You are on page 1of 12

COVER PAGE

Title: Probabilistic Damage Control Approach for Seismic Design of Bridges

Authors: Ashkan Vosooghi, Ph.D., P.E. (Contact person and presenter)


Bridge Engineer III
AECOM Transportation
2020 L Street, Suite 300
Sacramento, CA 95811
Phone: (916)404-5800
Fax: (916)404-1557
E-Mail: ashkan_vosooghi@yahoo.com

Amarjeet Saini
Doctoral Student
Department of Civil and Environmental Engineering/0258
University Of Nevada Reno
Reno, Nevada 89557
Phone: (408 )504-1713
E-mail:asaini@unr.edu

Mehdi Saiidi, Ph.D., P.E.


Professor
Department of Civil and Environmental Engineering/0258
University Of Nevada Reno
Reno, Nevada 89557
Phone: (775 )784-4839
Fax: (775)784-1390
E-mail:saiidi@unr.edu
Probabilistic Damage Control Approach for Seismic Design
of Bridges
A. Vosooghi1, A. Saini2, and M. Saiid Saiidi3

ABSTRACT

Probabilistic damage control approach (PDCA) is a new procedure for seismic design of
bridges subjected to earthquakes. The uncertainties in seismic response and seismic demand are
included in this method. As part of a study funded by the California Department of Transportation
(Caltrans), this innovative design methodology is developed by incorporating the extent of lateral
displacement nonlinearity defined by “Damage Index” (DI). The performance objective was defined
based on apparent damage states and the damage states were correlated to damage indices based on
a previous study at the University of Nevada, Reno. The correlation between DI and DS was
determined from a statistical analysis of measured data from 21 bridge column models subjected to
seismic loads. Extensive analytical modeling of seismic response of single column bents was
conducted. A wide range of variables was included in the study to address the effect of aspect ratio,
longitudinal steel ratio, site class, and distance to active faults. Each column was analyzed under
several near-field and far-field ground motions. A statistical analysis of the demand DIs was
performed to develop fragility curves and calculate the reliability index for each DS. A direct
method was developed to calibrate design DI to obtain a given reliability. The calculated reliability
indices and fragility curves showed that the proposed method can be effectively used in seismic
design of new bridges.

INTRODUCTION

Performance-based seismic design of bridges has become of interest to researchers and


structural engineers after Loma Prieta earthquake in 1989. Displacement of bridge components is
used in performance-based design methods as a measure of earthquake demand rather than forces
used in conventional seismic design methods (Priestley et al. 2007; Suarez and Kowalsky 2010).
There are uncertainties in both structural responses and earthquake demands because of scatter on
actual properties of structures and random nature of the earthquake. Vosooghi and Saiidi (2012)
developed fragility curves to correlate seismic response of bridge piers and apparent damage states
using a comprehensive database of large-scale shake table tests. The fragility curves provide a
probabilistic approach to account for uncertainties in the seismic response of bridge piers subjected
to earthquakes. Probabilistic Damage Control Approach (PDCA) is an innovative performance-
based seismic design method for bridges and takes into account the effects of uncertainties in
structural responses and earthquake demands. This method incorporates the extent of lateral plastic
displacement at different earthquake return periods. The extent of lateral plastic displacement was
quantified by Torzani et al. (2008) based on limited experimental results and engineering judgments

1
Ph.D., P.E., Bridge Engineer III, AECOM Transportation, 2020 L Street, Suite 300, Sacramento, CA 95881
2
Doctoral Student, Civil and Environmental Engineering Department, University of Nevada Reno, Reno, NV 89557
3
Professor, Civil and Environmental Engineering Department, University of Nevada Reno, Reno, NV 89557

1
for different bridge categories subjected to earthquakes with various return periods. The
uncertainties in earthquake demands and bridge responses were not included in their study.
In the present study, extensive nonlinear dynamic analyses were conducted to account for
earthquake demand uncertainties in PDCA. Column dimensions, steel ratios, column bent
configuration, site classes, and distance to active faults were of the parameters included in the
analysis. Fragility curves were developed to model the demand uncertainties. Having statistical
distribution of seismic response and demand, reliability analysis was conducted to calibrate DI to
reach a given reliability against failure consistent with AASHTO LRFD (AASHTO 2010) under the
design earthquake with a return period of 1000 years. The reliability against lower damage states
was also determined.

PERFORMANCE OBJECTIVE

Apparent damage states are used to define performance objective of bridges in PDCA. Six
distinct apparent damage states were used in this study as follows: DS-1: flexural cracks; DS-2:
minor concrete spalling and/or minor shear cracks; DS-3: extensive concrete spalling and/or
extensive shear cracks; DS-4: visible bars; DS-5: imminent failure in which the column is
approaching failure and damage has begun to penetrate the confined core, and there might be signs
of slight longitudinal bar bending without rupture (Vosooghi and Saiidi 2010); and DS-6: failure.
Based on discussion with Caltrans engineers, DS-3 was selected as the highest acceptable damage
state in a standard bridge subjected to a 1000-year earthquake. The bridge will not be serviceable to
public vehicle however will remain serviceable to emergency vehicle very restricted. The bridge
performance after reaching other damage states is listed in Table 1.

TABLE I. BRIDGE SERVICEABILITY AND DAMAGE STATES


Service to Service to
Damage States
Public Vehicles Emergency Vehicles
DS-1 Yes Yes
DS-2 Yes Yes
DS-3 No Yes, only 1 lane
DS-4 No Yes, only 1 lane
DS-5 No No
DS-6 No No

UNCERTAINTIES IN SEISMIC RESPONSE

There are uncertainties in seismic response of bridges because of scatter on actual properties
and bent configurations. Consequently a probabilistic approach is required to determine the seismic
response. Damage Index (DI) is used as seismic response parameter in PDCA and defined as follow
(Tourzani et al. 2008):

Dmax  DY
DI  (1)
Du  DY

Where Dmax is maximum displacement demand, and DY and DU are effective yield and
ultimate displacement, respectively.

2
A database of 21 large-scale bridge column bents tested on shake tables and under quasi-
static loads was used in this study. All column models were designed based on current seismic
design codes. This database consisted of column bents used in the previous study by Vosooghi and
Saiidi (2012) and shake table test results by Saiidi et al. (2013). Only column bents tested until
failure were used. The database included single column bents, two-column bents, and complete
bridge models. The columns had different aspect ratios, support conditions (fixed-pinned and fixed-
fixed), cross-sectional properties, and material properties and were subjected to different
earthquakes (far-field and near-field motions).
The columns were tested under gradually increasing loads in the original studies. DI was
calculated at each damage state using maximum measured displacement at that damage state. The
effective yield displacement was determined based on idealized elasto-plastic measured envelop
curves.
Measured damage indices at each damage state showed a scatter on data. For example, Fig.
1 shows the scatted of measured damage indices at DS-3. The horizontal and vertical axes show
measured DI and cumulative frequency, respectively. This figure indicates that a DI between 0.22
and 0.56 corresponds with extensive concrete spalling and/or extensive shear cracks (DS-3).
Cumulative lognormal distribution (fragility curve) was used to model the data scatter, and the
Smirnov-Kolmogorov goodness-of-fit test (Massey 1951) was used as the acceptance criterion for
lognormal distribution. In this criterion, the hypothesis that the data can be assumed to be
lognormally distributed is accepted if all test data points are placed between the lower and upper
confidence limits. In this study, the significance level of 10% was selected based on the Naeim et al.
(2005) recommendation. Figure 1 indicates that all data were placed between the limits. Similar
trend was observed for other damage states.

Figure 1. Fragility and limit curves for DI at DS-3

Figure 2 shows fragility curves (cumulative lognormal distribution) for five damage states.
It should be noted that normal distribution was used for DS-1 because some of measured DIs at DS-
1 had negative values.
As mentioned, DS-3 under design earthquake was selected as the performance objective.
Figure 2 shows that the median DI associated with DS-3 is 0.37. Tentatively the median DI was
rounded down to 0.35 and used in the rest of study.

3
Figure 2. Fragility curves for damage index at different damage states

UNCERTAINTIES IN SEISMIC DEMAND

There are uncertainties in seismic demand because of scatter in structural properties and
random nature of earthquakes. Damage Index was used as response parameter to study the scatter
on seismic demand. Twenty-four single column bents with different properties were designed, and
each column was subjected to 25 earthquakes using nonlinear dynamic analysis (NDA) to study the
uncertainties in seismic demand. The details are discussed in the following sections.

Column Design

Twenty-four single column bents with different properties were designed using Caltrans
design spectra (Caltrans 2012) for target DI of 0.35. The target DI was selected based on the median
DI associated with DS-3 (Fig. 2). The columns had different longitudinal steel ratios, heights, and
support conditions. The specified concrete strength of 3.6 ksi and steel Gr. 60 were used in the
design. For each column, the transverse steel was calculated to provide required confinement
pressure to meet the target DI. Damage index was calculated using elasto-plastic pushover curve
based on SDC (Caltrans 2010). Seismic displacement demands were determined based on the rule
of “equal displacement” and the effective stiffness of the column bents. Each column was designed
for two site classes: site class B/C (V s30=760 m/sec) and site class D (Vs30=270 m/sec).
Table 2 summarizes the column properties and design parameters. All columns had a
circular section with 6 ft diameter. The diameter of 6 ft was selected based on common practice by
Caltrans. The columns were not designed for shear because the intention of this article was to study
the scatter on seismic demand of the columns designed for a given DI. The columns with double -
curvature configuration and high longitudinal steel ratio reached a DI smaller than 0.35 with zero
confinement because they nearly remained elastic under the design earthquake. Similar trend was
observed in some of tall columns (H=60 ft) because of their relatively low stiffness. The columns
shown with italic font style in Table 2 could not reach the target DI=0.35 and were not included in
NDA.

4
TABLE 2: COLUMN PROPERTIES AND DESIGN PARAMETERS
H=30 ft H=60 ft
Site Steel Ratio Steel Ratio
Config. Period DY Dd DU Period DY Dd DU
Class (%) DI (%) DI
(sec.) (in.) (in.) (in.) (sec.) (in.) (in.) (in.)
Long. T rans. Long. T rans.
1.0 0.36 1.22 3.3 6.4 12.3 0.35 1.0 0.0 3.36 12.0 13.1 22.3 0.11
Cantilever Cantilever

B 2.0 0.23 1.04 3.7 5.7 9.5 0.35 2.0 0.0 2.83 13.2 11.7 21.7 -0.18
3.0 0.09 0.90 3.7 4.9 7.0 0.35 3.0 0.0 2.48 13.6 10.7 20.5 -0.42
1.0 1.05 1.23 3.5 10.1 22.5 0.35 1.0 0.44 3.39 12.7 24.7 47.2 0.35
D 2.0 0.84 1.04 4.0 8.6 17.0 0.35 2.0 0.30 2.84 14.0 21.7 35.8 0.35
3.0 0.66 0.91 4.0 7.3 13.3 0.35 3.0 0.20 2.49 14.2 19.5 29.5 0.35
1.0 0.17 0.64 1.8 3.1 5.7 0.35 1.0 0.0 1.72 6.3 8.25 12.7 0.30
Curvature
Double

B 2.0 0 0.55 2.1 2.6 4.6 0.22 2.0 0.0 1.47 7.1 7.3 12.3 0.04
3.0 0 0.48 2.1 2.2 4.5 0.08 3.0 0.0 1.29 7.3 6.6 11.5 -0.15
1.0 0.53 0.64 1.8 4.4 9.1 0.35 1.0 0.48 1.73 6.7 14.3 28.2 0.35
Curvature
Double

D 2.0 0.24 0.55 2.1 3.6 6.4 0.35 2.0 0.30 1.47 7.5 12.1 20.7 0.35
3.0 0.09 0.48 2.1 2.9 4.5 0.35 3.0 0.18 1.29 7.6 10.6 16.4 0.35

Ground Motion Selection

Sixteen out of 24 column bents met the target DI of 0.35 and were selected for nonlinear
dynamic analysis (NDA). Each column was subjected to 25 ground motions. PEER strong ground
motion database (PEER 2011) was used for ground motion selection. Shear wave velocity, distance
to the fault, and spectral acceleration at T=1 sec (S1) were the parameters used in selection of
ground motions. Ground motions recorded in a site with 500 m/s< Vs30 <1500 m/s and 200 m/s<
Vs30 <360 m/s were selected for site class B/D and site class D, respectively. Fifteen out of 25
motions were selected from near-field earthquakes and the rest were selected from far field motions.
The site-to-fault distance smaller than 15 km was used for near-field earthquakes and site-to-fault
distance between 15 and 30 km was used for far-field motions. Furthermore, to limit the scaling
factor, the ground motions were selected so that the scale factor calculated based on S1 was not
greater than 3.

Nonlinear Dynamic Analysis

Each column was analyzed under 25 ground motions including dynamic and material
nonlinearity effects. The ratio of response to design spectral accelerations at the fundamental period
of each column bents was used to scale the ground motions. The peak relative displacement under
each motion was determined and the associated demand damage index (DId) was calculated. Nearly
400 nonlinear dynamic analyses were conducted, and achieved demand damage indices were
plotted in Fig. 3. The figure shows that DId varied from zero to one even though the columns were
designed for DI=0.35. This indicates that scatter in seismic demand is significant, and DId should be
treated as a probabilistic variable. The scatter on DI d was modeled with cumulative lognormal
distribution (fragility curve) and plotted in Fig. 3 along with limit curves. It can be observed that all
data are placed between the limits, and there is a good correlation between the fragility curve and

5
calculated (DId)s. The fragility and limit curves were constructed using the method discussed
previously. Figure 4 illustrates the frequency of (DId)s and shows a good agreement between
lognormal distribution and the actual frequencies.

Figure 3. Fragility curve and calculated demand damage indices


Histogram of DId, Single Column Bents

50
- - - - - Lognormal Dist.
40
Frequency

30

20

10

0
0.00 0.25 0.50 0.75 1.00 1.25 1.50
DId
Figure 4. Frequency of demand damage indices

RELIABILITY ANALYSIS

As discussed, there is significant scatter on the seismic response and demand of column
bents, therefore the reliability of column bents needs to be determined. The reliability of column
bents was quantified by “reliability index” (β) and calculated as follow for lognormal distributions
(Ayyub and McCuen 2003):

6
   2 1 
ln  R L

  L  R2  1 
   (2)
  
ln   R2  1  L2  1 

Where L=load, R=resistance, μ=mean, σ=standard deviation, and δ=coefficient of variation


(σ/μ). The statistical parameters associated with fragility curves for seismic response (Fig. 2) used
for resistance parameters and those obtained from NDA (Fig. 3) used for loading parameters. The
reliability index was calculated for DS-3 to DS-6 and shown in Fig. 5. In this figure, β3 to β6 are
associated with DS-3 to DS-6, respectively. The reliability indices calculated by Eq. 2 were
modified to include the probability of occurrence of design earthquake during the bridge life time
(7% in 75 years). For example, the failure reliability (β6) of column bents designed for DI=0.35 was
calculated as 1.95 using Eq. 2. This reliability is associated with a failure probability of 0.026. The
product of the failure probability and the earthquake probability of occurrence gives the actual
failure probability of 0.026×0.07=0.0018. The actual failure probability of 0.0018 is associated with
a modified β6 of 2.89. The same procedure was applied to modify β3 to β5.
The graphs in Fig. 6 show the effect of longitudinal steel ratio, site class, support condition,
and height-to-diameter ratio (H/D) on reliability indices. The Figs. 6a and 6b indicate that the
reliability indices decrease by increasing longitudinal steel ratio and are independent of site class,
respectively. Figure 6c indicates that cantilever single column bents are more reliable than fixed-
fixed single column bents, and Figure 6d indicates that tall single columns are more reliable than
shorter single column bents.
4
Reliability index, β│DS

2.89
3 2.61
2.25
1.85
2

0
DS 3 DS 4 DS 5 DS 6
Damage State, DS
Figure 5. Reliability indices at different damage states

7
4.00 4.00

Reliability index, β│ DS
Reliability index, β│ DS 1% Long. Reinf. Site B/C
2% Long. Reinf.
3.00 3.00 Site D
3% Long. Reinf.

2.00 2.00

1.00 1.00

0.00 0.00
DS 3 DS 4 DS 5 DS 6 DS 3 DS 4 DS 5 DS 6
Damage State, DS Damage State, DS
(a) (b)

4.00 4.00

Reliability index, β│DS


Reliability index, β│DS

Cantilever H/D = 5
3.00 Fixed-Fixed 3.00 H/D =10

2.00 2.00

1.00
1.00
0.00
0.00 DS 3 DS 4 DS 5 DS 6
DS 3 DS 4 DS 5 DS 6
Damage State, DS
Damage State, DS
(c) (d)

Figure 6. The effect of different parameters on reliability indices at different damage states

CALIBRATION OF DESIGN DAMAGE INDEX

Target DI was tentatively specified as a median of experimental capacity fragility curve for
each damage state and resulted in failure reliability of 2.89. Target DI could be precisely determined
based on a given reliability index. A reliability index of 3.5 is used in AASHTO LRFD (AASHTO
2010) to calibrate gravity load factors. This reliability index was adopted in this study to calibrate
target DI. A direct method was developed in this study for DI calibration. This method is based on
two main assumptions: (1) displacement demand and (2) effective yield displacement are
independent of lateral steel ratio. Having these assumptions, it can be shown:

DI d Du  DY
  const.   (3a)
DI d Du  DY

 L
 (3b)
L

 L
1 (3c)
L

Where DId = demand damage index associated with tentative design damage index of DI;
DI’d = demand damage index associated with calibrated design damage index of DI’; Du and D’u

8
are ultimate displacements associated with DI and DI’, respectively; DY = effective yield
displacement; μ L=mean and δL=coefficient of variation associated with DId; μ’L=mean and
δ’L=coefficient of variation associated with DI’d.
The target reliability of  ′ can be written as follow:

   L2  1 
ln  R 
  L  R2  1 
'   (4)
  
ln   R2  1  L2  1 

Equation (4) can be solved for α and the calibrated damage index of DI’ can be determined
as follow:

DI    DI (5)

It should be noted that this method is applicable only if earthquake probability of occurrence
is excluded from reliability calculation.

Verification Example - A cantilever column was designed for a tentative damage index of 0.35
(DI=0.35) using ARS design spectrum for site class D. The column was analyzed under 25
earthquakes using nonlinear dynamic analysis. The structural reliability for DS-6 was calculated as
2.48 (β6=2.48). The mean value and standard deviation of demand damage indices are 0.301 and
0.177, respectively. Determine the calibrated design damage index (DI′) such that reliability for DS-
6 becomes 3.5 (β′6=3.5).
Substituting the parameters in Eq. (4), α is calculated as 0.57, and using Eq. (5), DI’ is
calculated as 0.20 (DI’=0.57×0.35=0.20).
The column was redesigned for damage index of 0.20 using the design spectrum and
reanalyzed under 25 earthquakes to find the exact reliability indices. The exact reliability indices
and those calculated using Eqs. (4) and (5) are listed in Table 3. This table indicates that design
damage index can be calibrated precisely using direct method.

TABLE 3: VERIFICATION OF DIRECT METHOD


Damage Direct Diff.
Exact
State Approach (%)
DS-3 1.52 1.47 -3.29
DS-4 2.39 2.38 -0.42
DS-5 3.01 3.03 0.66
DS-6 3.46 3.50 1.16
Average -0.47

The direct method was used to calibrate design damage index for a failure reliability of 3.5.
As discussed, the effect of earthquake probability of occurrence must be excluded when the direct
method is used. A failure reliability index of 3.5 is associated with a failure probability of 0.00023.
Excluding the earthquake probability of occurrence, β’6 is calculated as 2.74 which is associated
with the failure probability of 0.003 (0.00023/0.07=0.003). The same approach was used to

9
calculate β’3 to β’5. Using direct method, a single column bent should be design for a damage index
of 0.22 to reach a failure reliability index of 3.5. The other reliability indices associated with DI’ of
0.22 were calculated as of β’3=2.20, β’4=2.73, and β’5=3.17.

CONCLUSIONS

The following conclusions were drawn based on the study presented in this paper:
- There are uncertainties in seismic response of bridges because of scatter on actual
properties and bent configurations.
- There are uncertainties in seismic demand because of scatter in structural properties and
random nature of earthquakes.
- A probabilistic design approach is required to include both uncertainties in seismic
response and seismic demand of bridge column bents. PDCA is a new design method
including all uncertainties.
- The reliability index can be used to quantify the reliability of PDCA.
- Column properties such as longitudinal steel ratio and height-to-diameter ratio have an
effect on the reliability of column bents. Support conditions also can affect on the
reliability of column bents.
- The direct method can precisely calibrate the design damage index for a given reliability
index.

ACKNOWLEDGEMENTS

The research presented herein was sponsored by California Department of Transportation


(Caltrans) under Grant No. 65A0419. Mr. Abbas Tourzani, Dr. Amir Malek, Dr. Mark Mahan,
and Mr. Sam Ataya of Caltrans are thanked for their feedback and interest in different aspects of
the project. Special thanks due to Mr. Peter Lee, the Caltrans Research Program manager, for his
support and advice.

REFERENCES

AASHTO 2010, "AASHTO LRFD Bridge Design Specification," American Association of State Highway and
Transportation Officials, 5th Edition, Washington, D. C.
Ayyub, B. and McCuen, R. H. 2003. “Probability, Statistics, and Reliability for Engineers and Scientists,” CRC
Press LLC, Boca Raton, Florida.
Caltrans 2012, “ARS Online,” http://dap3.dot.ca.gov/shake_stable/v2/index.php, California Department of
Transportation, Sacramento, CA.
Caltrans 2010. “Seismic Design Criteria (SDC),” version 1.6, California Department of Transportation, Sacramento,
CA.
Fernandez, B. & Salas, J.D. 1999. “Return Period and Risk of Hydrologic Events I: Mathematical Formulation,”
Journal of Hydrologic Engineering, October
Massey, F.J. 1951. “The Kolmogorov-Smirnov Test for Goodness of Fit,” Journal of the American Statistical
Association, V. 46, No. 253, March, pp. 68-78.
Naeim, F., Hagie, S., Alimoradi, A., and Miranda, E. (2005). “Automated Post-Earthquake Damage Assessment and
Safety Evaluation of Instrumented Buildings,” Report No. 2005-10639, John A. Martin & Associates, Inc., Los
Angeles, CA.
Nowak, A.S. & Collins, K.R. 2000. “Reliability of Structures,” McGraw-Hills Companies

10
Nowak, A.S. and Zhou, J. H., 1985, "Reliability Models for Bridge Analysis," Report No. UMCE85-9, University of
Michigan, March.
Nowak, A.S. and Grouni, H.N. 1988, "Serviceability Considerations for Guide ways and Bridges," Canadian Journal
of Civil Engineering, Vol. 15, No. 4, August, pp. 534-538.
PEER 2011, “Users Manual for the PEER Ground Motion Database Web Application,” Beta Version, November,
http://peer.berkeley.edu/peer_ground_motion_database/.
Priestley, M. J. N., Calvi G. M., and Kowalsky M. J. 2007, “Displacement-Based Seismic Design of Structures,”
IUSS Press, Pavia, Italy.
Saiidi, M., Vosooghi, A., and Nelson, R.B. 2013. "Shake Table Studies of a Four-Span Reinforced Concrete
Bridge," Journal of Structural Engineering, ASCE, Special Issue on NEES: Advances in Earthquake
Engineering, In press, http://ascelibrary.org/doi/abs/10.1061/%28ASCE%29ST.1943-541X.0000790.
Suarez, V.A. and Kowalsky, M.J. 2010 “Direct Displacement Based Design as an Alternative Method for Seismic
Design of Bridges,” ACI Structural Journal, SP-271-4, May, pp. 63-78
Tourzani, A.M, Malek, A.M, Ataya, S, and Mahan, M 2008, “Probabilistic Damage Control Approach (PDCA) and
Performance–Based Design of Bridges,” Sixth National Seismic Conference on Bridges and Highways: Seismic
Technologies for Extreme Loads , Multidisciplinary Center for Earthquake Engineering Research .
Vosooghi, A. & Saiidi, M. 2010. "Seismic Damage States and Response Parameters for Bridge Columns," ACI
Special Publications, SP271-02, Structural Concrete in Performance-Based Seismic Design of Bridges, V. 271,
May 24, pp. 29-46.
Vosooghi, A. & Saiidi, M. 2012. "Experimental Fragility Curves for Seismic Response of Reinforced Concrete
Bridge Columns," ACI Structural Journal, V. 109, No. 6, November-December, pp. 825-834.
Yen, B. C. 1970. ‘‘Risks in Hydrologic Design of Engineering Projects,’’ Journal Hydrologic Division, ASCE,
96(4), 959–966.

11

You might also like