example, in case of reinforced concrete
buildings, the model should consider shear deformation and reinforcing bar bond slip. Williams
et al. noted that most of the available damage indices consider flexural yielding only and do not
consider the possibility of shear failure for example. These issues are significant in the damage
analysis of existing non-ductile structures.
On the negative side, the calculation of the proposed damage index involves more analysis and
effort than what is needed to evaluate some other damage models. In addition, the applicability
and reliability of the proposed damage assessment are affected by the limitations of the pushover
analysis technique.

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
RESPONSE-BASED DAMAGE ASSESSMENT OF STRUCTURES 87

5. BUILDING DESIGN

As examples, the damage analysis is performed for two typical low-rise three storey reinforced
concrete frame office buildings. One of the buildings is designed according to the 1963 ACI code
and the other according to the 1995 NBCC code. The current code designed structure and the
existing building designed to earlier code are considered to represent ductile and nonductile
moment resisting frames. The dimensions of the buildings are shown in Figure 2. The design live
load is taken as 2)4 kN/mm which is typical for an office building. The steel reinforcement details
in the existing frame includes: (a) light confinement reinforcement in the columns; (b) the beam
bottom longitudinal reinforcement is embedded 150 mm into the beam-column joint; (c) widely
spaced transverse reinforcement in beams and columns; (d) column lap splices are 20 times the
column longitudinal bar diameter and located just above the floor level; and (e) no transverse
reinforcement in the beam-column joint. Section dimensions for the columns and beams and the
area of reinforcement as a percentage of the cross-sectional area (reinforcement ratio) are shown
in Figure 2. The NBCC code designed building is assumed to be located in the city of Victoria
on Canada’s west Coast. The design base shear for the building is calculated to be equal to 0)147
times the weight of the building. The design loads on the building are further increased by the
effect of the accidental torsion provisions of the code to 0)177 times the weight of the building. The
sectional dimensions of the beams and columns of the current code designed frame and the
reinforcement ratios are shown in Figure 2.

6. THE ELEMENT MODEL

An inelastic single-component element was developed for modelling the beams and columns of
the frame. All essential characteristics of the hysteretic behaviour of reinforced concrete members
including stiffness degradation, pinching and strength deterioration, are explicitly taken into
account. The element was implemented in the general-purpose program DRAIN-2DX. The
hysteretic rules are shown in Figure 3. The element utilizes a non-symmetric bilinear curve in
conjunction with four variable user-specified parameters. The stiffness degradation parameter, c,
and the strength deterioration due to low cycle fatigue parameter, b, are similar to the ones
proposed by Park et al. The pinching effect is introduced in the loops by an input parameter, a,
similar to the parameter proposed by Chung et al.. The main characteristics of the element is in
its ability to model the softening behaviour of reinforced concrete members after reaching
a specified deformation corresponding to ultimate strength. The model is also capable of
representing reinforcement pullout and takes into account inadequate shear capacity of the
column. Details of the model are presented by Ghobarah et al..

7. DAMAGE ANALYSIS

The damage analysis was conducted in three steps: (a) a static nonlinear pushover analysis of the
frame was performed; (b) the frame response to an applied earthquake ground motion was
determined; and( c) a second pushover analysis was performed. The damage index is evaluated for
the whole structure and for each storey using the proposed approach as given by equations (2)
and (3), as well as three selected damage analysis approaches for comparison of the results.

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
88 A. GHOBARAH, H. ABOU-ELFATH AND A. BIDDAH

Figure 2. Dimensions and reinforcement of the three-storey office building

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
RESPONSE-BASED DAMAGE ASSESSMENT OF STRUCTURES 89

Figure 3. Hysteretic modelling of the moment—curvature relationship: (a) stiffness degradation; (b) pinching; (c) strength
deterioration; (d) softening

7.1. Pushover analysis

The first pushover analysis of the interior frame was conducted using an inverted triangular
load to determine the initial stiffness of the frame. The second pushover analysis was conducted
after subjecting the frame to the selected earthquake ground motion to determine the final
stiffness. In the second pushover analysis, the selected stiffness is the smaller of: (a) applying the
load to the frame from left to right; or (b) applying the load to the frame in the opposite direction
from right to left. Examples of the pushover analyses before and after the application of 1940 El
Centro earthquake record scaled to PGA of 0)3g, are shown in Figures 4(a) and 4(b) for the ACI
and NBCC code designed buildings, respectively. The analysis results shown in Figures 4(a) and
(b) do not include strength deterioration in order to enable the comparison between the different
damage indices on equal bases.

7.2. Dynamic analysis

The seismic behaviour of the interior frame was studied using scaled ground motion records.
The selected ground motions are El Centro record of the 1940 Imperial Valley earthquake (S00E

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
90 A. GHOBARAH, H. ABOU-ELFATH AND A. BIDDAH

Figure 4. Pushover analysis of the buildings when subjected to El Centro earthquake scaled to PGA of 0)3 g

component), the 1985 Mexico earthquake (Zihuatenejo, Guerrero Array, S00E component), the
1971 San Fernando earthquake (800 W. First St., L.A., N53W component) and the 1985 Nahanni
earthquake (Iverson-Site 1, N10E). These earthquake records were selected to represent a wide
range of frequency content as evaluated by the A/» ratio (A is the peak ground acceleration PGA
in g and » is the peak ground velocity in m/s). The two extreme cases are the far-field Mexico
earthquake record which is rich in low-frequency content and the near-field Nahanni record
which is rich in high frequencies. The fundamental period of the two frames are almost the same
and equals approximately 1)0 s. The dynamic analysis was carried out using a time step increment
of 0)005 s taking into account the P!* effect. Three of the discussed damage models were
calculated after the dynamic analysis for comparison with the proposed damage model. The three

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
RESPONSE-BASED DAMAGE ASSESSMENT OF STRUCTURES 91

models are: (a) Park and Ang;( b) Roufaiel and Meyer; and (c) DiPasquale and Cakmak final
softening damage model.

7.3. Results of the damage analysis

Figures 5 and 6 show the damage index versus PGA when the existing 1963 ACI and the 1995
NBCC code designed buildings are subjected to various earthquakes scaled to different PGA
levels. The ultimate rotations used in calculating Park and Ang’s index are obtained according to

Figure 5. Damage index variation with PGA for the existing frame

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
92 A. GHOBARAH, H. ABOU-ELFATH AND A. BIDDAH

Figure 5. Continued

the concrete material model given by Park and Paulay. The parameter b used in calculating

Park and Ang’s index is taken 0)1 according to Park et al. for nominal strength deterioration.
From the figures, there is a correlation between the PGA and the proposed damage level. As
expected, the current code designed ductile building shows much lower damage index values than
those of the existing non-ductile building when subjected to the same earthquake at the same level
of PGA.
The final softening index is the closest to the proposed pushover index since they are both
conceptually similar. In most cases, it gives lower values than the proposed index especially at
high PGA levels, as shown in Figure 6. In some cases, the final softening index was found to
decrease with increasing PGA level which is not physically correct. However, this is mainly due to
the analytical difficulties in determining the final period as discussed in Section 2.3. The final
fundamental period used in calculating the final softening index is determined after the last

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
RESPONSE-BASED DAMAGE ASSESSMENT OF STRUCTURES 93

increment of the earthquake loading with no control on the loading direction which may give
different loading or unloading stiffness. The figure shows that there is a degree of randomness in
the calculation of the instantaneous stiffness of the structure at the final step of the earthquake.
In the case of the response of the current code designed frame to the San Fernando earthquake
(Figure 6), the building was more damaged at PGA"0)3g in terms of local damage indices,
storey drift and the proposed damage index as compared to the damage state at PGA"0)2g.
However, the averaging procedure used in calculating Roufaiel and Meyer and Park and Ang
indices yielded global damage indices for the case of PGA"0)3g that is less than or almost equal

Figure 6. Damage index variation with PGA for the current code designed frame

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
94 A. GHOBARAH, H. ABOU-ELFATH AND A. BIDDAH

Figure 6. Continued

to the case of PGA"0)2g. These results confirm the earlier discussion that in some cases the
averaging procedure may give incorrect results. The difference between the results may be
tangible for one specific ground motion, as shown in Figure 6. However, these inconsistent
damage values almost disappear when the response to several earthquakes is averaged.
Figures 7 and 8 show the damage index variation with the maximum interstorey drift for the
existing and current code designed frames, respectively. The maximum interstorey drift occurred
at the first storey. From the figures, some correlation between the proposed damage index and the
maximum interstorey drift is observed. Comparing Figures 7 and 8, the interstorey drift limits are
dependent on the ductility class of the structures. Collapse occurs at maximum interstorey drift of
2)25 per cent for the existing nonductile building (Figure 7) and from 4 to 5 per cent for current
code designed ductile building (Figure 8). Figure 9 shows the proposed pushover damage index
variation with PGA of the El Centro record for each storey and for the whole existing frame. The

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
RESPONSE-BASED DAMAGE ASSESSMENT OF STRUCTURES 95

damage indices for the storeys are useful in determining the distribution of damage to the various
storeys and identifying which storey control the performance of the whole frame. In this
particular case, the first storey damage controlled the performance of the frame. Once a storey
index reaches 1)0, the damage index of the whole structure will be also 1)0.
Figures 10(a) and 11(a) show the correlation between the proposed damage model and Park
and Ang’s model calculated for the existing and current code designed frames when subjected to
the selected earthquakes scaled to various PGA levels. The figures indicate that the value of the
proposed damage index is about 0)66 of the value of the Park and Ang’s index in case of the

Figure 7. Damage index variation with the maximum interstorey drift for the existing frame

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
96 A. GHOBARAH, H. ABOU-ELFATH AND A. BIDDAH

Figure 7. Continued

non-ductile building design and is about 0)85 in case of the current code designed ductile building.
It is also interesting to note that for the existing building, failure of Park and Ang’s index
corresponds to 0)66 of the proposed index. Figures 10(b) and 11(b) show the relationship between
the proposed model and the weighted average of Roufaiel and Meyer’s index for the existing and
current code designed buildings, respectively. The weight factors are defined as functions of the
hysteretic energy dissipated by each element. The correlation with the proposed model is good in
case of the existing nonductile building. The value of the proposed damage index is about 0)8 the
value of the weighted average of Roufaiel and Meyer’s model in case of the existing building and is
1)06 in case of the current code designed ductile building. Figures 10(c) and 11(c) show that the
correlation between the final softening damage index and the proposed index is reasonable. In
general, there is more scatter in the results in case of the ductile frame (Figure 11) than the values

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
RESPONSE-BASED DAMAGE ASSESSMENT OF STRUCTURES 97

Figure 8. Damage index variation with the maximum interstorey drift for the current code designed frame

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
98 A. GHOBARAH, H. ABOU-ELFATH AND A. BIDDAH

Figure 8. Continued

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
RESPONSE-BASED DAMAGE ASSESSMENT OF STRUCTURES 99

Figure 9. Proposed damage index of the whole frame and of the different storeys versus the peak ground acceleration for
the existing frame

in the case of non-ductile frame (Figure 10). The reason for this behaviour is that the accuracy of
the indices in estimating the ultimate behaviour decrease with increased ductility. The compari-
sons shown in Figures 10 and 11 indicate that the consistency and agreement between several
response-based damage indices is good in the case of the non-ductile building. These good
correlations between the indices is remarkable considering that they are arrived at using
completely different definitions. However, these results are for the simple case of three-storey
frames and the correlation may not be as good in other cases of more complex response.
According to the damage states defined by Gunturi and the average relationship between the
proposed index and Park and Ang’s index, damage states for structures can be defined as listed
in Table I. Figures 12 and 13 show the results of comparing the relationship between
the proposed damage model and the maximum interstorey drift for the non-ductile and
ductile frames, respectively. The definition of damage to the three-storey frame for a given
interstorey drift presented by Sozen and the proposed damage states are included. It is observed
that the definition of the percentage of structural damage by Sozen is realistic for existing
(non-ductile) buildings and is overestimating the damage for the current code designed (ductile)
buildings. The existing building is expected to fail at interstorey drift of 2)5%. However, this limit
is lower than the expected interstorey drift at failure for code designed building which is
approximately 4—5%.

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
100 A. GHOBARAH, H. ABOU-ELFATH AND A. BIDDAH

Figure 10. Correlation between the damage indices for the existing frame

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
RESPONSE-BASED DAMAGE ASSESSMENT OF STRUCTURES 101

Figure 11. Correlation between the damage indices for the current code designed frame

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
102 A. GHOBARAH, H. ABOU-ELFATH AND A. BIDDAH

Table I. Proposed damage index for different damage


states

Range of proposed
Damage state damage index

Minor 0)0—0)15
Moderate (reparable) 0)15—0.3
Severe (irreparable) 0)3—0)8
Collapse '0)8

Figure 12. Proposed damage index variation with the maximum interstorey drift for the non-ductile frame

8. CONCLUSIONS

1. The proposed approach to evaluate the damage by performing two pushover analyses,
before and after subjecting the structure to an earthquake results in a simple and rational
damage indicator that is based solely on structural analysis. The damage model applies
equally to ductile and non-ductile structures.
2. For the studied cases, the proposed damage model at the storey level or for the whole
structure is a consistent and robust damage index that avoids the use of weighing functions
and other analytical difficulties. Local members information is available and can be easily
extracted from the pushover analysis.
3. There is good correlation between the response-based damage evaluation procedures that
account for maximum deformation and cumulative damage. This was demonstrated for the
simple case of three-storey ductile and non-ductile reinforced concrete frames.

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
RESPONSE-BASED DAMAGE ASSESSMENT OF STRUCTURES 103

Figure 13. Proposed damage index variation with the maximum interstorey drift for the ductile frame

REFERENCES
1. M. Grigoriu, ‘Damage models for seismic analysis’, ¹echnical Report 87-4, Department of Structural Engineering,
Cornell University, Ithaca, NY, 1987.
2. R. V. Whitman, J. M. Biggs, J. E. Brennan, C. A. Cornell, R. L. De Neufville and E. Vanmarcke, ‘Seismic design
decision analysis’, J. Struct. Div. ASCE 101(ST5), pp. 1067—1084, 1979.
3. T. Shiga, A. Shibata and T. Takahashi, ‘Earthquake damage and wall index of reinforced concrete buildings’, Proc.
¹ohoku District Symp., Architectural Institute of Japan, 1968, pp. 29—32.
4. Y. Yang, and L. Yang, ‘Empirical relationship between damage to multistorey brick buildings and strength of walls
during the Tangshan earthquake’, Proc. 7th ¼orld Conf. on Earthquake Engineering, Vol. 6, Istanbul, 1980,
pp. 501—508.
5. H. Aoyama, ‘A method for the evaluation of the seismic capacity of existing buildings’, Bull. New Zealand Nat’l Soc.
Earthquake Engng 14(3), 17—42, (1981).
6. H. Banon and D. Veneziano, ‘Seismic safety of reinforced members and structures’, Earthquake Engng. Struct. Dyn.
10(2), 179—193 (1982).
7. Y. J. Park and A. H.-S. Ang, ‘Mechanistic seismic damage model for reinforced concrete’, J. Struct. Div. ASCE 111(4),
722—739, (1985).
8. N. M. Newmark and E. Rosenblueth, Fundamentals of Earthquake Engineering, Prentice-Hall, Englewood Cliffs, NJ,
1971.
9. G. Ayala and Y. Xianguo, ‘Analytical evaluation of the structural seismic damage of reinforced concrete frames’, 7th
Canadian Conf. on Earthquake Engineering, Montreal, Canada, 1995, pp. 389—396.
10. M. A. Sozen, ‘Review of Earthquake response of reinforced concrete buildings with a view to drift control’,
State-of-the-Art in Earthquake Engineering, Turkish National Committee on Earthquake Engineering, Istanbul,
Turkey, 1981, pp. 383—418.
11. M. S. L. Roufaiel and C. Meyer, ‘Analysis of damaged concrete frame buildings’, ¹echnical Report No. NSF-CEE-81-
21359-1, Columbia University, New York, NY, 1981.
12. S. Toussi and J. T. P. Yao, ‘Hysteresis identification of existing structures’, J. Engng. Mech. ASCE 109(5), 1189—1203
(1982).
13. J. E. Stephens and J. T. P. Yao, ‘Damage assessment using response measurements’, J. Struct. Engng. ASCE 113(4),
787—801 (1987).
14. J. E. Stephens, ‘A damage function using structural response measurements’, Struct. Safety J. 5, 22—39 (1985).
15. S. K. Kunnath, A. M. Reinhorn and R. F. Lobo, ‘IDARC Version 3.0: a program for the inelastic damage analysis of
reinforced concrete structures’, ¹echnical Report NCEER-92-0022, National Center for Earthquake Engineering
Research, Buffalo, NY, 1992.

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)
104 A. GHOBARAH, H. ABOU-ELFATH AND A. BIDDAH

16. Y. S. Chung, C. Meyer and M. Shinozuka, ‘Seismic damage assessment of reinforced concrete members’, Report
NCEER-87-0022, National Center for Earthquake Engineering Research, State University of New York at Buffalo,
Buffalo, NY, 1987.
17. E. DiPasquale and A. S. Cakmak, ‘Identification of the serviceability limit state and detection of seismic structural
damage’, Report NCEER-88-0022, National Center for Earthquake Engineering Research, State University of New
York at Buffalo, NY, 1988.
18. S. Otani, V. W. Cheung and S. S. Lai, ‘Behaviour and analytical models of reinforced concrete columns under biaxial
earthquake loads’, Proc. 3rd Canadian Conf. on Earthquake Engineering, Montreal, Canada, 1979, pp. 1141—1168.
19. S. Otani and M. A., Sozen, ‘Behaviour of multi-storey reinforced concrete frames during earthquakes’, Structural
Research Series No. 392, Civil Engineering Studies, University of Illinois, Urbana, IL, 1972.
20. M. A. Sozen, ‘Drift-driven for earthquake resistance of reinforced concrete’, Report ºCB/EERC-97/05, the EERC-
CºREe Symposium in Honour of »itelmo Bertero, 1997.
21. V. Kilar and P. Fajfar, ‘Simple push-over analysis of asymmetric buildings’, Earthquake Engng. Struct. Dyn. 26,
233—249 (1997).
22. M. S. Williams, I. Villemure and R. G. Sexsmith, ‘Evaluation of seismic damage indices for concrete elements loaded
in combined shear and flexure’, ACI Struct. J. 94(3), 315—322 (1997).
23. ACI-318, ‘Building code requirements for reinforced concrete’, American Concrete Institute, Detroit, MI, 1963.
24. NBCC, ‘National building code of Canada’, The National Research Council of Canada, Ottawa, Ontario, Canada,
1995.
25. V. Prakash and G. H. Powell, ‘DRAIN-2DX — »ersion 1.02 — ºser Guide’, Department of Civil Engineering,
University of California, Berkeley, CA, 1992.
26. Y. J. Park, A. M. Reinhorn and S. K. Kunnath, ‘IDARC: inelastic damage analysis of frame shear-wall structures’,
¹echnical Report NCEER-87-0008, National Center for Earthquake Engineering Research, State University of New
York at Buffalo, NY, 1987.
27. A. Ghobarah, T. Aziz and H. Abou-Elfath, ‘Softening effects on seismic response of non-ductile concrete frames’, J.
Earthquake Engng. (1999) (to appear).
28. R. Park and T. Paulay, ‘Reinforced Concrete Structures’, Wiley, New York, 1975, 769 pp.
29. S. K. Gunturi, ‘Building specific earthquake damage estimation’, Ph.D. ¹hesis, Department of Civil Engineering,
Stanford University, Stanford, CA, 1992.

Copyright  1999 John Wiley & Sons, Ltd. Earthquake Engng. Struct. Dyn. 28, 79—104 (1999)

You might also like