Evaluation of the MPA Procedure

for Estimating Seismic Demands:

RC-SMRF Buildings
Hugo Bobadillaa) and Anil K. Chopra,b) M.EERI

The modal pushover analysis (MPA) procedure is extended for analysis of

reinforced concrete special moment resisting frame (RC-SMRF) buildings,
after demonstrating that the theory, assumptions, and approximations
underlying this procedure are valid for such systems. The principal extension
of the procedure is in the hysteretic model for modal SDF systems, chosen as
the peak-oriented model to represent the global monotonic and cyclic behavior
of such buildings, characterized by deterioration of stiffness and strength
under cyclic deformation. The median seismic demands for 4-, 8-, 12-, and
20-story RC-SMRF buildings—designed to comply with current codes—due
to an ensemble of 78 ground motions scaled to four intensity levels were
computed by MPA and nonlinear RHA, and compared. It is demonstrated that,
even for the most intense ground motions that deform the buildings far into the
inelastic range, the MPA procedure demonstrates an adequate degree of
accuracy that should make it useful for practical application in estimating
seismic demands for RC-SMRF buildings. In contrast the FEMA-356 force
distributions are inadequate in estimating seismic demands for the 8-, 12-, and
20-story buildings at all excitation intensities, from the weakest that causes
response essentially within the linearly elastic range, to the strongest that
drives the buildings far into the inelastic range. 关DOI: 10.1193/1.2945295兴

INTRODUCTION
According to the nonlinear static procedure (NSP)—also known as pushover
analysis—described in FEMA-356 and ATC-40 (FEMA 2000, ATC 1996) guidelines for
seismic evaluation of existing buildings, seismic demands may be computed by nonlin-
ear static analysis of the structure subjected to monotonically increasing lateral forces
with a specified, usually invariant, heightwise distribution until a predetermined target
displacement is reached. Also known as pushover analysis, these procedures are now
standard in structural engineering practice. They provide a better assessment of the ac-
tual capacity and expected performance of the structure than traditional linear static
analysis, and are attractive because they require much less computational effort com-
pared to rigorous nonlinear response history analysis (RHA).

a)
Graduate student, Dept. of Civil & Environmental Engineering, University of California, Berkeley, CA 94720
b)
Johnson Professor of Engineering, Dept. of Civil & Environmental Engineering, University of California,
Berkeley, CA 94720

827
Earthquake Spectra, Volume 24, No. 4, pages 827–845, November 2008; © 2008, Earthquake Engineering Research Institute
828 H. BOBADILLA AND A. K. CHOPRA

In recent years, several researchers have demonstrated the limitations of the NSP
(Krawinkler and Seneviratna 1998, Naeim and Lobo 1998, Gupta and Krawinkler 1999,
Kim and D’Amore 1999, Elnashai 2001, Fajfar 2002). In particular, seismic demands
(story drifts and plastic rotations) are underestimated in the upper stories of buildings
where higher-“mode” contributions to response can be significant.
Developing improved NSPs has also been the subject of much research (Bracci et al.
1997, Sasaki et al. 1998, Matsumori et al. 1999, Gupta and Kunnath 2000, Kunnath and
Gupta 2000, Elnashai 2001, Aydinoglu 2003). One such procedure is the modal push-
over analysis (MPA) (Chopra and Goel 2002). Based on structural dynamics theory,
MPA has been shown to achieve superior estimates of seismic demands for buildings
while retaining the conceptual simplicity and computational attractiveness of standard
NSPs. Analyses of several steel moment resisting frame buildings covering a range of
heights and a range of ground-motion intensities have demonstrated that the MPA pro-
cedure estimates the seismic demands for such buildings responding into the inelastic
range to a degree of accuracy that is comparable—only slightly worse—to that of the
standard response spectrum analysis (RSA) procedure for linearly elastic systems
(Chopra 2007, Section 19.8.3).
One of the principal objectives of this investigation is to evaluate the accuracy of the
MPA procedure for a different class of buildings: reinforced concrete special moment
resisting frame (RC-SMRF) buildings. Although such structural systems are designed to
have ductile behavior during earthquake excitations, they are characterized by deterio-
ration of strength and stiffness under cyclic deformations (Haselton and Deierlein 2007).
The second objective of this study is to comparatively evaluate the accuracy of MPA and
the FEMA-356 NSP in estimating seismic demands for RC-SMRF buildings. This paper
summarizes the results of a comprehensive investigation reported in Bobadilla and
Chopra (2007).

STRUCTURAL SYSTEMS AND MODELING ASSUMPTIONS

The structural systems considered in this study are 4-, 8-, 12-, and 20-story RC-
SMRF buildings. These structures were designed as part of a broader study of collapse
assessment of buildings (Haselton and Deierlein 2007), for the ATC-63 Project, “Quan-
tification of building system performance and response modification parameters” (ATC
2007).
The building plan and the elevation of a lateral resisting frame for the 4-story build-
ing are shown in Figure 1. The other three buildings have the same plan but they are
taller; for all four buildings, the first story is 15 ft-0 in. tall, while all the other stories
are 13 ft-0 in. high.
The selected building site represents a highly seismic region in the Los Angeles met-
ropolitan area, corresponding to the design category D of the IBC-2003 Code for soil
class Sd. The seismic loads were determined for the MCE design spectrum specified in
the IBC-2003 code, using a response modification factor of 8, permitted in ASCE
7-2002 code for RC-SMRF buildings. Gravity loads were determined according to the
ASCE 7-2002 code for office use: dead load= 175 psf, and live load= 50 psf.
EVALUATION OF THE MPA PROCEDURE FOR ESTIMATING SEISMIC DEMANDS: RC-SMRF BUILDINGS 829

Figure 1. Elevation and Plan of 4-story RC-SMRF building (adapted from Haselton and Deier-
lein 2007).

The reinforced concrete elements that are part of the moment resisting frames were
detailed according to Chapter 21 of the ACI 318-2002 code, incorporating several fea-
tures, including: capacity design of beams and columns to ensure a ductile response of
plastic hinge zones; and strong column-weak beam design. The static overstrength fac-
tor, defined as the ultimate base shear determined from the first-mode pushover curve
divided by the design base shear, varied from 1.65 to 1.87 for these four buildings.
The structural models selected for this investigation were those created by Haselton
and Deierlein (2007), with the objective of evaluating the seismic performance of build-
ings for a wide range of ground-motion intensities, responding from within the elastic
range to collapse. These structural systems were modeled in the OpenSees computer
program. Their fundamental vibration periods were 1.09, 1.67, 1.96, and 2.56 seconds
for the 4-, 8-, 12-, and 20-story buildings, respectively. Inelastic behavior of beams and
columns occurs at plastic-hinge zones located at the end of each element.
The peak-oriented model developed by Ibarra and Krawinkler (2005a) was selected
to represent the hysteretic behavior of the plastic hinge. The parameters of the peak-
oriented model were calibrated against experimental data for ductile RC elements
(Bobadilla and Chopra 2007).
830 H. BOBADILLA AND A. K. CHOPRA

GROUND MOTIONS

SELECTED EARTHQUAKE RECORDS

A total of 39 ground acceleration records from 14 different earthquakes with mag-
nitudes ranging from 6.5 to 7.6 were selected according to the following criteria (Hasel-
ton and Deierlein 2007):
• Fault type: strike-slip and thrust faults, consistent with earthquake mechanisms
present in California.
• Site class: hard rock to stiff soil, i.e., NEHRP site classes A to D (selected
records are all from site classes C or D).
• Peak ground acceleration, PGA ⬎ 0.2 g, and peak ground velocity, PGV
⬎15 cm/ sec.
• Average shear-wave velocity in upper 30 m of soil, Vs-30 ⬎ 180 cm/ sec.
• Lowest usable frequency ⬎ 0.25 Hz, to ensure that the low frequency content
was not removed by the ground-motion filtering process.
• Limit of six records from a single seismic event.
• No consideration of response spectral shape.
Available in Bobadilla and Chopra (2007) is the list of the selected records and their
relevant data. Each of the 39 records includes two orthogonal components of horizontal
ground motion, leading to a total of 78 ground motions.

GROUND-MOTION SCALING PROCEDURE

All records were scaled to represent the same seismic hazard defined by A共T1兲, the
pseudo-acceleration at the fundamental vibration period T1 of the structure. Both com-
ponents of a record were scaled by the same factor selected to match their geometric
mean, defined as A共T1兲gm = 冑A共T1兲comp1 ⫻ A共T1兲comp2, where A共T1兲comp1 and A共T1兲comp2
are the A共T1兲 values for the two horizontal components of the record to the selected seis-
mic hazard.
Table 1 lists the values of A共T1兲 selected to define ground motion ensembles for four

Table 1. Selected values of A共T1兲 corresponding to four ground-motion

intensities

Building Intensity 1 (g) Intensity 2 (g) Intensity 3 (g) Intensity 4 (g)

4-story 0.05 0.25 0.45 0.765

8-story 0.05 0.25 0.40 0.565
12-story 0.05 0.15 0.30 0.465
20-story 0.05 0.15 0.25 0.355

Note: Intensity 4 corresponds to the seismic hazard spectrum for 2% probability of ex-
ceedance in 50 years.
EVALUATION OF THE MPA PROCEDURE FOR ESTIMATING SEISMIC DEMANDS: RC-SMRF BUILDINGS 831

Figure 2. Seismic hazard spectrum for building site corresponding to 2% probability of ex-
ceedance in 50 years (solid line), and the median response spectrum of 78 scaled ground mo-
tions (dashed line): 4-, 8-, 12-, and 20-story buildings.

different intensities; the highest intensity chosen is A共T1兲2%/50, corresponding to the seis-
mic hazard spectrum with 2% probability of exceedance in 50 years (return period of
2475 years) for the building site.
Figure 2 shows the median response spectrum for the ensemble of 78 ground mo-
tions scaled to match A共T1兲2%/50 for each of the four buildings, and the seismic hazard
spectrum corresponding to 2% probability of exceedance in 50 years. As imposed by the
scaling criterion, the median pseudo-acceleration of the ensemble at the fundamental pe-
riod is matched to the seismic hazard spectrum; because T1 differs for each structure,
the scaling factors for ground motions and hence the median spectrum varies with the
building.

MODAL PUSHOVER ANALYSIS FUNDAMENTALS

The equilibrium equations governing the lateral displacements u of the N floors of a
symmetric-plan building due to horizontal ground acceleration üg共t兲 along one axis of
symmetry are

mü + cu̇ + fS共u,u̇兲 = − m␫üg共t兲 共1兲

where m and c are the mass and damping matrices, ␫ is the influence vector, and fS共u , u̇兲
describes the inelastic lateral force-deformation relation including P-⌬ effects. Although
modal analysis theory is strictly not valid for inelastic systems, it has been demonstrated
832 H. BOBADILLA AND A. K. CHOPRA

that elastic “modes” are coupled only weakly in the response of inelastic RC-SMRF sys-
tems (Bobadilla and Chopra 2007), thus validating the assumption underlying the MPA
procedure.
In this procedure, the effective earthquake forces given by,

peff共t兲 = − m␫üg共t兲 共2兲

are expanded into their modal components. The spatial distribution of these forces,
defined by the vector s = m␫, can be expanded as a summation of modal inertia force
distributions (Chopra 2007)
N
s= 兺 sn s n = ⌫ nm ␾ n 共3兲
n=1

where ␾n is the nth natural vibration “mode” of the system vibrating in its linear range
and ⌫n = ␾Tn m␫ / ␾Tn m␾n. Thus,

peff,n共t兲 = − snüg共t兲 共4兲

is the nth-mode component of effective earthquake forces.
In the MPA procedure, the peak response of the building to peff,n共t兲—or the peak
“modal” demand rn—is determined by a nonlinear static or pushover analysis using the
modal force distribution s*n = m␾n (based on the second part of Equation 3) at the peak
roof displacement urn associated with the nth-“mode” inelastic SDF system. The peak
modal demands rn are then combined by an appropriate modal combination rule to es-
timate total demand. This procedure is directly applicable to the estimation of deforma-
tion demands (e.g., floor displacements, story drifts, and plastic hinge rotations). The
procedure has been described in a convenient step-by-step form (Chopra 2007, Section
19.7.3).
Based on structural dynamics theory, the MPA procedure is computationally attrac-
tive because it avoids nonlinear RHA of the structure. Instead, computing each modal
demand rn requires one nonlinear static analysis of the structure and a nonlinear RHA of
a “modal” SDF system; and modal demands need to be determined only for the first few
(generally two or three) “modes” of the structure. A practical alternative to nonlinear
RHA of the “modal” SDF system is to estimate its deformation using empirical equa-
tions (Chopra and Chintanapakdee 2004).
The MPA procedure developed earlier for steel MRF buildings (Chopra and Goel
2002, Goel and Chopra 2004) has been extended here for RC-SMRF buildings that ex-
hibit deterioration in their stiffness and strength during cyclic deformations (Bobadilla
and Chopra 2007). The principal extension is in the hysteretic model for the nth-“mode”
inelastic SDF system now chosen to represent the global behavior of RC-SMRF build-
ings. The parameters of this hysteretic model, chosen herein as the peak-oriented model,
are determined by the procedure described next.
EVALUATION OF THE MPA PROCEDURE FOR ESTIMATING SEISMIC DEMANDS: RC-SMRF BUILDINGS 833

Figure 3. Modal pushover curves for the 20-story building: (a) first, second, and third “mode”
pushover curves; and (b) pushover curve (dashed line) and cyclic pushover curve (solid line) for
the first “mode.”

FORCE-DEFORMATION RELATION FOR MODAL SDF SYSTEMS

MODAL PUSHOVER CURVES

The force-deformation relation for inelastic modal SDF systems is determined from
the nth-“mode” pushover curve for the lateral force distribution s*n = m␾n. Modal push-
over curves for the first three “modes” of the 20-story building are presented in Figure
3a; for other buildings, see Bobadilla and Chopra (2007). The post-yield branch of the
first-“mode” pushover curve exhibits negative slope because of P-⌬ effects.
The global hysteretic behavior of the frames is described by the cyclic pushover
curves shown in Figure 3b. These were determined by nonlinear static analysis of the
frame subjected to the modal force distribution s*n with its magnitude varied and reversed
to cause cyclic roof displacement with gradually increasing amplitude. Comparison of
the cyclic and monotonic pushover curve in Figure 3b indicates cycle-to-cycle deterio-
ration of strength and stiffness in the RC-SMRF building.

MODAL SDF-SYSTEM PARAMETERS

The force-deformation relation for the modal SDF system is determined from the
base shear versus the roof displacement relation defined by modal pushover curves using
equations in Section 19.7.2 of Chopra (2007). The resulting force-deformation relation
for the first-“mode” SDF system, determined from the monotonic pushover curves (Fig-
ure 3a), is shown in Figure 4a.
The hysteretic force-deformation relation is idealized by the peak-oriented model
(Ibarra and Krawinkler 2005a, Ibarra et al. 2005b) with the monotonic curve idealized as
trilinear (Figure 4a), with its parameter values available in Bobadilla and Chopra (2007).
In the peak-oriented model, the cyclic behavior is described by a series of deterioration
rules. These deterioration parameters are determined from the cyclic force deformation
relation for the modal SDF system obtained from the cyclic pushover curve (Figure 4b).
Determined by an iterative trial-and-error procedure to obtain a best fit of the model to
834 H. BOBADILLA AND A. K. CHOPRA

Figure 4. First-“mode” force-deformation relation for the 20-story building: (a) pushover curve
(solid line) and its trilinear model (dashed line); and (b) cyclic pushover curve (solid line) and
its hysteretic model (dashed line).

the actual cyclic curve, these parameters are listed in Bobadilla and Chopra (2007) for
the first three “modes” of each building. The associated hysteretic model provides an
excellent representation of the global cyclic behavior of the building, as shown in
Figure 4b.

EVALUATION OF MPA PROCEDURE1

MODAL PUSHOVER CURVES AND ROOF DISPLACEMENTS

Figure 5 shows the pushover curves for the 20-story building associated with its first,
second, and third “modes,” respectively, and identifies the peak modal roof displacement
due to each of the 78 ground motions corresponding to the 2% in 50 years seismic haz-
ard, and their median value; results for other buildings are available in Bobadilla and

Figure 5. First-, second-, and third-“mode” pushover curves for the 20-story building. The roof
displacements due to 78 ground motions scaled to A共T1兲2%/50 yrs are identified and the median
values are noted.

1
For brevity, some of the figures could be presented only for one of the buildings; however, the observations and
discussion in the rest of the paper are based on results for all buildings in Bobadilla and Chopra (2007).
EVALUATION OF THE MPA PROCEDURE FOR ESTIMATING SEISMIC DEMANDS: RC-SMRF BUILDINGS 835

Chopra (2007). Excluded from the first-“mode” plot are roof displacements due to those
ground motions that caused collapse of the SDF-system: 12, 29, 24, and 13 excitations
in case of 4-, 8-, 12-, and 20-story buildings, respectively. Note that the data points in
Figure 5 for individual ground motions are not all located on the monotonic pushover
curve because of the strength deterioration in RC structures undergoing cyclic deforma-
tion (see Figure 3b).
The very intense ground motions drive all four buildings far beyond the yield dis-
placement in the first “mode,” and the median displacement exceeds the yield displace-
ment by a factor of 3.2 to 5.7. Almost all of the excitations drive the 8-, 12-, and 20-
story buildings well into the inelastic range in the second-“mode,” with the median
displacement equal to 2.7 to 3.2 times the yield displacement; however, only half of the
excitations drive the 4-story building into the inelastic range, and the median displace-
ment is 1.4 times the yield displacement. Almost half of the ground motions drive the 8-,
12-, and 20-story buildings beyond the elastic limit in the third “mode,” but the median
displacement is only slightly beyond the yield displacement. The 4-story building re-
mains essentially elastic in its third “mode” during almost all the ground motions.
The median displacements of SAC steel MRF buildings subjected to SAC ground
motions in “modes” higher than the first were either close to or exceeded the yield dis-
placement only by a modest amount (Goel and Chopra 2004). Such is not the case for
the second “mode” of the selected RC-SMRF buildings subjected to the 2% in 50 years
ground motions, indicating that the scaled set of ground motions drive these buildings
much farther into the inelastic range compared to SAC buildings studied earlier. Thus,
the selected combination of RC-SMRF buildings and ground motions represents an ex-
treme test for the MPA procedure.

HIGHER MODE CONTRIBUTIONS IN SEISMIC DEMANDS

Figure 6 shows the median values of story drift demands, including a variable num-
ber of “modes” in MPA superimposed with the “exact” result from nonlinear RHA, for
each building subjected to the four different levels of ground-motion intensity. The first
“mode” alone is adequate in estimating floor displacements; including higher “modes”
does not significantly improve the estimate (Bobadilla and Chopra 2007). Although the
first “mode” alone is inadequate in estimating story drifts, with a few “modes” included,
story drifts estimated by MPA are much better, especially in the upper stories, and re-
semble nonlinear RHA results over the entire range of building heights and excitation
intensities considered (Figure 6). Notable discrepancies remain for some of the taller
buildings or some of the higher-intensity ground motions, however. These discrepancies
are discussed next.

ACCURACY OF MODAL PUSHOVER ANALYSIS

The structural dynamics theory underlying the MPA procedure for inelastic systems
is based on two principal approximations: (1) neglecting the weak coupling of “modes”
in computing the peak “modal” response rn to peff,n共t兲; and (2) combining the rn by
modal combination rules, known to be approximate, to compute the peak value of the
836 H. BOBADILLA AND A. K. CHOPRA

Figure 6. Median story drifts for 4-story (row 1), 8-story (row 2), 12-story (row 3), and 20-
story (row 4) buildings determined by nonlinear RHA and MPA, with a variable number of
“modes.” Ground motions are scaled to four different values of A共T1兲.

total response. Because the latter is the only source of approximation in the RSA pro-
cedure, now standard for analysis of linearly elastic systems, the resulting error in the
response of these systems serves as a baseline for evaluating the additional approxima-
tions in MPA for inelastic systems.
EVALUATION OF THE MPA PROCEDURE FOR ESTIMATING SEISMIC DEMANDS: RC-SMRF BUILDINGS 837

Figure 7. Median story drifts due to ground motions scaled to A共T1兲2%/50 for: (a) linearly elas-
tic systems determined by RSA and RHA procedures, and (b) inelastic systems determined by
MPA and nonlinear RHA procedures. Results are for 4-, 8-, 12-, and 20-story buildings.

Figure 7 compares the accuracy of MPA in estimating the response of inelastic sys-
tems with that of RSA in estimating the response of elastic systems. These results were
obtained by including 2, 3, 4, and 5 “modes” in the analyses of 4-, 8-, 12-, and 20-story
buildings, respectively. For each of the four buildings, the results are organized in two
parts: (a) story drift demands for these buildings treated as elastic systems determined
by RSA and RHA procedures; and (b) demands for inelastic systems determined by
MPA and nonlinear RHA. These results are for the most intense ground motions con-
sidered (with 2% probability of exceedance in 50 years) that deform the buildings far
into the inelastic range (Figure 5). Therefore, the results of Figure 7 provide an extreme
test of the accuracy of MPA.
Observe that the RSA procedure underestimates the median response for all four
buildings (except in the lower stories of the 8-story building). This underestimation
tends to increase from the bottom to top of buildings, consistent with the heightwise
variation of contribution of higher “modes” to response (Chopra 2007, Chapter 18). The
heightwise average underestimation is 5%, 6%, 5%, and 11%, and the heightwise largest
underestimation is 15%, 19%, 23%, and 29% for the 4-, 8-, 12-, and 20-story buildings,
respectively. The discrepancy in the RSA procedure tends to increase for taller (or
longer-period) buildings because higher-mode contributions are known to be more sig-
nificant for such buildings (Chopra, 2007, Chapter 18). By pervasive use of commercial
software based on modal combination approximation, the profession tacitly accepts this
approximation. However, it appears that the research community has not recognized
838 H. BOBADILLA AND A. K. CHOPRA

Figure 8. Median story drifts for 20-story building determined by three procedures: (1) non-
linear RHA, (2) four FEMA-356 force distributions (upper boxes), and (3) MPA (lower boxes).
Results are presented for ground-motion ensembles scaled to four different values of A共T1兲.

fully that RSA may lead to such significant underestimation of response, especially for
taller buildings, or effectively communicated this discrepancy to the profession.
The additional errors introduced by neglecting modal coupling in the MPA proce-
dure, which are apparent by comparing parts (a) and (b) of Figure 7, are significant.
Having said that, even for the most intense ground motions that deform the buildings far
into the inelastic range, the MPA procedure offers an adequate degree of accuracy that
should make it useful for practical application in estimating seismic demands for build-
ings. The accuracy tends to improve as the intensity of the ground motion decreases
(Figure 8).

COMPARATIVE EVALUATION OF FEMA-356 AND MPA PROCEDURES

Commonly used for seismic assessment of existing buildings, the NSP in FEMA-
356 (FEMA 2000) requires development of a pushover curve by nonlinear static analysis
of the structure, subjected first to gravity loads, followed by monotonically increasing
lateral forces with a specified, invariant heightwise distribution. At least two force dis-
tributions must be considered. The first is selected from among the following: first-
“mode” distribution, equivalent lateral force (ELF) distribution, and RSA distribution;
the second distribution is either the “uniform” distribution or an adaptive distribution;
several options are allowed for the latter, which varies with change of deflected shape of
the structure. This study evaluates the first four lateral force distributions, which are de-
scribed in Section 22.4 of Chopra (2007).
EVALUATION OF THE MPA PROCEDURE FOR ESTIMATING SEISMIC DEMANDS: RC-SMRF BUILDINGS 839

Figure 8 shows the median story drift demands for the 20-story building due to
ground-motion ensembles scaled to the four intensity levels mentioned earlier; similar
results for other buildings are available in Bobadilla and Chopra (2007). The target dis-
placement for FEMA analysis was not determined by the empirical equations in FEMA-
356, but was taken equal to the MPA value to ensure a meaningful comparison of the
two sets of results. In the upper part of the figure, the FEMA-356 estimate of story drifts
are compared with the “exact” value determined by nonlinear RHA. In the lower part,
the MPA estimate (including all significant “modes”) of seismic demands is compared
with the “exact” value. It is obvious by comparing the two parts of this figure and of
other figures in Bobadilla and Chopra (2007) that MPA provides much superior results
for the 8-, 12-, and 20-story buildings for the entire range of excitation intensities. For
the 4-story building, MPA results are similar to the FEMA estimates.
FEMA force distributions underestimate the story drifts, especially in the upper sto-
ries, due to low-intensity ground motions—A共T1兲 = 0.05 g—that produce response
within the elastic range. Although the ELF and RSA distributions are intended to ac-
count for higher-“mode” responses, they do not provide satisfactory estimates of seismic
response even for buildings responding within their elastic range. The MPA procedure
estimates seismic demands much better than do FEMA force distributions.
For higher ground-motion intensities, FEMA force distributions generally underes-
timate story drifts in upper stories and overestimate them in lower stories, especially the
“uniform” distribution. The other force distributions provide story drifts similar to those
due to the first-“mode” force distribution, although the ELF and RSA force distributions
are intended to account for higher-“mode” response. In contrast, for all excitation inten-
sities, the MPA procedure provides a much better estimate of story drift demands in the
upper stories of the 8-, 12-, and 20-story buildings, because it includes higher-“mode”
contributions to the response; these higher-“mode” contributions are especially notice-
able for the 20-story building. In the lower stories, the MPA estimate is slightly better
than the FEMA-356 estimates. Because the response of the 4-story building is domi-
nated by the first-“mode,” the FEMA-356 force distributions are adequate and MPA does
not offer improvement in the demand estimate.
In summary, even for the most intense excitations, the MPA procedure estimates
seismic demands to a degree of accuracy useful for practical application in seismic
evaluation of buildings (see Figure 9). In contrast, the FEMA-356 force distributions are
inadequate in estimating seismic demands for the 8-, 12-, and 20-story buildings at all
excitation intensities, from the weakest to the strongest.

SDF-SYSTEM ESTIMATE OF ROOF DISPLACEMENT

How well can the roof displacement ur of a multistory building be determined from
the deformation of an SDF system, which is the concept underlying the current NSP? To
address this question, we compare the values of roof displacement determined by two
methods: the “exact” value 共ur兲MDF determined by nonlinear RHA of the building treated
as an MDF system; and the estimate based on the first-“mode” SDF system,
共ur兲SDF = ⌫1␾r1D1. The displacement ratio was then defined as 共ur*兲SDF = 共ur兲SDF / 共ur兲MDF.
840 H. BOBADILLA AND A. K. CHOPRA

Figure 9. Median story drifts for 4-, 8-, 12-, and 20-story buildings due to ground motions
scaled to A共T1兲2%/50 determined by three procedures: (1) nonlinear RHA, (2) four FEMA-356
force distributions (upper boxes), and (3) MPA (lower boxes).

The difference between the median value of the displacement ratio and unity indicates
the bias in the SDF-system estimate of median roof displacement.
The SDF system estimates the median roof displacement of multistory buildings to a
useful degree of accuracy. Figure 10 shows histograms of the 78 values of the displace-
ment ratio together with its range of values and the median value for the 20-story build-
ing; results are shown for an ensemble of ground motions scaled to match the four val-
ues of spectral acceleration A共T1兲 mentioned earlier; results for other buildings are
available in Bobadilla and Chopra (2007). The bias in the SDF-system estimate of me-
dian roof displacement is generally larger for the more intense ground motions that drive
the structure farther into the inelastic range. For example, the SDF system estimates the
median roof displacement of the 8- and 12—story buildings within 3% for the lower
three excitation intensities, but overestimates it by 10% and 16%, respectively, for the
most intense motions. The bias increases significantly for the 20-story building; it is
21% for the most intense motion and 15% for the second highest intensity. The histo-
grams indicate that increase in ground-motion intensity results, generally, in larger dis-
persion in the 共ur*兲SDF and a wider range of values.
Figure 11 shows histograms of the displacement ratio for all the four buildings sub-
jected to ground motions scaled to match A共T1兲 for the 2% probability of exceedance in
50 years. The bias in the SDF-system estimate of median roof displacement depends on
the vibration properties (or height) of the building. It is generally smaller for short-
EVALUATION OF THE MPA PROCEDURE FOR ESTIMATING SEISMIC DEMANDS: RC-SMRF BUILDINGS 841

Figure 10. Histograms of ratio 共ur*兲SDF for 20-story building subjected to 78 ground motions
scaled to four different values of A共T1兲. The range of values and the median value of this ratio
are noted.

period (or shorter) buildings because their higher “modes” contribute little to their re-
sponse, and increases for long-period (or taller) buildings because they respond signifi-
cantly in higher “modes” of vibration.
The SDF system may not estimate to a useful degree of accuracy the roof displace-
ment of a building due to an individual excitation. For the 4-story building, this SDF-
system estimate varies from 58% to 174% of the exact value, a surprisingly large dis-
crepancy for a first-“mode” dominated structure. The SDF-system estimate can be
alarmingly small (as low as 58% to 74% of the exact value among the four buildings) or
unexpectedly large (as large as 161% to 182% of the exact value among the four build-
ings). The errors are actually worse than indicated by Figure 11 because it does not in-
clude those cases where nonlinear RHA predicted collapse of the first-“mode” SDF
system.2 This large discrepancy arises because for individual ground motions the SDF
system may significantly underestimate or overestimate the yielding-induced permanent
drift in the response of the building; an assertion supported by response results (Boba-
dilla and Chopra 2007).

2
Data for records that caused collapse of the SDF system are excluded, reducing the number of data to 66 for
the 4-story building, 49 for the 8-story building, 54 for the 12-story building, and 65 for the 20-story building.
842 H. BOBADILLA AND A. K. CHOPRA

Figure 11. Histograms of ratio 共ur*兲SDF for 4-, 8-, 12-, and 20-story buildings subjected to 78
ground motions scaled to A共T1兲2%/50 yrs. The range of values and the median value of this ratio
are noted.

CONCLUSIONS
Recognizing that nonlinear RHA of RC-SMRF buildings for a large ensemble of
ground motions is computationally demanding, much of this work is focused on extend-
ing the MPA procedure to such buildings and evaluating the accuracy of this approxi-
mate procedure.
The median seismic demands for 4-, 8-, 12-, and 20-story RC-SMRF buildings de-
signed according to current building codes—due to an ensemble of 78 ground
motions—were computed by MPA and nonlinear RHA procedures and compared. These
ground motions were scaled to four different intensity levels to evaluate the accuracy of
the MPA procedure over a wide range of building responses from essentially within the
linearly elastic range to far into the inelastic range. The presented results have led to the
following conclusions:
1. The theory, assumptions, and approximations underlying the MPA procedure are
valid for RC-SMRF buildings, characterized by deterioration of stiffness and
strength under cyclic deformation. The principal extension in the procedure is in
the hysteretic model for modal SDF systems, which is now chosen to represent
the global monotonic and cyclic behavior of such buildings.
2. The peak-oriented model (Ibarra and Krawinkler 2005a) provides an excellent
representation of the monotonic and cyclic modal pushover curves for RC-
SMRF buildings. The force-deformation relation for the model is determined by
EVALUATION OF THE MPA PROCEDURE FOR ESTIMATING SEISMIC DEMANDS: RC-SMRF BUILDINGS 843

idealizing the monotonic pushover curve and the deterioration parameters for
the model are determined from the cyclic modal pushover curves.
3. The first-“mode” SDF system estimates the median roof displacement to a use-
ful degree of accuracy. The bias in this estimate is generally larger for the more
intense ground motions that drive the structures farther into the inelastic range.
The bias depends on the vibration properties (or height) of the building; it is
generally smaller for short-period (or shorter) buildings and increases for
longer-period (or taller) buildings because they respond significantly in higher
“modes” of vibration.
4. The first-“mode” SDF system may not estimate to a useful degree of accuracy
the roof displacement of a building due to an individual excitation.
5. With a couple of “modes” higher than the first “mode” included, the heightwise
distribution of story drifts estimated by MPA become generally similar to the
results of nonlinear RHA.
6. The “modal” combination approximation used in the RSA procedure for lin-
early elastic systems, a standard tool in structural engineering practice, may lead
to significant (15% to 29% for the four buildings) underestimation of story drift
demands in the upper stories.
7. Although neglecting modal coupling in MPA introduces additional errors, even
for the most intense ground motions that deform the buildings far into the in-
elastic range, the MPA procedure demonstrates an adequate degree of accuracy
that should make it useful for practical application in estimating seismic de-
mands for RC-SMRF buildings.
A comparison of the seismic demands computed by MPA, FEMA-356 NSP, and non-
linear RHA procedures has determined that even for the most intense excitations, which
represent a very severe test, the MPA procedure estimates seismic demands for RC-
SMRF buildings to a useful degree of accuracy. In contrast, the FEMA-356 force dis-
tributions are inadequate in estimating seismic demands for 8-, 12-, and 20-story build-
ings at all excitation intensities, from the weakest that causes response essentially within
the linearly elastic range, to the strongest that drives the buildings far into the inelastic
range.

ACKNOWLEDGMENTS
The first author would like to acknowledge the fellowship from the Chilean Fulbright
Commission to pursue an M.Eng. degree in Structural Engineering at the University of
California, Berkeley.
We are most grateful to Professor Greg G. Deierlein from Stanford University and
Professor Curt Haselton from California State University at Chico for providing the
structural models and ground motion data that served as the basis for this study.
844 H. BOBADILLA AND A. K. CHOPRA

REFERENCES
Applied Technology Council, 1996. Seismic evaluation and retrofit of concrete buildings. Vol-
ume 1, Report No. ATC-40, Applied Technology Council, Redwood City, CA.
Applied Technology Council, 2007. Quantification of building system performance and re-
sponse modification parameters, ATC-63, Applied Technology Council, Redwood City,
Calif.
Aydinoglu, M. N., 2003. An incremental response spectrum analysis procedure based on inelas-
tic spectral displacements for multi-mode seismic performance evaluation, Bull. Earthquake
Eng. 1, 3–36.
Bobadilla, H., and Chopra, A. K., 2007. Modal pushover analysis for seismic evaluation of re-
inforced concrete special moment resisting frame buildings, EERC Report 2007/01, Earth-
quake Engineering Research Center, University of California, Berkeley, CA.
Bracci, J. M., Kunnath, S. K., and Reinhorn, A. M., 1997. Seismic performance and retrofit
evaluation for reinforced concrete structures, J. Struct. Eng. 123, 3–10.
Chopra, A. K., 2007. Dynamics of Structures: Theory and Applications to Earthquake Engi-
neering, 3rd Edition, Pearson Prentice Hall, Upper Saddle River, NJ, 876 pp.
Chopra, A. K., and Chintanapakdee, C., 2004. Inelastic deformation ratios for design and evalu-
ation of structures: single-degree-of-freedom bilinear systems, J. Struct. Eng. 130, 1309–
1319.
Chopra, A. K., and Goel, R. K., 2002. A modal pushover analysis procedure for estimating seis-
mic demands for buildings, Earthquake Eng. Struct. Dyn. 31, 561–582.
Elnashai, A. S., 2001. Advanced inelastic static (pushover) analysis for earthquake applications,
Struct. Eng. Mech. 12, 51–69.
Fajfar, P., 2002. Structural analysis in earthquake engineering—a breakthrough of simplified
non-linear methods, in Proceedings of the 12th European Conference on Earthquake Engi-
neering, Elsevier Science Ltd., London, United Kingdom.
FEMA, 2000. Pre-standard and Commentary for the Seismic Rehabilitation of Buildings,
FEMA-356, Washington, D.C.
Goel, R. K., and Chopra, A. K., 2004. Evaluation of modal and FEMA pushover analyses: SAC
buildings, Earthquake Spectra 20, 225–254.
Gupta, A., and Krawinkler, H., 1999. Seismic demands for performance evaluation of steel mo-
ment resisting frame structures, Report No. 132, John A. Blume Earthquake Engineering
Center, Stanford University, Stanford, CA.
Gupta, B., and Kunnath, S. K., 2000. Adaptive spectra-based pushover procedure for seismic
evaluation of structures, Earthquake Spectra 16, 367–391.
Haselton, C. B., and Deierlein, G. G., 2007. Assessing seismic collapse safety of modern rein-
forced concrete moment-frame buildings, Report No. 156, John A. Blume Earthquake Engi-
neering Center, Stanford University, Stanford, CA.
Ibarra, L. F., and Krawinkler, H., 2005a. Global collapse of frame structures under seismic ex-
citations, Report No. 152, John A. Blume Earthquake Engineering Center, Stanford Univer-
sity, Stanford, CA.
Ibarra, L. F., Medina, R. A., and Krawinkler, H., 2005b. Hysteretic models that incorporate
strength and stiffness deterioration, Earthquake Eng. Struct. Dyn. 34, 1489–1511.
Kim, B., and D’Amore, E., 1999. Pushover analysis procedure in earthquake engineering,
Earthquake Spectra 15, 417–434.
EVALUATION OF THE MPA PROCEDURE FOR ESTIMATING SEISMIC DEMANDS: RC-SMRF BUILDINGS 845

Krawinkler, H., and Seneviratna, G., 1998. Pros and cons of a pushover analysis of seismic
performance evaluation, Eng. Struct. 20, 452–464.
Kunnath, S. K., and Gupta, B., 2000. Validity of deformation demand estimates using nonlinear
static procedures, in Proceedings of the U.S.-Japan Workshop on Performance-Based Earth-
quake Engineering Methodology for Reinforced Concrete Building Structures, Hokkaido,
Japan.
Matsumori, T., Otani, S., Shiohara, H., and Kabeyasawa, T., 1999. Earthquake member defor-
mation demands in reinforced concrete frame structures, in Proceedings of the U.S.-Japan
Workshop on Performance-Based Earthquake Engineering Methodology for Reinforced Con-
crete Building Structures, Maui, Hawaii.
Naeim, F., and Lobo, R. M., 1998. Common pitfalls in pushover analysis, in Proceedings of the
SEAOC Annual Convention, Reno, Nevada.
Sasaki, K. K., Freeman, S. A., and Paret, T. F., 1998. Multimode pushover procedure
(MMP)—A method to identify the effects of higher “modes” in a pushover analysis, in Pro-
ceedings of the 6th U.S. National Conference of Earthquake Engineering, Seattle,
Washington.
(Received 8 November 2007; accepted 23 April 2008兲