Peer P Bee Gun Aymo Salam

This article was downloaded by: [University of California, Berkeley]
On: 29 October 2013, At: 21:07

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered
office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Journal of Earthquake Engineering

Publication details, including instructions for authors and
subscription information:
http://www.tandfonline.com/loi/ueqe20
PEER Performance-Based Earthquake

Engineering Methodology, Revisited
a a
Selim Günay & Khalid M. Mosalam
a
Department of Civil and Environmental Engineering , University of
California , Berkeley , California , USA
Accepted author version posted online: 25 Mar 2013.Published
online: 18 Jun 2013.
To cite this article: Selim Günay & Khalid M. Mosalam (2013) PEER Performance-Based Earthquake
Engineering Methodology, Revisited, Journal of Earthquake Engineering, 17:6, 829-858, DOI:
10.1080/13632469.2013.787377
To link to this article: http://dx.doi.org/10.1080/13632469.2013.787377
PLEASE SCROLL DOWN FOR ARTICLE
Taylor & Francis makes every effort to ensure the accuracy of all the information (the
“Content”) contained in the publications on our platform. However, Taylor & Francis,
our agents, and our licensors make no representations or warranties whatsoever as to
the accuracy, completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content
should not be relied upon and should be independently verified with primary sources
of information. Taylor and Francis shall not be liable for any losses, actions, claims,
proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or
howsoever caused arising directly or indirectly in connection with, in relation to or arising
out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any
substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,
systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &
Conditions of access and use can be found at http://www.tandfonline.com/page/terms-
and-conditions
Journal of Earthquake Engineering, 17:829–858, 2013
Copyright © A. S. Elnashai
ISSN: 1363-2469 print / 1559-808X online
DOI: 10.1080/13632469.2013.787377
PEER Performance-Based Earthquake Engineering

Methodology, Revisited
SELIM GÜNAY and KHALID M. MOSALAM

Department of Civil and Environmental Engineering, University of California,
Berkeley, California, USA
Downloaded by [University of California, Berkeley] at 21:07 29 October 2013
A performance-based earthquake engineering (PBEE) methodology was developed at the Pacific

Earthquake Engineering Research (PEER) Center. The method is based on explicit determination
of performance, e.g., monetary losses, in a probabilistic manner where uncertainties in earth-
quake ground motion, structural response, damage, and losses are explicitly considered. There is an
increasing trend towards use of probabilistic performance-based design (PPBD) methods in practice.
Therefore, the International Federation for Structural Concrete (fib) initiated a task group to dissemi-
nate PPBD methods. This article is a contribution to this task group summarizing and demonstrating
the PEER PBEE methodology in a useful manner to practicing engineers.
Keywords Damage Analysis; Hazard Analysis; Loss Analysis; PEER Methodology; Performance-
Based Design; Structural Analysis; Total Probability; Uncertainty
1. Introduction
Traditional earthquake design philosophy is based on preventing structural and non struc-
tural elements of buildings from any damage in low-intensity earthquakes, limiting the
damage in these elements to repairable levels in medium-intensity earthquakes, and pre-
venting the overall or partial collapses of buildings in high-intensity earthquakes. After
the 1994 Northridge and 1995 Kobe earthquakes, the structural engineering community
realized that the amount of damage, economic loss due to downtime, and repair cost of
structures were unacceptably high, even though those structures complied with available
seismic codes based on traditional design philosophy [Lee and Mosalam, 2006].
The concept of performance-based earthquake engineering (PBEE) has its roots from
the above-mentioned realization. Vision 2000 report [SEAOC, 1995] is one of the early
documents of the first-generation PBEE in the U.S. In this article, performance-based earth-
quake design (PBED) is defined as a design framework which results in the desired system
performances at various intensity levels of seismic hazard (Fig. 1). The system perfor-
mance levels are classified as fully operational, operational, life safety, and near collapse,
while hazard levels are classified as frequent, occasional, rare, and very rare events. The
designer and owner consult to select the desired combination of performance and hazard
levels to use as design criteria (objectives). The intended performance levels corresponding
to different hazard levels are either determined based on the public resiliency requirements,
e.g., hospital buildings, or by the private property owners, e.g., residential or commercial
buildings. Subsequent documents of the first-generation PBEE—namely, ATC-40 [ATC,
Received 23 May 2012; accepted 11 March 2013.

Address correspondence to Khalid M. Mosalam, Department of Civil and Environmental Engineering,
University of California, 733 Davis Hall, Berkeley, CA 94720-1710, USA. E-mail: mosalam@ce.berkeley.edu
829
830 S. Günay and K. M. Mosalam
System Performance
Operational
Operational
Life Safety
Fully
Collapse
Near
Frequent
(43 years)
(Return Period)
Hazard Levels
Occasional : unacceptable performance

(72 years) : basic safety objective
Rare : essential hazardous objective
(475 years) : safety critical objective
Very rare
(949 years)
FIGURE 1 Vision 2000 recommended seismic performance objectives for buildings.
1996], FEMA-273 [BSSC, 1997], and FEMA-356 [ASCE, 2000]—express the design
objectives using a similar framework, with slightly different performance descriptions and
hazard levels. The element deformation and force acceptability criteria corresponding to
the performance are specified for different structural and non structural elements for linear,
nonlinear, static, and/or dynamic analyses. These criteria do not possess probability distri-
butions, i.e., the element performance evaluation is deterministic. The defined relationships
between engineering demands and component performance criteria are based somewhat
inconsistently on relationships measured in laboratory tests, calculated by analytical mod-
els, or assumed on the basis of engineering judgment [Moehle, 2003]. In addition, the
element performance evaluation is not tied to a global system performance.
Considering the shortcomings of the first-generation procedures, a more robust PBEE
methodology was developed in the Pacific Earthquake Engineering Research (PEER)
Center. One of the key features of PEER PBEE methodology is the explicit calculation of
system performance measures, such as monetary losses, downtime (duration correspond-
ing to loss of function), and casualties, which are expressed in terms of the direct interest
of various stakeholders. Unlike earlier PBEE methodologies, element forces and deforma-
tions are not directly used for performance evaluation. A key feature of the methodology is
the calculation of performance in a rigorous probabilistic manner without relying on expert
opinion. Accordingly, uncertainties in earthquake intensity, ground motion characteristics,
structural response, physical damage, and economic and human losses are explicitly consid-
ered in the method [Lee and Mosalam, 2006]. PEER performance assessment methodology
has been summarized in various publications [Cornell and Krawinkler, 2000; Krawinkler,
2002; Moehle, 2003; Porter, 2003; Krawinkler and Miranda, 2004; Moehle and Deierlein,
2004] and various benchmark studies have been conducted [Comerio, 2005; Krawinkler,
2005; Goulet et al., 2006; Kunnath, 2006; Kunnath et al., 2006; Mitrani-Reiser et al., 2006;
Bohl, 2009].
Due to the probabilistic nature of the methodology and its various analysis stages,
research on PEER PBEE methodology, examples of which are cited above, has been con-
ducted mostly by the academia with little attention from practicing engineers. However,
it is an accepted fact that the probabilistic PBED methods are gaining popularity and
they may be adopted by standard design codes in the near future. As an example, Rojas
et al. [2011] developed algorithms which utilize the PEER framework for optimized
and automated design of steel frame systems. The increasing trend to use probabilistic
PEER PBEE Methodology, Revisited 831
PBED as a design method is justified by the consequences of recent earthquakes, where

the traditional earthquake design philosophy has occasionally fallen short of meeting the
requirements of sustainability and resiliency. For example, a traditionally designed hospi-
tal building was evacuated immediately after the 2009 L’Aquila, Italy earthquake, while
ambulances were arriving with injured people [Günay and Mosalam, 2010]. Similarly,
some hospitals were evacuated due to non-structural and infill wall damages after the
2010, moment magnitude 8.8, Chile earthquake [Holmes, 2010]. In addition, some resi-
dents did not want to live in their homes after the 2010 Chile earthquake, although the
performance of these buildings was satisfactory according to available codes [Moehle,
2010].
The International Federation for Structural Concrete (fib) initiated a task group for
the dissemination of probabilistic-based design to a larger audience. This article, which
is an extended version of the authors’ contribution to this task group, is an attempt to

summarize and explain the PEER PBEE methodology in a simplified manner such that
it can be appreciated and easily adopted by practicing engineers. The application of the
method, as presented in this simplified manner, is demonstrated using two structures: an
idealized portal frame with and without infill wall and an existing shear wall building on
the campus of the University of California, Berkeley (UCB).
2. PEER PBEE Formulation

As schematically presented in Fig. 2, PEER PBEE methodology consists of four successive
analyses: hazard, structural, damage, and loss. The methodology focuses on the proba-
bilistic calculation of meaningful system performance measures to facility stakeholders by
considering these four analyses in an integrated manner, where uncertainties are explicitly
considered in all analyses. Although each analysis is described separately in the following
subsections, the uncertainties considered in each analysis can be summarized as follows.
Hazard analysis considers the uncertainty of the components that define the earthquake haz-
ard, such as fault locations, magnitude-recurrence rates, and level of attenuation. Structural
analysis takes into account the response of the structure to the earthquake hazard, con-
sidering the uncertainty from the structure, such as material and geometrical properties or
damping, or from the characteristics of earthquake excitation, such as the differences in
ground motion characteristics corresponding to the same hazard level. Damage analysis
defines the level of damage corresponding to the response of the structure, considering the
uncertainty in the corresponding capacity that defines the damage level as well as the uncer-
tainty in the pattern and history of the structural response. Loss analysis determines the
amount of economic loss corresponding to damage considering the uncertainty in the dis-
tribution of damage throughout the structure, such as component variations that may result
in the same damage level. In addition, uncertainty in economical values such as market
prices is also taken into consideration. Due to the above uncertainties, the intensity of haz-
ard, amount of structural response corresponding to hazard, level of damage corresponding
to structural response, and amount of economic loss corresponding to the level of damage
cannot be defined by deterministic values; instead, they are defined by various values with
different probabilities. PEER PBEE methodology combines these probabilities in a consis-
tent manner using the total probability theorem, where the end result is the probability of
exceedance (POE) of various values of a decision variable (DV) during the lifecycle of the
building due to earthquake hazard. It is noted that, DV is typically a system performance
measure in the interest of various stakeholders, such as monetary losses, downtime, or
casualties.
Facility Definition: Location and Design

Hazard Analysis
P (IM) in t years
p(IM) in t years
Intensity measure (IM) Intensity measure (IM)
Structural Analysis
For each value (IMm) of the
p( EDPj IMm)
intensity measure IM: PDFs
m = 1:
j = 1:
Conduct nonlinear time
history analyses with the : # of IMs
ground motions selected for : # of EDPs

IM = IMm Eng.demand param. (EDPj)
Damage Analysis
j = 1: # of damageable groups ( = # of EDP’s) i = 1: # of data points for EDPj
p(DM EDPj)
p(DM EDPji)
P(DM EDPji)
&
k fragility functions
k = 1: # of DM levels (n)
Eng.demand param. (EDPj) DM1 ... DMn DM1 DM2 ... DMn
Loss Analysis
Loss functions for individual

P(DV DM)
damageable groups of the facility
Decision variable (DV)
Combination of the Analyses Stages with Total Probability Theorem

Loss curve for the facility
P(DV)
Decision about Design and Location

P(X Y): Probability of exceedance of X given Y, P(X): probability of exceedance of X, p(X): probability of X
FIGURE 2 Analysis stages of PEER PBEE methodology.
2.1. Hazard Analysis

Hazard analysis is conducted to describe the earthquake hazard in a probabilistic manner,
considering nearby faults, their magnitude-recurrence rates, fault mechanism, source-
site distance, site conditions, etc., and employing attenuation relationships, such as next
generation attenuation (NGA) ground motion prediction equations [Power et al., 2008].
The end result of the hazard analysis is a hazard curve, which shows the variation
of a selected intensity measure (ground motion parameter) vs. mean annual frequency
(MAF) of exceedance [Bommer and Abrahamson, 2006]. Assuming the temporal occur-
rence of an earthquake is described by a Poisson’s model [Kramer, 1996], POE of an
intensity parameter in “t” years corresponding to the given MAF of exceedance is cal-
culated using Eq. (1), where “t” can be selected as the duration of lifecycle of the
facility:
exceedance in " t " years

Annual frequency of
Poisson
Probability of
model
exceedance
Intensity measure (IM) Intensity measure (IM)
FIGURE 3 Correspondence between MAF of exceedance and POE of IM.

P(IM) = 1 − e−λ(IM) t , (1)
where IM is the intensity measure, λ(IM) is the MAF of exceedance of IM, and P(IM) is the
POE of IM in “t” years (Fig. 3). Probability of each value of IM, i.e., p(IMm ), is calculated
algorithmically using Eq. (2) from the POE of IM:
for m = 1 : # of IM data points

p(IM m ) = P(IM m ) if m = # of IM data points. (2)
p(IM m ) = P(IM m ) − P(IM m+1 ) otherwise
Peak ground acceleration, PGA, peak ground velocity, PGV, and spectral acceleration at the
period of the first mode, Sa(T1 ), are examples of parameters that have been commonly used
as intensity measures. These parameters are generally used as IM because most of available
attenuation relationships, used in probabilistic seismic hazard analysis, are developed for
these parameters.
Hazard analysis also includes the selection of a number of ground motion time histo-
ries compatible with the hazard curve. For example, if Sa(T1 ) is utilized as IM, for each
Sa(T1 ) value in the hazard curve, an adequate number of ground motions should be selected
which possess that value of Sa(T1 ). Here, adequate number refers to the number of ground
motions needed to provide meaningful statistical data in the structural analysis. In order to
be consistent with the probabilistic seismic hazard analysis, selected ground motions should
be compatible with the magnitude and distance combination which dominates the hazard
for a particular value of IM [Sommerville and Porter, 2005]. In addition, it is important to
use the ground motions recorded in sites consistent with the site class of the considered
structure.
For practical purposes, instead of selecting ground motions for each IM value, an alter-
native way is to select a representative set of records and scale them for different IM values.
However, this alternative might lead to unrealistic ground motions when large scales are
needed. As an example for this situation and according to [Sommerville and Porter, 2005],
higher magnitudes correspond to higher Sa(T1 ) values. However, higher magnitudes also
correspond to longer durations. When the amplitude of a ground motion is scaled consider-
ing Sa(T1 ), the duration of the ground motion does not change which may make the ground
motion unrealistic. Therefore, it is beneficial to use unscaled ground motions, whenever
possible. This requires selection of a larger number of ground motions with respect to
the representative set mentioned above, since different ground motions need to be used for
each IM value. This selection process, which may seem to be demanding, is actually worth-
while, since it needs to be performed once by a structural engineer. The compiled database
can subsequently be used in a variety of the probabilistic PBED projects conducted by the
engineer.
2.2. Structural Analysis

Structural analysis is conducted to determine the response of a structure to various lev-
els and characteristics of earthquake hazard in a probabilistic manner. For this purpose,
a computational model of the structure is developed. Uncertainties in parameters defin-
ing the structural model, e.g., mass, damping, stiffness, and strength, are considered by
varying the relevant properties in the model. For each intensity level of earthquake haz-
ard, nonlinear time history analyses are conducted to estimate the structural responses
in terms of selected engineering demand parameters (EDP), using the ground motions
selected for that intensity level. EDPs may include local parameters such as element
forces or deformations, or global parameters such as floor acceleration and displace-
ment, and interstory drift. For the structural components, element forces (such as the
axial or shear forces in a non ductile column) or deformations (such as plastic rota-
tions for ductile flexural behavior) are more suitable, whereas global parameters (such
as floor acceleration) are better suited for non-structural components, e.g., building con-
tent or equipment. Interstory drift is a suitable parameter that can be used for the analyses
focusing on both structural and non-structural components. The PEER-PBEE formulation
requires a single value for EDP. Therefore, peak values of the above EDPs are generally
employed.
It is possible to use different EDPs for different damageable components of a structure
(denoted by EDPj in Fig. 4). For example, interstory drift can be used for the structural
system of a building [Krawinkler, 2005], while using floor acceleration for office or labora-
tory equipment [Comerio, 2005] of the same building. As a result of nonlinear time history
analyses, the number of data points for each of the selected EDPs (i.e., EDPj ) at an intensity
level is equal to the number of analyses conducted for that intensity level, i.e., the number
of used ground motions multiplied by the number of variations in the structural model.
For a structural engineer, the demanding part is the construction of different variations of a
model, unless parameter-based software such as the structural and geotechnical simulation
framework, OpenSees [2010], is used. It is worth mentioning that Lee and Mosalam [2006]
showed that ground motion variability is more significant than the uncertainty in structural
parameters in affecting the local EDPs, based on analyses conducted for one of the testbeds
of PEER PBEE methodology. In the same study, a method to determine the significance
of the uncertainty in the various parameters of a model is described by conducting a tor-
nado diagram analysis. Using a similar approach for an investigated facility, if the structural
engineer arrives to a similar conclusion, the effort of introducing variations in the model
can be reduced.
m = 1: # of IM data points
n simulations for a given IM and
p( EDPj IMm)
j = 1: # of EDP’s
x simulations with no global collapse
Calculate the parameters of the considered probability

distribution for each EDP (EDPj) from x simulations
Engineering demand parameter (EDPj)
FIGURE 4 Determination of the PDF of an EDP from structural analysis.

It is likely to observe global collapse at higher intensity levels. Global collapse is

treated separately in PEER PBEE methodology since its probability does not change
from a damageable group to another. In a simulation, global collapse corresponds to
infinite increase of response for small increases in input intensity, i.e., global dynamic
instability. This approach can be used to determine global collapse when a represen-
tative set of ground motions is selected and the ground motions are scaled for each
intensity level. However, for the case of using unscaled ground motions or the case
when the number of utilized scales is not adequate to accurately determine the initia-
tion of global dynamic instability, this approach of global collapse determination from
increasing intensities of the same ground motion is not feasible. A practical method to
determine global collapse is to calculate a value of a representative response param-
eter, e.g., interstory drift, corresponding to global collapse from a prior analysis, e.g.
pushover analysis, and compare this collapse value of the representative parameter with
the peak value from nonlinear time history analysis. For a more realistic representation of
global collapse, Talaat and Mosalam [2009] developed a progressive collapse algorithm
based on element removal and implemented it into OpenSees [2010], which is one of the
main tools utilized for the application of PEER PBEE methodology. The probability of
the global collapse event, p(C|IM), can be approximately determined as the number of
simulations leading to it divided by the total number of simulations conducted for the con-
sidered intensity level. Probability of having no global collapse is therefore p(NC|IM) =
1.0 – p(C|IM). These probabilities are employed in the loss determination as explained
later.
A suitable probability distribution, e.g. lognormal, is used for each considered EDP,
i.e., EDPj , by calculating the distribution parameters from the data obtained from simula-
tions with no global collapse (Fig. 4). The number of probability density functions (PDFs)
available as a result of structural analysis is α × β, where α is the number of IM data points
and β is the number of considered EDPs.
2.3. Damage Analysis

The main advantage of PEER PBEE methodology compared to the first-generation PBEE
methods is the determination of DVs meaningful to stakeholders, e.g., monetary losses,
downtime, casualties, rather than the determination of parameters meaningful to engineers
only, e.g., forces or displacements. After the determination of probabilities of EDPs in the
structural analysis, these probabilities are used to determine the POE or expected values of
the DVs. This is achieved by the inclusion of the damage and loss analyses as explained in
the following.
The purpose of the damage analysis is to estimate physical damage at the component or
system levels as a function of the structural response. Although other definitions are possi-
ble, damage measures, DMs, are typically defined in terms of damage levels corresponding
to the repair measures needed to restore the components of a facility to the original condi-
tions [Porter, 2003]. For example, Mitrani-Reiser et al. [2006] defined DMs of structural
elements as light, moderate, and severe (or collapse) corresponding to repair with epoxy
injections, repair with jacketing, and element replacement, respectively. They defined DMs
of non structural drywall partitions as visible and significant corresponding to patching and
replacement of the partition, respectively.
Damage level of a damageable group shows variance, even for the same value of EDP.
This is mainly due to two reasons. The first reason is the uncertainty in the material char-
acteristics which translate into the uncertainty in the corresponding EDP capacity term that
defines the damage level. The second reason is the differences in the pattern and history of
the structural response. EDPs are generally represented as peak quantities. However, dif-
ferences in the path of achieving the same peak value introduce differences in the observed
damage and these differences contribute to the variance of the DM corresponding to an
EDP. Accordingly, a specific value of EDP corresponds to various DMs with different
probabilities, as shown in Fig. 5.
The tool used to determine the above probabilities is the “fragility function,” as com-
monly referred to in the literature. A fragility function represents the POE of a damage
measure for different values of an EDP. Fragility functions of structural and non-structural
components can be developed for a facility using experimental or analytical models.
Alternatively, generic fragility functions of a structure or component type based on expert
opinion can be used. The damageable parts of a facility are divided into damageable groups
consisting of components that are affected by the same EDP in a similar manner, i.e., the
components in a group have the same fragility functions. For example, Bohl [2009] used
16 different groups for a steel moment frame building including the structural system, exte-
rior enclosure, drift-sensitive and acceleration-sensitive non structural elements, and office
content of each floor. For each damageable group (index j) and each EDP (index i) data
point (EDPj i ), the POE of a DM is available as a point on the related fragility curve (see
Fig. 5). Probability of a DM is calculated from the POE using the following equation:
for k = 1 : # of DM levels
p(DM k |EDPij ) = P(DM k |EDPij ) if k = # of DM levels . (3)
p(DM k |EDPij ) = P(DM k |EDPij ) − P(DM k+1 |EDPij ) otherwise
2.4. Loss Analysis

Loss analysis is the last stage of PEER PBEE methodology, where damage information
obtained from the damage analysis (see Fig. 5) is converted to the final DVs. These vari-
ables can be used directly by the structural engineer in the design process with the inclusion
of stakeholders for decision-making. Most commonly utilized DVs are as follows.
1. Fatalities: Number of deaths as a direct result of facility damage.

2. Economic loss: Monetary loss which is a result of the repair cost of the damaged
components of a facility or the replacement of the facility.
3. Repair duration: Duration of repairs during which the facility is not functioning.
4. Injuries: Number of injuries, as a direct result of facility damage.
The first three of these DVs are commonly known as deaths, dollars, and downtime
(the 3 Ds).
P(DM EDPj)
p(DM EDPji)
P(DM EDPji)
EDPji
Engineering demand parameter (EDPj) DM1 DM2 DM3 DM1 DM2 DM3
FIGURE 5 Probability of exceedance, P, and probability, p, of a damage level from

fragility curves.
In the loss analysis, the POE of the losses for different damageable groups at different
DMs (loss functions) is determined. It is not practical to define every individual structural
and non-structural component as a separate damageable group. Hence, the distribution of
damage within the damageable group may result in different values of DVs for the same
DM. If the DV is economic loss or repair cost, uncertainty that originates from the eco-
nomical values, such as fluctuation in the market prices, can be included. Accordingly, a
specific DM of a damageable group corresponds to various DVs with different probabili-
ties represented by the loss functions in Fig. 6. As shown in this figure, the total number of
loss functions present for a facility is γ × λ, where γ and λ are the number of DMs and
damageable groups, respectively.
2.5. Determination of the Loss Curve

The outcome of the PEER PBEE methodology is the loss curve, which describes the POE
of different values of a DV. This curve is obtained from the total probability theorem by
combining the probability and POE values from the hazard, structural, damage, and loss
analyses as follows:

P(DV n ) = P( DV n | IMm ) p(IMm ) (4a)
m
P( DV n | IMm ) = P( DV n | NC, IMm ) p( NC| IMm ) + P( DV n | C) p( C| IMm ) (4b)

P( DV n | NC, IM m ) = P DV nj NC, IM m (4c)
j
n
P DVjn NC, IMm = P DVj EDPij p EDPij IMm (4d)
i
n
P DVjn EDPij = P DVj DMk p DMk | EDPij , (4e)
k
where p(IMm ) is the probability of the mth value of IM (hazard analysis outcome, Fig.
3), p(EDPj i |IMm ) is the probability of the ith value of the jth EDP (EDP used in fragility
function of the jth damageable group) for the mth value of IM (structural analysis outcome,
Fig. 4), p(DMk |EDPj i ) is the probability of the kth DM when subjected to the ith value of
the EDP utilized for the fragility function of the jth damageable group (damage analysis
: Loss functions
P(DV DMk)
: # of DMs
: # of damageable groups
FIGURE 6 Loss functions for individual damageable groups of the facility.

outcome, i.e., fragility function, Fig. 5), and P(DVj n |DMk ) is the POE of the nth value of
the DV for the jth damageable group of the facility when the kth DM occurs (loss analysis
outcome, i.e., loss function, Fig. 6). Moreover, p(C|IMm ) and p(NC|IMm ) are the probabil-
ities of having and not having global collapse, respectively, under ground motion intensity
IMm . Finally, P(DVn |C) is the POE of the nth value of DV in the case of global collapse.
Krawinkler [2005] assumed a lognormal distribution for P(DVn |C) for the case of DV being
the economic loss.
The calculation of the loss curve in terms of DV, where DV = DVn ∀n, is summarized
as follows.
1. Determine the loss functions of the selected DV for each jth damageable group of
the facility undergoing each kth value of DM, P(DVj |DMk ).
2. Combine the loss and damage analyses results for each damageable group.
Determine the POE of DV for each damageable group of the facility for each ith
value of the EDP used in the fragility function of that group, P(DVj |EDPj i ), Eq. (4e),
considering loss functions of step 1, P(DVj |DMk ), and probability of each DM from
the fragility function, p(DMk |EDPj i ).
3. Combine the loss, damage, and structural analyses results for each damageable
group. Determine the POE of DV for each damageable group of the facility for
a given value of IM under the condition that global collapse does not occur,
P(DVj |NC,IMm ), Eq. (4d), considering the POE of step 2, P(DVj |EDPj i ) and prob-
ability of each value of EDP used in the fragility function of that group when
subjected to the mth ground motion compatible with the considered IM, p(EDPj i
|IMm ).
4. Combine the results of each damageable group. Determine the POE of DV for the
facility for a given value of IM under the condition that global collapse does not
occur, P(DV|NC, IMm ) by summing up the POE of DV for each damageable group,
P(DVj |NC,IMm ), Eq. (4c).
5. Combine the results of “no collapse” and collapse cases. Determine the POE of DV
for the facility for a given value of IM, P(DV|IMm ), by summing up the POE of DV
for “no collapse” and collapse cases weighted by the probabilities of these cases,
Eq. (4b).
6. Finally, include the hazard analysis results. Determine the POE of DV for the
facility, P(DV), by summing up the POE of DV for different IMs, P(DV|IMm ),
multiplied by the probabilities of these IMs, p(IMm ), Eq. (4a). P(DV) represents the
POE in “t” years, the duration for which the POE values are calculated for the IM
in the hazard analysis, refer to Eq. (1).
Some simplifications, variations and comments related to Eq. (4) can be stated as
follows.
● Order of the steps used for the determination of the loss curve are independent
of order of the conducted analyses shown in Fig. 2 and represent the order of the
calculations once all the four analysis stages are completed.
● Expected value of DV can be calculated with Eq. (4) by replacing the POE values in
Eq. (4) with the expected values, e.g., E(DVj |DMk ) instead of P(DVj |DMk ).
● The loss curve defined in Eq. (4a) considers all possible scenarios for hazard,
whereas in some applications of the PEER PBEE methodology, few IM values, e.g.,
for 2%, 10%, and 20% POE in 50 years, are considered separately. In this case,
Eq. (4a) is not used and the loss curves for individual IMs (different scenarios) are
defined with Eq. (4b).
● In Eq. (4b), the POE of the DV in case of collapse, P(DVn |C), is not conditioned on
the IM, whereas that in case of “no collapse,” P(DVn |NC,IMm ), is conditioned on the
IM, since “no collapse” case consists of different damage states and the contribution
of each of these damage states to the “no collapse” case changes for different IMs.
For example, the loss function for slight damage has the highest contribution for a
small value of the IM, whereas the loss function for severe damage has the highest
contribution for a large value of the IM.
● It is observed that the loss, damage, and structural analyses results are summed in
a straightforward manner in Eqs. (4d) and (4e). However, integration of the hazard
analysis into the formulation does not take place in this way because of the presence
of damageable groups and collapse and “no collapse” cases and therefore requires
Eqs. (4a)–(4c).
● In order to have a direct demonstration of the total probability theorem, the POE of
DV is presented in Eq. (5) for the case of only “no collapse” and single damageable
group (j = 1), where the equation is represented in the form of a triple summation:

P (DV n ) = P DV n DMk p DMk | EDPi p EDPi IMm p(IMm ) .
m i k
(5)
● Equation (5) has a direct resemblance to the well-known triple integration, Eq. (6),
referred to as the PEER PBEE framework equation in [Moehle, 2003; Krawinkler,
2005]:

λ(DV) = GDV|DM dGDM|EDP dGEDP|IM dλ(IM), (6)
where G represents the conditional probability and λ represents the MAF. It is to be

noted that Eq. (6) is restated here as originally published in the above documents.
● There are various ways that an engineer can use the PEER PBEE methodology,
which can be summarized as follows: (1) Evaluation of a traditional code-based
design in a performance-based probabilistic approach. This application is appropri-
ate in the current state of traditional code-based design if the engineer prefers to
introduce performance-based enhancements to the mandatory code-based design.
(2) Evaluation of the performance of an existing structure or the outcome of differ-
ent retrofit measures. (3) Use of the methodology directly as a design tool, e.g., for
decision-making amongst different design alternatives. This type of application is
expected to gain widespread usage when the probabilistic PBED methods start to be
employed as a standard design method.
3. Guidance on Uncertainty Quantification

This section provides concise guidance on the methods, tools, and applications that can
be used for the quantification of uncertainty for the various stages of the PEER PBEE
methodology. Here, the intent is not to provide a recipe for users, since it is against the
overall objective of the methodology, which is the elimination/minimization of prescrip-
tive requirements in earthquake engineering design and assessment. Rather, the intent is to
highlight and combine some of the key issues mentioned in the above sections.

There are various publicly available applications that can be used for hazard analysis, such
as OpenSHA [Field et al., 2003] or the online hazard curve application provided by USGS
[Peterson et al., 2008]. These applications can be used to calculate the hazard curve using
the geographical coordinates and the site class of the considered structure as inputs. It is
to be noted that OpenSHA application provides a greater degree of flexibility, in terms
of the selection of intensity measure, attenuation relationships (ground motion prediction
equations) and site class description.

OpenSees [2010] can be considered as the most suitable tool in terms of automation of the
analytical simulations required in the structural analysis stage of the methodology. Using
OpenSees, all of the required analyses can be conducted using a single input script without
additional programming effort. Up to date, OpenSees was generally utilized by the aca-
demic community, although it started to gain broader applications in the industry in the
recent years. Practicing engineers usually prefer the software they are most familiar with
for the structural analysis simulations. The commonly utilized structural analysis/finite ele-
ment software in the industry are generally based on graphical user interfaces instead of the
text or script based input files. Engineers can make use of automation tools for Windows
applications such as MacroExpress [Insight Sofware Solutions, 2013] to automate the
analyses for such platforms.
At the structural analysis stage, it is useful for an engineer to list all the random vari-
ables that may have an influence on the structural response. Then, the number of random
variables in this initial list can be reduced by investigation of the effect of each considered
random variable on the variability of structural response. The tornado diagram analysis
mentioned above is a suitable and practical tool that can be used for this purpose [Porter
et al., 2002, Lee and Mosalam, 2006].
In the first step of tornado diagram analysis, median and two extreme values (e.g., val-
ues with 10% and 90% POE) are determined for each considered random variable. In the
second step, two nonlinear time history simulations are conducted for each random vari-
able (resulting in a total number of simulations equal to 2N where N is the total number
of considered random variables). These two simulations are conducted with two extreme
values of the respective random variable while setting all the other random variables to
their median values. The difference between the EDP values calculated as a result of the
simulations with the two extreme values, called the swing, represents the variation in EDP,
due to the variation in the respective random variable. Accordingly, a variable with larger
effect on the EDP has larger swing than those with lesser effect. In the third step, a final
nonlinear time history simulation is conducted where all random variables are set to their
median values. EDP calculated from this simulation can be used as a common reference in
the tornado diagram. In the final step, swings are displayed in a descending order from top
to bottom, which resembles the shape of a tornado. The parameters that are decided to have
insignificant effect based on the tornado diagram can be treated as deterministic, thereby
eliminating the burden of unnecessary simulations.

The tool used in the damage analysis stage is the fragility function, represented in the form
of a fragility curve. Ideally, an engineer should develop the fragility curves for different
damageable groups of the investigated structure by considering the specifics of the struc-
ture. However, such development may require component testing or detailed finite element
simulations, which may not be practical for all projects. For such cases, the fragility curves
developed for similar structural and non-structural systems could be used. An example of
these fragility curves are those in HAZUS [2003], where the median and lognormal stan-
dard deviation of interstory drifts corresponding to different damage states of structural
components are provided for different building types, e.g. concrete, steel, and wood, to
be designed for different levels of seismic intensities. The fragility curves of HAZUS for
non structural components provide POE of DM as a function of an IM. Therefore, they
are not suitable to be used in the context of the PEER PBEE methodology. PACT software
developed as a result of ATC-58 project [2013] provides fragility curves for a broad range
of non structural damage groups including servers and network equipment or unanchored
bookcases, in the format suitable for PEER PBEE methodology. However, it is highly rec-
ommended that an engineer should have a brief survey regarding the development of the
fragility curves under consideration along with some basic checks to verify the suitability
of the selected fragility curves for the considered structure.
The damageable groups consist of components which are expected to have the same
fragility functions. Accordingly, each damageable group should incorporate elements that
are sensitive to the same EDP and that are expected to be affected from the considered EDP
in the same manner. This basic definition can be used by an engineer for the selection of
different damageable groups.
3.4. Loss Analysis

Loss function is the tool used in the loss analysis stage. Determination of a loss function
requires a probabilistic distribution, expected value, and a dispersion measure to calculate
the main descriptors of this distribution. Lognormal distribution is a suitable probabilis-
tic distribution to define the loss functions [Beck et al., 2002]. Regarding the expected
value, an engineer may be able to make reasonable estimates of the expected losses cor-
responding to different damage states if there is enough information about the occupancy
type, contents, etc. Otherwise, HAZUS [2003] provides expected values of a broad range
of DVs (e.g., repair cost, building recovery time) for different occupancy types and for
the damage states corresponding to structural components and drift and acceleration sensi-
tive non-structural components. However, there is not any information in HAZUS related
to a dispersion measure. Porter [2010] proposed a method for the determination of loss
coefficient of variation (COV) using the expected value information provided by HAZUS.
To determine the COV for the structure in consideration, engineers can obtain guidance
from professional cost estimators [Porter et al., 2002].
4. Application of PEER Formulation: Portal Frame

The application of the PEER PBEE methodology, as presented in a simplified manner,
is demonstrated on two structures: (1) an idealized portal frame with and without infill
wall and (2) an existing shear wall building on the UCB campus. Application of structural
and hazard analyses are demonstrated on the idealized portal frame in this section and
application of all analyses are demonstrated using the existing shear wall building in the
following section.
The geometry of the portal frame is based on the dimensions of a single-story rein-
forced concrete (RC) frame with unreinforced masonry infill wall tested on UCB shaking
table [Hashemi and Mosalam, 2006]. PEER PBEE methodology is applied on both bare
FIGURE 7 Portal frame with infill used for demonstration of PEER PBEE methodology.
and infilled frame configurations. The infill wall is modeled as a diagonal strut, as shown
in Fig. 7.

The idealized portal frame is assumed to be located at the North gate of UCB campus,
where the latitude is 37.877◦ , longitude is −122.264◦ , and the site class is assumed as
NEHRP D. One of the commonly used IM, Sa(T1 ), is chosen as the intensity measure.
Hazard curve is obtained by using the Hazard Curve Calculator application of OpenSHA
[Field et al., 2003], using the above site coordinates and site class. The hazard curves
obtained for the bare (T1 = 0.50 sec) and infilled (T1 = 0.10 sec) frames are plotted in
Fig. 8. It is observed that the MAF of exceedance of Sa(T1 ) is less for the infilled frame
compared to the bare frame for almost all the considered Sa(T1 ) values.
100
T = 0.1 sec
Mean annual frequency of exceedance
T = 0.5 sec
10–2
10–4
10–6
10–8
10–2 10–1 100 101
Sa (g)
FIGURE 8 MAF of exceedance of Sa(T1 ) for the assumed site of the portal frame.

The bare and infilled cases of the portal frame, Fig. 7, are modeled using OpenSees [2010].
Columns, 3.43 m [11.25 ft] long with a 305 × 305 mm [12 × 12 in] square section and
the beam, 4.88 m [16 ft] long with a 343 × 267 mm [13.5 × 10.5 in] rectangular sec-
tion are modeled using force-based beam-column elements, i.e., Nonlinearbeamcolumn
[OpenSees, 2010], with fiber discretized sections. Column section contains eight #6 rein-
forcing bars, while the beam section is symmetrically reinforced with three #6 bars at
the top and bottom of the section. The material property values reported in [Hashemi and
Mosalam, 2006] are employed for concrete (fc for beam: 37 MPa [5.4 ksi]; for columns:
38 MPa [5.6 ksi]) and steel (fy : 458 MPa [67 ksi]), where core and cover concrete are
modeled using Concrete02 and longitudinal reinforcing steel is modeled with Steel01 in
OpenSees. Transverse reinforcement consisting of #3 bars spaced at 95 mm [3.75 in] and

70 mm [2.75 in] in columns and beam, respectively, are used to determine the confined core
concrete properties [Hashemi and Mosalam, 2006].
Twenty ground motions [Lee and Mosalam, 2006] are used in nonlinear time history
analyses, which are also used for the shearwall building discussed in the next section. These
ground motions are scaled for each of the considered Sa(T1 ) values for simplicity. However,
use of unscaled ground motions should be the preferred method in a real-life application.
For brevity, only uncertainty in ground motion is considered and material uncertainty is not
taken into consideration. Therefore, the total number of analyses conducted for an intensity
level is 20.
Peak interstory drift ratio (IDR) and peak roof acceleration (RA) are considered as
the EDPs. For each IM value, number of ground motions leading to collapse is obtained
first. For this purpose, IDR value corresponding to collapse is determined from pushover
analysis and a ground motion applied at a specific intensity level is said to lead to collapse
if the peak IDR response of the ground motion is greater than the pre-determined collapse
IDR. This method is chosen for collapse determination rather than the global dynamic
instability approach, since there is no higher mode participation in the considered structure
and accordingly, the collapse IDR from pushover analysis provides a representative col-
lapse IDR value for nonlinear time history analyses. The probability of global collapse is
calculated by dividing the number of analyses leading to collapse by the total number of
conducted analyses. Probability of no global collapse is obtained by subtracting the prob-
ability of global collapse from 1.0. Probabilities of collapse and “no collapse” are plotted
in Fig. 9 for the bare and infilled frames, where collapse probability is much less for the
infilled frame case for all intensity levels. However, it is important to state that this finding is
specific for this portal frame. In a multistory, three-dimensional (3D) frame, the sudden fail-
ure of infill walls can lead to weak stories, which is usually followed by a global collapse.
In addition, sudden shear failure, another factor that can lead to collapse, is not modeled
in this illustrative application. Shear failure can be critical for the columns because of the
lateral component of the force transferred by the infill wall [Mosalam and Günay, 2010].
Therefore, a more realistic 3D modeling of a multistory frame with infill walls, which is
beyond the scope of this article, is expected to provide a different outcome, especially for
higher intensity levels.
For each Sa(T1 ) value, median and COV are calculated for both EDPs from the anal-
yses with no global collapse and the probability distribution (assumed lognormal) and the
POE of IDR and RA are determined. Probability and POE of RA of the infilled frame are
shown in Fig. 10 for different values of Sa(T1 ). It should be noted that the number of “no
collapse” cases decreases for larger IM values and can be as small as four (Sa = 2.0 g of
the bare frame). Use of such small number of data can introduce bias in the calculation of
0.9
Probability of Collapse and "No collapse"

0.8
0.7
0.6 Collapse: infilled frame

No collapse: infilled frame
0.5 Collapse: bare frame
No collapse: bare frame
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2
Sa (g)
FIGURE 9 Probability of global collapse and “no collapse” for bare and infilled frames.
0.1
Sa = 0.6g
Sa = 1.1g
Probability
Sa = 1.6g
0.05
0
0 0.5 1 1.5 2 2.5 3
1
Probability of Exceedance
Sa = 0.6g
0.8
Sa = 1.1g
0.6 Sa = 1.6g
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3
RA (g)
FIGURE 10 Probability and POE of RA of the infilled frame.
median and COV values, hence should be avoided whenever possible. However, it should
also be noted that the contribution of the “no collapse” cases to the probability calculation
decreases in such case (Eq. (4b)). Accordingly, the influence of the error introduced by
obtaining the statistical parameters from an insufficient number of data points is reduced.
However, a sensitivity study to quantify this effect is beyond the scope of this article.
4.3. Combination of Hazard and Structural Analyses

As stated previously, differences in the path of achieving the same peak value introduce
differences in the observed damage and these differences contribute to the variance of the
DM corresponding to an EDP. However, for this portal frame example, this variance is
expected to be minimal due to the small level of redundancy and redistribution. In addition,
for this idealized example, the loss analysis is not meaningful. Therefore, damage and loss
analyses are not conducted in this application. However, following the total probability
theorem, it is possible to combine the hazard and structural analyses with Eq. (7). Replacing
IM with Sa(T1 ) and EDP with RA and IDR, combination equations specific to this example
are represented with Eqs. (8) and (9), respectively:

P EDPi IMm p(IMm )
P EDPi = (7)
m
i
P RAi = P RA Sam p(Sam ) (8)
m

P IDRi = P IDR i Sam p(Sam ) . (9)
m
POE of RA and IDR, calculated with Eqs. (8) and (9), are plotted in Figs. 11 and 12,
respectively, for both bare and infilled frames. From Fig. 11, the POE of RA of the infilled
frame is observed to be larger than that of the bare frame. For small values of RA, this is
the case because of the initial periods of the structures (0.1 s for the infilled frame vs. 0.5 s
for the bare frame), where the acceleration response for 0.1 s is greater than that for 0.5 s
for the employed ground motions. For the larger RA, the peak response is dictated by the
lateral force capacity, which is larger for the infilled frame compared to the bare frame. It is
observed that the POE of RA of the two frames becomes closer to each other as the RA
1
Bare Frame
0.9 Infilled Frame
0.8
Probability of Exceedance of RA
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5 3
RA (g)
FIGURE 11 POE of RA for the portal frame.

1
Bare Frame
0.9 Infilled Frame
Probability of Exceedance of IDR

0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5 3 3.5
IDR (%)
FIGURE 12 POE of IDR for the portal frame.
increases (i.e., RA greater than 1.0 g), which is mainly because of the probability of Sa.
As observed from Fig. 8, p(Sa) of the infilled frame is smaller than that of the bare frame for
Sa greater than 0.4 g with the difference increasing as Sa increases. The POE of a large RA
value is mostly affected by the contribution from large intensities, i.e., large Sa. Although
the POE of a large value of RA of the infilled frame for a given value of Sa, P(RAi |Sam ),
is larger for a large Sa, the contribution from a large value of Sa, P(RAi |Sam )p(Sam ), to
the POE of RA, P(RAi ), becomes closer to that of the bare frame. This is explained by
the fact that the probability of Sa, p(Sam ), considered as a weighing factor, is smaller for
the infilled frame for a large value of Sa. This explanation provides an example of the
benefit of combining different analyses stages. If the hazard analysis was not combined
with the structural analysis, results of structural analysis alone would indicate larger POE
of the RA for the infilled frame than that for the bare frame for larger intensities. However,
combination of the two analyses indicates that the POE values of the RA for the bare and
infilled frames are comparable for large intensities.
Figure 12, where the POE of IDR is plotted, indicates that the POE of IDR of the bare
frame is much larger than that of the infilled frame, because of the significant contribution
of the infill wall in reducing the frame deformation response. However, as previously dis-
cussed, this result is specific to the portal frame analyzed in this application and the adopted
modeling assumptions. Therefore, such finding should not be generalized.
5. Application of PEER Formulation: UCS Building

In this section, PEER PBEE methodology is applied in detail to the University of
California Science (UCS) building located on the UCB campus. A MATLAB [Mathworks,
2010] script is developed to combine results from hazard, structural, damage, and loss
analyses as defined by Eq. (4). The considered building is a modern RC shear wall build-
ing which includes high technology research laboratories for organismal biology. Besides
the research laboratories, the building contains animal facilities, offices and related service
support spaces. The building consists of six stories and a basement. It is almost rectangu-
lar in plan with overall dimensions of approximately 93.27 m [306 ft] in the longitudinal
(north-south) direction and 32 m [105 ft] in the transverse (east-west) direction [Comerio,
2005]. A RC space frame carries the gravity loads of the building, and coupled shear walls
and perforated shear walls support the lateral loads in the transverse and the longitudinal
directions, respectively, as shown in Fig. 13. The floors consist of waffle slab systems with
solid parts acting as integral beams between the columns. The waffle slab is composed of a
114 mm [4.5 in] thick RC slab supported on 508 mm [20 in] deep joists in each direction.
The foundation consists of a 965 mm [38 in] thick mat.
This building is an example for which the non-structural components contribute to the
PBEE methodology in addition to the structural components due to the valuable building
contents, i.e., the laboratory equipment and research activities. Detailed information about
the content inventory and their importance can be found in [Comerio, 2005].

The UCS building, which is located in the southwest quadrant of the UCB campus, is
within a distance of 2 km [1.243 mile] from the Hayward fault [Comerio, 2005], an active
strike-slip fault with an average slip rate of 9 mm/yr [0.354 in/yr]. The latest rupture of its
southern segment (Fremont to somewhere between San Leandro and Berkeley) occurred on
October 21, 1868, producing a magnitude 7 earthquake. Frankel and Leyendecker [2001]
provided the MAF of exceedance of Sa at periods of 0.2, 0.3, and 0.5 s and B-C soil bound-
ary as defined by the International Building Code [International Code Council, 2000] for
the latitude and longitude of the UCS building site. Lee and Mosalam [2006] assumed a
lognormal distribution of Sa with the mean of 0.633 g and standard deviation of 0.526 g,
which is a good fit for the POE of Sa for t = 50 years obtained from Eq. (1) using the MAF
of exceedance suitable for the period (T = 0.38 s) and local site class C of the building (see
Fig. 14). Probability p(IMm ) in Eq. (4a) and POE values for discrete values of Sa between
0.05 g and 4.0 g with 0.05 g increments are shown in Fig. 15.

Although it is possible to select several damageable groups, only two of these groups
are considered for the UCS building for brevity. These are (1) structural components
FIGURE 13 Plan view of the UCS building [Lee and Mosalam, 2006].
100
Mean annual frequency of exceedance

10–1
10–2
10–3
10–4
10–5
10–2 10–1 100 101
Sa (g)
FIGURE 14 MAF of exceedance of Sa for the site of the UCS building.
0.08
Probability of Sa
0.06
0.04
0.02
0
Probability of Exceedance of Sa
0 0.5 1 1.5 2 2.5 3 3.5 4

1
0.8
0.6
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Sa (g)
FIGURE 15 Probability and POE of Sa in 50 years for the UCS building site.
and (2) non structural components. Maximum peak interstory drift ratio along the height
(MIDR) and peak roof acceleration (RA) are considered as the EDPs. Lee and Mosalam
[2006] conducted nonlinear analyses of the building using 20 ground motions, which are
selected as ground motions that have the same site class as the building site and distance
to a strike-slip fault similar to the distance of the UCS building to Hayward fault. Forty
different scales of these ground motions are employed corresponding to Sa (T1 ) values
between 0.1 g and 4.0 g with 0.1 g increments. For each value of Sa, lognormal distribution
is assumed for both of the considered EDPs with the obtained median and COV values
0.02 2.5
Median PRA (g)

0.015
Median MIDR
1.5
0.01
1
0.005
0.5
0 0
0 1 2 3 4 0 1 2 3 4
Spectral Acceleration (g) Spectral Acceleration (g)
Fit: all data

1.5 0.5
Fit: partial data
Data
COV MIDR
COV PRA 0.25
0.5
0 0
0 1 2 3 4 0 1 2 3 4
Spectral Acceleration (g) Spectral Acceleration (g)
FIGURE 16 Regression of median and COV data from [Lee and Mosalam, 2006].
TABLE 1 Spectral acceleration corresponding to different POE values for the UCS
building site
POE (%) 90 80 70 60 50 40 30 20 10
Sa (g) 0.18 0.25 0.32 0.39 0.47 0.57 0.71 0.90 1.39
plotted in Fig. 16. In addition to the median and COV values obtained directly from struc-
tural analyses, fitted linear or quadratic relationships are also plotted in Fig. 16. The solid
lines represent functions fitted to all of the calculated values, while dashed lines represent
functions fitted to a subset of the calculated values of Sa, where only ten Sa values are con-
sidered corresponding to POE of 10–90% with 10% increments from the hazard analysis,
as presented in Table 1 [Lee and Mosalam, 2006]. Effect of such fitting on the loss curve is
discussed later.
Probability of MIDR and RA are plotted in Fig. 17 for example values of Sa = 0.5 g,
2.0 g, and 3.0 g. These probabilities correspond to p(EDPj i |IMm ) in Eq. (4d). Cumulative
distributions of MIDR and RA for the same Sa values, obtained by the cumulative
summations of the probabilities, are plotted in Fig. 18.
Probability of global collapse is determined by using the probability distribution of
MIDR obtained from the structural analysis for each Sa. The median global collapse MIDR
is accepted as 0.018 based on the study by Hwang and Jaw [1990], who determined the
median capacity and logarithmic standard deviation of different limit states by analyzing
the test data of shear wall specimens. Probability of global collapse for each IM value is cal-
culated by summing up the probabilities of MIDR values greater than the median collapse
0.12
Sa = 0.5 g
0.1
Probability of MIDR
Sa = 1.0 g
0.08
Sa = 3.0 g
0.06
0.04
0.02
0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045
MIDR
0.03
Sa = 0.5 g
Probability of RA
Sa = 1.0 g
0.02 Sa = 3.0 g
0.01
0
0 0.5 1 1.5 2 2.5 3 3.5 4
RA (g)
FIGURE 17 Probability distributions of MIDR and RA of the UCS building.
1
Cum. Distr. of MIDR
Sa = 0.5 g
0.75 Sa = 1.0 g
Sa = 3.0 g
0.5
0.25
0
0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04 0.045
MIDR
1
Sa = 0.5 g
Cum. Distr. of RA
0.75 Sa = 1.0 g
Sa = 3.0 g
0.5
0.25
0
0 0.5 1 1.5 2 2.5 3 3.5 4
RA (g)
FIGURE 18 Cumulative distributions of MIDR and RA of the UCS building.
MIDR, i.e., shaded area in Fig. 19, for that IM. The resulting values of the probability of
global collapse and “no collapse” are plotted in Fig. 20. It is observed that the collapse and
“no collapse” probabilities do not follow smooth trends with Sa, but they are jagged. This
is attributed to sudden change of probability distributions between consecutive intensity
Median global
Probability of MIDR
collapse MIDR
Probability of
global collapse
0
0
MIDR
FIGURE 19 Probability of global collapse from the MIDR probability distribution.
0.9
Probability of Collapse and "No Collapse"
0.8 No Collapse
Collapse
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Sa (g)
FIGURE 20 Probability of global collapse and “no collapse” for the UCS building.
levels. As observed from Fig. 16, median of MIDR follows a smooth trend. On the other
hand, the COV shows sudden variations resulting in the mentioned sudden change. Since
the collapse probability is calculated using the probability distribution, sudden change of
these probability distributions results in jagged collapse and “no collapse” probabilities.

Fragility functions are obtained for the two damageable groups in damage analysis.
Damage levels considered for the structural components are slight, moderate, and severe
damage. On the other hand, damage levels of the non structural components are based
on the maximum sliding displacement experienced by the scientific equipment relative to

its bench-top surface [Chaudhuri and Hutchinson, 2005]. Sliding displacements of 5 cm
[0.2 in] and 10 cm [0.4 in] are accepted as the two damage levels for the non structural
components.
The probability of a damage level given a value of the EDP, p(DMk |EDPj i ), is assumed
to be lognormal. Median and COV values for the damage levels of structural and non-
structural components of the UCS building are shown in Table 2. Values for the structural
components are based on the work of Hwang and Jaw [1990] and those for the non
structural components are obtained from the study by Chaudhuri and Hutchinson [2005].
Resulting fragility curves for structural and non structural components are shown in Fig. 21.
5.4. Loss Analysis

Monetary loss is chosen as the DV for the UCS building application. The loss functions
are derived from the available reports on the considered building assuming that the proba-
bility distribution of monetary loss for a damage level is lognormal. The total value of the
TABLE 2 Median and COV of EDPs for different damage levels of the UCS building
Component Damage level EDP Median COV
Structural Slight MIDR 0.005 0.30
Moderate MIDR 0.010 0.30
Severe MIDR 0.015 0.30
Non structural DM = 5 cm PRA (g) 0.005 0.35
DM = 10 cm PRA (g) 0.010 0.28
1 1
0.9 Structural 0.9 Non structural

Probability of Exceedance of Damage
Components Components
0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 Slight damage 0.2

Moderate damage DM = 5 cm
0.1 0.1
Severe damage DM = 10 cm
0 0
0 0.01 0.02 0.03 0.04 0.05 0 1 2 3 4
MIDR PRA (g)
FIGURE 21 Fragility curves for structural and non structural components of the UCS
building.
scientific equipment is estimated to be $23.00 million [Comerio, 2005]. Median values cor-
responding to the damage levels of 5 cm and 10 cm of sliding displacements are assumed
to be $6.90 million (30% of the total value) and $16.10 million (70% of the total value),
respectively. The COV is assumed to be 0.2 for both of these non structural component
damage levels.
There is no available information about the monetary losses related to the structural
components. However, since the contents damage has more significance for the building
relative to the structural damage, median monetary losses for the slight, moderate, and
severe damage levels are assumed to be $1.15 million, $3.45 million, and $6.90 million,
respectively, which correspond to 5%, 15%, and 30% of the total value of the non structural
components. The COV is assumed to be 0.4 for all the damage levels, which is larger
than the COV for the non-structural components since a larger variation is expected due
to the lack of information. Resulting loss functions for the structural and non structural
components are shown in Fig. 22.
5.5. Determination of Loss Curve

In addition to the conducted four analyses, determination of the loss curve requires the
knowledge of the POE of the monetary loss in case of global collapse, P(DV|C), in Eq. (4b).
The probability of the monetary loss in case of global collapse is assumed to be lognormal
with the median of $30 million, which corresponds to the total value of structural and non
structural components, and COV of 0.2. The resulting loss function is shown in Fig. 23
together with the loss functions for the damage levels of the structural damageable group,
shown previously in Fig. 22. The difference between the loss function for collapse and
other damage levels emphasizes the importance of the non structural building contents of
the UCS building.
1 1
Slight damage DM = 5 cm
Probability of Exceedance of Economic Loss
0.9 Moderate damage 0.9 DM = 10 cm

Severe damage
0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
Structural Non structural
0.1 Components 0.1 Components
0 0
0 10 20 30 40 0 10 20 30 40
Economic Loss (million $) Economic Loss (million $)
FIGURE 22 Loss functions for structural and non structural components of the UCS
building.
0.9
0.8
0.7
0.6
0.5
0.4
Slight damage
0.3
Moderate damage
0.2 Severe damage
Collapse
0.1
0
0 5 10 15 20 25 30 35 40
Economic Loss (million $)
FIGURE 23 Loss functions of structural damageable group including collapse of the UCS
building.
0.04
Total
0.035 Collapse
No collapse
0.03
0.025
0.02
0.015
0.01
0.005
0
0 5 10 15 20 25 30 35 40
FIGURE 24 UCS building loss curve with collapse and “no collapse” deaggregated
contributions.
The resulting total loss curve obtained using Eq. (4) is plotted in Fig. 24. The POE of
the monetary loss is deaggregated to the POE due to global collapse and “no collapse” cases
in this figure. It is observed that the global collapse case is more dominant on the total loss
curve and all the loss is attributed to the global collapse case for monetary losses greater
than $25 million. The “no collapse” plot can be interpreted as the loss curve for a hypo-
thetical case where collapse is prevented for all intensity levels. The significant reduction
in economic loss as a result of the elimination of collapse shows the effect of the “collapse
prevention” mandated by the seismic codes from an economical perspective.
The effect of the uncertainty introduced by fitting functions to the statistical descrip-
tors of the calculated EDPs, as shown in Fig. 16, is investigated in Fig. 25. The case
where the functions are fitted to the median and COV corresponding to only ten Sa val-
ues represents a situation where all analyses results are not available and extrapolation
is used to represent the missing data. This case results in underestimated loss curve and
indicates that extrapolation of structural analyses results should be avoided. Although dif-
ference is smaller, the case where functions are fitted to all the median and COV values also
results in underestimation. Considering that this underestimation can lead to an incorrect
decision-making about the considered facility, it is preferable to obtain the structural anal-
ysis outcome directly as a result of the analyses conducted for all the considered intensity
levels in order to use the PEER PBEE methodology in an effective and accurate manner.
6. Summary and Concluding Remarks

In an attempt to disseminate probabilistic performance-based seismic design to a larger
audience, PEER PBEE methodology is summarized and explained in a simplified manner
to facilitate its appreciation and adoption by practicing engineers. The application of the
method, as presented in this simplified manner, is demonstrated using two structures: an
idealized portal frame with and without infill wall and an existing shear wall building on
the UCB campus. Concluding remarks are stated as follows.
0.04
Data
0.035 Fit: all data
Fit: partial data
0.03
0.025
0.02
0.015
0.01
0.005
0
0 5 10 15 20 25 30 35 40
FIGURE 25 Effect of smoothing structural analysis results of the UCS building on the
loss curve.
● PEER PBEE methodology is based on the total probability theorem, which requires
only an elementary knowledge of probabilistic concepts and therefore it can be
easily adopted by practicing engineers.
● Effects of the combination of different analyses are demonstrated, where the incor-
poration of hazard analysis with the structural analysis improves the determination
of POE of EDP.
● Among the various factors that determine the performance of a structure subjected
to earthquake hazard, a structural engineer has the main control on the structural
design. PEER PBEE methodology provides a powerful tool to the structural engineer
to evaluate the design at hand with the other influencing factors, such as the hazard or
loss, which are usually known by the engineer but without control. Hence, a designer
can improve the design by considering all the influencing factors in an integrated
manner, an example of which is mentioned for the case of collapse prevention.

● A structural engineer can utilize PEER PBEE methodology to introduce
performance-based enhancements to the traditional code-based designs.
● PEER PBEE methodology provides an effective tool not only for the conventional
structural types, such as moment resisting frames with unreinforced masonry infill
walls, but also for the innovative and sustainable design and retrofit methods such
as base isolation, rocking foundations, and self-centering systems. Accordingly, a
structural engineer can employ PEER PBEE methodology for the evaluation of the
performance of an existing conventional structural type and the outcome of different
retrofit measures.
● A structural engineer can make use of the PEER PBEE methodology directly as
a design tool, e.g., for decision making among different design alternatives. This
article provides a contribution towards more common use of the method as a direct
design tool.
Acknowledgments
The authors thank Mr. L. Lombard for the assistance in the portal frame application. The
efforts of Prof. Tae-Hyung Lee, Department of Civil Engineering, Konkuk University,
Korea, in the analysis of the UCS building are greatly appreciated. The authors would
like to express their gratitude to Prof. P. Gardoni, Department of Civil and Environmental
Engineering, University of Illinois, Urbana-Champaign, for his careful review of an earlier
version of the article.
References
American Society of Civil Engineers (ASCE) [2000] Prestandard and Commentary for the Seismic
Rehabilitation of Buildings, Report No. FEMA-356, Washington, D.C.
Applied Technology Council (ATC) [1996] Seismic Evaluation and Retrofit of Concrete Buildings,
Report No. ATC-40, Volume 1–2, Redwood City, California.
Applied Technology Council (ATC) [2013] “Development of next generation performance-based
seismic design procedures for new and existing buildings.” Retrieved from https://www.atcouncil.
org/Projects/atc-58-project.html.
Beck, J. L., Porter, K. A., Shaikhutdinov, R., Moroi, T., Tsukada, Y., and Masuda, M. [2002]
Impact of Seismic Risk on Lifetime Property Values, Final Report, CUREE-Kajima Joint Research
Program Phase IV Consortium of Universities for Earthquake Engineering Research, Richmond,
California.
Bohl, A. [2009] “Comparison of performance based engineering approaches,” Master thesis, The
University of British Columbia, Vancouver, Canada.
Bommer, J. J. and Abrahamson, N. A. [2006] “Why do modern probabilistic seismic-hazard analyses

often lead to increase hazard estimates?,” Bulletin of Seismological Society of America 96(6),
1967–1977.
Building Seismic Safety Council [1997] NEHRP Guidelines for the Seismic Rehabilitation of
Buildings, Report No. FEMA-273, Federal Emergency Management Agency, Washington, D.C.
Chaudhuri, S. M. and Hutchinson, T. C. [2005] Performance Characterization of Bench- and Shelf-
Mounted Equipment, Pacific Earthquake Engineering Research Center, PEER, Report No. PEER
2005/05.
Comerio, M. C., Ed. [2005] PEER Testbed Study on a Laboratory Building: Exercising Seismic
Performance Assessment, Pacific Earthquake Engineering Research Center, PEER, Report No.
PEER 2005/12.
Cornell, C. A. and Krawinkler, H. [2000] “Progress and challenges in seismic performance
assessment,” PEER News, April.
Field, E. H., Jordan, T. H., and Cornell, C. A. [2003] “OpenSHA: A developing community-modeling
environment for seismic hazard analysis,” Seismological Research Letters 74(4), 406–419.
Frankel, A. and Leyendecker, E. V. [2001] Uniform Hazard Response Spectra and Seismic Hazard
Curves for the United States, US Geological Survey, Menlo Park, California.
Goulet, C., Haselton, C. B., Mitrani-Reiser, J., Deierlein, G. G., Stewart, J. P., and Taciroglu, E. [2006]
“Evaluation of the seismic performance of a code-conforming reinforced-concrete frame build-
ing – part I: ground motion selection and structural collapse simulation,” 8th National Conference
on Earthquake Engineering (100th Anniversary Earthquake Conference), San Francisco, CA,
April 18–22.
Günay, M. S. and Mosalam, K. M. [2010] Structural Engineering Reconnaissance of the April 6,
2009, Abruzzo, Italy, Earthquake, and Lessons Learned, Pacific Earthquake Engineering Research
Center, PEER, Report No. 2010/105.
Hashemi, A. and Mosalam, K. M. [2006] “Shake-table experiment on reinforced concrete struc-
ture containing masonry infill wall,” Earthquake Engineering and Structural Dynamics 35(14),
1827–1852.
Hwang, H. H. M. and Jaw, J.–W. [1990] “Probabilistic damage analysis of structures,” Journal of
Earthquake Engineering 116(7), 1992–2007.
Holmes, W. [2010] “Reconnaissance report on hospitals,” (oral presentation), Chile EERI/PEER
Reconnaissance Briefing at UC Berkeley, March 30, Berkeley, California.
Insight Software Solutions [2013] MacroExpress. Retrieved from http://www.macros.com
International Code Council [2000] International Building Code 2000, International Conference of
Building Officials, Whittier, California, pp. 756.
Kramer, S. [1996] Geotechnical Earthquake Engineering, Prentice-Hall, Upper Saddle River, NJ.
Krawinkler, H. [2002] “A general approach to seismic performance assessment,” Proceedings of
International Conference on Advances and New Challenges in Earthquake Engineering Research,
ICANCEER, Hong Kong, August 19–20.
Krawinkler, H. and Miranda, E. [2004] “Performance-based Earthquake Engineering,” in Earthquake
Engineering: from Engineering Seismology to Performance-based Engineering, eds. Y. Bozorgnia
and V. V. Bertero (CRC Press, Boca Raton, FL), Ch. 9.
Krawinkler, H., Ed. [2005] Van Nuys Hotel Building Testbed Report: Exercising Seismic Performance
Assessment, Pacific Earthquake Engineering Research Center, PEER, Report No. PEER 2005/11.
Kunnath, S. K. [2006] Application of the PEER PBEE Methodology to the I-880 Viaduct, Pacific
Earthquake Engineering Research Center, PEER, Report No. 2006/10.
Kunnath, S. K., Larson, L., and Miranda, E. [2006]. “Modeling considerations in probabilistic
performance-based seismic evaluation: case study of the I-880 viaduct,” Earthquake Engineering
and Structural Dynamic 35(1), 57–75.
Lee, T.-H. and Mosalam, K. M. [2006] Probabilistic Seismic Evaluation of Reinforced Concrete
Structural Components and Systems, Pacific Earthquake Engineering Research Center, PEER,
Report No. 2006/04.
MathWorks Inc. [2010] MATLAB - Version 2010b.
McKenna, F. [2010] Opensees User’s Manual. Retrieved from http://opensees.berkeley.edu.
Mitrani-Reiser, J., Haselton, C. B., Goulet, C., Porter, K. A., Beck, J., and Deierlein, G. G. [2006]
“Evaluation of the seismic performance of a code-conforming reinforced-concrete frame build-
ing - part II: loss estimation,” 8th National Conference on Earthquake Engineering (100th
Anniversary Earthquake Conference), San Francisco, California, April 18–22.
Moehle, J. P. [2003] “A framework for performance-based earthquake engineering,” Proceedings
of ATC-15-9 Workshop on the Improvement of Building Structural Design and Construction
Practices, Maui, HI, June.
Moehle, J. P. and Deierlein, G. G. [2004] “A framework for performance-based earthquake engi-
neering,” Proceedings of 13th World Conference on Earthquake Engineering, Paper No 679,
Vancouver, Canada.
Moehle, J. P. [2010] “27 March 2010 offshore Maule, Chile Earthquake,” (oral presentation), Chile
EERI/PEER Reconnaissance Briefing at UC Berkeley, March 30, Berkeley, California.
Mosalam, K. M. and Günay, M. S. [2010] “Seismic Retrofit of Non-Ductile Reinforced Concrete
Frames Using Infill Walls as a Rocking Spine,” in Advances of Performance-Based Earthquake

Engineering, ed. M. N. Fardis (Springer, Dordrecht) Ch. 33.
National Institute of Building Sciences and Federal Emergency Management Agency (NIBS and
FEMA) [2003] Multi-hazard Loss Estimation Methodology, Earthquake Model, HAZUS®MH
Technical Manual, Federal Emergency Management Agency, Washington, DC, 690 pp.
Petersen, M. D., Frankel, A. D., Harmsen, S. C., Mueller, C. S., Haller, K. M., Wheeler, R. L.,
Wesson, R. L., Zeng, Y., Boyd, O. S., Perkins, D. M., Luco, N., Field, E. H., Wills, C. J., and
Rukstales, K. S. [2008] Documentation for the 2008 Update of the United States National Seismic
Hazard Maps U.S. Geological Survey Open-File Report 2008-1128, 61 p.
Porter, K. A., Beck, J. L., and Shaikhutdinov, R. V. [2002] “Sensitivity of building loss estimates to
major uncertain variables,” Earthquake Spectra 18(4), 719–743.
Porter, K. A. [2003] “An overview of PEER’s performance-based earthquake engineering method-
ology,” Conference on Applications of Statistics and Probability in Civil Engineering (ICASP9),
Civil Engineering Risk and Reliability Association (CERRA), San Francisco, California, July
6–9.
Porter, K. A. [2010] “Cracking an Open Safe: Uncertainty in HAZUS-based seismic vulnerability
functions,” Earthquake Spectra 26(3), 893–900.
Power, M., Chiou, B., Abrahamson, N., Bozorgnia, Y., Shantz, T., and Roblee, C. [2008] “An
overview of the NGA project,” Earthquake Spectra 24(1), 3–21.
Rojas, H. A., Foley, C., and Pezeshk, S. [2011] “Risk-based seismic design for optimal structural and
nonstructural system performance,” Earthquake Spectra 27(3), 857–880.
SEAOC Vision 2000 Committee [1995] Performance-based Seismic Engineering, Structural
Engineers Association of California, Sacramento, California.
Sommerville, P. and Porter, K. A. [2005] “Hazard analysis,” in PEER Testbed Study on a Laboratory
Building: Exercising Seismic Performance Assessment, ed. M. C. Comerio, Pacific Earthquake
Engineering Research Center, PEER, Report No. PEER 2005/12, Ch. 3
Talaat, M. and Mosalam, K. M. [2009] “Modeling progressive collapse in reinforced concrete build-
ings using direct element removal,” Earthquake Engineering and Structural Dynamics 38(5),
609–634.

Peer P Bee Gun Aymo Salam

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Peer P Bee Gun Aymo Salam

Uploaded by

Copyright:

Available Formats

This article was downloaded by: [University of California, Berkeley]

On: 29 October 2013, At: 21:07

Journal of Earthquake Engineering

PEER Performance-Based Earthquake

To link to this article: http://dx.doi.org/10.1080/13632469.2013.787377

PLEASE SCROLL DOWN FOR ARTICLE

PEER Performance-Based Earthquake Engineering

SELIM GÜNAY and KHALID M. MOSALAM

A performance-based earthquake engineering (PBEE) methodology was developed at the Pacific

Received 23 May 2012; accepted 11 March 2013.

Occasional : unacceptable performance

FIGURE 1 Vision 2000 recommended seismic performance objectives for buildings.

PBED as a design method is justified by the consequences of recent earthquakes, where

is an extended version of the authors’ contribution to this task group, is an attempt to

2. PEER PBEE Formulation

Facility Definition: Location and Design

ground motions selected for : # of EDPs

Loss functions for individual

damageable groups of the facility

Decision variable (DV)

Combination of the Analyses Stages with Total Probability Theorem

Decision variable (DV)

Decision about Design and Location

FIGURE 2 Analysis stages of PEER PBEE methodology.

2.1. Hazard Analysis

exceedance in " t " years

Intensity measure (IM) Intensity measure (IM)

FIGURE 3 Correspondence between MAF of exceedance and POE of IM.

P(IM) = 1 − e−λ(IM) t , (1)

for m = 1 : # of IM data points

2.2. Structural Analysis

Calculate the parameters of the considered probability

FIGURE 4 Determination of the PDF of an EDP from structural analysis.

It is likely to observe global collapse at higher intensity levels. Global collapse is

2.3. Damage Analysis

2.4. Loss Analysis

1. Fatalities: Number of deaths as a direct result of facility damage.

FIGURE 5 Probability of exceedance, P, and probability, p, of a damage level from

2.5. Determination of the Loss Curve

P( DV n | IMm ) = P( DV n | NC, IMm ) p( NC| IMm ) + P( DV n | C) p( C| IMm ) (4b)

Decision variable (DV)

FIGURE 6 Loss functions for individual damageable groups of the facility.

where G represents the conditional probability and λ represents the MAF. It is to be

3. Guidance on Uncertainty Quantification

3.1. Hazard Analysis

3.2. Structural Analysis

3.3. Damage Analysis

3.4. Loss Analysis

4. Application of PEER Formulation: Portal Frame

4.1. Hazard Analysis

4.2. Structural Analysis

OpenSees. Transverse reinforcement consisting of #3 bars spaced at 95 mm [3.75 in] and

Probability of Collapse and "No collapse"

0.6 Collapse: infilled frame

FIGURE 10 Probability and POE of RA of the infilled frame.

4.3. Combination of Hazard and Structural Analyses

FIGURE 11 POE of RA for the portal frame.

Probability of Exceedance of IDR

FIGURE 12 POE of IDR for the portal frame.

5. Application of PEER Formulation: UCS Building

5.1. Hazard Analysis

5.2. Structural Analysis

Mean annual frequency of exceedance