Structural System Selection Using Performance-Based Design

Structural Engineering Research Frontiers © 2007 ASCE
Structural System Selection Using Performance-Based Design

Downloaded from ascelibrary.org by California State Polytechnic University, Pomona on 08/25/13. Copyright ASCE. For personal use only; all rights reserved.
Authors:
Farzin Zareian, Assistant Professor, University of California Irvine, Irvine, CA 92697,

zareian@uci.edu
Helmut Krawinkler, Professor of Civil Engineering, Stanford University, Stanford, CA
94305, krawinkler@stanford.edu
ABSTRACT
This paper focuses on providing a practical approach for Performance-Based Design
(PBD). The process for probabilistic Performance-Based Assessment (PBA), as
developed recently in various organizations (e.g., Pacific Earthquake Engineering
Research Center), implies that, given a building and its location, one can rigorously
calculate losses (e.g., expected monetary loss at a specified hazard level or mean annual
loss). This process is mathematically correct and powerful, but its implementation in
practice is in many cases too complex and impractical. Also, it is neither practical nor
efficient to use a rigorous probabilistic approach for conceptual Performance-Based
Design (PBD), in which the building first needs to be created before it can be evaluated.
In the simplified PBD approach proposed in this paper, performance objectives are
defined by placing limits on direct (monetary) loss, downtime loss, and life loss (or
tolerable probability of collapse). The proposed design process incorporates different
performance objectives up front, before the structural system is created, and assists
engineers in making informed decisions on the choice of an effective structural system
and its stiffness (period), base shear strength, and other important global structural
parameters. The tools needed to implement this design process are (1) hazard curves for a
specific ground motion intensity measure, (2) mean loss curves for structural and
nonstructural subsystems, (3) structural response curves that relate, for different structural
systems, a ground motion intensity measure to the engineering demand parameter (e.g.,
interstory drift or floor acceleration) on which the subsystem loss depends, and (4)
collapse fragility curves. The proposed process is illustrated in an example to
demonstrate its practicality.
INTRODUCTION
Performance-Based Earthquake Engineering is the art of applying scientific and
mathematical principles, along with experience and judgment, in order to either estimate
the consequences of an earthquake event on existing facilities (PBA) or provide society
with facilities that accommodate certain needs (expressed by owners, users, and society)
during and after an earthquake event (PBD).
Copyright ASCE 2007 Structures Congress 2007

Structural Engineering Research Frontiers
Development of methodologies for achieving the goals of PBEE has been the core of
research programs in many countries and research centers around the globe. The Pacific
Earthquake Engineering Research (PEER) Center has developed a detailed methodology
for PBA capable of predicting building performance in a probabilistic format. The
primary motivation for the study summarized in this paper was to develop a methodology
for PBD that is easy-to-use and can be implemented by engineers in their day-to-day
design practice. Throughout this paper, the detailed PBA and simplified PBD
methodologies are discussed. The simplified PBD methodology is exercised in the design
of an 8-story building to demonstrate its practicality.
PERFORMANCE-BASED EARTHQUAKE ENGINEERING

The work presented in this paper is based on the PBA methodology developed by
researchers of the Pacific Earthquake Engineering Research (PEER) center, which is
documented in many publications such as Cornell & Krawinkler (2000), Krawinkler
(2002), Deierlein (2004), etc. This PBA approach is general and can consider different
sources of uncertainty in estimating the performance of a building, however, it is rigorous
and involves elaborate formulation that makes it difficult to implement for ordinary
buildings. Also, this PBA methodology can hardly be used to explore design alternatives
to accomplish an efficient PBD. These drawbacks of the detailed PBA methodology were
the motivation to conduct research on the development of a simplified PBEE
methodology that can be used easily but with an acceptable level of accuracy (Zareian
2006).
The PEER PBA methodology is a rigorous process to quantify the performance of a
building and involves a chain of four random variables:
• Decision Variable (DV): is the representation of building performance and is a
quantitative expression of monetary loss, downtime loss, and life loss. DV is the
outcome of the PEER PBA methodology that can be computed at discrete hazard
levels or on a Mean Annual Frequency (MAF) basis.
• Damage Measure (DM): is the representation of the state of damage in building
components, such as cracks in partition walls or breakage of equipment. In order to
obtain relation between DMs and repair cost, DMs are determined based on a repair
strategy plus many other considerations, some explained in ATC-58 (2005).
• Engineering Demand Parameter (EDP): is a response parameter such as maximum
story drift, maxIDR, or peak floor acceleration, PFA. EDPs, for a given building and
location, are estimated through nonlinear response analysis using a representative
structural model and a set of ground motion records (Medina & Krawinkler, 2003). A
representative structural model should be capable of capturing important deterioration
modes in structural component properties while subjected to earthquake motion
(Ibarra & Krawinkler, 2005).
• Intensity Measure (IM): is the representation of ground motion intensity at a given
hazard level. Traditionally, scalar IMs such as spectral acceleration at the first mode
period of the structural system, Sa(T1), or peak ground acceleration, PGA, are used.
Recently, vector based IMs (Baker and Cornell, 2004; Tothong and Cornell, 2006)
have proven to be more efficient and sufficient (Luco, 2002) for estimating structural
response parameters at different hazard levels.
2
The PEER PBA methodology is completed by following a flow from IM to DV.

Equations (1) and (2) portray the mathematical format for quantification of DV at a
discrete hazard level (represented by IM) and on a MAF of exceedance basis,
respectively.
G ( DV | IM ) = ∫ ∫ G ( DV | DM ) dG ( DM | EDP ) dG ( EDP | IM ) (1)

EDP DM
λ DV ( DV ) = ∫ G ( DV | IM ) λ ( IM )
IM (2)
IM
In Equations (1) and (2), G functions represent complementary cumulative

distribution functions, and the function λ is the mean annual frequency of exceedance of
its corresponding variable. In other words, λDV(DV) and λIM(IM) are MAF of exceedance
of DV and IM, respectively. For instance, the average annual monetary loss for a given
building can be estimated as follows: IMs, such as Sa(T1), are determined from seismic
hazard analysis; relevant engineering demand parameters, EDPs, (e.g., story drifts, floor
accelerations, plastic deformations) are predicted from structural analysis for given
values of IMs and representative ground motions using a representative mathematical
model of the building; damage states of building components are developed from repair
strategies, and DM fragility curves are developed for each building component; and
finally, predictions are made on DVs using (1) and/or (2). This rigorous procedure has
been exercised, among others, on a 7-story reinforced concrete moment-resisting frame
building in Van Nuys, California (Krawinkler ed., 2005)
SIMPLIFIED APPROACH FOR PERFORMANCE-BASED DESIGN

In order to complete PBA, using (1) or (2), the building model needs to be available in
order to obtain G(EDP|IM). This methodology can not be used for conceptual PBD
because the building’s model is not yet available. One could perform an iterative PBA in
which an initial design is developed based on the engineer’s judgment and is refined in
several PBA iterations. This approach is not practical because the detailed PBA
procedure provides a probabilistic representation for losses based on probabilistic
representations of IMs, EDPs|IMs, DMs|EDPs, and DVs|DMs, each of which requires
extensive computational effort and comprehensive databases (Krawinkler, ed., 2005).
Furthermore, a judgmental initial design does not necessarily result in an efficient final
design even if it is modified through many detailed PBA iterations.
The simplified PBD proposed in this paper provides a solution to the aforementioned
shortcomings of rigorous PBA for designing buildings. In this approach, several
performance objectives (limits on DVs) are considered up front to select an efficient
structural system. The major structural parameters of the selected structural system are
proportioned using expected (mean) values of the random variables. This preliminary and
conceptual design can be considered as a good starting point for detailing the structural
3
components and may be followed by a detailed PBA for verification and fine-tuning.
Simplified PBD is easy to implement and is practical for design of common structures
that comprise most of the structural engineering work in consulting offices. A by-product
of the simplified PBD proposed here is a simplified PBA by which an estimate of DVs, in
the mean sense, for a structure can be provided without performing the detailed PBA.
To further facilitate the simplified PBD, a semi-graphical tool denoted as DDSS

(Design Decision Support System) is proposed (Zareian, 2006). This tool incorporates
mean values of all important variables in the PBA process (IM, EDP, DM, DV) and
provides a rapid assessment of relevant design alternatives (choice of structural system
and associated parameters). The DDSS comprises three domains: Hazard Domain, Loss
Domain, and Structural System Domain, as shown in Figure 1. Each domain is populated
with relationships between mean estimates of associated random variables. Due to the
significance of collapse phenomena in the decision-making and its effect on losses
(Krawinkler ed. 2005), the Structural System Domain and Loss Domain are partitioned
into a no-collapse (NC) and a collapse (C) sub-domain.
Hazard Structural System Domain

Domain No Collapse Collapse
Mean Hazard
Mean IM-EDP Curves Collapse Fragility Curves
Curve(s) for Design
for Design Alternatives for Design Alternatives
Alternatives
λ(IM) E ( EDP | IM & NC ) P ( C | IM )
Loss Domain
No Collapse Collapse
Mean Loss Curves Mean Loss due to Collapse

E ( Loss | EDP & NC ) E ( Loss | C )
-Monetary Loss =$loss -Downtime Loss = loss -Life Loss = D loss
FIGURE 1
FRAMEWORK AND SYMBOLIC REPRESENTATION OF SIMPLIFIED PBEE (KRAWINKLER ET AL.,
2005)
Hazard Domain
The Hazard Domain contains the relation between the ground motion intensity IM and its
mean annual frequency of exceedance λIM (i.e., hazard curve), plus associated
representative ground motion records. An essential characteristic of an IM is that it
should be feasible to find the relation between seismic hazard and the selected IM for the
location of the structure. This relation is the link between the seismicity of the structure’s
location and the IM. Such a relationship is obtained by performing Probabilistic Seismic
Hazard Analysis (Cornell, 1968). Usually, Sa(T1) is considered as the ground motion
intensity measure, however, several other scalar IMs and vector IMs have been
introduced recently, as was mentioned earlier.
4
The selected ground motion records link the Hazard Domain to the Structural System
Domain as they are used to obtain the IM-EDP relationships and the IM level at which
the building collapses (denoted as collapse capacity) for each individual record. These
ground motions should be representative of the seismicity of the structure’s location and
of the hazard level at which design or assessment is performed. An important aspect of
ground motion selection is the potential impact of special ground motions such as near-
fault records (Alavi & Krawinkler, 2001) and soft soil records.
Loss Domain
The Loss Domain is divided into two sub-domains, one containing losses conditioned on
collapse not occurring (i.e., NC sub-domain), and the other containing losses conditioned
on collapse (i.e., C sub-domain). The NC sub-domain contains the relations between
induced losses and structural response parameters that correlate well with these losses.
Obtaining these relations involves consideration of two sets of information: 1)
relationships between structural response parameters of building components and
corresponding damage states (component damage fragility curves), and 2) relations
between building component damage states and the component loss (component loss
functions). For a given value of EDP of a building component, the value of loss in that
component is obtained by integrating the associated losses weighted by the probability of
being in different damage states conditioned on the value of the response parameter.
We assume that the total loss conditioned on collapse not happening can be
disaggregated into losses in different subsystems. A subsystem is an assembly of building
components that contribute significantly to building total loss and in which loss is
sensitive to a common EDP. We also assume that the loss in each subsystem is
independent of the loss in any other subsystem, and that the total loss can be obtained as
the sum of the losses in the subsystems. For example, the monetary loss of a building can
be disaggregated into the following three subsystems: Nonstructural Drift Sensitive
Subsystem (NSDSS) (e.g., partitions), Nonstructural Acceleration Sensitive Subsystem
(NSASS) (e.g., much of the content), and Structural Subsystem (SS) (e.g., beams and
columns). The choice/art of dividing the building into different subsystems is based on
many parameters such as the functional use of the subsystems, spatial distribution of loss
in the building, sensitivity of subsystem loss to a common EDP, and the availability of
data.
In order to develop mean loss curves for a subsystem, E[loss|EDP&NC], damage
fragility curves have to be developed for each damage state k of component j in the
subsystem, as a function of the appropriate EDP, P(DSjk|EDP&NC). Presuming that the
expected cost of repair for each damage state is known, E[lossj| DSjk&NC], the total loss
in the subsystem, given EDP and conditioned on collapse not occurring is computed by
summing the weighted expected losses for all components of the subsystem as shown in
(3). The value of building loss conditioned on collapse is simply the value of the building
plus the cost of demolition/clean-up and design of a new building.
n mj
E[loss | EDP & NC] = ∑∑E ⎡⎣loss j | DS jk & NC⎤⎦ ⋅ P( DS jk | EDP & NC) (3)
j =1 k =1
5
Structural System Domain
The Structural System Domain contains the information about EDPs and the probability
of collapse for selected design alternatives. It provides the link between the Hazard
Domain and the Loss Domain. In general, the Structural System Domain is the domain in
which decisions are made on a suitable structural system and structural system
parameters. The effect of such decisions at discrete hazard levels (i.e., link to Hazard
Domain) is observed in the Loss Domain (i.e., link to Loss Domain). In order to complete
the link between the sub-domains of the Loss Domain and the Hazard Domain, the
Structural System Domain is also divided into a NC sub-domain (No-Collapse sub-
domain of the Structural System Domain) and a C sub-domain (Collapse sub-domain of
the Structural System Domain).
The NC sub-domain of the Structural System Domain includes information about
relations between building subsystem EDPs and the IM. For each subsystem, mean
(expected) IM-EDP curves for various design alternatives are presented. The IM is the
intensity measure employed in the Hazard Domain and the EDP is the one that correlates
well with the loss in a specific subsystem. Mean IM-EDP curves are obtained by
subjecting structural system alternatives with specific properties to sets of ground
motions representative of specific IM values, as provided in the Hazard Domain. If it can
be assumed that the frequency content of the ground motions is insensitive to magnitude
and distance within the IM range of primary interest, then Incremental Dynamic Analysis
(IDA) (Vamvatsikos & Cornell, 2002) can be used to obtain mean IM-EDP curves.
Information on mean IM-EDP curves has been summarized in (Zareian, 2006) for a
comprehensive set of generic moment-resisting frames and generic reinforce concrete
shear walls for a set of 40 ordinary ground motions. In this study, structures from 4 to 18
stories are considered, with periods equal to 0.1N, 0.15N, and 0.2N for moment-resisting
frames, and 0.1N, 0.075N, and 0.05N for shear wall structures. Three variation of base
shear coefficient γ = Vy/W is considered for each combination of number of stories and
building fundamental period. For moment-resisting frames, three variations of strength
and stiffness along the height, and three cases of relative column strength to beams
strength, are considered. Structural components of the generic structures are modeled by
elastic elements and rotational springs that describe the inelastic moment-rotation
behavior at plastic hinge locations. The hysteretic properties of these rotational springs
incorporate stiffness and strength deterioration properties of the type discussed in (Ibarra
et al. 2005).
Figure 2A shows the process for obtaining mean IM-EDP curve for an 8-story generic
shear wall structure subjected to the set of 40 ground motions. In this example, the EDP
is the maximum roof drift ratio (RDR), and the intensity measure is Sa(T1). Individual
gray lines show the IM-EDP curve for each of 40 ground motions and the black line
shows the mean of EDP|IM conditioned on no collapse. The solid circles show the last
point for which analysis has converged prior to collapse for individual ground motions.
The increase in the slope of the mean IM-EDP curve at large IM values is due to the
exclusion of cases in which collapse has occurred at smaller IM values.
The collapse sub-domain of the structural system domain includes collapse fragility
curves for design alternatives. The process for obtaining collapse fragility curves is
shown in Figure 2B for the same 8-story generic shear wall structure. Collapse fragility
6
curves are developed by fitting a cumulative lognormal distribution function to the

projected last converged points of the IDA curves on the vertical axis as shown at the
right side of Figure 2B (note the difference between Figure 2A and Figure 2B is the scale
of the IM axis). The collapse fragility curve is shown with a solid gray line at the right
side of Figure 2B (Zareian and Krawinkler, 2007a).
EDP-IM Curves (SW)

N = 8, T = 0.8,γ = 0.25, Stiff = Unif. Str. = -0.05My,base , ξ = 0.05
Obtaining the collapse fragility curve (SW)
θp = 0.02, θpc/θp = 1, λ = 20, Mc/My = 1.1 N = 8, T = 0.8, γ = 0.25, Stiff = Unif. Str. = -0.05My,base, ξ = 0.05
θp = 0.02, θpc/θp = 1, λ = 20, Mc/My = 1.1
2
7
Individual IDA curves
Mean of no-collapse
6
last point before
1.5 collapse
Individual IDA curves 5 Projection of last point
last point used before before collapse
IM[Sa(T1)]
IM[Sa(T1)]
collapse
4 Collapse fragility curve
1
Mean no-collapse
3
2
0.5
1
0 0
0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05
EDP[max.RDR] EDP[max.RDR]
(A) (B)
FIGURE 2
IMPLEMENTATION OF IDA TO OBTAIN: A) MEAN IM|EDP CURVE, B) COLLAPSE FRAGILITY
CURVE
Framework for a Design Decision Support System
The main focus of conceptual PBD is to provide the designer with an effective
combination of structural system and component parameters based on performance
objective that are defined at discrete hazard levels. Such a process is illustrated
schematically in Figure 3A. In this figure, for simplicity, only one subsystem is shown in
the NC sub-domain of the structural system and loss domains. Losses in different
subsystems can be assessed simultaneously, or the focus could be placed on the one
subsystem that contributes most to the loss value of the building, and other subsystem
losses could be evaluated subsequently.
The lower central portion of Figure 3A shows the mean loss curve for the single
subsystem conditioned on collapse does not occur, E(loss | EDP & N C ), and the lower
right portion shows the expected value of loss conditioned on collapse occurs, E(loss | C).
The upper left portion of Figure 3A shows the mean hazard curve for the building period
and location. The upper central portion shows mean IM-EDP curves for several design
alternatives, and the upper right portion shows collapse fragility curves for the same
design alternatives. The design decision process starts at the lower left portion of Figure
3A where E(loss | EDP & N C ) is shown. The designer enters this graph with a value of
acceptable loss (at a specific hazard level) and obtains the associated EDP on the mean
loss curve. The designer then enters the hazard curve with the hazard level at which the
loss is acceptable and obtains the associated IM. The intersection of a horizontal line at
this IM value and a vertical line at the previously obtained EDP value in the Structural
System domain, which contains mean IM-EDP curves of design alternatives, can be
7
viewed as a “design target” point. All design alternatives, represented by individual mean
IM-EDP curves, that intersect the IM line to the left of the design target point have an
expected loss that is lower than the acceptable value, therefore, are “feasible” solutions.
Hazard Domain Structural System Domain Hazard Domain Structural System Domain
Mean Hazard Mean Hazard

Mean IM-EDP Curves Mean IM-EDP Curves Collapse Fragility Curves
Collapse Fragility Curves Curve(s) for
Curve(s) for for the given building for the given building
for Design Alternatives for Design Alternatives the building location
Design Alternatives
IM IM
IM IM
λ ( IM ) EDP P (C | I M ) λ ( IM ) EDP P (C | I M )
E($ loss| E D P & N)C
E($loss | EDP & NC)

E ( $loss | C ) E ( $loss | C )
EDP EDP
Mean $Loss Curve(s) (No Collapse) Mean $Loss Value (Collapse) Mean $Loss Curve(s) (No Collapse) Mean $Loss Value (Collapse)
Loss Domain Loss Domain
(A) (B)
FIGURE 3
DDSS FOR ESTIMATION OF MONETARY LOSS AT DISCRETE HAZARD LEVEL: A) DDSS FOR
PBD (KRAWINKLER ET AL., 2005), B) DDSS FOR PBA
This process continues to the C sub-domain of the structural system domain where
the designer finds the probability of collapse, P(C | IM), for design alternatives at the
hazard level of interest. By continuing vertically, the loss associated with collapse is
obtained. The total expected loss of each design alternative, at the hazard level of interest,
can then be expressed by summation of losses in each sub-domain as shown in (4). In this
equation, E(loss | IM) is the total expected loss of a design alternative at intensity level
IM, and P(NC | IM) is the probability of no-collapse conditioned on the value of IM and is
equal to 1- P(C | IM). E(loss | IM & C ) is the total loss of the building in case of collapse.
A detailed discussion of this approach along with approximations involved can be found
in Zareian (2006).
⎛ ⎞
E( loss | IM ) = ⎜ ∑ E( loss | IM &NC) ⎟× P(NC| IM) + E( loss | IM &C) × P(C| IM) (4)
⎜ all ⎟
⎜ subsystems ⎟
⎝ ⎠
Collapse fragility curves can also be used to provide assistance in conceptual design
for collapse performance targets. Such a target could be expressed as a tolerable
probability of collapse at a specific hazard level. The intersection of the line denoting the
IM value at the specified hazard level with the line denoting the tolerable probability of
collapse, can be viewed as “collapse design target”, which divides the design alternatives
into a feasible and an unfeasible solution space. All design alternatives whose collapse
fragility curve intersects the IM line to the left of the collapse design target point have a
lower probability of collapse and are “feasible” alternatives. Clearly, the hazard level of
interest for design against collapse usually is different from that used to establish design
targets for acceptable monetary loss.
8
The presented PBD process can be utilized also as a simplified PBA process with a
reversal in the flow of information in the three-domain approach presented for the DDSS.
Given the building, its location and characteristics, the mean information in the Hazard
Domain, Structural System Domain, and Loss Domain can be generated. Starting the
flow of information in the Hazard Domain, the user can estimate the expected value of
DV and probability of collapse of the building at a discrete hazard level. This approach is
illustrated schematically in Figure 3B, in which we demonstrate how to estimate the
expected value of monetary loss of a given structure at a specific hazard level.
The simplified PBEE process can be taken to the next level and address performance
objectives defined at an annualized mean level. In the previously discussed approach, for
each design alternative (PBD) or for any given structure (PBA), an IM versus expected
DV curve can be obtained by determining the expected total loss for a number of hazard
levels. Then, the expected annual loss can be computed by numerical integration of the
expected (mean) DV|IM curve over the hazard curve, as shown in (5).
E ( loss) = ∫ E ( loss | im) dλIM ( im) (5)

IM
An equivalent approach can be employed to obtain the Mean Annual Frequency of

collapse, which is estimated by integrating the collapse fragility curve of each design
alternative over the hazard curve as expressed mathematically in (6). Jalayer (2003) has
introduced a closed form solution to (6), which is shown on the right hand side of this
equation. λIM (ηC ) is the MAF of the ground motion intensity associated with the median of
the collapse fragility curve (ηc), k is the slope of the hazard curve in vicinity of ηc, and
βRC is the dispersion of the collapse fragility curve. Details of design for collapse safety
can be found in Zareian and Krawinkler (2007b).
λColllapse = ∫ P ( C | IM ) dλIM ( im) = ⎣⎡λIM (ηC ) ⎦⎤ ⎡⎣exp (1 2 k 2 βRC 2 )⎤⎦ (6)

IM
IMPLEMENTATION EXAMPLES
The effectiveness of the simplified (mean-based) performance based design methodology
for selection of appropriate structural system for an 8-story office building is illustrated in
this section. This building is located in the Los Angeles area (34.221° north latitude,
118.471° west longitude). The site hazard in terms of spectral accelerations at the 50/50,
10/50, and 2/50 hazard levels is provided for different periods (Krawinkler ed., 2005).
We assume that the expected monetary value of this building is roughly equal to
$12,000,000. The performance targets are an acceptable expected monetary loss of 10%
of the building value at the 50/50 hazard level, a tolerable probability of collapse of 15%
at the 2/50 hazard level, and a tolerable mean annual frequency (MAF) of collapse of
0.0002 (less than 1% probability of collapse in 50 years)
Among the many structural systems that could be used for the aforementioned
buildings, we only investigate regular reinforced concrete shear wall structures and
moment-resisting frame structures. We assume that the building is regular in plan and in
9
height. Figure 4 shows the design decisions process using DDSS to obtain a practical
structural system for the 8-story office building. In order to keep the process simple yet
comprehensive, we will discuss the advantages and disadvantages of three design
alternatives, two moment-resisting frame structures (denoted as OF and OF2) and one
reinforced concrete shear wall structure (denoted as OW). Values presented in this
example are for illustration only.

The total expected loss in this building is disaggregated into building level expected
loss in the NSDSS, NSASS, and SS subsystems. Such subsystems could be defined also
at the story level, which would increase the accuracy, but at the cost of more elaborate
computational work (Zareian, 2006). We assume that the total value of NSDSS and
NSASS in the 8-story office building is independent of the choice of structural system
and are equal to $7,000,000 and $4,000,000, respectively. The higher value for NSDSS
compared to NSASS is due to large number of non-structural drift sensitive components
(e.g., partitions, cladding) in a typical office building. To construct an appropriate mean
loss curve for NSDSS we assume that the EDP that correlates best with loss in this
subsystem is the average of the maximum interstory drift ratios over the height, (max
IDR)avg. Also, we assume that the EDP that correlates best with the building level
NSASS loss is the average of peak floor accelerations over the height, (PFA)max.
The current state of knowledge in obtaining such curves is limited and more research
and development is needed to obtain reasonable estimates of generally applicable loss
curves. To complete this design example, we have developed mean loss curves based on
judgment and some limited data (Aslani, 2005; Taghavi and Miranda, 2003). Such curves
are shown in the lower-central portion of Figure 4 for the NSDSS and NSASS in the 8-
story office building. Each mean loss curve consists of three parts: an initial part in which
the expected value of loss is zero for small values of EDP, an intermediate part in which
the expected value of loss linearly increases to the total value of the subsystem, and a
final part where loss is equal to the value of the subsystem.
The value of the structural subsystem (SS) in any building, and the EDP that
correlates well with this subsystem loss, depends on the type and properties of the
selected structural system. In this example, we have assumed that for shear wall systems
the EDP that best represents loss in the SS is the maximum plastic hinge rotation at the
base. For a moment-resisting frame system, the appropriate EDP is assumed to be the
average of the maximum plastic story drift angles over the height of the structure. One
can not plot SS mean loss curves prior to selection of design alternatives due to their
dependence on the selected structural system. However, in order to converse about the
schematic shape of such curves, we direct the reader’s attention to the SS section in the
Loss Domain of Figure 4. As seen, the cost of the SS subsystem is relatively small and
not dominant, except if the EDP exceeds a certain threshold value after which the SS loss
curve jumps to a loss value equal to the total value of the building. This threshold value
represents the EDP in which the owner decides to demolish the building rather than
repair it.
10
Hazard Structural System Domain

Domain
Hazard Mean IM-EDP Curves Collapse Fragility
Curves for for Design Alternatives Curves for Design
Design E ( EDP | IM & NC ) Alternatives
Alternatives
P(C | IM )
λ(IM) 8-story Frame, T1 = 0.8 sec., γ = 0.25, Stiff. & Str. = Shear, SCB = 2.4,2.4
θp = 3%, θpc/θp = 5, λ = 20
OF
8-story Wall, T1 = 0.8 sec., γ = 0.25, Stiff. = Unif., Str. = -0.05My,base / floor
OW
θp = 2%, θpc/θp = 1, λ = 20
8-story Frame, T1 = 1.6 sec., γ = 0.25, Stiff. & Str. = Shear, SCB = 2.4,2.4
OF2
θp = 3%, θpc/θp = 5, λ = 20
Sa(T1)/g Sa(T1)/g Sa(T1)/g Sa(T1)/g

2.0 2.0 2.0 2.0
T1=0.8sec.
1.5 1.5 1.5 1.5
T1=1.6sec.
1.0 1.0 1.0 1.0
0.5 0.5 0.5 0.5
50/50 10/50 0.0 0.01 0.02 0.0 0.5 1.0 1.5 2.0 0% 10% 20%
2/50
λ(Sa(T1)/g) (maxIDR)avg. (PFA)avg. (g) EDP for SS P(C|Sa(T1)/g)
NSDSS NSASS SS
Expected Total $Loss at
Collapse ( in millions)
Expected Subsystem
$Loss ( in millions)
12.0 Bldg. level 12.0 Bldg. level 12.0 Bldg. level 12.0
10.0 10.0 10.0 10.0
Expected
8.0 8.0 8.0 8.0
Total $Loss
6.0 6.0 6.0 6.0
(in millions)
4.0 4.0 4.0 4.0
2.0 2.0 2.0 2.0
0.0 0.0
0.0 0.005 0.01 0.015 0.02 0.025 0.0 0.5 1.0 1.5 2.0
EDP=Avg. of max. EDP=Avg. of max.
story drift ratios, floor accelerations, EDP for SS
(maxIDR)avg. (PFA)avg.(g)
E ( $ Loss | EDP & NC ) E ( $ Loss | C )

$Loss Curves $Loss Value
(No Collapse) (Collapse)
$Loss Domain
FIGURE 4
IMPLEMENTATION OF DDSS IN CONCEPTUAL DESIGN OF 8-STORY OFFICE BUILDING WITH
PERFORMANCE TARGETS AT DISCRETE HAZARD LEVELS, USING BUILDING-LEVEL
SUBSYSTEMS
Given the mean loss curves for NSDSS and NSASS, various structural system
alternatives can be evaluated using the DDSS for the 8-story office building. For each
design alternative, the Hazard Domain, the Structural System Domain, and SS section of
11
the Loss Domain need to be populated with appropriate graphs. In the Hazard Domain, a
hazard curve for the location of the building and the fundamental period of each design
alternative is obtained using seismic hazard analysis. In the Structural System Domain,
appropriate mean IM-EDP curves and associated collapse fragility curve for each design
alternative are obtained from the database of such relationships developed for this study.
Finally, the mean loss curve for the SS is obtained from previous data (or judgment) and
plotted in the appropriate place in the Loss Domain.
The first performance criterion is to limit the average total loss value to $1,200,000
(10% of building value) at the 50/50 hazard level. The largest contributor to loss is the
NSDSS, therefore, we will focus on this subsystem while keeping an eye on losses in
other subsystems. As the first design decision, we assume that design alternatives have a
fundamental period of 0.8 seconds. Hazard values for the location of this building and for
period of 0.8 seconds are obtained and plotted in the Hazard Domain of Figure 4, and a
hazard curve is fit to the data points. Entering the NSDSS mean loss curve with a value of
$1,200,000 and entering the hazard curve associated with T1 = 0.8sec at the 50/50 hazard
level, we obtain the design point. Figure 4 shows this process with heavy arrow lines and
shows the design point with a solid circle in the NSDSS section of the Structural System
Domain.
Two design alternatives are investigate, a moment-resisting frame with T1 = 0.8 sec.
and γ = 0.25 (denoted as OF), and a reinforced concrete shear wall with T1 = 0.8 sec. and
γ = 0.25 (denoted as OW). In the OF design alternative, we assumed that: member design
is controlled by stiffness (straight line deflected shape under NEHRP lateral load pattern),
member strength is proportional to its stiffness (denoted as “Stiff. & Str. = Shear”), and
the ratio of columns strength to beam strength is equal to 2.4 at all joints (denoted as
“SCB = 2.4,2.4”). Plastic hinge rotation capacity θp, post capping rotation capacity ratio
θpc/θp, and the cyclic deterioration parameter λ of each element in this structure is set to
3%, 5, and 20, respectively. In the OW design alternative, it is assumed that wall
dimensions along the height are constant (denoted as “Stiff. = Unif.”), and the wall
bending strength decreases at the rate of 0.05My,base per floor (denoted as “Str. = -
0.05My,base/floor”). θp, θpc/θp, and λ for each story of the shear wall are set to 2%, 1, and
20, respectively. Detailed information about these parameters can be found in Zareian
(2006).
IM-EDP relationships and collapse fragility curves for OF and OW are shown with
solid gray lines and solid black lines, respectively. Keeping the focus on NSDSS in the
Structural System Domain, it is evident that both structural systems will satisfy the
monetary loss performance objective. It can be seen that structural systems with a period
significantly larger than 0.8 sec. will not satisfy this performance objective. This is
demonstrated by investigating a third design alternative, that being a moment-resisting
frame similar to OF but with a fundamental period T1 = 1.6 sec. (denoted as OF2). The
IM-EDP relationship for this structural system is depicted with gray dashed lines in the
NSDSS portion of the Structural System Domain, and the associated hazard curve (for a
period of 1.6 seconds) is plotted with gray dashed lines in the Hazard Domain. The
intersection of the IM-EDP relationship for OF2 and the horizontal line associated with
IM at 50/50 hazard level falls to the right of the design point, therefore, OF2 is not a
desirable design alternative.
12
The monetary loss for each subsystem in the OF and OW design alternatives can be
evaluated at the 50/50, 10/50, and 2/50 hazard levels using the DDSS. In addition, the
probability of collapse at these hazard levels can be found by reading the probability of
collapse of each structural system from the right hand side of Figure 4 in the collapse
sub-domain of the Structural System Domain. Overall, both OF and OW provide an
acceptable performance by limiting the total monetary loss at the 50/50 hazard level to
less than $1,200,000 and the probability of collapse at the 2/50 hazard level to less than
15%.
The final check is to investigate if the design alternatives satisfy the last performance
objective, that is, a MAF of collapse less than 0.0002. Using the collapse fragility curves
for OF and OW, we obtain that the median and dispersion of the collapse fragility curve,
ηc & βRC, are equal to 3.56g & 0.4 and 3.10g & 0.5, respectively. Using the hazard curve
for the fundamental period of these design alternatives, we obtain that the mean annual
frequency of exceedance of ηc, i.e., λIM(ηc), is equal to 0.00009 and 0.00012,
respectively. The slope of the hazard curve in the vicinity of 3.56g and 3.10g is estimated
to be 2.16. Thus, from Equation (6) we find that the MAF of collapse for OF and OW is
equal to 0.00009*exp(0.5x2.162x0.42) = 0.00013 and 0.00012*exp(0.5x2.162x0.52) =
0.00021, respectively.
Based on the calculated DVs for the OF and OW design alternatives, the OF
alternative provides better seismic performance. Whether a moment frame structure with
T1 = 0.8 is indeed a more effective solution than a shear wall structure with the same
period is a matter of a cost-benefit analysis that should include up-front construction
costs as well as architectural/functional considerations.
CONCLUSIONS
In this paper we presented a simplified PBEE methodology and illustrated its potential
for conceptual PBD and quick PBA through a semi-graphical tool denoted as DDSS
(Design Decision Support System). The DDSS incorporates three domains: Hazard
Domain, Structural System Domain, and Loss Domain, in which relationships between
mean values of three basic random variables: IM, EDP, DV, are evaluated to perform a
conceptual design. The DDSS can be used as a design aid for selection of an appropriate
structural system and associated parameters that fulfill multiple performance objectives.
Such a conceptual design could be followed by a detailed PBA if dispersions due to
uncertainties need to be considered for the final design. Furthermore, the DDSS can be
used for a quick performance assessment of a given building, avoiding the extensive
computational effort involved in accounting for uncertainties and their propagation from
hazard and ground motion modeling up to loss estimation and decision making.
The authors tried to demonstrate in a design example that the advantage of the
proposed procedure lies in its simplicity, its potential to disaggregate total loss into
subsystem losses, and its ability to evaluate the effects of subsystem losses on the
aggregated loss for various design alternatives and at various hazard levels. It was shown
that the proposed procedure can provide guidance for informed decision making on
structural systems and system parameters based on performance objectives that may vary
significantly depending on the use and importance of the building.
13
At this time the focus of the DDSS is on monetary loss and collapse, and even there
the information (particularly on loss curves) is far from comprehensive. The simplified
PBEE methodology presented in this paper, in concept, is equally applicable to downtime
loss. The challenge is to find appropriate EDPs that correlate well with damage resulting
in downtime loss and to develop loss curves that apply to performance objectives based
on this decision variable.
ACKNOWLEDGEMENT
This study was supported by the NSF sponsored Pacific Earthquake Engineering
Research (PEER) Center and was carried out at Stanford University’s John A. Blume
Earthquake Engineering Center as part of a comprehensive effort to develop basic
concepts for PBEE and supporting data on seismic demands and capacities.
REFERENCES
[1] ATC-58, “Guidelines for seismic performance assessment of buildings-25% complete draft”, Applied Technology Council,
Redwood City, California. http://www.atcouncil.org/pdfs/ATC-58-25percentDraft.pdf
[2] Alavi, B., Krawinkler, H, “Effects of near-fault ground motions on frame structures”, Report No. 138, 2001, The John A. Blume
Earthquake Engineering Center, Department of Civil & Environmental Engineering, Stanford University, Stanford, California.
[3] Aslani, H, “Probabilistic earthquake loss estimation and loss deaggregation in buildings”, Ph.D Dissertation (under supervision
of Prof. Eduardo Miranda) 2005; Stanford University, Stanford California.
[4] Baker, JW, Cornell, CA., “Vector-valued ground motion intensity measure consisting of spectral acceleration and epsilon”,
Earthquake Engineering & Structural Dynamics, 34(10), 2005, 1193-1217
[5] Cornell CA, “Engineering seismic risk analysis”, Bulletin of the Seismological Society of America, 58, 1968, 1583-1606.
[6] Cornell, CA, Krawinkler, H, “Progress and challenges in seismic performance assessment”, PEER News, April 2000.
[7] Deierlein, G, “Overview of a comprehensive framework for earthquake performance assessment”, Proceedings of International
Workshop on Performance-Based Seismic Design – Concepts and Implementation, Bled, Slovenia, 2004.
[8] Ibarra, LF, Krawinkler H, “Global collapse of frame structures under seismic excitations”, Report No. PEER 2005/06. 2005;
Pacific Earthquake Engineering Research Center, University of California at Berkeley, Berkeley, California.
[9] Ibarra, LF, Medina, RA, and Krawinkler, H, “Hysteretic models that incorporate strength and stiffness deterioration”, Journal
for Earthquake Engineering and Structural Dynamics, 34(12), 2004, 1489-1511.
[10] Jalayer, F, “Direct probabilistic seismic analysis: implementing non-linear dynamic assessments”, Ph.D. dissertation (under
supervision of Prof. Allin Cornell), 2003, Stanford University, Stanford, California.
[11] Krawinkler, H, “A general approach to seismic performance assessment”, Proceedings of International Conference on Advances
and New Challenges in Earthquake Engineering Research, ICANCEER, Vol. 3, 2002, 173-180.
[12] Krawinkler, H, Miranda, E, Performance-based earthquake engineering, Earthquake Engineering: from engineering seismology
to performance-based engineering, Chapter 9, Bozorgnia Y, Bertero VV (eds); CRC Press: Boca Raton, 2004, 9-1 to 9-59.
[13] Krawinkler, H, Zareian, F, Medina, RA, Ibarra, LF, ”Decision support for conceptual performance-based design”, Earthquake
Engineering and Structural Dynamics, 35(1), 2005, 115-133.
[14] Krawinkler, H, (ed). “Van Nuys office building testbed report: Exercising seismic Performance Assessment”, Report No. PEER
2005/11, Pacific Earthquake Engineering Research Center, University of California at Berkeley, Berkeley, California, 2005
[15] Luco, N, “Probabilistic seismic demand analysis, SMRF connection fractures, and near source effects”, Ph.D. dissertation
(under supervision of Prof. Allin Cornell), 2002, Dept. of Civil & Environmental Engineering, Stanford University, Stanford,
California.
[16] Taghavi, S, Miranda, E, “Response Assessment of Nonstructural Building Elements”, Report No. PEER 2003/05, Pacific
Earthquake Engineering Research Center, University of California at Berkeley, Berkeley, California, 2003
[17] Tothong, P, “Improved probabilistic seismic demand analysis using advanced scalar ground motion intensity measures,
attenuation relationships, and near-fault effects” Ph.D. dissertation (under supervision of Prof. Allin Cornell), 2006, Department
of Civil & Environmental Engineering, Stanford University, Stanford, California.
[18] Vamvatsikos, D, Cornell, CA, “Incremental Dynamic Analysis”, Earthquake Engineering and Structural Dynamics, 31(3), 2002,
491-514.
[19] Zareian, F, “Simplified Performance-Based Earthquake Engineering”, Ph.D Dissertation (under supervision of Prof. Helmut
Krawinkler) 2006; Stanford University, Stanford California.
[20] Zareian, F, Krawinkler, H, “Sensitivity of Collapse Potential of Buildings to Variations in Structural Systems and Structural
Parameters”, ASCE Structures Congress, Long Beach, CA., 2007(a)
[21] Zareian, F, Krawinkler, H, “Assessment of probability of collapse and design for collapse safety”, Earthquake Engineering and
Structural Dynamics, In review, 2007(b)
14

Structural System Selection Using Performance-Based Design

Uploaded by

Structural System Selection Using Performance-Based Design

Uploaded by

Structural Engineering Research Frontiers © 2007 ASCE

Structural System Selection Using Performance-Based Design

Farzin Zareian, Assistant Professor, University of California Irvine, Irvine, CA 92697,

Copyright ASCE 2007 Structures Congress 2007

PERFORMANCE-BASED EARTHQUAKE ENGINEERING

The PEER PBA methodology is completed by following a flow from IM to DV.

G ( DV | IM ) = ∫ ∫ G ( DV | DM ) dG ( DM | EDP ) dG ( EDP | IM ) (1)

In Equations (1) and (2), G functions represent complementary cumulative

SIMPLIFIED APPROACH FOR PERFORMANCE-BASED DESIGN

To further facilitate the simplified PBD, a semi-graphical tool denoted as DDSS

Hazard Structural System Domain

Mean Loss Curves Mean Loss due to Collapse

Structural System Domain

curves are developed by fitting a cumulative lognormal distribution function to the

EDP-IM Curves (SW)

Framework for a Design Decision Support System

Mean Hazard Mean Hazard

E($loss | EDP & NC)

Loss Domain Loss Domain

E ( loss) = ∫ E ( loss | im) dλIM ( im) (5)

An equivalent approach can be employed to obtain the Mean Annual Frequency of

λColllapse = ∫ P ( C | IM ) dλIM ( im) = ⎣⎡λIM (ηC ) ⎦⎤ ⎡⎣exp (1 2 k 2 βRC 2 )⎤⎦ (6)

example are for illustration only.

Hazard Structural System Domain

Sa(T1)/g Sa(T1)/g Sa(T1)/g Sa(T1)/g

0.5 0.5 0.5 0.5

E ( $ Loss | EDP & NC ) E ( $ Loss | C )

on this decision variable.

You might also like