COMPARISON OF FRAGILITY ANALYSIS FOR A RC FRAME BY

TWO MAJOR APPROACHES

Haran Pragalath D C1, Pradip Sarkar2 and Robin Davis3

Department of Civil Engineering, National Institute of Technology Rourkela
Odisha - 769008, India

ABSTRACT

Performance-based seismic design is the modern approach to earthquake resistant design for
buildings. Many parameters involved in seismic design have uncertainty associated with
them. Characterizing the probabilistic nature of these parameters can be done through the use
of Fragility Curves. A fragility analysis assesses the probability that the seismic demand
placed on the structure exceeds the capacity conditioned on a chosen IM representative of the
seismic loading. Demand (D) and capacity (C) are assumed to follow a lognormal
distribution, and the probability of exceeding a specific damage state for a particular
component can be estimated with the standard normal cumulative distribution function as per
Cornell et. al. (2002). There is a different approach for fragility analysis using Incremental
Dynamic Analysis (IDA) as given by ATC 63 (2008). A two-dimensional building model is
analyzed for a set of 30 ground motions, and in each case the peak demand measures (e.g.,
inter-storey drift etc.) are recorded. Fragility curves are drawn based on the above mentioned
two approaches and comparison is done based on the results obtained. Thirty spectrum
consistent (IS 1893:2002) time-history data are selected and used for the analyses. A four
storeyed (G+3) RC frame designed as per relevant Indian standard is chosen for this study.

1. INTRODUCTION
Due to the randomness in nature of seismic analysis for structures, it is not always possible to
do deterministic approach to get accurate results. Randomness/Uncertainty caused may be due
to change in Material properties, Time history data, loading profiles etc. A probabilistic based
approach is the most appropriate to account the uncertainties. Fragility curve is probabilistic
based approach to represent the safety of the structure incorporating the uncertainties
involved.

1
Research Scholar, NIT Rourkela, Orissa 769 008, India. email: haran5441@gmail.com
2
Associate Professor, NIT Rourkela, Orissa 769 008, India. email: sarkarp@ nitrkl.ac.in
3
Assistant Professor, NIT Rourkela, Orissa 769 008, India. email: robind@nitrkl.ac.in
Mathematically, fragility curves can be defined as the probability of exceedance of damage at
various levels of ground acceleration, which is considered as an Intensity Measure (IM). Out
of the various existing methodologies for development of fragility curves, a method based on
nonlinear time history analysis and the probabilistic demand model suggested by Cornell et. al
(2002) is discussed. Recently ATC 63 (2008) has introduced a method to derive fragility
curve by conducting Incremental Dynamic Analysis for evaluating seismic collapse safety of
buildings. In the present study, a comparison of fragility curves based on the above mentioned
approaches is presented.
2. EXAMPLE FRAMES
The building frame considered for numerical analysis in the present study is assumed to be
located in Indian seismic zone V with medium soil conditions. The design peak ground
acceleration (PGA) of this zone is specified as 0.36g. The frame is designed as per prevailing
practice in India. Seismic loads are estimated as per IS 1893 (2002) and the design of the RC
elements are carried out as per IS 456 (2000) standards. The characteristic strength of
concrete and steel were taken as 25MPa and 415MPa. The buildings are assumed to be
symmetric in plan, and hence a single plane frame may be considered to be representative of
the building along one direction. Typical bay width and column height in this study are
selected as 3m and 3.2m respectively, as observed from the study of typical existing
residential buildings. A configuration of 4 stories and 4 bays (4s4b) is considered in this
study. fig.1(a) shows the geometry and reinforcement details of the building frame
considered. The dead load of the slab (3 m x 3 m panel), including floor finishes, is taken as
2.5 kN/m2 and live load as 3 kN/m2. The design base shear (VB) is calculated as per IS 1893
(2002).

Z I Sa
V W (1)
2 R g
where, seismic zone factor, Z = 0.36, Importance factor I = 1.0, Response reduction factor R =
3.0.
3. MODELLING
Bare frame is modelled in the program Seismostruct (2007) for Time history analysis and
Incremental Time history analysis. Seismostruct uses fiber based spread plasticity elements
for frame elements as shown in fig.2(b). Manders (1988) confinement model is used for
concrete. For reinforcement steel, Menegotto and Pinto (1973) stress-strain relationship with
Filippous isotropic hardening rule is used. Rayleigh damping model is used with 3%
damping in the first mode and 5% damping in the third mode.

@ 3.0 m

@ 3.2 m =

(a) Elevation of frame (b) Fiber based element discretization

fig 1. 2D frame with section detailing (Seismostruct, 2013)

4. EARTHQUAKE DATA
The number of ground motions required for an unbiased estimate of the structural response is
3 or 7 as per ASCE 7-05. However, ATC 58(2009) 50% draft recommends a suite of 11 pairs
of ground motions for a reliable estimate of the response quantities. ASCE/SEI 41 (2005)
suggests 30 recorded ground motions to meet the spectral matching criteria for NPP
infrastructures. A set of thirty IS 1893 (2002) spectrum compatible ground motions are
generated using WavGen (Mukherjee and Gupta, 2002) from the thirty far field natural time
histories collected from Haselton(2006).

5. FRAGILITY CURVES
The fragility function represents the probability of exceedance of the selected Engineering
Demand Parameter (EDP) for a selected structural limit state (DS) for a specific ground
motion intensity measure (IM). Fragility curves are cumulative probability distributions that
indicate the probability that a component/system will be damaged to a given damage state or a
more severe one, as a function of a particular demand. A fragility curve can be obtained for
each damage state.
5.1 Method I
The fragility can be expressed in closed form using Eq. 2,

P (C-D 0|IM) = (2)
|

where, C is the drift capacity, D is the drift demand, Sd is the median of the demand and Sc is
the median of the chosen damage state (DS). d/IM and c are dispersion in the intensity
measure and capacities respectively. Eq. 2 can be rewritten as Eq. 3 for component fragilities
(Nielson, 2005) as,

P (DS|IM) = (3)

where, IMm = exp , a and b are the regression coefficients of the probabilistic

Seismic Demand Model (PSDM) and the dispersion component, is given as,

|
= (4)

5.1.1 Probabilistic Seismic Demand Model

It has been suggested by Cornell et. al (2002) that the estimate of the median engineering
demand parameter (EDP) can be represented by a power law model as given in Eq. 5.

= ( ) (5)

5.2 Method II
According to ATC-63 (2008), the collapse fragility of each index archetype is defined by the
random variable, SCT (or Sa), assumed to be equal to the product of the median value of the
collapse ground motion intensity, SCT[T] (or Sa[T]), as calculated by nonlinear dynamic
analysis, and the random lognormal variable, TOT.

S CT S CT T TOT

Where, TOT is a lognormal random variable with a median value of unity and a lognormal
standard deviation of TOT. The lognormal random variable is assumed to be the product of
four component random variables:

TOT RTR DR TD MDL

where RTR MDL DR, and TD are lognormal random variables with median values of unity,
and lognormal standard deviation parameters, RTR , DR , TD and MDL respectively. Since
these parameters are assumed to be statistically independent, the lognormal standard deviation
parameter, TOT, describing total collapse uncertainty, is given by:

TOT RTR DR TD MDL

where,
TOT = total system collapse uncertainty
RTR = record-to-record collapse uncertainty
DR = design requirements related collapse uncertainty
TD = test data related collapse uncertainty
MDL = modeling related collapse uncertainty

6. MATERIAL UNCERTAINTY
To consider the uncertainty in the material properties, the characteristic strength of concrete,
fck and the yield strength of the steel, fy
are taken as the random variable. The statistical details (Table 2) of the parameters, fck and fy
have been taken from Ranganathan (1999). A set of thirty statistically equivalent analytical
models are generated using the Latin Hypercube Sampling (LHS) scheme.

Table 1: Details of random variables used in LHS scheme

Variable Mean COV (%) Distribution Remarks
Concrete fck (MPa) 30.28 21.0 Normal Uncorrelated
Steel fy (MPa) 468.90 10.0 Normal Uncorrelated

7. CAPACITY MODEL
Limit states define the capacity of the structure to withstand different levels of damage. It can
be represented qualitatively (HAZUS) or quantitatively (FEMA 356). In this study, the inner
storey drift limits suggested by Ghobarah (2004) has been taken as the median values of the
capacity, Sc. The median inter-storey drifts for light repairable damage (IO), moderate
repairable damage (LS) and near collapse (CP) for moment concrete frames are listed in Table
2.
The dispersion in capacity, c is dependent on the building type and construction quality. For
c whereas ATC 58(2006) 50% draft suggests 0.10, 0.25 and 0.40 depending on the quality of
construction. In this study, dispersion in capacity has been assumed as 0.25. The details of the
capacity model considered in this study are listed in Table 2. In this present study, maximum
of inter-storey drift () at the floor levels is taken as the engineering damage parameter
(EDP) and peak ground acceleration (Sa[T1]) as the intensity measure (IM). For IDA results
it is assumed that the failure reaches when inter-storey drift reaches 4%.
Table 2: Damage limits and dispersion associated with various structural performance levels

Limit Performance level Median Interstorey Dispersion, c

states Drifts Sc for
designation Concrete frames, (%)
IO Light repairable damage 1 0.25
LS Moderate repairable 2 0.25
damage
CP Near collapse 4 0.25

6. DISCUSSIONS AND CONCLUSIONS

Figure 4, shows the probabilistic seismic demand model and the corresponding fragility
curves developed by method I for three performance levels namely, IO, LS and CP. The
variation probability of exceedance of the inter-storey drift with the Sa[T1] is shown. It can be
seen that, for any specific Sa[T1] the exeedance probability for the performance level IO is
higher than that for the LS, which is higher than that of CP, as expected.

Fig 4. PSDM model and Fragility Curve (Method I)

Fig. 5 shows Incremental Dynamic analysis (IDA) curves obtained as part of the Method II.
An Inter-storey drift of 4% is considered as the drift at collapse for the IDA curve. The
spectral acceleration, Sa(T) corresponds to the median collapse capacity is marked in the Fig.
Fragility curve developed corresponding to two extreme capacity standard deviations using
the Method II is shown in Fig. 6. The two extreme standard deviations are corresponding to
two different qualities of constructions. As the standard deviations increases the exceedance
probability in the initial region, i.e. upto a spectral acceleration of 1.2g, the exceedance
probability increases.
2

1.5
Sa[T] (median)
Sa (T1), g

0.5

0
0 2 4 6 8 10 12 14
% of inter storey drift

Fig 5. IDA curves for the frame considered

Fig 6. Fragility Curves according to Fig 7. Fragility Curves comparison b/w

ATC-63 (Method II) Method I and Method II

Fig. 7 shows the comparison of fragility curves for two different methods. It can be seen that
the exceedance probabilities predicted by two different methods (Method I and II), although
the approaches are different, is not so much different in terms of fragility curves. However,
ATC-63 explicitly describes the uncertainties involved and a procedure to take into account of
the same in the fragility analysis. The computational cost of IDA is slightly higher than that of
Method I.
REFERENCES
[1] ATC 58 50% Draft, 2009. Guidelines for Seismic Performance Assessment of Buildings,
Applied Technology Council, Redwood City, CA.
[2] ATC-63, 2008. Applied Technology Council, Quantification of Building Seismic
Performance Factors - 90% Draft, Report No. ATC-63, 2008, prepared by the Applied
Technology Council, Redwood City, California for the Federal Emergency Management
Agency, Washington, D.C.
[3] Cornell, C. A., F. Jalayer, R.O. Hamburger and D.A. Foutch, 2002. The Probabilistic
Basis for the 2000 SAC/FEMA Steel Moment Frame Guidelines, Journal of Structural
Engineering 128(4), 526-533.
[4] Ghobarah A., 2001. Performance-based Design in Earthquake Engineering: State of
Development, Engineering Structures 23, 878-884.
[5] Haselton., C., B.,(2006) Assessing seismic collapse safety of modern reinforced concrete
moment frame buildings, PhD Dissertation, December 2006, Stanford University.
[6] IS 1893 Part 1 (2002) Indian Standard Criteria for Earthquake Resistant Design of
Structures. Bureau of Indian Standards. New Delhi. 2002.
[7] IS 456, (2000). Indian Standard for Plain and Reinforced Concrete - Code of Practice,
Bureau of Indian Standards, New Delhi. 2000.
[8] Mander J.B., Priestley M.J.N., Park R, 1988. Theoretical stress-strain model for confined
concrete, Journal of Structural Engineering, Vol. 114, No. 8, pp. 1804-1826.
[9] Menegotto M., Pinto P.E, 1973. Method of analysis for cyclically loaded R.C. plane
frames including changes in geometry and non-elastic behaviour of elements under
combined normal force and bending, Symposium on the Resistance and Ultimate
Deformability of Structures Acted on by Well Defined Repeated Loads, International
Association for Bridge and Structural Engineering, Zurich, Switzerland, pp. 15-22.
[10] Mukherjee, S and V.K. Gupta, 2002. Wavelet-based Generation of Spectrum
Compatible Time histories, Soil dynamics and Earthquake engineering 22, 799-804.
[11] Nielson, B.G, 2005. Analytical Fragility Curves for Highway Bridges in Moderate
Seismic zones, Ph.D. Thesis, Georgia Institute of Technology.
[12] Ranganathan, R. (1999) Structural Reliability Analysis and Design, Jaico Publishing
House, Mumbai.
[13] Seismostruct, (2012). SeismoStruct - A Computer Program for Static and Dynamic
Nonlinear Analysis of Framed Structures [online]. < http://www.seismosoft.com/ > (Feb.
02, 2012).