First India Disaster Management Congress, New Delhi, 28-30, November 2006

Seismic risk analysis of reinforced concrete framed structures

K. Balaji Rao and M.B. Anoop

Scientists, Structural Engineering Research Centre,
CSIR Campus, Taramani, Chennai 600 113
e-mail: [email protected]

Abstract
A methodology for estimation of statistical properties (viz. mean and standard deviation) of the expected damage to a
corrosion-affected reinforced concrete framed structure, subjected to stochastic seismic excitation, over a specified
reference time (typically the service life of the structure) is proposed in this paper. The damage to the structure
under the earthquake loading is characterised by the damage index, determined using the modified Park & Ang
damage model. The proposed methodology is illustrated through an example problem and the effect of soil type on
the statistical properties of expected damage to the structure is also studied. The results obtained indicate that there
is a need to consider the effect of corrosion and the type of soil on the estimation of expected damage.

Introduction
The estimation of expected damage due to earthquakes for structures located in seismically active
regions is a matter of concern for design engineers and decision-making authorities in pub lic- and
private-sector organizations. Estimation of expected damage to reinforced concrete structures
located in coastal areas should take into account the degradation in strength and stiffness of the
structure due to corrosion of reinforcement. A methodology for estimation of statistical properties
(viz. mean and standard deviation) of the expected damage to a corrosion-affected reinforced
concrete framed structure, subjected to stochastic seismic excitation, over a specified reference
time (typically the service life of the structure) is proposed in this paper. The methodology is
formulated within the framework of Monte Carlo simulation. The service life of the structure is
divided into reference times, and the statistical properties of the expected damage to the structure
during these times are determined. The time to corrosion initiation and rate of corrosion
propagation are considered as random variables to take into account the uncertainties in the
material properties and variations in the exposure condition. The earthquake occurrences during
the reference time period are modelled as a Poisson process, and the stochastic seismic excitation
is represented by an ensemble of acceleration time histories. The damage to the structure under
the earthquake loading is determined from the results of an inelastic dynamic analysis. The
inelastic damage analysis program IDARC 2D is used for this purpose. The damage to the
structure under the earthquake loading is characterised by the damage index, determined using
the modified Park & Ang damage model. The paper also considers the effect of soil type on the
statistical properties of expected damage to the structure. The proposed methodology is illustrated
through an example problem. Determination of statistical properties of expected damage will be
useful in carrying out vulnerability analysis of the structure and in the regional risk assessment.

Proposed methodology for expected damage estimation

Modelling chloride-induced corrosion of reinforcement
For RC structures located in marine and other aggressive environments, chloride- induced
corrosion of reinforcement is identified as a major mechanism of resistance degradation. The

1
 First India Disaster Management Congress, New Delhi, 28-30, November 2006

process of chloride- induced corrosion of reinforcement in concrete can be considered as a two-

stage process, namely, the corrosion initiation and the corrosion propagation.

Time for corrosion initiation: The time for corrosion initiation is determined from the solution to
Fick's second law of diffusion as
−2
c 2  − 1  cs − ccr  (1)
ti = erf   
4D   cs  
where
c clear cover to reinforcement
D diffusion coefficient for chlorides in concrete
cs surface chloride concentration
ccr critical chloride concentration
erf error function

To account for variations in workmanship and exposure conditions, the diffusion coefficient at
any given time, surface chloride concentration, critical chloride concentration and the clear cover
to reinforcement should be considered as random variables. Thus, the time to corrosion initiation
obtained using Eq. 1 will be a random variable.

The surface chloride concentration decreases as the distance to the sea increases and also
decreases with height [French, 2003]. Hence, in multi- storeyed RC structures, the structural
members in the ground floor will be more susceptible to corrosion than that at the uppermost
floor in a given time period. Thus the time to corrosion initiation should be determined for
different storey levels separately taking into account the variation of surface chloride
concentration with height.

Rate of corrosion: Researchers have proposed different models for determining the rate of
corrosion of reinforcing bar. From a brief review of these models, it is found that the model
proposed by Rodrigues et al [Rodriguez et al, 1996] is widely accepted [Balaji Rao et al, 2000].
Using this model, the rate of corrosion can be determined as:
rcorr = 0.0115 I corr a (2)
where
rcorr rate of corrosion, in mm/year
Icorr average value of corrosion current density, in µA/cm2
α factor for including the effect of highly localised pitting normally
associated with chloride-induced corrosion (varies from 4 to 8)
0.0115 factor which converts µA/cm2 to mm/year
The remaining diameter of the reinforcing bar at any time t, Φ(t), can be obtained as
F(t) = F(0) − rcorr (t − t i ) (3)
where
Φ(0) initial diameter of the reinforcing bar, in mm
ti time required for corrosion initiation, years

2
 First India Disaster Management Congress, New Delhi, 28-30, November 2006

Determination of earthquake damage

Modelling of the Structure [Valles et al, 1996]: The beam elements are modelled as flexural
elements with coupled shear deformations. The hysteretic flexural behaviour is modelled using
the three parameter Park model. The column elements are modelled considering flexural, shear
and axial deformations. The spread plasticity model, available in IDARC, is considered to include
the variation of the section flexibility. The element stiffness matrix is constantly varied
throughout the analysis according to the spread plasticity model and the hysteretic model.

The damage (D) to the structure under the earthquake loading is characterised by the damage
index. The modified Park & Ang damage model is considered in this study for determining the
damage index. A brief description of this model is given below.

The modified Park & Ang Damage Model [Valles et al, 1996]: The Park & Ang damage index
for a structural element is defined as:

δm β
δ u δ u Py ∫ h
DI = + dE (4)

where δ m is the maximum experienced deformation, δ u is the ultimate deformation of the

element, Py is the yield strength of the element, ∫dEh is the hysteretic energy absorbed by the
element during the response history; and β ?is a model constant parameter. A value of 0.1 for the
parameter β ?has been suggested for nominal strength deterioration. The Park & Ang damage
model accounts for damage due to maximum inelastic excursions, as well as damage due to the
history of deformations. Both components of damage are linearly combined.

Three damage indices are computed using this damage model:

1. Element damage index: columns, beams or shear wall elements.
2. Story damage index: vertical and horizontal components and total story damage.
3. Overall building damage index (ODI).

The Park & Ang damage model has been calibrated with observed structural damage of
reinforced concrete buildings (see Table 1).

Procedure for estimating the expected damage to corrosion-affected rc structures due to

earthquakes
Two cases are considered. In the first case, risk to individual structures is considered. Here the
aim is to determine the statistical properties of the damage to the structure given that an
earthquake has occurred. This information will be useful for the designers for performance based
design in which the earthquake types are specified depending upon the design basis. In the second
case, structural risk taking into account the regional risk is considered. In this case, the aim is to
determine an indicator of damage to the nominally similar reinforced concrete structures in a
region over a period of time considering uncertainty in the occurrence of earthquakes and also the
stochasticity in seismic loading. This case hereafter would be referred to as overall (regional) risk
analysis. This information will be useful to the decision makers in the allocation of funds.

3
 First India Disaster Management Congress, New Delhi, 28-30, November 2006

Structural Risk Analysis: The structural system is considered to be deterministic, while the
stochasticity in seismic loading is considered by generating an ensemble of accelerograms using
the method proposed by Deodatis and Shinozuka [1988]. The steps involved are:
1. Generate an ensemble of hundred acceleration time histories for the earthquake
considered using Kanai-Tajimi spectrum
2. Carry out the nonlinear dynamic analysis of the structure and determine the damage for
the ensemble of earthquake acceleration time histories generated
3. Determine the statistical properties of the damage

Overall (regional) risk analysis: The procedure for estimating the expected damage to structures
in a given region is presented in Balaji Rao et al [Balaji Rao et al, 2003]. The stochastic damage
assessment model is based on Poisson process occurrence of earthquakes. This procedure is
modified for including the effect of corrosion on the performance of the structure and is used in
the present study. The procedure consists of the following steps.

1. Determine of the mean recurrence rate (ν) of the earthquake with the specified magnitude for
the region in which the structure is located.
2. Generate an ensemble containing one hundred accelerograms for the earthquake considered
using Kanai- Tajimi spectrum
3. Determine of statistical properties of ti at the different storey levels.
4. Define the reference time (TREF) for damage estimation.
5. Generate one thousand Poisson random variables representing the number of occurrences of
earthquakes during the period TREF with ν as the parameter.
6. For each realisation of number of occurrences of earthquake
6.1. Determine the time of occurrence of each earthquake (Teq)
6.2. Compare Te q with ti to check whether corrosion is initiated.
6.3. Carry out the nonlinear dynamic analysis of the structure and determine the damage
for the ensemble of earthquake acceleration time histories generated.
6.4. Determine the expected overall damage index at the end of TREF using Monte Carlo
simulation.
7. Determine the statistical properties of expected overall damage index for one thousand
realisations of number of earthquakes.

The proposed methodology is illustrated through an example problem in the next section.

Example
The two-bay six-storey reinforced concrete frame (Fig. 1) whose behaviour under seismic loading
has been studied using model testing in laboratory by Skjærbæk et al [Skjærbæk et al, 1998] has
been considered. The frame is assumed to be new at the beginning of reference time period and is
assumed to be located in a coastal region in Peninsular India at a distance of about 50m from
coast. The exposure condition for the frame is characterised as severe. From a database of D
values created at SERC, Chennai, by collecting the values of D reported by various researchers
based on different laboratory tests and field exposure tests [Anoop et al, 1999], the value of D for

4
 First India Disaster Management Congress, New Delhi, 28-30, November 2006

concrete with grade of 20MPa is taken as 5.0x10-8 cm2 /s. The values for cs, ccr and rcorr are
selected as 0.25% by weight of concrete, 0.12% by weight of concrete and 0.18 mm/year, which
are representative of the values reported in similar environments and for similar type and grade of
concrete. To account for variations in workmanship and exposure conditions, the diffusion
coefficient and surface chloride concentration are treated as lognormally distributed random
variables, with mean as the above values and an assumed coefficient of variation (cov) of 0.20.

To study the effect of soil type on the earthquake damage to a structure, two soil types, namely,
alluvium and alluvium on rock are considered in the study. For these two soil types, the values of
spectral acceleration coefficient, for the frame considered (natural period of the frame = 0.47
sec), determined using the acceleration response spectrum given in IS 1893-2002 [IS 1893, 2002]
would be the same. This indicate that the structure will be designed for the same design seismic
base shear for these two soil types, assuming that the structures are in the same seismic zone and
having the same importance. Thus the designs would be the same for both soil types. The
parameters of the Kanai-Tajimi spectrum for these soil types are given in Table 2. From the
records of earthquake magnitudes in the region considered, the mean recurrence rates of
earthquakes with magnitude > 6.0 is determined as 0.0086 yr -1 . One of the important aspects of
prediction of earthquakes in a given peninsular region is the paucity of acceleration time histories
corresponding to various earthquakes, especially for earthquakes of larger magnitudes. In this
study, the 1940 El Centro earthquake of magnitude 6.9 (PGA = 0.348g) is considered. A typical
realisation of acceleration time history is shown in Fig. 2. The values of expected damage and the
statistical properties of expected damage for different reference time periods are determined using
the proposed procedure.

Results and discussion

Structural Risk Analysis:
1. The values of expected damage index for the columns and beams at different storey levels
for the frame with uncorroded reinforcement for the soil types considered are shown in
Figs. 3 and 4. From these figures, it is noted that the columns and beams in the second and
third storey levels are showing the maximum damage. This information will be useful in
identifying the storeys that will suffer maximum damage in the event of an earthquake.

2. The values of expected overall damage index (ODI) and the cov of ODI for the
uncorroded frame for the two soil types considered are given in Table 3. As mentioned
above, designs would be the same for the two soil types, assuming that the structures are
in the same seismic zone and having the same importance. But from the values given in
Table 3, it is noted that the expected damage to a structure located in alluvium soil is
higher than that for the similar structure located in alluvium on rock soil. This can be
attributed to the higher energy content of the Kanai- Tajimi spectrum near the structural
natural frequency (2.11 Hz) for alluvium soil when compared to that for alluvium on rock
soil. This observation suggests the need for seismic microzonation.

3. From Table 3, it is noted that the cov of ODI is higher for the structure located on
alluvium on rock soil, indicating that damage is distributed over a larger range when
compared to that for the structure on alluvium soil. While this may be due to the lesser

5
 First India Disaster Management Congress, New Delhi, 28-30, November 2006

degree of peakedness of the Kanai- Tajimi spectrum for the alluvium on rock soil,
indicated by a lower value of ε (see Table 2), further investigations, looking more into the
probability distributions of ODI, are required.

4. The expected values of damage index for columns and beams at different storey levels for
the uncorroded frame and the frame exposed to severe environment for 30 years, located
on alluvium soil, are shown in Figs 5 and 6. From these figures, it is noted that, as
expected, in most of the storey levels, the expected damage to the beams and columns are
higher for the corroded structure. It is also noted that while the magnitude of expected
damage to the beams in different storey levels for the corroded structure follow the similar
pattern as that of the uncorroded structure, there is a change in the magnitude of expected
damage to the columns at different storey levels when corrosion is considered. For
instance, maximum expected damage for columns is at the third storey for the corroded
structure, while it was at the second storey for the uncorroded structure. This indicates the
importance of considering the effect of corrosion for realistically estimating the expected
damage to the structure considered after a period of time.

Overall (regional) risk analysis:

1. The mean of expected ODI for the two soil types with and without considering the effect
of corrosion are shown in Fig. 7. From this figure, it is noted that the mean of expected
ODI increases with time. This is expected, because the probability of having (one or
more) earthquakes within the reference time increases with increase in reference time,
thus increasing the expected damage. From Fig. 7, it is noted that as expected, the damage
to the structure in alluvium soil is higher than that for the structure in alluvium on rock
soil, indicating the importance of considering the type of soil in seismic risk analysis. It is
also noted that as the reference time increases, the effect of corrosion on expected damage
become significant. This is expected since with increase in reference time, more and more
structures would have corroded and the level of corrosion in already corroded structures
increase, thus increasing the overall damage.

2. The cov of expected ODI for the soil type alluvium with and without considering the
effect of corrosion are given in Table 4. From Table 4, it is noted that the values of cov of
expected ODI are very high when compared to the cov of ODI obtained from structural
risk analysis (see Table 3), but decreases with increase in reference time. The high values
of cov are due to the superposition of uncertainty in the occurrence of earthquakes. This
uncertainty dominates for smaller reference time periods, but decreases with increase in
reference time period, thus reducing the cov of expected damage. It is also noted that the
cov values are higher for the corroded structure when compared to that for the uncorroded
structure, especially at larger reference time periods. This is due to the increase in
uncertainty on corrosion level with increase in time.

From the results obtained for structural risk analysis and overall (regional) risk analysis, it is
noted that the effect of corrosion and the type of soil have significant effects on the expected
damage to the structure. Thus, there is a need to explicitly consider the type of soil and effect of
corrosion in seismic risk analysis of reinforced concrete structures.

6
 First India Disaster Management Congress, New Delhi, 28-30, November 2006

Summary and conclusions

A methodology for estimation of statistical properties (viz. mean and standard deviation) of the
expected damage to a corrosion-affected reinforced concrete framed structure, subjected to
stochastic seismic excitation, over a specified reference time (typically the service life of the
structure) is proposed. The proposed methodology is illustrated through an example problem, also
studying the effect of soil type on the statistical properties of expected damage to the structure. In
the present study, the effect of corrosion is limited to the reduction in area of reinforcement. But,
also the deterioration of bond and the cracking of cover concrete due to the corrosion of
reinforcement need to be taken into account. Also, only one earthquake source is considered,
while there can be more than once source of earthquake which need to be taken into account.

Acknowledgement
This paper is published with the kind permission of the Director, Structural Engineering Research
Centre, Chennai. The authors are grateful to Dr T.V.S.R. Appa Rao, former Director, SERC,
Chennai, for giving the encouragement.

References
Anoop, M. B., Balaji Rao, K. and Appa Rao, T. V. S. R., (1999) Durability of reinforced concrete
with respect to chloride ingress - A critical review, Project Report No. SS-OLP 07741-RR-99-2,
Structural Engineering Research Centre, Chennai.
Balaji Rao, K., Anoop, M.B. and Appa Rao, T.V.S.R., (2000) A critical review of corrosion
propagation model and results of probabilistic analysis of resistance degradation of reinforced
concrete flexural members due to corrosion, Project Report No. SS-OLP 07741-RR-2000-1,
Structural Engineering Research Centre, Chennai.
Balaji Rao, K., Lakshmanan, N., Anoop, M. B., Gopalakrishnan, S. and Manjuprasad, M., (2003)
A Stochastic Model for Regional Seismic Risk Assessment, SERC Project Report No. SS-
OLP09441-RR-2003-2, Structural Engineering Research Centre, Chennai.
Deodatis, G. and Shinozuka, M., (1988) Auto –Regressive Model for Non- Stationary Stochastic
Processes, Journal of Engineering Mechanics, ASCE, Vol. 114, No. 11, pp.1985-2012.
Elghadamsi, F. E., Mohraz, B., Lee, C. T. and Moayyad, P., (1988) Time-Dependent Power
Spectral Density of Earthquake Ground Motion, International Journal of Soil Dynamics and
Earthquake Engineering, Vol. 7, No. 1, 15-21.
French, C., (2003) Durability of concrete structures, Structural Concrete, Vol. 4, No. 3, pp. 101-
107.
IS 1893-2002, (2002) Criteria for Earthquake Resistant Design of Structures: Part 1 – General
Provisions and Buildings IS 1893-2002, Bureau of Indian Standards, New Delhi.
Park, Y.J., Ang, A.H.-S. and Wen, Y.K., (1985) Seismic Damage Analysis of Reinforced
concrete Buildings, Journal of Structural Engineering, ASCE, Vol. 111, No. ST4, pp. 740-757.

7
 First India Disaster Management Congress, New Delhi, 28-30, November 2006

Rodriguez J., Ortega L. M., Casal J. and Diez J. M. (1996), Assessing structural conditions of
concrete structures with corroded reinforcement. in Concrete repair, rehabilitation and protection,
R. K. Dhir and M. R. Jones (Eds.), E&FN Spon, London, pp. 65-78.
Skjærbæk, P.S., Nielsen, S.R.K., Kirkegaard, P.H. and Cakmak, A.S., (1998) Damage
localization and quantification of earthquake excited RC- frames, Earthquake Engineering and
Structural Dynamics, Vol. 27, pp. 903-916.
Valles, R.E., Reinhorn, A.M., Kunnath, S.K., Li, C. and Madan, A., (1996) IDARC 2D Version
4.0: A program for the inelastic damage analysis of buildings, Technical Report NCEER-96-
0010, National Center for Earthquake Engineering Research, State University of New York at
Buffalo.
VanMarcke, E. H., (1983) Random fields: analysis and synthesis, Cambridge, MA: MIT Press.

TABLE 1 INTERPRETATION OF OVERALL DAMAGE INDEX USING PARK & ANG

MODEL [PARK et al, 1985]
Degree of Physical appearance Damage State of Building
Damage Index
Collapse Partial or total collapse of building > 1.0 Loss of building
Severe Extensive crushing of concrete; 0.4 – 1.0 Beyond repair
disclosure of buckled reinforcement
Moderate Extensive large cracks; spalling of < 0.4 Repairable
concrete in weaker elements
Minor Minor cracks; partial crushing of - -
concrete in columns
Slight Sporadic occurrence of cracking - -

TABLE 2 KANAI-TAJIMI SPECTRUM PARAMETERS FOR DIFFERENT SOIL

TYPES
ε**
*
Soil Type Ground frequency Ground damping* Ground intensity*
(Hz) (1/Hz)
Alluvium 2.92 0.34 0.102 0.94
Alluvium on rock 3.64 0.30 0.078 0.91
(Note *- from Elghadamsi et al [1988]
** - defined as in VanMarcke [1983] )

8
 First India Disaster Management Congress, New Delhi, 28-30, November 2006

TABLE 3 EXPECTED VALUE AND COV OF ODI FOR THE UNCORRODED FRAME
SUBJECTED TO STOCHASTIC SEISMIC EXCITATION
Soil Type Expected value cov
Alluvium 0.30 1.04
Alluvium on rock 0.18 1.31

TABLE 4 VALUES OF COV OF EXPECTED ODI AT DIFFERENT REFERENCE

TIMES FOR THE STRUCTURE ON ALLUVIUM SOIL
Reference time Without considering Considering
(years) corrosion corrosion
5 5.59 5.59
10 3.30 3.39
20 2.23 2.47
30 1.71 2.70

(6)
Concrete grade: M20
2.75 m
(5)

0.065 m
2.75 m
(4) 30 mm Φ bars 0.3 m

300 kg/ m
2.75 m
0.25 m
(3)
BEAM
2.75 m
(2) 0.065 m

30 mm Φ bars 0.3 m
2.75 m

(1)
0.25 m
2.75 m
COLUMN

6.0 m 6.0 m

FIG. 1 SIX STORY BUILDING SUBJECTED TO BASE EXCITATION (FROM

SKJAERBAEK et al, 1998)

9
 First India Disaster Management Congress, New Delhi, 28-30, November 2006

0.4

0.3

0.2

0.1

PGA (g)
0
0 5 10 15 20 25 30
-0.1

-0.2

-0.3

-0.4
time (sec)

FIG. 2 TYPICAL SIMULATED ACCELERATION TIME HISTORY FOR

ALLUVIUM TYPE SOIL (PGA = 0.348 g)

6
storey 6
5

4
Floor level

3
Alluvium

2 Alluvium on rock

1
storey 1
0
0 0.05 0.1 0.15 0.2 0.25 0.3
expected damage index
FIG. 3 VALUE OF EXPECTED DAMAGE INDEX FOR COLUMNS AT DIFFERENT
STOREY LEVELS OF RC FRAME WITH UNCORRODED REINFORCEMENT
SUBJECTED TO STOCHASTIC SEISMIC EXCITATION

10
 First India Disaster Management Congress, New Delhi, 28-30, November 2006

6
storey 6
5 Alluvium

4 Alluvium on rock
Floor level

1
storey 1
0
0 0.05 0.1 0.15 0.2
expected damage index
FIG. 4 VALUE OF EXPECTED DAMAGE INDEX FOR BEAMS AT DIFFERENT
STOREY LEVELS OF RC FRAME WITH UNCORRODED REINFORCEMENT
SUBJECTED TO STOCHASTIC SEISMIC EXCITATION

6
storey 6
5

4
Floor level

uncorroded structure
3
structure exposed to severe
2 environment for 30 years

1
storey 1
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
expected damage index
FIG. 5 EXPECTED DAMAGE INDEX FOR COLUMNS AT DIFFERENT STOREY
LEVELS FOR RC FRAME LOCATED ON ALLUVIUM SOIL AND SUBJECTED TO
STOCHASTIC SEISMIC EXCITATION

11
 First India Disaster Management Congress, New Delhi, 28-30, November 2006

6
storey 6
5

4
Floor level

3
uncorroded structure

2 structure exposed to severe

environment for 30 years
1
storey 1
0
0 0.05 0.1 0.15 0.2 0.25
expected damage index

FIG. 6 EXPECTED DAMAGE INDEX FOR BEAMS AT DIFFERENT STOREY LEVELS

FOR RC FRAME LOCATED ON ALLUVIUM SOIL AND SUBJECTED TO
STOCHASTIC SEISMIC EXCITATION

0.14

Alluvium - considering corrosion

0.12
Alluvium - without considering corrosion

0.1 Alluvium on rock - considering corrosion

Alluvium on rock - without considering corrosion

<E[ODI]>

0.08

0.06

0.04

0.02

0
0 5 10 15 20 25 30 35
reference time (years)
FIG. 7 VARIATION IN MEAN OF EXPECTED ODI WITH REFERENCE TIME

12