Post earthquake availability of damaged structures.

Application to Highway
Bridges

Paolacci F., Giannni R., Alessandri S.

Department of Structures, University of Rome Tre, Italy

ABSTRACT:
A procedure for the evaluation of the seismic risk of damaged structures due aftershocks is here proposed with
the aim to provide a tool for deciding whether, after a damaging mainshock, a structure can still be open for
either emergency or ordinary traffic. In particular the mean annual rate of collapse of a mainshock-damaged
structure due to aftershocks is calculated using the probabilistic aftershock hazard theory and calculating the
fragility curves with IDA technique. The novelty of the procedure consists in using the in-situ inspection results
in order to updated the fragility curves by Bayes theorem. For this purpose the correlation values between the
observed damage and the drift available in literature has been used. An application to a highway r.c. bridge has
shown the effectiveness of the procedure.

Keywords: Time-Varying Collapse Risk, aftershock hazard, fragility curves, damaged structure, bridges

1. INTRODUCTION

In the days immediately after a strong seismic event, the engineers are called to assume some very
rapid decisions on the usability of buildings and other constructions that have suffered limited
damages caused by the mainshock. These decisions are usually taken on the basis of few data, using
only synthetic evaluations based on the experience. The highest risk to which the damaged
constructions are exposed depends on the possibility that one or more aftershocks make worse the
situation up to the collapse. In fact, it is well known that a high magnitude earthquake is followed by
numerous aftershocks, generally of smaller intensity, but that sometimes can also equalize the
magnitude of the mainshock; consequently a damaged structure could collapse, both for its reduced
vertical and lateral load carrying capacity and local effects of the aftershocks. Recently some authors
(Yeo and Cornell, 2005, 2009), (Franchin and Pinto, 2009) have tried to introduce more rational and
analytical procedures that can guide the decision on the usability of damaged constructions
immediately after an earthquake through an aftershocks probabilistic hazard analysis (APSHA). The
application of analytical methods for a quick estimation of the usability of a building, introduces high
difficulties, both for the complexity of the structures, and for the definition of significant damage
levels, which have also to account for the behavior of some non structural elements; the application to
the bridges (particularly to the highway viaducts) seem to be more affordable, both because are
simpler structures and for the possibility of using few damage parameters.

The bridges are strategic structures because their unavailability can cause the interruption of the road
ways, prevent the ordinary means to reach the zone struck by the seismic event. This condition
becomes particularly penalizing during the post-earthquake emergency, when high amount of
transportation means have to be employed for the rescue operations. The decision to close to the traffic
a bridge of an important roadway, which connect the most struck zones, is therefore very delicate, and
it should be taken on the basis of the more possible rational and objective criteria. In a recent paper
(Mackie and Stojadinovic, 2006) the authors have studied the problem to correlate the intensity of the
event with the vertical load carrying capacity (traffic) of the bridge. The results are given in terms of
fragility curves for different percentages of traffic reduction. Nevertheless the correlation of the
residual vertical load carrying capacity with the seismic damage is very difficult and uncertain, both
for the limits of the current models and because the live loads often represent a limited fraction of the
bridge weight.
More promising seems to be the way proposed by (Franchin and Pinto, 2009) to relate the decision of
the usability of a bridge to the risk of aftershocks. In any case, a preliminarily evaluation of the
damage state of the structure is needed, adopting one or more parameters of the numerical models
used for the fragility curves of the damaged structure; a mostly used quantity is, for instance, the ratio
between the maximum horizontal deflection and the height (drift). Unfortunately, using only a in-situ
inspection it is impossible to evaluate the maximum drift due to the mainshock, unless a specific
instrumentation has been previously installed in the structures.

In practice, an engineer involved in the evaluation of the damage will be able to ascertain only the
effects of the deformations suffered at the base of the piers as cracks, spalling, buckling of the steel
bars. These levels of damage can be certainly correlated to analytical parameters as the drift (Berry
and Eberhard, 2006), even if the dispersion is very high, with important consequences on the
reliability of the results. Alternatively, being the epicenter and magnitude of the mainshock already
known (and possibly also the position and the length of the fault), the intensity level at the site of the
seismic event can be evaluated by attenuation laws, in terms of PGA or spectral ordinates. Starting
from this information, and using the fragility curves of the undamaged structure for different levels of
damage, the maximum drift can be calculated; nevertheless this result is affected by high uncertainties,
due to the dispersion of both attenuation laws and structural response; moreover this measure,
exclusively based on the analysis, has the conceptual defect of not accounting for the observation of
the facts, with the risk it behaves as those people that " They don't believe in the facts, they believe
only themselves. When it comes to the point The facts must go by the board " (B. Brecht, Praise of
doubt).
In this paper, both the above means are synergically used. The expected damage, calculated on the
basis of the seismological data and fragilities curves of the structure, can be used not only for
identification of the more probable damaged structures, but also as prior probability to be updated with
a Bayesian approach, using the observed damages and considering the conditional probability as
likelihood function, obtained by a statistic analysis on the results of laboratory tests (Berry and
Eberhard, 2006).
 
 
2. DESCRIPTION OF THE METHOD

2.1 Evaluation of the fragility curves

The first phase of the procedure, to be carried out before the earthquake occurs, consists of evaluating
the fragility curves of undamaged and damaged structure. As a parameter of damage measure is
adopted (in the following the maximum drift will be adopted, but obviously other choices are
possible), a series of levels of increasing intensity is established ( ϕ1 , ϕ 2 , K ); then a wide sample of
accelerograms and a representative parameter of the local intensity (e.g. PGA or spectral ordinate) are
chosen. By the analysis on a proper model the fragility curves of the undamaged structure are
determined, for instance, using the IDA procedure (Vamvatsikos and Cornell 2002), in order to
calculate the probabilities functions p0 (ϕ i | y ) of the damage ϕi , versus the local intensity y of the
earthquake. It is also possible to determine for each accelerogram the intensity y jk for which the
accelerogram kth produces in the structure the damage ϕ j . It is possible, for each accelerogram, to
induce in the structure the damage level jth, and then, starting from that point, to apply another series
of accelerograms with increasing intensity, in order to simulate the effects of the aftershocks. It is
necessary to repeat the IDA procedure for a number of times equal to the product between the number
of levels of damage and the number of the accelerograms selected for simulating the mainshock.
Altogether, the required number of analysis is rather large: nan = ndl × nam × naa × nil , in which ndl is
the number of the considered damage levels, n am is the number of the accelerograms used for the
mainshock, naa is the number of the accelerograms used for the aftershocks and nil is the number of
the intensity levels considered in the IDA procedure. Usually, nan is high; nevertheless, since the
structures of the bridges can often be modeled with simple structural schemes, the time of calculation
remains still acceptable. The extension of the procedure to more complex structures, for which longer
times of calculation are necessary, will require a more efficient procedure then IDA. At the end of this
preliminary phase, we have both the fragility curves of the undamaged structure p0 (ϕ i | y ) , and the
( )
probability FD ϕ i | ϕ j , y that the structure, with an initial damage level ϕ j reaches or exceeds the
damage level ϕi , as function of the local intensity measure y of the aftershocks.
2.2 Damage evaluation

After a seismic event, the magnitude m0 and the position r of the epicenter are quickly available; in this
paper we will suppose that the direction of the fault line is known as well, whereas, the length of the
rupture zone, if not directly known, can be estimated as function of the magnitude as indicated in
(Wells and Coppersmith, 1994). With such a data and using a proper attenuation law (Sabetta-
Pugliese, 1987), (Ambraseys et al., 1996), (Ambraseys et al., 2005), it is possible to find the average
value and the standard deviation of the select parameter of local intensity as function of the magnitude
and the epicentral distance (or minimum distance from the fault). Assuming for y a log-normal
distribution, the conditional probability density of the intensity y is:

⎡ (ln y − μ 0 )2 ⎤
p y ( y | m0 , r0 ) =
1
exp⎢− ⎥ (2.1)
2π ζ 0 y ⎢⎣ 2ζ 02 ⎥⎦

( (
in which ζ 0 = ln 1 + σ y / y ) ) and μ0 = ln y − ζ 02 / 2 .
2

Consequently the probability of exceeding the damage level ϕi can be calculated by integration of the
product between seismic hazard and fragility:
∞
p0 (ϕ i ) = ∫ p0 (ϕ i | y ) p y ( y | m0 , r0 ) dy (2.2)
0

The previous function provides the probability of the expected damage in the structure, conditioned to
the occurrence of an event of magnitude m0 and distance r0. This (probabilistic) damage estimation
does not account for what is really happened to the structure. As stressed above, the post-earthquake
observation of the damage does not allow, at least for moderate intensity events, to evaluate the
damage index ϕi used in the analytical model without a permanent instrumentation placed on the
structure. On the basis of laboratory tests, Berry e Eberhard (Berry and Eberhard, 2006) have
developed relationships that provides, as function of the characteristics of the reinforced concrete
element (normal force, sectional area, length, percentage of reinforcement, etc..), the average value
and the coefficient of variation of the damage index ϕi (drift) corresponding to the observed damage
θk, ( k = 1, 2, 3 ), where θ1 = spalling, θ2 = buckling of the steel bars, θ3 = failure of the steel bars.
Assuming that the probability of ϕ conditioned to the damage θk can be expressed as a lognormal, the
previous data permit to determine the CDF of the drift ϕ conditioned to θk , Fϕ (ϕ |θ k ) . The probability
p (θ k | ϕ ) can be obtained observing that, if ϕk is the drift for the damage level θk, then ϕ < min k (ϕ k ) =
no damage, ϕ1 ≤ ϕ < min (ϕ 2 , ϕ 3 ) = damage θ1 , max (ϕ1 , ϕ 2 ) ≤ ϕ < ϕ 3 =damage θ2, and finally
max k (ϕ k ) ≤ ϕ = damage θ3. Consequently, if θ0 stands for no damage condition, we can assume:

[
p(θ 0 | ϕ ) = Pr min (ϕ k ) > ϕ =
k
] ∏ [1 − Fϕ (ϕ | θ )]
3

k =1
k k

∏ ( ) ∏ [1 − Fϕ (ϕ | θ )]
i 3
p(θ i | ϕ ) = Fϕ1 ϕ | θ j k k (i = 1,2) (2.3)
j =1 k = i +1

[
p(θ 3 | ϕ ) = Pr max(ϕ k ) ≤ ϕ =
k
] 3

∏ Fϕ (ϕ | θ )
k =1
k
k

where p (θ k | ϕ ) is the probability of the observed damage θk conditioned to the drift ϕ. If we consider
p0 (ϕ ) a prior probability distribution, estimated before the direct observation of the damage, and θk is
the level of the real observed damage, applying the Bayes theorem using p (θ k | ϕ ) as likelihood
function, the updated estimation of the probability can be obtained as:

p (θ k | ϕ ) p0 (ϕ )
p (ϕ | θ k ) = (2.4)
∫ p(θ k | ϕ ) p0 (ϕ ) dϕ
This probability function of the damage index ϕ combines the results of a predictive analysis carried
out on the numerical model of the structure with the information can be gathered from the visual
observation of the residual damage of the structure.

2.3 Aftershock hazard analysis

Since it is supposed that the main event has occurred and then magnitude, epicenter distance and
(possibly) fault line and position are known, it is possible to carried out an Aftershock Probabilistic
Hazard Analysis (APSHA), similarly to the procedure illustrated in (Yeo and Cornell, 2009). Unlike
the standard hazard analysis (PSHA), in which Poisson process of the seismic events is assumed to be
stationary, the aftershocks process is highly non-stationary; the average number of events at time t
with magnitude greater or equal to m following a mainshock of magnitude mm, is given by the
“modified Omori Law” (Utsu T.,1995) .

10 a + b (mm − m )
γ (t , m, m m ) = (2.5)
(t + c ) p
where a,b,c, and p are regional parameters that can be estimated on the basis of the past seismic events
(Reasemberg and Jones, 1989), (Lolli and Gasperini, 2003). Moreover, by introducing a truncation of
the magnitude m such that m1<m<mm (Reasemberg and Jones, 1989), the average number of daily
events with intensity greater than m is given by (Yeo and Cornell, 2009):

e − βm − e − βmm
γ (t , m, mm ) = γ 0 (t , m1 , mm ) = γ 0 (t , m1 , mm )[1 − FM (m )] (2.6)
e − βm1 − e − βmm

where β = b ln 10 and γ 0 (t , m1 , mm ) is the average number of the daily events at time t with magnitude
m1<m<mm.

10 a + b (m m − m1 ) − 10 a
γ 0 (t , m1 , m m ) = (2.7)
(t + c ) p
Given an aftershock of magnitude m and epicenter P, the local intensity at the site is deduced by the
attenuation law: ln Y = ψ ( m, r ) + ε , in which ψ is a function of magnitude m and distance r between the
structure and the epicenter (or the minimum distance from the fault), whilst ε is a zero-mean Gaussian
variable with standard deviation σε. The probability density of Y for a given event is then:

⎛ ln y −ψ (m, r ) ⎞
pY ( y | r ) = ⎟⎟ f M (m ) dm
1 mm

y ∫m1
φ ⎜⎜
⎝ σε ⎠
(2.8)

where the φ(.) is a standard normal PDF and f M (m ) = dFM (m ) / dm = βe βm / e − βm1 − e − βmm is the ( )
probability density of the magnitude of the aftershocks. As frequently happens, when ψ is a linear
function of m, the integral (2.8) can be developed in closed form. If R denotes the region where the
aftershocks can occur (in the following it will be assumed that this zone coincides with the length of
mainshock fault, similarly to (Yeo and Cornell, 2009)), and f R (r ) is the probability density function
of the distances r from the site, the probability of having an event with intensity y, conditioned to the
occurring of the event, is pY ( y ) = ∫R pY ( y | r ) f R (r ) dr .

2.4 Evaluation of the risk of aftershock

( )
The previously defined functions FD (ϕ i | ϕ j , y ) e p ϕ j | θ k have the following meaning; the first one
provides the probability that the structure, stricken by the mainshock with a damage level ϕj, could
exceeds the damage level ϕi ≥ ϕj because of an aftershock, while the second one is the probability that
the structure, stricken by the mainshock, has suffered the damage ϕi. Therefore, if an aftershock
occurred, the probability that a damage ϕ ≥ ϕi is:

φ ∞
(
Pr ( φ ≥ φi ) = ∫ i ∫ FD ( φi | φ j , y )pY ( y ) p φ j | θk dyd φ j
0 0
) (2.9)

At time t the daily mean number of the shaking that induce in the structure a level of damage greater
or equal to ϕi is therefore γ (ϕ i | t ) = γ 0 (t , m1 , mm ) Pr (ϕ ≥ ϕ i ) . If an aftershock cluster is represented by a
Poisson process, the probability (risk) that the damage is exceeded in the time [t , t + T ] is:

Pr (ϕ ≥ ϕ i | t , T ) = 1 − e − N a (ϕ i ,t ,T ) (2.10)

where:
t +T t +T
N a (ϕ i , t , T ) = ∫ γ (ϕ i | τ ) dτ = Pr (ϕ ≥ ϕ i | s ) ∫ γ 0 (τ , m1 , mm ) dτ =
t t

10 a + b (mm − m1 ) − 10 a
[ ]
(2.11)
= Pr (ϕ ≥ ϕ i | s ) (c + t )1− p − (c + t + T )1− p
p −1

3. DECISION-MAKING ANALYSIS

The probability evaluated with equation (2.10) represents the risk, due to an aftershock, that a damage
level ϕi in the structure is exceeded during the time interval T, starting from the time t; in the
following, ϕi will be assumed as collapse condition. This data is useful to assume a rational decision
on the usability of the construction (e.g. the transitability of a bridge). A reasonable criterion to orient
such a decision, consists of comparing the risk due to the aftershocks (provided by eq. (2.10)) with the
risk of the constructions due to the mainshock. Nevertheless, this comparison introduces considerable
conceptual difficulties.

In fact, in the Probabilistic Seismic Risk Analysis (PSRA), the process of the events is assumed as
stationary, therefore the mean number of the events that produce the damage ϕi in the time T is simply,
γ m (ϕ i )T , where γ m (ϕ i ) is the mean number of events in the unit time (generally one year). The
comparison among these two risks can simply be done comparing the frequencies, independently from
the used time unit, because of the simple proportionality with the time T. In case of aftershocks, the
Omori law shows that the process is strongly non-stationary, and the average frequency decades as
∼1/t; for instance, using the values of the coefficients of the Omori law obtained by (Lolli and
Gasperini, 2003) for the Italian earthquakes, after 7 days the activity is reduced to less than the 17% of
that of the first day, while after one month it is a little bit more than 4%. Similar results, but even more
pronounced, are found using the values of Reasemberg and Jones for the southern California.

The dependency of the aftershocks risk from the initial time t can be employed for establishing a date
over which the risk decrease up to a level comparable with the mainshock risk and therefore
considered as acceptable; for this purpose the problem of establishing for what time T the risk has to
be measured, must be overcome. For instance, following (Yeo and Cornell, 2009), T=1 year could be
assumed, in order to make Na directly comparable with the annual frequency of collapses due to
mainshocks. But if a lower value of T is assumed (e.g. 6 months), Na is reduced, at least in the first ten
days, less than 20% (Lolli and Gasparini dates), while the expected mean number of the mainshocks
halves. The choice of T=1 year appears therefore arbitrary and related only to common criterions.

More realistic appears the approach proposed by (Franchin and Pinto, 2009) consisting in using the
maximum duration of the aftershocks sequence expressed as t u = 60 + 60(m m − 4 ) days (Lolli and
Gasparini, 2003). In this way however the duration of the time period T = tu − t varies with t. For a
direct comparison between this quantity and the annual stationary frequency, they have calculate an
"equivalent" annual frequency expressed as: [365 / (tu − t )] N a (ϕi , t , tu − t ) .

In the present paper a more direct approach is proposed, which is based on the comparison between
the average number of exceedance in a short time (for example one day); in this way the effect of the
variation in the time of the frequency of the aftershocks becomes negligible. Therefore, the daily
frequency of the aftershocks can be compared with the analogous frequency of the mainshocks. The
results of the example shown in the following will be then compared and discussed with the results
obtainable using other strategies.

4. APPLICATION

The developed procedure has been applied to a simply supported highway viaduct (Vallone del Duca)
of the A16 Napoli-Canosa highway, composed by two independent roadways, each one composed by
three simply supported reinforced concrete beams with a span of 32 m (Figure 1). The columns range
in total height from 5.5m to 16 m; their cross section is a rectangle of 1.40 m by 2.70 m with two way
steel reinforcement of φ28 mm diameter bars and a transverse reinforcement of φ10 mm diameter
hoops spacing of 10 cm. The pier analyzed in this work is the pile n.1, with a total height of 8.5m. The
material strengths, specified in original design drawings and report, are: concrete compressive strength
fc=45.5 MPa and steel yield strength fyd=370 MPa.
The dynamic analysis have been executed by the non-linear analysis program Opensees (McKenna et
al., 2007). The pier have been modeled as cantilever with the total mass placed at the top. A nonlinear
beam column element has been used including uniaxial bending and axial force. The bending response
has been simulated by the uniaxial hysteretic material implemented in Opensees. This model allows to
account for the degradation of r.c. structures and the residual displacements after the mainshock, for a
reliable assessment of the seismic response of the damaged structure to aftershocks. A typical cyclic
response for a given axial force, here assumed as constant, is given in Figure 2, in which the numerical
model is also shown.

F
F,δ

Figure 1 – Transversal section of the bridge Figure 2 – Model and cyclic behaviour of the pier

1 -1
10 10

-2
10
0
10 φ=0.02
Daily rates of exceeding site PGA

-3
Cumulative Distribution Function

10
φ=0.04
-4 φ=0.044
-1 10
10
-5
10
-2
10 -6
10

-7
1 days from mainshock 10
-3
10 5 days from mainshock
-8
10 days from mainshock 10 fragility aftershock
15 days from mainshock fragility mainshock
-4 -9
10 -1 0
10
10 10 0.4 0.5 0.6 0.7 0.8 0.9 1
pga(g) pga (g)

Figure 3 – Daily rate of exceeding site PGA Figure 4 – Fragility curves

The application refers to a site in the Campania region of South Italy, close to the epicentre of the
1980 Iripina earthquake. In order to evaluate the aftershock hazard of this site, a mainshock of
magnitude M = 7, with epicentre placed at a distance r = 10 km from the site, has been here
considered. The fault length has been calculated as function of the magnitude M (Wells and
Coppersmith, 1994). For the rate of aftershock events (eq. 2.7), the parameters for the considered area
are taken from the study by (Lolli and Gasperini 2003): a = −1.676, b = 0.96, c = 0.0295, p = 0.93 and
m1 = 4.5. A Montecarlo simulation, using 1000 samples, has been used for the integration over the
distance r. In order to compare the aftershock and mainshock risk, the mainshock mean annual
frequency has been evaluated using the most recent description and data for seismogenetic zones for
the Italian territory (Gruppo di Lavoro, 2004) implemented in conventional PSHA program
Mathazard. For the fragility calculation, a set of 10 ground motion records has been selected from the
PEER database, with the parameters: 6.5 ≤ Magnitude ≤ 7.5, Source-Site distance ≤ 25km; the records
have been normalized to the peak ground acceleration and scaled so as to match in mean the uniform
hazard spectrum, as defined by the Italian Code for the bridge localization and the average return
period of 2475 years. The records have been selected so that their average spectrum minimize the
difference respect to the uniform hazard spectrum. Regarding the choice of the aftershock records, in
this work the same set of records used to represent the mainshocks has been used for the aftershocks.
In Figure 3 the daily rate of exceeding site PGA of the aftershocks is illustrated as function of the
elapsed time from the initial rupture t (1, 5, 10, 15 days from mainshock) and for different levels of
PGA. As expected, the initial number of the aftershocks is high, even if it decreases very rapidly with
t. Following the indications of section 2.1, a first cycle of analysis (IDA), representing the mainshock
event, has been carried out to evaluate the fragility curves for the undamaged structure and the PGA
leading to predefined damage levels. A second cycle of analysis (IDA) has been performed to evaluate
the fragility curves of the damaged structure due to the aftershocks leading to collapse. Figure 4 shows
the fragility curves of the undamaged structure due to the mainshock (p0) in comparison with fragility
curves due to aftershocks FD (ϕ i | ϕ j , y ) , for three levels of damage caused by mainshock (ϕ1 =0.02,
ϕ3 =0.04, ϕ3 =0.044). The effect of the initial damage on the probability of failure due to an aftershock
generally increases the probability of failure for each value of PGA.
0
10
1
aftershock: in T
0.9 aftershock: daily
-1
mainshock
10
0.8
Cumulative Distribution Function

Mean annual rates of collapse

0.7
-2
0.6 10

0.5
-3
0.4 10

0.3
-4
0.2 10

0.1 P( φ |M,r)
updated P( φ |M,r)
-5
0 10 0 1 2 3
0 0.01 0.02 0.03 0.04 0.05 10 10 10 10
Drift (f) t (days)

Figure 5 – CDF of the drift for an observed Figure 6 – Mean annual frequency of collapse as
damage θ=θ1 a function of time elapsed from the
mainshock

In Figure 5 the effect of the bayesian updating on the probability function of the damage (drift)
produced by the mainshock is shown; the dashed line is the CDF of the drift due to the seismic event,
evaluated using the attenuation law and the fragility curve of the undamaged structure. The continuous
curve represents the updated function [eq. (2.4)] assuming an observed damage θ1 (spalling) and using
the correlation values between the observed damage and the drift carried out by (Berry, M., Eberhard,
M. 2003), reported in Table 4.1, and assuming a lognormal distribution. A high sensitivity to the
Bayesian updating, considering the limited level of damage here adopted, can be noticed. In particular,
the median value of drift moves from 1.4% to 2% while the corresponding coefficients of variation
reduce from 0.64 to 0.48. This result shows as the bayesian updating can strongly modify the damage
curve, when the damage predicted by the numerical analysis does not correspond to the in-situ one,
although the high dispersion of the parameters of Table 4.1.

Table 4.1. Mean value and correlation coefficient of the drift by Berry et al. 2003
Observed damage θ1 θ1 θ1
Drift mean value 0.02 0.04 0.044
Correlation coefficient 0.43 0.25 0.2

Figure 6 shows the results of the analysis in terms of risk. The continuous line represents the mean
daily rate of collapse, calculated using equation (2.11) assuming T = 1 day, but transformed in annual
rate by multiplication by 365. In the same graph the equivalent mean annual rate of collapse,
computed using the variable observation time T = tu-t, as suggested by (Franchin and Pinto 2009), is
also shown. In this equation tu is the time duration of the aftershocks (Lolli and Gasperini, 2003).
Immediately after the mainshock the difference of the mean annual rate of collapse predicted by the
two curves is quite high but become coincident when t = tu −1 day. The comparison with the mean
annual rate of collapse due to the mainshock (the dotted horizontal line) shows that the risk structure
change dramatically when a aftershock sequence strike the bridge. Only after the duration tu the risk is
reduced of 1/100, even if after 1000 days (3 years) the difference between aftershock and mainshock
risk remains considerable. The analyzed bridge is not particularly vulnerable being very low the mean
annual frequency of collapse due to the mainshock. Nevertheless, the effect of an aftershock sequence
can be still appreciable.
5. CONCLUSIONS

A critical issue in the emergency management after the earthquake is the functionality of the main
infrastructure (hospitals, road network, etc.) and the decision on their usability just after the main
shock. At present, a decision is take on the basis of a structure in-situ inspection; an analytical
assessment is in contrast with the lack of time and data. For this reason in this paper a method that
rationally combines information from the analytical approach and the in situ inspection is proposed. In
particular an effective tool to speed-up the decision making phase concerning the evaluation of the
seismic risk of mainshock-damaged structures due to aftershocks has been proposed. The risk is
calculated combining the aftershock hazard using the Omori law and the fragility curves of the
structure calculated using IDA technique and updated using in-situ inspection data. The procedure has
been applied to a highway r.c. bridge. The results have highlighted a high sensitivity to the Bayesian
updating especially when the damage predicted by the numerical analysis does not correspond to the
real damage. The mean annual rates of collapse provided by the method has shown that the risk
structure change dramatically when an aftershock sequence strike the bridge and that this risk
decreases with time allowing the authorities to decide if and when re-open the bridge to traffic.

AKCNOWLEDGEMENT
The work presented in this paper has been partially funded by a grant of Italian Ministry of University and
Research (MIUR) in the scope of Research Programs of National Interest (PRIN 2007).

REFERENCES

Ambraseys, N. N. , Simpson, K. A., Bommer, J. J. (1996). Prediction of horizontal response spectra in Europe.
Earthquake Engineering and Structural Dynamics, 25, 371 -400.
Ambraseys, N. N., Douglas, J., Sarma, S. K., Smit, P. M. (2005). Equations for the estimation of strong ground
motions from shallow crustal earthquakes using data from Europe and the Middle East: horizontal peak
ground acceleration and spectral acceleration. Bulletin of Earthquake Engineering 3, 1–53.
Berry, M., Eberhard, M. (2003). Performance models for flexural damage in reinforced concrete columns. PEER
Report 2003/18.
Franchin, P., Pinto, P. (2009). Allowing traffic over mainshock-damaged bridges. Journal of Earthquake
Engineering, 13, 585–599.
Gruppo di Lavoro (2004) Redazione della mappa di pericolosità sismica prevista dall’OPCM 3274. Rapporto
conclusivo per il DPC, INGV, Milano-Roma.
Kumitani, S., Takada. T. (2008). Probabilistic assessment of buildings damage considering aftershocks of
earthquakes. 14th World Conference on Earthquake Engineering. Beijing, China.
Lolli, B., Gasperini, P. (2003). Aftershocks hazard in Italy Part I: Estimation of time-magnitude distribution
model parameters and computation of probabilities of occurrence. Journal of Seismology 7, 235–257.
Mackie, K. R., Stojadinovic B. (2006). Post-earthquake functionality of highway overpass bridges. Earthquake
Engng Struct. Dyn.; 35,77–93.
Maffei, J., Telleen, K., Nakayamab, Y. (2008). Probability-based seismic assessment of buildings, considering
post-earthquake safety. Earthquake Spectra, 24:3, 667–699.
McKenna, F., Fenves, G. L., and Scott, M. H. (2007) OpenSees: open system for earthquake engineering
simulation. PEER, University of California, Berkeley, CA.
Reasenberg, P.A., Jones, L. M. (1989). Earthquake hazard after a mainshock in California. Science 243, 1173-
1176 .
Sabetta, F., Pugliese, S. (1987). Attenuation of peak horizontal acceleration and velocity from italian strong-
motion records. Bul. of the Seism. Soc. of Am. 77:5, 1491-1513.
Utsu T. (1995) The centenary of the Omori formula for a decay law of aftershock activity. Journal of the Physics
of the Earth 43, 1–33.
Vamvatsikos, D., Cornell C. A. (2002). Incremental dynamic analysis. Earthquake Eng. Struct. Dyn. 31, 491–
514.
Wells, D. L., Coppersmith, K. J. (1994). New empirical relationships among magnitude, rupture length, rupture
width, rupture area, and surface displacement. Bul. of the Seism. Soc. of Am. 84:4, 974-1002.
Yeo, G. L., Cornell, C. A. (2009). A probabilistic framework for quantification of aftershock ground-motion
hazard in California: Methodology and parametric study. Earthquake Engng Struct. Dyn. 2009; 38, 45–60.
Yeo,G. L., Cornell, C. A. (2005). Stochastic characterization and decision bases under time-dependent
aftershock risk in performance-based earthquake engineering. PEER Report 2005/13.
Zhong, J., Gardoni, P., Rosowsky, D. (2009). Bayesian updating of seismic demand models and fragility
estimates for reinforced concrete bridges with two-column bents. Journal of Earthquake Eng. 13, 716–735.