Você está na página 1de 10

Vienna Congress on Recent Advances in

Earthquake Engineering and Structural Dynamics 2013 (VEESD 2013)


C. Adam, R. Heuer, W. Lenhardt & C. Schranz (eds)
28-30 August 2013, Vienna, Austria
Paper No. 311

An analytical procedure for the derivation of state-dependent fragility


curves for masonry buildings
A. Penna1,2, M. Rota2, A. Mouyiannou3, F. Graziotti3, G. Magenes1,2
1

Department of Civil Engineering and Architecture, University of Pavia, Via Ferrata 3, 27100 Pavia, Italy
European Centre for Training and Research in Earthquake Engineering, Via Ferrata 1, 27100 Pavia, Italy
3
ROSE program, UME School - IUSS Pavia, via Ferrata 1, 27100 Pavia, Italy
2

Abstract. The seismic capacity of a structure is a function of the characteristics of the system and of its state,
which is mainly affected by previous damage and deterioration. The effect of repeated shocks (for example
during a seismic sequence or due to multiple events affecting an unrepaired building stock) on the seismic
vulnerability of masonry buildings is considered in this work by developing state-dependent fragility curves.
These curves only take into account cumulated seismic damage, neglecting possible ageing effects. An analytical
procedure for the derivation of fragility curves is proposed, which is based on nonlinear dynamic analyses with a
large set of unscaled natural records. To speed up computational time, analyses are carried out using an
equivalent single-degree-of-freedom system, properly calibrated to reproduce the cyclic dynamic response of the
selected building prototype.
Keywords: Fragility curves; Masonry; State-dependent; Equivalent SDOF system

1 INTRODUCTION
The seismic safety of a system can be measured by its capacity to withstand a certain intensity of the
seismic action. Hence, the seismic capacity of the structure can be expressed in terms of intensity
measures, the latter being expressed either by scalar or vector parameters.
Since seismic safety has to be evaluated for different limit states, the seismic capacity can be
expressed by a set of values of the intensity measures corresponding to the levels of seismic action
inducing the attainment of the considered limit states, i.e. the attainment of given thresholds for
selected engineering demand parameters. The seismic capacity is a function of the characteristics of
the system and of its state, which is mainly affected by damage and deterioration.
Damage state is considered as the cumulated effect of previous seismic events. It mainly depends on
the seismic history of the site since the system was established (construction time in case of structures)
and on the seismic vulnerability. For sake of simplicity the effect of possible damages other than those
caused by seismic event are not considered in this work as well as ageing effects are not considered.
If repair, retrofitting or maintenance interventions are not considered, the seismic safety is assumed to
monotonically decrease with time. The damage accumulation due to seismic events is by nature a
discontinuous process. In general, the effect of the event intensity on the damage evolution and hence
on the reduction of the residual capacity is highly nonlinear and only events whose intensity at the site
exceed a certain threshold level affect damage accumulation and hence the state of the structures. Such
an intensity threshold can be also regarded as a function of the previous damage state.
Based on the aforementioned considerations, the problem of defining time-dependent vulnerability of
systems can be split into two separate sub-problems:
1. Defining the system state as a function of time, e.g. considering the combined effect of ageing
and cumulated seismic damage;
2. Defining state-dependent fragility curves representing the seismic vulnerability of the system
given the system state (at a given time).

A. Penna, M. Rota, A. Mouyiannou, F.Graziotti, G.Magenes / VEESD 2013

This work concentrates on the latter aspect and in particular on the definition of state-dependent
fragility curves for masonry buildings. The first aspect is explicitly considered in other works (e.g.
Iervolino et al., 2012), while the second has been considered in some works for reinforced concrete
structures (e.g Luco et al.,2004, Rveillre et al.,2012)
In order to derive state-dependent fragility functions for systems with a given (cumulated) seismic
damage, the following steps should be carried out:
1. Selection of seismic input for time-history analysis;
2. Selection of representative models for each system typology (e.g. modelling strategies,
representative building prototypes);
3. Experimental basis for mechanical properties;
4. Selection of relevant engineering demand parameters (EDPs);
5. Identification of relevant Limit States and corresponding EDP thresholds (based on
experimental results);
6. Analytical derivation of fragility functions starting from a predefined state of damage (EDP
values).
2 PROCEDURE FOR DERIVING STATE-DEPENDENT FRAGILITY CURVES
The state dependent fragility curves are derived based on the results of time history analysis with a
large number of unscaled records obtained from a selected database (e.g. Smerzini and Paolucci,
2011). In order to examine the effect of pre-existing damage in the building, models with different
levels of damage should be analysed. The procedure scheme is presented in Figure 1.
Earthquake record selection
from database

MODEL DEFINITION

Calibration of MDOF model


according to experimental
data
SDOF derivation

MDOF

SDOF

Damage due to previous


seismic events
Cyclic pushover until
various levels of
displacement

Representative intensity
measure
Modified Housner Intensitiy
PGA

9
SDOF
models

Definition of limit states

OLS operational
DLS damage limitation
ULS/NC ultimate/near collapse

Thresholds expressed in terms


significant displacements attained from
pushover analyses

Time history
analyses with the
selected unscaled
records

DERIVATION OF
FRAGILITY CURVES
Discretization of the intensity
measure in intervals with
adequate lengths
Derivation of DPMs
Lognormal fitting

Figure 1. Scheme of the procedure for the derivation of state-dependent fragility curves

A. Penna, M. Rota, A. Mouyiannou, F.Graziotti, G.Magenes / VEESD 2013

The procedure starts with the model definition. A multiple degree of freedom (MDOF) model can be
created (calibrated according to experimental data) and a single degree of freedom (SDOF) model can
be derived with respect to the results of analyses with the MDOF mode. The MDOF model should be
used for the identification of limit states. The limit state thresholds are expressed in terms of
significant displacements of the capacity curve, resulting from pushover analysis. In the present study,
three limit states are considered, namely: operational limit state (OLS), damage limitation limit state
(DLS) and ultimate limit state/near collapse (ULS). If the derivation of an adequate representative
SDOF model for the building considered is possible, then the SDOF model can be used in the rest of
the procedure allowing the performance of time history analyses with more earthquake records in a
significantly smaller amount of time.
The derivation of models with different levels of pre-existing damage is realised with subjecting the
SDOF model to cyclic analyses up to different predefined levels of displacement. The levels of
displacement can be expressed as a percentage of the displacement threshold corresponding to ULS,
for brevity reported as damage percentage of ULS.
Finally in order to perform nonlinear dynamic analyses the time histories should be selected. The
adequate number of the earthquake records as well as the representative intensity measure for the
fragility curves should be decided. In this study the majority of records of SIMBAD database
(Smerzini and Paolucci, 2011) are selected and two intensity measures are considered, namely: the
peak ground acceleration and the Housner intensity modified for masonry structures.
Once the time history analyses are performed with all the models characterized by different levels of
pre-existing damage, the intensity measures are discretized in an appropriate number of ranges, the
cumulative damage distributions are calculated and the fragility curves are lognormally fitted to the
distributions.
3 APPLICATION EXAMPLE
3.1 Building prototype: modelling, mechanical parameters, calibration of SDOF model
The procedure can be fulfilled either with the use of a MDOF model or a SDOF. In this example, a
MDOF model of the building is used for the derivation of the considered limit states in terms of
displacement thresholds resulting nonlinear static analyses, while a SDOF model is used for the
derivation of the various levels of the pre-shock damage and the nonlinear dynamic analyses with a
selection of natural accelerograms. The MDOF as shown in Figure 2, was calibrated in accordance to
the results obtained by the experimental campaign with a similar building model. The building model,
shown in Figure 3, was subjected to a cyclic quasi-static test at the University of Pavia in 1994
(Magenes et al., 1995).

Figure 2. MDOF model: Plan view and 3-D model.

A. Penna, M. Rota, A. Mouyiannou, F.Graziotti, G.Magenes / VEESD 2013

Figure 3. Plan view and side views of the experimental specimen (Magenes et al., 1995).

The results of the nonlinear analyses with the MDOF model were also used for the calibration of an
equivalent SDOF model. The equivalence of the SDOF was validated through dynamic analyses with
natural accelerograms (Graziotti et al.,2013). The calibrated SDOF model shown in Figure 4 is able to
interpret the seismic response of the MDOF model, in a synthetic way representing the entire building.
Axial rigid link
F1
Fh

F2

Figure 4. Simplified SDOF model used for the nonlinear static and dynamic analyses (Graziotti et al.,2013).

3.2 Identification of Limit States


As mentioned before the displacement thresholds corresponding to the attainment of the significant
limit states considered in the fragility curve derivation were evaluated from nonlinear static analyses
with the complete MDOF model. From the resulting capacity curve and its bilinear idealization
according to C7.3.4.1 (NTC2008, c. 617/09) as shown in Figure 5 (see also Mouyiannou et al., 2013,
and Penna et al.,2013b), the following thresholds have been considered:
the end of the bilinear elastic range at dMy = 0.23 cm, to represent the threshold of OLS:
operational limit state
the attainment of maximum base shear at dVmax = 0.38 cm, to represent the threshold of DLS:
damage limitation limit state
the degradation of base shear to 80% of its pick value at dMU = 1 cm, to represent the threshold of
ULS: ultimate limit state.
the attainment of displacement 20% greater than dMU, dMC=1.2 cm to represent the threshold of
COL: near collapse limit state.

A. Penna, M. Rota, A. Mouyiannou, F.Graziotti, G.Magenes / VEESD 2013

700

FMax
600

Base shear [kN]

500

FMy
70%FMax

400
300
200
100

dMU

dMy

0
0

0.2

0.4

0.6

0.8

1.2

Top displacement [cm]

Figure 5. Identification of significant displacement thresholds from the pushover curve and its bilinear
idealization.

3.3 Damage representation: 8 levels of pre-existing damage


In order to study the impact of the level of pre-existing damage in the seismic performance of the
masonry, 8 different levels of damage were created. Each level can be described by one level of
displacement. In order to obtain the damaged model, the undamaged model was subjected to a cyclic
pushover analysis up to the displacement level characterizing the damage level. The levels of damage
can be expressed as percentages of the near collapse limit state displacement. Since 1.2 cm was
assumed to represent the threshold of COL, the damaged model after imposed to cyclic pushover with
maximum displacement for example 0.24cm is called DAM20%COL. Figure 6 includes the derived
cyclic pushover curves for all the damage levels. The undamaged structure is referred as NODAM:
zero initial damage. In addition to the undamaged structure analyses with the models with damage
levels 20,30,40,50,60,70,80 and 90% of COL were performed.
800
600

Base Shear [kN]

400
200
20%COL
30%COL
40%COL
50%COL
60%COL
70%COL
80%COL
90%COL

0
-200
-400
-600
-800
-1000
-1

-0.5

0
Displacement [cm]

0.5

Figure 6. Generation of various levels of pre-shock damage with pushover analyses

3.4 Selection of earthquake records for the nonlinear dynamic analyses


The earthquake records used for the nonlinear dynamic analyses were selected from the SIMBAD
database. The database includes 467 records of 3 components of acceleration. All the earthquakes
have magnitudes in the range of Mw 5 and 7.3 and epicentral distances Repi not greater than 30 km.
The records would be used for time history analyses without scaling with respect to PGA. When IDA
analysis is performed PGA can be adopted as a possible representative intensity measure for the

A. Penna, M. Rota, A. Mouyiannou, F.Graziotti, G.Magenes / VEESD 2013

fragility curve derivation. This consideration is reasonable since the same records are scaled to
increasing levels of PGA so higher PGA values correspond to stronger earthquakes and consequently
higher levels of structural damage. When unscaled records are used there is no immediate relation
between the PGA of the earthquake record to its structural response. Hence the maximum acceleration
of an earthquake record cannot capture the damage induced by the total earthquake. When a large
number of records is used is more likely to obtain higher levels of structural damage for higher levels
of PGA, but is not an assumption that can be valid for each single earthquake record.
For this study two intensity measures were selected for the derivation of state dependent fragility
curves. The first being Housner Intensity modified for masonry structures and the second PGA.
Housner Intensity is an energy based intensity measure, calculated for each earthquake record by the
integral of the pseudo-velocity spectrum for periods from 0.1 to 2.5s (Housner, 1959). The integration
limits were modified in accordance to the representative range of periods for masonry structures which
considers periods from 0.1 to 0.5s. It has to be mentioned that the pseudo-velocity spectrum was
evaluated for 5% damping.
In total 520 earthquake records were used for the dynamic analyses with the SDOF model. Records of
both EW and NS direction were selected from the database, 263 and 257 respectively. Figure 7 shows
the level of Modified Housner Intensity plotted against the number of the selected records and the
PGA of the selected earthquakes (example for the NS direction). The different ranges of Housner
Intensity are shown with different colours. As it can be observed, the two intensity measures are not
proportionally related. A high value of PGA may correspond to a low value of Housner Intensity and
vice versa.
0.7

25
IH=0-4cm
IH=4-8cm
IH=8-12cm
IH=12-16cm
IH=16-20cm
IH=20-24cm

Peak Ground Acceleration [g]

Housner Intensity Is [cm]

20

IH=0-4cm
IH=4-8cm
IH=8-12cm
IH=12-16cm
IH=16-20cm
IH=20-24cm

0.6

15

10

0.5

0.4

0.3

0.2

0.1

50

100
150
Earthquake Record Number

a)

200

250

50

100
150
Earthquake record number

200

250

b)

Figure 7. Housner intensities and peak ground accelerations of the selected earthquake records (NS direction of
records).

It has to be noted that all the available earthquake records were used for Housner intensity values
greater than 4 cm. A significant portion of records in the database correspond to events with modified
Housner intensities not greater than 4cm. After analyses with records of that level of Housner intensity
(0-4cm) it is shown that no response exceeds the operational limit state threshold and so only a limited
number of records in that range was used for the analyses for fragility curve derivation.
It has also to be mentioned that only the uncertainty in the seismic input was considered in this
application example. As it has been shown in the several cases the record to record variability is one of
the dominant sources of uncertainty (e.g. Rota et al., 2010).
3.5 Time history analyses
Dynamic analyses were performed on the 9 single degree of freedom models with the 520 earthquake
records using the TREMURI program (Penna et al., 2013a, Lagomarsino et al.,2013).

A. Penna, M. Rota, A. Mouyiannou, F.Graziotti, G.Magenes / VEESD 2013

The maximum top displacement of the SDOF model was selected as the representative damage
indicator for the fragility curve derivation. The maximum obtained displacement is compared to the
displacement thresholds assigned to the previously described limit states (section).
As shown in the following figures, the displacement demand can be described as a function of the
intensity measure. In Figure 8, the displacement demand is plotted against modified Housner Intensity
(subplots a and c) and PGA (subplots b and d), for the structure with no pre-existing damage (above)
and for level of pre-existing damage 60% of ULS (below). The displacement thresholds corresponding
to the considered limit states are also included in the figures.

1.5

Disp.demand NODAM
OLS
DLS
ULS

Displacement Demand [cm]

Displacement Demand [cm]

1
0.5
0
0

5
10
15
Modified Housner Intensity [cm]

1.5
1
0.5
0
0

20

Disp.demand NODAM
OLS
DLS
ULS

0.1

a)

1.5

Disp.demand 60%
OLS
DLS
ULS

1
0.5
0
0

5
10
15
Modified Housner Intensity [cm]

c)

0.3

0.4

0.3

0.4

b)

Displacement Demand [cm]

Displacement Demand [cm]

0.2
PGA [g]

20

1.5

Disp.demand 60%
OLS
DLS
ULS

1
0.5
0
0

0.1

0.2
PGA [g]

d)

Figure 8. Displacement demand Vs Intensity measures: a) and b) for undamaged structure as a function of
modified Housner intensity and PGA, respectively; c) and d) for an already damaged structure (60% of
ULS).

It can be noticed from the plots that the scatter obtained in the displacement demands is greater when
PGA is selected. The demands for Modified Housner Intensity are more concentrated around some
mean values of displacement demand. Clearly, for Modified Housner Intensity, increasing the level of
pre-existing damage diminishes the scatter of the displacement demand for a given level of the
intensity measure. This shows that Modified Housner Intensity has a higher correlation with the
displacement demand.
3.6 Derivation of fragility curves
For each of the considered limit states the maximum displacements were compared to the
displacement thresholds of the limit state (dLS). The probability of exceedance of dLS was monitored

A. Penna, M. Rota, A. Mouyiannou, F.Graziotti, G.Magenes / VEESD 2013

for both each range of modified Housner Intensity and PGA. The cumulative distributions are fitted
with lognormal curves, as shown in Figure 9 for the three limit states considered (OLS, DLS and
ULS). The fragility curves for Modified Housner Intensity are shown to the left side of the figure,
while the fragility curves for PGA are shown in the right side. Both the intensity measures were
divided into 10 ranges, a relatively good compromise with respect to the available number of records.
The fragility curves for 20 sub-intervals were also derived with no significant variation with the curves
derived by the division in 10 ranges.

0.4
0.2
0
0

Probability of exceeding DLS

1
0.8
0.6
0.4

1
Probability of exceeding ULS

NODAM
20%COL
30%COL
40%COL
50%COL
60%COL
70%COL
80%COL
90%COL

0.8
0.6
0.4
0.2
0
0

5
10
15
20
Modified Housner Intensity [cm]

0.6
0.4
0.2

0.1

0.2

0.3
PGA [g]

0.4

0.5

0.6

NODAM
20%COL
30%COL
40%COL
50%COL
60%COL
70%COL
80%COL
90%COL

0.8
0.6
0.4
0.2
0
0

5
10
15
20
Modified Housner Intensity [cm]
NODAM
20%COL
30%COL
40%COL
50%COL
60%COL
70%COL
80%COL
90%COL

NODAM
20%COL
30%COL
40%COL
50%COL
60%COL
70%COL
80%COL
90%COL

0.8

0
0

5
10
15
20
Modified Housner Intensity [cm]

0.2
0
0

Probability of exceeding OLS

0.6

Probability of exceeding DLS

0.8

NODAM
20%COL
30%COL
40%COL
50%COL
60%COL
70%COL
80%COL
90%COL

1
Probability of exceeding ULS

Probability of exceeding OLS

0.8
0.6
0.4

0.1

0.2

0.3
PGA [g]

0.4

0.5

0.6

0.2

0.3
PGA [g]

0.4

0.5

0.6

NODAM
20%COL
30%COL
40%COL
50%COL
60%COL
70%COL
80%COL
90%COL

0.2
0
0

0.1

Figure 9. Time dependent fragility curves for various limit states (top to bottom: OLS, DLS and ULS/NC).

The results can be summarized as follows:


- For OLS, increasing the damage level of the pre-existing damage is increasing the probability
of exceeding the OLS displacement threshold. The increase is similar for the two intensity

A. Penna, M. Rota, A. Mouyiannou, F.Graziotti, G.Magenes / VEESD 2013

measured considered. The attainment of OLS is mainly associated with the variation of initial
stiffness which appears to be independent on the intensity measure.
- For DLS, the evolution of damage due to the increasing levels of pre-existing damage can be
captured by PGA (and possibly other spectral ordinates) and slightly differs from the
evolution expressed by Housner Intensity.
- For ULS, which is controlled by the limited ductility of the system, Housner Intensity seems
to be a better indicator for the evolution of damage in the highly nonlinear range.
An interesting picture for the synthetic evolution of the intensity measure values for which the
response of the structure exceeds the predefined limit states as a function of the pre-existing damage
can be observed in Figure 10. The pre-existing damage is represented by a scalar EDP, the ratio
between peak displacement and the ULS displacement threshold. In the figure the first three plots
correspond to the evolution of modified Housner Intensity thresholds for exceeding (from left to the
right) OLS, DLS and ULS given the pre-existing damage level, while the second row of plots is
showing the plots similarly obtained for PGA thresholds.
The median values of both intensity measures at all limit states seem to be the same up to 30% of
collapse pre-existing damage, whilst for higher levels of pre-existing damage, there is a decrease of
the values. The decrease is more noticeable in the dispersion especially for the plots of Housner
Intensity.
20

IH95

18

IH50

16

16

IH05

14

12
10
8

IH [cm]

IH [cm]

14
12
10
8

12
10
8

10

20

30

40

50

60

70

80

90

100

10

Damage level (% of COL)

20

30

40

50

60

70

80

90

100

IH95
IH50
IH05
0

10

PGA50

0.6

PGA05

40

50

60

70

80

90

100

0.6

PGA05

0.5

PGA [g]

0.5

PGA [g]

0.5

0.3

30

PGA95

PGA50

0.4

20

Damage Level (% of COL)

Damage level (% of COL)

PGA95
0.6

PGA [g]

18

IH50

16

IH05

14

IH [cm]

20

20

IH95

18

0.4

0.3

0.4

0.3

0.2

0.2

0.2

0.1

0.1

0.1

PGA95
PGA50
PGA05
0

10

20

30

40

50

60

70

80

Damage level (% of COL)

90

100

10

20

30

40

50

60

70

80

Damage level (% of COL)

90

100

10

20

30

40

50

60

70

80

90

100

Damage Level (% of COL)

Figure 10. Evolution of IM values necessary to attain a given limit state given a pre-existing damage state,
represented by a scalar EDP. Plots for IH (top) and for PGA (bottom) values, from left to right: OLS,
DLS and ULS IM thresholds.

4 CONCLUSIONS
This study focused on setting up a procedure to create state-dependent fragility curves for masonry
structures, considering especially the level of pre-existing damage due to previous earthquake events.
The procedure was applied to one building example and the results are presented. These preliminary
results show that the methodology can be used for time-dependent risk assessment, if combined with
studies on the evolution of the state of the structure with time. The fulfilment of the aforementioned
goal will be completed in the future within the framework of the EC-FP7 REAKT project.

A. Penna, M. Rota, A. Mouyiannou, F.Graziotti, G.Magenes / VEESD 2013

10

ACKNOWLEDGEMENTS
This work was carried out with the financial support of the EC-FP7 REAKT Project - Strategies and
tools for Real Time EArthquake RisK ReducTion (www.reaktproject.eu).
REFERENCES
Graziotti, F., Penna, A. and Magenes, G. (2013). Use of equivalent SDOF systems for the evaluation of
displacement demand for masonry buildings. Proc. of VEESD 2013, 28-30 August, 2013, Vienna, Austria
Housner, G. W. (1959). Behavior of structures during earthquakes, Journal of the Engineering Division, ASCE,
Vol. 85, No. EM14, 109-129.
Iervolino, I., Chioccarelli, E., Giorgio, M. (2012). Time-dependent seismic reliability of damage-cumulative
non-evolutionary bilinear systems. 15th WCEE , Lisboa, Portugal.
Italian Building Code (2008). DM 14.01. 2008: Norme Tecniche per le Costruzioni.
Lagomarsino, S., Penna, A., Galasco, A. and Cattari S. (2013). TREMURI program: an equivalent frame model
for the nonlinear seismic analysis of masonry buildings. Engineering Structures (under review).
Luco, N., Bazzurro, P. and Cornell, A. (2004). Dynamic versus Static Computation of the Residual Capacity of a
Mainshock-Damaged Building to withstand an Aftershock. 13thWCEE ,Vancouver, Canada .
Magenes, G., Calvi, G.M. and Kingsley, G.R. (1995). Seismic Testing of a Full-Scale, Two-Story Masonry
Building: Test Procedure and Measured Experimental Response. University of Pavia, Italy.
Mouyiannou, A., Rota, M., Penna, A., Magenes, G. (2013) Identification of suitable limit states from nonlinear
dynamic analyses of masonry structures, Journal of Earthquake Engineering (in press)
Penna, A., Lagomarsino, S. and Galasco, A. (2013a). A nonlinear macro-element model for the seismic analysis
of masonry buildings. Earthquake Engineering and Structural Dynamics, in press.
Penna A., Rota, M., Mouyiannou A. and Magenes G. (2013b). Issues on the use of time-history analysis for the
design and assessment of masonry structures. Proc. COMPDYN2013, Kos Island, Greece, Paper No. 1327.
Rveillre, A., Gehl, P., Seyedi, D., Modaressi, H. (2012). Development of seismic fragility curves for
mainshock-damaged reinforced-concrete structures. 15thWCEE, Lisboa, Portugal.
Rota, M., Penna, A., Magenes, G. (2010). A methodology for deriving analytical fragility curves for masonry
buildings based on stochastic nonlinear analyses. Engineering Structures, Vol.32, 1312-1323.
Smerzini, C. and Paolucci, R. (2011). Research Project DPC RELUIS 2010-2013 .SIMBAD: a database with
Selected Input Motions for displacement-Based Assessment and Design 2nd release. Politecnico di Milano.

Você também pode gostar