Fajfar and Dolšek (2012) - A Practice Oriented Estimation of The Failure Probability of Building Structures

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS
Earthquake Engng Struct. Dyn. 2012; 41:531–547

Published online 1 July 2011 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/eqe.1143
A practice‐oriented estimation of the failure probability of

building structures
Peter Fajfar*,† and Matjaž Dolšek

University of Ljubljana, Faculty of Civil and Geodetic Engineering, Slovenia
SUMMARY
In the presented practice‐oriented probabilistic approach for the seismic performance assessment of building
structures, the SAC‐FEMA method, which is a part of the broader PEER probabilistic framework and
permits probability assessment in closed form, is combined with the pushover‐based N2 method. The most
demanding part of the PEER probabilistic framework, that is incremental dynamic analysis, is replaced by
the much simpler N2 method, which requires considerably less input data and much less computational
time, but which can, nevertheless, often provide: acceptable estimates for the mean values of the structural
response. Using some additional simplifying assumptions that are consistent with seismic code procedures,
an explicit equation for a quick estimation of the annual probability of “failure” (i.e. the probability of
exceeding the near collapse limit state) of a structure can be derived, which is appropriate for practical
applications, provided that predetermined default values for the dispersion measures are available. In the
paper, this simplified approach is summarized and applied to the estimation of the “failure” probability of
reinforced concrete frame buildings representing both old structures, not designed for earthquake resistance,
and new structures designed according to Eurocode 8. The results of the analyses indicate a high probability
of the “failure” of buildings, which have not been designed for seismic loads. For a building designed
according to a modern code, the conservatively determined probability of “failure” is about 30 times less but
still significant (about 1% over the lifetime of the structure). Copyright # 2011 John Wiley & Sons, Ltd.
Received 12 July 2010; Revised 18 April 2011; Accepted 18 April 2011
KEY WORDS: probabilistic seismic assessment; reinforced concrete frame; simplified nonlinear analysis;
N2 method; SAC‐FEMA method; Eurocode 8
1. INTRODUCTION
Earthquake engineering is characterized by large aleatory (random) and epistemic (knowledge)

uncertainties. Both types of uncertainty are discussed, for example, in [1, 2]. Performance‐based
earthquake engineering seeks to improve seismic risk decision‐making through new design and
assessment methods that have a probabilistic basis and try to properly take into account the inherent
uncertainty. At the Pacific Earthquake Engineering Research (PEER) Center, a probabilistic
framework has been developed [3, 4], which has in recent years been widely used worldwide. One
of the methods developed for the seismic risk evaluation of structures is the SAC‐FEMA method,
which permits probability assessment in closed form [5].
Within the PEER probabilistic framework, the relationship between the seismic intensity measure
(IM) and the engineering demand parameter (EDP) is usually determined by the incremental dynamic
analysis (IDA) (Vamvatsikos and Cornell [6]). IDA requires a large number of inelastic response‐
history analyses (and corresponding detailed data on the ground motion time histories and hysteretic
behavior of structural elements), and is thus very time‐consuming. It is often possible to create mean
*Correspondence to: Peter Fajfar, FGG, Jamova 2, SI‐1000 Ljubljana, Slovenia.

†
E‐mail: pfajfar@ikpir.fgg.uni‐lj.si
Copyright # 2011 John Wiley & Sons, Ltd.

532 P. FAJFAR AND M. DOLŠEK
IDA curves with less input data and less effort, but with still acceptable accuracy. One possible
approach is to determine the seismic demand for multiple levels of seismic intensity using the N2
method [7, 8], which is a practice‐oriented nonlinear method based on pushover analysis and inelastic
response spectrum. The N2 method has been implemented in Eurocode 8 (EC8) [9]. Such an approach
yields an incremental N2 (IN2) curve [10, 11], which is intended to approximate a mean IDA curve.
By using the IN2 curve, together with default values for dispersion measures, the probabilistic
approach can be substantially simplified. The simplified probabilistic approach has been presented by
the authors in [11, 12]. It seems, however, that further simplifications, clarifications, and guidelines
would facilitate the application of probability‐based methods in practice. As a matter of fact, in many
cases, additional simplifying assumptions can be adopted, which are consistent with seismic code
procedures. As a result, an approach for a quick estimation of the annual probability of the “failure” of
a structure can be derived, which is appropriate for practical applications. Note that “failure” has been
conservatively defined as the state when the near collapse (NC) limit state is attained at the first critical
element, that is, in a column or wall (see Section 5).
In the paper, the practice‐oriented approach is summarized and applied to the estimation of the
“failure” probability of RC frame buildings. First, a four‐storey structure, which was not designed for
earthquake resistance, is investigated. Then, two variants of an asymmetric three‐storey structure are
analyzed. The first variant represents a building designed for vertical loads only, without consideration
of seismic codes, that is the “SPEAR building”, which was tested and analyzed within the European
project SPEAR [13, 14], whereas the second variant represents a building with the same geometry but
designed according to EC8.
2. DETERMINISTIC DETERMINATION OF SEISMIC DEMAND AT “FAILURE” BY THE

N2 METHOD
The N2 method combines pushover analysis of an MDOF model with the response spectrum analysis
of an equivalent SDOF model. The method is presented in [7, 8]. Here, only the equations needed for
the deterministic determination of seismic demand at “failure” are summarized.
First, a pushover analysis is performed. The lateral load distribution vector Ψ, employed in
pushover analysis, is related to the assumed displacement shape vector Φ (Φn = 1, where n denotes the
roof level) by
Ψ ¼ MΦ (1)
where M is the diagonal mass matrix. There are no fixed rules for the choice of the displacement
shape. Some guidelines may be given in the regulatory documents.
In the next step, the deformation capacity at “failure”, DC, is determined. DC depends on the
definition of “failure”.
The base shear–top displacement relationship obtained by pushover analysis has to be idealized by
bilinear (elasto‐plastic) idealization. In this way, the yield force Fy, and the yield displacement Dy, are
determined. The ductility at “failure” is defined as
DC
μC ¼ (2)
Dy
The transformation to an equivalent SDOF model is made by dividing both the base shear and top
displacement of the MDOF model with a transformation factor Γ, which is, for a planar model, defined
as
ΦT M 1 m
Γ¼ ¼ (3)
ΦT M Φ L
where 1 is the unity vector and m* is the effective mass. The elastic period of the idealized system is
defined as
Copyright # 2011 John Wiley & Sons, Ltd. Earthquake Engng Struct. Dyn. 2012; 41:531–547
DOI: 10.1002/eqe
ESTIMATION OF THE FAILURE PROBABILITY OF BUILDING STRUCTURES 533
sffiffiffiffiffiffiffiffiffiffiffi
m Dy
T ¼ 2π (4)
Fy
Because the same transformation factor Γ is used for both forces and displacements, it does not
matter if the yield force and yield displacement corresponding to the MDOF or the equivalent SDOF
systems are used in Equation (4). The capacity curve, which can be plotted in the acceleration–
displacement (AD) format and visually compared with demand spectra plotted in the same format, is
determined from the idealized force–deformation curve of the equivalent SDOF system by dividing
the force by m*. The relation between the (spectral) acceleration and the yield force Fy of the MDOF
system is thus defined as
Fy
Say ¼ (5)
Γ m
The reduction factor Rμ due to energy dissipation capacity is defined as the ratio of the acceleration
demand Sae in terms of the elastic spectral acceleration, for the period T, to the acceleration capacity
Say ( i.e. spectral acceleration corresponding to the yield force, Equation (5))
Sae
Rμ ¼ (6)
Say
Note that the reduction factors used in seismic codes (e.g. the behavior factor q used in EC8) also
take into account a reduction caused by overstrength, and are thus not equivalent to the reduction
factor Rμ defined in Equation (6). The relationships between the reduction factor Rμ, the ductility μ,
and the period T (the Rμ–μ–T relations) depend on the type of structural system, and are provided in
the literature. The simplest relation is the “equal displacement rule”, which is often used for ordinary
structures with periods in the medium‐period and long‐period ranges. It will also be used in this paper.
According to the equal displacement rule, the reduction factor Rμ is equal to the ductility factor at
“failure” μC
Rμ ¼ μ C (7)
Consequently, the seismic demand at “failure” in terms of the elastic spectral acceleration Sae,C can
be determined from Equations (6) and (7)
Sae;C ¼ μC ⋅Say (8a)

or from the well‐known relation between the elastic displacement and the (pseudo‐) acceleration
spectrum
4π2 DC
Sae;C ¼ (8b)
T2 Γ
Graphically, in the AD format, Sae,C is defined as the intersection of the line corresponding to the
period T and a vertical line at the displacement capacity (see the figures in Section 5). Sae,C represents
one point of the elastic demand spectrum (in AD format) corresponding to the “failure” state. Knowing
this point and the shape of the elastic demand spectrum, the whole spectrum can be defined, including
the peak ground acceleration at “failure”, PGAC.
For buildings that are asymmetric in plan, an extended version of the N2 method has been developed
[15]. Torsional influences are taken into account by correction factors, which are based on the results
of a standard elastic modal analysis. The correction factors for different structural macro‐elements
DOI: 10.1002/eqe
(e.g. frames, walls) depend on the location of the element in plan. They are applied as amplification
factors to the results of the usual pushover analysis. A similar idea was used for extending the
applicability of the N2 method to medium‐rise and high‐rise buildings where higher modes in elevation
may have an important effect in the upper part of the building [16].
3. FRAMEWORK FOR PROBABILISTIC PERFORMANCE ASSESSMENT
A probabilistic assessment in closed form, upon which the SAC‐FEMA steel moment frame guidelines
are based [17], is defined by Equations (9–12). The x confidence level estimate of the annual
probability (mean annual frequency, for terminology see [18]) of the exceedance of a given limit state,
PLS,x, can be determined as [5]
PLS;x ¼H̃ðsa; C˜ ÞCH Cf Cx (9)

CH ¼ exp 0:5 ½ β 2H k2 2
β
; Cf ¼ exp 2b2 DR þ β 2

CR ; (10)
"
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi #
k2 2
Cx ¼ exp Kx 2
β DU þ β 2CU
b
where H̃ðsa; C̃ Þ is the median value of the hazard function at the seismic intensity sa; C̃ , which
causes a selected limit state. It provides a median estimate of the annual probability that the seismic
intensity will be equal to or exceed the level sa; C̃ , that is, the seismic intensity “corresponding” to
the median displacement capacity D̃C . The seismic IM is typically expressed by means of the
spectral acceleration or of peak ground acceleration. In this section, it will be denoted assa, k is a
parameter of the hazard function (curve), idealized in the form
H̃ ðsa Þ ¼ ko ⋅ðsa Þ−k (11)

where k0 depends on the site under investigation, b (in Equation 10) is a parameter of the function
relating to the EDP (typically expressed in terms of top (roof) displacement or storey drift, denoted
as D and referred to as “displacement”) to the IM, that is, the so‐called IDA curve. The IDA curve
is idealized in the form
D̃ðsa Þ ¼ a⋅ðsa Þb (12)

where a depends on the structural system. Kx (in Equation 10) is the standardized normal variate
associated with the probability x of not being exceeded. For example, the values Kx = 0, 1, and 1.28
are associated with 50, 84, and 90% confidence levels, respectively. βH is the dispersion measure
related to the hazard function. The product H̃ðsa; C̃ Þ⋅CH represents the mean value of the

hazard function Hðsa; C̃ Þ

Hðsa; C̃ Þ ¼ H̃ðsa; C̃ Þ⋅CH (13)
Cornell et al. [5] suggested that by using the mean hazard function, one can assume that the
epistemic uncertainty in estimation of seismic hazard (e.g. the uncertainty related to different ground
motion prediction equations) is incorporated. β DR and βCR are the dispersion measures for the aleatory
uncertainty in displacement demand and capacity, respectively, and β DU and β CU are the dispersion
measures for the epistemic uncertainty in displacement demand and capacity, respectively [5]. The
total dispersion of the EDP caused by record‐to‐record variability (aleatory uncertainty, randomness)
can be obtained as β TR = √ (β DR2 + β CR2). This variability is associated with frequency content and
DOI: 10.1002/eqe
other attributes of the ground motion, and applies to a given intensity. (Note that the uncertainty in
ground motion intensity, which is usually the largest uncertainty involved in seismic analysis, is
incorporated in the hazard curve.) Similarly, β TU = √ (β DU2 + β CU2) is the total dispersion caused by
the uncertainty related to structural modeling and analysis (epistemic uncertainty). For practical
applications, predetermined default values for the dispersion measures, based on statistical studies of
typical structural systems, are needed.
The described approach uses an EDP (“displacement”) as the major quantity for measuring demand
and capacity. It can be denoted as the EDP‐based approach. The approach that uses the ground motion
intensity (usually peak ground acceleration or spectral acceleration) directly for measuring demand
and capacity can be denoted as the IM‐based approach [19, 20]. In the case of the IM‐based approach,
Cf and Cx are defined as

Cf ¼ exp 0:5 k2 β 2RC ; Cx ¼ exp ½Kx k β UC (14)
Compared with expressions (10), the slope parameter b that defines the relationship between the
displacement‐based demand and the IM (Equation (12)) is not needed. The dispersion measures
related to displacement demand βDR and βDU are also not needed. This is because the IM‐approach
does not employ displacement‐based demand as an intermediate variable [19]. Note that in the IM‐
based approach, the dispersion measures are related to spectral acceleration, whereas in the EDP‐based
approach, they are related to displacements, so that a different notation (different indices) is used for
the dispersion measures related to capacity, β RC and β UC.
The approximate approach for combining the effect of aleatory and epistemic uncertainties, used in
Equations (9) and (10) or (14), can be called the “confidence level” approach. In this approach, the
effect of epistemic uncertainty is that the estimate of the median of capacity is decreased according to
the dispersion caused by epistemic uncertainty and the confidence level sought (more than 50%) [21].
Although this approach is conceptually appealing, the resulting structural performance predictions
become highly dependent on the level of confidence chosen [22].
Recently, another approach for combining aleatory and epistemic uncertainties, that is the “mean”
approach [5], has become more popular [20–22]. This approach is also used in the collapse assessment
procedure described in FEMA P695 [23]. In the “mean” approach, it is assumed that the effect of
different aleatory and epistemic sources of variability can be concentrated in a single dispersion measure
β TOT, which is determined as the Square Root of the Sum of the Squares (SRSS) of the individual
dispersion measures related to both aleatory and epistemic uncertainty β TOT = √ (β TR2 + β TU2) (or
β TOT = √ (β RC2 + β UC2) if the IM‐based approach (Equation 14) is used). Using the mean hazard

function ( Hðsa; C̃ Þ, Equation (13), and the “total” dispersion measure β TOT , Equation (9) becomes

k2 2
PLS ¼ Hðsa; C̃ Þ exp β (15)
2 b2 TOT
The results derived from the “mean” approach are not sensitive as to whether individual
uncertainties are classified as aleatory or epistemic, which is helpful when the classification of a
particular uncertainty is not obvious [22].
The original formulation of the SAC‐FEMA method was developed for steel structures. One of the
first applications to RC structures was presented in [24].
4. A SIMPLIFIED PROCEDURE USING INCREMENTAL N2 INSTEAD OF IDA
An important simplification of the probabilistic seismic assessment can be introduced by employing

the pushover‐based N2 method instead of response‐history based IDA analysis for the determination
of the relation between the IM and the EDP. In such a case, the IDA curve is replaced by an IN2
curve [10–12]. The whole IN2 curve can be determined by repeating the N2 approach for
DOI: 10.1002/eqe
increasing ground motion intensity until “failure” occurs. In general, the shape of the IN2 curve
depends on the relation between the reduction factor, ductility, and period (the Rμ–μ–T relation),
which defines the inelastic spectra to be used in the N2 method for the determination of seismic
demand. Knowing the IN2 curve, the EDP can be easily linked to the corresponding seismic IMs.
An IN2 curve is intended to approximate a mean IDA curve, and is not calculated for a single
ground motion. The proposed simplified approach is not intended for the determination of
dispersion. Default values for the dispersion measures, therefore, have to be used in order to
determine the probability of exceedance of a given limit state.
Note that some other authors have also proposed some simplifications of the probabilistic approach
described in Section 3. The developers of the IDA curves observed some relations between the IDA
and pushover curves [25]. They developed the software tool SPO2IDA, which provides a direct
connection between these two types of curves [26]. SPO2IDA was used to approximate the IDA
capacity curve from the appropriately post‐processed results of the static pushover results in [27]. Han
and Chopra [28] estimated IDA curves with modal pushover analysis [29]. Recently, Han et al. [30]
explored the potential of modal pushover analysis in investigating the probability of collapse of
buildings. A pushover‐based approach for predicting the median and dispersion of collapse capacity of
moment resisting frame and shear wall structural systems subjected to seismic excitations was
proposed in [31].
5. ADDITIONAL PRACTICE‐ORIENTED SIMPLIFICATIONS
5.1. Assumptions
In this section, further simplifications are introduced, which facilitate the application of the
probabilistic approach. The analyst may decide to apply these simplifications only partly, that is, only
for determination of some parameters, whereas the other parameters can be determined more
accurately if appropriate data exist.
It is assumed that the elastic spectral shape does not change with the intensity of the ground motion.
This approximation is used in the majority of seismic codes and is also consistent with the approach
used for the determination of the IDA curves developed in [6]. Using this assumption, there is a
constant factor between the elastic spectral acceleration Sae corresponding to any period and the peak
ground acceleration PGA. PGA can thus be used as the IM, that is sa.
Proper k0 and k estimates can be obtained by fitting the actual hazard curve in the “region of
interest”; the region of interest can be determined based on the limit state for which the performance
assessments are done [32]. The parameter k in the hazard function (Equation (11), which relates the
rate of change in ground motion intensity with the change in probability, has, in some regions, been
determined by seismological studies. The values of k are usually in the range from 1 to 3
(exceptionally to 4). If two seismic hazard maps are available, e.g. for a 10 and 2% probability of
exceedance in 50 years (i.e. for 475 and 2475 year return periods), the parameter k can be easily
estimated (see e.g. FEMA 350 [17]). Alternatively, according to [17], a default value of k = 3.0 can be
used for Alaska, California, and the Pacific Northwest. The same value applies approximately also to
some other parts of the world, for example Slovenia [12,33] and Egypt [34]. If the value of k, specific
for the region, is not known, a value of k = 3.0 may be an option. It is near the upper bound of k values,
and it mostly provides conservative estimates for probabilities of “failure”, unless the parameter k0 is
determined by using a hazard map for a return period, which is much lower than the return period of
PGAC corresponding to the structure under investigation. For the determination of the parameter k0, at
least one value of the PGA (or Sae) corresponding to a return period should be known for the location
under consideration. This value can be typically obtained from the seismic hazard map, which applies
to a specified probability of exceedance of ground motion H(PGA), for example 2 or 10% in 50 years
or, alternatively, to a specified return period (e.g. 2475 or 475 years). Knowing H(PGA) and PGA or
Sae, the parameter k0 can be obtained from Equation (11). For example, if the hazard map is provided
for PGA for the return period 475 years, k0 is obtained as the annual probability of exceedance H
(PGA) = 1/475 = 0.0021 divided by PGA−k 475 , where PGA475 is the value from the hazard map, taking
into account the soil factor if applicable.
DOI: 10.1002/eqe
For the majority of medium‐period and long‐period structures, that is structures with the
fundamental period longer than the characteristic period of the ground motion TC, the “equal
displacement rule” applies, that is the inelastic displacement is assumed to be equal to the elastic
displacement of the system with the same stiffness and mass, but with unlimited strength. In such a
case, the IN2 curve is a straight line (with its origin at the point (0,0)) until “failure” occurs. The IN2
curve is defined by the point corresponding to “failure”, which can be determined by the N2 method
(see Section 2). It is conservatively assumed that the structure “fails” after the NC limit state is
attained. Thus, after the NC limit state is reached, the IN2 curve is horizontal. Equation (12), with
b = 1.0, defines the first part of the IN2 curve (up to the “failure” point). The “equal displacement rule”,
which has been widely accepted and at least implicitly adopted in many seismic codes, is used also in
the proposed approach (see also Section 2), because it greatly simplifies the assessment procedure and
provides fairly accurate results for a large number of structures. Because b = 1, there is no difference
between the EDP‐based approach and the IM‐based approach. The “equal displacement rule” is the
main assumption used in the proposed procedure.
At the level of the structure, there is a lack of a generally accepted definition about the NC limit
state. One conservative possibility is to assume that the most critical vertical element (column or
structural wall) controls the state of the structure, that is that the NC limit state at the level of the
structure corresponds to the NC limit state at the most critical vertical element. It is therefore assumed
that “failure” occurs when the NC limit state is attained at the first column or wall.
In a practice‐oriented approach, default values for the dispersion measures have to be used. Reliable
data for large populations of buildings are not yet available. Such values will have to be determined in
the future, based on extensive parametric studies. In the following text, the dispersion measures for
aleatory and epistemic uncertainty will be denoted as βR and βU, respectively. For aleatory uncertainty
(record‐to‐record variability), values ranging from 0.35 to 0.45 have been reported in the literature
(e.g. [20,22,35–37]. The results obtained in [38] are in the range from 0.30 to 0.40. Based on these
results, a value of βR = 0.40 can be reasonably assumed, if more accurate data for a specific case are
not available. The same value is used also in FEMA P695 [23].
For the dispersion measure βU, a much larger range of results has been observed than in the case of
the dispersion measure for record‐to‐record variability. The dispersion depends on the quality of the
data and modeling. Based on recent investigations, it is reasonable to use the values βU = 0.40 [20,36]
or βU = 0.45 [22,37]. In FEMA P695 [23], a large range of values for individual dispersion measures
is foreseen. Depending on the quality of the data and on the modeling and design procedure, βU
values are in the range between 0.35 and 1.13. In this paper, it will be assumed that βU = 0.45.
Considering βR = 0.40, the total dispersion measure to be used in the “mean” approach amounts to
βTOT = √ (βR2 + βU2) = 0.60.
The results obtained in [39] suggest that there is a difference between the dispersion measures
related to EDP approach and to IM approach. The values obtained in the case of the EDP approach are
similar to the values indicated in previous two paragraphs, whereas the values related to the IM
approach are larger (βR = 0.68, βU = 0.52). Both Dolšek [40] and Vamvatsikos and Fragiadakis [38]
noticed that βU increases rapidly as the structure approaches global dynamic instability.
5.2. Simplified equations

Considering the assumptions listed in Section 5.2, simplified forms of Equations (9) and (10) or (14) are
obtained. First, only aleatory uncertainty will be taken into account. In such a case, Cx in Equation (9) is
equal to 1.0 (or β TOT = β R in Equation 15). If only the assumptions, which are basic for the proposed
simplified approach (fixed spectral shape, equal displacement rule: b = 1) are taken into account, the
annual probability of “failure” Pf can be estimated as

Pf ¼ exp 0:5 k 2 β 2R H ðPGAC Þ ¼ exp 0:5 k 2 β 2R k0 PGA−k
C (16)
In the case when specific data for k and βR are not available, and the values k = 3 and βR = 0.40 are
used, Equation (16) can be written as
DOI: 10.1002/eqe
Pf ¼ 2 H ðPGAC Þ ¼ 2 k0 PGA−3
C (17)
PGAC is the seismic intensity “corresponding” to the displacement capacity DC (at “failure”)
determined by the N2 method (see Section 2). Equations (16) and (17) do not take into account
epistemic uncertainty. H(PGAC) is the mean value of the hazard function at the seismic intensity PGAC.
It provides an estimate of the annual probability that the seismic intensity will be equal to or exceed the
level PGAC. k0 is a parameter of the hazard curve (Equation 11), which depends on the location of the
structure. Equation (16) implies that, if there was no record‐to‐record variability, the probability of
“failure” would be found simply by substituting the PGA “corresponding” to “failure”, that is PGAC,
into the hazard curve. As a result of record‐to‐record variability, this probability increases by a factor
(Cf in Equations 10 or 14), which depends on the slope of the hazard curve (k) and on the magnitude of
the dispersion (βR). The value Cf = 2 in Equation (17) corresponds to k = 3 and βR = 0.40. Equation (17)
suggests that, in a typical case, the annual probability of the “failure” of the structure is twice the
annual probability that the seismic intensity will be equal to or exceed the level PGAC. If another k
value is used instead of k = 3, for example k = 1, which represents a lower bound according to FEMA
350 [17], a much smaller value of the factor Cf = 1.08 is obtained from Equation (10). Note, however,
that k does not influence only the factor Cf but also the slope of the hazard function, which represents
the second factor in Equation (16).
If the epistemic uncertainty is taken into account in addition to the aleatory uncertainty, the
estimated annual probability of “failure” increases. The epistemic uncertainty is defined by the
dispersion measure βU. In the case of the “confidence level” approach, the amplification factor Cx
(Equation 10) has to be used in addition to Cf

Pf ;x ¼ exp 0:5 k2 β 2R exp ½kx k β U H ðPGAC Þ ¼ exp ½0:5 k 2 β 2R exp ½kx k β U k0 PGA−k
C (18)
In the “mean” approach, Equation (15) becomes

Pf ¼ exp 0:5 k 2 β 2TOT H ðPGAC Þ ¼ exp 0:5 k 2 β 2TOT k0 PGA−k
C (19)
As explained in Section 5.1, in this paper, it will be assumed that the dispersion measure βU = 0.45.
Considering βR = 0.40, the total dispersion measure to be used in the “mean” approach amounts to
β TOT = √ (βR2 + βU2) = 0.60. In the special case, when k = 3.0, as in Equation (17), the amplification
factor in Equation (17) increases from 2 to 5, and 8 in the case of “mean” and “confidence level” (with
84 % confidence) approach, respectively.
Probabilistic approaches are relatively new in practice, and it may not be easy to understand the
meaning of the results. An explanation adapted from Ellingwood and Kinali [41] may help to better
understand the difference between the analyses involving only aleatory variability (Equations 16 and
17), and analyses involving both aleatory and epistemic uncertainty (Equations 18 and 19). In the first
case, the analysis answers the question “What is the failure probability?” In the second case, the
analysis also provides an indirect answer to the question “How confident one can be that the failure
probability is, in fact, what has been estimated.”
One of the reviewers referred to a conference paper by Cornell [42], in which practically the same
explicit equation for failure probability was developed. Unfortunately, this reference has not received
enough attention. It is hoped that this paper, where the explicit equation for the estimation of failure
probability was “reinvented” in an alternative way, will accelerate the implementation of the proposed
approach in practical applications.
6. PROBABILISTIC PERFORMANCE ASSESSMENT OF A FOUR‐STOREY RC

FRAME BUILDING
6.1. Description of the example structure and mathematical model

To demonstrate the proposed simplified probabilistic performance assessment approach, the annual
probability of “failure” (defined as the NC limit state of the most critical vertical element – see Section 5)
DOI: 10.1002/eqe
will be determined for a four‐storey RC frame. The frame is shown in Figure 1, together with the
typical reinforcement in the columns. All the beams are 0.25 m wide and 0.50 m deep. The slabs are
0.15 m thick. The longitudinal reinforcement is reduced in the top two storeys in columns B and C.
The structure had been designed to reproduce the design practice in European and Mediterranean
countries about half a century ago. Capacity design principles (especially the strong column–weak
beam concept) and up‐to‐date detailing were not considered in the design. The masses amount to 46 t
in the bottom storeys and 40 t in the top storey, resulting in a total mass of 178 t. The design base shear
coefficient amounts to 0.08.
The most common mathematical model was used. For the elements of the RC frame, one‐
component lumped plasticity elements were used, consisting of an elastic beam and two inelastic
rotational hinges (defined by a moment–rotation relationship). The yield and the maximum moment
were calculated taking into account the axial forces caused by the vertical loading on the frame. The
potential reduction in strength (moment) as a result of insufficient anchorage length of the reinforcing
bars was not considered when determining the moment–rotation envelopes.
The characteristic rotations, which describe the moment–rotation envelope of a plastic hinge, were
determined according to the procedure described in [14]. Note that the envelope is trilinear and
includes a small post‐yield stiffness followed by a strength degradation. The zero moment point was
assumed to be at the mid‐span of the columns and beams. The ultimate rotation Θu in the columns at
the NC limit state, which corresponds to a 20% reduction in the maximum moment, was estimated by
means of the conditional average estimator (CAE) method [43]. For the beams, the Eurocode 8–3
(EC8‐3) [44] formulas were used, the parameter γel being assumed to be equal to 1.0. Because of the
absence of seismic detailing, the ultimate rotations were multiplied by a factor of 0.85. The calculated
ultimate rotations varied from 0.031 to 0.034 rad for all the columns with the exception of column C,
whose ultimate rotation varied between 0.042 in the first storey and 0.057 rad in the top storey. The
calculated ultimate rotations in the beams varied from 0.027 to 0.043 rad. The mathematical model
was validated by comparing the results of nonlinear dynamic analysis with the experimental results
obtained in pseudo‐dynamic tests. OpenSees [45] was used for the calculations. More details about the
structure and the mathematical model, and a comparison between the calculated and experimental
response‐histories, can be found in [46].
6.2. Estimate of the probability of “failure”

In this section, the annual probability of “failure” will be evaluated. First, the seismic intensity
“corresponding” to the displacement capacity (at “failure”), PGAC, will be determined by the N2
method (see Section 2).
The lateral loads for the pushover analyses are determined based on the first mode shape, for which
the effective mass amounts to about 80% of the total mass. The pushover curve is presented in
Figure 2(a), together with the idealized force–displacement relationship and with marked NC limit
state. The top displacement at the NC limit state, DC, amounts to 12.1 cm (drift = 1.12 %). The critical
Figure 1. The elevation, plan view, and typical reinforcement in the columns of the four‐storey reinforced
concrete frame.
DOI: 10.1002/eqe
Figure 2. The normalized base shear–top displacement relationship. (a) Four‐storey frame (height of the
structure = 10.8 m, weight = 1746 kN). (b) Two variants of the SPEAR building‐X direction (height of the
structure = 9 m, weight W = 1900 kN for Test structure, and W = 2900 kN for EC8 H structure).
element is column B in the third storey (Figure 1). The inter‐storey displacement of the third storey at
the NC limit state amounts to 8.3 cm (storey drift = 3.1%).
In Figure 3(a), the capacity diagram, inelastic spectra, and elastic spectra (EC8 shape, soil type C)
corresponding to the NC limit state are plotted. The transformation from the MDOF system to the
equivalent SDOF system is made by using the transformation factor Γ = 1.29 (Equation 3). Using the
equal displacement rule, the elastic spectral acceleration corresponding to the NC capacity is defined
as the intersection point of the diagonal line representing the period of the structure (in the
acceleration–displacement format) and the vertical line at the displacement capacity. Knowing this
point and the spectral shape, the complete elastic spectrum can be defined. From Figure 3(a), it can be
seen that the structure will “fail” at a ground motion with PGAC = 0.30 g. The inelastic spectrum
(ductility demand μ = 3.50) according to the N2 method is also plotted in Figure 3(a) for illustration.
Note, however, that it is not needed for analysis.
A graphical presentation of capacity and demand allows visualization of the important structural
and ground motion characteristics, and is beneficial for the understanding of the analysis procedure.
Analytically, the spectral acceleration corresponding to the NC limit state, Sa,C(T), can be determined
according to Equations (8a) or (8b). Using the period of the idealized system T = 0.83 s, the
displacement (of the MDOF system) at the NC limit state DC = 12.1 cm, and the transformation factor
Γ = 1.29, Equation (8b) yields Sa,C(T) = 0.55 g. Considering the EC8 spectral shape for soil type C, the
corresponding peak ground acceleration PGAC = 0.30 g is obtained.
IN2 curves represent, in the proposed method, the relation between PGA and the top displacement
D. As a consequence of the equal displacement rule, the IN2 curve is in the investigated case a straight
line from the origin up to the “failure” point, which is defined by PGAC and DC (Figure 4(a)). After the
“failure” point is reached, the IN2 curve is a horizontal line. As in the case of inelastic spectra, the IN2
curve is not needed for calculations, but it helps to understand the procedure.
Figure 3. Elastic and inelastic demand spectra corresponding to “failure” (i.e. the NC limit state) and
capacity diagram for the idealized SDOF system in acceleration–displacement (AD) format. (a) Four‐storey
frame. (b) Two variants of the SPEAR building‐X direction.
DOI: 10.1002/eqe
Figure 4. The IN2 curves. (a) Four‐storey frame. (b) Two variants of the SPEAR building‐X direction.
In the next step of the analysis, the coefficient k0, which defines the seismic hazard function
(Equation 11), is determined. It is assumed that the structure is located in a moderate seismic region on
soil type C (“medium” soil according to EC8). According to the seismic hazard map for a 475‐year
return period, the peak ground acceleration amounts to 0.25 g, and the soil factor for soil type C
amounts to 1.15. Consequently, the seismic demand for the 475‐year return period amounts to
PGA475 = 0.25 g · 1.15 = 0.29 g. The 475‐year return period corresponds to 10% probability of
exceedance in 50 years or to 0.21% (or 0.0021) annual probability of exceedance. It is assumed that
k = 3 and βR = 0.40. The coefficient k0 can be obtained from Equation (11) (note that PGA is used as
IM, that is sa): k0 = 0.0021/(0.29 g)−3 = 0.048 (the unit for acceleration is m/s2, the same unit should be
used in Equations 16–19).
The annual probability of “failure” (considering only aleatory uncertainty), determined by using
Equation (17), is presented in Table I, together with the probability of “failure” in 50 years (the average
lifetime of structures), which is determined as 1‐(1‐Pf)50, and with the corresponding return period TR.
A large probability of “failure” of the structure, which has not been designed for earthquake resistance,
can be observed.
If epistemic uncertainty is taken into account with βU = 0.45 (βTOT = 0.60), the “mean” approach
(Equation 19) yields Pf = 0.0095 (38% in 50 years), and the “confidence level” approach (Equation 18)
yields Pf,84 = 0.0152 (53% in 50 years).
7. PROBABILISTIC PERFORMANCE ASSESSMENT OF TWO VARIANTS OF THE SPEAR

BUILDING
7.1. Description of the example structures and the mathematical model

As the second example, the annual probability of “failure” will be determined for two variants of an
asymmetric three‐storey RC frame building (referred to in the following text as the SPEAR building). The
basic data of the SPEAR building are shown in Figure 5. The building was conceived as representative
of older construction in Southern European countries, but without engineered earthquake resistance. It
was designed for vertical loads only. The building was tested pseudo‐dynamically within the European
project SPEAR [13]. In addition to the original SPEAR building (denoted as “Test”), a variant of
the building was designed in compliance with EC8 (denoted as “EC8 H”) [47]. Realistic values were
Table I. The peak ground acceleration PGAC “corresponding” to the displacement capacity DC at “failure”,
the corresponding spectral acceleration at the fundamental natural period, the mean value of the hazard
function at PGAC, the annual probability of “failure” (50% confidence), the probability of “failure” in
50 years (in %), and the “failure” return period.
PGAC (g) Sa,C (g) H(PGAC) Pf % in 50 years TR (years)
0.30 0.55 0.0019 0.0038 17 260
DOI: 10.1002/eqe
Figure 5. Elevation and plan view of the SPEAR building, and the typical cross‐sections of the beams and
columns of the variants Test and EC8 H.
used for the permanent loads. As a result, the total mass of this variant was 45% greater than the
total mass of the Test structure. The geometry of the whole structure remained the same as in the
case of the Test structure, but the dimensions of individual load‐bearing elements were changed.
In order to comply with EC8 requirements, the dimensions of the columns were increased, and
the dimensions of the beams were adjusted. The structure was designed for a design ground
acceleration of 0.25 g (on rock). Ground type C (soil factor S = 1.15) was assumed, resulting in a
peak ground acceleration PGA = 0.29 g. The high ductility class was selected. For analysis with
SAP2000 [48], the most common mathematical model was used, similar to the model applied in the
first example (Section 6). For details, see [47]. The ultimate rotation Θu in the columns and beams at
the NC limit state, which corresponds to a 20% reduction in the maximum moment, was determined
by using the EC8‐3 [44] formulas, the parameter γel being assumed equal to 1.5. Note that γel = 1.5,
which is used for primary elements according to EC8‐3, yields conservative values (roughly the
mean minus the standard deviation) of the plastic chord rotation capacity. In the case of the Test
building, because of the absence of seismic detailing, the plastic parts of the ultimate chord rotations
were multiplied by a factor of 0.825. The calculated ultimate rotations in columns in the first two
DOI: 10.1002/eqe
storeys varied from 0.018 to 0.027 rad for the Test structure, and from 0.024 to 0.034 rad for the EC8 H
structure. The values for the beams varied from 0.015 to 0.030 rad for the Test structure, and from 0.027
to 0.038 rad for the EC8 H structure. The computed difference between the rotation capacities of the
columns in the two structures is relatively small. An alternative estimation of the rotation capacities
made by the non‐parametric empirical CAE method [43], yields smaller values for the Test structure and
larger values for the EC8 H structure [47]. Details of both variants of the building (denoted as “SPEAR”
and “EC8 H” of the modeling and of the analysis procedure) are provided in [47]. The mathematical
model for the Test structure had been earlier validated by comparing the results of nonlinear dynamic
analysis with the experimental results obtained in pseudo‐dynamic tests. A comparison between the
calculated and experimental response‐histories can be found in [14].
7.2. Estimate of the probability of “failure”

The same procedure as in the case of the first example (Section 6) will be applied. First, the seismic
intensity “corresponding” to the displacement capacity (at “failure”), PGAC, is determined by the N2
method. The details are presented in [47]. Here, only the main results are summarized for the X‐
direction, which proved to be the critical one.
The lateral loads for the pushover analyses for the X‐direction are determined based on the first
mode shape, which is relevant for the X‐direction. The pushover curves are presented in Figure 2(b).
Marked are the points corresponding to the NC limit state, that is the deformation at which the NC
limit state (expressed in terms of the ultimate chord rotation) is attained at the first column. The
torsional influences were taken into account with correction factors based on elastic modal analysis,
using the extended N2 method [15]. The pushover curves clearly demonstrate larger stiffness, strength,
and ductility of the structure designed according to EC8. The roof displacements at the NC limit
state, DC, amount to 11.1 cm (drift = 1.2 %) and to 22.4 cm (drift = 2.5 %) for the Test and EC8
structures, respectively. The critical element is column C1 at the top of the second storey, for the
Test structure, and the same column at the base for the EC8 H structure. In the case of the Test
structure, a plastic mechanism is formed in the lower two storeys. Almost all the columns in the first
and second storeys yield at both joints, whereas most of the beams remain in the elastic region. A
favorable global plastic mechanism, where all the beams yield, as well as the columns at their fixed
base, occurs in the case of the EC8 H structure. The maximum inter‐storey displacements at the NC
limit state occur at the second storey. They amount to 6.2 cm (storey drift = 2.1 %) and to 8.2 cm
(storey drift = 2.7 %) for the Test and EC8 H structures, respectively.
The idealized pushover curves are shown in Figure 2(b). In Figure 3(b), the capacity diagrams and
elastic spectra (EC8 shape, soil type C) corresponding to the NC limit state are plotted. The
transformation factor Γ amounts to Γ = 1.26 and Γ = 1.28, for the Test structure and the EC8‐H
structure, respectively. Using the same procedure as in the first example, it can be seen from Figure 3(b)
that the Test structure would “fail” at a ground motion with PGAC = 0.25 g, whereas the EC8 H
structure is able to resist a much more severe ground motion (PGAC = 0.77 g). The latter result
indicates that the ground motion corresponding to “failure” is 2.7 times more severe than the design
ground motion. The inelastic spectra (ductility demand μ = 3.2 and μ = 6.5 for the Test structure and the
EC8 H structure, respectively) according to EC8 (based on the N2 method) are also plotted in Figure 3(b),
for illustration.
Analytically, the spectral acceleration corresponding to the NC limit state, Sa,C(T), can be
determined from Equations (8a) or (8b). Considering T = 0.94 s and T = 0.61 s for the Test structure and
the EC8 H structure, respectively, the spectral values Sa,C(T) amount to 0.40 g and to 1.89 g for the
former and latter structure, respectively. Taking into account the EC8 spectral shape for soil type C,
the corresponding peak ground accelerations PGAC = 0.25 g and PGAC = 0.77 g are obtained. The IN2
curves for both structures are shown in Figure 4(b).
It is assumed that both variants of the building are located at the same location as the building
analyzed in Section 6 (the EC8 spectrum for ground type C normalized to PGA475 = 0.29 g is used),
and that the same seismic hazard curve applies: k0 = 0.048 (the unit for acceleration is m/s2) and k = 3.
The annual probabilities of failure (50% confidence), determined by using Equation (17), are
presented in Table II, together with the probabilities of “failure” in 50 years (the average lifetime of the
DOI: 10.1002/eqe
Table II. The peak ground acceleration PGAC “corresponding” to the displacement capacity DC at “failure”,
the corresponding spectral acceleration at the fundamental natural period, the mean value of the hazard
function at PGAC, the annual probabilities of “failure” (50% confidence), the probabilities of “failure” in
50 years (in %), and the return periods.
Structure PGAC (g) Sa,C (g) H(PGAC) Pf % in 50 years TR (years)
−2 −2
Test 0.25 0.40 0.325 × 10 0.65 × 10 28 160
EC8 H 0.77 1.89 1.11 × 10−4 2.22 × 10−4 1.1 4500
structures), which is determined as 1‐(1‐ Pf,50)50, and with the corresponding return periods. The
results indicate a very large probability of “failure” of the structure, which was not designed for
seismic resistance. The probability of “failure” of the structure designed according to a modern code is
much smaller, but still quite considerable. The probability would increase by a factor of about 2.5 or
about 4.0 if epistemic uncertainty was considered with βU = 0.45 (βTOT = 0.60) by the “mean” and
“confidence level” (84% confidence) approaches, respectively. The resulting probabilities of “failure”
of the Test structure in 50 years are 56 and 73% for the “mean” and “confidence level” approach,
respectively. In the case of the EC8 H structure, the corresponding probabilities are 2.7 and 4.3%. It
should be noted that, in the presented analyses, “failure” was defined in a quite conservative way, so
the computed probabilities may be near an upper bound of possible estimates.
The probabilities of “failure” have to be compared with the tolerable probabilities of failure, which
have not yet been clearly defined in the earthquake engineering community.
8. CONCLUSIONS
The proposed method is basically the probabilistic seismic assessment method in closed form,
developed by Cornell and co‐authors [5], upon which the SAC‐FEMA steel moment frame guidelines
are based (FEMA 350 [17]). The most demanding part of the PEER probabilistic methodology, that is
IDA, is replaced by the IN2 method, which is based on pushover analysis and the response spectrum
approach [10–12]. Default values for dispersion measures are needed. Using a number of simplifying
assumptions, which are consistent with the assumptions used in usual code analyses, an explicit
formula for the annual probability of “failure” can be developed. The most important assumption is the
“equal displacement rule.” The proposed method yields the same results for the EDP‐based approach
and the IM‐based approach. Both the “confidence level” approach and the “mean” approach can be
used if epistemic uncertainty is taken into account.
The computational procedure consists of two steps. First, the spectral and the peak ground
acceleration representing the capacity of the structure at “failure”, Sae,C and PGAC, respectively, are
determined, using the procedure described in Section 2. In the second step, the annual probability of
“failure” is determined from Equation (16), if only record‐to‐record variability is taken into account. If
epistemic uncertainty is also considered, Equations (18) and (19) apply to the “confidence level” and
“mean” approach, respectively.
The proposed approach was applied to two RC frame buildings. The results show that the first
(four‐storey) building, which was not designed for earthquake resistance, has a large probability of
“failure” in its lifetime, if located in a moderate seismic region, in spite of a considerable
deterministically determined intensity of ground motion associated with its ultimate capacity. The
same applies to the original three‐storey SPEAR building. The ultimate capacity of the variant of the
SPEAR building designed according to a modern standard (EC8) is, of course, much larger. A
deterministic analysis (using some conservative assumptions) indicates that the structure designed for
PGA = 0.29 g would “fail” if subjected to a ground motion with PGA = 0.77 g. However, the
probability of “failure” in the lifetime is still considerable (1.1% for 50% confidence). On the other
hand, the SPEAR building, which was not designed for seismic resistance, would fail at PGA = 0.25 g.
Its probability of “failure” is about 30 times larger than that corresponding to the same structure
designed according to a modern code. Note that, as a result of the conservative assumptions related to
the definition of “failure”, the calculated probabilities of “failure” can be considered as upper bound
DOI: 10.1002/eqe
values. Similar results were obtained by Liel et al. [22,49], whereas Goulet et al. [50] report smaller
probabilities for code‐conforming buildings.
The proposed method has, like other simplified methods, limitations, which are basically the same
as those which apply to the basic N2 method (see [7, 8]) and to the SAC‐FEMA method (see [5]). The
“equal displacement rule” is used, which is not valid for short period structures and for structures,
which do not maintain positive post‐yield stiffness. Some other simplifying assumptions have also
been adopted, which are consistent with seismic code procedures. Reliable default dispersion measures
for aleatory and epistemic uncertainty are not yet available. Nevertheless, such simplified approaches
can facilitate the introduction of probabilistic considerations in practice. For designers and other
stakeholders, it is important to have at least a rough idea about the probability of “failure”. Typically, a
designer would consider a structure as overdesigned, if a deterministic analysis indicated that the
structure was able to resist ground motions almost three times stronger than the design ground motion.
On the other hand, however, many designers and other stakeholders would find a 1% probability of
“failure” within the lifetime of a code‐designed structure excessive. The “tolerable” probabilities of
“failure” and collapse have not yet been agreed upon. In any case, the values around 10−4 for the
lifetime of a structure suggested in Eurocode [51] for “ultimate limit state” cannot be attained for
common structures in seismic regions.
The absolute values of the “failure” probability obtained by the proposed approach are sensitive to
the input data and simplifying assumptions. Comparisons between different structures, however, are
more reliable. Even if the absolute values may be only a very crude approximation, the results of
comparative probabilistic analyses provide valuable additional data needed for decision‐making.
ACKNOWLEDGEMENTS
The authors are grateful to Professor H. Krawinkler for stimulating discussions and for the review of the
manuscript. The constructive comments of two anonymous reviewers helped to improve the paper. The
results presented in this paper are based on work continuously supported by the Slovenian Research
Agency. This support is gratefully acknowledged.
REFERENCES
1. McGuire RK. Seismic Hazard and Risk Analysis. Earthquake Engineering Research Institute: Oakland, CA, 2004.
2. Der Kiureghian A, Ditlevsen O. Aleatory or epistemic? Does it matter? Structural Safety 2009; 31:105–112.
3. Cornell CA, Krawinkler H. Progress and challenges in seismic performance assessment. PEER Center News 2000;
3(2). Available from: http://peer.berkeley.edu/news/2000spring/performance.html.
4. Deierlein G. Overview of a comprehensive framework for earthquake performance assessment. Proceedings of the
International Workshop on Performance‐Based Seismic Design – Concepts and Implementation, Fajfar P,
Krawinkler H (eds.). Bled: Slovenia, 2004. PEER Report 2004/05:15–26, Berkeley, CA. Available from: http://peer.
berkeley.edu/publications/peer_reports/reports_2004/0405.pdf.
5. Cornell CA, Jalayar F, Hamburger RO, Foutch DA. Probabilistic basis for 2000 SAC Federal Emergency Management
Agency Steel Moment Frame Guidelines. Journal of Structural Engineering ASCE 2002; 128(4):526–533.
6. Vamvatsikos D, Cornell CA. Incremental dynamic analysis. Earthquake Engineering and Structural Dynamics
2002; 31:491–514.
7. Fajfar P. Capacity spectrum method based on inelastic demand spectra. Earthquake Engineering and Structural
Dynamics 1999; 28(9):979–993.
8. Fajfar P. A nonlinear analysis method for performance‐based seismic design. Earthquake Spectra 2000; 16(3):573–592.
9. CEN. European standard EN 1998–1:2004. Eurocode 8: Design of structures for earthquake resistance. Part 1:
General rules, seismic action and rules for buildings, European Committee for Standardization, Brussels, 2004.
10. Dolšek M, Fajfar P. IN2 – A Simple Alternative for IDA. Proceedings of the 13th World Conference on Earthquake
Engineering, Vancouver, Canada, 2004; Paper 3353.
11. Dolšek M, Fajfar P. Simplified probabilistic seismic performance assessment of plan‐asymmetric buildings.
Earthquake Engineering and Structural Dynamics 2007; 36(13):2021–2041.
12. Dolšek M, Fajfar P. Effects of masonry infills on the seismic response of a four storey reinforced concrete frame – a
probabilistic assessment. Engineering Structures 2008; 30:3186–3192.
13. SPEAR. Seismic performance assessment and rehabilitation of existing buildings, European Laboratory for
Structural Assessment, Ispra, Italy. Available from: http://elsa.jrc.ec.europa.eu/term_activity.php?id=2.
14. Fajfar P, Dolšek M, Marušić D, Stratan A. Pre‐ and post‐test mathematical modelling of a plan‐asymmetric
reinforced concrete frame building. Earthquake Engineering and Structural Dynamics 2006; 35(11):1359–1379.
15. Fajfar P, Marušić D, Peruš I. Torsional effects in the pushover‐based seismic analysis of buildings. Journal of
Earthquake Engineering 2005; 9(6):831–854.
DOI: 10.1002/eqe
16. Kreslin M, Fajfar P. The extended N2 method taking into account higher mode effects in elevation. Earthquake
Engineering and Structural Dynamics Early View. DOI: 10.1002/eqe.1104.
17. FEMA. Recommended seismic design criteria for new steel moment frame buildings. Report No. FEMA 350, SAC
Joint Venture, Federal Emergency Management Agency, Washington, DC, 2000.
18. McGuire RK, Cornell CA, Toro GR. The case for using mean seismic hazard. Earthquake Spectra 2005;
21(3):879 – 886.
19. Jalayer F, Cornell CA. A Technical Framework for Probability‐Based Demand and Capacity Factor Design (DCFD)
Seismic Formats. Report No. PEER 2003/08. Pacific Earthquake Engineering Research Center, University of
California at Berkeley, Berkeley, CA, 2003.
20. Zareian F, Krawinkler H, Ibarra L, Lignos D. Basic concepts and performance measures in prediction of collapse of
buildings under earthquake ground motions. Structural Design of Tall and Special Buildings 2010; 19:167–181.
21. Zareian F, Krawinkler H. Assessment of probability of collapse and design for collapse safety. Earthquake
Engineering and Structural Dynamics 2007; 36(13):1901–1944.
22. Liel AB, Haselton CB, Deierlein GG, Baker JW. Incorporating modeling uncertainties in the assessment of seismic
collapse risk of buildings. Structural Safety 2009; 31(2):197–211.
23. FEMA. Quantification of Building Seismic Performance Factors. Report No. FEMA P695, Federal Emergency
Management Agency, Washington, DC, 2009.
24. Lupoi G, Lupoi A, Pinto P. Seismic risk assessment of RC structures with the ‘2000 SAC/FEMA’ method. Journal
of Earthquake Engineering 2002; 6(4):499–512.
25. Vamvatsikos D, Cornell CA. Applied incremental dynamic analysis. Earthquake Spectra 2004; 20:523–553.
26. Vamvatsikos D, Cornell CA. Direct estimation of the seismic demand and capacity of oscillators with multi‐linear
static pushovers through IDA. Earthquake Engineering and Structural Dynamics 2006; 35(9):1097–1117.
27. Fragiadakis M, Vamvatsikos D. Fast performance uncertainty estimation via pushover and approximate IDA.
28. Han SW, Chopra AK. Approximate incremental dynamic analysis using the modal pushover analysis procedure.
Earthquake Engineering and Structural Dynamics 2006; 35:1853–1873.
29. Chopra AK, Goel RK. A modal pushover analysis procedure for estimating seismic demands for buildings.
Earthquake Engineering and Structural Dynamics 2002; 31:561–582.
30. Han SW, Moon KH, Chopra AK. Application of MPA to estimate probability of collapse of structures. Earthquake
Engineering and Structural Dynamics 2010; 39(11):1259–1278.
31. Shafei B, Zareian F, Lignos DG. A simplified method for collapse capacity assessment of moment‐resisting frame
and shear wall structural systems. Engineering Structures 2011; 33:1107–1116.
32. Jalayer F, Cornell CA. Alternative non‐linear demand estimation methods for probability‐based seismic assessments.
33. Kramar M, Isaković T, Fischinger M. Seismic collapse risk of precast industrial buildings with strong connections.
34. El Howary HA, Mehanny SSF. Seismic vulnerability evaluation of RC moment frame buildings in moderate seismic
zones. Earthquake Engineering and Structural Dynamics 2011; 40(2):215–235.
35. Dolšek M, Fajfar P. Inelastic spectra for infilled reinforced concrete frames. Earthquake Engineering and Structural
Dynamics 2004; 33(15):1395–1416.
36. Ibarra LF, Krawinkler H. Global collapse of frame structures under seismic excitations. Report No. PEER 2005/06.
Pacific Earthquake Engineering Research Center, University of California at Berkeley, Berkeley, CA, 2005.
37. Haselton CB. Assessing Seismic Collapse Safety of Modern Reinforced Concrete Moment Frame Buildings. Ph.D.
Dissertation, Department of Civil and Environmental Engineering, Stanford University, 2006.
38. Vamvatsikos D, Fragiadakis M. Incremental dynamic analysis for estimating seismic performance sensitivity and
uncertainty. Earthquake Engineering and Structural Dynamics 2010; 39(8):141–163.
39. Dolšek M. Incremental dynamic analysis with consideration of modeling uncertainties. Earthquake Engineering and
Structural Dynamics 2009; 38(6):805–825.
40. Dolšek M. Estimation of seismic response parameters through extended incremental dynamic analysis. Computational
Methods in Earthquake Engineering, Papadrakakis et al. (eds.). Springer, 2011, ISBN 978‐94‐007‐0052‐9, 285–304.
41. Ellingwood BR, Kinali K. Quantifying and communicating uncertainty in seismic risk assessment. Structural Safety
2009; 31:179–187.
42. Cornell CA. Calculating building seismic performance reliability: a basis for multi‐level design norms. Eleventh
World Conference on Earthquake Engineering, Mexico City, Mexico, 1996; Paper No.2122.
43. Peruš I, Poljanšek K, Fajfar P. Flexural deformation capacity of rectangular RC columns determined by the CAE
method. Earthquake Engineering and Structural Dynamics 2006; 35(12):1453–1470.
44. CEN. European standard EN 1998–3:2005. Eurocode 8: Design of structures for earthquake resistance. Part 3:
Assessment and retrofitting of buildings, European Committee for Standardization, Brussels, 2005.
45. McKenna F, Fenves GL. Open system for earthquake engineering simulation. Pacific Earthquake Engineering
Research Center, Berkeley, CA, 2007. Available from: http://opensees.berkeley.edu.
46. Dolšek M, Fajfar P. Effects of masonry infills on the seismic response of a four storey reinforced concrete frame – a
deterministic assessment. Engineering Structures 2008; 30:1991–2001.
47. Rozman M, Fajfar P. Seismic response of a RC frame building designed according to old and modern practices.
Bulletin of Earthquake Engineering 2009; 7:779–799.
DOI: 10.1002/eqe
48. CSI. SAP2000. Analysis Reference Manual, Computers and Structures, Inc., Berkeley, CA, 2002.
49. Liel AB, Haselton CB, Deierlein GG. Seismic collapse safety of reinforced concrete buildings. II: comparative
assessment of nonductile and ductile moment frames. Journal of Structural Engineering ASCE 2011; 137(4):
492–592.
50. Goulet CA, Haselton CB, Mitrani‐Reiser J, Beck JL, Deierlein CG, Porter KA, Stewart JP. Evaluation of the seismic
performance of a code‐conforming reinforced‐concrete frame building – from seismic hazard to collapse safety and
economic losses. Earthquake Engineering and Structural Dynamics 2007; 36:1973–1997.
51. CEN. European Standard EN 1990:2002. Eurocode: Basis of structural design, European Committee for
Standardization, Brussels, 2002.
DOI: 10.1002/eqe

Fajfar and Dolšek (2012) - A Practice Oriented Estimation of The Failure Probability of Building Structures

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

Fajfar and Dolšek (2012) - A Practice Oriented Estimation of The Failure Probability of Building Structures

Uploaded by

Copyright:

Available Formats

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2012; 41:531–547

A practice‐oriented estimation of the failure probability of

Peter Fajfar*,† and Matjaž Dolšek

Received 12 July 2010; Revised 18 April 2011; Accepted 18 April 2011

Earthquake engineering is characterized by large aleatory (random) and epistemic (knowledge)

*Correspondence to: Peter Fajfar, FGG, Jamova 2, SI‐1000 Ljubljana, Slovenia.

Copyright # 2011 John Wiley & Sons, Ltd.

2. DETERMINISTIC DETERMINATION OF SEISMIC DEMAND AT “FAILURE” BY THE

Sae;C ¼ μC ⋅Say (8a)

3. FRAMEWORK FOR PROBABILISTIC PERFORMANCE ASSESSMENT

PLS;x ¼H̃ðsa; C˜ ÞCH Cf Cx (9)

H̃ ðsa Þ ¼ ko ⋅ðsa Þ−k (11)

D̃ðsa Þ ¼ a⋅ðsa Þb (12)

4. A SIMPLIFIED PROCEDURE USING INCREMENTAL N2 INSTEAD OF IDA

An important simpliﬁcation of the probabilistic seismic assessment can be introduced by employing

5. ADDITIONAL PRACTICE‐ORIENTED SIMPLIFICATIONS

5.2. Simpliﬁed equations

6. PROBABILISTIC PERFORMANCE ASSESSMENT OF A FOUR‐STOREY RC

6.1. Description of the example structure and mathematical model

6.2. Estimate of the probability of “failure”

7. PROBABILISTIC PERFORMANCE ASSESSMENT OF TWO VARIANTS OF THE SPEAR

7.1. Description of the example structures and the mathematical model

7.2. Estimate of the probability of “failure”

You might also like