Вы находитесь на странице: 1из 16

Soil Dynamics and Earthquake Engineering 92 (2017) 472487

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering


journal homepage: www.elsevier.com/locate/soildyn

Estimation of oor response spectra using modied modal pushover


analysis
crossmark

Xiaolan Pana, , Zhi Zhenga, Zhenyu Wanga,b
a
School of Civil Engineering, Harbin Institute of Technology, Harbin 150090, China
b
Key Lab of Structures Dynamic Behavior and Control of the Ministry of Education, Harbin Institute of Technology, Harbin, Heilongjiang 150090, China

A R T I C L E I N F O A BS T RAC T

Keywords: A common approach to designing nonstructural components against seismic excitations involves the use of oor
Floor response spectra response spectra (FRS). FRS can be accurately computed only through a nonlinear time-history analysis of the
Nonstructural components structure subjected to a specic earthquake ground motion. However, for multi-storey structures, which are
Modal pushover analysis usually modeled as multi-degree-of-freedom (MDOF) systems, this analysis becomes computation-intensive
Nonlinear dynamic analysis
and time-consuming and is not suitable for adopting in seismic design guidelines. An alternative method of
estimating FRS on MDOF systems is presented here. The proposed method uses multiple generalized or
equivalent single degree of freedom (ESDOF) systems to estimate FRS on a MDOF system within the context of
a modal pushover analysis (MPA). This is a modied version of the previous MPA procedure as it considers the
contribution of the rst mode to yielding of higher modes when obtaining multiple ESDOF systems. FRS values
for each mode are obtained through nonlinear dynamic analysis of each ESDOF system and then the total FRS
values are calculated for the considered modes according to the SRSS combination rule. The eciency of the
modied procedure is tested by comparing FRS based on this method with results from nonlinear dynamic
analyses of MDOF systems, as well as estimates based on ESDOF systems built from the traditional MPA
method, for several ground motion scenarios. Three steel moment frame structures, of 3-, 9-, and 20-storey
congurations, are selected for this comparison. Bias statistics that show the eectiveness of the modied
method are presented.

1. Introduction nonstructural components by directly amplifying the ground response


spectrum [3]. The time-history analysis method is to obtain FRS
Architectural, electrical, and mechanical components may be sub- converted from total acceleration time histories of a specic attachment
jected to signicant dynamic forces when the supporting structure is location of a structure. Currently, there have been a large number of
excited by earthquake ground motions. Observations following recent reported studies on evaluating the seismic performance of nonstruc-
major earthquakes have indicated that a considerable portion of the tural components through time history analysis method. Early in the
total cost of damage is often associated with such nonstructural 1980s, Bumpus et al. [4] and Kennedy et al. [5] excited dierent
components [1,2]. Based on the types of damage and failure mechan- nuclear power plant models using real earthquake records to study the
isms associated with nonstructural components, they are classied as eects of structural nonlinearity on the in-structure response. In 1985,
either displacement-sensitive or acceleration-sensitive. Examples of based on ten ground motion records, Lin and Mahin [6] investigated
acceleration-sensitive nonstructural components include suspended the eects of structural ductility and hysteretic model on FRS. After
ceiling systems, re sprinkler piping, lighting xtures, mechanical that, most studies continued to study the characteristics of component
equipment, etc. Seismic damage to these light components is caused acceleration demands employing real or articial ground motion
by the inertia forces due to oor accelerations that may be larger than records [719]. In order to estimate FRS for the design of nonstruc-
those of the ground. tural components, the application of nonlinear inelastic dynamic time-
A common procedure for the design of these acceleration-sensitive history analysis is suggested, for accuracy reasons. However, one has to
nonstructural components involves the use of oor response spectra apply this detailed method (nonlinear response-history analysis of a
(FRS). FRS can be obtained mainly in two ways: direct method or time- multi-degree-of-freedom (MDOF) model) for each individual structure
history analysis method. The direct method predicts the response of separately, which makes this method unsuitable for incorporating in a


Corresponding author.
E-mail address: xiaolan_pan26@163.com (X. Pan).

http://dx.doi.org/10.1016/j.soildyn.2016.10.024
Received 21 May 2016; Received in revised form 13 October 2016; Accepted 17 October 2016
0267-7261/ 2016 Elsevier Ltd. All rights reserved.
X. Pan et al. Soil Dynamics and Earthquake Engineering 92 (2017) 472487

general purpose design methodology based on FRS. Similar problems ment with 1% strain hardening is used for modeling beams and
in force- or displacement-based design approaches were solved by the columns. The joint panel zones are assumed to be rigid. The stiness
use of response spectra and equivalent single-degree-of-freedom contribution of gravity frames (interior frames) is not considered. The
(ESDOF) models of a structure. Nevertheless, there has not been any gravity loads and P-Delta eects are considered in the following
detailed work trying to obtain ESDOF systems for estimating FRS. analysis. Damping ratios of 5% are assumed for the rst two modes.
Instead of using a single ESDOF system, the modal pushover analysis
(MPA) employs several ESDOF systems for estimating the response of 2.2. The coupling eect on yielding of higher-mode systems
a structure. This is very important to estimate FRS as the amplitude of
the peaks in FRS is strongly dependent on modal periods of the The ESDOF systems proposed in this work are based on the
structure and FRS at higher modal periods are not negligible. nonlinear static pushover analysis of the MDOF model of the structure.
In order to be able to estimate FRS accurately for the case of MDOF The basic formulation of a generalized or equivalent system starts from
systems, a suitable nonlinear equivalent ESDOF system must rstly be the dynamic response of a planar MDOF cantilever-type structure
dened. In the nonlinear inelastic area, contrary to the elastic area, the subjected to horizontal base motion, ug :
exact denition of a nonlinear ESDOF system is impossible. In order to mu + cu + r = mug (1)
dene approximately ESDOF systems, many attempts had been made
by researchers in the past [2026]. Chopra and Goel [27,28] presented where m is the mass matrix, c is the damping matrix, u is the lateral
a complete calculation procedure to establish nonlinear ESDOF displacement vector, is the inuence vector, and r is the restoring
systems and to conduct modal pushover analysis. In order to estimate force vector.
displacement demands more accurately, Makarios [29] further pro- The displacement,u , of a linear N-DOF system can be expressed as a
posed a new denition of nonlinear ESDOF systems through the superposition of modal contributions:
mathematical and iterative procedure. However, using the previous N
ESDOF systems to estimate acceleration demands such as FRS, the u (t ) = n qn (t )
results are not always satisfying since these systems mainly focused on n =1 (2)
total seismic demands and did not take into account the coupling eect where n has the normalized mode shape with the roof degree of
among dierent modes on yielding of the higher-mode system. For freedom rn =1 for all modes and the modal coordinate,qn (t ), is
example, due to the coupling eect between the rst mode and any governed by:
higher mode, the higher-mode system may reach the yielding state
earlier while they still remain elastic in the traditional modal pushover qn + 2n n qn + n2 qn = n ug (t ) (3)
curve. In other words, once the rst-mode system yields, it will in which n , n , and n are the natural vibration frequency, the damping
accelerate yielding of higher-mode systems. This is particularly im- ratio, and the modal participation factor of the nth mode. The n can be
portant in estimating acceleration demands for higher modes and is expressed by:
demonstrated in the following section.
The present article proposes a multiple equivalent systems-based n = Tn m /nT mn (4)
approach for estimating FRS on a structure. A modied MPA proce-
The qn can be represented by:
dure is adopted to determine modied ESDOF systems through
studying the coupling eect on yielding of higher-mode systems. qn (t ) = n Dn (t ) (5)
After obtaining modied ESDOF systems, FRS values for each mode where Dn (t ) is governed by the equation of motion for an ESDOF
are computed through nonlinear dynamic analysis of each ESDOF system, with natural frequency, n , and damping ratio, n , subjected to
system and then the total FRS values are obtained for all considered ug (t ):
modes according to the SRSS combination rule. For comparison, FRS
values based on multiple ESDOF systems built from the traditional Dn + 2n n Dn + n2 Dn = ug (t ) (6)
MPA method are also calculated and presented. Besides, the results
Substituting Eq. (5) into Eq. (2) gives oor displacements:
obtained by nonlinear response-history analysis (NLRHA) of MDOF
N
structures are treated as exact solutions and are compared with those
obtained by other procedures. As practical examples, the above three
u (t ) = n n Dn (t )
n =1 (7)
analytical procedures are used to estimate FRS for three steel moment
frame structures (3-, 9-, and 20-storey) under various ground motion Making use of Eq. (2) and the relation of kn = n2 mn , the
scenarios. The nonlinear analysis program OpenSees [30] is used to equivalent static force fs (t ) = ku (t ) can be expressed as
process various nonlinear dynamic and static analyses. N
f s (t ) = n2 mn qn (t )
n =1 (8)
2. Modied ESDOF system
In the traditional MPA method, for the nth mode, the base-shear
2.1. Description of test structures roof-displacement (Vbnurn) pushover curve for the force distribution
mn , which is equivalent to n2 mn qn (qn as a constant), is developed.
The test structures considered in the following sections are the Notice that the traditional MPA procedure is an approximate method
pre-Northridge design of the 3-, 9-, and 20-storey SAC Steel and Chopra and Goel [27,28] demonstrated that the errors would be
moment frame buildings from Los Angeles, USA. These structures introduced because the coupling among modal coordinates, qn (t ),
meet the seismic code requirements as per UBC-94 [31] and represent arising from yielding of the system is neglected. That is to say, the
typical low-, medium-, and high-rise buildings designed for Los modal pushover curve is obtained separately without taking into
Angeles at that time. These frames have been used in various similar account the coupling eect among dierent modes. However, obtaining
research works in the recent past [28,3235], and the details, that are an appropriate and reasonable ESDOF system is very important to
available in a report [36], are avoided here for brevity. For each predict seismic demands. One important factor due to the coupling
building, a two-dimensional model of the NorthSouth moment frame eect between the rst mode and any higher mode is that the coupling
representing the building is considered. This frame is modeled for may change the yielding mechanism for the higher mode. In other
nonlinear static and dynamic analyses in OpenSees [30], based on its words, due to the contribution of the rst mode, any higher-mode
centerline dimensions. A simple bilinear inelastic beam-column ele- system may reach the yielding state earlier while they still remain

473
X. Pan et al. Soil Dynamics and Earthquake Engineering 92 (2017) 472487

Fig. 1. Pushover curve for 3-storey steel moment frame structure.

elastic in the traditional higher-mode system. Since the displacement- mode, develop the base-shearroof-displacement pushover curve
response contribution of the fundamental mode is far larger than a (Vb1nur1n) for the combined force distribution
higher mode and the coupling eect due to a higher mode may have an 12 m1q1 + n2 mn qn . Extract the nth-mode pushover curve
advantageous eect on the yielding mechanism of the rst-mode (Vbnurn) from the pushover curve (Vb1nur1n): Vbn is obtained
system, the rst-mode pushover curve is assumed remaining un- by using Vb1n corresponding to ur1n of q1+qn to minus Vb1 corre-
changed without considering the coupling eect. Based on this sponding to ur1 of q1 and urn is calculated by scaling ur1n by
assumption, the higher-mode pushover curves considering dierent qn /(q1 + qn ). The equivalent system parameters such as the yield
q1/ qn values are obtained and investigated. Firstly, develop the rst- displacement, the yield force, and the strain hardening stiness ratio
mode base-shearroof-displacement pushover curve (Vb1ur1) for the can be easily obtained through a bilinear approximation of the push-
force distribution m1. Secondly, for the nth mode apart from the rst over curve (Vbnurn). It is noted that the superposition of inertial

Fig. 2. Pushover curve for 9-storey steel moment frame structure.

474
X. Pan et al. Soil Dynamics and Earthquake Engineering 92 (2017) 472487

modal-forces is adopted only to obtain more reasonable ESDOF coupling eect due to yielding of the structure tends to lead to the
systems because ESDOF systems established from the traditional reduction of the yield force and the yield displacement signicantly.
MPA method have the inherent error of neglecting the coupling eect. This is very important in estimating acceleration demands because
In this step, the superposition is not used to obtain nal FRS and each acceleration demands are very sensitive to the reduction of the yield
modal response for FRS is not examined and determined. force and the yield displacement. It should also be mentioned that the
For each mode, the pushover analysis is carried out to a maximum pushover curve (Vbnurn) after yielding is based on the assumption
interstory drift of 2.5%, as suggested by previous studies [34,35,37]. that the rst-mode displacement and the nth-mode displacement will
Here, the maximum interstory drift is selected as the limiting value, increase according to the ratio of q1/qn . This is problematic since the
instead of the global drift, because for higher modes limiting the global sum of the displacement for two modes after yielding does not satisfy
drift may not necessarily limit the story drifts to realistic values and the linear elastic superposition principle, but it can be considered
that may lead to instability. In our investigation, the limit 2.5% is only reasonable because the yield force is reasonable and the strain hard-
used to induce yielding of structures in order to obtain equivalent ening stiness ratio generally does not change signicantly compared
system parameters such as the yield force and the yield displacement. to the values obtained from the traditional pushover curves.
More importantly, using other drift limits for the bilinear approxima-
tion does not have any signicant eect on equivalent system para-
2.3. Modied ESDOF system
meters [38], which indicates that the obtained equivalent system
parameters are reasonable. Note that in the case of wall systems or
The modied ESDOF system is developed as follows:
other similar systems, if the limit 2.5% will result in abrupt failure or
instability of the structure, then lower limit values should be applied.
(1) Compute the natural frequencies, n , and modes, n , for linearly-
But this study is concentrated on three typical steel moment frames
elastic vibration of the building.
and the wall system is beyond the scope of this investigation. Figs. 13
(2) For the rst mode, develop the base-shearroof-displacement
present the obtained pushover curves from using the above procedure
(Vb1ur1) pushover curve for the force distribution m1.
and the traditional MPA method (q1/ qn = 0 ) for three steel moment
(3) For the nth mode apart from the rst mode, develop the base-
frame structures. The modal pushover curves with dierent q1/ qn values
shearroof-displacement (Vb1nur1n) pushover curve for the
have the same elastic stiness, which indicates that the modal systems combined force distribution 12 m1q1 + n2 mn qn . For a linear
are valid and independent when the structure is in the elastic stage. It structure under a specic ground motion, the modal coordinate
can be seen that for higher modes, with lower q1/ qn values, the higher- ratio q1/ qn changes with time t, and can be expressed by from Eq.
mode pushover curve is close to the traditional pushover curve, (5):
indicating that the rst-mode system does not have a signicant eect
on yielding of the higher-mode system. With the increase ofq1/ qn , the q1 (t )/qn (t ) = 1 D1 (t )/n Dn (t ) (9)

Fig. 3. Pushover curve for 20-storey steel moment frame structure.

475
X. Pan et al. Soil Dynamics and Earthquake Engineering 92 (2017) 472487

Table 1 the error estimate to the prediction of FRS, it is an eective way to take
ESDOF parameters for the case study frames. into account the eect of coupling on yielding of higher-mode systems.
Besides, the deformation response spectrum can be easily obtained and
Mode Tn (s) Mn (kNs2/m) Kn (103 kN/m) n n
conveniently utilized in the practical design stage.
3-storey 1 0.988 757.8 30.68 1.267 0.05 (4) Extract the higher-mode pushover curve (Vbnurn) by using the
2 0.319 1852.7 628.1 0.331 0.05 pushover curve (Vb1nur1n) from step (3) to minus the rst-mode
pushover curve (Vb1ur1) from step (2).
9-storey 1 2.184 2056.5 15.20 1.357 0.05
2 0.820 1786.1 94.86 0.515 0.05 (5) Idealize the pushover curve from step (2) and (4) as a bilinear
3 0.477 2811.5 443.4 0.231 0.070 curve. The base-shearroof-displacement pushover curve is ap-
proximated by a bilinear function by equating the areas under-
20-storey 1 3.735 2319.0 6.553 1.380 0.05 neath the curves. The bilinear curve gives the elastic stiness, K,
2 1.291 1921.5 45.45 0.580 0.05
the yield displacement, Dy, the yield force, Vy, and the strain
3 0.747 1760.9 124.2 0.332 0.072
4 0.527 2030.7 287.6 0.222 0.096 hardening stiness ratio, , from which critical parameters for the
nth-mode equivalent system are obtained.

Table 2 Table 1 shows sample values of ESDOF parameters for the three
The yield force and the strain hardening stiffness ratio with the increase of q1/qn . test structures. The Mn , Kn , and fn can be calculated by Eqs. (10)(12):
Mode q1/qn 3-storey 9-storey 20-storey n = nT mn (10)
Vy (kN) Vy (kN) Vy (kN) n = KnT fn (11)
1 0 3371.3 0.074 5745.7 0.116 3186.1 0.14
fn = mn / Tmn (12)
2 0 4113.4 0.065 3610.4 0.101 3715.3 0.073
5 3960.9 0.056 3774.6 0.048 2206.1 0.072
Note that n is normalized so that the roof displacement is equal to
10 2312.6 0.040 1808.9 0.043 1183.3 0.009 1 in the calculation of ESDOF parameters. Besides, n is the damping
15 1568.6 0.032 1354.9 0.0097 712.7 0.004 ratio for the corresponding ESDOF system and can be calculated by:
50 527.1 0.011 342.0 0.004 186.8 0.006
n = /(2n ) + n /2 (13)
3 0 3453.5 0.134 2707.2 0.070
The Rayleigh damping coecients, and , are determined by
5 1700.1 0.017 1435.1 0.049
10 1666.4 0.063 1660.6 0.017 specifying the damping ratio of 5% in the rst two modes for the real
15 1796.3 0.038 1348.0 0.024 structure.
50 831.1 0.006 445.5 0.006 Table 2 presents the yield force and the strain hardening stiness
ratio, , with the increase of q1/ qn . In the application of the modied
4 0 2484.5 0.108
10 2712.0 0.083
method, for simplicity, the yield force and the strain hardening stiness
15 2000.3 0.057 ratio, , can be obtained through the linear interpolation method
50 690.9 0.017 between two q1/ qn ratios.
100 398.9 0.0005

3. Ground motions
In this study, the maximum elastic displacement response is
To judge the utility of the multiple equivalent systems-based
assumed for each mode to take into account the coupling eect for
method for evaluating FRS, the response-history analysis is performed
simplicity, that is, Dn (t ) D (Tn, n ), which is the deformation response
using earthquake histories generated for the specic site. A total of 20
spectrum ordinate corresponding to the natural period, n , and damp-
strong ground motion records are developed for earthquake magnitude
ing ratio, n (n =1, 2, , N). Although this assumption may introduce
from 6.0 to 8.0, rupture distance (Rrup) from 0 to 40 km, and VS30

Table 3
Detailed information of selected earthquake ground motions.

No. Ground motion Station Comp. Vs30 (m/s) Mag. (Ms) Rrup (km) Scale factor PGA

1 Northern Calif03 Ferndale City Hall 044 219.31 6.5 27.02 1.7549 0.287
2 Managua_ Nicaragua01 Managua_ ESSO 090 288.77 6.24 4.06 1.8337 0.682
3 Managua_ Nicaragua01 Managua_ ESSO 180 288.77 6.24 4.06 1.5786 0.520
4 Gazli_ USSR Karakyr 000 259.59 6.8 5.46 0.8792 0.617
5 Gazli_ USSR Karakyr 090 259.59 6.8 5.46 0.9195 0.794
6 Imperial Valley06 Aeropuerto Mexicali 045 259.86 6.53 0.34 1.4667 0.450
7 Imperial Valley06 Bonds Corner 140 223.03 6.53 2.66 1.0721 0.642
8 Imperial Valley06 Chihuahua 282 242.05 6.53 7.29 1.7047 0.433
9 Imperial Valley06 Delta 262 242.05 6.53 22.03 1.6168 0.381
10 Imperial Valley06 Delta 352 242.05 6.53 22.03 1.3577 0.475
11 Imperial Valley06 El Centro Array #11 140 196.25 6.53 12.56 1.4306 0.525
12 Imperial Valley06 El Centro Array #11 230 196.25 6.53 12.56 1.6139 0.612
13 Imperial Valley06 El Centro Array #8 140 206.08 6.53 3.86 1.1374 0.694
14 Imperial Valley06 El Centro Array #8 230 206.08 6.53 3.86 1.349 0.629
15 Corinth_ Greece Corinth L 361.4 6.6 10.27 1.989 0.471
16 Coalinga01 Parkfield - Fault Zone 14 000 246.07 6.36 29.48 1.4368 0.377
17 Coalinga01 Parkfield - Fault Zone 14 090 246.07 6.36 29.48 1.8246 0.500
18 Coalinga01 Pleasant Valley P.P. - bldg 045 257.38 6.36 8.41 1.6534 0.497
19 Coalinga01 Pleasant Valley P.P. - yard 045 257.38 6.36 8.41 1.1097 0.668
20 Coalinga01 Pleasant Valley P.P. - yard 135 257.38 6.36 8.41 1.6367 0.860

476
X. Pan et al. Soil Dynamics and Earthquake Engineering 92 (2017) 472487

probability of being exceeded (2/50 set) and 50% probability of being


exceeded in 50 years (50/50 set), factors of 1.505 and 0.48 are used to
scale ground motions with 10% probability to represent the corre-
sponding earthquake hazard level. Acceleration spectra for the result-
ing 20 ground motions and the target spectral values with 10/50
hazard level are shown in Fig. 4. It should be noted that 20 records may
not be sucient to avoid the sensitivity of the result statistics to record
selection. However, considering that the primary objective of the case
studies presented here is to check the eectiveness of the multiple
equivalent systems-based method for estimating FRS for a variety of
records, the sample results presented in this paper are deemed
adequate. Besides, no pulse-like ground motions are included in this
investigation.
Fig. 4. Acceleration spectra for 20 ground motions and the target spectral values with
10/50 hazard level.
4. Three test cases for estimating FRS

4.1. Estimation of FRS

from 182 to 366 m/s from the PEER Center. The complete list of these Based on the modied ESDOF systems, this study attempts to
earthquakes appears in Table 3. These records are identied later using provide a simple but eective means to estimate FRS. Following is a
only the serial number that appears corresponding to each record in stepwise description of the calculation of FRS using multiple ESDOF
this table. Each earthquake ground motion is slightly scaled to systems:
minimize the sum of the squared error between the target spectral
values (10% probability of being exceeded in 50 years, referred to as the (1) The mode shape vectors, n , are obtained from an eigenvalue
10/50 set, provided by [39]) and the spectral ordinates. The weights analysis of the elastic two-dimensional frame model. The general-
ascribed to the four-period points are 0.1 at the 0.3 s period point and ized mass, Mn , and modal participation factor, n , for each mode
0.3 for the other three period points. For ground motions with 2% are also calculated based on these mode shapes.
(2) Nonlinear pushover analysis is carried out for each mode to

Fig. 5. FRS values of the roof oor for 3-storey steel moment frame structure under three ground motions with 2/50 hazard level.

477
X. Pan et al. Soil Dynamics and Earthquake Engineering 92 (2017) 472487

N
establish multiple ESDOF systems using the modied method
proposed in Section 2.3. For comparison, multiple ESDOF systems u (t ) + xg (t ) = n n (Dn (t ) + xg (t ))
n =1 (14)
are also established using the traditional MPA method.
(3) The pushover curve for each mode is approximated by a bilinear
curve, and force-deformation parameters such as the elastic
stiness, K, the yield displacement, Dy, the yield force, Vy, and According to step (4), for both the FRSMPA procedure and the
the strain hardening stiness ratio, , are obtained from this FRSmodied procedure, FRS values for each modal system under each
approximation. earthquake are obtained through nonlinear dynamic analysis of the
(4) Acceleration demands (total acceleration demands) for both tradi- corresponding ESDOF system, indicating that both procedures to
tional and modied ESDOF systems, corresponding to each mode, estimate FRS are not a static method in nature. However, the modied
are obtained using nonlinear time-history analysis as shown in Eq. procedure would be conveniently used for the design of nonstructural
(14). FRS values for each mode are converted from the total components since it adopts the concept of equivalent systems and
acceleration response using the damping ratio of 5%. These allows the estimation of FRS based on SDOF systems. It is noteworthy
analyses are carried out for each building for 20 real earthquake that the concept of multiple equivalent systems is nothing new and
scenarios presented in Table 3. equivalent systems were also proposed for estimating seismic demands
(5) For each earthquake, FRS values for all considered modes are in structures. For example, a multiple ESDOF systems-based approach
added using the SRSS combination rule at each period point, since was used to estimate absorbed energy (elastic energy plus hysteretic
the modes for the test structures are found to be well-separated energy) demands [35,40]. Besides, Ghosh et al. [41] performed an
[28]. For the FRS values obtained using modied ESDOF systems, estimation of the Park-Ang damage index for planar multi-storey
this total is denoted as FRSmodied. The FRS values computed on frames using equivalent single-degree systems. However, to date, there
the basis of ESDOF systems built using the traditional MPA has not been any detailed work trying to obtain multiple ESDOF
method are denoted as FRSMPA. In addition, for the MDOF model systems for estimating FRS. Therefore, the authors attempt to bridge
of the actual structure, the calculated FRS values based on non- this gap by providing a multiple equivalent systems-based method to
linear response-history analysis are denoted as FRSNL-RHA. estimate FRS on the basis of modied ESDOF systems.

Fig. 6. FRS values of the roof oor for 9-storey steel moment frame structure under three ground motions with 2/50 hazard level.

478
X. Pan et al. Soil Dynamics and Earthquake Engineering 92 (2017) 472487

Fig. 7. FRS values of the roof oor for 20-storey steel moment frame structure under three ground motions with 2/50 hazard level.

4.2. Results of three test cases Table 4


Bias factor statistics for the 3-storey building under 2/50 hazard level.
Figs. 57 present the obtained FRS values of the roof oor under
No. T1 T2
three ground motions with 2/50 hazard level for 3-storey, 9-storey, and
20-storey steel moment frame structures, respectively. It can be seen NModied NMPA NModied NMPA
that FRS values for both the modied ESDOF system and the
traditional ESDOF system have the similar trend with those for the 1 0.89 0.89 0.97 0.67
2 0.86 0.86 1.23 0.84
real structure under nonlinear response-history analysis. For the three
3 0.91 0.91 0.86 0.66
structures, larger FRS values are observed for nonstructural compo- 4 0.91 0.91 1.13 0.81
nents with periods near the modal period of the supporting structure. 5 0.88 0.88 1.27 0.84
This implies that peak acceleration demands for nonstructural compo- 6 0.93 0.93 1.26 0.79
7 0.97 0.97 0.84 0.58
nents would be amplied when the period of nonstructural components
8 0.90 0.91 1.02 0.73
is close to one of modal periods of the primary structure. Moreover, the 9 0.94 0.94 1.21 0.94
reduction of FRS values due to the inelasticity of the primary structure 10 0.93 0.93 1.02 0.95
is more sensitive when the period of the nonstructural component is 11 0.94 0.94 1.10 0.81
close to one of modal periods of the primary structure [11]. Therefore, 12 0.90 0.90 1.07 0.57
13 0.91 0.91 1.23 0.62
in this study, responses (i.e. FRS) corresponding to modal periods of
14 1.00 0.97 1.20 1.16
the primary structure are selected to evaluate the eectiveness of 15 0.93 0.93 1.05 0.81
dierent methods for estimating FRS. 16 0.92 0.92 0.89 0.76
To evaluate how well the FRSmodied procedure works in predicting 17 0.96 0.96 1.07 0.58
acceleration demands, a bias factor is dened as: 18 0.93 0.93 0.96 0.74
19 0.86 0.85 1.19 0.74
FRSNLRHA 20 0.71 0.71 1.13 0.79
NModified = Mean 0.91 0.91 1.08 0.77
FRSModified (15)
SD 0.06 0.06 0.13 0.14
CoV 0.07 0.06 0.12 0.19
This bias factor is calculated for each earthquake and provides a
quick way to evaluate how much the modied procedure overpredicts
or underpredicts FRS values. By comparison, the bias factor between
FRSNL-RHA and FRSMPA has also been calculated:

479
X. Pan et al. Soil Dynamics and Earthquake Engineering 92 (2017) 472487

Table 5 Table 6
Bias factor statistics for the 9-storey building under 2/50 hazard level. Bias factor statistics for the 20-storey building under 2/50 hazard level.

No. T1 T2 T3 No. T1 T2 T3 T4

NModied NMPA NModied NMPA NModied NMPA NModied NMPA NModied NMPA NModied NMPA NModied NMPA

1 0.90 0.91 1.50 0.88 0.81 0.55 1 0.51 0.50 0.83 0.66 1.14 1.04 0.80 0.77
2 0.61 0.60 0.84 0.72 1.05 0.64 2 0.88 0.88 0.98 0.96 1.00 0.69 0.76 0.70
3 0.80 0.80 0.88 0.75 0.92 0.67 3 0.87 0.86 0.91 0.90 0.99 0.85 0.75 0.72
4 0.96 0.96 0.80 0.71 1.07 0.84 4 0.86 0.86 0.81 0.60 1.16 0.93 0.80 0.62
5 0.86 0.86 0.94 0.69 0.95 0.69 5 0.85 0.85 1.13 0.83 1.05 0.85 0.83 0.66
6 0.88 0.89 1.13 0.85 1.03 0.76 6 0.85 0.85 1.17 0.92 0.75 0.66 0.67 0.61
7 0.88 0.88 0.99 0.75 1.24 0.83 7 0.90 0.90 0.83 0.60 0.91 0.82 0.95 0.86
8 0.88 0.87 1.13 0.59 1.44 1.02 8 0.82 0.82 0.85 0.64 1.15 1.09 1.12 0.98
9 0.90 0.90 0.91 0.66 1.07 0.92 9 0.81 0.81 0.55 0.24 1.13 0.64 0.93 0.60
10 0.84 0.83 1.24 0.68 1.53 0.76 10 0.65 0.66 0.79 0.64 1.12 0.88 0.95 0.80
11 0.99 0.98 0.76 0.65 1.02 0.75 11 0.77 0.77 0.66 0.54 0.92 0.74 0.67 0.61
12 0.77 0.77 1.09 0.98 0.86 0.65 12 0.81 0.81 1.01 0.75 0.91 0.80 0.55 0.53
13 0.72 0.72 0.96 0.82 1.08 0.82 13 0.94 0.94 0.75 0.74 1.27 0.89 0.97 0.68
14 0.94 0.94 1.12 0.96 1.28 1.04 14 0.87 0.87 0.82 0.75 1.08 1.05 0.70 0.68
15 0.93 0.93 0.83 0.73 0.87 0.78 15 0.38 0.38 0.86 0.80 1.11 0.79 0.69 0.60
16 1.05 1.04 0.82 0.78 0.91 0.70 16 0.48 0.48 0.74 0.65 0.93 0.72 0.62 0.56
17 0.76 0.76 0.86 0.63 1.10 0.83 17 0.74 0.74 0.82 0.63 1.12 1.01 0.72 0.71
18 0.88 0.87 0.83 0.59 1.04 0.79 18 0.87 0.87 0.89 0.67 0.90 0.67 0.73 0.67
19 0.85 0.85 1.10 0.74 1.14 0.71 19 0.86 0.85 0.75 0.68 0.82 0.81 0.70 0.67
20 0.85 0.85 0.94 0.87 1.08 0.78 20 0.67 0.67 0.74 0.70 0.86 0.82 0.95 0.94
Mean 0.86 0.86 0.98 0.75 1.08 0.78 Mean 0.77 0.77 0.85 0.70 1.02 0.84 0.79 0.70
SD 0.10 0.10 0.18 0.11 0.19 0.12 SD 0.15 0.15 0.15 0.16 0.14 0.13 0.14 0.12
CoV 0.11 0.11 0.18 0.15 0.17 0.16 CoV 0.20 0.20 0.17 0.23 0.13 0.16 0.18 0.17

FRSNLRHA 5. Interpretation of results


NMPA =
FRSMPA (16)
Tables 46 and scatterplots in Figs. 810 provide outcomes of the
Tables 46 provide the bias factors considering all 20 earthquakes three test cases under 2/50 hazard level. For the rst modal period,
with 2/50 hazard level for 3-storey, 9-storey, and 20-storey buildings, averages and coecients of variation do not show greater eectiveness
respectively. The average, standard deviation and coecient of varia- to estimate FRS for the two methods. This indicates that dierent
tion for each column are shown at the bottom of the table. In addition, higher-mode ESDOF systems would not have a signicant eect on
simple scatter plots (Figs. 810) are also provided here for an easy rst-mode acceleration demands. For higher-mode periods, although
comparison between the two procedures. A scatter plot provides a coecients of variation are larger for the FRSmodied method in some
comparison for all the earthquakes considered for a selected structure. cases compared with the FRSMPA method of estimating FRS values,
The diagonal line across a scatter plot represents a perfect agreement averages indicate that the FRSmodied method of estimating FRS
between the nonlinear response-history analysis and the approximate provides more accurate (that is, closer to FRSNL-RHA) estimates than
analysis technique. The linear t of FRS values reects the distribution the FRSMPA method. The mean bias for the FRSmodied method
of data and the eectiveness in estimating FRS. Due to the randomness (NModied) is closer to its ideal value of 1.0 than the mean bias of the
of earthquakes, median FRS values can be seen as a reasonable FRSMPA method (NMPA) in all the three cases. The scatterplots and
estimation of FRS and are corresponding to designing and evaluating linear ts of FRS values in Figs. 810 provide a quick and easy way to
nonstructural components. Therefore, Figs. 1113 present the median compare the two methods, and these also show very clearly the
spectra under 50/50, 10/50, and 2/50 hazard levels for the 3-storey FRSModied estimates to be closer to the ideal diagonal line, compared
building using the above three methods. Figs. 1419 show the median with the FRSMPA estimates.
spectra with three hazard levels for 9-storey and 20-storey buildings, The FRS values for each mode obtained in these analyses show that
respectively. it is not only the fundamental mode that is contributing to nonstruc-

Fig. 8. Scatterplots comparing FRSModied and FRSMPA with FRSNL-RHA for the roof oor of 3-storey building.

480
X. Pan et al. Soil Dynamics and Earthquake Engineering 92 (2017) 472487

Fig. 9. Scatterplots comparing FRSModied and FRSMPA with FRSNL-RHA for the roof oor of 9-storey building.

tural component demands on the actual structure. Hence, the FRS buildings under 2/50 hazard level, the largest overestimation using the
values due to dierent modal periods of the primary structure must be FRSMPA method is 46.0% for 2/3 height and 52.4% for roof oor,
taken into account. In this study, only two modes, three modes, and respectively, while the error estimate corresponding to the same period
four modes for 3-storey, 9-storey, and 20-storey buildings, respectively, can be decreased to 0.0% and 18%, respectively. For other period
are considered to obtain the FRS values as the results for more modes points, the FRSmodied method can also provide consistently a better
would not have a larger dierence than the previous one. estimate of FRS. It can be demonstrated that under larger earthquake
It can be noted from the linear t of FRS scatter points as well as intensity level, the FRSMPA method overestimates FRS signicantly,
from Tables 46 that the FRSMPA method overestimates the actual particularly for lower-storey buildings, while the FRSmodied method
response (FRSNL-RHA) for most of the earthquake records considered. generally provides a good estimate for FRS. With the decrease of
This is attributed to the neglecting of coupling of higher modes in the earthquake intensity level, the eectiveness of evaluating FRS for the
inelastic domain because such inelastic coupling cannot be properly FRSmodied method is reduced because the ESDOF systems, particularly
accounted for through analyses of independent modal equivalent for the higher mode, tend to remain elastic under earthquake ground
systems. In comparison with the FRSMPA method, the FRSmodied motions. Take the 3-storey building under 10/50 hazard level for
method employs a modied ESDOF system to consider the coupling example, using the FRSMPA method may introduce the error estimate
eect between the rst mode and any higher mode on yielding of the within the interval [9.4%, 20.4%] while the error estimate for the
higher-mode system. The obtained results for the modied method FRSmodied method is within the interval [5.3%, 14.7%]. The obtained
show a better agreement with the actual response from the real results for other period points show the similar tendency. With 50/50
structure. hazard level, it can be seen that not only the FRSmodied method but
Figs. 1113 present the median spectra for the 3-storey building also the FRSMPA method can provide accurate FRS values in compar-
under 50/50, 10/50, and 2/50 hazard levels, respectively. Figs. 1419 ison with the actual response from the real structure with dierent
give the median spectra with three hazard levels for 9-storey and 20- heights.
storey buildings, respectively. For the 3-storey building under 2/50 In addition, it should be noted that the coupling among dierent
hazard level, the FRSMPA method may overestimate the FRS corre- modes is so complex that the coupled modes cannot be easily and
sponding to the period of T2 by 48%, 111% and 97% for roof oor, 2/3 totally decoupled. Therefore, this study only considers the coupling
height, and 1/3 height, respectively. Using the FRSmodied method, the eect between the rst mode and any higher mode. Moreover, the error
error estimate of the FRS corresponding to the same period can be estimate introduced by the FRSmodied method may be due to the
reduced to 10%, 20.3%, and 6.8%, respectively. For 9- and 20-storey following two assumptions: the modal coordinate ratio of the rst-

481
X. Pan et al. Soil Dynamics and Earthquake Engineering 92 (2017) 472487

Fig. 10. Scatterplots comparing FRSModied and FRSMPA with FRSNL-RHA for the roof oor of 20-storey building.

Fig. 11. Median spectra comparing FRSModied and FRSMPA with FRSNL-RHA for 3-storey building under 50/50 hazard level.

482
X. Pan et al. Soil Dynamics and Earthquake Engineering 92 (2017) 472487

Fig. 12. Median spectra comparing FRSModied and FRSMPA with FRSNL-RHA for 3-storey building under 10/50 hazard level.

Fig. 13. Median spectra comparing FRSModied and FRSMPA with FRSNL-RHA for 3-storey building under 2/50 hazard level.

483
X. Pan et al. Soil Dynamics and Earthquake Engineering 92 (2017) 472487

Fig. 14. Median spectra comparing FRSModied and FRSMPA with FRSNL-RHA for 9-storey building under 50/50 hazard level.

Fig. 15. Median spectra comparing FRSModied and FRSMPA with FRSNL-RHA for 9-storey building under 10/50 hazard level.

484
X. Pan et al. Soil Dynamics and Earthquake Engineering 92 (2017) 472487

Fig. 16. Median spectra comparing FRSModied and FRSMPA with FRSNL-RHA for 9-storey building under 2/50 hazard level.

Fig. 17. Median spectra comparing FRSModied and FRSMPA with FRSNL-RHA for 20-storey building under 50/50 hazard level.

485
X. Pan et al. Soil Dynamics and Earthquake Engineering 92 (2017) 472487

Fig. 18. Median spectra comparing FRSModied and FRSMPA with FRSNL-RHA for 20-storey building under 10/50 hazard level.

Fig. 19. Median spectra comparing FRSModied and FRSMPA with FRSNL-RHA for 20-storey building under 2/50 hazard level.

486
X. Pan et al. Soil Dynamics and Earthquake Engineering 92 (2017) 472487

mode system to a higher-mode system is seen as a constant; the modal 2006;28(14):192740.


[10] Villaverde R. Simple method to estimate the seismic nonlinear response of
participation factor and the modal shape do not change in the nonstructural components in buildings. Eng Struct 2006;28(8):120921.
calculation of FRS. [11] Sankaranarayanan R, Medina RA. Acceleration response modication factors for
nonstructural components attached to inelastic moment-resisting frame structures.
Earthq Eng Struct Dyn 2007;36(14):2189210.
6. Conclusions [12] Chaudhuri SR, Villaverde R. Eect of building nonlinearity on seismic response of
nonstructural components: a parametric study. J Struct Eng 2008;134(4):66170.
The following conclusions are drawn based on the study of [13] Miranda E, Taghavi S. A comprehensive study of oor acceleration demands in
multi-story buildings. Improv Seism Perform Exist Build Other Struct
estimation of FRS using multiple modal equivalent systems: 2009:61626. http://dx.doi.org/10.1061/41084(364)5.
A simplied ESDOF system, which takes into account the eect of [14] Oropeza M, Favez P, Lestuzzi P. Seismic response of nonstructural components in
coupling on yielding of higher-mode systems, for the nonlinear seismic case of nonlinear structures based on oor response spectra method. Bull Earthq
Eng 2010;8(2):387400.
evaluation of planar building frames, has been proposed. Based on the
[15] Shooshtari M, Saatcioglu M, Naumoski N, Foo S. Floor response spectra for seismic
proposed ESDOF system, this paper provides a simple but eective design of operational and functional components of concrete buildings in Canada.
means of estimating FRS on a structure without going through a Can J Civ Eng 2010;37(12):15909.
computation-intensive nonlinear response-history analysis of the [16] Chaudhuri SR, Hutchinson TC. Eect of nonlinearity of frame buildings on peak
horizontal oor acceleration. J Earthq Eng 2011;15(1):12442.
MDOF system. [17] Wieser J, Pekcan G, Zaghi AE, Itani A, Maragakis M. Floor accelerations in yielding
special moment resisting frame structures. Earthq Spectra 2013;29(3):9871002.
Based on the three case studies of 3-, 9-, and 20-story buildings, the [18] Lucchini A, Mollaioli F, Bazzurro P. Floor response spectra for bare and inlled
reinforced concrete frames. J Earthq Eng 2014;18(7):106082.
FRSmodied procedure based on the modied ESDOF system is found [19] Petrone C, Magliulo G, Manfredi G. Seismic demand on light acceleration-sensitive
to provide consistently better estimates of FRS than the FRSMPA nonstructural components in European reinforced concrete buildings. Earthq Eng
method on the basis of the traditional ESDOF system. Struct Dyn 2015;44(8):120317.

The FRSmodied method is expected to make engineers design


[20] Saiidi M, Sozen MA. Simple nonlinear seismic analysis of r/c structures. J Struct
Division (ASCE) 1981;107:93752.
nonstructural components using FRS-based design procedures. As [21] Fajfar P, Fischinger M. Non-linear seismic analysis of RC buildings: implications of
long as functions of median or mean FRS for each ESDOF system a case study. Eur Earthq Eng 1987;1:3143.
[22] Fajfar P, Fischinger M. N2-A method for non-linear seismic analysis of regular
have been established, the FRS-based design procedure for non-
buildings. In: Proceedings of the ninth world conference in earthquake engineering;
structural components can be available and used conveniently. 5; 1988, p. 111116.
[23] Uang C, Bertero VV. Evaluation of seismic energy in structures. Earthq Eng Struct
Dyn 1990;19:7790.
Although the modal approach to estimating FRS looks very
[24] Qi X, Moehle JP. Displacement design approach for reinforced concrete structures
promising, additional studies are needed to conrm its robustness. subjected to earthquakes. Berkeley, California: Earthquake Engineering Research
Specically, a larger set of earthquakes must be considered, as well as Center, College of Engineering/University of California; 1991.
alternate building congurations. Furthermore, future research should [25] Rodriguez M. A measure of the capacity of earthquake ground motions to damage
structures. Earthq Eng Struct Dyn 1994;23:62743.
include considerations of issues such as buildings having the non- [26] Fajfar P, Gaspersic P. The N2 method for the seismic damage analysis of RC
symmetrical plan, change of modal shape due to structural nonlinear- buildings. Earthq Eng Struct Dyn 1996;25:3146.
ity, eects of strain hardening, etc. [27] Chopra AK, Goel RK. A modal pushover analysis procedure to estimate seismic
demands of buildings: theory and preliminary evaluation. Report No. PEER 2001/
03. Berkeley (CA): Pacic Earthquake Engineering Research Center, University of
Acknowledgements California; 2001.
[28] Chopra AK, Goel RK. A modal pushover analysis procedure for estimating seismic
demands for buildings. Earthq Eng Struct Dyn 2002;31(3):56182.
This research has been supported by National Natural Science [29] Makarios TK. Optimum denition of equivalent non-linear SDF system in pushover
Foundation of China (Grant nos. 51278150 and 51478143), and procedure of multistory r/c frames. Eng Struct 2005;27(5):81425.
National Key Basic Research Program of China (973 Program) with [30] OpenSees. 2010.[retrieved from]http://opensees.berkeley.edu.
[31] International Conference of Building Ocials (ICBO). Uniform Building Code.
Grant no. 2012CB026200. This research has been supported by the ICBO: Whittier, CA, USA.; 1994.
China Scholarship Council Program for joint Ph.D. student. [32] Chopra AK, Goel RK. A modal pushover analysis procedure to estimate seismic
demands for unsymmetric-plan buildings. Earthq Eng Struct Dyn
2004;33(8):90327.
References
[33] Chase JG, Barroso LR, Hunt S. The impact of total acceleration control for
semiactive earthquake hazard mitigation. Eng Struct 2004;26(2):2019.
[1] Filiatrault A, Christopoulos C, Stearns C. Guidelines, specications, and seismic [34] Ghosh S, Collins KR. Merging energy-based design criteria and reliability based
performance characterization of nonstructural building components and equip- methods: exploring a new concept. Earthq Eng Struct Dyn 2006;35(13):167798.
ment. Report No. PEER 2002/05. Berkeley (CA), Pacic Earthquake Engineering [35] Prasanth T, Ghosh S, Collins KR. Estimation of hysteretic energy demand using
Research Center. Berkeley, California: University of California; 2002. concepts of modal pushover analysis. Earthq Eng Struct Dyn 2008;37(6):97590.
[2] Naeim F. Learning from structural and nonstructural seismic performance of 20 [36] Gupta A, Krawinkler H. Seismic demands for the performance evaluation of steel
extensively instrumented buildings. In: Proceedings of the 12th world conference moment resisting frame structures [Ph.D. Thesis, Department of Civil and
on earthquake engineering. Auckland, New Zealand; 2000. Environmental Engineering]. Stanford, California: Stanford University; 1999.
[3] Biggs JM, Roesset JM. Seismic analysis of equipment mounted on a [37] Rathore M, Chowdhury ARoy, Ghosh S. Approximate methods for estimating
massiveStructure. Cambridge: Massachusetts Inst. of Tech; 1970. hysteretic energy demand on plan-asymmetric buildings. J Earthq Eng
[4] Bumpus SE, Johnson JJ, Smith PD. Best Estimate Method vs Evaluation Method: a 2011;15(1):99123.
comparison of two techniques in evaluating seismic analysis and design. CA (USA): [38] Ghosh S. Two alternatives for implementing performance-based seismic design of
Lawrence Livermore National Lab; 1980. buildings-life cycle cost and seismic energy demand [Ph.D. Thesis, Department of
[5] Kennedy RP, Short SA, Newmark NM. The response of a nuclear power plant to Civil and Environmental Engineering]. Ann Arbor, Michigan: University of
near-eld moderate magnitude earthquakes. Netherlands: North-Holland Publ Co; Michigan; 2003.
1981. [39] Somerville P, Smith N, Punyamurthula S, Sun J. Development of ground motion
[6] Lin J, Mahin SA. Seismic response of light subsystems on inelastic structures. J time histories for phase 2 of the FEMA/SAC steel project. SAC Background
Struct Eng (ASCE) 1985;111(2):40017. Document, Report No. SAC/BD-97/04; 1997.
[7] Singh MP, Moreschi LM, Suarez LE, Matheu EE. Seismic design forces. I: rigid [40] Chou CC. An energy-based seismic evaluation procedure for moment-resisting
nonstructural components. J Struct Eng (ASCE) 2006;132(10):152432. frames [pH.D. Thesis, Department of Structural Engineering]. San Diego, La Jolla,
[8] Singh MP, Moreschi LM, Suarez LE, Matheu EE. Seismic design forces. II: exible CA, U.S.A: University of California; 2001.
nonstructural components. J Struct Eng (ASCE) 2006;132(10):153342. [41] Ghosh S, Datta D, Katakdhond AA. Estimation of the ParkAng damage index for
[9] Medina RA, Sankaranarayanan R, Kingston KM. Floor response spectra for light planar multi-storey frames using equivalent single-degree systems. Eng Struct
components mounted on regular moment-resisting frame structures. Eng Struct 2011;33(9):250924.

487

Вам также может понравиться