Bull Earthquake Eng (2011) 9:1137–1155

DOI 10.1007/s10518-010-9241-3

ORIGINAL RESEARCH PAPER

Simplified estimation of seismic risk for reinforced

concrete buildings with consideration of corrosion
over time

Daniel Celarec · Dimitrios Vamvatsikos ·

Matjaž Dolšek

Received: 15 February 2010 / Accepted: 15 December 2010 / Published online: 24 February 2011
© Springer Science+Business Media B.V. 2011

Abstract Throughout the world, buildings are reaching the end of their design life and
develop new pathologies that decrease their structural capacity. Usually the ageing process is
neglected in seismic design or seismic risk assessment but may become important for older
structures, especially, if they are intended to be in service even after they exceed their design
life. Thus, a simplified methodology for seismic performance evaluation with consideration
of performance degradation over time is presented, based on an extension of the SAC/FEMA
probabilistic framework for estimating mean annual frequencies of limit state exceedance.
This is applied to an example of an older three-storey asymmetric reinforced concrete build-
ing, in which corrosion has just started to propagate. The seismic performance of the structure
is assessed at several successive times and the instantaneous and overall seismic risk is esti-
mated for the near collapse limit state. The structural capacity in terms of the maximum base
shear and the maximum roof displacement is shown to decrease over time. Consequently,
the time-averaged mean annual frequency of violating the near-collapse limit state increases
for the corroded building by about 10% in comparison to the typical case where corrosion
is neglected. However, it can be magnified by almost 40% if the near-collapse limit state is
related to a brittle shear failure, since corrosion significantly affects transverse reinforcement,
raising important questions on the seismic safety of the existing building stock.

Keywords Seismic risk · Capacity degradation · Corrosion · Reinforced concrete

frame · Performance-based earthquake engineering · Static pushover

This article is based on short paper presented at the COMPDYN 2009 Conference (Rhodes, Greece).

D. Celarec · M. Dolšek (B)

University of Ljubljana, Ljubljana, Slovenia
e-mail: [email protected]

D. Vamvatsikos
University of Cyprus, Nicosia, Cyprus

123
1138 Bull Earthquake Eng (2011) 9:1137–1155

1 Introduction

Structures are exposed to aggressive environmental conditions which may cause different
types of structural damage. For example, wind, waves, corrosive environment, extreme tem-
peratures and earthquakes are the influences that can impact many existing structures every
day. Such environmental conditions can cause corrosion or material fatigue that may lead
to the extensive deterioration of mechanical properties of structural elements. Consequently,
the structural capacity degrades over time and considerable costs have to be incurred just to
maintain the serviceability of a structure and to assure its resistance to the loads that it was
designed for.
Driven by the frequent failures of bridge structures, the influence of corrosion on their
traffic load capacity has been widely researched. There, the effects of ageing are more severe
since the entire structure is exposed to the environment. Different studies (Val et al. 1998;
Estes and Frangopol 2001) show that deterioration of performance resulting from reinforce-
ment corrosion could have a significant effect on both serviceability and ultimate limit state
of bridge structures, and thus, have to be properly considered in system reliability assess-
ments. Until recently, most work has focused on the assessment of aseismic bridges, but the
latest studies (Choe et al. 2009; Kumar et al. 2009) demonstrate that the effect of corrosion
becomes even more meaningful if the bridges are subjected to the seismic load. Less has
been done for buildings, especially to quantify their degrading performance under seismic
loads (e.g. Berto et al. 2009). Thus, we propose to investigate the effect of environmental
corrosion on the seismic behaviour of reinforced concrete (RC) buildings in performance-
based earthquake engineering terms. This effort becomes especially significant for older RC
structures designed and constructed in the 1950–1960 era that are nearing the end of their
nominal design life. The fundamental understanding of the effect of weathering on our ageing
infrastructure will help us actually understand the performance of the structures during their
entire life, not just when they are still intact.
The corrosion of reinforcement, which arises from carbonation phenomena and chloride-
induced penetration, is one of the most important sources of deterioration for RC members
(Val and Stewart 2009). The deterioration process related to the corrosion of reinforcement
in general comprises two parts, that is, the corrosion initiation and corrosion propagation.
The corrosion initiation is the process of diffusion and direct ingress of aggressive agents
(e.g. chloride or carbon dioxide) through protective cover and cracks, while the corrosion
propagation, which starts when the concentration of those agents at bar surface exceeds
a threshold level, is related to formation of different damage in structural elements, such
as loss of cross-sectional area of reinforcing steel, reduction of ductility and mechanical
properties of reinforcing, reduction of bond and crack propagation like spalling and delami-
nation of concrete protective cover caused by extensive corrosion products (Val and Stewart
2009).
Although models that consider all of the above-mentioned phenomena do exist, they can
be cumbersome. For the purposes of our study, several simplifications were made in order to
provide a simple and efficient estimate of the influence of corrosion on the seismic risk of RC
structures. Therefore, a uniform corrosion was adopted along the longitudinal and transverse
reinforcement bars of exposed structural elements, a simple model compared to more accu-
rate spatial non-homogeneous pitting corrosion (Stewart 2009). The concrete spalling and
reduction of maximum bond stress between concrete and reinforcement (Berto et al. 2009)
bar were not included in the model. In other words, corrosion only influences the diameter of
the steel bar. Also, the evaluation of the time to corrosion initiation, which depends on a large
number of parameters such as the composition of concrete, its porosity and microstructure,

123
Bull Earthquake Eng (2011) 9:1137–1155 1139

the degree of pore saturation and the exposure conditions (Val and Stewart 2009), was not
considered; we are focusing on an existing structure in which the corrosion process has just
started to propagate. Therefore, the seismic risk was estimated for a time period of 50 years,
starting from the initiation of corrosion in the structure. Such a simplified approach can be
used for estimating the seismic risk in existing structures, which are expected to be in service
well beyond the time that corrosion has set in, as typically happens when they exceed the
lifetime that they were designed for.
Based on these simplifying assumptions, it would be attractive to use a closed-form expres-
sion to evaluate the performance of an ageing structure in terms compatible with current
performance-based earthquake engineering concepts, i.e. in terms of the mean annual fre-
quency (MAF) of exceeding a given limit state. One candidate is the engineering demand
parameter (EDP) based methodology introduced by Torres and Ruiz (2007) that will be
used for structural reliability evaluation in combination with the simplified seismic perfor-
mance assessment method (Fajfar 2000; Dolšek and Fajfar 2007, 2008). It is based on the
SAC/FEMA probabilistic framework proposed by Cornell et al. (2002), in addition to which,
the structural capacity is considered to change in time. Similarly, an intensity measure (IM)
based formulation is also considered as an alternative (Vamvatsikos and Dolšek 2010). Both
are able to provide an expected number of limit-state exceedance events and the overall time-
averaged MAF over the period of interest that can be compared to typical acceptable rates of
exceedance, e.g. the ubiquitous 2 or 10% in 50 years.
In the following sections, the proposed methodology is applied to an existing three-storey
asymmetric non-ductile RC frame building. Our aim is to walk the reader through all the steps
of a practical application on a realistic structure, taking shortcuts and making simplifications
where appropriate to derive a basic result that can help us determine whether corrosion is
worth considering when estimating the seismic performance of a given structure. Let us begin
by briefly summarising the probabilistic framework.

2 Summary of probabilistic framework

Existing methods for structural performance assessment, such as that proposed by Cornell
et al. (2002), are usually based on the estimation of mean annual frequency (MAF) of violat-
ing the designated limit states or performance goals, where the resulting MAFs are assumed
to remain constant throughout the design life of the structure. The MAF of exceedance of
predefined limit states during any given time interval is then equal to the instantaneous MAF
at any time. However, this cannot be the case under severe weathering conditions, where
significant performance degradation is expected to take a place and as a consequence the
MAF increases with time. Thus, additional effort has to be done to account for performance
degradation over time.
Since the numerical evaluation of the time integral adds another layer of complexity,
simplified methods for structural reliability evaluation considering the performance degra-
dation over time have recently been developed. In the present study we primarily use the
reliability framework introduced by Torres and Ruiz (2007). Their method is a straight-
forward extension of the probabilistic framework proposed by Cornell et al. (2002), with
the added assumption that the structural capacity C (τ ) is a random variable with a prob-
ability density function f C ( c| τ ), which changes with time. Therefore the conditional
probability of failure P [C (τ ) < D| y, τ ] is also a function of time and the expected
number of limit state exceedance events within the time interval (t0 , t0 + t) can be
expressed as:

123
1140 Bull Earthquake Eng (2011) 9:1137–1155

t0+t∞ ∞  
 d H (y) 
E [η (t0 , t)] =  
 dy  P [C (τ ) < D| y, τ ] f C ( c| τ ) dy dc dτ , (1)
t0 0 0

where P [C (τ ) < D| y, τ ] is the conditional probability of seismic capacity being less than
demand at a given seismic intensity y and time instant τ, f C ( c| τ ) is the conditional prob-
ability density function of seismic capacity C at time instant τ , and |d H (y)/dy| is the
absolute value of the derivative of the seismic hazard curve. The analytical solution of
Eq. (1) provides two different formulations, which were both used in our study, that is,
the Engineering Demand Parameter (EDP) based formulation and the Intensity Measure
(IM) based formulation, which differentiate on how the seismic demands (D) and structural
capacities (C) are defined. The formulations are explained in more detail in the following
sections.

2.1 The EDP based formulation

The EDP based formulation, proposed by Torres and Ruiz (2007), is based on the engineer-
ing demand parameter, i.e. a measure of the structural response. In that case, Eq. (1) can be
expressed in closed form, similarly as derived by Cornell et al. (2002). The expected number
of limit state exceedance events ηedp over the time interval (t0 , t0 + t) and the average
MAF λAVGedp can then be written as:
   2
k  2 
ηedp (t0 , t) = H âg,ls × exp
0
β + βCR 2
+ βDU
2
+ βCU
2
2 · b2 DR
ηedp
× (t0 , t) , λAVG
edp = , (2)
t
where H (·) is the mean seismic hazard function, k is the parameter of the power-law approx-
imation to the hazard curve, βDR is the dispersion measure for aleatory randomness in dis-
placement demand, βCR is the dispersion measure for aleatory randomness in displacement
capacity, βDU and βCU are, respectively, the dispersion measures for epistemic uncertainty
in displacement demand and capacity, and b is the exponent of the approximate power-law
0 is the seismic intensity measure, in our case
relationship between the IM and EDP. âg,ls
the peak ground acceleration, related to the median of the limit state capacity Ĉ (t0 ) at the
beginning of the evaluation time interval (t0 , t0 + t), and  is a derived parameter covering
the specified time interval:
1
Ĉ (t0 ) b
0
âg,ls = (3)
a

1− k 
(α + β · t0 ) b β · t b
 (t0 , t) = 1+ −1 . (4)
β (b − k) α + β · t0

The simplified approach presented above is based on several assumptions. The hazard curve
0 , and the relationship
has to be approximated considering an appropriate interval around âg,ls
between the IM and the EDP has to be fitted around the Ĉ (t0 ), both accomplished by the
following power-law expressions:
 
H âg = k0 · âg −k , D̂ (t) = a · âg b . (5)

123
Bull Earthquake Eng (2011) 9:1137–1155 1141

Additionally, the median capacity Ĉ (t) is assumed to vary linearly with time:

Ĉ (t) = α + β · t, α> 0 and β ≤ 0, (6)

where α in β are the parameters defining the linear function (note, that parameter β with-
out subscripts has no relation with those defining dispersion measures in Eq. (2)). It is also
assumed that all dispersion measures and the parameters of the relationship between the IM
and the EDP (a, b) are constant over the integration time interval t. The demand D(t) and
capacity C(t) are assumed to be lognormally distributed. The dispersion measures related to
demand and capacity are therefore defined as the standard deviation of the logarithm of the
demand and capacity, respectively.
The assumption of time-independent dispersion measures is not entirely accurate. It is
expected in reality that dispersion increases over time, since prediction of seismic response
parameters of an aging structure becomes more uncertain. However, assumption of time-
independent dispersion measure leads to a simple solution with a low error since average
dispersion measures, which incorporate the impact of aging process, can be estimated and
used in the process of risk assessment. Also an assumption that the parameters of the demand
are constant over time can be argued for the case of presented methodology. However, the
parameter b practically does not vary over time. Namely, the “equal displacement rule”
applies for wide range of structures in the case of the N2 method. In addition, also parameter
a is practically constant due to the fact that the corrosion of reinforcement, as modelled in
the case of presented example, practically does not affect the stiffness of the structure.
Among the above issues, the methodology is subject to some other well known limita-
tions (Cornell et al. 2002), namely, (a) the assumption of constant response dispersion for
all intensity levels, and (b) the power-law approximations for the seismic hazard curve and
for the relationship between seismic intensity and structural demand (Eq. (5)). Such errors
can be reduced by estimating the parameters (β-dispersions, a, b and k, k0 ) separately for
each individual limit state (see Sect. 3.4). Further, the assumption that the seismic capacity
changes linearly with time is justified only for systems for which the effect of degradation
does not become overwhelming over time. The latter issue and the problem associated with
the assumption that the dispersion measures in seismic demand and capacity are constant
over the whole integration time interval can be partly resolved by the improved version of
the equations available in Vamvatsikos and Dolšek (2010).

2.2 The IM based formulation

Unlike the case of the EDP formulation, the IM based approach (e.g. Cornell et al. 2002)
requires structural capacity expressed in terms of the earthquake intensity measure, which
can be for example, the spectral acceleration Sa or the peak ground acceleration ag . This alter-
native formulation has several theoretical advantages compared to the standard EDP-based
formulation. For example, an approximation of the relationship between the seismic intensity
measure and engineering demand parameter is not needed, and the IM based formulation also
does not require any correlation assumptions between demand and capacity.
Starting again from Eq. (1) and using similar assumptions to those presented earlier,
(Vamvatsikos and Dolšek 2010) have derived the expression for the expected number of
limit state exceedance events ηedp over the time interval (t0 , t0 + t) and the average MAF
λAVG
im as follows:

123
1142 Bull Earthquake Eng (2011) 9:1137–1155

   2 
k
ηim (t0 , t) = H âg,ls
0
· exp βRag
2
+ βUag
2
2
⎡ ⎤
0 1−k
âg,ls
⎣ 1+ γ · t ηim
×· − 1⎦ , λAVG
im = (7)
γ · (1 − k) 0
âg,ls t

where H (·) is the mean seismic hazard function, k is the parameter of the hazard curve
approximation, βRag and βUag are the dispersion measures in intensity measure for random-
0 is the median peak ground acceleration corresponding to the
ness and uncertainty, and âg,ls
0 of the Eq. (2)
predefined limit state (e.g. near collapse limit state). Note that parameters âg,ls
0
and Eq. (7) are in general different, since in the first case the âg,ls is determined indirectly
through EDP-capacity whereas in the later case it is determined directly through IM-capacity.
Expression (7) was derived considering the assumption that structural capacity Ĉ(t) var-
ies linearly in time. For mathematical convenience it was assumed that value of that linear
function at time t0 corresponds to the structural capacity at the same time, which is a further
simplification compared to the Torres and Ruiz (2007) formulation. Thus, the linear function
of structural capacity Ĉ (t) is defined with a single parameter γ only:

Ĉ (t) = Ĉ (t0 ) + γ · t, γ ≤ 0, (8)

where γ defines the slope of the linear function.

It is worth to emphasize that the IM-based formulation has an advantage in comparison
to the EDP-based formulation, since in the case of the IM-based formulation the capacity is
expressed in terms of the seismic intensity, which is determined by consideration of both the
degradation of deformation capacity over time and its impact on displacement demand that
increased due to the aging process, the latter being neglected in the case of the EDP-based
formulation. Further explanation on that issue is given in Sect.3.4.

2.3 Simplified method for seismic performance assessment of structures

Determination of the relationship between the seismic intensity measure and the engineering
demand parameter may become extremely time-consuming. This is especially the case if the
structural response is estimated with nonlinear dynamic analysis, which, even for a single
instant in the building’s lifetime, has to be performed for several intensity measures and dif-
ferent ground motion records, e.g. with incremental dynamic analysis (IDA) (Vamvatsikos
and Cornell 2002), in order to capture randomness due to earthquakes. In addition, analyses
will have to be performed for different times within the lifetime of interest to capture the effect
of corrosion. Therefore, for practical application, simplified analysis methods become very
attractive. In our study, the time-dependent relationship between the seismic intensity mea-
sure and the engineering demand parameter was determined with the incremental N2 analysis
(IN2) (Fajfar 2000; Dolšek and Fajfar 2007). It is a simplified nonlinear method for seismic
performance assessment of structures and represents an alternative to IDA (Vamvatsikos and
Cornell 2002).
The procedure for determination of the IN2 curve is explained elsewhere (Dolšek and
Fajfar 2007). For common structural systems with moderate or long fundamental period(s)
the IN2 curve results in a straight line. In this case the “equal displacement rule” applies, i.e.
b = 1.0 (Eq. 5), up to the “failure” point, which is conservatively represented by the near
collapse limit state. After “failure”, the IN2 curve becomes a horizontal line. In more general
cases, the IN2 curve can be approximated in the same way as a mean IDA curve, i.e. from

123
Bull Earthquake Eng (2011) 9:1137–1155 1143

two points of the actual curve, e.g. from the points representing the damage limitation and
near collapse limit states, or by regression over several discrete points.
Since the IN2 curve is only a “central” (mean or approximately median) result, it does
not contain any dispersion information. Therefore the dispersion values for randomness in
displacement demand βDR and displacement capacity βCR cannot be determined. A possible
alternative is using SPO2IDA instead, a simple-to-use tool that employs complex R-μ-T
relationships to provide the median and the 16th and 84th fractiles of demand or capac-
ity (Vamvatsikos and Cornell 2006). This allows the estimation of both the needed median
dispersion of capacity for complex pushover shapes and will provide accurate results. Alter-
natively, they can be estimated directly by performing IDA for the equivalent SDOF system
as also shown in Dolšek and Fajfar (2007). Still, our pushover shapes are practically
elasto-plastic, allowing the use of existing results from other works, (e.g. Dolšek and Faj-
far 2007; Ruiz-Garcia and Miranda 2003), who have found that the coefficient of variation
for the displacement of such SDOF systems varied from 0.4 for structures with a moderate
or long natural period, to 0.7 for structures with a short predominant period. The determi-
nation of dispersion measures for uncertainty in displacement demand βDU and capacity
βCU is in general possible to determine with the IN2 method, but requires a probabilis-
tic structural model, i.e. a model with appropriate consideration of parameter uncertainties,
which is not within the scope of this paper. For a simplified approach, it is convenient to
predetermine the dispersion measures for uncertainty. For example, dispersions for steel
frames have been proposed in the FEMA 350 (2000). They may serve as rough estimates
also for some other structural systems. For example, the total uncertainty dispersion mea-
 2 
2 0.5 = 0.35 was applied for a global inter-story drift evaluation
sure βUT = βDU + βCU
in the case of low-rise buildings within the SAC/FEMA seismic performance evaluation
(Yun et al. 2002). The dispersions in intensity measure due to randomness and model-
ling uncertainties for different types of structural systems and for the IM based approach
were recently proposed in FEMA P695 (2009). The values for record-to-record and mod-
elling-related variability are in range of 0.35 to 0.45 for randomness and 0.2 to 0.65 for
uncertainties.

3 Example of the three-storey RC frame building

As an example of our methodology, we will estimate the seismic risk for an existing frame
building, in which the reinforcement corrosion has just initiated. The structure is located in
a region of moderate seismic hazard and high level of reinforcement corrosion risk area, a
typical scenario for most coastal areas in the Mediterranean. The objective of the analysis is
to estimate the increase in seismic risk for two predefined near-collapse limit states as the
associated structural performance degrades with time.
The structure was analysed at different time instants (t0 + t), in which t0 relates to the
initial condition of the existing structure, when corrosion attack was detected, and t is the
time interval during which the propagation of corrosion is taking place. Since the evaluation
interval is taken to be [0, 50yrs], t0 was set to zero in Eqs. (2), (4) and (8). For the discus-
sions to follow, the term “initial structure” will relate to the structure at time t0 , at which it
was assumed that reinforcement corrosion did not affect the structural capacity yet, and the
term “degraded structure” will relate to the structure affected by corrosion at selected time
instants t.

123
1144 Bull Earthquake Eng (2011) 9:1137–1155

Fig. 1 a The elevation and plan view of example structure, and b typical cross-sections and reinforcement in
columns and beams

3.1 Description of the structure and model

The example structure is a three-storey asymmetric reinforced concrete frame building. This
structure was pseudo-dynamically tested within the European research project SPEAR (Seis-
mic performance assessment and rehabilitation of existing buildings, M. Fardis and P. Negro,
coordinators) and analysed in previous studies (e.g. Fajfar et al. 2006). The elevation, the
plan of the building and the reinforcement of typical cross sections of the columns and beams
are shown in Fig. 1. The structure was designed for gravity loads only.
The columns and beams of the structure were modelled by one-component lumped plas-
ticity elements, which consist of an elastic beam-column element and two inelastic rotational
hinges at the ends, defined by a moment-rotation relationship. These relationships were deter-
mined for the columns by properly taking into account their axial load and its interaction
with the moment capacity. Gravity loads for this RC structure amounted to 6.3 kN/m2 and
6.2 kN/m2 for first two stories and top storey, respectively.
Rigid diaphragms were assumed at the floor levels due to monolithic RC slabs. Conse-
quently the masses were lumped at the mass centres. The lumped masses and the corre-
sponding mass moments of inertia amounted to 65.5 t and 1,196 tm2 for the first two stories,
and 64.1 t and 1,254 tm2 for the top storey, respectively. The centreline dimensions of the
elements were used with the exception of beams which are connected eccentrically to the
column C6. Using centreline dimensions, the storey heights of 2.75 and 3.0 m, respectively,
for the first and upper two storeys, were assumed.
A schematic moment-rotation relationship is shown in Fig. 2. The yield (My ) and max-
imum moment (Mm ) was determined from appropriate section analysis. The characteristic
rotations, which describe the moment-rotation envelope of a plastic hinge, were determined
according to the procedure described by Fajfar et al. (2006). The zero moment point was

123
Bull Earthquake Eng (2011) 9:1137–1155 1145

Fig. 2 Schematic
moment-rotation relationship
of a plastic hinge in columns
and beams

assumed to be at the mid-span of the columns and beams. Therefore the rotations θy for
moment-rotation envelopes of columns and beams were calculated using the formula:

My · lspan
θy = , (9)
6 · E · Ieff

where lspan is the length of a beam or column, E is the modulus of elasticity and Ieff is the
effective moment of inertia of the element (0.5I ).
The near collapse rotation θnc,c in the columns, which corresponds to a 20% reduction
in the maximum moment, was estimated by means of the Conditional Average Estimator
(CAE) method (Peruš et al. 2006), whereas the near collapse rotation for hinges in beams
θnc,b were determined using the formula defined in Eurocode 8 (CEN 2005):
 0.35  
1   max (0.01; ω) 0.225
LV αρsx
f yw  
θnc,b = 0.016· 0.3ν · fc · · 25 fc
· 1.25100ρd .
γel max (0.01; ω) h
(10)

where γel is equal to 1.0 (mean values), parameter ν is the normalised axial load (for beams
ν = 0), ω and ω’ are the mechanical reinforcement ratios of the tension and compression lon-
gitudinal reinforcement, respectively, f cm and f yw are the mean strength of concrete ( MPa)
and yield strength of steel ( MPa), respectively, ρsx is the ratio of transverse steel parallel
to the direction of loading, ρd is the steel ratio of diagonal reinforcement in each diagonal
direction and α is the confinement effectiveness factor. All beams are defined as members
without detailing for earthquake resistance. Therefore the rotations at near collapse limit state
θnc,b are multiplied by 0.825. The post-capping or negative-stiffness part of moment-rotation
envelope is determined under the assumption that the ratio between the ultimate rotation θu,b
and rotation at maximum moment θm,b is 3.5.
The degradation of capacity over time was modelled only via the simplified model of cor-
rosion of longitudinal and transverse reinforcement in the external (i.e. exposed) elements of
the structure. In general, the corrosion decreases the diameter of reinforcement and the bond
stress between the concrete and steel bars. The later phenomenon and as well as spalling of
concrete were not considered in this stage of the study. Therefore in our model the corrosion
influences only the diameter of the steel bar. The reduced diameter Drb (t) of a reinforcing
steel bar with initial diameter of Db ( mm), which is subjected to corrosion for a time period
(years) t = t − t0 (t0 relates to the initial condition of the building and is equal to 0) is,
according to Pantazopoulou and Papoulia (2001):

Drb = Db − 0.023 · i corr · t, (11)

123
1146 Bull Earthquake Eng (2011) 9:1137–1155

where i corr represent the mean annual corrosion current per unit anodic surface area of steel
(μA/cm2 ). In our analysis the corrosion current i corr = 1.2 μA/cm2 was considered, which
corresponds to a high level of reinforcement corrosion (Pantazopoulou and Papoulia 2001).
In order to determine the structural capacity as a function of time, six structural models
M0 , M10 , M20 , M30 , M40 and M50 were prepared on the OpenSees platform (McKenna and
Fenves 2004) in combination with the PBEE toolbox (Dolsek 2010). These correspond to
the initial condition of the building at time t0 = 0 and to the degraded structure (affected by
corrosion) after 10, 20, 30, 40 and 50 years of corrosion propagation, respectively. The first
three modal periods of the initial 3D structure are: T1 = 0.80 s, T2 = 0.67 s and T3 = 0.54 s
and only slightly change with time as the corrosion insignificantly reduces the stiffness.
Therefore, the increase of the natural periods over time is negligible. Note that the period
T2 and the associated mode shape actually correspond to the first translational mode in the
Y direction. Therefore the second mode shape will be exclusively used for applying the N2
method.

3.2 Definition of the limit states and nonlinear static analyses

The MAF of exceedance was determined for a near collapse limit state and for additional
limit state which is related to the potential brittle failure and defined by a storey shear
demand/capacity ratio (DCR) equals 0.5. The reason for such a definition of the second limit
state is described later on in this Section.
According to European standard Eurocode 8 (CEN 2005) the near collapse (NC) limit
state at the element level is defined with the ultimate rotation θnc , which corresponds to 20%
drop of moment in the softening range of the moment-rotation relationship (Fig. 2), and it is
related to a ductile collapse mechanism. The NC limit state can also be defined on the basis
of the shear strength, which is related to a brittle collapse mechanism. In this case the shear
strength (capacity) of an element was calculated according to Eurocode 8 (CEN 2005):
   
1 h−x pl
VR = min (N ; 0.55Ac f c ) + 1 − 0.05 min 5; μ ·
γel 2L V
 
LV 
0.16 max (0.5; 100ρTOT ) 1 − 0.16 min 5; f c Ac + VW , (12)
h
where γel was assumed 1.0 in order to get the mean value of the shear strength, h is the depth of
cross-section, xis the height of compressive zone, L v is the moment-shear ratio (M/V ), Ac is
the cross-section area calculated as bw · d, f c is the concrete compressive strength, ρtot is the
total longitudinal reinforcement ratio and Vw is the contribution of transverse reinforcement
to shear resistance, taken as equal to:
Vw = ρw · bw · z · f yw , (13)
where ρw is the transverse reinforcement ratio, bw is the width of the rectangular web of
the cross-section, z is the length of the internal lever arm and f yw is the yield stress of the
transverse reinforcement.
Since the limit state at the building level is not defined in the European standard it was
simply assumed that the near collapse limit state appears whether the base shear strength
of structure, in relation to its static pushover curve, reduces to 20% of its maximum value,
or the storey shear demand/capacity ratio (DCR) at any storey is equal or exceeds 1. Such
definition is needed since the elastic model of shear was used in the analysis. In this place it
is worth to emphasize that the structure under investigation is not sensitive to a brittle failure

123
Bull Earthquake Eng (2011) 9:1137–1155 1147

Fig. 3 a the pushover curves for different time instants and the points indicating the NC and potential brittle
limit state and b the damage in columns and beams at NC limit state for initial condition (t0 )
Table 1 The values for the maximum base shear Fmax and the roof displacements corresponding to the NC
and potential brittle limit state

Condition Time instant Max. load NC limit state Potential brittle limit state
t (y) Fmax (kN) Dnc (m) Dpb (m)

Initial 0 278 0.244 0.053

Corroded 10 272 0.242 0.046
20 266 0.240 0.041
30 260 0.238 0.037
40 254 0.236 0.034
50 248 0.235 0.031
(Fmax,50 − Fmax,0 )/Fmax,0 −10.8% – –
(Dls,50 – Dls,0 )/Dls,0 a – −3.7% −42%
aD
ls corresponds to the displacements at NC or potential brittle limit state

(shear DCR never exceeds 1) although corrosion significantly affects transverse reinforce-
ment. However, we wanted to show how much the reduction of transverse reinforcement due
to corrosion affects the MAF of violating the potential brittle limit state. Therefore, the MAF
was estimated for an additional limit state (i.e. a potential brittle limit state), which is defined
by a storey shear demand/capacity ratio (DCR) at any storey equals 0.5. Note that this limit
state limit state is not related to near collapse limit state, but is a good indicator of the effect
of corrosion on a potential shear failure.
In general, nonlinear static analysis is performed independently in the X and Y direction.
For brevity, results are presented only for pushover analysis in the Y direction of the global
coordinate system that is in the positive direction of the strong side of column C6 (Fig. 1).
The influence of the unsymmetrical plan of the structure on the results of the analyses is prac-
tically negligible. The imposed horizontal loads were determined by the product of storey
masses and modal shape (Fajfar 2000), and are presented later in Sect. 3.3. The displacement
of the structure was monitored at the mass centre at the roof.
The nonlinear static (pushover) analyses were performed for six structural models (M0 ,
M10 , . . ., M50 ), starting with the conditions of the initial structure, and then simulating struc-
tural response after every 10 years, to provide the capacity of the structure as a function of
time. The results of the nonlinear static analysis (pushover curves) are presented in Fig. 3.
The values for the maximum load resistance Fmax and the associated displacements Dnc and
Dpb , which correspond to the NC limit state and potential brittle limit state, are presented in
Table 1.

123
1148 Bull Earthquake Eng (2011) 9:1137–1155

The base shear strength versus weight ratio starts at 14% for the initial structure and
decreases with age under the influence of the corrosion. After 50 years of corrosion prop-
agation the strength of the structure is reduced by about 11%. The difference in the roof
displacement in the NC limit state due to ageing of the structure is not as important as the
difference in the strength of the structure as it only amounts to 3.7%. The reason is that area of
the longitudinal reinforcement does not have an important influence on the ultimate rotation
in plastic hinges (Fig. 3). On the other hand the roof displacement that corresponds to the
potential brittle limit state is significantly reduced by about 42%.

3.3 Seismic capacity at different time instants

The definition of the structural capacity as a function of time differs depending on the prob-
abilistic formulation. The EDP based formulation requires the structural capacity expressed
in terms of an appropriate engineering demand parameter, whereas the IM based formula-
tion depends on the structural capacity being expressed via the intensity measure, e.g. peak
ground acceleration or spectral acceleration.
In the case of the EDP-based formulation, the structural capacity is defined in terms of
the maximum roof displacement that corresponds to the predefined limit states, whereas for
the IM-based formulation, the structural capacity is defined with the lowest peak ground
acceleration that causes violation of each limit state. The relation between the maximum
roof displacement and the peak ground acceleration is computed with the N2 method (Fajfar
2000) for the seismic load, which is defined with the elastic response spectrum according to
Eurocode 8 (CEN 2004) for soil class C (S = 1.15, TC = 0.6s).
The roof displacements that correspond to the separate limit states were already reported
in Sect. 3.2. In this Section determination of peak ground acceleration capacity ag,nc is explic-
itly demonstrated only for the NC limit state and for the initial building condition, i.e. for
model M0 , while for degraded structures and for the potential brittle limit state only the final
results are presented.
Once the results of pushover analysis are available, the pushover curve has to be idealized
as shown in Fig. 4a in order to determine the properties of the equivalent single degree of
freedom (SDOF) model. The results of this idealization are the yield force Fy and yield
displacement Dy . The properties of the SDOF system are then determined by dividing the
corresponding properties of the MDOF system by the transformation factor :
m SDOF 
=  = 1.27; m SDOF = m i · φi = 128.6 t, (14)
m i · φ i2
where m SDOF is the mass of SDOF system, m = {65.6, 65.6, 64.1} is a vector of storey
masses in tons, and φY = {0.28, 0.70, 1.00} is the mode shape vector (normalized by its
roof component) of the predominant translational mode shape in the Y direction, which in
our example corresponds to T2 = 0.67s. The yield point of the equivalent SDOF system is
obtained simply as dy∗ = Dy /  = 0.025m and f y∗ = Fy /  = 220kN. The period of the
SDOF system is calculated as follows:

m SDOF · dy∗
TSDOF = 2 · π · = 0.753 s. (15)
f y∗

The periods of all equivalent SDOF systems over all building ages are obviously within
the medium-period range of the spectrum (Fig. 4) and exceed the characteristic period TC ,
which is the corner period between the constant acceleration and constant velocity ranges

123
Bull Earthquake Eng (2011) 9:1137–1155 1149

Fig. 4 a Idealization of the pushover curve for model M0 and b the AD format for the equivalent SDOF
system

in an idealized Newmark-Hall type spectrum. Therefore, the equal displacement rule can be
applied for the determination of the mean (or approximately median) spectral acceleration,
which corresponds to the NC limit state. Therefore the Sae,nc is:
2
∗ 2·π
Sae,nc = dnc · = 13.4m/s2 (16)
TSDOF

where dnc∗ = D / , and the mean/median peak acceleration a

nc g,nc corresponding to the NC
limit state is determined from the elastic spectrum as follows:

Sae,nc (TSDOF ) TSDOF

ag,nc = · = 5.87 m/s2 = 0.598 g, (17)
S · η · 2.5 TC

where S = 1.15 is a soil factor for soil type C and η is a damping correction factor with
value η = 1 for 5% viscous damping. The above evaluation of the ag,nc can be presented in
the acceleration-displacement (AD) format together with the capacity diagram of the SDOF
system (Fig. 4b).
This procedure was repeated to evaluate the peak ground acceleration capacity that corre-
sponds to the NC and potential brittle limit states and for all degraded structures (models M10
to M50 ). The resulting peak ground acceleration capacities ag,nc and ag,pb are presented in
Table 2. The reduction in peak ground acceleration capacities for both limit states is similar
to that shown before for the maximum roof displacement (Sect. 3.2, Table 1) and is about
4.5% for the NC limit state and 40% for the potential brittle limit state.
The results presented in Table 2 are used as input data for the definition of the seismic
capacity as an approximately linear function of time (Eqs. 6 and 8). tHE parameters α, β and
γ that approximate the capacity as a linear function were calculated using linear regression
(i.e. the method of least squares) for the EDP and IM-based format. The results are presented
in Table 3. The roof displacements corresponding to each limit state defined at each time
instant are presented in Fig. 5. The fitted lines approximate the linear decrease of the median
capacity with time after corrosion initiation.

3.4 Instantaneous and overall seismic risk

The expected number of exceedance events and average MAFs of the NC and potential brittle
limit state were calculated by means of Eq. (2) and (7) for the initial conditions at time t0 = 0,

123
1150 Bull Earthquake Eng (2011) 9:1137–1155

Table 2 The peak ground acceleration capacities ag,nc and ag,pb for the NC and potential brittle limit states

Condition Time instant NC limit state Potential brittle limit state

t (y) ag,nc (g) ag,pb (g)

Initial 0 0.598 0.131

Corroded 10 0.592 0.114
20 0.586 0.103
30 0.581 0.093
40 0.575 0.086
50 0.571 0.079
(ag,ls,50 − ag,ls,0 )/ag,nc,0 a − 4.5% −40%
aa
g,ls corresponds to the peak ground acceleration capacities for NC or potential brittle limit state

Table 3 The parameters defining the linear functions for structural capacity in time

Formulation Units NC limit state Potential brittle limit state

EDP-based m, year α = 0.244m, β = −1.86 × 10−4 m/y α = 0.051m, β = −4.0 × 10−4 m/y
IM-based g, year γ = −5.60 × 10−4 g/y γ = −1.1 × 10−3 g/y

Fig. 5 The values of maximum roof displacement and peak ground accelerations capacities (ag,ls ) at prede-
fined building ages and their linear approximation for the NC and potential brittle limit state. The results are
presented for a EDP- and b IM-based format, respectively

which corresponds to initial structure, and for different time periods, i.e. 10, 20, 30, 40 and
50 years.
The moderate seismic hazard typical for the South-East part of Slovenia (Dolšek and
Fajfar 2008), was adopted in the procedure for the estimation of the seismic risk. This was
approximated by a two-parameter seismic hazard function, derived separately for both limit
states. The associated parameters k and k0 were determined by locally fitting the hazard curve
with the function H (ag ) = k0 · (ag )−k (Cornell et al. 2002). The fitting was performed over
the interval from 0.25 ag,ls (t0 ) to 1,25 ag,ls (t50 ) (Dolšek and Fajfar 2008), where ag,ls (t0 ) is
the median peak ground acceleration capacity of the initial structure for the selected limit
state, and ag,ls (t50 ) is the corresponding capacity of the degraded structure after 50 years, that
is, at the end of the considered time interval. Since ag,ls (t) differs for the NC and the potential
brittle limit state, as marked on the curve on Fig. 6, the parameters of the hazard curves also
differ for both defined limit state, a fact that helps improve the accuracy of the closed form

123
Bull Earthquake Eng (2011) 9:1137–1155 1151

Fig. 6 The seismic hazard curve

and the median points for when
the NC and the potential brittle
limit state is achieved

approximation in Cornell et al. (2002). The parameters are k = 3.50, k0 = 6.40 × 10−6 for
the NC limit state and k = 1.36, k0 = 2.73 × 10−4 for the potential brittle limit state.
Note that parameter k need not be used in Eqs. (2) and (7) since the hazard corresponding
to ag,ls (t) can be determined directly from the hazard curve. The seismic hazard curve used
for peak ground acceleration is presented in Fig. 6. The intensities for return periods 225,
475 and 2,475 years are 0.130, 0.181 and 0.325 g, respectively, and the mean/median peak
ground acceleration capacities derived from IN2 for the initial structure, are 0.530 and 0.121
g for the NC and potential brittle limit state. As corrosion propagates, these will decrease
with time.
The dispersions for randomness in displacement demand (βDR ) and capacity (βCR ) were
considered to be equal to 0.4 and 0.2, respectively. These values are based on a previous study
(Dolsek 2009), where dispersion measures for βDR and βCR were determined based on the
result of IDA analysis. For example, dispersion measures βDR and βCR , which were estimated
for the near-collapse limit states, amounted to 0.41 and 0.17, respectively. The dispersions
for uncertainty in roof displacement demand (βDU ) and capacity (βCU ) were considered the
same and amount to 0.25. These values were chosen to produce total uncertainty dispersion,
 2 
2 0.5 = 0.35 as suggested in FEMA 350 (2000). Note that
βUT , equal to βUT = βDU + βCU
very similar values for βDU and βCU were calculated also in Dolsek (2009). Unlike the EDP-
based formulation, the IM-based approach requires only the dispersion in intensity, which
was assumed to be equal to 0.40 for both randomness and uncertainty. The latter value is
recognized as a recommendation of FEMA P695 (2009) although it can significantly vary
from case to case.
The values for the expected number of exceedance events of the NC and potential brit-
tle limit state at different times and the corresponding values of the average MAFs, using
both the EDP and IM formulation, are collected in Tables 4 and 5. The results are com-
pared with the case, in which the degradation of structural performance was neglected. In
the latter case, the instantaneous MAFs of exceedance are all equal to the initial MAF at
time t0 , and obviously equal to the time-average MAF as silently assumed in all typical per-
formance-based earthquake engineering calculations. Note that values corresponding to this
comparison are marked bold in Tables 4 and 5.
The expected numbers of exceedance events for the potential brittle limit state per time
interval of 50 years, using the EDP and IM based formulation, is estimated at 0.39 and 0.40,
respectively, and it is about 39 and 43% higher than the case in which the degradation was not
considered. Note that it is assumed that the potential brittle limit state is detected when storey
shear demand/capacity ratio at any storey equals to 0.5. On the other hand, the corrosion has

123
1152 Bull Earthquake Eng (2011) 9:1137–1155

Table 4 The expected number of exceedance events ηedp and the average MAFs λAVG
edp for the NC and
potential brittle limit state using the EDP-based formulation

Condition Time interval NC limit

 state   Potential
 brittle
 limit state
 
t (y) ηedp ×10−2 λAVG
edp ×10−4 ηedp ×10−2 λAVG
edp ×10−4

No degradation 50 1.38 2.77 27.9 5.59

Corroded 10 0.28 2.80 5.91 5.91
20 0.57 2.84 12.5 6.27
30 0.86 2.88 20.1 6.69
40 1.17 2.92 28.7 7.18
50 1.48 2.96 38.8 7.77
(xdeg,50 − xnodeg,50 )/xnodeg,50 a + 7.1% + 7.1% + 39% + 39%
a x corresponds to η orλ

Table 5 The expected number of exceedance events ηim and average MAFs λAVG
im for the NC and potential
brittle limit state using the IM based formulation

Condition Time interval NC limit

 state   Potential
 brittle
 limit state
 
t (y) ηim ×10−2 λAVG
im ×10−4 ηim ×10−2 λAVG
im ×10−4

No degradation 50 1.34 2.68 27.8 5.56

Corroded 10 0.27 2.73 5.90 5.90
20 0.55 2.77 12.6 6.30
30 0.84 2.82 20.2 6.76
40 1.14 2.87 29.3 7.30
50 1.46 2.92 39.8 7.97
*(xdeg,50 − xnodeg,50 )/xnodeg,50 a + 8.8% + 8.8% + 43% + 43%
a x corresponds to η orλ

only slightly influenced the moment capacity of the structural elements. Thus, the expected
number of exceedance events for the NC limit state per time interval of 50 years is increased
by only for 7.1 and 8.8%, if considering the performance degradation over time and amounts
to 1.48 × 10−2 and 1.46 × 10−2 , depending on which formulation (EDP or IM) is used.
The results of the EDP-based formulation are also presented in Fig. 7. The continuous
curve represents the expected number of exceedance events for both limit states considering
the performance degradation over time and the dashed line represents the case when the
degradation was neglected. Note that the Fig. 7a relates only to the NC limit state, while
in Fig. 7b, the expected number of exceedance events for the potential brittle limit state is
compared to that for the NC limit state, in order to highlight the differences between the two
limit states. Namely, the expected number of exceedance events for the potential brittle limit
state exceeds that for the NC limit state by about 26 times.
Note that insignificant difference in the results obtained by the EDP- or IM-based for-
mulation is also the consequence that the corrosion process, according to the model used
in this study, does not significantly influence the displacement demand as long as it is not
very close to the global dynamic instability of the structure. In the case if the aging process
significantly influences the displacement demand for a given seismic intensity, then it is rec-
ommended to use the IM-based formulation, since in this case the time-dependant demand is

123
Bull Earthquake Eng (2011) 9:1137–1155 1153

Fig. 7 The expected number of exceedance events of a NC limit state and b potential brittle limit state using
the EDP based methodology

indirectly incorporated in the capacity estimation, which in the case of IM-based formulation
is expressed in terms of the intensity measure.

3.5 Discussion of the results

One important outcome of the study is the comparison between the EDP and IM-based for-
mulations. The difference in the results amounts to 2 and 10% for NC and potential brittle
limit state, respectively, and it is practically negligible. Basically, the only sources of differ-
ence are the values of the dispersion measures, which in our case were assumed as average
values of the reported dispersions from the literature. All other parameters (see Eqs. (2) and
(7)) practically do not contribute to the differences in the final results obtained by the EDP-
or IM-based approach, especially, since the equal displacement rule applies (e.g. b = 1 in
Eq. (5b)). This confirms the sound basis of both formulations. Still, we expect them to dif-
fer more when the power law approximation of Eq. (5b) is no longer accurate, or when we
approach global dynamic instability where the EDP formulation becomes inaccurate (see
Vamvatsikos and Dolšek 2010 and references therein). Finally, the selection of the value of
the dispersion measure becomes more significant when the value of the hazard curve slope k
is high, for example, for the ductile limit state (k = 3.5), something that may cause different
results in the two formulations.
For the potential brittle limit state the corrosion significantly increases the seismic risk
since the corrosion of the shear reinforcement has a relatively greater influence on the shear
capacity. This becomes apparent in the static pushover curves, where, consequently, the roof
displacement that corresponds to that limit state decreases by about 4.0 mm per decade or
2.2 cm in 50 years. This is a relatively large reduction compared to the initial condition,
in which the maximum roof displacement at potential brittle limit state amounts to 5.3 cm.
However, this is not true for the case of the NC limit state, since the corrosion of longitudinal
reinforcement practically does not have any influence on the calculated ultimate rotations
in the plastic hinges. As a result, the maximum roof displacements at the ductile NC limit
state for different instants of time decreases by just a few percents, that is 0.9 cm in 50 years
(Table 1).
In addition to the averaged results in Tables 4 and 5, the instantaneous MAFs were calcu-
lated (Cornell et al. 2002), which via the EDP-based formulation can be estimated as
 2
k  2 
λedp (t) = H âg,ls (t) · exp 2
βDR + βCR2
+ βDU
2
+ βCU
2
, (18)
2b

123
1154 Bull Earthquake Eng (2011) 9:1137–1155

where the seismic intensity measure âg,ls (t) is related to the selected limit state capacity
at time t. Such values may be of interest as they represent the MAFs derived by a perfor-
mance evaluation (based on current probabilistic frameworks) that takes place at that instant
in the building’s lifetime. Therefore the instantaneous MAFs may heavily influence future
rehabilitation decisions. For example, the MAFs of exceedance of the NC and potential
brittle limit states for the initial structure using the EDP-based formulation are estimated at
0.0276 × 10−2 and 0.559 × 10−2 , respectively. Otherwise, the corresponding instantaneous
return periods are around 3620 and 180 years, respectively. For comparison, the instantaneous
return periods for exceedance of the two limit states after 50 years of corrosion propagation
decrease to 2930 and 75 years, respectively.

4 Conclusions

A simplified methodology has been presented for estimating the seismic performance of
ageing RC structures. Considering the deterioration of longitudinal and shear reinforcement
due to corrosion, and utilizing simplified analysis techniques within a SAC/FEMA-like prob-
abilistic framework, we are able to estimate the changing mean annual frequency (MAF) of
limit state exceedance as it worsens with time. Finally, the time-average of the MAF of limit
state exceedance is quantified over a continuous time period, providing us with a cumulative
single measure of the structure’s performance as ageing sets in. Two different approaches
were demonstrated to achieve these estimates, based on the EDP and IM formulation of the
SAC/FEMA probabilistic framework, the latter being suitable for all limit states, even close
to global collapse.
In our case-study of a 3-story non-ductile RC structure, both approaches were shown to
produce similar results, as long as we properly assign the values of dispersion, especially,
if the hazard curve slope k is relatively steep. Thus, corrosion is shown to have moder-
ate influence on the moment capacity of beams and columns, while their shear capacity
was heavily degraded. However, the later issue is not reflected in the average MAF of vio-
lating the NC limit state, since the structure is not sensitive to brittle failure. Therefore,
we defined a potential brittle limit state, which indicates shear capacity degradation, and
observed around 40% increase of the average MAF of exceedance events at which the storey
shear demand/capacity ratio exceeds 50%. This is a significant increase that simply cannot
be ignored if structures contain shear-critical members, something that tends to be the norm
in older RC buildings.
It is envisaged that further refinement of our corrosion model with inclusion of concrete
spalling and bond degradation will additionally increase the estimated seismic risk of ageing
RC structures. Thus, further verifications of such results are needed in order to better under-
stand the actual risks faced by our ageing infrastructure and appropriately amend our design
codes.

Acknowledgments The authors wish to acknowledge the support of the Cyprus Research Promotion Agency
under grant CY-SLO/407/04 and of the Slovenian Research Agency under grant BI-CY/09-09-002.

References

Berto L, Vitaliani R, Saetta A, Simioni P (2009) Seismic assessment of existing RC structures affected by
degradation phenomena. Struct Saf 31:284–297

123
Bull Earthquake Eng (2011) 9:1137–1155 1155

CEN (2004) Eurocode 8: design of structures for earthquake resistance. Part 1: general rules, seismic actions
and rules for buildings, EN 1998-1. European committee for standardisation, Brussels, December 2004
CEN (2005) Eurocode 8: design of structures for earthquake resistance. Part 3: strengthening and repair of
buildings. EN 1998-3, European committee for standardisation, Brussels, March 2005
Choe D, Gardoni P, Rosowsky D, Haukaas T (2009) Seismic fragility estimates for reinforced concrete bridges
subject to corrosion. Struct Saf 31:275–283
Cornell CA, Jalayer F, Hamburger RO, Foutch DA (2002) Probabilistic basis for 2000 SAC federal emergency
management agency steel moment frame guidelines. J Struct Eng 128:526–533
Dolsek M (2009) Incremental dynamic analysis with consideration of modeling uncertainties. Earthq Eng
Struct Dyn 38(6):805–825
Dolsek M (2010) Development of computing environment for the seismic performance assessment of rein-
forced concrete frames by using simplified nonlinear models. Bull Earth Eng 8(6):1309–1329
Dolšek M, Fajfar P (2007) Simplified probabilistic seismic performance assessment of plan-asymmetric build-
ings. Earthq Eng Struct Dyn 36:2021–2041
Dolšek M, Fajfar P (2008) The effect of masonry infills on the seismic response of a four storey reinforced
concrete frame–a probabilistic assessment. Eng Struct 30:3186–3192
Estes AC, Frangopol DM (2001) Bridge lifetime system reliability under multiple limit states. J Bridge Eng
6(6):523–528
Fajfar P (2000) A nonlinear analysis method for performance-based seismic design. Earthq Spec 16(3):
573–592
Fajfar P, Dolšek M, Marušić D, Stratan A (2006) Pre- and post-test mathematical modeling of a plan-asym-
metric reinforced concrete frame buildings. Earthq Eng Struct Dyn 35:1359–1379
FEMA 350 (2000) Recommended seismic design criteria for new steel moment frame buildings. SAC joint
venture, federal emergency management agency, Washington, DC
FEMA P695 (2009) Quantification of building seismic performance. Prepared by the advanced technology
council for the federal emergency management agency, Washington, DC
Kumar R, Gardoni P, Sanchez-Silva M (2009) Effect of cumulative seismic damage and corrosion on the life
cycle cost of reifirced concrete bridges. Earthq Eng Struct Dyn 38:887–905
McKenna F, Fenves GL (2004) Open system for earthquake engineering simulation, Pacific Earthquake Engi-
neering Research Center, Berkeley, California http://opensees.berkeley.edu/
Pantazopoulou SJ, Papoulia KD (2001) Modeling cover-cracking due to reinforcement corrosion in RC struc-
tures. J Eng Mech 127(4):342–351
Peruš I, Poljanšek K, Fajfar P (2006) Flexural deformation capacity of rectangular RC columns determined
by the CAE method. Earthq Eng Struct Dyn 35:1453–1470
Ruiz-Garcia J, Miranda E (2003) Inelastic displacement ratios for evaluation of existing structures. Earthq
Eng Struct Dyn 32:1237–1258
Stewart MG (2009) Mechanical behavior of pitting corrosion of flexural and shear reinforcement and its effect
on structural reliability of corroding RC beams. Struct Saf 31:19–30
Torres MA, Ruiz SE (2007) Structural reliability evaluation considering capacity degradation over time. Eng
Struct 29:2183–2192
Val DV, Stewart MG, Melchers RE (1998) Effect of reinforcement corrosion on reliability of highway bridges.
Eng Struct 20:1010–1019
Val DV, Stewart MG (2009) Reliability assessment of ageing reinforced concrete structures – current situation
and future challenges. Struct Eng Inter 19(9):211–219
Vamvatsikos D, Cornell CA (2002) Incremental dynamic analysis. Earthq Eng Struct Dyn 31:491–514
Vamvatsikos D, Cornell CA (2006) Direct estimation of the seismic demand and capacity of oscillators with
multi-linear static pushovers through incremental dynamic analysis. Earthq Eng Struct Dyn 35(9):1097–
1117
Vamvatsikos D, Dolšek M (2010) Equivalent constant rates for performance-based seismic assessment of
ageing structures. Struct Saf 33(1):8–18
Yun S-Y, Hamburger RO, Cornell CA, Foutch DA (2002) Seismic performance evaluation for steel moment
frames. J Struct Eng ASCE 128(4):534–545

123