You are on page 1of 8

Method of Developing Fragility Curves - a Case Study for Seismic

Assessment of Masonry Buildings in IRAN


A. Bakhshi. Assisstant Professor. bakhshi(shariI.edu
Department oI Civil Engineering. ShariI University oI Technology. Tehran. Iran.
and
K. Karimi. Graduate Student. kkarimi(mehr.shariI.edu
Department oI Civil Engineering. ShariI University oI Technology. Tehran. Iran.
Abstract
In this paper. fragilitv curves and the method of developing them are discussed. Then.
this method of seismic assessment of buildings is applied to evaluate seismic
vulnerabilitv of unreinforced masonrv buildings. the tvpe of buildings which are so
common in IRAN. Fragilitv curves show the probabilitv of exceeding a certain damage
state versus a seismic intensitv parameter. In pervious woks done in this field. usuallv
PGA (Peak Ground Acceleration) is chosen to show the intensitv of earthquake while.
this parameter is not appropriatelv related to damage of structures. Therefore. in this
research. CAJ (Cumulative Absolute Jelocitv) is used instead of PGA in obtaining such
curves. The uncertaintv in structural parameters is considered in compressive and
tensile strength of masonrv. compressive strength of concrete and tensile strength of
reinforcement. Monte Carlo simulation method is applied to produce random
parameters based on their assigned probabilitv distribution functions. Different models
are analvzed with IDARC 4.0 nonlinear analvsis program and fragilitv curves are
obtained based on the average and standard deviation of results. Bv comparing these
curves before and after retrofitting of buildings. the effect of vertical and horizontal ties
and reinforcements on the seismic performance of masonrv buildings is shown.
Keywords: Fragilitv curves. Monte Carlo simulation. Damage index. Unreinforced
masonrv buildings
1- Introduction
Some destructive earthquakes occur periodically in diIIerent countries in the world.
This leads to damage oI many vulnerable buildings and loss oI lives each year.
ThereIore. evaluation oI seismic vulnerability oI buildings beIore the occurrence oI an
earthquake is an essential step in preventing Iinancial costs and loss oI lives due to
earthquakes. Developing Iragility curves is a helpIul tool to accomplish this goal. It is a
probabilistic method to estimate the probability that a building will exceed a certain
state oI damage Ior the speciIic amount oI seismic intensity. Seismic intensity parameter
can be PGA. MMI. I
a
. CAV and so on. In previous works done. usually PGA (Peak
Ground Acceleration) is used as the seismic intensity parameter while. in this paper.
CAV (Cumulative Absolute Velocity) has been chosen Ior this purpose. because it is
Iound to be well associated with structural damage |1|.
Fragility curves are either empirical oI analytical. Empirical Iragility curves are
developed using real damage data obtained Irom previously occurred earthquakes. while
analytical Iragility curves are constructed based on the nonlinear dynamic analysis.
Many researchers have developed Iragility curves Ior some types oI building such
as bridges. reinIorced concrete structures. and high-rise buildings |2. 3. 4. 5. 6 and 7|.
but less work is done Ior masonry buildings specially the traditional type which is
common in Iran. Hence. this method is applied to estimate seismic vulnerability oI
these buildings and evaluate the eIIect oI vertical and horizontal ties and reinIorcements
in improving seismic perIormance oI these buildings.
2- Damage index
Various damage indices have been proposed by diIIerent researchers. They can be
categorized to local and global damage indices |8|. Local damage indices are either
cumulative or non-cumulative. Some local damage indices are based on deIormation.
energy or both. Global damage indices combine local ones to obtain a weighted average
oI them or they use the change in modal parameters such as the natural period oI the
structure beIore and aIter the earthquake to estimate the global damage to the building.
Two oI the most recent modiIied damage indices are proposed by Bozorgnia and
Bertero (2001) |9.10|. They are deIined Ior a general inelastic system with one degree
oI Ireedom and consist oI both deIormation and energy eIIects:
( )( ) ( ) ( )
1 1 1 .
1 / 1 /
e mon H H mon
DI E E o u u u o = +

(1)
( )( ) ( ) ( )
1
2
2 2 2 .
1 / 1 /
e mon H H mon
DI E E o u u u o = +

(2)
where:
max
/
v
u u u = is deIormation-based ductility. /
e elastic v
u u u = is the maximum potion oI
elastic deIormation divided by
v
u . 1
e
u = Ior inelastic behavior and 1
e
u = . iI the
behavior oI the structure remains elastic ( 1 u s ).
mon
u is the monotonic deIormation
capacity oI the structure.
H
E is the demanded hysteretic energy by earthquake.
. H mon
E
is the hysteretic energy capacity oI the structure under monotonic lateral deIormations
and
1
o and
2
o are constants (
1
0 1 o s s .
2
0 1 o s s ).
In this paper the proposed damage index in IDARC 4.0 is used in developing
Iragility curves. It is composed oI three diIIerent models oI damage:
2.1- Park and Ang model of damage
This model was Iirst suggested by Park and Ang in 1984 and then modiIied by
Kunnath et a. in 1992 as Iollows:
m r
h
u r v u
DI E
M
u u |
u u u

= +

(3)
Where.
m
u is the maximum rotation obtained during loading.
u
u is the ultimate
rotational capacity oI the section.
r
u is the returnable rotation aIter unloading.
v
M is
the yielding moment and
k
E is the energy absorbed by the section.
The story damage and the global damage oI the building are calculated using
weighted coeIIicients based on the hysteretic energy in the members and story levels:
( ) ( ) ( )
( ) ( ) ( )
;
;
i
storv i i i
component component component
i
component
i
overall i i i
storv storv
storv
i
storv
E
DI DI
E
E
DI DI
E


| |
= =
|
|
\ .
| |
= =
|
|
\ .


(4)
i
is the weighted coeIIicient based on hysteretic energy and
i
E is the total energy
absorbed by the story or member i.
2.2- Fatigue based damage model
This model was proposed by Reinhorn and Valles in 1995. It is based on the
principal response parameters oI the structure and low-cycle Iatigue law. This damage
index is:
( )
1
1
4
a v
u v
h
u v v
DI
E
F
o o
o o
o o

=
| |
|
|

\ .
(5)
a
o is the maximum obtained deIormation. rotation or curvature.
v
o is the yielding
deIormation capacity.
u
o is the ultimate deIormation capacity.
v
F is the capacity oI
yielding Iorce and
k
E is the cumulative absorbed hysteretic energy.
2.3- Global damage model
The soItening damage index proposed by DiPasquale and Cakmak in 1988 is used
in IDARC:
( )
( )
0
0
1
initial
equivalent
T
DI
T
= (6)
This damage index is based on the variation oI the Iundamental period oI the
structure during earthquake and due to development oI damage.
3- Damage states
Damage indices are deIined diIIerently in various texts. The deIinition and
classiIication oI them varies based on the type oI structure and damage index used. Park
and Ang |12| calibrated their proposed damage index with data obtained Ior damage to
9 reinIorced concrete structures in several past earthquakes. Based on their work. they
determined the limits oI damage index according to each damage index. It is shown in
Table 1 and we have also used this classiIication Ior damage states in this paper.
Table 1: Damage stated based on Park & Ang damage index
4- Fragility curves
To develop Iragility curves Ior a speciIic type oI structures Iirst. appropriate values
Ior uncertain structural parameters are produced by Monte Carlo simulation method
based on their assigned probability distribution Iunctions. Then diIIerent structural
models Ior variant values oI these parameters are analyzed under some appropriate
earthquake records. These records should be scaled according to each value oI seismic
intensity parameter in plotting Iragility curves. The values oI damage indices in each
model and Ior a speciIic value oI seismic intensity are obtained through nonlinear
dynamic analysis.
As previously mentioned. cumulative absolute velocity is used to represent the
seismic intensity oI earthquake since PGA does not estimate the intensity oI earthquake
appropriately. This parameter is deIined as:

=
T
dt t a CAJ
0
) ( (7)
ii
PF . the probability oI exceeding the i th damage state Ior an earthquake with CAV
equal to
i
v . is calculated Irom the Iollowing equation and the values oI damage index
Irom analysis:
( )
( ) ( )
1 1
ii i i
i i DT i i
PF Prob DT DT CAJ v
Prob DT DT CAJ v F DT CAJ v
= > =
= < = = =
(8)
where:
i
DT is the damage index relevant to the i th damage state and ( )
DT
F is the
probability distribution Iunction oI DT . Fragility curve Ior the i th damage state can be
plotted using
ii
PF values Ior each level oI CAV.
4.1- Damage probability matrix
Damage probability matrix shows the probability oI occurrence oI damage due to
earthquake in each damage state.
ii
PDS . the probability that the damage to the structure
due to an earthquake with
i
CAJ v = be located in the i th damage state. is:
( )
( )
1
4
5
ii i i
ii
ii
PF PF i
PDS
PF i
+
s

=

=

(9)
4.2- Seismic damage cost
By seismic damage cost. we mean the direct cost oI repair oI the building aIter the
earthquake. It does not include the cost to the contents oI the building and the indirect
cost due to loss oI building Iunction. To estimate damage cost. the central damage cost
ratio (
i
CDR ) corresponding to the i th damage state is deIined as the ratio oI average
repair cost oI the building in the i th damage state to the cost oI replacing the building.
The mean damage cost ratio ( i DR ) due to an earthquake with
i
CAJ v = is
calculated as Iollows:
5
1
i
ii i
i
DR PDS CDR
=
=

(10)
Then the damage cost (
i
DC ) assuming the occurrence oI an earthquake with
i
CAJ v = would be:
i
i
DC DR RPC = (11)
RPC is the cost oI replacing the building. Assuming the occurrence oI an
earthquake. the expected annual earthquake loss is calculated as Iollows:
1
.
2 2
N
a
i J i J i
i
v v
AEL DC F v F v
=
A A | | | |
= +
| |
\ . \ .

(12)
Where. ( )
J
F is the probability distribution Iunction oI cumulative absolute
velocity ( CAJ ) in one year.
5- Applying the method to an example of URM
1
In this section. we are going to apply the method oI developing Iragility curves to a
common type oI buildings in Iran. the unreinIorced masonry one. These buildings are
made oI masonry units and mortar. They are usually constructed without ties or any
earthquake resistant systems and thereIore. so vulnerable in earthquakes. Fragility
curves and damage probability matrices are developed Ior a sample one-story masonry
building beIore and aIter using horizontal and vertical ties and reinIorcing. The building
is 5m5m in plan 3m high. The thickness oI masonry walls is assumed 0.3m.
Horizontal and vertical ties are provided according to the Iranian Seismic Code |13|.
AIter designing masonry walls to resist seismic loading. the required percentage oI
horizontal and vertical reinIorcement in each wall turns into 0.3. The buildings are
assumed to be located on Iirm soil and 12 earthquake records with diIIerent
characteristic are used in devolving the Iragility curves Ior these buildings.
5.1- Uncertainty in structural parameters
In developing Iragility curves. uncertainty is considered in some structural
parameters. The uncertain parameters considered in this paper with their assumed mean
value. probability distribution Iunction and coeIIicient oI variation is shown in Table 2.
The properties oI masonry material in this table are Ior the common type in Iran.
1
UnreinIorced Masonry Buildings
Table 2: Uncertaintv in material properties with their average value.
tvpe of distribution function and coefficient of variation
5.2- Monte Carlo simulation method
Monte Carlo simulation method is used to generate random values corresponding
to each structural parameter in Table 2. based on their probability distribution Iunctions
|14|.
II
1
U and
2
U are two independent standard uniIorm variables. the Iollowing
Iunctions constitute a pair oI independent normal variables with mean value oI u and
standard deviation oI o :
( )
( )
1 1 2
2 1 2
2ln cos 2
2ln sin 2
x U U
x U U
u o t
u o t
= +
= +
For the log-normal random variable X with parameters and . . ln X would be a
normal random variable with the mean value equal to and standard deviation equal to
. . Hence. iI x' has a normal distribution as ( ) . N . .
x
x e
'
= would be a random
variable having log-normal distribution with corresponding parameters and. .
5.3- Structural Modeling
IDARC 4.0 nonlinear analysis program is used to model masonry buildings in this
study |11|. This program can model and analyze steel and concrete structures and also
inIill masonry panels. It should be mentioned that the behavior oI inIill masonry panels
are completely diIIerent Iorm masonry walls without Irames. Since masonry units and
concrete behaves similarly (both are strong in compression and week in tension). the
idea developed to model unreinIorced masonry walls as shear walls with a negligible
amount oI reinIorcement in IDARC program. It is done by setting some parameters in
masonry stress-strain diagram equal to corresponding parameters in that oI concrete.
5.4- The resulted fragility curves
15 diIIerent values are generated Ior each uncertain structural parameter using
Monte Carlo simulation method. Structural model are made based on the values oI these
parameters and analyzed under earthquake records. By applying the method mentioned
in section 4. Iragility curves are developed Ior the sample URM buildings beIore and
aIter using ties and retroIitting by reinIorcements. These curves and their corresponding
probability damage matrices are shown in Figures (1). (2) and (3).
Figure 1: Fragilitv curves and damage probabilitv matrix for a one-storv URM building
without ties
Figure 2: Fragilitv curves and damage probabilitv matrix for a one-storv URM building with
horizontal and vertical ties
Figure 3: Fragilitv curves and damage probabilitv matrix for a one-storv Reinforced Masonrv
Building
6- Conclusion
Developing Iragility curves is a helpIul method to estimate seismic vulnerability oI
structures. These diagrams give a probability basis to predict damage to structures in
earthquakes with diIIerent intensities. Cumulative absolute velocity is an appropriate
parameter to show seismic intensity oI ground motion. By comparing Iragility curves in
Figures (1). (2) and (3). the eIIect oI ties and reinIorcements in improving seismic
behavior oI URM buildings can be clearly observed. Using horizontal and vertical ties
empowers seismic resistance oI these buildings and decreases the probability oI
exceeding a speciIic damage state Ior each value oI cumulative absolute velocity
(CAV). Rehabilitating the building using horizontal and vertical reinIorcements
improves its seismic perIormance apparently and signiIicantly decreases exceeding
probabilities. as it is seen in Figure (3).
7- References
|1| Bakhshi. A.. Tavallali. H.. and Karimi. K.. 'Structural Damage Estimation Using
Seismic and Spectral Parameters. Proceedings of the Tenth International Conference
on Civil. Structural and Environmental Engineering Computing. Rome. Italy.
September-August 2005.
|2| Shinozuka. M.. et al.. 'Statistical analysis oI Iragility curves. Journal of
Engineering Mechanics. Vol. 126. No. 12. 2000. pp. 1224-1231.
|3| Mackie. K.. and Stoiadinovic. B.. 'Fragility Curves Ior ReinIorced Concrete
Highway Overpass Bridges. Proceedings of the 13
th
World Conference on Earthquake
Engineering. Vancouver. B.C.. Canada. 2004.
|4| Dumova-Jovanoska. E.. 'Fragility curves Ior reinIorced concrete structures in
Skopie (Macedonia) region. Journal of Soil Dvnamics and Earthquake Engineering.
Vol. 19. 2000. pp. 455-466.
|5| Singhal. A.. and Kiremidiian. A.S.. 'A method Ior earthquake motion-damage
relationship with application to reinIorced concrete Irames. Technical Report NCEER-
97-0008. State University oI New York at BuIIalo. 1997.
|6| Abrams. D.P.. and Shinozuka. M.. 'Loss assessment oI Memphis Buildings.
Technical Report NCEER-97-0018. State University oI New York at BuIIalo. 1997.
|7| Tantala. M.W.. and Deodatis. G.. 'Development oI seismic Iragility curves Ior tall
buildings. Proceedings of the 15
th
ASCE Engineering Mechanics Conference.
Publishing Company. New York. NY. 2002.
|8| Williams. M.S.. and Sexsmith. R.G. "Seismic damage indices Ior concrete
structures: a state-oI-the-art review". Journal of Earthquake Spectra. Vol. 11. No. 2.
May 1995. pp. 319-349.
|9| Bozorgnia. Y.. and Bertero. V.V. "Evaluation oI Damage Potential oI Recorded
Earthquake Ground Motion". 96
th
Annual Meeting of Seismological Societv of America.
Seismol. Res. Lett.. 2001. 74. 312.
|10| Bozorgnia. Y.. and Bertero. V.V. "Improved Shaking and Damage Parameters Ior
Post-Earthquake Applications". Proceedings of the SMIP01 Seminar on Utilization of
Strong-Motion Data. Los Angeles. 2001. pp. 1-22.
|11| Valles. R.E.. Reinhorn. A.M.. Kunnath. S.K.. Li. C.. and Madan. A. "IDARC 2D
version 4.0: a program Ior inelastic damage analysis oI RC structures". Technical
Report NCEER-96-0010. State University oI New York at BuIIalo. NY. January 1996.
|12| Park. YJ.. Ang. AH-S. and Wenn. YK. "Seismic damage analysis oI reinIorced
concrete buildings". Journal of Structural Engineering. ASCE. Vol. 111. No. ST4.
1985. pp. 740-757.
|13| 'Iranian code of practice for seismic resistant design of buildings. Standard No.
2800. 2
nd
edition-1999.
|14| Rubinstein. R.Y.. "Simulation and the Monte Carlo Method". John Wiley & Sons.
1989. New York.

You might also like