See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/298568129

Seismic fragility estimates of high-rise buildings controlled by MR dampers

using performance-based design

Article  in  Engineering Structures · June 2016

DOI: 10.1016/j.engstruct.2016.02.055

CITATIONS READS

10 311

2 authors:

Young- Jin Cha Jong-Wha Bai

University of Manitoba California Baptist University
70 PUBLICATIONS   704 CITATIONS    32 PUBLICATIONS   152 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Development of Protocol for 100 year service life of SYN-FRCP based on ASTM C1818 View project

Bridge Scour Monitoring View project

All content following this page was uploaded by Young-Jin Cha on 08 October 2017.

The user has requested enhancement of the downloaded file.

 Engineering Structures 116 (2016) 192–202

Contents lists available at ScienceDirect

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

Seismic fragility estimates of a moment-resisting frame building

controlled by MR dampers using performance-based design
Young-Jin Cha a,⇑, Jong-Wha Bai b
a
Department of Civil Engineering, University of Manitoba, Winnipeg, MB R3T 5V6, Canada
b
Department of Civil Engineering, California Baptist University, Riverside, CA 92504, USA

a r t i c l e i n f o a b s t r a c t

Article history: Seismic fragility was estimated for a controlled high-rise building using 200 kN magnetorheological (MR)
Received 25 April 2014 dampers with direct performance-based design (DPBD) to assess seismic vulnerability and to validate the
Revised 27 February 2016 performance of the DPBD which was previously developed. The DPBD offers multiple control design lay-
Accepted 29 February 2016
outs for various performance levels subjected to different hazard levels using multi-objective optimiza-
tion approaches. These multiple control design layouts for the given performance levels need to be
validated using random seismic excitations because those performance-based designs (PBD) had been
Keywords:
devolved based on the specific strength of design objective earthquakes (i.e., hazard levels) from the
Fragility analysis
Performance-based design
DPBD. In order to evaluate those PBD cases using MR dampers, two different approaches for fragility esti-
Magnetorheological damper mation of the four PBD cases under two hazard levels are conducted: traditional approach using the over-
Multi-objective all maximum interstory drift and system reliability approach which considers multiple limit states
Genetic algorithm associated with the maximum interstory drift for stories within the entire system. The results are com-
High-rise building pared using 41 earthquake ground motions. From this study, overall seismic fragility relations have been
derived from extensive fragility analyses in terms of broad range of hazard levels for multiple perfor-
mance levels which were achieved by new direct performance-based design using MR dampers.
Moreover, it is observed that the multiple performance-based control design cases obtained from
DPBD clearly show significant reduction in seismic vulnerability compared to the uncontrolled case. It
also shows different seismic fragility estimates against seismic hazards reflecting the performance
enhancement based on the initial objective of the DPBD. Based on the results, the system reliability
approach can identify the stories that have close interstory drifts to the overall maximum value allowing
for more accurate estimates of the seismic fragility of multi-story buildings.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction defined as total weight of a structure. The structural member sizes

Seismic hazard mitigation of civil structures has been one of the
hot topics and challenging research areas during the last several
decades. Performance-based design (PBD) is one of the most
well-investigated approaches to carry out structural seismic
designs to sustain structural safety about a specific hazard level.
The basic concepts of the PBD is that size, type, or material proper-
ties of structural members are designed and determined to satisfy
the predefined performance level about the predefined hazard
level. Most PBD approaches are focused on a single performance
level about single hazard level [1–4]. The single performance level
should be achieved in terms of single objective which is generally
⇑ Corresponding author.
E-mail addresses: young.cha@umanitoba.ca (Y.-J. Cha), jbai@calbaptist.edu
(J.-W. Bai).
are optimized to minimize the objective and satisfy performance
level using optimization algorithms [4]. Multi-objective optimiza-
tion algorithms also applied for the PBD for the moment-
resisting civil steel structures such as multi-objective genetic algo-
rithms [5], simulated annealing methods [6], tabu search methods
[7], and the combinational approaches [4]. In order to solve engi-
neering problems, reliability-oriented single and multi-objective
optimization techniques have been developed and applied [8–
10]. In all the studies cited above, the performance-based designs
are only limited to structural member optimization to satisfy the
predefined performance levels.
As another approach for the seismic mitigation of civil struc-
tures, structural control devices have been developed with 4 differ-
ent categories: passive control devices [11,12], active control
devices [13], semi-active control devices [14–16], and hybrid con-
trol approaches [17]. These control devices have been also used for

http://dx.doi.org/10.1016/j.engstruct.2016.02.055
0141-0296/Ó 2016 Elsevier Ltd. All rights reserved.
 Y.-J. Cha, J.-W. Bai / Engineering Structures 116 (2016) 192–202
PBD to satisfy performance level by installing control devices to
reduce structural responses instead of changing structural member
sizes and material types [18–20]. However, most of these investi-
gations are limited to satisfy a predefined performance level about
a specific hazard level, and locations and number of the control
devices were not optimized, even though the performance of the
control devices is highly dependent on the location and number
of devices in each story of the building [21–23]. Thus, in order to
achieve enhancement in performance of the PBD using control
devices with minimum cost, optimal locations and number of con-
trol devices should be considered as variables of the performance-
based design. Furthermore, these optimal design layouts are also
highly dependent on the objective performance level and its haz-
ard level. Thus, Cha et al. [24] proposed a cost-effective direct
performance-based design (DPBD) method using multi-objective
optimization to find multiple optimal control device layouts which
satisfy multiple performance levels (PLs) subjected to multiple
hazard levels (HLs) simultaneously with Pareto optimal concept.
This DPBD uses one objective earthquake to find optimal layouts
of the control devices for the performance-based design. However,
extensive vulnerability evaluation of the cases of the performance-
based design has not been carried out based on random seismic
events.
Seismic fragility analysis can be a tool to estimate vulnerability
of structural systems under natural hazards. It plays an important
role in estimating seismic losses and in the decision making pro-
cess based on structural performance during seismic events. A
number of research studies related to seismic fragility analysis
and the methodology of developing fragility curves for building
structures have been actively conducted [25–29]. To develop seis-
mic fragility curves, structural capacity limits and demand models
are needed. For constructing demand models, the overall maxi-
mum interstory drift over the height of a building has been widely
used as typical demand measure in these studies. The overall max-
imum interstory drift is a convenient measure to describe the
structural response of a building to lateral loads. However, for
multi-story buildings, fragility estimates developed using only
the overall maximum interstory drift may not reflect the actual
vulnerability of a building, especially when the interstory drifts
for one or more stories are close to the maximum value [30]. This
is because there is only one limit state function defined based on
the overall maximum interstory drift. To assess the probability that
any interstory drift exceeds a specified limit for a given structural
performance level, it is important to include more limit state func-
tions associated with the specific drift demand for stories within
the entire system. To estimate the reliability of a system that is
composed of multiple components, not only the reliabilities of
individual components, but also the uncertainties and correlation
that may exist among the components should be considered [31].
Hence, in this study, overall fragility relationships in terms of
broad hazard levels use a system reliability approach which can
address multiple stories of controlled structures equipped by MR
dampers based on newly recently proposed PBD method which
offers design sets satisfying multiple performance levels subjected
to multiple hazard levels.
2. Backgrounds
2.1. Semi-active magnetorheological damper
To date, the performance of the magnetorheological (MR) dam-
pers are well proved from large-scale real time hybrid simulations
[32–33]. Moreover, damping performances under various types of
time-delay in MR dampers (including control systems) were exten-
sively studied [34]. Consequently, for the DPBD approach, MR dam-
193
pers are used as a control device. The capacity of the nominal
control force is approximately 200 kN at a control current 2.5 A
to the damper wire coils. High nonlinear behavior of the MR dam-
per is described by the phenomenological model developed by
Spencer et al. [35]. The detailed information of the optimal param-
eters of this model can be found in Phillips and Spencer [36], and
MR damper is expressed as,
c1 y_ ¼ az þ k0 ðxd  yÞ þ c0 ðx_ d  yÞ
_ ð1Þ
z_ ¼ cjx_ d  yjzjzj
_ n1
 bðx_ d  yÞjzj
_ n1
þ Aðx_ d  yÞ
_ ð2Þ
f ¼ az þ c0 ðx_ d  yÞ
_ þ k0 ðxd  yÞ þ k1 ðxd  x0 Þ ð3Þ
where az is force determined by the evolutionary variable z, f is the
restoring force, c1 is included to produce the roll-off observed at low
velocities, c0 indicates the viscous damping, k0 represents the stiff-
ness at large velocities, k1 is related to the nominal damper force
due to damper accumulator, x0 is the initial displacement of the
spring of the MR damper model. For the maximum current of MR
damper, 2.5 A is used for the Passive-On (PSON) control.
2.2. Direct performance-based design procedure
Unlike the traditional performance-based design method, direct
performance-based design (DPBD) [24] can find multiple control
design layouts to satisfy multiple performance levels subjected to
multiple hazard levels. Using a smart damping system (i.e., MR
dampers), the multiple performances can be achieved by optimally
installing without any change of the member sizes of the structure
subjected to multiple hazard levels. A set of the PBD options that
satisfy multiple performance objectives is obtained by multi-
objective optimization genetic algorithms (MOGAs) [21–23] which
allow a tradeoff among all predefined conflicting design objectives
in a Pareto-optimal sense. The procedure of the DPBD is,
1. Development of numerical models of the structure and control
device to carry out dynamic nonlinear time history analyses.
2. Determination of the multiple performance levels under multi-
ple hazard levels based on spectral response acceleration
parameters from seismic hazard maps. By modifying spectral
acceleration based on site class effects, response spectrum is
developed. As an objective earthquake, El Centro earthquake
is used. The objective earthquake for each hazard level is scaled
to maximum considered earthquake and design-based earth-
quake using spectral acceleration parameters calculated based
on site class and damping ratio of the structure location.
3. In order to find Pareto-optimal solutions (i.e., multiple layouts
of the control devices) satisfying multiple design objectives
and multiple performance levels simultaneously, multi-
objective optimizations are carried out using gene manipulation
genetic algorithms (GMGAs) developed by Cha et al. [19]. Paral-
lel computing methodologies are used for multiple hazard
levels to explore Pareto-optimal solutions. The two objective
functions are formulated using nonlinear dynamic equations
of motions of the structure including location and numbers of
MR dampers.
4. Integration of multiple Pareto-optimal curves for multiple haz-
ard levels into one Pareto-optimal solution set. Using the com-
bined Pareto-optimal solutions, the final performance-based
designs will be selected to satisfy multiple performance levels
and hazard levels.
5. GMGAs are continued until all predefined performance levels
are satisfied.
6. Validation of performance of the selected design solution by
nonlinear time history analyses using a set of historical earth-
quakes having diverse frequency content and magnitudes.
194 Y.-J. Cha, J.-W. Bai / Engineering Structures 116 (2016) 192–202
2.3. Seismic fragility analysis and probabilistic demand model
development using a system reliability approach
Seismic fragility is defined as the conditional probability of
attaining or exceeding a specific damage limit state during an
earthquake event with a given intensity measure (IM), such as
spectral acceleration (Sa) or peak ground acceleration (PGA). In gen-
eral, the fragility can be written as
FðIM; HÞ ¼ P½gðIM; HÞ 6 0 jIM ð4Þ
where g(IM;H) = C  D(IM;H) is a limit state function used to define
the failure event, for which C and D represent the capacity and seis-
mic demand imposed on the building, respectively, and H is a vec-
tor of unknown parameters that contribute to uncertainties in the
demand model. For frame building structures, demand models
and capacity limits are expressed in terms of limiting interstory
drifts which are defined as relative horizontal displacements
between the top and the bottom of each floor. The overall maxi-
mum interstory drift over the height of a building is a convenient
measure to describe the structural response of a building to lateral
loads. However, the structural behavior of a multi-story building
tends to be more complex. Based on the traditional approach which
only considers the overall maximum interstory drift, the seismic
fragility may be underestimated, particularly if the interstory drifts
for one or more stories are close to the overall maximum interstory
drift of the building. Therefore, to assess the conditional probability
of the building more accurately, multiple drift demands within the
structure should be evaluated. In this paper, probabilistic demand
models using a system reliability approach are developed based
on the several groups of stories for a 9-story high-rise building as
the maximum interstory drift for a given Sa.
Probabilistic demand models have been developed to describe
the relationship between earthquake intensity and the maximum
interstory drift of selected stories. Eq. (5) shows the model form
of a probabilistic linear model:
DðIM; HÞ ¼ h0 þ h1 lnðIMÞ þ re ð5Þ
where D(IM;H) = ln[d(IM;H)] = natural logarithm of the drift
demand for a given earthquake intensity IM, H = (h0, h1, r) is a vec-
tor of unknown parameters; e is a random variable representing the
error in the model with zero mean and unit standard deviation, and
r is the standard deviation of the model error. The logarithmic
transformation is used to approximately satisfy the normality
assumption (i.e., e has the Normal distribution) and the
homoscedasticity assumption (i.e., r is constant). The unknown
parameters in this equation are estimated using the Bayesian
updating rule [37].
f ðHÞ ¼ cLðHÞpðHÞ ð6Þ
where p(H) is the prior distribution of H, which is based on previ-
ous knowledge before obtaining the observation; L(H) is the likeli-
hood function representing the objective information on H, which
is proportional to the conditional probability for given values of
H; c is a normalizing factor; and f(H) is the posterior distribution
of H that incorporates the prior information in p(H) and the infor-
mation from the observation. The mean vector, MH, and the covari-
ance matrix, RHH, can be obtained once f(H) is known. In the
analysis presented in this paper, a non-informative prior distribu-
tion is assumed to reflect that there is little or no information avail-
able about H before collecting the observation.
Using the system reliability approach, the entire data is
grouped into three height levels: 1st – 3rd, 4th – 6th, and 7th
– 9th stories. One linear model is developed for each group so
that there are three demand models and a total of twelve
unknown parameters are used including correlations between
standard deviation. The posterior statistics of the unknown model
parameters are obtained using an adaptive Markov Chain Monte
Carlo (MCMC) simulation method, the Delayed Rejection Adaptive
Metropolis (DRAM), which combines the Delayed Rejection (DR)
method and the Adaptive Metropolis (AM) [38]. Markov chains
are generated with the likelihood formulation of the demand
models based on the initial points and non-informative prior dis-
tribution until a convergence criterion is met. More details on the
formulation of the likelihood formulation can be found in Gardoni
et al. [39].
3. Application to a civil structure
3.1. Case study building
Because the focus of this study is limited to fragility estimates
of controlled structure using DPBD, a 9-story structural control
benchmark model by Ohtori et al. [40] is selected for the
performance-based design carried out by Cha et al. [24]. This build-
ing was designed for Los Angeles, California and used steel
moment-resisting frames (MRFs). In-plane finite element models
of the north–south MRFs have been used for the performance-
based design with bilinear material nonlinearities to characterize
the nonlinear bending stiffness of the structural members as
shown in Fig. 1 [24]. First five natural frequencies of the building
are 0.443, 1.18, 2.04, 3.08, and 4.26 Hz. In detailed properties are
described in Ohtori et al. [40]. It is noted that uncertainties are
considered in the fragility estimation to reflect possible
limitation since the results of this study are based on a single case
study.
3.2. Direct performance-based design using large-scale MR dampers
Direct performance-based design (DPBD) method which is pro-
posed by Cha et al. [24] is used for performance-based design. The
performance levels (PLs) have been determined by the maximum
interstory drift (MID) based on current state of the art [1,3,41] as
shown in Table 1 [24]. Based on this table, the objective perfor-
mance levels have been determined for a 9-story MRF building
subjected to two different hazard levels. The hazard levels (HLs)
are design-based earthquake (DBE) and maximum-considered
earthquake (MCE). Four maximum interstory drift values associ-
ated with two HLs are considered based on the DPBD control
design. Table 2 summarizes all four cases with corresponding max-
imum interstory drifts and hazard levels: (1) 0.7% MID under the
DBE, (2) 1.0% MID under the DBE, (3) 1.3% MID under MCE, and
(4) 1.7% MID under MCE.
Two objective functions are used to minimize control cost and
minimize maximum interstory drift subjected to defined hazard
levels. The first objective function is maximum interstory drift sub-
jected to J 1 6 PLðiÞ; i ¼ 3; 4; . . . ; 5 and normalized by the uncon-
trolled maximum interstory drift during nonlinear time history
analysis:
0
1
jdz ðtÞj
maxt; z
hz
J1 ¼ @ max
A ð7Þ
dn
where dz(t) is maximum interstory drift ratios for the 9-story build-
ing, z is 1; 2; . . . ; n; n is total number of story, hz is the height of the
max
associated floor dz(t) and dn is uncontrolled interstory drifts of the
building structures. The second objective function is the total num-
ber of MR dampers installed in entire stories of the building:
X
nf
J2 ¼ 2 ðrj Þ ð8Þ
j¼1
 Y.-J. Cha, J.-W. Bai / Engineering Structures 116 (2016) 192–202 195

A-A
9th W24x68

W14x257
W27x84
8th
Possible
W30x99 locations of
7th

W14x283
MR damper
W36x135 N
6th

W36x135
5th
W14x370
W36x135 Building Plan A
4th A

W36x135
3rd
W14x455

W36x160
2nd

W36x160
1st
W14x500

W36x160
Ground

B-1

Fig. 1. 9-story benchmark building north–south moment-resisting frame (Cha et al. [24]).

Table 1 The optimal layouts of the MR dampers for the performance-

Performance levels, damage states, and maximum interstory drifts (Cha et al. [24]). based control design cases of the DPBD is used to carry out exten-
Performance levels Damage states Maximum interstory drift (%) sive fragility analyses to investigate vulnerability using random
1 None D < 0.2
seismic excitations. The two Pareto-optimal curves were obtained
2 Slight 0.2 < D < 0.5 and plotted as shown in Fig. 2 [24]. From this figure, 4 different
3 (Immediate Occupancy) Light 0.5 < D < 0.7 control design cases are chosen to carry out seismic fragility esti-
4 Moderate 0.7 < D < 1.5 mations from the each area ‘‘A”, ‘‘B”, ‘‘C”, and ‘‘D” with minimum
5 (Life Safety) Heavy 1.5 < D < 2.5
total number of control devices to select cost efficient design solu-
6 Major 2.5 < D < 5.0
7 (Collapse Prevention) Collapsed 5.0 < D tions from Fig. 2. Table 3 [24] shows the optimal layouts of MR
dampers for each case of control design and total number of dam-
(): FEMA 356 structural performance levels.
pers are quite different. For high levels of performance (such as
0.7% and 1.3%) under the same hazard level, it requires a higher
Table 2
number of MR dampers than low levels of performance (such as
Case study buildings based on the DPBD. 1.0% and 1.7%), because to achieve high performance levels, more
control devices are required. In terms of the hazard levels, to
Cases Hazard levels Performance levels
achieve same performance, more number of control devices is also
Control design 1 DBE 0.7% required to dissipate earthquake energy.
Control design 2 1.0%
Control design 3 MCE 1.3%
Control design 4 1.7% 3.3. Seismic capacity model

where rj is the number of MR dampers installed in jth floor and nf is
the total number of floors in the building. Thus, J 1 and J 2 ; are simul-
taneously minimized by multi-objective optimization approaches.
The spectral acceleration parameters, Ss (short period spectral
acceleration) and S1 (1 s period spectral acceleration) are assumed
as 1.0 g, and 0.4 g, respectively for 9-story building. Using these val-
ues, design spectral response acceleration parameters at short and
long periods are calculated based on FEMA 450 [42] by considering
site class to establish design response spectrum. Thus, the spectral
acceleration values corresponding to two different hazard levels
are 0.118 g and 0.177 g, respectively. The El Centro earthquake is
chosen as an objective earthquake and has been scaled to the design
spectrum to find Pareto-optimal control device layouts [24].
Although computational time of a GA is highly dependent on the
type of the GA, our parallel computing simulations required about
two weeks of computational time to find near-optimal Pareto
curves of the direct performance based design of the nonlinear 9-
story building model by using a general dual-CPU quad-core PC.
The structural capacity relates to a specific damage limit state,
and it can be defined in terms of a maximum force or dynamic
response characteristics, such as maximum displacement, maxi-
mum interstory drift, maximum acceleration, or maximum veloc-
ity. The practice of performance-based seismic design requires
defining the acceptable damage limit states for a given earthquake
intensity. These limits are based on input from project stakehold-
ers and consideration of the building’s contents and function. In
this study, capacity limits are defined based on the target perfor-
mance objectives (0.7%, 1.0%, 1.3%, and 1.7%) to be consistent with
DPBD objectives.
3.4. Earthquake ground motions
A suite of 40 synthetic ground motions (i.e., from LA1 to LA40)
from the SAC project is used [43]. It includes 20 pairs of horizontal
ground motions for Los Angeles with probabilities of exceedance of
196 Y.-J. Cha, J.-W. Bai / Engineering Structures 116 (2016) 192–202

A: Possible solutions for

Cost: Number of MR dampers

C: Possible solutions for Hazard level: DBE
PL3 under DBE PL4 under MCE

installed in entire MRFs

Hazard level: MCE
Uncontrolled result under DBE
56 D: Possible solutions for
PL5 under MCE
46

Uncontrolled result under MCE

24
20 B: Possible solutions for
PL4 under DBE
10
0.7(PL3) 1.0(PL4) 1.3(PL4) 1.46 1.7(PL5) 2.09
Performance levels: Interstory drift (%)

Fig. 2. Multiple Pareto-optimal curves of DPBD (Cha et al. [24]).

Table 3
Optimal MR damper layouts for different cases (Cha et al. [24]).

Cases HLs Number of dampers in different floors Total number of dampers

1st 2nd 3rd 4th 5th 6th 7th 8th 9th
Control design 1 DBE 8 6 4 8 8 2 8 8 4 56
Control design 2 8 0 0 0 0 0 6 6 0 20
Control design 3 MCE 8 8 8 8 0 0 2 8 4 46
Control design 4 8 2 4 0 0 0 0 8 2 24

2% and 10% in 50 years for the nonlinear time history analyses. In
addition, El Centro motion that is used for DPBD is added.
3.5. Nonlinear time history analysis
In order to calculate the maximum interstory drift values with
41 ground motions including objective earthquake of the
performance-based design using the DPBD, the Newmark-b time-
step integration method has been used in the simulation package
for the benchmark model [40]. The governing equation of motion
for the case study structure is expressed as,
€ þ Cs DU_ þ Ks DU ¼ Ms GD€xg þ PDf þ DFerr
Ms DU
where Ms ; Cs ; and Ks are reduced order mass, damping, and stiff-
ness matrices of the 9-story building, DU is incremental response
vector, and G is loading vector for the earthquakes, D€xg is ground
acceleration increment, P is loading vector for the control forces
layouts presented in Table 3 for four different objective perfor-
mances about two different hazard levels achieved from the DPBD,
Df are incremental control force, and DFerr is vector of the unbal-
anced forces which represent the nonlinear hysteretic forces of
the structural member, respectively. In Eq. (9), matrix P; is given as
2 3
r1 0 0 0 0 0 0
6 0 r2 r 2 0 0 0 0
6 7
P¼6
6 .. ..
4 0 0 0   . .
0 0 0 0    rn r n
where ri is the number of MR dampers installed in ith floor in one
side of the building, n is number of the story. P is regulating matrix
to distribute MR damper force to possible locations of MR dampers
which are connected between two adjacent floors. Using this
dynamic equation of motion, nonlinear time history analyses have
been carried out and the maximum interstory drifts about 41 earth-
quake ground motions are presented in Table 4.
Fig. 3 shows the relationship between the overall maximum
interstory drift values based on four performance objectives and
the corresponding spectral acceleration of 41 ground motions.
Because the magnitude of the interstory drift is incrementally
changed, both data points are transformed into logarithmic spaces.
As shown in Fig. 3, there is no significant difference on demands
using different objectives of performance-based design. It means
that the impact of the performance-based design may not be
reflected well based only on the overall maximum interstory drift
values. Thus, in order to assess seismic vulnerability of the case
study structure equipped with MR dampers for the performance-
based design, a system reliability approach (SYS) using multiple
limit states and the traditional approach based on a component
reliability approach (COM) using the overall maximum interstory
drift response are compared. Both approaches use a linear regres-
ð9Þ sion formulation for developing demand models. While a compo-
nent reliability approach (COM) requires only three unknown
parameters with the overall maximum interstory drift based on
the Eq. (5), a system reliability approach (SYS) requires three
parameters for each linear function and the corresponding correla-
tion between model errors is also included.
For a component reliability approach (COM), the overall maxi-
mum interstory drift for a given ground motion record is used so
that only three unknown parameters in Eq. (5) are needed for a
demand model. For the system reliability approach (SYS) in this
study, the entire data of maximum interstory drift for all nine sto-
ries is divided into three groups due to the computational chal-
7
lenges with large number of variables: maximum interstory drift
7
7 ð10Þ demands for (1) 1st – 3rd stories, (2) 4th – 6th stories, and (3)
5
7th – 9th stories. Therefore, a total of 12 unknown parameters
are used: three parameters (h0, h1, and r) used in Eq. (5) for each
height group, and three additional correlation values (q) between
model errors (r). Fig. 4 shows one example of drift demand sets
for a particular performance level (Control design 4: 1.7% MCE).
As shown in Fig. 4, the range of drift demands for the first two
groups (1st – 3rd and 4th – 6th stories) is close while the maxi-
mum interstory drifts for the last group (7th – 9th stories) is differ-
ent. Therefore, true conditional probability of having or exceeding
the given capacity limits would be different from that using the
component reliability approach (COM) based only on the overall
maximum interstory drift because other data points except the
 Y.-J. Cha, J.-W. Bai / Engineering Structures 116 (2016) 192–202 197

Table 4 prior distribution until a convergence criterion is met. To check

Maximum interstory drift results of the nonlinear time history analyses. the convergence of the simulated Markov chains, the Geweke con-
Earthquakes Control Control Control Control Uncontrolled vergence criterion is used [44]. It is based on the comparison
design 1 design 2 design 3 design 4 between the mean values of the first 10% and last 50% of the sam-
El Centro 0.7250 1.0103 0.8667 1.0489 1.5194 ples. If the difference of the mean values is less than 5%, the MCMC
LA01 1.5893 1.9216 1.6393 1.8492 2.0816 simulation is terminated.
LA02 1.8400 2.2348 1.9040 2.1497 2.2518 The demand models in Fig. 4 show one example of the relation-
LA03 2.3067 2.6807 2.3215 2.5964 2.8544
LA04 1.3488 1.7501 1.3518 1.6637 2.0779
ship between maximum interstory drift values and the corre-
LA05 2.2341 2.6123 2.2495 2.5448 2.9198 sponding Sa values for a particular performance objective (1.7%
LA06 1.1620 1.3858 1.1619 1.3368 1.6148 MCE). The maximum interstory drift values obtained from the sim-
LA07 1.0761 1.2337 1.1497 1.2260 1.7337 ulations are used to indicate the level of damage. The regression
LA08 1.1588 1.3222 1.1970 1.2767 1.7795
lines are defined by Eq. (5). The predicted demand (solid line) is
LA09 2.1019 2.3110 2.3723 2.4599 2.7311
LA10 1.1570 1.3817 1.2214 1.3544 1.6684 shown along with the one standard deviation confidence interval
LA11 2.2006 2.3708 2.2376 2.3615 2.2948 (dashed line) in the logarithmic space. A total of 41 points are con-
LA12 0.7717 1.1448 0.7448 0.9989 1.9208 sidered, where each data point represents the demand relationship
LA13 1.2657 1.5007 1.2727 1.4447 2.0403 for one ground record.
LA14 1.5323 1.7130 1.6120 1.6723 2.0700
Tables 5 and 6 show the posterior statistics of the unknown
LA15 2.0755 2.4074 2.2661 2.3552 2.8894
LA16 2.6793 2.9208 2.8274 2.9260 3.7251 parameters for the case study structure with different performance
LA17 2.2230 2.3790 2.3273 2.3092 2.8034 objectives. Since there are two different seismic hazard levels (DBE
LA18 1.9177 1.9291 1.9034 1.8987 2.1423 and MCE) used for PBD, Table 5 contains 0.7% and 1.0% DBE results
LA19 0.8542 1.2217 0.9913 1.1485 2.1110
(control designs 1 and 2), while 1.3% and 1.7% MCE (control design
LA20 1.9617 2.0038 2.0208 1.9925 2.2758
LA21 2.8754 3.1200 2.9660 3.0593 3.5698 3 and 4) are compared in Table 6. Table 7 shows the correlations
LA22 3.0781 3.5036 3.3657 3.5520 4.0552 between model errors for both DBE and MCE. Based on Tables 5
LA23 1.5962 1.8457 1.7873 1.8406 2.2596 and 6, there is the similar tendency that the slopes (h1) of the
LA24 4.9665 5.6990 4.9063 5.4585 6.0323 demand models having higher performance objectives (1.0% DBE
LA25 2.6401 2.8114 2.7413 2.7253 3.0669
and 1.7% MCE) are smaller than those with lower performance
LA26 3.1155 3.2480 3.4985 3.4555 3.8295
LA27 3.2257 3.3841 3.0374 3.2833 3.3877 objectives (0.7% DBE and 1.3% MCE). It is because the maximum
LA28 3.0042 3.1095 2.8650 3.0556 3.2189 interstory drift values are slightly higher when the control devices
LA29 1.6311 1.9367 1.6691 1.8184 2.5524 are designed with the lower performance levels, in particular,
LA30 3.3510 3.4702 3.2273 3.3883 3.6329
when the earthquake intensity is low. It shows that the DPBD cases
LA31 2.4110 2.7282 2.8032 2.9305 2.9097
LA32 2.8147 2.9262 2.7045 2.7851 3.1487
of 0.7% DBE and 1.3% MCE used 56 and 46 MR dampers for the pre-
LA33 4.0401 4.4779 4.0436 4.3956 4.4505 defined hazard levels provide less seismic vulnerability compared
LA34 3.9070 4.3383 3.9138 4.2453 4.3498 to the other two DPBD cases (i.e., 1.0% DBE and 1.7% MCE with
LA35 5.8629 6.0606 5.6698 5.9263 6.4025 20 and 24 numbers of MR dampers) using the 41 ground motions.
LA36 6.1784 6.4401 6.0391 6.3289 6.9654
Therefore, the control design using the suggested DPBD method
LA37 4.4642 4.9492 4.6813 4.8667 4.9273
LA38 6.4436 6.9630 6.6264 6.8927 6.8480 works properly to reduce seismic risk.
LA39 2.1465 2.2323 2.2152 2.2298 2.6853
LA40 4.8137 5.1555 4.8758 5.0176 5.0691 4.2. Seismic fragility curves
Standard 1.4695 1.5230 1.4498 1.4992 1.4381
deviation The seismic fragility curves for the case study structure with
Mean 2.6036 2.8740 2.6652 2.8261 3.1919
four control designs are developed using Monte Carlo simulations.
Median 2.2230 2.4074 2.3215 2.4599 2.8544
As shown in Eq. (4), probabilistic models of the structural capacity
limits and seismic demand are needed to develop seismic fragility
curves. As described in the previous section, the structural capacity
very maximum value are excluded in the fragility estimation. Bai
limits are determined from the target performance objectives
et al. [30] compared the fragility estimates using both COM and
(0.7%, 1.0%, 1.3%, and 1.7%) to be consistent with DPBD objectives.
SYS approaches and proved that COM approach may not reflect
The capacity limit is defined as a random variable having a normal
the actual vulnerability of multi-story buildings when the inter-
distribution in the logarithmic space with the median capacity
story drifts for one or more stories are close to the maximum value.
given above and a standard deviation of 0.3 based on Wen et al.
Therefore, the system reliability approach (SYS) using multiple
[45]. Uncertainties associated with demand models summarized
demand models have more accuracy, in particular, when the inter-
in Tables 5 and 6 are also included in the fragility formulation. It
story drift values are close to the overall maximum values.
is noted that no modeling uncertainty is included in the Monte-
Carlo simulation.
4. Seismic fragility estimates Fig. 5 shows the resulting fragility curves based on the DBE and
MCE seismic hazard levels with four capacity limits. The fragility
4.1. Probabilistic demand models curves using the system reliability approach (SYS) are compared
with those using the traditional component reliability approach
For the probabilistic demand model using a system reliability (COM) with the overall maximum response. As shown, considering
approach (SYS), a linear regression model based on the maximum the possibility of failure of multiple stories significantly increases
interstory drift values from for each story-group is used to predict the overall fragility for the case study structure. It means that fra-
the seismic demand relationship as defined by Eq. (5). The poste- gility based on the overall maximum interstory drift approach,
rior statistics of the unknown model parameters are obtained using which is a traditional approach, is highly underestimated com-
an adaptive MCMC simulation method. Details of the posterior pared to those based on the system reliability approach. This is
statistics of the unknown parameters are summarized in Tables because the formation of the demand model using the system reli-
5–7. These are generated with the likelihood formulation of the ability approach captures the possibility of failure of multiple sto-
demand models based on the initial points and non-informative ries and reduces uncertainty in fragility estimates. Therefore, the
198 Y.-J. Cha, J.-W. Bai / Engineering Structures 116 (2016) 192–202

2 2
(a) (b)
1.5 1.5

1 1

0.5 0.5

0 0

-0.5 -0.5
-2 -1 0 -2 -1 0

2 2
(c) (d)
1.5 1.5

1 1

0.5 0.5
El Centro
0 0 10% in 50 yrs
2% in 50 yrs
-0.5 -0.5
-2 -1 0 -2 -1 0

Fig. 3. Overall maximum interstory drift values in a logarithmic scale for the case study structure with different performance objectives. (a) Control design 1, (b) Control
design 2, (c) Control design 3, (d) Control design 4.

2 2 2
1.5 1.5 1.5
1 1 1
0.5 0.5 0.5
0 0 0
-0.5 -0.5 -0.5
-2 -1 0 -2 -1 0 -2 -1 0

Fig. 4. Maximum interstory drift values with probabilistic demand models using system reliability approach for Control design 4. (a) 1st – 3rd stories, (b) 4th – 6th stories, (c)
7th – 9th stories.

Table 5
Posterior statistics of unknown parameters for the demand models with the DBE hazard levels.

Cases Height Levels Parameters

h0 h1 r
Control design 1 1st – 3rd stories Mean 1.56 0.951 0.192
Standard Deviation 0.0339 0.0317 0.0142
4th – 6th stories Mean 1.51 0.913 0.199
Standard Deviation 0.0351 0.0327 0.0139
7th – 9th stories Mean 1.06 0.560 0.297
Standard Deviation 0.0568 0.0509 0.0233
Control design 2 1st – 3rd stories Mean 1.57 0.872 0.195
Standard Deviation 0.0362 0.0370 0.0112
4th – 6th stories Mean 1.59 0.848 0.186
Standard Deviation 0.0319 0.0288 0.0124
7th – 9th stories Mean 1.11 0.476 0.280
Standard Deviation 0.0421 0.0370 0.0191

system reliability approach better reflects seismic fragility of
multi-story buildings when the interstory drifts for one or more
stories are close to the overall maximum interstory drift of the
building.
Figs. 6 and 7 show the fragility estimates for each capacity limit
based on the DBE and MCE seismic hazard levels, respectively. The
spectral acceleration corresponding to the fundamental period at
each hazard level (0.118 g for DBE and 0.177 g for MCE) is shown
as dashed vertical lines along with the fragility curves in Figs. 6
and 7. For a consistent comparison, all the fragility curves are
developed using the system reliability approach (SYS). As shown
in Fig. 6, the fragility estimates for the case study structure with
 Y.-J. Cha, J.-W. Bai / Engineering Structures 116 (2016) 192–202 199

Table 6
Posterior statistics of unknown parameters for the demand models with the MCE hazard levels.

Cases Height Levels Parameters

h0 h1 r
Control design 3 1st – 3rd stories Mean 1.50 0.911 0.188
Standard Deviation 0.0328 0.0306 0.0120
4th – 6th stories Mean 1.53 0.892 0.194
Standard Deviation 0.0341 0.0320 0.0133
7th – 9th stories Mean 1.20 0.603 0.291
Standard Deviation 0.0482 0.0440 0.0218
Control design 4 1st – 3rd stories Mean 1.55 0.889 0.198
Standard Deviation 0.0363 0.0342 0.0153
4th – 6th stories Mean 1.59 0.858 0.189
Standard Deviation 0.0330 0.0315 0.0147
7th – 9th stories Mean 1.19 0.535 0.281
Standard Deviation 0.0501 0.0485 0.0225

Table 7 Table 8. It is because fragility estimates are compared with four

Posterior mean of correlations for the demand models with the capacity limits (0.7%, 1.0%, 1.3%, and 1.7%) which are consistent
DBE and MCE hazard levels. with DPBD objectives, for control design layouts based on two haz-
Parameters DBE MCE ard levels (DBE and MCE).
As shown in Table 8, the seismic fragilities with performance-
Control Control Control Control
design 1 design 2 design 3 design 4 based design using DPBD are quite less than the fragilities with
uncontrolled structure case in two different hazard levels such as
q13;46 0.819 0.779 0.690 0.744
q13;79 0.198 0.0547 0.140 0.111
DBE and MCE. It means that the application of the control devices
q46;79 0.393 0.272 0.453 0.378 based on DPBD method is quite effective to reduce seismic fragility
and damages from random earthquake excitations, even though
the DPBD was carried out to satisfy specific performance levels
under specific hazard levels.
the target performance objective of 0.7% DBE are lower than those
Furthermore, under the same hazard levels, the seismic fragili-
with 1.0% DBE for a given earthquake intensity for all the capacity
ties with various performance objectives are significantly different.
limits. This proves that the seismic vulnerability of structures with
Fragility estimates for the performance objective of 0.7% DBE (Con-
lower target interstory drift values (i.e., high performance levels) is
trol design 1) have 8%, 25%, 29% and 11% less than those for the
less so that the corresponding losses due to seismic events are
performance objective of 1.0% DBE (Control design 2) under four
expected to be less. For the buildings designed based on the MCE
capacity limits, respectively. In particular, for the capacity limits
seismic hazard level, the same tendency is observed from the fra-
of 1.0% and 1.3% which corresponds to the ‘‘Moderate” damage
gility curves in Fig. 7. Table 8 summarizes the fragility estimates
state, enhancement in seismic fragility due to the difference of per-
at the spectral acceleration corresponding to the fundamental per-
formance objective is more than 25%. This confirms that the
iod of four cases with DBE and MCE seismic hazard levels. A total of
performance-based control design with lower target interstory
8 combinations of hazard levels and capacity limits are listed in

1 1
(a) (b)
0.8 0.8
Control design 1: (COM)
0.6 Control design 1: (SYS) 0.6
Control design 2: (COM)
0.4 Control design 2: (SYS) 0.4
Control design 3: (COM)
0.2 Control design 3: (SYS)
0.2
Control design 4: (COM)
Control design 4: (SYS)
0 0
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5

1 1

0.8
(c) 0.8
(d)

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5

Fig. 5. Seismic fragility curves for the case study structure with four capacity limits. (a) Capacity limit 0.7%, (b) Capacity limit 1.0%, (c) Capacity limit 1.3%, (d) Capacity limit
1.7%.
200 Y.-J. Cha, J.-W. Bai / Engineering Structures 116 (2016) 192–202

(a) (b)

(c) (d)

Fig. 6. Comparison of seismic fragility curves for the case study structure with DBE hazard levels. (a) Capacity limit 0.7%, (b) Capacity limit 1.0%, (c) Capacity limit 1.3%, (d)
Capacity limit 1.7%.

(a) (b)

(c) (d)

Fig. 7. Comparison of seismic fragility curves for the case study structure with MCE hazard levels. (a) Capacity limit 0.7%, (b) Capacity limit 1.0%, (c) Capacity limit 1.3%, (d)
Capacity limit 1.7%.

Table 8
Fragility estimates of different control design cases with DBE and MCE hazard levels.

} } } }

} } } }
a
Differences in seismic fragility estimate between two performance objectives within the same hazard level.
 Y.-J. Cha, J.-W. Bai / Engineering Structures 116 (2016) 192–202
drift values provides less fragility and the corresponding damage
and losses are expected to be less. For the hazard level of MCE, fra-
gility estimates for the performance objective of 1.3% MCE (Control
design 3) have 1%, 6%, 13%, 16% less than those for the performance
objective of 1.7% MCE (Control design 4) under four capacity limits,
respectively. The reduction in fragility is less than those for the
hazard level of DBE because those control design cases use differ-
ent total number of MR dampers. However, these fragilities are
already reduced significantly compared to the uncontrolled one.
For example, control design 4 which requires 24 MR dampers
reduces seismic fragility for the capacity limit of 1.7% from 77%
to 41%. This will affect to estimate corresponding losses for future
events.
5. Conclusions
In this study, the optimal damper layouts of the direct
performance-based control design (DPBD) which was developed
by Cha et al. [21] are used to reduce seismic damages of a 9-
story MRF building using large-scale magnetorheological (MR)
dampers. In order to evaluate performance of the proposed
performance-based design method against random seismic excita-
tions in terms of seismic fragility, two different approaches for
developing fragility curves have been applied. The followings con-
clusions are made based on the results of this study:
 Seismic fragility relationships were derived from extensive fra-
gility analyses of 9-story building controlled by MR dampers
which are installed optimally to satisfy multiple performance
levels subjected to multiple hazard levels. It means that the per-
formance of the semi-active control system including MR dam-
pers and DPBD methodology was evaluated quantitatively in
terms of broad hazard levels.
 System reliability approach for developing fragility curves more
properly reflects vulnerability of multiple stories for the
performance-based design than the traditional approach using
only the overall maximum response. It is because the system
reliability approach considers multiple failure mechanisms that
can contribute to fragility estimates, which is consistent with
the previous result from Bai et al. [30].
 The seismic fragilities are quite reduced by application of the
MR dampers based on DPBD method than those of uncontrolled
structure under two different hazard levels using 41 earth-
quakes. It means that the control devices and DPBD method
are quite effective to reduce seismic damages from random
earthquake ground motions.
 Within the same seismic hazard level, the seismic vulnerabili-
ties of structures with lower performance objectives are quite
less than those with higher performance objectives. It is
because more MR dampers are required to satisfy the target
performance levels of the DPBD. Because of this, the corre-
sponding losses due to seismic events are expected to be less.
 The DPBD method provides a significant difference on seismic
performance with four performance levels in two hazard levels,
in which the seismic vulnerability results from the fragility
analyses are quite well reflected by the initial performance level
of the DPBD. It means that the DPBD approach is quite reason-
able to encounter various random levels of earthquakes.
It should be noted that this study focuses on a comparison of
fragility estimates for a 9-story building with four different control
design layouts based on given hazard levels. For future work, com-
parisons of fragility for a given building configuration with differ-
ent design specifications are planned to be included to provide a
more complete estimate of the fragility due to seismic events.
201
References
[1] Ganzerli S, Pantelides CP, Reaveley LD. Performance-based design using
structural optimization. Earthq Eng Struct Dyn 2000;29:1677–90.
[2] Liu M, Burns SA, Wen YK. Multiobjective optimization for performance-based
seismic design of steel moment frame structures. Earthq Eng Struct Dyn
2005;34:289–306.
[3] Ghobarah A. Performance-based design in earthquake engineering: state of
development. Eng Struct 2001;23:878–84.
[4] Fragiadakis M, Lagaros N, Papadrakakis M. Performance-based multiobjective
optimum design of steel structures considering life-cycle cost. Struct
Multidiscip Optimiz 2006;32:1–11.
[5] Choi SW, Park HS. Multi-objective seismic design method for ensuring beam-
hinging mechanism in steel frames. J Constr Steel Res 2012;74:17–25.
[6] Ohsaki M, Kinoshita T, Pan P. Multiobjective heuristic approaches to seismic
design of steel frames with standard sections. Earthq Eng Struct Dyn
2007;36:1481–95.
[7] Kargahi M, Anderson JC, Dessouky MM. Structural weight optimization of
frames using tabu search I: optimization procedure. J Struct Eng 2006;132
(12):1858–68.
[8] Thampan CPV, Krishnamoorthy CS. System reliability-based configuration
optimization of trusses. J Struct Eng 2001;127(8):947–56.
[9] Mathakari S, Gardoni P, Agarwal P, Raich A, Haukaas T. Reliability-based
optimal design of electrical transmission towers using multi-objective genetic
algorithms. Computer-Aided Civil and Infrastructure Engineering 2007;22
(4):282–92.
[10] Dong F, Liu Y, Su H, Zou R, Guo H. Reliability-oriented multi-objective optimal
decision-making approach for uncertainty-based watershed load reduction.
Sci Total Environ 2015;515:39–48.
[11] Soong TT, Dargush GF. Passive energy dissipation systems in structural
engineering. New York, NY: John Wiley & Sons Inc; 1997.
[12] Housner GW, Bergman LA, Caughey TK, Chassiakos AG, Claus RO, Masri SF,
Skelton RE, Soong TT, Spencer Jr BF, Yao TP. Structural control: past, present,
and future. J. Eng. Mech. 1997;123(9):897–971.
[13] Soong TT. Active structural control: theory and practice. Essex, U.K.: Longman
Scientific; 1990.
[14] Spencer Jr BF, Nagarajaiah S. State of the art of structural control. J. Struct. Eng.
2003;129(7):845–56.
[15] Dyke SJ, Spencer BF, Sain MK, Carlson JD. Modeling and control of
magnetorheological dampers for seismic response reduction. Smart Mater
Struct 1996;5:565–75.
[16] Dyke SJ, Spencer Jr BF, Sain MK, Carlson JD. An experimental study of MR
dampers for seismic protection. Smart Mater Struct 1998;7:693–703.
[17] Housner GW, Masri SF Chassiakos AG, editors. In: Proceedings of the First
World Conference on Structural Control, Pasadena, CA; 1994.
[18] Kurata N, Kobori T, Koshika N. Performance-based design with semi-active
structural control technique. Earthq Eng Struct Dyn 2002;31:445–58.
[19] Lee KS, Ricles J, Sause R. Performance-based seismic design of steel MRFs with
elastomeric dampers. J Struct Eng 2009;135(5):489–98.
[20] Karavasilis TL, Sause R, Ricles JM. Seismic design and evaluation of steel
moment resisting frames with compressed elastomer dampers. Earthq Eng
Struct Dynam 2012;41:411–29.
[21] Cha Y-J, Agrawal AK, Kim Y, Raich AM. Multi-objective genetic algorithms for
cost-effective distributions of actuators and sensors in large structures. Exp
Syst Appl 2012;39(9):7822–33.
[22] Cha Y-J, Raich AM, Barroso LR, Agrawal AK. Optimal placements of active
control devices and sensors in frame structures using multi-objective genetic
algorithms. Struct Contr Health Monitor 2013;20(1):16–44.
[23] Cha Y-J, Kim Y, Raich AM, Agrawal AK. Multi-objective formulation for actuator
and sensor layouts of actively controlled 3-D buildings. J Vib Contr 2013;19
(6):942–60.
[24] Cha Y-J, Agrawal AK, Phillips B, Spencer Jr BF. Direct performance-based design
with 200 kN MR dampers using multi-objective cost effective optimization for
Steel MRFs. Eng Struct 2014;71:60–72.
[25] Rossetto T, Elnashai A. Derivation of vulnerability functions for European-type
RC structures based on observational data. Eng Struct 2003;25(10):1241–63.
[26] Hueste MBD, Bai JW. Seismic retrofit of a reinforced concrete flat-slab
structure: Part II—Seismic fragility analysis. Eng Struct 2007;29(6):1178–88.
[27] Kinali K, Ellingwood BR. Seismic fragility assessment of steel frames for
consequence-based engineering: a case study for Memphis TN. Eng Struct
2007;29(6):1115–27.
[28] Kim YS, Bai J–W. Fragility estimates of smart structures with sensor faults.
Smart Mater Struct 2012;22:125012.
[29] Barnawi W, Dyke SJ. Fragility based analysis of a 20-story benchmark building
with smart device implementation. In: Earth and space conference 2008:
proceedings of the 11th aerospace division international conference on
engineering, science, construction, and operations in challenging
environments; 2008. p. 323.
[30] Bai J-W, Gardoni P, Hueste MBD. Story-specific demand models and seismic
fragility estimates for multi-story buildings. Struct Saf 2011;33(1):96–107.
[31] Straub D, Der Kiureghian A. Improved seismic fragility modeling from
empirical data. Struct Saf 2008;30(4):320–36.
[32] Cha Y-J, Zhang J, Agrawal A, Dong B, Friedman A, Dyke S, Ricles J. Comparative
studies of semi-active control strategies for MR dampers: pure simulation and
real-time hybrid tests. J. Struct. Eng. 2013;139(7):1237–48.
202 Y.-J. Cha, J.-W. Bai / Engineering Structures 116 (2016) 192–202
[33] Cha Y-J, Agrawal AK, Friedman A, Phillips B, Ahn R, Dong B, Dyke SJ, Spencer Jr
BF, Ricles J, Christenson R. Performance validations of semi-active controllers
on a large-scale moment resisting frame equipped with a 200 kN MR damper
using real-time hybrid simulations. J Struct Eng (ASCE) 2014;140
(10):04014066.
[34] Cha Y-J, Agrawal AK, Dyke SJ. Time delay effects on large-scale MR damper
based semi-active control strategies. Smart Mater Struct 2013;22(1):015011.
[35] Spencer Jr BF, Dyke SJ, Sain ML, Carlson JD. Phenomenological model of a
magnetorheological damper. J Eng Mech ASCE 1997;123(3):230–8.
[36] Phillips BM, Spencer BF Jr. Model-based feedforward-feedback tracking control
for real-time hybrid simulation. NSEL Report No. 028. 2011. University of
Illinois at Urbana-Champaign.
[37] Box GEP, Tiao GC. Bayesian inference in statistical analysis
reading. MA: Addison-Wesley; 1992.
[38] Laine M. Adaptive MCMC methods with applications in environmental and
geophysical models Ph.D. dissertation. Lappeenranta, Finland: Lappeenranta
University of Technology; 2008.
[39] Gardoni P, Der Kiureghian A, Mosalam KM. Probabilistic capacity models and
fragility estimates for RC columns based on experimental observations. J Eng
Mech 2002;128(10):1024–38.
View publication stats
[40] Ohtori Y, Christenson RE, Spencer Jr BF, Dyke SJ. Benchmark control problems
for seismically excited nonlinear buildings. J Eng Mech 2004;130(4):366–85.
[41] Federal Emergency Management Agency. Prestandard and commentary for the
seismic rehabilitation of buildings. Report No. FEMA-356, 2000. Washington,
DC.
[42] Building Seismic Safety Council (BSSC). NEHRP recommended provisions for
seismic regulations for new buildings and other structures. Rep. No. FEMA 450.
2003. Washington, D.C.
[43] Somerville PG, Smith NF, Punyamurthula S, Sun JI. Development of ground
motion time histories for phase 2 of the FEMA/SAC Steel Project. SAC/BD-97/
04, SAC Joint Venture, Sacramento, CA; 1997.
[44] Geweke J. Evaluating the accuracy of sampling-based approaches to the
calculation of posterior moments. Bayesian Stat 1992;4:169–93.
[45] Wen YK, Ellingwood BR, Bracci JM. Vulnerability function framework for
consequence-based engineering. DS-4 Technical Report. Mid-America
Earthquake Center, University of Illinois at Urbana-Champaign; 2004.