Mohebi Et Al. - A New Damage Index For Steel MRFs Based On Incremental Dynamic Analysis
Mohebi Et Al. - A New Damage Index For Steel MRFs Based On Incremental Dynamic Analysis
Mohebi Et Al. - A New Damage Index For Steel MRFs Based On Incremental Dynamic Analysis
a r t i c l e i n f o a b s t r a c t
Article history: The present study introduced a damage index based on a relationship between two values: the maximum
Received 7 August 2018 interstory drift ratio and spectral acceleration at the first-mode period of the structure. The proposed damage
Received in revised form 17 January 2019 index was defined so as it equals 0 in the yield point of the structural system, while it equals 1 in the collapse
Accepted 6 February 2019
point. The incremental dynamic analysis was used to compute the constituent parameters of the damage
Available online 15 February 2019
index in both yield and collapse points. It was also used to observe the change trend of the index through varying
Keywords:
intensities of the earthquake ground motion records. A number of steel models, including single degree of free-
Damage index dom models, stick models and moment resisting frames served as different examples to exhibit the performance
Incremental dynamic analysis of the damage index as well as to compare it with a number of recognized damage indices. Non-linear models
Twisting pattern were developed in OpenSees. Stiffness and strength deterioration in these models were included using the
Sharp softening pattern concentrated plasticity model. The results indicated that unlike the comparative damage indices, the proposed
Steel moment resisting frame damage index was not influenced by successive hardening and instantaneous softening. In addition, once the
intensity of ground motion records increased, the index maintained its ascending trend. Besides, increasing the
period as well as the number of floors along with the presence of some levels of inconsistency in structures
caused a decline in the correlation between the proposed damage index and the comparative indices.
© 2019 Published by Elsevier Ltd.
https://doi.org/10.1016/j.jcsr.2019.02.005
0143-974X/© 2019 Published by Elsevier Ltd.
138 B. Mohebi et al. / Journal of Constructional Steel Research 156 (2019) 137–154
provided by that DI. Some researchers have proposed GDIs Researchers Research summary
based on the pushover analysis [29,30]. This method of analysis H.R. Ahmadi et al. [44] A new algorithm called EFTFD was proposed to detect
is built upon incorrect assumptions that do not take into account damage in bridge piers. The algorithm was defined
the duration and cyclic effects. Hence, it cannot be an appropriate using a square time-frequency distribution. In
method to study the damage under seismic and dynamic loads. addition, two new DIs were proposed based on
time-frequency matrices.
The non-linear dynamic analysis under earthquake effects can
M. Zhang and R. Schmidt A new approach was proposed to detect the damage
provide more accurate EDP values in DI equations. By extending [45] based on the auto correlation function. The relative
the results of the non-linear dynamic analysis at various intensity change of a normalized vector called Auto Correlation
levels, conducted by the Incremental Dynamic Analysis (IDA) Function at Maximum Point Value Vector (AMV) was
[31], the overall behavior of the structure can be studied compre- used as the DI to locate the damage before and after
damage.
hensively from the elastic state to the ultimate collapse [32]. As A. Eraky et al. [46] Dynamic computer simulation techniques were used
an inherent flaw, these DIs cannot detect the location and extent to develop non-destructive damage detection
of damage in the structure effectively, while calculating the DIs methodology for beam and plate structures. The
for the whole structure leads to more reliable results than those proposed procedure was built upon the comparison of
modal strain energy for different structure conditions
obtained through weighting. As a result, they may not provide
from which the DI can be calculated.
an accurate estimation of how the imposed damage affects an el- J.P. Amezquita-Sanchez A new methodology was presented for detecting,
ement and how it is distributed across the structure. To estimate and H. Adeli [47] locating, and quantifying the damage severity in a
the deterioration in a structure effectively, using a combination smart high-rise building structure. Fractality
of the structural global response and other damage assessment dimension (FD) was employed to detect features to be
used for damage detection. In addition, a new
tools, such as LDIs is proposed [33].
structural DI was proposed based on the estimated FD
values. Further, the damage was located using the
The current study presents a DI for earthquake-excited structures changes of the estimated FD values.
based on a specific combination of general response and local damage Z.X. Tan et al. [48] A technique was presented for predicting the
early-stage damage for both single and multiple
parameters of the structure. These parameters include the spectral ac-
damage scenarios in a steel beam. This technique uses
celeration at the first-mode period of the structure (Sa (T1)) and the the modal strain energy-based DI β to treat single
maximum interstory drift ratio (MIDR) in the structure, respectively. damage scenarios and Artificial Neural Network (ANN)
Expressly, the proposed DI equation includes the MIDR in the structure incorporating β to quantify damage severities under
during the non-linear dynamic analysis and Sa (T1) as the main param- multiple damage scenarios.
eters. In addition, the IDA analysis is applied to study the change trend
in the proposed DI at different Ground Motion (GM) intensities, and
to specify the parameters of the DI when the structure yields and col-
lapses. The DI is proposed to explore the damage in steel structures,
since the models used in this study consider stiffness and strength dete-
rioration only via backbone curves associated with steel structures. The Table 2
behavior of the proposed DI when examining damage in reinforced con- Applications of DIs in Performance-Based Seismic Design (PBSD).
structural response. Therefore, it is essential to explore the association during the design phase, or in case of post-earthquake reliability as
between seismic intensity parameters and damage criteria [54]. Accord- well as problems with repairability. To serve these purposes, GDIs
ingly, identification of an optimal intensity measure, which sufficiently should be both observable for practical purposes and a non-decreasing
correlates with an appropriate Engineering Demand Parameter (EDP), function of time unless the structure is repaired or strengthened [62].
is of high prominence. Many studies attempted to provide the best A GDI could be obtained from special combinations of local damage
seismic parameters to represent potential structural damage. Table 3 measures. Using a weighting scheme is the simplest technique for com-
indicates some research studies on the issue. bining LDIs. The weighting factor can reflect the replacement cost and/
or the relative importance of the member or substructure in maintain-
3. The proposed DI and the comparative DIs ing the integrity of the structure. For example, the lower story of a
building might be allocated more importance than the upper ones.
To assess the behavior of the proposed DI, a number of commonly The weighting factor for any story could also depend on the magnitude
used DIs were opted to be compared with the proposed DI. These DIs of the corresponding DI, so that severely damaged stories are weighted
(labeled comparative DIs) and their equations are defined in detail in more heavily [28].
Section 3.1. In addition, Section 3.2, which introduces the proposed DI, Park et al. [25] and Bracci et al. [26] have proposed weighted averag-
states the necessities of proposing this index initially, and then the ing methods to integrate the damage of single elements for each indi-
corresponding equation and its calculation procedure for a structural vidual story followed by measuring the global damage for the overall
system are discussed in detail. structure. The method proposed by Park et al. [25] is based on the
value of hysteretic energy dissipated by elements, whereas Bracci
3.1. Quantitatively compared DIs et al. [26] designated the gravity load as the averaging operative.
In the current paper, the Park et al. [25] method was employed as the
In the current study, three DIs were picked to be compared with the averaging method, to achieve the damage level in each story (Eq. (1))
proposed DI quantitatively. These included the DIs of Cosenza et al. [13], and the overall structure (Eq. (3)). In Eq. (2), E denotes the total energy
Kunnath et al. [21] and Ghobarah et al. [28]. The first two are LDIs and dissipated by the element and n is the number of elements of an individ-
that of Ghobarah et al. [28] is global. Generally, a DI is local when it re- ual story. In addition, in Eq. (4), E represents the total energy absorbed
fers to a single point, section, member, or structural part, whereas it is by the story and n is the number of stories. Finally, DIcomponent, DIstory
considered global when it defines the state of the entire structure [60]. and DIoverall are the damage level in a component, a story and the
Local damage at a cross-section of the structure can be measured whole structure respectively.
adequately by the degradation of bending stiffness and moment capac-
ity of that cross-section [19]. The analysis of LDIs depicts weak or vul- X
n
DI story ¼ λi;component :DIi;component ð1Þ
nerable elements that should be retrofitted [61]. However, it is i¼1
challenging to get a clear idea of a structural response to a given input
" #
ground motion from a long list of element DIs [28]. Stiffness and Ei
strength deterioration of the whole structure comprise the overall effect ðλi Þcomponent ¼ n ð2Þ
∑k¼1 Ei component
of local damage at various locations. A GDI can then be defined as a func-
tion of such continuously distributed local damage, characterizing the n n o
X
overall damage state as well as serviceability of the structure. GDIs are DI overall ¼ ðλi Þstory ðDIi Þstory ð3Þ
response quantities, determining the damage state of the structure i¼1
after an earthquake excitation and can be used for decision-making " #
Ei
ðλi Þstory ¼ n ð4Þ
∑k¼1 Ei story
Table 3
Applications of DIs in finding appropriate Intensity Measures.
The DI proposed by Cosenza et al. [13], is presented in Eq. (5):
Researchers Research summary
Y. Lieping et al. [55] The correlations between some existing intensity indices μ−1
DI ¼ ð5Þ
and major seismic responses of structures (as μ u;mon −1
representatives of the degree of structural damage) were
studied based on an analysis of elastoplastic SDOF and
MDOF systems, leading to several recommendations.
where μ is the maximum ductility during the loading history and μu,mon
M. Kumar et al. [56] The impact of frequency content and properties of the is the maximum allowable value of ductility equals to uu,mon/ uy while uu,
structure on global and interstory drift for steel frames mon is the ultimate displacement under monotonic loading and uy is the
designed in accordance with Eurocode 8 provisions was yield displacement. For flexure-resisting components μ, μu,mon, uu,mon
investigated using the incremental dynamic analysis.
and uy are replaced with μθ, μθ,mon, θu,mon and θy, respectively. μθ, is the ro-
V.V. Cao and H.R. 1040 far-fault motions were applied to obtain the
Ronagh [57] maximum interstory drifts and the Park-Ang [20] DIs in a tation ductility during the loading history and μθ,mon is the maximum al-
structure. Then, the correlation coefficient was calculated lowable value of rotation ductility under monotonic loading, while θu,
to attain the degrees of correlation between the damage to mon and θy are the ultimate and the yield rotation respectively.
the structure and the seismic parameters. Taking into account both deformation and hysteretic energy,
K. Kostinakis et al. The correlation between 19 ground motion IMs and the
[58] damage state of four medium-rise 3D RC buildings was
Kunnath et al. [21] proposed a damage model which is a modified
estimated and significant information relative to the form of the Park and Ang [20] DI. Eq. (6) presents the formulation of
efficiency of the commonly used damage indicators as well Kunnath et al. [21] DI:
as the influence of the seismic orientation with regard to
structural axes was provided. θm −θy E
A. Massumi and F. Regression coefficients were computed to acquire DI ¼ þβ h ð6Þ
θu −θy M y θu
Gholami [59] expedient ground motion parameters that well
characterize the potential damage caused by earthquakes.
Analysis was conducted on four RC frames under a set of where θm is the maximum experienced rotation of an element in a
85 ground motion records. The structural damage system subjected to an earthquake, θy is the yield rotation under mono-
considered was interstory drift, roof drift and the Park and tonic loading, θu is the ultimate rotation under monotonic loading, Eh is
Ang [20] index.
the hysteretic energy dissipated by the element, My is the yield moment
140 B. Mohebi et al. / Journal of Constructional Steel Research 156 (2019) 137–154
K final
DI ¼ 1− ð7Þ
K initial
where Kinitial is the initial slope of the base shear-top displacement curve
obtained from the pushover analysis of the structure before subjecting it
to the earthquake excitation while Kfinal is the initial slope of the same
curve after subjecting the structure to the earthquake GM. The value
of this DI is between 0 and 1. A value of 0 indicates the structure is linear
and a value of 1 indicates the structure is in danger of collapse. However, Fig. 2. The sharp softening pattern in an IDA curve.
collapse may occur at lower damage values or within the domain of lin-
ear behavior of the structure because of the construction faults. Fig. 1 illustrates the twisting pattern. The pattern is, in fact, the repeti-
tion of the hardening and softening phenomenon in a succession. The
3.2. The proposed DI cause of the hardening phenomenon is that an increase in the scale fac-
tor of the record makes the small cycles of the structural response so
In this section, the proposed DI is introduced. To calculate the pro- strong at the beginning of the time history that the structure is dam-
posed DI, from among the known EDPs, Sa (T1) and the MIDR in the aged, and as a result, the structural characteristics are changed in the
structure under any excitement applied to the structure are required. next strong cycles [31]. The occurrence of severe hardening phenome-
The IDA analysis has been introduced as an appropriate tool for studying non can prevent or even reverse the accumulation of damage instanta-
the structural behavior, and to calculate the EDPs, which considers the neously, therefore, can locally cause the curve to move toward a smaller
parameters from the elastic state to the yield point, non-linear domain, damage measure. Expressly, under a greater intensity of seismic load-
and finally the dynamic instability [32]. The IDA analysis involves ing, a damaged structural system displays the same or smaller response
performing a number of non-linear time history analyses where the compared to the preceding step. The phenomenon can be observed in
intensity of GM record, selected for the study of damage, is gradually in- segment BD in Fig. 1. In the following, assuming the possibility of creat-
creased until it reaches the global collapse capacity point of the struc- ing a collapse mechanism in the structural model, the final softening
ture. The IDA analysis has been proved as a time-consuming method segment is seen in the curve, indicating the onset of the dynamic insta-
due to high volumes of calculations. However, nowadays, the analysis bility (segment EF in Fig. 1).
can be performed at very high speeds owing to computing advances. The equations related to DIs provided by Kunnath et al. [21] and
The curve derived from the IDA analysis demonstrates an intensity Cosenza et al. [13] indicate that in Fig. 1, in the hardening segment
parameter of the GM (such as the spectral acceleration response, Sa (BD) associated with a decrease in the maximum rotation and displace-
(T1)) against a damage parameter (such as the MIDR) called Intensity ment, the level of these DIs also decreases. It is also expected that
Measure (IM) and Damage Measure (DM), respectively [31]. Therefore, Ghobarah et al. [28] DI, which is related to the initial stiffness of the
the curve obtained from the IDA analysis can have a positive effect on structure after the earthquake, will decrease in the hardening and rever-
the calculation process of the proposed DI. Figs. 1 and 2 represent exam- sal domain of the IDA curve slope. This phenomenon causes these indi-
ples of IDA analysis curves for a Single Degree Of Freedom (SDOF) struc- ces to exhibit fewer values toward the end of the hardening segment
ture. In the following section, the detailed account of the design and compared to the beginning of this segment. However, as the hardening
modeling of this structure and other structures is provided. As men- segment ends and with a slight increase in Sa (T1), the structure reaches
tioned earlier, in the current study, the spectral acceleration response point E and enters the final softening step, or specifically, the dynamic
of the structure, Sa (T1), was selected as IM and the MIDR as DM. instability.
The IDA curves reveal different behaviors of the structures under The sharp softening is another common pattern in IDA curves that
various GM records. Successive hardening and softening (the twisting can occur for a variety of reasons, including the presence of construction
pattern) and instantaneous dynamic instability of the structure (the faults in the structure, formation of a weak story, and the P-Delta effect.
sharp softening pattern) in structures are among these behaviors. The sharp softening pattern is seen in Fig. 2. As the figure depicts, the
structure suffers from severe softening after passing point G, and a
sharp increase in the MIDR occurs with a small increase in the value of
Sa (T1). If the above indices are used for seismic assessment of this pat-
tern, they are expected to provide small values at lower GM intensities,
be accompanied by a sudden increase at higher intensities of GM, and
reach their maximum value. Consequently, when these DIs are used in
the seismic assessment of structures and the current status of the struc-
ture under the intended GM record is followed by the presence of a
twisting pattern or sharp softening, they can provide a misleading
view of the status and capacity of the structure, especially when ap-
proaching the dynamic instability. In the current paper, an index is pre-
sented to resolve this problem. The index is completely independent of
the weighted averaging method and consists of a local damage param-
eter and a global response parameter of the structure, so as to be able to
interpret the status of the structure correctly with an appropriate com-
bination of these two parameters in a non-linear dynamic analysis
Fig. 1. The twisting pattern in an IDA curve. under each GM record. The MIDR as one of the most commonly used
B. Mohebi et al. / Journal of Constructional Steel Research 156 (2019) 137–154 141
seismic assessment factors for structures was selected as the local dam- explanations provided, the steps for calculating the proposed DI under
age parameter, based on which important DIs have been presented any GM record are as follows:
[64,65]. Sa (T1) was used to be applied in the proposed DI among overall
response parameters of the structural system. The reason for using Sa 1. Determining the spectral acceleration at the first-mode period of the
(T1) is that, among the global response parameters of the structure, it structure (Sa (T1)).
has reported the highest correlation with Park and Ang [20] DI (one of
2. Performing the non-linear dynamic analysis under the GM record
the most widely used and reliable LDIs) [54,57]. The proposed DI is rep-
and obtaining the MIDR (d) in the structure.
resented by Eq. (10):
3. Performing the IDA analysis through selecting parameters Sa (T1)
and d, as IM and DM, respectively.
d−dy
DI1 ¼ ð8Þ 4. Specifying the yield (Say,dy) and collapse (Sau,du) points on the IDA
du −dy
curve and calculating the DI through Eq. (10).
Sa−Say
DI2 ¼ ð9Þ
Sau −Say Fig. 3 clarifies the sequence of these steps. The equation for the
proposed DI is defined such that if the structure does not reach the
8 9
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi < If DI1 and DI2 ≤0 then : DI ¼ 0 = yield point (A), the damage level equals 0. Besides, if the structure
DI ¼ DI1 DI 2 ð10Þ crosses the collapse point (E), the damage level equals 1, indicating
: ;
If DI1 and DI2 ≥1 then : DI ¼ 1 the dynamic instability of the structure. Therefore, the proposed DI ex-
amines the condition of the structure in the segments between the
where Sa and d are the spectral acceleration at the first-mode period of yield point and the collapse point. Given that, no significant change
the structure (Sa (T1)) and MIDR for each accelerogram, respectively. occurs in the Sa (T1) level by passing the collapse point, and this is
In fact, these parameters are obtained by conducting the non-linear the MIDR that has a sudden increase, DI2 does not exceed 1. As can
dynamical analysis under the selected accelerogram. Say and dy denote be seen, the geometric mean of Eqs. (8) and (9) was used to attain
the spectral acceleration at the first-mode period and the MIDR in the the final equation of the DI. Of the advantages of geometric mean
structure at the yielding state, respectively. In other words, parameters compared to other averaging methods, is that its result is less influ-
Say and dy correspond to the first point where the IDA curve has no enced by too much or too little data [66]. This feature can help pre-
longer the linear state indicated by point A in Fig. 1. Sau and du also vent high number of fluctuations in the values of the DI during
represent the spectral acceleration and MIDR in the structure at the col- severe hardening or softening.
lapse state of the structure (point E in Fig. 1), respectively. After this In order to show that the problems associated with the indices intro-
point, the IDA curve flattens out in a plateau. According to the duced in Section 3.1 are removed by the proposed DI, the value of this DI
Fig. 4. Comparison of the comparative DIs with the proposed one for the IDA curve of Fig. 1 Fig. 5. Comparison of the comparative DIs with the proposed one for the IDA curve of Fig. 2
(the twisting pattern). (the sharp softening pattern).
142 B. Mohebi et al. / Journal of Constructional Steel Research 156 (2019) 137–154
Table 4
Qualitative Comparison of DIs.
Proposed Ghobarah Kunnath Cosenza et al. Amezquita-Sanchez Rajeev & Wijesundara Yu. Chen & Yi. Chen
et al. et al. & Adeli
DI typea global global Local Local Local and global Local Local
c
Loading or Analysis type IDA Push Over Monotonic Monotonic loading FD Analysis Continuous wavelet Monotonic loading
Analysis Analysis loading / THb / TH Analysis transform and IDA and Cyclic loading
Analysis
Considering the stiffness and strength YES NO NO NO NO YES YES
deterioration
Influenced by successive hardening/ NO YES YES YES NO NO YES
the sharp softening patterns
more accurate estimation near the YES NO NO NO NO NO NO
collapse state
Considering the design and YES NO NO NO NO YES NO
construction faults
A combination of global response and YES NO NO NO YES YES YES
local damage parameters
Influenced by Weighted averaging NO NO YES YES NO NO YES
methods
Considering the frequency content YES NO NO NO YES YES NO
effect
a
Defined in Section 3.1.
b
Time History.
c
Fractality Dimension.
was computed for each step of the curves of Figs. 1 and 2 and then com- At this step, a qualitative comparison of the proposed DI and other
pared with the comparative DIs in Figs. 4 and 5, respectively. DIs could improve demonstrating the advantages to the proposed DI.
As indicated in Figs. 4 and 5, when hardening or softening occur All the comparative DIs, along with DIs of Rajeev and Wijesundara
(segment BD in Fig. 1 and segment GH in Fig. 2, respectively), the pro- [17], Yu. Chen and Yi. Chen [24], and Amezquita-Sanchez and Adeli
posed DI values are not influenced and can provide the necessary infor- [47] were compared with the proposed one. The effort was made to con-
mation on reaching the collapse point of the structure in the segments sider numerous comparison fields and to choose different types of DIs.
near this point. However, the comparative DIs are influenced by the The effect of stiffness and strength deterioration, hardening and sharp
IDA curve fluctuations. For example, point C in Fig. 1, whose correspond- softening phenomenon, proximity to the collapse state, and construc-
ing point in Fig. 4 is specified as C′, is located in the hardening segment. tion faults were discussed. Furthermore, the efficiency of each DI with
Fig. 4 depicts that all the DIs of Kunnath et al. [21], Ghobarah et al. [28], regard to considering both local and global damages, the frequency con-
and Cosenza et al. [13] provide a lower damage value for that level of Sa tent, and the use of weighted averaging methods were investigated.
(T1) compared to the proposed DI, and also the damage level at this Table 4 presents the assessment of the mentioned DIs and compares
point is less than that in the hardening starting point (B). The IDA the characteristics thereof.
curve indicates that a slight increase in Sa (T1) to reach point E exposes
the structure to collapse. Thus, the three comparative DIs may not be 4. Structures and non-linear modeling
able to warn about reaching this point. With regard to uncertainties in
modeling and the assumed parameters, in reality, the value of Sa (T1) For the purpose of the present paper, the analyzed structures are di-
could be slightly greater than that obtained in the analysis. As shown vided into three types: A) Single Degree Of Freedom (SDOF) models,
in Fig. 2, this slight increase leads to a high increase in the damage pa- B) stick models [67,68] and C) moment resisting frames, the results of
rameters of the structure or even to global instability of the structure. which are discussed in the following section. All the structures were
While none of the introduced DIs, which do not consider a little ahead made of steel and designed in accordance with the procedures in
and a little behind the current point, could correctly estimate the dam- ASCE/SEI 7–10 [69]. Furthermore, the structures were assumed being
age level, i.e. far less than it may actually occur by taking the precise con- located in an area having a Risk-Targeted Maximum Considered
dition of the structure into account. (MCER) and with site class D. For the design of all the structures, the im-
Therefore, the use of DIs that are influenced by the fluctuations in the portance factor, I = 1, and the seismic design parameters, SDS = 1.25 g
IDA curve may provide limited understanding of the structure's condi- and SD1 = 0.6 g were included.
tions when exposed to different GM records. As fluctuations of the IDA A) SDOF models were assumed with periods of 0.5 s, 1.0 s, and 2.0 s
curve have much less effect on the results of the proposed DI in each and a height of 3000 mm. Also, the ductility ratio, miu = δ / δy, for these
step of this analysis, one can examine the approaching of the global in- structures was assumed 2, in that these structures represent steel ordi-
stability of the structure more accurately using the proposed DI. nary moment resisting frames. The schematic view of the SDOF models
is presented in Fig. 6. Williamson [70] used a number of SDOF models
under various GM records to study the role of the P-Delta effect in re-
sponse of non-elastic systems and concluded that the P-Delta effect is
very important in the structural response and has a significant effect
on the behavior of the analyzed structure. Therefore, for all type
(A) models, the P-Delta effect was included.
B) This group includes structures assigned to the stick model, which
are used as a simplified model of Multi Degree Of Freedom (MDOF)
structures and moment resisting frames. Two-dimensional stick
models, which were 2-, 6-, and 10-story, were applied for damage as-
sessment. The period of each of them equals 0.2 N, where N is consid-
Fig. 6. The schematic view of the SDOF models. ered as the number of stories [67,68]. The height of the stories in all of
B. Mohebi et al. / Journal of Constructional Steel Research 156 (2019) 137–154 143
the models was assumed 3600 mm. The typical floor plan of the struc-
tures is shown in Fig. 7. Gravity loads were applied to all floors as the
dead load of 5.5 kN/m2 and the live load of 2 kN/m2. Given the increase
in the number of stories in comparison to type (A) models and, conse-
quently, the increased impact of the P-Delta effect, for this type of
models, the P-Delta effect was also taken into account. Thus, according
to the ASCE/SEI 7–10 [69], the gravity load of DL + 0.25LL was applied
to all the structures where DL is the dead load and LL is the live load.
2- and 6-story models were assumed steel intermediate moment
resisting frames and the ductility ratio of the members for these models
was considered to be 4. Therefore, according to ASCE/SEI 7–10 [69], the
value of the response modification factor (R) for them is 4.5. Besides, a
steel special moment resisting frame was selected for the 10-story
model, in which the ductility ratio of the members is 6 and the response
modification factor (R) is 8. Fig. 8 illustrates the 2-, 6- and 10-story stick
models. As mentioned earlier, the P-Delta effect was included in all of
the models, but it was excluded from the analysis of this structure by
Fig. 7. Typical floor plan of the structures used for creating the stick models.
creating a new model in order to further examine the P-Delta effect
on the results of the DIs in the 10-story stick model. Then the results
were compared to those of the condition in which the P-Delta effect
had been included.
C) A number of steel moment resisting frames with different geo-
metric and strength characteristics designed by Kitayama and
Constantinou [71] were analyzed. The results are discussed in the next
section. At first, two 3- and 6-story steel moment resisting frames
were selected as the benchmark models, both of which were steel spe-
cial moment resisting frames. W-type sections were used for their
cross-sections. In both models, the height of the first story was assumed
4420 mm and the height of the remaining stories was considered
4304 mm. The frames had three spans, each with a length of 8230 mm.
Fig. 9a and b represent the elevations of 3- and 6-story frames, re-
spectively. In modeling of these structures, in order to consider the P-
Delta effect, leaning columns were used, which is a common method
for taking into account the P-Delta effect in two-dimensional frames.
The loads applied to the structure, as well as the details of the design of
these models, are presented in Kitayama and Constantinou [71]. Then,
two changes were made in the benchmark frames, each aimed at model-
ing the possible faults in the design and construction of structures. This
was done to investigate the damage estimation capability by the pro-
posed DI in structures, which are not fully complying with the codes.
One of these changes is an increase in the height of the first story of the
studied frames, which increases the potential for weak story formation
in the structure. This change was applied to both the 3- and 6-story
frames. The elevation of these frames is shown in Fig. 10a and b.
The second change is the redesign of the frames so that the strong
Fig. 8. The 2-, 6- and 10-story stick models. column-weak beam principle is not observed in the structure, the
Fig. 9. Illustration of the benchmark moment resisting frames: (a) the 3-story model (b) the 6-story model [71].
144 B. Mohebi et al. / Journal of Constructional Steel Research 156 (2019) 137–154
Fig. 10. Illustration of the moment resisting frames with a changed column height: (a) the 3-story model, (b) the 6-story model.
columns have less resistance than the beams, and it is more likely Fig. 13 illustrates the backbone curve of this model. In this figure, the
that plastic hinges are created in columns. The redesign was per- strength parameters include: My = effective yield moment, Mc = cap-
formed in both structures according to ASCE/SEI 7–10 [69], and no ping moment strength, Mr = residual moment, and the deformation pa-
changes were made to the geometric characteristics, loading way, rameters include: θy = yield rotation, θp = pre-capping plastic rotation
and design method. for monotonic loading (difference between yield rotation and rotation
The cross-sections of the redesigned structures are shown in Fig. 11a at maximum moment), θpc = post-capping plastic rotation (difference
and b. Finally, the modified models were compared with the benchmark between rotation at maximum moment and rotation at the point of
models and the results were interpreted. Each of the models in groups complete loss of strength), θu = ultimate rotation capacity. Also, Ke =
A, B and C were labeled according to their characteristics (Table 5). My / θy is the elastic stiffness.
The distributed plasticity and the concentrated plasticity are two In order to find the value of deformation parameters, the equations
common methods for modeling the plasticity of the elements in non- presented by Lignos and Krawinkler [76] were used, in which geometric
linear analyses. Here the concentrated plasticity method was applied. characteristics and material type of the member are determinant. Lignos
In this method, modeled by OpenSees software [72], in each of the and Krawinkler [76] obtained these equations by analyzing the database
beams and columns, an Elastic Beam Column element is used in the cen- results found in Newell and Uang's studies [79].
ter, two Zero Length Elements are used at the two ends, and the plastic To enhance the reliability of the IDA analysis, the structural models
behavior occurs merely in the Zero Length Elements. The stiffness of were analyzed under records with various specifications. For type
Elastic Beam Column Element and Zero Length Elements are considered (A) and (B) structures, 22 pairs of GM records were used according to
serial [73]. Fig. 12 displays the schematic representation of the concen- FEMA P695 [80]. All records were scaled to different intensity levels
trated plasticity model in a single element. until collapse occurred in the structure. For simplicity, details of all
In order to define the plastic behavior (here, the moment-rotation), in GM records are not provided. The IDA analyses were performed for
Zero Length Elements, the cyclic degradation of stiffness and strength type (C) structures using eight records from among the listed records
was included. Many researchers have used deteriorating material models (Table 6). The Hunt and Fill [81] algorithm was used to reach the col-
to study the structural damage [74–78]. In the current study, the modi- lapse point of structures in each IDA analysis. To do this, MATLAB [82]
fied deterioration model of Ibarra and Krawinkler [75] was used. and OpenSees [72] were used simultaneously.
Fig. 11. Illustration of the redesigned moment resisting frames: (a) the 3-story model, (b) the 6-story model.
B. Mohebi et al. / Journal of Constructional Steel Research 156 (2019) 137–154 145
Table 5
Labels of the models.
Table 6
Characteristics of the GM records used for type (C).
Fig. 14. IDA curves for the SDOF-0.5 model under 22 GM record pairs mentioned in FEMA
P695 [52]. Fig. 17. Correlation between the proposed DI and that of Cosenza et al. [13] for the SDOF-
0.5 model.
proposed DI. This can lead to a more precise judgment on the forward
and backward steps by the proposed DI. In other words, the proposed
DI can warn about the approaching of the collapse point through its
gradual increase.
Figs. 16-18 represent the correlation between the results of the
proposed DI and those of the comparative DIs for the SDOF-0.5 model.
To represent the numerical value of correlation, the coefficient of deter-
mination (R2) was used. Besides, a linear relationship with the y = m.x
pattern was obtained for correlation with each index by performing a
linear regression analysis and using the least square method, between
the proposed DI and the comparative DIs. In this equation, x is the
value of the proposed DI, y is the value of the compared index and m
is the line slope, which is another parameter used to evaluate the corre-
lation of the results.
Fig. 15. Comparison of the comparative DIs with the proposed one for the SDOF-0.5 model If parameters R2 and m become closer to unity, the greater correla-
under the GM record No.1 of FEMA P695 [80]. tion between the results of the proposed DI and the comparative DIs is
achieved. Figs. 16-18, which demonstrate the results obtained from
correlation parameters. The important point in Fig. 15 is that the hard- the IDA analysis under 22 pairs of GM records, suggest that the pro-
ening phenomenon in the structures led to a decline in the DIs of posed DI exhibits a high correlation with the comparative DIs for the
Kunnath et al. [21], Cosenza et al. [13] and Ghobarah et al. [28]. It was SDOF-0.5 model. Studying the correlation between the results of the
also observed that the final softening, which leads to the collapse of proposed DI and those of the comparative DIs showed that the results
the structure, is shown with a steep increase in slope, while, for the pro- of the SDOF-1.0 and SDOF-2.0 models were in accordance with the re-
posed DI, the descending branch is not observed in the hardening seg- sults of the SDOF-0.5 model. In Table 7, the results of parameters R2
ments, and the corresponding graph reaches its maximum value from and m are provided for all three models, namely SDOF-0.5, SDOF-1.0,
0 with a gradual increase. Besides, the final softening segment was not and SDOF-2.0.
associated with a sudden increase in the DI. Therefore, according to The results indicate that in type (A) models, as the period of the
the observations, the successive hardening and softening can be con- structure increased, the value of both parameters R2 and m for all the
cluded to have a very lower effect on the damage estimation by the existing indices decreased. For example, R2 value between the proposed
Fig. 16. Correlation between the proposed DI and that of Kunnath et al. [21] for the SDOF- Fig. 18. Correlation between the proposed DI and that of Ghobarah et al. [28] for the SDOF-
0.5 model. 0.5 model.
B. Mohebi et al. / Journal of Constructional Steel Research 156 (2019) 137–154 147
Table 7
Parameters R2 and m between the proposed DI and the comparative DIs for type
(A) models.
R2 m
DI and Cosenza et al. [13] DI is 0.973, 0.968 and 0.862 for the SDOF-0.5,
SDOF-1.0 and SDOF-2.0 models, respectively, indicating that a 4.6% de-
crease occurred for the SDOF-1.0 model and a reduction of about
11.4% occurred for the SDOF-2.0 model compared to the SDOF-0.5
model. In addition, to examine Cosenza et al. [13] DI, parameter m for
the SDOF-0.5 model equals 0.872, while this parameter for the SDOF- Fig. 20. Comparison of the comparative DIs with the proposed one for the ST-10 model
1.0 model is 0.862, and for the SDOF-2.0 model, it is 0.753, representing under the GM record No.16 of FEMA P695 [80].
1.1% and 13.6% reductions in correlation compared to that of the SDOF-
0.5 model. Accordingly, it can be inferred that as the period in type type (A) models, the ascending trend in the proposed DI occurs from
(A) models increased, the value of results obtained from the proposed 0 to 1, while in those ranges of Sa (T1), where hardening and softening
DI diverged from the results of the comparative DIs. The reason can be occur on the IDA curves, fluctuations are perceived on the curves of
the greater influence of P-Delta effect on higher periods, which in- Kunnath et al. [21], Cosenza et al. [13], and Ghobarah et al. [28] DIs.
creases the likelihood of dynamic instability in structures. The weaker This confirms that as hardening and sharp softening phenomena occur
consideration of the P-Delta effect by the comparative DIs can be the in stick models, the proposed DI reports a lower level of affectability. It
outcome of this phenomenon. means that a more accurate judgment can be made on the capacity of
the structural system in these segments, since previous and subsequent
5.2. Stick models steps are also included for exploring the structure status, along with ex-
amining the DIs in this segment. The same phenomenon was observed
This section deals with examining the results of the stick models. At for the other GM records in the other models of type (B).
first, models with 2, 6, and 10 stories that have the periods of 0.4 s, 1.2 s Figs. 21-23 indicate the correlation between the results of the pro-
and 2.0 s, respectively, were analyzed by the same 22 pairs of GM re- posed DI and those of the three comparative DIs for the ST-10 model.
cords mentioned in the previous sections. In addition, the results of In these figures, which were produced using the IDA analysis results
the DIs in these models were studied. Next, to study the role of the P- from the above-mentioned 22 pairs of the GM records, coefficients R2
Delta effect, the results of the DIs in a 10-story model, in which the P- and m between the proposed DI and the comparative DIs are listed in
Delta effect was excluded (ST-10-NP), were examined and compared addition to the graphs for each index. Moreover, coefficients R2 and m
with the results obtained from the ST-10 model. are presented in Table 4 for the quantitative representation of the corre-
lation between the proposed DI and the comparative DIs for the ST-2,
5.2.1. The effect of number of stories ST-6, and ST-10 models as well as to examine the effect of increased
Among the models studied in this section, the IDA curves of the ST- number of stories on the correlation between the results of the DIs.
10 model are shown in Fig. 19. As seen, when each graph flattens, the According to Figs. 21-23, a lower correlation is seen between the re-
collapse of the structural system occurs in the IDA analysis under the sults of the proposed DI and the comparative DIs for the ST-10 model.
GM record associated with that curve. Low correlation, especially in parameter m (the slope of regression
Fig. 20 is presented to examine and compare the change trend in the line), which represents the ratio of the results of the compared index
proposed DI and the comparative ones by increasing the spectral accel- to those of the proposed DI, is clearly visible for all the indices. According
eration and also to examine the sudden hardening and softening phe- to Table 8, the correlation established for the ST-2, ST-6 and ST-10
nomena in the stick models. models between the proposed DI and the DIs of Cosenza et al. [13],
The DIs are presented as a function of Sa (T1) under the GM record Kunnath et al. [21], and Ghobarah et al. [28], reveals that an increase
No.16 of FEMA P695 [80] for the ST-10 model. As with the results of in the number of floors in the stick models, leads to a decrease in both
Fig. 19. IDA curves for the ST-10 model under 22 GM record pairs mentioned in FEMA Fig. 21. Correlation between the proposed DI and that of Kunnath et al. [21] for the ST-10
P695 [80]. model.
148 B. Mohebi et al. / Journal of Constructional Steel Research 156 (2019) 137–154
Table 8
Parameters R2 and m between the proposed DI and the comparative DIs for the ST-2, ST-6
and ST-10 models.
R2 m
Fig. 25. Comparison of the comparative DIs with the proposed one for the ST-10-NP model Fig. 28. Correlation between the proposed DI and that of Ghobarah et al. [28] for the ST-10-
under the GM record No.7 of FEMA P695 [80]. NP model.
Table 9
Parameters R2 and m between the proposed DI and the comparative DIs for the ST-10 and
ST-10-NP models.
R2 m
DIs. Then, the models of Table 5, which were included for comparison
with the 3- and 6-story benchmark models (MRF-3-RD, MRF-3-CH,
MRF-6-RD, MRF-6-CH), were subjected to the same eight records, by
the application of the IDA analysis. Similar to the process applied for
the benchmark models, the damage level was calculated for each
index in these models as well. As mentioned in Section 4, the changes
Fig. 26. Correlation between the proposed DI and that of Kunnath et al. [21] for the ST-10-
made in the benchmark models were implemented in order to investi-
NP model. gate the performance of the proposed DI and the other DIs, if there are
faults related to the design and construction of the structure. In fact,
the 3- and 6-story frames were examined as case studies in Sections
5.3. Moment resisting frames 5.3.1 and 5.3.2, respectively.
Fig. 27. Correlation between the proposed DI and that of Cosenza et al. [13] for the ST-10-
NP model. Fig. 29. IDA curves for the MRF-3 model under the eight GM records of Table 6.
150 B. Mohebi et al. / Journal of Constructional Steel Research 156 (2019) 137–154
Fig. 30. Comparison of the comparative DIs with the proposed one for the MRF-3 model Fig. 32. Correlation between the proposed DI and that of Cosenza et al. [13] for the MRF-3
under the GM record No.1 of Table 6. model.
In Fig. 30, the DIs of Section 3.1 along with the proposed DI are
shown as a function of Sa (T1) for the MRF-3 model, using the IDA anal-
ysis under the GM record No.1 of Table 6. In this curve, the fluctuation
due to hardening can be observed in the curve of Kunnath et al. [21], Co-
senza et al. [13] and Ghobarah et al. [28] DIs, while, a gradual increase in
the curve of the proposed DI occurs similar to previous examples, and
the value of this index rises from 0 to 1. This behavior occurred in the
other models of this section and for the other GM records as well.
Figs. 31-33 display the correlation between the proposed DI and the
comparative DIs for the MRF-3 model under all the eight records of
Table 6. According to these figures, parameters R2 and m show satisfac-
tory values for the MRF-3 model. In fact, these figures indicate that the
results of the proposed DI and the comparative DIs are very close and
highly correlated for this 3-story model, which was designed with com-
plete compliance with the code criteria.
In the following, the MRF-3-RD and MRF-3-CH models were sub-
Fig. 33. Correlation between the proposed DI and that of Ghobarah et al. [28] for the MRF-3
jected to the same eight records mentioned above, by the IDA analysis,
model.
and the values of the proposed DI and the comparative DIs were calcu-
lated. In order to compare the results from these models and the bench-
mark model (MRF-3), coefficients R2 and m were calculated between from the results of these two indices, and the ratio of values of these
the proposed DI and the comparative DIs for these two models. The re- two indices to the proposed DI decreased. A reduction in parameter m
sults are mentioned in conjunction with the results of the MRF-3 model may not have the same interpretation for all the indices. As seen, the
in Table 10. value of m between the proposed DI and Ghobarah et al. [28] DI de-
Comparison of the results of the MRF-3 and MRF-3-RD models in creases from 1.22 to 0.93. This means that, totally, the ratio of Ghobarah
Table 10 indicates that by changing the design and excluding the strong et al. [28] DI values to the proposed DI values decreases. Nevertheless,
column-weak beam principle in the design of the 3-story model, values despite the reduction in m, the results converge as this parameter is ap-
R2 and m decreases from the proposed DI to the Kunnath et al. [21] and proaching 1.
Cosenza et al. [13] DIs.This might mean that by making this change in The results of the MRF-3 and MRF-3-CH models in Table 10 indicate
the quality of design, the response rate of the proposed DI diverged that, as the height of the first column increases, or, expressly, as the po-
tential for formation of weak story enhances, the value of R2 between
the proposed DI and Kunnath et al. [21], Cosenza et al. [13], and
Ghobarah et al. [28] DIs increases up to 3%, 3.2%, and 2.4%, respectively.
Since the difference of R2 between the models of MRF-3 and MRF-3-CH
is small, the results of this parameter can be assumed approximately
equal, while parameter m decreases by 17.7%, 16.4%, and 8.2%, respec-
tively. As is shown, the decrease in m is more significant than the in-
crease in R2. The sharp decline in m shows that, in general, with an
increase in the height of the columns of the first story in the 3-story
Table 10
Parameters R2 and m between the proposed DI and the comparative DIs for the 3-story
models.
R2 m
Fig. 34. IDA curves for the MRF-6 model under the eight GM records of Table 6. Fig. 36. Correlation between the proposed DI and that of Kunnath et al. [21] for the MRF-6
model.
model, the ratio of the results of the comparative DIs to those of the pro-
posed DI dropped. This decreased ratio in Kunnath et al. [21] and Co-
senza et al. [13] DIs means that the results are also diverging, since the
value of m distances from unity. However, like the previous example
of this section, here, Ghobarah et al. [28] DI exhibited a different behav-
ior, and the decrease in the value of m between this index and the pro-
posed DI led to the convergence of the results of these two indices, as
the values of this parameter for the MRF-3 and MRF-3-CH models are
1.22 and 1.12 respectively. This indicates that the results of the pro-
posed DI and Ghobarah et al. [28] DI are more correlated in the MRF-
3-CH model. Therefore, according to the information provided in this
section, it can be inferred that parameters R2 and m were usually re-
duced by making changes, which were pertinent to the design and con-
struction weaknesses. However, this phenomenon does not seem to be
true for Ghobarah et al. [28] DI, as the correlation increased with
changes in the structural model. Given that the results of Ghobarah
et al. [28] DI, which is a GDI, and the proposed DI are obtained using
the pushover analysis and the IDA analysis, respectively, the results of Fig. 37. Correlation between the proposed DI and that of Cosenza et al. [13] for the MRF-6
the proposed DI can be more reliable. The comparative DIs underesti- model.
mate the actual damage level in the structure with the high potential
of weak story formation (the MRF-3-CH model). For further discussion,
the results of the 6-story models are examined in Section 5.3.2. Fig. 35 demonstrates the proposed DI and the comparative DIs ver-
sus Sa (T1) for the MRF-6 model under the GM record No.7 of Table 6.
Figs. 36-38 illustrate the correlation of the results for the MRF-6
5.3.2. 6-Story Frames model. In these figures, the results of the proposed DI are compared
In this section, the IDA analyses were performed for the 6-story with those of the comparative DIs for each step of the IDA analyses,
models (MRF-6, MRF-6-RD, MRF-6-CH) by the eight GM records of and the values of parameters R2 and m are listed alongside each figure.
Table 6. As with the 3-story models, the proposed DI and the compara- Furthermore, The MRF-6-RD and MRF-6-CH models, whose details
tive DIs were calculated for these models in each step of the analysis. were presented in Section 4, were subjected to the above-mentioned
Fig. 34 displays the IDA curves for the MRF-3 model (the benchmark eight GM records by the IDA analysis and the values of all the indices
model), under all the eight records. were calculated for them in order to compare the correlation of results
Fig. 35. Comparison of the comparative DIs with the proposed one for the MRF-6 model Fig. 38. Correlation between the proposed DI and that of Ghobarah et al. [28] for the MRF-6
under the GM record No.7 of Table 2. model.
152 B. Mohebi et al. / Journal of Constructional Steel Research 156 (2019) 137–154
Table 11 and a sudden increase of the DI with a small increase in Sa (T1) value.
Parameters R2 and m between the proposed DI and the comparative DIs for the 6-story The occurrence of these phenomena in calculating the damage level
models.
can provide a misleading view of the status and capacity of the struc-
R2 m ture, especially when approaching the dynamic instability. Another im-
Kunnath Cosenza Ghobarah Kunnath Cosenza Ghobarah portant point is that the use of the weighted averaging method to
expand LDIs to the whole structure can also lead to erroneous results.
MRF-6 0.811 0.774 0.861 0.666 0.630 1.049
MRF-6-RD 0.769 0.747 0.707 0.651 0.615 0.773 For the same reasons, the proposed DI was defined such that it can re-
MRF-6-CH 0.746 0.730 0.847 0.555 0.543 0.816 move the effects of the above-mentioned issues on the values of DIs.
Further, it can warn about approaching the structure collapse via exam-
ining previous and subsequent steps creating an ascending trend in
of each model with those of the MRF-3 model. As mentioned in the pre- their values. For this reason, a more accurate judgment can be made
vious sections, this is done in order to evaluate the effects of design and about the capacity of the structural system in these segments. In order
construction faults. Table 11 shows the parameters R2 and m between to examine the level of correlation between the results of the proposed
the proposed DI and the comparative DIs for the MRF-6, MRF-6-RD, DI and the comparative DIs, two parameters of R2 and m were used. It
and MRF-6-CH models. was seen that as the period and/or the number of floors increased, the
The trends of results are similar to those of Section 5.3.1, but in con- results of the proposed DI diverged from those of the comparative DIs.
tradiction to them, no exception in the values of R2 and m is seen for any In the higher structures, the role of overstrength and the P-Delta effect
of the models and the indices studied. Expressly, by making both the is more significant. Therefore, as explained earlier, the precision of the
intended changes in the benchmark model, the correlation between comparative DIs decreased. In particular, this phenomenon is more pal-
the proposed DI and all the comparative DIs dropped. pable at the collapse state. In addition, as the P-Delta effect was ex-
According to the results of Section s 5.3.1 and 5.3.2, it can be con- cluded, the correlation between the proposed DI and the comparative
cluded that in general, if a kind of inconsistency exists in the structural DIs increased. Moreover, the comparative DIs cannot offer an accurate
system indicating the weaknesses of design or construction, the correla- description of a sudden damage mechanism, a phenomenon related to
tion between the results of the proposed DI and the comparative DIs the instability of the whole structure.
mentioned in Section 3.1, may decrease. This suggests that the results Accordingly, due to the conceivable faults in the structure, the
of the proposed DI and the comparative DIs are diverging. By examining weighted averaging method could not result in the correct estimation
the frames with various inconsistency types, it is observed that there of the damage state. In other words, if the structural system complies
were various scenarios for different situations of the plastic hinges and with the current codes, a relatively high correlation between the value
eventually a sudden damage mechanism in the structure. In fact, differ- of the proposed DI and that of the comparative DIs would be observed.
ent collapse mechanisms and arrangement of plastic hinges can cause However, changing the design approach and excluding the strong
structural instability or only local damage to a structure under seismic column-weak beam principle in the design, which occasionally leads
loads. The changed models were designed so that the probability for to the formation of plastic joints with various arrangements in different
the occurrence of an instability in these models is more than the bench- points of the structure and global instability, decreases the correlation
mark models due to a weakness in columns. When dynamic instability parameters between the proposed DI and the comparative DIs. The pro-
occurred, that was more likely to be found in the changed models, the posed DI has the ability to overcome these shortcomings and address
correlation between the results of the proposed DI and those of the more accurate damage levels. It can be used to achieve a better estima-
comparative DIs, compared to the benchmark models, decreased. tion of the structure performance after earthquakes.
Cosenza et al. [13] and Kunnath et al. [21] DIs that calculate the damage
in the whole structure by weighting the damage created in the ele-
ments, could not consider the sudden occurrence of a collapse mecha- References
nism, which is a phenomenon related to the instability of the whole
[1] A. Ghobarah, Performance-based design in earthquake engineering: state of devel-
structural system. In addition, due to utilizing the pushover analysis, opment, Eng. Struct. 23 (2001) 878–884.
Ghobarah et al. [28] DI could not perform an accurate examination of [2] FEMA, Federal Emergency Management Agency, Next-generation performance
how dynamic instability occurs in structures. based seismic design guidelines: FEMA 445, Washington D. C, 2006.
[3] D. Grecea, F. Dinu, D. Dubina, Performance criteria for MR steel frames in seismic
zones, J. Constr. Steel Res. 60 (2004) 739–749.
6. Conclusions [4] M.N. Fardis, Advances in Performance-Based Earthquake Engineering, Geotechnical,
Geological and Earthquake Engineering, Springer Science & Business Media:
Dordrecht, Netherlands, 13, 2010.
In recent decades, applications of DIs have been proposed for making [5] A. Tenchini, C. M. D'Aniello, R. Rebelo, L.S. Landolfo, L. Lima da Silva, Seismic perfor-
numerical relationships between damage levels and performance ob- mance of dual-steel moment resisting frames, J. Constr. Steel Res. 101 (2014)
jectives. The introduced indices are considered based on the type of 437–454.
[6] E. Brunesi, R. Nascimbene, G.A. Rassati, Response of partially-restrained bolted
EDPs and divided into local and global groups that calculate the damage
beam-to-column connections under cyclic loads, J. Constr. Steel Res. 97 (2014)
level for a member of the structure and the whole structure, respec- 24–38.
tively. In the current study, a DI was proposed based on a certain com- [7] K. Wijesundara, R. Nascimbene, G.A. Rassati, Modeling of different bracing configu-
bination of MIDR as a local damage parameter and Sa (T1) as a global rations in multi-storey concentrically braced frames using a fiber-beam based ap-
proach, J. Constr. Steel Res. 101 (2014) 426–436.
response. Three DIs, namely Kunnath et al. [21], the Cosenza et al. [8] E. Brunesi, R. Nascimbene, G.A. Rassati, Seismic response of MRFs with partially-
[13], and Ghobarah et al. [28] DIs, were selected to be compared with restrained bolted beam-to-column connections through FE analyses, J. Constr.
the proposed DI in different structures. Among the various behaviors Steel Res. 107 (2015) 37–49.
[9] SEAOC, Structural Engineers Association of California, Performance-Based Seismic
of structures, successive hardening and softening (twisting pattern), Engineering of Buildings, Technical Report, Structural Engineers Association of Cali-
and sharp softening (dynamic instantaneous instability) can be men- fornia, Vision 2000: Sacramento, California, 1995.
tioned. The equation of the DIs used in this study represents lower [10] F.M. Mazzolani, R. Montuori, V. Piluso, Performance based design of seismic-
resistant MR frames, in: F.M. Mazzolani, R. Tremblay (Eds.),Behaviour of Steel Struc-
values at the end of the hardening segment compared to the beginning tures in Seismic Areas, Proceedings of the Third International Conference STESSA
of this segment, while as this segment ends, with a slight increase in Sa 2000 2000, pp. 611–618 , 21–24 August, Montreal, Canada. Rotterdam, A.A.
(T1), the structure suffers from dynamic instability. In addition, in the Balkema.
[11] H. Banon, J.M. Biggs, H.M. Irvine, Seismic damage in reinforced concrete members, J.
instantaneous softening phenomenon, which could be due to construc- Struct. Eng. ASCE 107 (9) (1981) 1713–1729.
tion faults, the P-Delta effect or the formation of a weak story, the struc- [12] M.S.L. Roufaiel, C. Meyer, Analytical modelling of hysteretic behaviour of RC frames,
ture suffers from sharp softening leading to a sharp increase in MIDR J. Struct. Eng. ASCE 113 (3) (1987) 429–443.
B. Mohebi et al. / Journal of Constructional Steel Research 156 (2019) 137–154 153
[13] E. Cosenza, G. Manfredi, R. Ramasco, The use of damage functionals in earthquake [45] M. Zhang, R. Schmidt, Sensitivity analysis of an auto-correlation-function based
engineering: a comparison between different methods, Earthq. Eng. Struct. Dyn. damage index and its application in structural damage detection, J. Sound Vib.
22 (10) (1993) 855–868. 333 (26) (2014) 7352–7363.
[14] J.E. Stephens, J.T.P. Yao, Damage assessment using response measurements, J. Struct. [46] A. Eraky, A.M. Anwar, A. Saad, A. Abdo, Damage detection of flexural structural sys-
Eng. ASCE 113 (1987) 787–801. tems using damage index method-Experimental approach, Alex. Eng. J. 54 (3)
[15] M.L. Wang, S.P. Shah, Reinforced concrete hysteresis model based on the damage (2015) 497–507.
concept, Earthq. Eng. Struct. Dyn. 15 (1987) 993–1003. [47] J.P. Amezquita-Sanchez, H. Adeli, Synchrosqueezed wavelet transform-fractality
[16] S.L. McCabe, W.J. Hall, Assessment of seismic structural damage, J. Struct. Eng. ASCE model for locating, detecting, and quantifying damage in smart high-rise building
115 (1989) 2166–2183. structures, Smart Mater. Struct. 24 (6) (2015).
[17] P. Rajeev, K.K. Wijesundara, Energy-based damage index for concentrically braced [48] Z.X. Tan, D.P. Thambiratnam, T.H.T. Chan, H. Abdul Razak, Detecting damage in steel
steel structure using continuous wavelet transform, J. Constr. Steel Res. 103 beams using modal strain energy based damage index and Artificial Neural Net-
(2014) 241–250. work, Eng. Fail. Anal. 79 (2017) 253–262.
[18] C. Castiglioni, R. Pucinotti, Failure criteria and cumulative damage models for steel [49] A. Sharifi, Mahmoud-R. Banan, Mohammad-R. Banan, A strain-consistent
components under cyclic loading, J. Constr. Steel Res. 65 (2009) 751–765. approach for determination of bounds of ductility damage index for different
[19] H. Banon, D. Veneziano, Seismic safety of reinforced concrete members and struc- performance levels for seismic design of RC frame members, Eng. Struct. 37
tures, Earthq. Eng. Struct. Dyn. 10 (1982) 179–193. (2012) 143–151.
[20] Y.-J. Park, A.H.-S. Ang, Mechanistic seismic damage model for reinforced concrete, J. [50] M.S. Alhaddad, Kh.M. Wazira, Y.A. Al-Salloum, H. Abbas, Ductility damage indices
Struct. Eng. ASCE 111 (4) (1985) 722–739. based on seismic performance of RC frames, Soil Dyn. Earthq. Eng. 77 (2015)
[21] S.K. Kunnath, A.M. Reinhorn, R.F. Lobo, IDARC Version 3.0: A Program for the Inelas- 226–237.
tic Damage Analysis of Reinforced Concrete Structures, Technical Report NCEER-92- [51] G. Nie, Ch-x. Zhang, X-d. Zhi, J. Dai, Damage quantification, damage limit state
0022, National Center for Earthquake Engineering Research, State University of criteria and vulnerability analysis for single-layer reticulated shell, Thin-Walled
New-York, Buffalo NY, 1992. Struct. 120 (2017) 378–385.
[22] Y. Bozorgnia, V.V. Bertero, Evaluation of damage potential of recorded earthquake [52] W. Huang, M. Zou, J. Qian, Zh. Zhou, Consistent damage model and performance-
ground motion, 96th annual meeting of Seismological Society of America, Seismol. based assessment of structural members of different materials, Soil Dyn. Earthq.
Res. Lett. 74 (2001) 312. Eng. 109 (2018) 266–272.
[23] P. Fajfar, Equivalent ductility factors, taking into account low-cycle fatigue, Earthq. [53] R. Pang, B. Xu, X. Kong, D. Zou, Seismic fragility for high CFRDs based on deformation
Eng. Struct. Dyn. 21 (10) (1992) 837–848. and damage index through incremental dynamic analysis, Soil Dyn. Earthq. Eng. 104
[24] Y. Chen, Y. Chen, The strain-weighted energy damage model for structural steel (2018) 432–436.
under cyclic loading, J. Constr. Steel Res. 139 (2017) 449–456. [54] A. Elenas, Correlation between seismic acceleration parameters and overall struc-
[25] Y.-J. Park, A.H.-S. Ang, Y.K. Wen, Seismic damage analysis of reinforced concrete tural damage indices of buildings, Soil Dyn. Earthq. Eng. 20 (2000) 93–100.
buildings, J. Struct. Eng. ASCE 111 (4) (1985) 740–757. [55] Y. Lieping, M. Qianli, M. Zhiwei, G. Hong, Zh. Yan, Numerical and comparative study
[26] J.M. Bracci, A.M. Reinhorn, J.B. Mander, S.K. Kunnath, Deterministic Model for Seis- of earthquake intensity indices in seismic analysis, Struct. Des. Tall and Spec. Build.
mic Damage Evaluation of RC Structurese, Technical Report NCEER-89-0033, Na- 22 (4) (2013) 362–381.
tional Center for Earthquake Engineering Research, State University of New-York, [56] M. Kumar, P.J. Stafford, A.Y. Elghazouli, Influence of ground motion characteristics
Buffalo NY, 1989. on drift demands in steel moment frames designed to Eurocode 8, Eng. Struct. 52
[27] E. DiPasquale, A.S. Cakmak, Identification of the serviceability limit state and detec- (2013) 502–517.
tion of seismic structural damage, Technical Report NCEER-88-0022, National Cen- [57] V.V. Cao, H.R. Ronagh, Correlation between seismic parameters of far-fault motions
ter for Earthquake Engineering Research, State University of New-York, Buffalo and damage indices of low-rise reinforced concrete frames, Soil Dyn. Earthq. Eng. 66
NY, 1988. (2014) 102–112.
[28] A. Ghobarah, H. Abou-Elfath, A. Biddah, Response based damage assessment of [58] K. Kostinakis, A. Athanatopoulou, K. Morfidis, Correlation between ground motion
structures, Earthq. Eng. Struct. Dyn. 28 (1) (1999) 79–104. intensity measures and seismic damage of 3D R/C buildings, Eng. Struct. 82
[29] T.-H. Kim, K.-M. Lee, Y.-S. Chung, H.M. Shin, Seismic damage assessment of rein- (2015) 151–167.
forced concrete bridge columns, Eng. Struct. 27 (4) (2005) 576–592. [59] A. Massumi, F. Gholami, The influence of seismic intensity parameters on structural
[30] S.A. Diaz, L.G. Pujades, A.H. Barbat, Y.F. Vargas, D.A. Hidalgo-Leiva, Energy damage damage of RC buildings using principal components analysis, Appl. Math. Model. 40
index based on capacity and response spectra, Eng. Struct. 152 (2017) 424–436. (3) (2016) 2161–2176.
[31] D. Vamvatsikos, C.A. Cornell, Incremental dynamic analysis, Earthq. Eng. Struct. Dyn. [60] A.D. Hanganu, E. Onate, A.H. Barbat, A finite element methodology for local/global
313 (2002) 491–514. damage evaluation in civil engineering structures, Comput. Struct. (80 (20–21)
[32] R. Villaverde, P.E. Mjlk, ASCE, Methods to assess the seismic collapse capacity of (2002) 1667–1687.
building structures: state of the art, J. Struct. Eng. ASCE 133 (1) (2007) 57–66. [61] S. Amziane, J.F. Dubé, Global RC structural damage index based on the assessment of
[33] M.S.L. Roufaiel, C. Meyer, Reliability of RC frames damaged by earthquakes, J. Struct. local material damages, J. Adv. Concr. Technol. 6 (3) (2008) 459–468.
Eng. ASCE 113 (3) (1987) 445–457. [62] H.U. Koyluoglu, S.R.K. Nielsen, A.Ş. Çakmak, P.H. Kirkegaard, Prediction of global and
[34] J.P. Amezquita-Sanchez, H. Adeli, Signal processing techniques for vibration- localized damage and future reliability for RC structures subject to earthquakes,
based health monitoring of structures, Arch. Comput. Methods Eng. 23 (1) Earthq. Eng. Struct. Dyn. 26 (4) (1997) 463–475.
(2016) 1–15. [63] E. Cosenza, G. Manfredi, Classificazione e comportamento sismico di modelli ciclici
[35] Z. Li, H.S. Park, H. Adeli, New method for modal identification and health moni- degradanti, Proc. of Workshop on Danneggiamento Ciclico e Prove Pseudo-
toring of super high-rise building structures using discretized synchrosqueezed Dinamiche 1994, pp. 59–74.
wavelet and hilbert transforms, Struct. Des. Tall and Spec. Build. 26 (3) (2017) [64] M.A. Sozen, Review of Earthquake Response of Reinforced Concrete Buildings with a
1–16. View to Drift Control, State of the Art in Eartquake Engineering, Turkish National
[36] C.A. Perez-Ramirez, J.P. Amezquita-Sanchez, H. Adeli, M. Valtierra-Rodriguez, D. Committee on Earthquake Engineering, Istanbul, Turkey, 1981.
Camarena-Martinez, R.J. Romero-Troncoso, New methodology for modal parame- [65] A. Ghobarah, On Drift Limits Associated with different damage Levels, Proceedings
ters identification of smart civil structures using ambient vibrations and of International Workshop on Performance-Based Seismic Design, Department of
synchrosqueezed wavelet, Eng. Appl. Artif. Intell. 48 (2016) 1–12. Civil Engineering, McMaster University, 2004 , Bled, 28 June-1 July.
[37] J.P. Amezquita-Sanchez, H. Adeli, A new music-empirical wavelet transform meth- [66] D. Clark-Carter, Geometric Mean, Encyclopedia of Statistics in Behavioral Science, 2,
odology for time-frequency analysis of noisy nonlinear and non-stationary signals, 2005 744–745.
Digital Signal Proc. 45 (2015) 55–68. [67] V. Gicev, M.D. Trifunac, Transient and permanent shear strains in a building
[38] J.P. Amezquita-Sanchez, H. Adeli, Feature extraction and classification techniques for excited by strong earthquake pulses, Soil Dyn. Earthq. Eng. 29 (10) (2009)
health monitoring of structures, Sci. Iran. Trans. A: Civil Eng. 22 (6) (2015) 1358–1366.
1931–1940. [68] Z.W. Yu, H.Y. Liu, W. Guo, Q. Liu, A general spectral difference method for calculating
[39] Y.Z. Lin, Z.H. Nie, H.W. Ma, Structural damage Detection with Automatic Feature- the minimum safety distance to avoid the pounding of adjacent structures during
extraction through Deep Learning, Comput. Aid. Civil Infrastr. Eng. 32 (12) (2017) earthquakes, Eng. Struct. 150 (2017) 646–655.
1025–1046. [69] ASCE, American Society of Civil Engineers, Minimum Design Loads for Buildings and
[40] B.K. Oh, K.J. Kim, Y. Kim, H.S. Park, H. Adeli, Evolutionary Learning based Sustainable Other Structures, ASCE/SEI, Reston (VA), 2010 7–10.
Strain Sensing Model for Structural Health monitoring of High-rise buildings, Appl. [70] E.B. Williamson, Evaluation of damage and P- effects for systems under earthquake
Soft Comput. 58 (2017) 576–585. excitation, J. Struct. Eng. 129 (8) (2003) 1036–1046.
[41] G.F. Sirca Jr., H. Adeli, System Identification in Structural Engineering, Scientia [71] S. Kitayama, M.C. Constantinou, Probabilistic collapse resistance and residual drift
Iranica-Transaction a, Civ. Eng. 19 (6) (2012) 1355–1364. assessment of buildings with fluidic self-centering systems, Earthq. Eng. Struct.
[42] B.K. Oh, D. Kim, H.S. Park, Modal response-based visual system identification and Dyn. 45 (12) (2016) 1935–1953.
model updating methods for building structures, Comput. Aid. Civil Infrast. Eng. [72] OpenSees, Open system for earthquake engineering simulation Pacific Earthquake
32 (1) (2017) 34–56. Engineering Research Center (PEER), http://opensees.berkeley.edu 2006.
[43] E. Poskus, G.W. Rodgers, C. Zhou, J.G. Chase, Damage identification for hysteretic [73] L. Eads, Pushover Analysis of 2-Story Moment Frame, from OpenSees Wiki, http://
structures using a mode decomposition method, Comput. Aid. Civil Infrast. Eng. 33 opensees.Berkeley.edu 2009.
(2) (2018) 97–109. [74] L.F. Ibarra, H. Krawinkler, Global Collapse of Frame Structures under Seismic Excita-
[44] H.R. Ahmadi, F. Daneshjoo, N. Khaji, New damage indices and algorithm based on tions, Pacific Earthquake Engineering Research Center, 2005.
square time–frequency distribution for damage detection in concrete piers of rail- [75] D. Lignos, H. Krawinkler, Sidesway collapse of deteriorating structural systems
road bridges, Struct. Control. Health Monit. 22 (1) (2014) 91–106. under seismic excitations, in: A. John (Ed.), Report No. TB 172, Blume Earthquake
154 B. Mohebi et al. / Journal of Constructional Steel Research 156 (2019) 137–154
Engineering Research Center, Department of Civil and Environmental Engineering. Construction, Inc, Department of Structural Engineering, University of California,
Stanford University, Stanford, CA, 2009. San Diego, 2006.
[76] D. Lignos, H. Krawinkler, Deterioration modeling of steel components in support of [80] FEMA, Federal Emergency Management Agency, Quantification of Building Seismic
collapse prediction of steel moment frames under earthquake loading, J. Struct. Eng. Performance Factors: FEMA P695, Redwood City, California, 2009.
ASCE 137 (11) (2010) 1291–1302. [81] D. Vamvatsikos, Performing incremental dynamic analysis in parallel, Comput.
[77] A. Pourgharibshahi, T. Taghikhany, Reliability-based assessment of deteriorating Struct. 89 (1) (2011) 170–180.
steel moment resisting frames, J. Constr. Steel Res. 71 (2012) 219–230. [82] MathWorks, Inc, MATLAB (R2016a): the Language of Technical Computing. Desktop
[78] M. Bosco, L. Tirca, Numerical simulation of steel I-shaped beams using a fiber-based Tools and Development Environment, MathWorks, 2016.
damage accumulation model, J. Constr. Steel Res. 133 (2017) 241–255. [83] H. Tavakoli, M. Moradi, A, Robustness analysis of steel structures with various lateral
[79] J. Newell, C.M. Uang, Cyclic Behavior of Steel Columns with Combined High Axial load resisting systems under the seismic progressive collapse, Eng. Fail. Anal. 83
Load and Drift Demand, Report No. SSRP-06/22, American Institute of Steel (2017) 88–101.