You are on page 1of 13

Engineering Structures 52 (2013) 654666

Contents lists available at SciVerse ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

An adaptive pushover procedure based on effective modal mass


combination rule
Reza Abbasnia , Alireza Tajik Davoudi, Mohammad M. Maddah
Department of Civil Engineering, Iran University of Science and Technology, Tehran 1684613114, Iran

a r t i c l e

i n f o

Article history:
Received 22 April 2012
Revised 13 March 2013
Accepted 19 March 2013
Available online 24 April 2013
Keywords:
Adaptive pushover
Higher modes
Sign reversal
Modal mass combination rule

a b s t r a c t
In order to overcome the major drawbacks of conventional pushover methods, researchers have recently
been motivated to develop adaptive pushover procedures by which effect of higher modes as well as progressive damage accumulation are taken into account. In spite of their vigorous theory, these novel methods suffer from the quadratic modal combination rules, in which the sign reversals of load vectors in
higher modes are neglected and consequently lead to a positive load pattern. In this paper, a displacement-based adaptive modal pushover method, called APAM, based on effective modal mass combination
rule is developed in order to include the sign reversals in the load vectors. In this combination rule a modication factor associated to each mode of interest is determined and applied to the corresponding load
vector. The modied modal load vectors are algebraically added and subtracted and result in a range of
load pattern and thus, multiple pushover analysis is required. These load patterns are independently
applied to the structure within an adaptive framework and the envelope of demand values is considered.
These modication factors are updated proportional to the instantaneous dynamic characteristic of structure in each step. Another novel aspect of the proposed method is that the target displacement is estimated during the analysis by implementing the concept of capacity spectrum method recommended
by ATC 40. In order to assess the accuracy of this method in predicting the seismic responses, the proposed methodology is applied to three different moment-frame buildings. The obtained results demonstrate that APAM procedure provides well estimation of important seismic demand parameters.
2013 Elsevier Ltd. All rights reserved.

1. Introduction
Earlier versions of pushover methods presented in different
code provisions such as FEMA-356 [1], EuroCode-8 [2] and ATC40 [3], are limited to fundamental mode of the structures. In these
procedures, the structure is subjected to an invariant load pattern
until a predetermined target displacement is achieved or collapse
occurs. There are two major drawbacks in the conventional pushover methods: (I) neglect of the higher mode effects [410] and
(II) neglect the changes in the dynamic properties of the structures
that leads to a continuously altered loading pattern [1114]. Due to
these important shortcomings, conventional pushover cannot
accurately predict the response of structures when the higher
mode effects are considerable [410].
Extensive research has been conducted in recent years to overcome the aforementioned deciencies. Some researchers are limited to taking into account the higher mode effects while the
modal load vectors are constant during the analysis (rst category).
Some others consider both the higher mode effects as well as
changing in the dynamic properties of the structure which is so
Corresponding author. Tel.: +98 9122186903; fax: +98 2177240398.
E-mail address: abbasnia@iust.ac.ir (R. Abbasnia).
0141-0296/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2013.03.029

called adaptive pushover analysis (second category). In the former


category, the earliest attempt was conducted by Paret et al. [4] who
introduced the multi-modal pushover procedure. One of the most
famous procedures in this category of pushover procedures is modal pushover analysis (MPA) of building was developed by Chopra
and Goel [8]. In this method, which is a multi-run procedure, the
structure is subjected to different load vectors (proportion to each
mode) and the modal responses are combined with quadratic modal combination rules such as the square root of the sum of the
squares (SRSS) and the complete quadratic combination (CQC).
Since the responses of the structure subjected to higher mode load
vectors are within the elastic range, the modied modal pushover
analysis (MMPA) procedure [15] was proposed in which the response of higher modes can be calculated by a response spectrum
analysis.
Although the methods that consider the higher mode contributions present a better estimation of seismic response of structure in
comparison to the conventional pushover methods, the effect of
the progressive damage accumulation and the subsequent modication of the load pattern is not taken into account.
In order to reect the progressive stiffness degradation of structure on the load vectors in an inelastic analysis, the adaptive pushover analysis methods have been introduced in recent years. The

R. Abbasnia et al. / Engineering Structures 52 (2013) 654666

earliest version of adaptive pushover was presented by Reinhorn


[11] and Bracci et al. [12] in which the load pattern is updated from
the instantaneous base shear and story resistance of the previous
load step. Gupta and Kunnath [13] presented a force-based adaptive pushover procedure (FAP) in which the applied load pattern
is a function of mass-normalized mode shape, modal participation
factor, spectral amplication of considered modes and weight of
story. In this regard, an eigenvalue analysis is performed to calculate mode shapes in each step. The load vector related to each
mode is applied to the structure independently and the response
in each step is calculated by the quadratic mode combination rule.
Elnashai [14] and Antoniou and Pinho [16] elaborated the previous
forced-based adaptive methodologies and developed a single-run
full adaptive pushover method.
In spite of the conceptual superiority of forced-based adaptive
pushover methods, the prediction of seismic responses has not
been signicantly improved [1618]. The main reason for this inaccuracy can be related to the use of the quadratic modal combination rules [1618]. In these rules, the sign reversals in the modal
load vectors are neglected which results in a constantly positive
load pattern.
Antonio and Pinho [19] have developed a displacement-based
adaptive pushover method (DAP) to improve the prediction of
adaptive pushover methods. The general concept of this method
is the same as FAP procedure. The only important difference is
the implementation of the displacement vector instead of the force
vector. It is conrmed by several researches that DAP procedure
has improved the prediction of seismic demands of the structures
in comparison to FAP method [20,21]. Although the accuracy of the
results has became more satisfactory, the aforementioned problems arising from SRSS or CQC, still exist. Kalkan and Kunnath
[22] proposed a multi-run adaptive pushover, which is called adaptive pushover combination AMC. The authors combined the MPA
method, capacity spectrum method (CSM) and adaptive modal
procedure in which the target displacement is computed during
the analysis. The seismic demand is predicted more accurately in
comparison to the conventional methods [22,23]. However, interaction between modes in the inelastic range is not considered
and the responses are computed based on SRSS or CQC rules. Recently, Shakeri et al. [24] have proposed a story shear-based adaptive pushover method where the applied load pattern is derived
from the instantaneous combined modal story shear prole. The
lateral load pattern is calculated by subtracting the combined
modal shear of consecutive stories. This method underestimates
the crucial seismic demand such as drift.
As mentioned above, quadratic modal combination is a serious
problem in the adaptive modal pushovers which leads to decreasing the precision of seismic demand prediction. In the following,
the research related to the modal combination rules rather than
quadratic modal combinations in a pushover analysis are
presented.

2. Alternative modal combination rules


The crucial shortcoming of the quadratic modal combination
rule is that the signs of modal load vectors are suppressed during
the combination. In other words, the possible negative sign of
modal load vectors is eliminated and inevitably leads to monotonically increasing load vectors [1618].
Matsumori et al. [25] used the alternative modal combination
rule to estimate the response of structure. In this methodology
two independent story shear patterns including the sum and the
difference of two modal story shears, are utilized. Afterward, Kunnath [26] presented a pushover method in which multiple invariant load patterns are calculated by adding and subtracting modal

655

story loads. The following expression (Eq. (1)) is used to compute


the story forces in this methodology:

Fi

k
X

 Rj  Cj  M  /ij  Saj fj ; T j

j1

where Fi is the lateral force to be applied at story level i and j stand


for the mode number; Rj, a modication factor to scale the contribution of each mode; Saj, the spectral acceleration at period Tj for
mode j and corresponding to the damping ratio fj. /ij, the ith component of the jth eigenvector (mode shape) and Cj is the modal participation factor for the jth mode.
The obtained story forces (Fi) are applied to the structure independently and the envelope of responses is considered as the seismic demand. The challenging matter in this method is how to
determine the modication factor (Rj). However, the author does
not present a distinct rule to compute this modication factor
[27]. Park et al. [27] used the concept of the Kunnath procedure
and developed a novel method whereby modal combination factor
of each mode is presented. The modal combination factors (Rj) are
calculated from a comprehensive set of elastic time history analyses. Although this procedure can accurately predict the seismic demand of the structure in the elastic range, it cannot signicantly
improve the accuracy of the predictions in the inelastic range.
In the present study, in order to overcome the above mentioned
shortcomings arising from the quadratic modal combination rules,
an adaptive pushover analysis based on Modal Mass Participation
(APAM) is developed. The proposed method, which is a displacement-based adaptive procedure, employs an effective modal mass
combination rule (EMMC) to construct the applied load pattern. In
EMMC rule, the sign of each modal load vector is maintained and
unlike the quadratic modal combination rules, the sign reversals
in the load vectors are included. In addition, the proposed technique employs the concept of CSM [3] and AMC methodology
[22] to estimate the target displacement during the analysis and
therefore, the challenge related to estimating a predened target
displacement is eliminated. It should be noted that this method
is intended to estimate the seismic demands in the building
frames.
3. Adaptive pushover analysis based on Modal Mass
Participation
The main advantages of APAM procedure, as mentioned previously, rely upon the methodology which is implemented in dening the load pattern as well as determination of the target
displacement during the analysis.
In the following the basic elements of the proposed procedure
are discussed in detail.
3.1. Load pattern
In the APAM procedure the effects of higher modes, the changes
in the dynamic characteristics of structure in the inelastic range
and the effect of the frequency content of a specic response spectrum on the load pattern are taken into account. In addition, in order to overcome the known drawbacks in the quadratic
combination rule, herein the EMMC rule for multi-story buildings
is presented. In this combination rule a modication factor associated to each mode of interest is determined and applied to the corresponding load vector. The modied modal load vectors are
algebraically added and subtracted which result in a range of load
pattern and thus, multiple pushover analyses are required. These
load patterns are independently applied to the structure within
an adaptive framework and the envelope of demand values are
considered. These modication factors are updated proportional

656

R. Abbasnia et al. / Engineering Structures 52 (2013) 654666

to the instantaneous dynamic characteristic of structure in each


step. The process for determining the applied load pattern at one
step of the proposed method (APAM) is schematically depicted in
Fig. 1.
The story displacement associated to each mode (Fig. 1a) is calculated at each step by the following equation:

Dij Cj  /ij  Sdj fj ; T j

where Sdj is the spectral displacement corresponding to the jth


mode and Dij is modal displacement at the story i related to jth
mode.
Now, EMMC rule is implemented to obtain the applied displacement Di at the story i through:

Di

n
X

 Rj  Cj  /ij  Sdj fj ; T j

Rj

aj
amax

amax Maxa1 ; a2 ; . . . ; aj

where aj is the modal mass coefcient corresponding to the jth


mode and amax is the greatest value between modal mass coefcients for all modes. aj is obtained using Eq. (6):

aj

f/ij gT mflg2
P
f/ij gT mf/ij g j mj

where {l} and [m] are the inuence and mass matrixes, respectively.
If the rst three modes are considered, regarding to Eq. (3), the
following four load patterns would be used (Fig. 1c):

j1

In Eq. (3), Rj is a relative mode contribution factor which is dened


for each mode using the following equations:

1  Di R1  C1  /i1  Sd1 f1 ; T 1 R2  C2  /i2  Sd2 f2 ; T 2 R3


 C3  /i3  Sd3 f3 ; T 3

Fig. 1. The process for determining the applied load pattern at one step of the proposed method (APAM) in comparison to DAP method. (a) Mode shapes and modal load
vectors. (b) Load pattern in DAP method. (c) Load pattern in APAM method.

R. Abbasnia et al. / Engineering Structures 52 (2013) 654666

2  Di R1  C1  /i1  Sd1 f1 ; T 1 R2  C2  /i2  Sd2 f2 ; T 2  R3


 C3  /i3  Sd3 f3 ; T 3

3  Di R1  C1  /i1  Sd1 f1 ; T 1  R2  C2  /i2  Sd2 f2 ; T 2 R3


 C3  /i3  Sd3 f3 ; T 3

4  Di R1  C1  /i1  Sd1 f1 ; T 1  R2  C2  /i2  Sd2 f2 ; T 2


 R3  C3  /i3  Sd3 f3 ; T 3

10

Finally, these load patterns are applied to the structure independently, and the envelope of demand values are obtained.
Since using the modal interstory drift for each mode (/ij  /(i1)j),
instead of modal displacement (/ij), leads to much more improved
results [19], herein an interstory drift-based scaling technique is
used as shown in Eq. (11) instead of Eq. (3)

Di

i
X
Di
k1

Di

k
X

 Rj  Cj  /ij  /i1j  Sdj fj ; T j

11

In this paper, the proposed methodology is conceptually analogous in approach to Kalkans method [22] for determining the target displacement, which was described above, but in a single-run
framework. In this regard, a set of inelastic response spectra in
ADRS format with various ductility level are determined. An interval of 0.5 (Dl = 0.5) is generally adequate to generate these response spectra [22]. Then spectral acceleration is determined
versus spectral displacement curve of ESDOF system according to
the energy approach [28]. In this energy approach, the incremental
displacement of ESDOF system is dened by dividing the incremental work done due to lateral force at step k by the base shear.
In this regard, Eq. (12) through Eq. (14) are utilized for determining
the peak displacement of ESDOF system.

DEk

DDk

DEk

3.2. Determination of target displacement

12

Vb
n
X

j1

It should be noted that for updating the load vectors at each


step, the algorithm proposed by Antonio and Pinho [19] is used.
In this algorithm, considering the stiffness of the structure at the
end of previous step, an Eigenvalue analysis is carried out and periods and mode shapes are calculated. Based on these mode-shapes,
the load vector for each mode of interest is computed according to
Eq. (2) in each step.

657

F i  Ddi

13

i1

k1

Sd S d

DDk

14

where F i is the existing force in the story i at step k; DE(k) the


k
increment of work done by lateral forces; V b , the base shear at step
k
k
k; Ddi , the incremental displacement in the story i at step k and Sd
is the displacement of the ESDOF system at step k.
The spectral acceleration of ESDOF system is computed at each
step k by Eq. (15).
k

Determination of the target displacement is a basic element in


each pushover procedure. The ATC-40 document [3] presents an
equivalent linear method (CSM) in which the capacity of the structure (in the form of a pushover curve) is compared with the demands on the structure (in the form of a response spectrum).
This iterative procedure is limited to the fundamental mode of
vibration [22]. In order to account for the higher modes in the estimation of target displacement, Chopra and Goel [8] employed the
concept CSM procedure and developed modal pushover analysis
(MPA) procedure in which the target displacement of MDOF system is estimated through a series of bilinear equivalent single degree of freedom (ESDOF) systems corresponding to the modes of
interest and consequently, using the roof displacement as a conversion parameter the maximum inelastic displacement of each
ESDOF system transformed back to the roof displacement. The signicant restriction is that the increase of the roof displacement is
not proportional to the other stories in the higher modes and it
is only meaningful for the rst mode. In order to overcome this
limitation, an energy-based method is developed by HernandezMontes et al. [28] in which the incremental displacement of ESDOF
system is dened by dividing the incremental work done due to
lateral force at step k by the base shear.
Kalkan and Kunnath [22] developed an innovative procedure to
determine the target displacement. In their procedure, a capacity
curve for each mode of interest is computed using the energy approach. Afterwards, a series of predetermined ductility-level response spectra, which are in acceleration displacement response
spectrum (ADRS) format (i.e., spectral acceleration versus spectral
displacement), are determined. The intersection of the capacity
curve of each mode at the current step of analysis and the
predetermined inelastic response spectra is considered as dynamic
target displacement if this point represents the same ductility level
of intersected response spectrum curve (Fig. 2). This methodology
is repeated for all modes considered and the target displacement
related to each mode is calculated.

Sk
a

Vb
an W

15

where Sak is the spectra acceleration at step k; W, the total weight;


an is the modal mass coefcient which is obtain by Eq. (6).
Since the load pattern in the proposed method, which is a single
run adaptive pushover, is representative of the contribution of all
modes of interest, this load pattern cannot be attributed to any
particular dynamic mode shape. In other words, none of the natural mode shapes can be visualized in Eqs. (6) and (15) since these
fundamental modes shapes are not compatible with the current
load pattern. In order to solve this problem Casarotti and Pinho
[29] and Shakeri et al. [24] introduced an assumed equivalent fundamental mode shape which is derived from the load pattern at
each step using Eq. (16):

f/gk m1 ff gk

16

where ff gk is the vector of the applied force to the structure and


f/gk is the assumed mode shape at step k.
Therefore using this equivalent mode shape, the spectral acceleration of ESDOF system is computed by Eq. (15).
The obtained capacity curve of ESDOF system in ADRS format is
intersected to the predened ductility level response spectra and
similar to the procedure proposed by Kalkan and Kunnath [22]
the target displacement is determined. After determining the target displacement of ESDOF, it shall be back transformed to the
MDOF using Eq. (17):

ur Ck  /k  Skd

17

where ur is the roof displacement and /k is the assumed mode


shape obtained by Eq. (16).
In the following, a step by step procedure is presented to estimate the seismic demands of a multi-story building as well as
the target displacement, using APAM method:

658

R. Abbasnia et al. / Engineering Structures 52 (2013) 654666

Fig. 2. Determination of the target displacement.

3.3. APAM procedure


k

(1) The modal shapes f/ij g and the natural frequencies fxj g
of the structure at the current state of analysis are computed
for all the considered modes.
(2) The modal story load corresponding to the jth mode of the
structure is calculated, {Dij}, using Eq. (2).
k
(3) The relative mode contribution factors (Rj ) are calculated
based on EMMC rule, using Eqs. (4)(6). It should be mentioned that these coefcients are updated at each step of
analysis.
k
(4) The load patterns fDi g corresponding to all considered
modes are constructed using Eqs. (7)(10).
(5) Nonlinear static analyses (NSA) are independently performed employing the load patterns which are computed
in the previous step. If the rst three modes are considered,
four NSA analyses are required.
k
(6) The spectral displacement of ESDOF system (Sd ) corresponding to the kth step of the analysis are computed, using
Eqs. (12)(14).
(7) The assumed mode shape f/gk and the spectral acceleration
of ESDOF system Sak are computed by Eqs. (16) and (15),
respectively.
(8) If the response is inelastic at kth step of the analysis, the
approximate global system ductility l(k) is calculated, using
Eq. (18):

lk

k
Sd
yield
Sd

18

(9) The response spectra of a given ground motion are extracted


in ADRS format (spectral acceleration, Sa l; f; k, versus spectral displacement, Sd l; f; k) for a series of predened ductility levels.
k
(10) Sak versus Sd (steps 6 and 7) is plotted together with the
inelastic response spectra at different ductility levels (step
ip
9). The target displacement, Sd , is considered as the intersection of the capacity curve of ESDOF system and the
response spectrum corresponding to the global system ductility l(k) obtained from step 8.
Steps 110 are repeated until the target displacement is
calculated with a reasonable approximation. When the target

displacement is determined, the responses of the MDOF structure


are calculated using the assumed equivalent fundamental mode
shape concept (Eq. (17)). As mentioned previously, if the rst three
modes are taken into account, four pushover analyses are independently performed employing the load patterns of Eqs. (7)(10) and
the envelopes of obtained results are considered as seismic demands. It is noteworthy to mention that for estimation of the
capacity curve and the target displacement, only the rst load
pattern (Eq. (7)) is adequate. The reason is discussed later in the
Section 4.3.1.
4. Validation of the proposed method
The proposed procedure is veried for three typical concrete
moment-resisting frames with different heights and twenty strong
ground motion records. The seismic response of the mentioned
structures is predicted by the APAM procedure, the displacement-based adaptive pushover (DAP) method, and also the conventional pushover analysis with triangular and uniform load
pattern. These seismic responses are then compared to benchmark
results obtained from incremental dynamic analysis (IDA). The total drift, inter-story drift and capacity curve are considered to evaluate the accuracy of each procedure in comparison to IDA method.
4.1. Numerical models
In this paper a 3-story building as a low rise, a 9-story building
as a mid rise and a 20-story building as a high rise are selected to
evaluate the accuracy of the proposed procedure. The plans and
elevation views of these buildings are shown in Fig. 3. The total
heights of the structures are 9.6 m, 28.8 m and 64 m for 3, 9, and
20 story buildings, respectively. These buildings are designed for
site class C according to ACI 318-08 [30] and ASCE 7-05[31]. The
specied compressive strength of concrete and the yield stress
for all reinforcement are assumed as 30 MPa and 400 MPa, respectively. The uniform dead and live loads equal to 6 kN/m2 and 2 kN/
m2 are applied to all oors, respectively. For the seismic design of
these buildings, the response modication coefcient for special
reinforced concrete moment frames is 8 according to ASCE 7-05
[28]. Also, the importance factor is considered to be 1. Effective
seismic weight includes total dead load and no contribution of
the live load. Accidental torsion equal to ve percent of dimension

R. Abbasnia et al. / Engineering Structures 52 (2013) 654666

659

Fig. 3. (a) Section properties of beam and columns. (b) Elevation view of the structural frames studied. (c) Structural plan of 3 story building. (d) Structural plans of 9 and 20
story buildings (units in cm).

of the structure perpendicular to the direction of the applied forces


is considered for the seismic design of buildings.
Since the structural systems are regular in plan and elevation
[31], it is permitted to analyze one typical frame (two planer models) in each main direction. In this regard, a typical frame in X
direction is selected. In the numerical analyses, the column bases
are assumed xed. The distributed vertical load on beams is
30 kN/m by considering a tributary width of 5 m. Nonlinear pushover and incremental dynamic analyses are performed using the
Opensees software [32]. The model takes into account geometrical
nonlinearity and material inelasticity. Material inelasticity is
explicitly considered by employing a ber modeling approach.
Beams and columns are modeled as nite elements with distributed inelasticity. The proposed model by Mander et al. [33] is
employed for conned concrete zone to take into account the

conning effect. A bilinear model with kinematic strain hardening


is employed to represent the reinforcing steel bars. Five Gauss integration points are used for each element. The rst three elastic
periods of the structural frames are T = (0.79, 0.2, 0.09) s, (1.6,
0.54, 0.3) s and (2.5, 0.93, 0.54) s for 3-story, 9-story and 20-story
buildings, respectively.
4.2. Ground motion data base
In the present work, a total of twenty strong earthquake ground
motions are compiled in order to develop a reliable set of benchmark responses. The main purpose of selecting this ensemble is
to take into account the variation in term of magnitude, source
to site distance and PGA. These motions were recorded during seismic events with moment magnitude 5.8 < M < 7.6 at closest fault

660

R. Abbasnia et al. / Engineering Structures 52 (2013) 654666

Table 1
Ground motion properties.
No.

Earthquake

Year

Station

Closest distance to the fault (km)

PGA (g)

PGV (cm/s)

Site condition

1
2
3
4
5
6

Cape Mondenico
Imperial Valley
Chi-Chi
Erizincan
Imperial Valley
Chi-Chi

1992
1940
1999
1992
1979
1999

7.1
7
7.6
6.9
6.5
7.6

89156 Petrolia
El Centro
TCU065
95 Erzincan
6621 Chihuahua
NST

9.5
8.3
0.98
2
28.7
36.95

0.662
0.313
0.603
0.515
0.27
0.388

89.7
29.8
78.8
83.9
24.9
26.9

C
C
C
C
C
C

7
8
9
10

Northridge
Cape Mondenico
Loma Prieta
Kocaeli

1994
1992
1989
1999

6.7
7.1
6.9
7.4

90019 San Gabriel


Cape Mendocino
47379 Gilroy Array #1
Izmit

41.7
8.5
11.2
4.8

0.256
1.497
0.473
0.22

9.8
127.4
33.9
29.8

A
A
A
A

11
12
13
14
15

Northridge
Victoria
Trinidad
CHI-CHI
Kocaeli

1994
1980
1980
1999
1999

6.7
6.4
7.2
7.6
7.4

24278 Castaic
6604 Cerro Prieto
1498 Rio Dell Overpass-W Ground
TCU089
Sakarya

22.6

8.22
3.1

0.514
0.587
0.156
0.333
0.376

52.2
19.9
11.2
30.9
79.5

B
B
B
B
B

16
17
18
19
20

Kobe
Loma Prieta
Imperial Valley
Superstitn Hills(B)
Westmorland

1995
1989
1979
1987
1981

6.9
6.9
6.5
6.7
5.8

Kakogawa
58117 Treasure Island
5057 El Centro Array #3
5062 Salton Sea Wildlife Ref.
5062 Salton Sea Wildlife Ref.

26.4
82.9
9.3
27.1

0.251
0.1
0.266
0.167
0.199

18.7
15.6
46.8
18.3
16.4

D
D
D
D
D

Fig. 4. Displacement spectra of scaled strong ground motions.

distance 1 < R < 83 km and belonging to USGS site classication A,


B, C and D. The PGA varies between 0.1 g and 1.5 g. The characteristics of these strong ground motions are summarized in Table 1.
The records are available in the Pacic Earthquake Engineering Research (PEER) site; http://peer.berkeley.edu/smact. These time histories are scaled in such a way that the displacement spectra of
records become identical. In this regard, the time history record,
which has the greatest spectral displacements, is considered as target record and the scale factors of other records are computed by
the following equation:

PT B

actual
T  Stdarg et T
TT A Sd
PT B
actual
T2
TT A Sd

19

where c is the scale factor of time history record; Stdarg et T, the spectral displacement of target record corresponding to period T;

Sactual
T, the spectral displacement of considered record corred
sponding to period T. TA and TB are equal to 20% and 150% of the fundamental period of structure, respectively.
The displacement spectra of scaled time histories are shown in
Fig. 4. The average of these displacement spectra is used in DAP
and APAM pushover analysis to reect the effect of frequency content of records in the response.
4.3. Comparative evaluation
The numerical model of each structural frame is subjected to
the scaled strong ground motions listed in Table 1 and the mean
seismic response of each structural frame is computed. These
mean results which are considered as benchmark response, are
compared with obtained results from different pushover
methods.

R. Abbasnia et al. / Engineering Structures 52 (2013) 654666

Fig. 5. Capacity curves obtain by different pushover analyses in comparison to the dynamic envelops.

Fig. 6. Observed errors in the different pushover methods for the buildings studied.

Fig. 7. The capacity curves obtained by different load patterns for buildings studied.

661

662

R. Abbasnia et al. / Engineering Structures 52 (2013) 654666

Fig. 8. The target displacement determination in the APAM method.


Table 2
Target displacement obtained by APAM method in comparison with the NTH method.
ESDOF target displacements (cm)
NTH
Buildings
3 Story
9 Story
20 Story

Rec.1
20.27
37.91
64.21

Rec.2
21.22
31.07
66.78

Rec.3
17.76
44.48
62.29

Rec.4
19.76
39.67
66.51

Rec.5
18.76
42.24
51.07

Rec.6
16.76
40.69
47.56

Rec.7
17.28
33.76
55.42

Rec.8
20.72
46.94
52.24

Rec.9
17.06
49.62
58.01

Rec.10
19.77
44.95
62.67

Buildings
3 Story
9 Story
20 Story

Rec.11
24.37
44.25
65.09

Rec.12
22.86
32.41
63.37

Rec.13
23.85
32.06
66.14

Rec.14
21.39
40.34
64.96

Rec.15
20.89
39.07
47.2

Rec.16
17.32
46.44
57.01

Rec.17
22.01
49.35
64.41

Rec.18
21.78
33.68
66.93

Rec.19
19.61
44.82
69.35

Rec.20
21.71
38.16
60.16

Buildings

APAM

Error (%)

3 Story

18.6

3.30

9 Story

38.9

3.40

20 Story

55.1

3.46

Fig. 9. Mean peak inter-story drift proles resulting from the different NSPs and the IDA analysis for the buildings studied.

4.3.1. Capacity curve


Fig. 5 illustrates the capacity curves of three case studies, which
are obtained by incremental dynamic analysis (IDA) procedure and
different pushover methods. In this study, capacity curve consists
of the maximum total drift versus maximum base shear.
The capacity curve discrepancy factor (CCDF) is used to measure
the accuracy of each pushover method [17]. This parameter denes
the difference between the ordinates of points on the pushover
curve in comparison to the corresponding IDA points. The CCDF
is computed by the following equation.



k
k
n abs V
PUSH  V IDA
1X
CCDF
n k1
V kIDA

20

where V kIDA and V kPUSH are maximum base shears at step k obtained
from IDA method and different pushover procedures, respectively.
It should be mentioned that the CCDF is computed for control
node displacements ranging between zero and 150% of the target
displacement [1].
As observed in Fig. 5, APAM and DAP procedures provide a closer t to the dynamic analyses envelops than those of the conventional methods in all studied buildings. Also, the error of the APAM
method is less than the error of the DAP procedure in all structural
frames (Fig. 6). As anticipated, the triangular load pattern provides
an admirable estimate of capacity curve only for three story building where the effect of higher modes is negligible. The accuracy of
this method is signicantly decreased as the higher mode effects
are increased in 9 and 20 story buildings. The uniform load pattern

R. Abbasnia et al. / Engineering Structures 52 (2013) 654666

663

Fig. 10. Observed mean errors of peak inter-story drift in different NSPs for the buildings studied.

Fig. 11. Mean peak total drift proles resulting from the different NSPs and the IDA analysis for the buildings studied.

Fig. 12. Observed mean errors of total drift in different NSPs for the buildings studied.

could not provide suitable results in all studied buildings except in


20 story building where results are somewhat similar to IDA
outputs.
Another important observation is that load pattern 1 (Eq. (7))
dominate the responses in APAM method for estimation of the
capacity curve in all case studies. As it can be seen from Fig. 7,
the capacity curve due to load pattern 1 lay beyond the other
capacity curves due to load patterns 24, throughout the entire
deformation range. Therefore, in order to determine the capacity
curve of a structure in APAM method, it is only required that the
rst load pattern (Eq. (7)) apply to the structure. It should be noted
that the drop observed in the capacity curve of 20 story building is
due to the signicant contribution of the higher modes in load
combinations 3 and 4 which will be explained in Section 4.3.5.

4.3.2. Target displacement


As mentioned previously, one of the main advantages of the
proposed methodology is determination of the target displacement
during the analysis. Fig. 8 illustrates the target displacements of
ESDOF systems and the pertinent ductility ratios. In order to evaluate the APAM procedure efciency in estimating the target displacement, these target displacements are compared with the
maximum inelastic displacement of ESDOF system obtained from
the nonlinear time history (NTH) analysis. In this regard, the derived capacity curve of ESDOF system is idealized as a bilinear
curve. A NTH analysis was performed for each record using this
bilinear curve and the maximum inelastic displacement is computed as target displacement of ESDOF system. The target displacements, which are obtained by NTH analyses as well as APAM

664

R. Abbasnia et al. / Engineering Structures 52 (2013) 654666

Fig. 13. The variation of relative mode contribution (Rj) of each mode within the analysis.

Fig. 14. (a) The relative mode contribution of each mode corresponding to the peak interstory drift of each story for 20-story building. (b) The interstory drift prole related to
each load combinations of APAM method.

R. Abbasnia et al. / Engineering Structures 52 (2013) 654666

method for each building and each record, are presented in Table 2.
The mean errors of target displacement are calculated by the following equation.

1
Error % 100 
n

s
Xn TDNTH  TDAPAM 2
i1
TDNTH

21

where TDNTH and TDAPAM are target displacements which are computed by NTH analysis and by APAM procedure, respectively.
As it can be seen from Table 2, APAM estimates the target displacement with an admirable accuracy. Therefore, the inter-story
drift and total-drift of structure are evaluated at the target displacement which is computed by APAM procedure.
4.3.3. Inter-story drift
The peak inter-story drift prole for the buildings studied is
presented in Fig. 9. Where it can be seen that APAM produces
structural response which is similar to the IDA results in all buildings. In order to measure the accuracy of different nonlinear static
pushover (NSPs) methods, Eq. (22) is used [24].

1
Error % 100 
n

s
Xn Di-PUSH  Di-IDA 2
i1
Di-IDA

665

a constant value. Fig. 13 illustrates the variation of relative mode


contribution (Rj) for the rst three modes.
As seen from this gure, the relative mode contribution of the
rst mode in the load combinations 1 and 2 is increased, while
in the load combinations 3 and 4, the contribution of the rst mode
is signicantly decreased. Inversely, the contributions of the second and the third modes are increased in load combinations 3
and 4. Also, in Fig. 14 the peak interstory drift proles predicted
by different load combinations of the APAM method for 20-story
building (Fig. 14a) as well as the corresponding relative mode contribution of each mode (Fig. 14b) are depicted. As shown, the load
combinations 1 and 2 provide well estimation in the lower stories.
In these load combinations, the contribution of the rst mode is
about 90% and the higher mode contributions are not signicant
(less than 10%). On the other hand, the load combinations 3 and
4 provide reasonable estimations in the upper stories. In this case,
the contribution of the second mode is increased to about 35%
while the contribution of the rst mode is decreased to about 60%.
These observations are compatible with the relative contribution of each mode during a NTH analysis. The study of the relative
contribution variations of each mode in the NTH analyses revealed
the following phenomena [23]:

22

In this expression Di-PUSH and Di-IDA are inter-story drifts at level i


obtained by pushover analysis and IDA, respectively. The mean observed errors in the different NSPs are revealed in Fig. 10. In the case
of 3-story building, APAM, triangular pattern and DAP procedure
provide well estimation of inter-story drift. This observation is justiable since the effects of higher modes are negligible in this lowrise building and the governing mode is the rst mode. Therefore
the triangular load pattern can predict this parameter accurately.
In the case of 9-story and 20-story buildings the errors of APAM
procedure are signicantly less than the errors of the other procedures. Since the effect of higher modes in the 9-story and particularly, in the 20-story building is signicant, the accuracy of
triangular method is signicantly decreased. On the other hand,
the APAM method takes into account the contribution of higher
modes as well as the sign reversals in the modal load vectors and
results in well estimations of responses. Although DAP method provides well estimation of inter-story drift in comparison to the conventional pushover, this procedure almost underestimates the
inter-story drift in all stories for 9 and 20 story buildings.
4.3.4. Total drift
In Fig. 11 the total drift obtained by NSPs and IDA methods for
studied buildings are depicted. In order to compute the errors of
the different NSPs, the Eq. (22) is used. As presented in Figs. 11
and 12, the accuracy of APAM procedure is much more than the
accuracy of other methods. The triangular pattern provides reasonable responses in all buildings particularly in 3 and 9 story buildings. DAP method underestimates the total drift in all studied
buildings except in 3-story building. It is surprising that the errors
of triangular pattern is less than that of DAP method.
4.3.5. Discussion
As stated in the previous sections, the APAM method provides
admirable estimation of seismic demands for all case studies. The
much more improved predictions by the APAM method in comparison with those of the DAP method, is due to employing the EMMC
rule. In the EMMC rule, not only the relative mode contribution (Rj)
is computed based on the modal mass contribution, but also it is
updated throughout the analysis proportional to the instantaneous
dynamic properties of structures. This implies that Rj for each
mode is continuously altered within the inelastic range rather than

I. Approximately, the peak interstory drifts in the lower stories


occur in the rst inelastic excursion in the system. In this situation the rst mode dominates the responses and the contribution of higher modes are not signicant.
II. The peak interstory drifts in the upper stories occur later and
the effects of higher modes signicantly increase.
Therefore, the EMMC rule can estimate the relative contribution
of each mode in a manner consistent with the NTH analysis.
5. Conclusions
In the present study, an alternative displacement-based adaptive pushover is developed based on the effective modal mass combination rule. In addition to the higher mode effects consideration
and the progressive changes in the dynamic characteristics of the
structures, this procedure utilize an effective modal mass combination rule in order to take into account the sign reversals of the applied load vector in the higher modes. In this regard, a relative
mode contribution factor, which is updated proportional to the
instantaneous dynamic characteristic of structure, is applied to
each modal load vector. The modied modal load vectors are algebraically added and subtracted and consequent load patterns are
independently applied to the structure within an adaptive framework and the envelope of demand values are considered. Also,
the proposed methodology can estimate the target displacement
during the analysis using the energy approach and CSM concept
proposed in ATC40. The accuracy of APAM method is evaluated
through three concrete moment-resisting frames under four
near-fault different ground motions.
The obtained results illustrate that APAM method can capture
the results of IDA analysis with a reasonable accuracy in all case
studies. Comparisons indicate that APAM method is able to reproduce the capacity curve obtained by IDA method with enough
accuracy. Moreover, well estimation of inter-story drift proles, a
critical parameter in seismic evaluation, as well as total drift proles feature the high ability of APAM method to reproduce IDA
envelops.
References
[1] FEMA. Prestandard and commentary for the seismic rehabilitation of buildings,
FEMA 356. Washington (DC): Federal Emergency Management Agency; 2000.

666

R. Abbasnia et al. / Engineering Structures 52 (2013) 654666

[2] CEN. Eurocode 8: design of structures for earthquake resistance, Part 1: PrEN1998-1, Final Draf. Brussels (BEL): European Committee for
Standardization; 2003.
[3] ATC. Seismic evaluation and retrot of concrete buildings. Report ATC-40.
Redwood City (CA): Applied Technology Council; 1996.
[4] Paret TF, Sasaki KK, Elibeck DH, Freeman SA. Approximate inelastic procedures
to identify failure mechanism from higher mode effects. In: Proceeding of the
eleventh world conference on earthquake engineering; 1996.
[5] Krawinkler H, Seneviratna GDPK. Pros and cons of a pushover analysis of
seismic performance evaluation. Eng Struct 1998;20(46):45264.
[6] Kim S, DAmore E. Pushover analysis procedures in earthquake engineering.
Earthq Spectra 1999;15(3):41734.
[7] FEMA. Improvement of nonlinear static seismic analysis procedures, FEMA
440. Washington (DC): Federal Emergency Management Agency; 2005.
[8] Chopra AK, Goel RK. A modal pushover analysis procedure for estimating
seismic demands for buildings. Earthq Eng Struct Dyn 2002;31:56182.
[9] Kim S, Kurama Y. An alternative pushover analysis procedure to estimate
seismic displacement demands. Eng Struct 2008;30:3793807.
[10] Poursha M, Khoshnoudian F, Moghadam AS. A consecutive modal pushover
procedure for estimating the seismic demands of tall buildings. Eng Struct
2009;31:5919.
[11] Reinhorn AM, Inelastic analysis techniques in seismic evaluation. In:
Proceedings of the international workshop on seismic design methodologies
for the next generation of codes; 1997.
[12] Bracci JM, Kunnath SK, Reinhorn AM. Seismic performance and retrot
evaluation of reinforced concrete structures. J Struct Eng, ASCE
1997;123(1):310.
[13] Gupta B, Kunnath SK. Adaptive spectra-based pushover procedure for seismic
evaluation of structures. Earthq Spectra 2000;16(2):36791.
[14] Elnashai AS. Advanced inelastic static (pushover) analysis for earthquake
applications. Struct Eng Mech 2001;12(1):5169.
[15] Chopra AK, Goel RK, Chinatanapakdee C. Evaluation of a modied MPA
procedure assuming higher modes as elastic to estimate seismic demands.
Earthq Spectra 2004;20(3):75778.
[16] Antoniou S, Pinho R. Advantages and limitations of adaptive and non-adaptive
force-based pushover procedures. J Earthq Eng 2004;8(4):497522.
[17] Papanikolaou VK, Elnashai AS, Pareja JF. Evaluation of conventional and
adaptive pushover analysis II: comparative results. J Earthq Eng
2006;10(1):12751.
[18] Lopez-Menjivar MA, Pinho R. A review of existing pushover methods for 2-D
reinforced concrete buildings. Pavia (Italy): Rose School; 2004.

[19] Antonio S, Pinho R. Development and verication of a displacement-based


adaptive pushover procedure. J Earthq Eng 2004;8(5):64361.
[20] Pinho R, Antoniou S, Peierta D. A displacement-based adaptive pushover
for seismic assessment of steel and reinforced concrete buildings. In:
Proceedings of the 8th US national conference on, earthquake engineering;
2006.
[21] Ferracuti B, Pinho R, Savoia M, Francia R. Verication of displacement-based
adaptive pushover through multi-ground motion incremental dynamic
analyses. Eng Struct 2009;31:178999.
[22] Kalkan E, Kunnath SK. Adaptive modal combination procedure for nonlinear
static
analysis
of
building
structures.
J
Struct
Eng,
ASCE
2006;132(11):172131.
[23] Kalkan E, Kunnath SK. Assessment of current nonlinear static procedures for
seismic evaluation of buildings. Eng Struct 2007;29(3):30516.
[24] Shakeri K, Shayanfar MA, Kabeyasawa T. A story shear-based adaptive
pushover procedure for estimating seismic demands of buildings. Eng Struct
2010;32:17483.
[25] Matsumori T, Otani S, Shiohara H, Kabeyasawa T. Earthquake member
deformation demands in reinforced concrete frame structures. In:
Proceeding of the USJapan workshop on performance-based earthquake
engineering methodology for RC building structure; 1999.
[26] Kunnath SK. Identication of modal combination for nonlinear static analysis
of building structures. Comput-Aided Civ Eng 2004;19:24659.
[27] Park HG, Eom T, Lee H. Factored modal combination for evaluation of
earthquake load proles. J Struct Eng, ASCE 2007;133(7):95668.
[28] Hernandez-Montes E, Kwon O-S, Aschheim MA. An energy based formulation
for rst and multiple-mode nonlinear static Pushover analyses. J Earthq Eng
2004;8:6988.
[29] Casarotti C, Pinho R. An adaptive capacity spectrum method for assessment of
bridges subjected to earthquake action. Bull Earthq Eng 2007;5(3):37790.
[30] ACI. Building code Requirements for structural concrete with commentary. ACI
318. American Concrete Institute; 2008.
[31] American Society of Civil Engineers (ASCE). Minimum design loads for
buildings and other structures. ASCE Standard No. ASCE/SEI 7-05; 2006.
[32] OpenSees Development Team (Open Source Project). OpenSees: Open system
for earthquake engineering simulation. Berkeley (CA): Pacic Earthquake
Engineering Research Center, University of California; 2008. <http://
opensees.berkeley.edu/>.
[33] Mander JB, Priestley MJN, Park R. Theoretical stressstrain model for conned
concrete. J Struct Eng 1988;114(8):180426.

You might also like