Seismic Behaviour and Analysis of Continuous Reinforced Concrete Bridges

Download as pdf or txt
Download as pdf or txt
You are on page 1of 356

Seismic Behaviour and Analysis of Continuous

Reinforced Concrete Bridges


by
Payam Tehrani

Supervisor:
Professor Denis Mitchell

Department of Civil Engineering and Applied Mechanics,


McGill University
Montreal, Quebec, Canada
April 2012

A manuscript-based thesis submitted to McGill University in partial


fulfillment of the requirements of the degree of Doctor of Philosophy

Payam Tehrani, 2012

Abstract
This study focuses on the seismic analysis and behaviour of continuous 4-span bridges.
Different methods of analyses including linear multi-mode analysis, inelastic time history
analysis and incremental dynamic analysis (IDA) are used for the seismic evaluations of
bridges in this study.
This thesis includes two main parts. In the first part the seismic behaviour of bridges with
different column heights (i.e., irregularity due to different column stiffnesses) is studied.
The seismic evaluations are carried out in the transverse and longitudinal directions of
bridges to recognize the important aspects which influence the seismic behaviour.
Parametric studies were carried out for a number of bridges in the transverse and the
longitudinal directions. To perform a large number of designs and analyses, a computer
program was developed to design the bridges, perform the modelling and extract and
evaluate the analysis results. The effects of different column heights, different column
diameters, different superstructure mass and stiffness, as well as different abutment
conditions on the seismic response of bridges were studied using elastic and inelastic
analyses. The results from the elastic and inelastic analyses were compared to
demonstrate the limitations of the linear analyses for the seismic design and evaluation of
irregular bridges. The effects of including nonlinear abutment models with different
stiffness and strengths were also studied in the longitudinal response of the bridges.
Seismic ductility demands and concentration of ductility demands were evaluated and the
maximum demand to capacity ratios were predicted for a wide range of bridges studied.

II

The use of different regularity indices to predict the seismic response of bridges was also
investigated.
In the second part of the thesis, the use of incremental dynamic analysis for seismic
evaluation of bridges is studied. The influence of different record selection
methodologies including the UHS-based, CMS-based and epsilon-based methods on the
predictions of the IDA results is investigated. In addition, the effects of different
earthquake types including crustal, subduction interface and subduction inslab
earthquakes on the IDA results are studied.
Three large record sets were selected for three earthquake types and a fast algorithm was
developed for the incremental dynamic analysis to evaluate the collapse capacity of
different bridge configurations subjected to different earthquake types. The IDA results
were also predicted for different subsets of records with specific characteristics. The
effects of spectral shapes and epsilon values were also considered using seismic hazard
deaggregation results.

III

Rsum
Cette tude se concentre sur lanalyse sismique et sur le comportement des ponts 4
portes continues. Diffrentes mthodes danalyse, telles que la mthode multimode
linaire, la mthode temporelle non linaire et la mthode danalyse dynamique
incrmentale (ADI), sont utilises pour lvaluation sismique de ponts.
Cette thse se divise en deux parties principales. Dans la premire partie, le
comportement sismique des ponts composs de colonnes de diffrentes hauteurs (cest-dire, irrgularit cause par diffrentes raideurs de colonne) est tudi. Les valuations
sismiques sont ralises dans les directions transversal et longitudinal des ponts afin de
considrer les aspects importants qui influencent le comportement sismique. Des tudes
paramtriques furent ralises pour un certain nombre de ponts (cest--dire 648 ponts
dans la direction transversale et plus de 2500 cas dans la direction longitudinale). Afin
deffectuer un grand nombre de dimensionnements et danalyses, un programme
informatique fut dvelopp pour dimensionner des ponts, effectuer la modlisation et
extraire et valuer les rsultats danalyse. Les effets de diffrentes hauteurs de colonne,
de diffrents diamtres de colonne, de diffrentes masses et raideurs de la superstructure,
et de diffrentes conditions de bute sur la rponse sismique des ponts furent tudis en
utilisant des analyses lastiques et inlastiques. Les rsultats des analyses lastiques et
inlastiques furent compars afin de dmontrer les limitations des analyses linaires pour
le dimensionnement et lvaluation sismique des ponts irrguliers. Les effets sur la
rponse longitudinale des ponts de modles de bute non linaires considrant diffrentes
rsistances et raideurs (incluant diffrentes longueurs despace de joint et diffrents
nombres de piles) furent galement tudis. Les demandes sismiques en ductilit et la
concentration des demandes en ductilit furent values et la demande maximale des
ratios en capacit fut prdite pour un large ventail de ponts tudis. Lutilisation de
diffrents indices de rgularit pour prdire la rponse sismique des ponts fut aussi
examine.
Dans la seconde partie de la thse, lutilisation de lADI pour lvaluation sismique des
ponts est tudie. Linfluence de diffrentes mthodes de slection denregistrement
(incluant les mthodes bases sur lala sismique, le spectre moyen conditionnel et

IV

lepsilon) sur les prdictions obtenues avec lADI est examine. De plus, leffet de
considrer diffrents types de tremblement de terre (incluant des tremblements de terre de
surface et de subduction) sur les rsultats de lADI est tudi. Prsentement, seulement
les tremblements de terre de surface sont utiliss pour lvaluation de la performance
sismique des structures. Les procdures actuelles ne sont pas ncessairement appropries
pour les rgions soumises des tremblements de terre de subduction. Trois ensembles
denregistrement furent slectionns pour trois types de tremblement de terre (cest--dire
un total de 3 x 78 = 234 enregistrements). Un algorithme calcul rapide fut dvelopp
pour lADI afin dvaluer la capacit leffondrement de diffrentes configurations de
pont soumises diffrents types de tremblement de terre. Les rsultats de lADI furent
galement prdits pour diffrents sous-ensembles denregistrements ayant des
caractristiques spcifiques (cest--dire des valeurs depsilon positives, des faibles
facteurs dchelle, etc.). Les effets des spectres de rponse et des valeurs depsilon furent
aussi considrs en utilisant les rsultats de dsagrgation du risque sismique.

Preface
I would like to express my deep and sincere gratitude to my supervisor, Professor Denis
Mitchell, for his guidance and patience during my PhD study and for providing me with
freedom on my research. His nice personality and his friendly and inspiring treatment
during my study are greatly appreciated.

The financial support provided by the Natural Sciences and Engineering Research
Council of Canada (NSERC) for the Canadian Seismic Research Network is gratefully
acknowledged. The International Doctoral Award (MIDAs) provided by the McGill
University is also appreciated.

I would like to thank Dr. William Cook for installing the computer equipments and the
remote control connections which expedited performing the large volume of the computer
analyses in this research.

I also thank Dr. Alejandro de la Puente for providing his program Ruaumoko helper
which inspired the development of the computer programs in this research.

I would like to also thank the anonymous reviewers of the manuscripts for their
comments which helped the improvement the papers.

I would like to thank the European Laboratory for Structural Assessment (ELSA) for
providing the test data which was used to validate the numerical models, as presented in
Appendix A of this thesis.

I would like to thank the following professors/scholars for their responses to my short
questions on different subjects, through emails, and/or for providing me with the
requested papers and references: Jack Baker, Matej Fischinger, Nigel Priestley, Athol

VI

Carr, Dimitrios Vamvatsikos, Yuk Hon Chai, Amr Elnashai, Kazuhiko Kawashima, Kent
A. Fogleman , Saiid Saiidi and Murat Saatcioglu.

I also thank the entire Staff of the Department of Civil Engineering and Applied
Mechanics, who helped solving those little problems that students have.

I would like to thank my friends Ali Ghafari Oskoei and Hossein Azimi who helped me
the most during my first days in Canada and also thank all of my office colleagues and
friends, especially Hamed Layssi and Farshad Habibi for making my Ph.D. study an
enjoyable experience. I also thank my friend Iman Ghorbani for the mutual scientific
discussions we had together.

Last, but certainly not least, I would like to express my special gratitude to my family in
Iran, especially my beloved spouse Solmaz and my parents for their endless patience,
love, encouragement and support during my studies.

Contributions of Authors

I would like to express my sincere thanks to Prof. Katsu Goda for providing the raw
seismic hazard data for Vancouver and Montreal and information for the Japanese
earthquakes which was used in this research (in Chapters 5 and 6). His support on the
seismological subjects are also gratefully acknowledged. I would like to also thank Prof.
Gail Atkinson and Prof. Luc Chouinard for reviewing the manuscript in Chapter 5 and
providing me with constructive comments.

VII

Dedicated to my parents, dear Hossein and Shohreh, to my wife, Solmaz and


to my brothers, Pouya and Peyman.

VIII

Table of Contents
Abstract ............................................................................................................................... II
Rsum .............................................................................................................................. IV
Preface............................................................................................................................... VI
List of Tables ...................................................................................................................XV
List of Figures ................................................................................................................ XIX
1

Introduction and Literature Review ............................................................................ 1


1.1

Introduction .......................................................................................................... 1

1.2

Research objectives .............................................................................................. 4

1.3

Thesis organization .............................................................................................. 4

1.4

Background and literature review ........................................................................ 8

1.4.1

Bridges with column stiffness irregularity .................................................... 8

1.4.2

Modal pushover analysis (MPA) ................................................................ 12

1.4.3

Analysis of bridges with different column stiffnesses ................................ 13

1.5

Code provisions .................................................................................................. 17

1.5.1

Caltrans (2006) and AASHTO Guidelines (2009)...................................... 17

1.5.2

Eurocode ..................................................................................................... 18

1.5.3

CSA-S6-06 and AASHTO-04..................................................................... 19

1.5.4

Performance objectives for bridges............................................................. 20

1.6

Problems associated with the force-based design approach............................... 21

1.6.1

Interdependency of strength and stiffness and relationship between Strength

and ductility demand ................................................................................................. 22


1.6.2

Structures with unequal column heights ..................................................... 23

IX

1.6.3
1.7
2

Structures with dual (Elastic and Inelastic) load paths ............................... 25

Introduction to direct displacement-based design method ................................. 25

Incremental dynamic analysis applied to seismic performance and risk assessment of

RC bridges ........................................................................................................................ 28


2.1

Introduction ........................................................................................................ 28

2.2

Incremental dynamic analysis ............................................................................ 29

2.2.1

Introduction ................................................................................................. 29

2.2.2

Damage measures and intensity measures .................................................. 30

2.2.3

IDA Curves ................................................................................................. 30

2.2.4

Summarizing IDA results ........................................................................... 33

2.2.5

Characteristics of median IDA curves ........................................................ 35

2.2.6

Defining limit-States on IDA curves .......................................................... 36

2.3

Different IDA procedures and algorithms .......................................................... 39

2.3.1

Regular IDA algorithms .............................................................................. 39

2.3.2

Time-efficient algorithms ........................................................................... 40

2.3.3

ATC-63 procedure ...................................................................................... 40

2.3.4

Fast IDA procedures ................................................................................... 41

2.4

Structural modelling for IDA ............................................................................. 42

2.5

Scaling of records in IDA................................................................................... 46

2.6

Prediction of damage states ................................................................................ 47

2.6.1

Study by Dutta and Mander (1998) ............................................................ 48

2.6.2

Theoretical method by Priestley et al. (2007) ............................................. 49

2.6.3

Equations by Berry and Eberhard (2007) ................................................... 51

2.6.4

Study by Fardis and Biskinis (2003) ........................................................... 53

2.6.5

Study by Haselton et al. (2007)................................................................... 53

2.7

Treatment of uncertainties in IDA ..................................................................... 54

2.7.1

Types of variability ..................................................................................... 54

2.7.2

Sources of uncertainty and combining uncertainties .................................. 55

2.8

Development of fragility curves using IDA results............................................ 57

2.9

Seismic risk assessment ..................................................................................... 60

2.10

Framework for performance-based earthquake engineering (PBEE) ............. 64

2.11

Ground motion prediction equations .............................................................. 67

2.12

Record selection for IDA ................................................................................ 69

2.12.1 Spectral shapes of the records and epsilon values ...................................... 69
2.12.2 Record selection strategies .......................................................................... 73
2.12.3 Conditional mean spectrum (CMS) ............................................................ 76
2.12.4 Simplified method to include the spectral shape effects ............................. 78
2.13
3

Evaluation of the seismic performance .......................................................... 79

Effects of Column and Superstructure Stiffness on the Seismic Response of Bridges

in the Transverse Direction ............................................................................................... 82


3.1

Preface ................................................................................................................ 82

3.2

Abstract .............................................................................................................. 84

3.3

Introduction ........................................................................................................ 84

3.4

Regularity and irregularity indices ..................................................................... 86

3.5

Code Provisions.................................................................................................. 87

3.5.1

Caltrans (2006) and AASHTO Guidelines (2009)...................................... 87

3.5.2

Eurocode ..................................................................................................... 88

3.5.3

CSA-S6-06 and AASHTO-04..................................................................... 89

3.6

Methods to improve the seismic behaviour of irregular bridges ........................ 90

3.7

Modelling and evaluation of the bridges ............................................................ 91

XI

3.7.1

Limit states and corresponding strain limits ............................................... 91

3.7.2

Calculating ductility capacities and modelling assumptions ...................... 93

3.7.3

Range of parameters studied ....................................................................... 96

3.8

3.8.1

Effect of column stiffness ratios ................................................................. 99

3.8.2

Effect of superstructure to substructure stiffness ratio.............................. 102

3.8.3

Prediction of regularity indices ................................................................. 104

3.9

4

Predicted response of bridges with restrained abutments .................................. 99

Predictions of responses of Bridges with unrestrained abutment conditions ... 107

3.9.1

Effect of maximum column stiffness ratios .............................................. 107

3.9.2

Effect of superstructure to substructure stiffness ratios ............................ 108

3.10

Displacements from elastic versus inelastic analysis ................................... 109

3.11

Demand to capacity ratios ............................................................................ 110

3.12

Summary and Conclusions ........................................................................... 111

Effects of Column Stiffness Irregularity on the Seismic Response of Bridges in the

Longitudinal Direction .................................................................................................... 115


4.1

Preface .............................................................................................................. 115

4.2

Abstract ............................................................................................................ 117

4.3

Introduction ...................................................................................................... 117

4.4

Modelling and analysis of the bridges.............................................................. 118

4.5

Range of parameters studied ............................................................................ 119

4.6

Modelling the abutments .................................................................................. 123

4.7

Evaluation of maximum ductility demands...................................................... 127

4.7.1

P-Delta effects ........................................................................................... 131

4.8

Predictions from inelastic versus elastic analysis............................................. 131

4.9

Concentration of seismic demands ................................................................... 133

XII

4.10

Drift ratios..................................................................................................... 136

4.11

Ductility demand versus ductility capacity .................................................. 138

4.12

Normalized ductility demand ....................................................................... 140

4.13

Demand to capacity ratios considering transverse and longitudinal responses


141

4.14
5

Conclusions .................................................................................................. 143

Effects of Different Record Selection Methods and Earthquake Types on the

Transverse Response of Bridges ..................................................................................... 146


5.1

Preface .............................................................................................................. 146

5.2

Abstract ............................................................................................................ 149

5.3

Introduction ...................................................................................................... 149

5.4

Seismic hazard analysis and conditional mean spectrum................................. 151

5.4.1

Conditional mean spectrum (CMS) .......................................................... 151

5.4.2

Preliminary record selection ..................................................................... 152

5.4.3

Seismic hazard analysis and target scenarios............................................ 153

5.5

Bridge properties, design and modelling assumptions ..................................... 157

5.5.1

Bridge properties ....................................................................................... 157

5.5.2

Incremental dynamic analysis ................................................................... 158

5.5.3

Damage states and performance indicators............................................... 160

5.5.4

Bridge modelling ...................................................................................... 162

5.5.5

Collapse modes and prediction of probability of collapse ........................ 163

5.6

Seismic evaluation using different record selection methods .......................... 165

5.6.1

Seismic evaluation using crustal records .................................................. 166

5.6.2

Seismic evaluation considering all event types......................................... 172

5.6.3

Sensitivity of the results to the number of records ................................... 179

5.7

Summary and conclusions ................................................................................ 180


XIII

6

Seismic Response of Bridges Subjected to Different Earthquake Types Using IDA


183

7

6.1

Preface .............................................................................................................. 183

6.2

Abstract ............................................................................................................ 185

6.3

Introduction ...................................................................................................... 185

6.4

Epsilon-based method and spectral shape issue ............................................... 186

6.5

Fast-IDA analysis ............................................................................................. 189

6.6

Record selection for IDA ................................................................................. 191

6.7

Bridge properties .............................................................................................. 193

6.8

Modelling of the bridges for IDA .................................................................... 198

6.9

Seismic hazard analysis .................................................................................... 199

6.10

Results obtained using different record sets and subsets .............................. 200

6.11

Combining the results from different earthquake types ............................... 207

6.12

1 and SSF values using different sets and subsets of records ..................... 208

6.13

Effect of number of records (44 records vs 78 records) ............................... 211

6.14

Comparison of the bridge responses for different configurations ................ 213

6.15

Comparing design using R/I=3.33 and R/I=5 ............................................... 215

6.16

Conclusions .................................................................................................. 216

Conclusions ............................................................................................................. 219


7.1

Summary and main conclusions ....................................................................... 219

7.1.1

Conclusions regarding seismic response and evaluation of bridges with

column stiffness irregularities ................................................................................. 219


7.1.2

Conclusions regarding the use of incremental dynamic analysis (IDA) for

seismic analysis and evaluation of bridges ............................................................. 222


7.2

Future research ................................................................................................. 224

7.3

Statement of original contributions .................................................................. 226

XIV

Appendix A: Verification of the models, preliminary studies and other modelling details
......................................................................................................................................... 228
Appendix B: Computer program for designing, modelling, running the analyses,
extracting and post processing the results ....................................................................... 252
Appendix C: Probabilistic seismic hazard analysis and the program developed for seismic
deaggregation .................................................................................................................. 259
Appendix D: Computer program for different record selection methods and predictions of
CMS ................................................................................................................................ 273
Appendix E: Dynaplot helper ......................................................................................... 277
Appendix F: Different earthquake types ......................................................................... 278
Appendix G: Detailed information of the ground motion records used ......................... 287
Appendix H: Pushover analysis of the bridge in Chapter 5 ............................................ 296
Appendix I: Some of the results from Chapter 6 ............................................................ 298
References ....................................................................................................................... 317

XV

List of Tables
Table 1.1. Proposed seismic design performance criteria by the Seismic Subcommittee of
the CHBDC ....................................................................................................................... 21
Table 2.1. The recommended drift limits by Dutta and Mander (1998) at different damage
states .................................................................................................................................. 49
Table 2.2. Maximum strain in concrete and steel at different damage states (Priestley et
al., 2007) ........................................................................................................................... 51
Table 2.3. Summary of the equations by Berry and Eberhard (2007) for cover-spalling,
bar buckling, and bar failure, and equation by Mackie and Stojadinovic (2007) to estimate
drift at failure .................................................................................................................... 52
Table 2.4. Loss ratio for different damage states (Dhakal and Mander, 2006) ................ 65
Table 2.5. Summary of the adopted ground motion prediction equations (GMPE) (Table
adapted from Atkinson and Goda, 2011) .......................................................................... 69
Table 3.1. Deformations at different damage states for a column with D=1.5m, H=5m,
s=2.0%, and v =1.2% a) Using moment-curvature analyses (Priestley et al., 2007); b)
Using experimentally derived equations (Berry and Eberhard, 2007) ............................. 93
Table 5.1. Seismic deaggregation results for T= 0.7 sec (epsilon values using the BA08
and Z06 GMPEs) ............................................................................................................ 154
Table 5.2. Seismic deaggregation results for T= 1.3 sec (epsilon values using the BA08
and Z06 GMPEs) ............................................................................................................ 155
Table 5.3. Deformations at different damage states for the central column of the bridge:
a) Using the theoretical approach by Priestley et al., (2007), b) Using the experimental
equations by Berry and Eberhard (2007). ...................................................................... 162
Table 5.4. Results of IDA using a 3D analysis for different abutment stiffness and
strength using 22 pairs of records applied twice at different principal directions (44 cases)
......................................................................................................................................... 171
Table 5.5. IDA results (CMS-All-based method, T1=0.7) for a) Z06 GMPE................. 175
Table 5.6. Epsilon-based method for probability of exceedance in 50 years of a) 2% and
b) 0.5% probability of exceedance in 50 years ............................................................... 175

XVI

Table 5.7. IDA results at T1=1.3 sec using 44 records for a) CMS-All-based method
(CMR=2.2) and b) CMS-Event-based method (CMR=2.14) ......................................... 178
Table 5.8. IDA results using 78 fixed records for each event type for a)T=0.7 sec and
b)T=1.3 sec ..................................................................................................................... 180
Table 6.1. Criteria used for record selection for different earthquake types .................. 193
Table 6.2. Summary of bridges with different configurations studied (a total of 30
bridges) ........................................................................................................................... 195
Table 6.3. Periods (sec) and (%M at the fundamental period) for different bridge
configurations studied in the case of a) R/I=3.33 and b) R/I=5 ...................................... 196
Table 6.4. Percentage of longitudinal reinforcement in the bridge columns C1,C2 and C3
for different bridge configurations .................................................................................. 196
Table 6.5. The deaggregation results for mean epsilon and percentage of contribution of
different event types at 0.5% probability of exceedance in 50 years (using BA08 and Z06
GMPE). ........................................................................................................................... 200
Table 6.6. Different record subsets considered to compute the predicted median collapse
capacity ........................................................................................................................... 202
Table 6.7. Comparison of the median collapse capacity (g) and (CMR/ CMRacceptable)
ratios using different subsets for configuration 1with R/I = 3.33 and a) D = 1.5 m and b)
D = 2.5 m ........................................................................................................................ 203
Table 6.8. Comparison of the median collapse capacity (g) and (CMR/ CMRacceptable)
ratios using different subsets and methods for configuration 2 (R/I = 3.3) and a) D = 1.5
m and b) D = 2.5 m ......................................................................................................... 203
Table 6.9. Comparison of the median collapse capacity (g) and (CMR/ CMRacceptable)
ratios using different subsets and methods for configuration 3 (R/I = 3.3) and a) D = 1.5
m and b) D = 2.5 m ......................................................................................................... 204
Table 6.10. Comparison of the median collapse capacity (g) and (CMR/ CMRacceptable)
ratios using different subsets and methods for configuration 4 (R/I = 3.3) and a) D = 1.5
m and b) D = 2.5 m ......................................................................................................... 204
Table 6.11. Comparison of the median collapse capacity (g) and (CMR/ CMRacceptable)
ratios using different subsets and methods for configuration 5 (R/I = 3.3) and a) D = 1.5
m and b) D = 2.5 m ......................................................................................................... 205

XVII

Table 6.12. An example of Statistics of unmodified collapse capacity predictions for


different earthquake types and for different subsets of records in each
subset(Configuration 4, D = 2.5 m, and R/I = 3.33). ...................................................... 210
Table 6.13. Comparison of the modified median collapse capacities obtained using 44
records and 78 records (in the parentheses) for different earthquake types and combined
earthquake types.............................................................................................................. 212
Table 6.14. Modified and unmodified median collapse capacities and corresponding
collapse margin ratios (CMR) and (CMR /CMRacceptable) using subset 3 of records. (D
= 2 m, and R/I = 3.33) ..................................................................................................... 213
Table 6.15. Modified and unmodified median collapse capacities and corresponding
CMR and CMR /CMRacceptable using subset 3 of records. (D = 2.5 m, and R/I = 5) . 214
Table 6.16. Modified median collapse capacities and corresponding collapse margin
ratios (CMR) using subset 3 of records for R/I=3.33 and R/I=5. (D = 2.0 m) ............... 215

XVIII

List of Figures
Fig. 1.1.An example of the concentration of the ductility demands in the stiffest column
(bridge at the bottom). The results obtained using a pushover analysis in SAP2000. The
columns in the regular bridge (bridge at the top) have uniform ductility demands. ........... 2
Fig. 1.2. Influence of strength on moment-curvature diagram: a) design assumption
(constant stiffness); b) realistic condition (constant yield curvature) (adapted from
Priestley et al., (2007)) ...................................................................................................... 22
Fig. 2.1. An example of an IDA curve developed using Sa(T1) as IM and maximum drift
ratio as DM. The results of nonlinear dynamic analysis for each IM are shown by dots
and linear interpolation is used to fit a curve though these points. The last point of the
IDA curve corresponds to an infinite drift ratio (i.e., dynamic instability). ..................... 32
Fig. 2.2. An example of IDA curves developed using 44 crustal records for a 4-span
bridge (percentile 16%, 50% and 84% IDA curves are shown by heavy lines) (Adapted
from Tehrani et al. (2012))................................................................................................ 34
Fig. 2.3. Characteristics of median IDA curves (adapted from ATC-62, 2008) ............... 36
Fig. 2.4. Prediction of different limit states on IDA curves .............................................. 38
Fig. 2.5. Backbone curve parameters (adapted from ATC-63 (2008)) ............................. 43
Fig. 2.6. Differences between cyclic and in-cycle strength degradation (from FEMA
P440A (ATC-62), 2008). .................................................................................................. 43
Fig. 2.7. Development of fragility curves for different limit states using IDA results ..... 58
Fig. 2.8. Effect of uncertainty on the shape of fragility curves......................................... 59
Fig. 2.9. Hazard curve for Vancouver at T=1 sec. using the updated seismic hazard data
by Goda et al. (2010) ........................................................................................................ 60
Fig. 2.10. An example of  values computed for the case of Imperial Valley (1979) record
using BA08 GMPE ........................................................................................................... 71
Fig. 2.11. The normalized average response spectra of crustal ground motion records for
different  values at a) T=0.3 sec and b) T=1 sec. ............................................................ 72
Fig. 2.12. The normalized average response spectra of interface ground motion records
for different  values at a) T=0.3 sec and b) T=1 sec........................................................ 72

XIX

Fig. 2.13. The normalized average response spectra of inslab ground motion records for
different  values at a) T=0.3 sec and b) T=1 sec. ............................................................ 72
Fig. 2.14. An example of the regression analysis to determine the 1factor..................... 79
Fig. 3.1. Modified Takeda hysteresis loop (adapted from Carr (2009)) ........................... 96
Fig. 3.2. Bridge configurations studied with different column heights ............................ 97
Fig. 3.3. Average response spectrum of 7 records used for inelastic time history analysis
matching the 2010 NBCC spectrum (2% in 50 years) ...................................................... 98
Fig. 3.4. Effect of column stiffness ratio on maximum ductility demand for bridges with
various configurations ..................................................................................................... 100
Fig. 3.5. Effect of column stiffness ratio on maximum to minimum ductility ratio ....... 100
Fig. 3.6. Effect of column stiffness ratio on the ratio of the displacements obtained using
elastic and inelastic analysis ........................................................................................... 100
Fig. 3.7. Comparison of the displacement envelopes obtained using elastic and inelastic
analysis for a) a regular bridge and b) an irregular bridge .............................................. 102
Fig. 3.8. Effect of superstructure to substructure stiffness on a) maximum ductility
demand and b) maximum to minimum ductility demand ............................................... 103
Fig. 3.9. Effect of superstructure/substructure stiffness on the results from elastic and
inelastic analysis ............................................................................................................. 104
Fig. 3.10. Different regularity indices versus maximum ductility demand .................... 105
Fig. 3.11. Different regularity indices versus maximum to minimum ductility demand
ratio ................................................................................................................................. 105
Fig. 3.12. Different regularity indices versus the ratio of the displacements obtained using
inelastic and elastic analysis ........................................................................................... 106
Fig. 3.13. Influence of column stiffness ratios on a) maximum to minimum ductility
demands and b) results obtained from elastic and inelastic analyses (Bridges with
unrestrained abutment conditions) .................................................................................. 108
Fig. 3.14. Effect of supersructure to substructure stiffness ratio on a) maximum ductility
demands, b) maximum to minimum ductility demands ratio, and c) results obtained from
elastic and inelastic analyses ........................................................................................... 109
Fig. 3.15. The maximum capacity to demand ratios for a) restrained abutments and b)
unrestrained abutments ................................................................................................... 111

XX

Fig. 4.1. Modified Takeda hysteresis loop (adapted from Carr (2009)) ......................... 119
Fig. 4.2. Bridge properties .............................................................................................. 120
Fig. 4.3. Structural modelling of the bridges .................................................................. 121
Fig. 4.4. Average response spectrum for 7 records used for inelastic time history analysis
matching the 2010 NBCC spectrum (2% in 50 years) .................................................... 122
Fig. 4.5. Schematic view of the seat-type abutment and its components ....................... 123
Fig. 4.6. Simplified abutment model for the longitudinal response................................ 124
Fig. 4.7. a) Hysteresis model with gap and nonlinear spring used to model abutment
response (Carr, 2009) b) typical nonlinear response from analysis ................................ 125
Fig. 4.8. Influence of abutment stiffness: a) effective abutment stiffness (Caltrans, 2006);
b) the influence of the ratio of the total stiffness of columns to effective abutment
stiffness on the maximum ductility demands.................................................................. 127
Fig. 4.9. Effects of total stiffness of columns on the maximum displacement ductility
demands: a) R=3, =0.5 and =0; b) R=5, =0.5 and =0 ............................................. 128
Fig. 4.10. Effects of maximum column stiffness ratio on the maximum displacement
ductility demands: a) R=3, =0.5 and =0; b) R=5, =0.5 and =0. .............................. 129
Fig. 4.11. Effects of columns total stiffness times max stiffness ratio on the maximum
displacement ductility demands: a) R=3, =0.5 and =0; b) R=5, =0.5 and =0......... 129
Fig. 4.12. Effects of columns total stiffness times max stiffness ratio on the maximum
displacement ductility demands considering the influence of abutments (Gap=50 mm,
R=5, =0.5 and =0) ....................................................................................................... 129
Fig. 4.13. Effects of hysteresis parameters,  and , on the maximum displacement
ductility demands for R=5 and: a) =0 and =0.6; b) =0.3 and =0.3 ......................... 130
Fig. 4.14. Possible instability due to P-Delta effects for bridges that were designed
neglecting P-Delta effects. .............................................................................................. 131
Fig. 4.15. Effects of total stiffness of columns on the ratio of the displacements obtained
using inelastic and elastic analysis for R=5 and: a) =0 and =0.6; b) =0.5 and =0; c)
=0.3 and =0.3; d) considering the influence of the abutments on seismic response (no
piles, gap=50 mm, =0.5 and =0) ................................................................................. 132
Fig. 4.16. Effects of maximum column stiffness ratio on the Max/Min ductility ratio
(R=5, =0.5 and =0) and predictions using Eq.[4.11]. ................................................. 134

XXI

Fig. 4.17. Maximum drift ratio of columns for: a) R=5 and Takeda hysteresis model with
=0.5 and =0; b) R=3 and Takeda hysteresis model with =0.5 and =0 .................... 136
Fig. 4.18. Maximum drift ratio of columns considering the influence of the abutments in
nonlinear response (no piles, gap=50 mm, R=5, =0.5 and =0) ................................... 137
Fig. 4.19. Maximum drift ratio versus maximum ductility demand using R=5, =0.5 and
2

=0 for: a) maximum drift ratio versus maximum ductility demand; b) ( D) / H versus
maximum ductility demand ............................................................................................ 137
Fig. 4.20. Maximum ductility demand to ductility capacity ratios obtained for: a) R=5,
=0.5 and =0; b) R=5, =0.3 and =0.3; c) R=3, =0.5 and =0; d) R=3, =0.3 and
=0.3 ............................................................................................................................... 139
Fig. 4.21. Effects of increasing the superstructure mass to 300 KN/m on the maximum
demand to capacity ratios obtained for: a) R=5, =0.5 and =0; b) R=3, =0.5 and =0
......................................................................................................................................... 140
Fig. 4.22. Maximum ductility demand to ductility capacity ratios obtained when abutment
effects were considered in modelling (Gap=50 mm, R=5, =0.5 and =0) ................... 140
Fig. 4.23. Analysis results using R=5 and Takeda hysteresis model with =0.5 and =0
for: a) normalized ductility demands; b) normalized demand to capacity ratios ............ 141
Fig. 4.24. Ductility demand to ductility capacity ratios for different bridge configurations
considering transverse and longitudinal responses based on the 100%/30% rule for: a)
Bridges with restrained transverse movements; b) Bridges with unrestrained transverse
movements ...................................................................................................................... 142
Fig. 5.1. PSHA results for Vancouver: a) Seismic hazard deaggregation for 2%
probability of exceedance in 50 years at T=0.7 sec b) Uniform hazard spectra ............. 155
Fig. 5.2. CMS for T=0.7 sec: a) for BA08 and Z06 GMPEs, b) for BA08 and AB03
GMPEs ............................................................................................................................ 156
Fig. 5.3. CMS for T=1.3 sec: a) for BA08 and Z06 GMPEs, b) for BA08 and AB03
GMPEs ............................................................................................................................ 156
Fig. 5.4. Bridge properties .............................................................................................. 158
Fig. 5.5. Backbone curve parameters .............................................................................. 160

XXII

Fig. 5.6. Effects of hysteresis parameters on the collapse capacity: a) Modified Takeda
hysteresis loops; b) The collapse capacity predictions using the ATC-63 record set for
different Takeda hysteresis loop parameters (alpha and beta). ....................................... 163
Fig. 5.7. Normalized average spectra from different methods at a) T=0.7 sec, b) T=1.3 sec
......................................................................................................................................... 165
Fig. 5.8. IDA results using the epsilon-based method (Crustal event,T1=0.7 sec,
CMR=3.5) ....................................................................................................................... 167
Fig. 5.9. IDA results using the ATC-63 record set at T1=0.7 sec (CMR=2.531.26=3.2)
......................................................................................................................................... 167
Fig. 5.10. IDA results using the epsilon-based record selection method at T1=1.3 sec
(CMR=3.8) ...................................................................................................................... 168
Fig. 5.11. IDA results using the ATC-63 record setat T1=1.3 sec (CMR= 1.36 * 2.8 = 3.8)
......................................................................................................................................... 169
Fig. 5.12. IDA results in the longitudinal direction using 44 crustal records ................. 170
Fig. 5.13. IDA results (CMS-Event based method, T1=0.7) for Z06 GMPE (CMR=2.7)
......................................................................................................................................... 174
Fig. 5.14. IDA results ( CMS-Event-based method, T1=0.7) for AB03 GMPE (CMR=2.8)
......................................................................................................................................... 174
Fig. 5.15. IDA results (CMS-Event based method, T1=0.7) for Z06 GMPE using 78
records ............................................................................................................................. 175
Fig. 5.16. IDA results for Epsilon-based method using 78 records (CMR=3.5) ............ 176
Fig. 5.17. IDA results based on the UHS-based record selection at T1=0.7 sec
(CMR=2.0) ...................................................................................................................... 176
Fig. 5.18. IDA results using the UHS-based record selection method at T1=1.3 sec
(CMR=1.8) ...................................................................................................................... 177
Fig. 5.19. IDA results using epsilon-based record selection method at T1=1.3 sec
(CMR=3.8) ...................................................................................................................... 177
Fig. 5.20. IDA results for Epsilon-based method using 78 records (CMR=2.62) .......... 178
Fig. 5.21. IDA results using CMS-Event-based method by using 78 records (CMR=2.02)
......................................................................................................................................... 179
Fig. 6.1. Bridge configurations studied ........................................................................... 197

XXIII

Fig. 6.2. Backbone curve parameters (adapted from ATC-63(2008)) ............................ 198
Fig. 6.3. IDA results: a) full IDA curves for the bridge with configuration 1, D=1.5 m and
R/I=3.3 using 78 crustal records (unmodified results) b) statistics of the IDA results for
different damage states.................................................................................................... 201
Fig. 6.4. Regression analysis to compute 1 for the case of configuration 1 with D = 2 m
using BA08 GMPE ......................................................................................................... 209
Fig. 6.5. Regression analysis for 1and period (T1) considering all records with removed
outliers (i.e., subset 3) for a) crustal records, b) interface records and c) inslab records 211

XXIV

1 Introduction and Literature Review

1.1

Introduction

Earthquake-resistant bridges should be designed to have regular configurations where


possible to ensure that the seismic behaviour is simple and predictable and above all,
inelastic energy dissipation is distributed almost uniformly in yielding components such
as columns. In regular structures the seismic demand does not tend to concentrate in a
few elements and all of the columns and/or fuse elements contribute to the seismic
response of the structure. However this ideal situation is often not achievable in bridge
construction due to irregularities imposed by site topography and required bridge
geometry (e.g., ramps).

It is recognised that in practice, bridges with certain configurations are more vulnerable
to earthquakes than others. Experience indicates that a bridge is most likely to be
vulnerable if (1) excessive deformation demands occur in a few brittle elements, (2) the
structural configuration is complex, or (3) a bridge lacks redundancy (Chen and Duan,
2000).

A common form of irregularity arises when a bridge traverses a basin or a valley


requiring columns with different lengths. Bridges with different column lengths can have
undesirable seismic behaviour in earthquake events. Stiffness irregularities in this type of
bridges results in considerable concentration of seismic forces in the shorter columns
which are usually the stiffer parts of the lateral resisting system. As a result very high
shear and moment forces arise in these columns.

Regular

Irregular

Fig. 1.1.An example of the concentration of the ductility demands in the stiffest column
(bridge at the bottom). The results obtained using a pushover analysis in SAP2000. The
columns in the regular bridge (bridge at the top) have uniform ductility demands.

Although the response of the superstructure in such irregular bridges may be relatively
uniform, the deformation demands on the individual substructure piers can be highly
irregular; with the largest strains imposed on the shortest columns. In some cases, the
deformation demands on the stiffer columns can induce their failure before longer, more
flexible columns can fully participate (e.g., Fig.1.1). The effects identified above can be
exacerbated in long-span bridges due to spatial and temporal variations in the ground
motions and out of phase displacements of the adjacent frames with different dynamic
characteristics which induce large relative displacements between adjacent bridge
columns (Chen and Duan, 2000).

Another problem arises in analysis of these irregular bridges where the sequential
yielding of ductile members may result in substantial deviations of the results from linear
analyses performed with the assumption of a global force reduction factor R from those
of the nonlinear response of the bridge structure. This problem is due to the fact that the
plastic hinges which appear first usually develop the maximum inelastic strains, which
may lead to concentration of unacceptably high ductility demands in these hinges.
Furthermore, following the formation of the first plastic hinges (normally in the stiffer
members) the distribution of stiffness and hence of forces may change from that predicted
2

by the equivalent linear analysis. This may lead to a substantial change in the assumed
pattern of plastic hinges (CEN, 2005). One may try to solve the problem by reducing the
value of the force modification factor, R, in which case it is certainly possible to reduce
the ductility demand in the stiffer piers, but the resulting overall design may be very
costly. Differentiated behavioural factors could then be proposed as a function of bridge
geometry (Calvi et al., 1994).

To improve the seismic performance of the bridges with varying column length several
methods such as the use of foundation sleeves for piers with appropriate depths to
equalize the effective length and stiffness of the columns (Priestly et al., 1996) , the use
of isolation devices (such as elastomeric and sliding bearings) with appropriate stiffness
to adjust the stiffness distribution and to improve the damping level (Calvi and
Pavese,1997) and the use of in-span hinges and abutments with sacrificial shear keys
(Saiidi et al. , 2001) can be beneficial.

In the past, earthquake damage to these types of bridges (e.g., failure of the shorter
columns), has been reported (Mitchell et al., 1995; Broderick and Elnashai, 1995 and
Chen and Duan, 2000). Due to the vulnerability of irregular bridges and lack of research
on this subject, research is needed to investigate the seismic behaviour and methods of
analysis of bridges with varying column stiffnesses.

The application of incremental dynamic analysis (IDA) for bridge structures is another
important issue which will be studied in this research. The use of different record
selection methodologies in IDA for the seismic evaluation of bridge structures is an
important issue which will be studied. In addition, the influence of considering different
earthquake types on the seismic evaluation of bridge structures using IDA is also studied.
Currently very limited research is available on these subjects which clearly require more
attention. More details and background information regarding incremental dynamic
analysis and different record selection methods are provided in Chapter 2. More details
and background information regarding different earthquake types are available in
Appendix F.

1.2 Research objectives

The main objectives of this research are:

To study the effects of different column heights and column stiffnesses on the
seismic response, and safety of bridges designed according to the 2006 Canadian
Highway Bridge Design Code provisions.

To investigate the use of the elastic multi-mode method for the seismic analysis of
the irregular bridges with different column and superstructure stiffnesses.

To evaluate the effects of different abutment conditions and abutment modeling


on the seismic response of bridges.

To evaluate the seismic performance of bridges using Incremental Dynamic


Analysis (IDA) considering the spectral shapes and epsilon effects.

To investigate the effects of using different record selection methodologies on the


seismic assessment of the bridges using IDA.

To compare the effects of including different earthquake types (i.e., subduction


interface and inslab earthquakes) on the seismic performance assessment of
bridges.

1.3 Thesis organization


This thesis is organized in seven chapters:

Chapter 1: Introduction and literature review presents an introduction to the problems


associated with irregular bridges with different column stiffnesses as well as an outline of
the thesis and a literature review of the previous studies concerning irregular bridges.

Chapter 2: provides background information and a literature review of the use of


incremental dynamic analysis (IDA) for seismic evaluation of structures, the use of
different record selection methods in IDA, different methods for prediction of the
capacity at different damage states, the development of fragility curves and seismic risk
assessment using IDA results. Chapter 2 has been summarized into a book chapter :
Tehrani, P. and Mitchell, D. Incremental Dynamic Analysis Applied to Seismic
Performance and Risk Assessment of RC Bridges in the publication Seismic Risk
Analysis and Management of Civil Infrastructure Systems, Edited by Tesfamariam, S.
(UBC, Canada) and Goda, K. (Bristol, UK), Woodhead Publishing Limited, submitted in
March 2012.

Chapter 3 : This chapter includes the paper: Tehrani, P. and Mitchell, D. Effects of
Column and Superstructure Stiffness on the Seismic Response of Bridges in the
Transverse Direction, accepted for publication in the Canadian Journal of Civil
Engineering, Manuscript 2011-0516, This chapter presents the study of the transverse
seismic response of bridges with column stiffness irregularity. The important parameters
affecting the seismic behaviour of bridges in this direction, including column stiffnesses,
abutment conditions and super structure stiffness have been studied. The use of different
regularity indices to determine the irregularity of bridges has been also studied.

Chapter 4: This chapter includes the paper: Tehrani, P. and Mitchell, D. Effects of
Column Stiffness Irregularity on the Seismic Response of Bridges in the Longitudinal
Direction, submitted in March, 2012 to the Canadian Journal of Civil Engineering
(Manuscript 2012-0091),. This chapter presents the study of the seismic response of
bridges with column stiffness irregularity in the longitudinal direction. The effects of
abutment conditions on the seismic response of bridges are also presented in this chapter.

Chapter 5: This chapter includes the paper: Tehrani, P., Goda, K., Mitchell, D., Atkinson,
G.M. and Chouinard, L.E. Effects of Different Record Selection Methods and
Earthquake Types on the Transverse Response of Bridges, submitted in December 2011
to the Journal of Earthquake Engineering and Structural Dynamics (Manuscript EQE-110079, revised version),. This chapter presents the use of incremental dynamic analyses for
the seismic assessment of bridges. The effects of using different record selection methods
and different earthquake types on the IDA results are studied.

Chapter 6: This chapter includes the paper: Tehrani, P. and Mitchell, D.Seismic
Response of Bridges Subjected to Different Earthquake Types using IDA, submitted in
January 2012 to the Journal of Earthquake Engineering (Manuscript UEQE-2012-1345).
In this chapter the incremental dynamic analysis has been used to evaluate the seismic
response of different bridge configuration using a large set of records including 234
ground motion records from three different earthquake types. The effects of including
different earthquake types and using different subsets of records in the seismic
assessments of bridges are studied.

Chapter 7: Conclusions summarizes the main conclusions from this study and provides
some recommendations for future studies.

Appendices:

Appendix A: provides more details regarding structural modelling, the validation of


computer models and preliminary results.

Appendix B: provides details regarding the computer program developed for design and
modelling of the bridges and extracting and processing the analyses results.

Appendix C: provides more information regarding the seismic hazard deaggregation and
presents more details regarding the computer program developed to compute the seismic
deaggregation results from the seismic hazard data. The seismic deaggregation results for

Vancouver are also given in some tables for different deaggregation methods and
different ground motion prediction equations.

Appendix D: provides more details regarding the computer program developed for record
selection and prediction of the conditional mean spectrum (CMS).

Appendix E: provides information regarding a program developed to monitor the details


of the analysis results from RUAUMOKO program.

Appendix F: provides more information regarding the subduction (i.e., interface and
inslab) earthquakes.

Appendix G: provides the list and detailed information regarding the ground motion
records selected for different earthquake types in Chapter 6.

Appendix H: provides a pushover analysis of the bridge studied in Chapter 5.

Appendix I: provides some analysis results for some bridges studied in Chapter 5 and 6.

1.4

Background and literature review

1.4.1

Bridges with column stiffness irregularity

Existing research on bridges with different column stiffnesses is limited, particularly in


North America, and most of the available studies have been carried out in Europe based
on the Eurocode provisions which are different from the North America provisions.

The first attempt in this regard was made by Calvi et al., (1994) and Calvi and Pinto
(1996). A simple, multi-degree-of-freedom, nonlinear dynamic model was used, after
some verification of the reliability of the results in comparison with more refined
simulations. The pier reinforcement was designed according to EC8/2, with some
modifications of the minimum longitudinal reinforcement percentage (assumed as low as
0.25% as commonly adopted for standard structural design rather than 1 %). The
maximum concrete compressive strain was assumed equal to 0.6%, with a
parabolic/constant stress-strain curve; the steel stress-strain curve was assumed bilinear,
with no maximum elongation capacity. The model used in the parametric simulations
assumes that each pier is hinged at the top and fixed at the base, with a bilinear global
force-displacement curve, resulting from a plastic hinge lumped at the base. The
numerical analyses indicated that the EC8/2 design approach may result in lower than
expected safety levels, or higher than expected ductility demands, when stiff piers are
coupled with flexible piers.

A first attempt to define a parameter of regularity, was made in this study which is a
measure of the difference between the mode shape of the whole bridge and of the deck
alone. The regularity parameter is expressed in Eq. 1.1.

()
[1.1]

R1

B
i

M ) Dj ) 2

j 1

[ 1.2]

R2 1

[(1  G

) ) iB M ) Dj ]

ij

i 1 j 1

A new index, R2 (given in Eq. 1.2), was then proposed by subtracting the norm of the
products of the off-diagonal terms to increase the sensitivity of the index. Where in Eqs.
1.1 and 1.2, ) iB , )Dj and M are the modal shapes of the bridge and the modal shapes and
the mass matrix of the deck alone, respectively. A summary of different indices of
regularity are given by Maalek et al. (2009):

1- Use of Modal Assurance Criterion-MAC (Ewins, 2000):


D
i

[1.3a]

^) ` ^) `
^) ` ^) ` ^) ` ^) `

MACi ( D, B)

D
i

B
i

D
i

B
i

B
i

(MAC ( D, B))

[1.3b]

i 1

MAC ( D, B, n)

2- Use of Modal Scale Factor-MSF (Ewins, 2000):

^) ` ^) `
^) ` ^) `
D

[1.4a]

MSFi

(MSF ( D, B))

[1.4b]

MSF ( D, B, n)

i 1

3-Difference Ratio of Mode Shapes-DRMS (Fischinger and Isakovic, 2003; Mackie and
Stojadinovic, 2003 and Maalek et al., 2009):

[1.5a]

DRMSi ( D, B)

)D  )B
)D

d d
d
D

SD  SB
SD

(DRMS (D, B))

i 1

DRMS (D, B, n)

[1.5b]

To improve the seismic performance of irregular bridges some design approaches for the
use of isolation systems have been proposed namely the equal (minimum) strength
approach and the maximum strength approach (Calvi and Pavese, 1997). In the equal
strength design approach the design will be based on the strength of the weakest column,
while in the maximum strength method different strength and stiffness will be assigned to
isolation systems considering the strength of each column and based on the maximisation
of a proposed regularity index given in Eq. [1.6a]. To account for the strengths of the
piers in the index, a more efficient (but complex) regularity index was proposed which
was based on the measure of the difference between a vector containing the product of
modal mass and spectral amplification and a vector containing the coefficients producing
the target deformed shape.

[1.6a]

R3

[1.6b]
[1.6c] Q

yi

1
1
2

(y  z )
i

i 1

(y )
i

i 1

(z )

'
)
i
ei ei

n
2

Vj

j 1

(V  k
i 1

i 1

)Ti MQ
)Ti M )i
m

y)
i

i 1

Where zi

Ei S di , E i are the modal participating factors,Q is the vector of the imposed

displacements, )i are the eigenvectors of the whole bridge, and Sdi are the spectral
displacements, Vi are the yield strengths of the piers (or of the isolators, if present, i.e.,
85% of the critical strength of the, corresponding pier), kei are the elastic stiffness of the
isolators and ' ei are the yielding displacements of the isolators.

10

Four reinforced concrete bridge models (1:2.5 scale) have been tested in the
European Laboratory for Structural Assessment (ELSA) using the pseudo-dynamic test
method (Pinto et al., 1996). Two bridge configurations were considered including a
regular (bridge B232) and another irregular bridge which included three alternative
design solutions (bridges B213A, B213B and B213C). In the case (B213B), the
reinforcement percentage of the shortest pier was increased (from 0.92% to 1.69%), with
the aim to reduce the high ductility demand which was anticipated in the analysis. In the
other case (B213C), the strength of the taller piers was increased (from 0.50% to 1.15%),
so that the forces attracted by the shortest pier can, in principle, be reduced.

Results showed that the absorbed energy in the regular bridge was almost equally shared
between the three piers while in the irregular bridge it was concentrated in the short
middle pier, which dissipated more than 70% of the total energy. Furthermore, safety
against collapse of the irregular bridges was quite low compared to the safety of the
regular bridge. In fact, despite the comparable demands obtained for the design
earthquake, the regular bridge was able to withstand twice the design loads without loss
of capacity and with a homogenous damage pattern. On the other hand, the irregular
bridges, tested with an input signal 1.2 times the design earthquake, suffered quite
important damage concentrated in the short central pier. Therefore a minor variation in
the peak ground acceleration may result in significant damage or high probability of
failure.

A comparison between the test results obtained for the three irregular bridges confirmed
that the strategy adopted for the design of the bridge B213B, which corresponds to the
use of a lower modification factor, results in performance that is not as bad as the other
bridges and may lead to a suitable design approach for these bridges with larger amounts
of longitudinal reinforcement in the short column. The short pier of bridge B213C,
containing less reinforced pier showed essentially flexural cracking while the pier from
bridge B213B had inclined shear cracking.

11

1.4.2

Modal pushover analysis (MPA)

Elastic analysis cannot predict the failure mechanisms or the redistribution of forces that
follow plastic hinge development, while pushover analysis identifies the locations of
structural inelasticity and failure mechanisms. Nonlinear static (pushover) analysis is a
popular tool for the seismic assessment of buildings. Although it is simpler compared to
nonlinear dynamic time-history analysis, its application is restricted to structures wherein
the response is governed by a single mode. To eliminate this limitation, pushover
analyses to consider higher modes effects have recently been developed for the case of
building structures. However little research has been focused on bridges in which the
higher modes effects are even more pronounced in the seismic response of the structure.

In an early work (Sasaki et al., 1998), the multi-mode pushover procedure was used to
identify the effects of higher modes in pushover analysis of buildings and separate
pushover curves were derived for each significant mode, however no attempt was made
to combine modal responses. In addition to MPA ,as another improved pushover method,
adaptive multi-mode pushover analysis method was developed (Bracci et al., 1997;
Gupta and Kunnath ,2000 and Antoniou et al., 2002) involving updating the loading
pattern, according to the current displaced shape of the structure at each step, which is
determined by modal combination rules at each stage of the response.

Modal pushover analysis (MPA), proposed by Chopra and Goel (2002), and subsequently
improved by them (Chopra and Goel, 2004), is a method in which the pushover analyses
are carried out separately for each significant mode, and the modal components
(displacements, drifts, etc.) are combined using an appropriate combination rule.
Considering inelastic time history analysis (ITHA) as a benchmark, errors in MPA are
typically smaller than in the case of superposition at the level of loading (with fixed
loading pattern), as recommended in the FEMA356 Guidelines, for example (FEMA 356,
2000).

12

Modal pushover analysis has been extended for application to bridge structures (Kappos
et al., 2004; Fischinger et al., 2004; Pinho et al., 2005; Paraskeva et al., 2006; Isakovic and
Fischinger, 2006 and Isakovic et al., 2008). On the basis of the results obtained for the

long curved irregular bridge studied by Paraskeva et al. (2006), MPA seems to be a
promising approach that yields more accurate results compared to the standard pushover
analysis, without requiring the higher modelling effort and computational cost, as well as
the other problems involved in ITHA and above all this method, unlike the adaptive
analyses, can be easily implemented using available standard software tools. Some
common computer programs such as SAP2000 are capable of performing modal
pushover analysis. The application of this method in bridge codes is then expected in the
near future after further confirmation of this method for a wide range of bridge structures.
It is thus concluded that more research is required, to further investigate the application of
the MPA to bridge structures with different configurations, degree of irregularity, and
dynamic characteristics (e.g., in terms of higher mode significance), since MPA is
expected to be even more valuable for the assessment of the seismic behaviour of bridges
with contribution of higher modes (Paraskeva et al., 2006).

1.4.3

Analysis of bridges with different column stiffnesses

The use of different elastic and inelastic analysis methods such as single mode (SM),
multi mode (MM) and inelastic time history analysis (ITHA) to predict the response of
regular and irregular bridges were investigated by Fischinger et al., (1997) on a few fourspan viaducts with different abutment conditions. The results indicated that the elastic
single mode procedure completely failed to identify the critical elements of many of the
viaducts that were analyzed and sometimes yielded qualitatively different results than the
MM and time history analyses. In addition, elastic analysis failed completely to predict
the displacement shape of some of the analyzed irregular viaducts. It was also concluded
that the results were sensitive to the ratio of torsional to translational stiffness rk = (
Kt / K y ), where Kt and Ky are defined in Eqs. [1.7a] and [1.7b], respectively. The strength

of the columns was another important factor affecting the results.

13

[1.7a]

Kt

( K ci xi ) / r 2

[1.7b]

Ky

ci

where K ci , xi and r are the flexural stiffness of pier i, the distance of pier i from the centre
of mass and the radius of inertia respectively. K t is a measure of the sensitivity of the
structure to rotations in the horizontal plane. The influence of higher modes was not
important in the case of a symmetric viaduct or when the ratio of the maximum column
stiffness to the deck stiffness, rk , was less than approximately 15. The complex
behaviour with important contribution of several response modes was typical in the
situations when the deck is relatively flexible in the transverse direction in comparison
with piers, the ratio of the quasi-torsional to translational stiffness of the structure is
small, the supports at the abutments are free in the transverse direction and the end
cantilevers are long and the viaduct is asymmetric. In the case of free supports at the
abutments, the results of the MM and SM methods differed even much more than in the
case of pinned abutments.

Fischinger and Isakovic (2003) have proposed an irregularity index as a global numerical
measure to help designers decide about the suitability of the single-mode pushover
analysis for bridges. This index represents relative differences between the areas bounded
by the normalized displacement lines of the first and second iteration of the single-mode
pushover method. For the first iteration a uniform load pattern is used to perform the
pushover analysis. The resulting displaced pattern of the bridge from the first iteration
will then be used as a new load pattern for the second iteration of pushover. It was
roughly estimated that bridges with an irregularity index, IRI, less than 5% could be
analysed with the simpler analysis procedures such as single-mode pushover method.
n 1

[1.8]

IRI

I  I 'x .100
I 'x
i ,1

i ,2

i ,1

i ,1

 Ii ,2

i 1

.100

n 1

i ,1

i 1

14

Where Ii ,1 is the normalised displacements obtained within the first iteration of the
pushover method, Ii ,2 is the normalised displacements obtained within the second
iteration of the pushover method and 'xi is the distance between two points where
displacements were calculated. A similar IRI has been also proposed by Mackie and
Stojadinovic (2003) in which mode shapes of the bridge and the deck alone are used
instead of the first and the second iteration of the pushover method, respectively.

Based on the results from the previous works on a few simple 4-span bridges, it can be
found that the elastic analysis and simple pushover method should be used with care for
the following cases (Fischinger and Isakovic, 2003):

a) Viaducts with great eccentricity, defined as a distance between the center of stiffness
of the supporting elements and the center of mass.
b) Torsionally flexible viaducts (due to the abrupt changes in the dynamic properties of
the bridge once the columns yield).
c) Viaducts with a relatively flexible deck in comparison with stiff and strong columns.
d) Viaducts with a very stiff central pier. This situation is more pronounced for viaducts
with roller supports at the abutments. These effects depend also on the relative stiffness
of the central columns in comparison with the deck.
e) For viaducts with very stiff and relatively strong end columns the influence of higher
modes can be expected even if they are symmetric.

In a study by Isacovic and Fischinger (2006), the influence of the higher modes and their
consideration in the pushover analysis of 6 four-span reinforced concrete single column
bent viaducts was investigated. Typical multimode pushover-based methods (modal
pushover analysis, modal adaptive non-linear static procedure and incremental response
spectrum analysis) are addressed and compared with a single mode procedure and
inelastic time history analysis for three types of viaducts including regular, slightly
irregular, and highly irregular viaducts. Based on the results from this study in most
cases, all the analysed multimode pushover-based methods have given results comparable

15

with those obtained from time history analysis, with exception of the cases where the
torsional sensitivity varied during the response.

According to this study, if the substructure is flexible in comparison with the


superstructure in the transverse direction, the influence of the higher modes will be small.
It was confirmed that the level of irregularity and contribution of the higher modes in the
response are considerably influenced by the position of stiff columns (particularly when
they are positioned close to the centre of the bridge), since certain configuration of stiff
columns enlarge the torsional sensitivity of the viaduct. Thus, for the analysis of irregular
bridges having short and slightly damaged columns, the multimode methods are needed.
The results also showed that for the case of a lower ductility demand, the pushover
procedures underestimate the response while in the region of higher ductility demand or
at higher earthquake intensity, in which plastic hinges develop in all columns, they yield
similar results as the ITHA.

The applicability of a typical single-mode pushover method and two typical multi-mode
pushover methods (the modal pushover analysis (MPA) and incremental response
spectrum analysis (IRSA) methods (Aydinoglu, 2004)) for the analysis of single column
bent viaducts in the transverse direction was then studied considering some longer
viaducts with more bents (Isacovic et al., 2008). According to the results the single-mode
pushover method is accurate enough for bridges where the effective modal mass of the
fundamental mode is at least 80% of the total mass. In the case of the moderately
irregular long viaducts the MPA method performed well.

The results from a study by Pinho et al. (2007) indicate that the use of single-run
pushover analysis might still be feasible even for irregular bridge configurations,
provided that a displacement-based adaptive version of the method is employed.

Research by Kappos et al. (2002) on an irregular bridge showed that soil-structure


interaction (SSI) had small effects on the response of the bridge and even smaller

16

displacements and less reinforcement were obtained when soil-structure interaction (SSI)
was considered in the models.

A comparison between the modelling using fibre cross-sections and beam elements in the
analyses was carried out in some studies. Although the beam model with lumped
plasticity was simple, it yielded results which were quite in good agreement with more
detailed models (Calvi et al., 1994 and Fischinger and Isakovic, 2003).

A comparison of the results from elastic response spectra and inelastic time history
analyses for an existing irregular bridge in a study by Jones et al., (2001) showed that the
effects of velocity pulses, especially when the structure is located in sufficient proximity
to faults to experience near-field effects from a fault rupture, can substantially change the
bridge response. They also recommended that a drift limit may be a useful preliminary
design tool, because when drifts exceeded certain limits then the results from the elastic
and inelastic analyses were not consistent.

1.5 Code provisions

1.5.1 Caltrans (2006) and AASHTO Guidelines (2009)


The requirements in Caltrans (2006) and the AASHTO Guidelines (2009) compare
the ductility demands with the ductility capacities in the energy dissipating members
(displacement-based design) instead of using an overall modification factor for design
(force-based design). Although the problem of different ductility demands of columns in
irregular bridges can be better addressed using this method, the ductility demands which
are based on the linear analysis methods such as response spectrum analysis (which is
permitted for design of irregular bridges) usually underestimate the ductility demands in
critical members and can lead to unsafe designs. Some conservative limitations are
recommended in Caltrans and the AASHTO Guidelines for the ratio of the column
stiffnesses in a frame or bent to balance the bent or pier stiffness along the bridge. For

17

any two bents or any two columns within a bent this stiffness ratio is recommended be
greater than 0.5. For adjacent bents or adjacent columns within a bent this ratio is
recommended to be less than 0.75. Some of the consequences of not meeting these
relative stiffness indicators include increased damage in the stiffer elements and an
unbalanced distribution of inelastic response throughout the structure.

If project constraints make it impractical to satisfy the stiffness requirements, a careful


evaluation of the local ductility demands and capacities shall be required for bridges in
regions of high seismicity.

1.5.2

Eurocode

In the Eurocode (EN 1998-2, 2005) a force-based design approach is used for the
design of bridges. The reduction factor, q (similar to the modification factor R), for
reinforced concrete columns depends on the shear span to depth ratio and on whether the
member is inclined or vertical. The q factor will then be modified in the cases of high
axial loads and irregular seismic behaviour of the bridge. In this regard the local force
reduction factor ri associated with member i is defined as:

[1.9a] ri = q (MEd,i / MRd,i)

[1.9b]

 = rmax / rmin  o

[1.9c]

qr = q (o / )  1.0

where MEd,i and MRd,i are the maximum values of design moment at the intended plastic
hinge location of ductile member i from the seismic analysis and the design flexural
resistance of the section with the actual reinforcement, respectively. The bridge is
considered to have regular seismic behaviour when Eq. 1.9b is satisfied. Where rmax and
rmin are the maximum and minimum values of ri, respectively and o is a limit value

18

selected so as to ensure that sequential yielding of the ductile members will not cause
unacceptably high ductility demands on the member. The recommended value for o is
2.0. The force reduction factor is then reduced to qr for the irregular seismic behaviour
according to Eq. 1.9c. To capture the actual seismic behaviour of an irregular bridge
where the ductility demands concentrate in a few elements and the distribution of the
forces deviate from that predicted by the linear analysis, a combination of an equivalent
linear analysis with a non-linear static analysis is recommended.

1.5.3

CSA-S6-06 and AASHTO-04

The seismic design provisions in the Canadian Highway Bridge Design Code
(CHBDC) (CSA 2006) are based on the AASHTO Specifications (AASHTO 2004). S6
uses a force modification factor, R, and an importance factor, I, of 1.0, 1.5 and 3.0 for
other, emergency-route and lifeline bridges, respectively. The AASHTO
Specifications combine the force modification factor for ductility with the importance
factor to give one force modification factor for other, essential and critical bridges.
Nevertheless, the resulting design forces are similar. The CHBDC requires that the MM
(multi-mode spectral method and also referred as response spectrum analysis in this
paper) be used for the analysis of irregular bridges, even in seismic performance zone 4.
However for irregular lifeline bridges, time-history analysis is required in zones 3 and 4.
Concerning the column stiffness ratios for regular bridges, the 2004 AASHTO and the
2006 CHBDC specifications indicate that the maximum bent or pier stiffness ratio from
span to span should not exceed 4.0 for bridges up to 4 spans, 3.0 for 5 spans and 2.0 for
6 span bridges. These stiffness limits are for bridges with a continuous superstructure or
multiple simple spans with longitudinal restrainers and transverse restraint at each
support or a continuous deck slab, otherwise this ratio shall not exceed 1.25.

The modification factors, R, are taken conservatively and lower than the expected
displacement ductility capacities, since the procedure is intended to apply to a wide
variety of bridge geometries. Where possible, pier stiffnesses should be adjusted to
attempt to achieve uniform yield displacements and ductility demands on individual

19

piers. In cases where attempts to regularize the structure are impractical, suitable
analyses need to be developed to account for localized, rather than simultaneous, yielding
of piers. In some cases, it may be possible to use stiff piers with energy dissipating
bearings to alleviate the problem as discussed in the CHBDC Commentary (CSA, 2006).

1.5.4

Performance objectives for bridges

Based on the CHBDC provisions, bridges are classified into three importance categories
including lifeline bridges, emergency-route bridges and other bridges.
Emergency-route bridges are generally those that carry or cross over routes that should, at
a minimum, be open to emergency vehicles and for security/defence purposes
immediately after the design earthquake (CHBDC, 2006).

The Seismic Subcommittee of the CHBDC has proposed the following performance
criteria as presented in Table 1.1 for the next version of the CHBDC. According to Table
1 the emergency-route bridges should be repairable after the occurrence of an earthquake
with 10% probability of exceedance in 50 years. However, all bridges should not collapse
when subjected to earthquakes with 2% probability of exceedance in 50 years.

20

Table 1.1. Proposed seismic design performance criteria by the Seismic Subcommittee of
the CHBDC
Seismic Ground motion Probability

Service Level

Damage Level

2% in 50 years (2475 years)

Possible loss of service

Significant (No collapse)

5% in 50 years (975 years)

Limited

Repairable

10% in 50 years (475 years)

Immediate

Minimal

2% in 50 years (2475 years)

Possible loss of service

Significant (No collapse)

10% in 50 years (475 years)

Limited

Repairable

Possible loss of service

No collapse

of Exceedance (return period)


Lifeline Bridges

Emergency-Route Bridges

Other Bridges
2% in 50 years (2475 years)

1.6 Problems associated with the force-based design approach


The forced-based design method involves some underlying assumptions which may not
be valid particularly for certain types of structures. Some problems associated with
irregular bridges are due to some unrealistic assumptions made in the force-based design
method.

In addition, the degree of protection provided against damage under a given seismic
intensity is non-uniform from structure to structure, when the force-based design
approach is used. Thus, the concept of "uniform risk" which is implicit in the formulation
of current seismic design codes is not achieved in the structural design. Using
displacement-based approaches, such as the Direct Displacement-Based Design (DDBD)
method (Priestley et al., 2007), a more rational and realistic design is achieved. The main
problems associated with the use of the force-based design approach have been

21

summarized by Priestley (Priestley et al., 2007). Some of these problems which are
related to the subject of this research are summarized below:
1.6.1

Interdependency of strength and stiffness and relationship between Strength


and ductility demand

In force-based design the member stiffness is traditionally assumed to be independent of


strength, for a given member section. The flexural rigidity can be estimated from the
moment curvature relationship in accordance with the beam equation as EI

M n / )y .

Where Mn is the nominal moment capacity and y is the yield curvature based on the
equivalent bi-linear representation of the moment-curvature curve. The assumption of
constant member stiffness implies that the yield curvature is directly proportional to
flexural strength. Detailed analyses and experimental evidence show that this assumption
is invalid in that stiffness is essentially proportional to strength, and the yield curvature is
essentially independent of strength, for a given section (Priestley et al., 2007).

M3

a)

b)

M2

Moment

Moment

M2
M1

M3

M1

Curvature

Curva ture

Fig. 1.2. Influence of strength on moment-curvature diagram: a) design assumption


(constant stiffness); b) realistic condition (constant yield curvature) (adapted from
Priestley et al., (2007))

Furthermore, according to the common force-based assumption that stiffness is


independent of strength, increasing the strength of a structure by reducing the forcereduction factor improves its safety. However, experimental and analytical results

22

indicate that the displacement capacity decreases as the strength increases (Priestley et
al., 2007).

1.6.2 Structures with unequal column heights


The ductility capacity of a cantilever bridge column,
, can be calculated using Eqs.
1.10a to 1.10c.

[1.10a]

'y

)y H 2 / 3

[1.10b]

'P

)pLpH

[1.10c]

P'

'y  'p
'y

1 3

)pLp
)y H

Where y and p are the yield and plastic curvature, y and p are the yield and plastic
displacement of the column, Lp is the plastic hinge length and H is the height of the
column. Therefore based on Eq.1.10c, the displacement ductility capacity reduces as the
height increases (Priestley et al., 2007). Thus the concept of uniform displacement
ductility capacity, and hence of a constant force-reduction factor may not be appropriate
even for this very simple class of structure in that the influence of structural geometry on
displacement capacity of columns with identical cross-sections, axial loads and
reinforcement details is not considered.

In conventional force-based design a force-reduction factor, reflecting the assumed


ductility capacity is applied to determine the seismic design lateral force, which is then
distributed to the piers in proportion to their stiffness. Implicit in this approach is the
assumption of equal displacement ductility demand for all columns. If the columns have
the same cross section dimensions, as is likely to be the case for architectural reasons, the
design shear forces in the columns will be in inverse proportion to H i 3 , since the stiffness
of column i is given by Ki

C1 EI i.e
,where I i.e is the effective cracked-section stiffness
Hi3

23

of column i.. The consequence of this design approach is that the design moment at the
base of the piers will be M Bi

C2Vi H i

C1C2 EI i.e
that is, in inverse proportion to the
Hi2

square of the column heights. C1 and C2 are constants dependent on the degree of fixity
at the pier top. Consequently the shortest piers will be allocated much higher flexural
reinforcement contents than the longer piers. This has three undesirable effects. First,
allocating more flexural strength to the short piers will increase their elastic flexural
stiffness, EI i .e , even further with respect to the more lightly reinforced longer piers.
Second, allocating a large proportion of the total seismic design force to the short piers
increases their vulnerability to shear failure. Third, the displacement capacity of the short
piers will clearly be less than that of the longer piers, since the displacement capacity of
heavily reinforced columns is reduced as the longitudinal reinforcement ratio increases
and hence the force-based design approach will tend to reduce the displacement capacity
(Priestley et al., 2007).

In DDBD approach in which the stiffness of the pier is based on the effective secant
stiffness at the maximum displacement, the distributed seismic forces are inversely
proportional to the height of the columns (see Eq. [1.11b]) assuming that the columns
have equal reinforcement ratios (Kowalsky, 1997).

[1.11a]

[1.11b]

K eff

K cr

K eff ' m

3EI cr ' y
H 3 'm
EI cr ) y
H

EI cr ) y
H 'm

Fi

1
Hi
1
H
i

where Fi is the shear force in column i, Icr is the column cracked section moment of
inertia, m is the maximum displacement of the column and
is the displacement
ductility demand in the column. The other parameters are defined before. The use of a
force-reduction factor which does not reflect the different ductility demands and
capacities clearly result in structures having different safety levels (Priestley et al., 2007).

24

1.6.3

Structures with dual (Elastic and Inelastic) load paths

In case of bridges with fixed abutments subjected to transverse seismic excitation, the
primary seismic resistance is provided by bending of the piers, which are designed for
inelastic response. However, since the abutments are restrained against the transverse
movements, another load path is also developed by the elastic bending of the
superstructure. As a result, the application of a force-reduction factor for design is not
rational, since part of the loads is carried by elastic action in the superstructure (Priestley
et al., 2007). Determination of the correct design solution for the transverse response of
bridges is considerably more onerous than for longitudinal design (Priestley et al., 2007).

1.7 Introduction to direct displacement-based design method


Damage can be directly related to deformation. Hence designing structures to achieve a
specified displacement limit implies designing for a specified risk of damage, which is
compatible with the concept of uniform risk applied to determining the design level of
seismic excitation. Therefore different structures designed with this approach will
(ideally) have the same risk of damage, rather than a variable risk associated with current
design approaches. Using state-of-the-art detailing/deformation relationships, structures
with uniform risk of collapse, as well as of damage can theoretically be achieved. The
fundamental difference from force-based design is that the Direct Displacement Design
(DDBD) characterizes the structure to be designed by a single-degree-of-freedom
(SDOF) representation of performance at peak displacement response, rather than by its
initial elastic characteristics (Priestley et al., 2007). This procedure was initially proposed
by Priestley (1993), with the objective of designing a structure which would achieve,
rather than be bounded by, a given performance limit state under a given seismic intensity
(Priestley, 1993; Kowalsky, 1997, Priestley et al., 2007). The method utilizes the
Substitute Structure approach (Gulkan and Sozen, 1974) to model the inelastic structure
as an equivalent elastic single-degree-of-freedom (SDOF) system. Calvi and Kingsley
(1995) extended this methodology to multiple degree of freedom bridge structures.

25

Dwairi and Kowalsky (2006) investigated the displacement patterns of bridges subjected
to transverse seismic excitations using nonlinear time-history analysis. The variables
considered in this study included the bridge geometry, the superstructure stiffness, the
substructure strength and the stiffness and abutment support conditions. A series of three
inelastic displacement patterns were identified: (1) a rigid body translation (2) a rigid
body translation with rotation and (3) a flexible pattern. A relative stiffness index, RS
(Eq. [1.12]), that is a function of the superstructure and substructure stiffness was shown
to be a key variable in determining the type of displacement pattern a bridge is likely to
follow.

[1.12]

RS

32I s
5L3s
n I
ci

Ks
n

K
i 1

ci

h
i 1

3
ci

where n is the number of piers, index i refers to the pier number, Ic is the pier cracked
section moment of inertia, hc is the pier height, Is is the deck gross moment of inertia, and
Ls is the deck total length. Based on the analyses results, a rigid body translation pattern
was identified for symmetric bridges with free abutments. In addition, a rigid body
translation with rotation was identified for asymmetric bridges with free abutments. The
majority of bridges with abutment restraint in the transverse direction had flexible
displacement patterns. The DDBD procedure was then evaluated for some multi-span
bridge design cases. The evaluation process showed that all of the four and five-span
asymmetric bridge cases had displacements less than the target displacements and the
target displacement profile was accurately predicted. However the target displacements
were exceeded in more than 50% of the design cases for 6, 7 and 8 span bridges for
symmetric and asymmetric bridge configurations. In a limited number of cases with very
irregular configurations and flexible superstructures, the design procedure failed to
predict the target-displacement profile. The failure is attributed to the inability of the
effective mode shapes to estimate the target-displacement profile of a highly irregular
MDOF structure.

26

In some studies at the ROSE School (Botero, 2004 and Restrepo, 2006), the
displacement-based procedure has been used to design different series of bridge
configurations with restrained abutments. The resulting designs were subjected to
acceleration time histories to assess the accuracy of the method in terms of reaching the
target design objectives, represented by target displacements. Satisfactory results were in
general obtained from the assessment of the procedure using inelastic time-history
analyses. Good results were also obtained for irregular Bridge configurations which
showed that the implementation of the DDBD method to more complex structures is fully
feasible. Nevertheless, there are still some problems associated with the displacement
pattern of very stiff bridge configurations, like those with very short central pier (or piers)
and taller exterior piers .In those types of bridges generally the first elastic and inelastic
mode shapes significantly differ and clearly more studies should be focused on these
types of bridges.

27

2 Incremental dynamic analysis applied to seismic


performance and risk assessment of RC bridges
P. Tehrani and D. Mitchell, McGill University, Canada

2.1 Introduction
In Chapter 1 the problems associated with the bridges with different column stiffnesses
were explained. This thesis also focuses on the use of incremental dynamic analyses for
the seismic evaluation of bridges. This chapter provides an introduction to the
incremental dynamic analysis (IDA) method, different methods for record selection,
concepts of epsilon and spectral shape, different ground motion prediction equations and
the application of IDA method for development of fragility curves and risk assessments.
The use of different theoretical and experimental models to estimate the capacity of
bridge columns is also explained. The concepts and background information presented in
this chapter are used in Chapters 5 and 6 to study the use of IDA for the seismic
assessment of bridges.

Chapter 2 was summarized into a book chapter : Tehrani, P. and Mitchell, D.


Incremental Dynamic Analysis Applied to Seismic Performance and Risk Assessment of
RC Bridges, and has been submitted in March 2012 to the publication Seismic Risk
Analysis and Management of Civil Infrastructure Systems, Editors: Tesfamariam, S.
(UBC, Canada) and Goda, K. (Bristol, UK), Woodhead Publishing Limited.
The authors are permitted to use the material from the book chapter in this thesis for noncommercial purposes, with acknowledgement to the original source.

28

2.2 Incremental dynamic analysis

2.2.1

Introduction

IDA, developed by Vamvatsikos and Cornell (2002), is an analysis method which can be
used for more detailed seismic performance predictions of structures subjected to
different seismic excitation levels. IDA involves numerous inelastic time history analyses
performed using one or a set of ground motion record(s), each scaled (up or down) to
study different seismic intensity levels.

The IDA procedure has been adopted by some guidelines including the ATC-63
provisions (ATC-63, 2008) to determine the seismic performance, collapse capacity and
fragility assessment of buildings. Some of the main objectives of the multi-purpose IDA
analysis can be summarized as follows (Vamvatsikos and Cornell, 2002) :

1. To provide a thorough evaluation of the seismic responses (seismic demands) for a


wide range of seismic intensity levels,
2. To provide a better understanding of the seismic response of structures due to rare and
more severe ground motions at ultimate performance levels, including collapse,
3. To provide valuable insight into the changes in the structural response as the intensity
of the ground motion changes
4. To provide estimates of the dynamic capacity of the structural systems, also known as
the collapse capacity of the structure, defined as the point of global dynamic instability,
5. To compare the seismic responses due to different possible ground motion records and
to provide estimates of the variability (uncertainty) in the seismic responses, when a set
of records with different characteristics are used,
6- To provide information regarding possible structural responses, required for the
probabilistic seismic performance assessment of structures and seismic risk analysis (e.g.,
development of fragility curves and prediction of the annual rate of collapse).

29

2.2.2

Damage measures and intensity measures

The IDA results are commonly presented using an intensity measure (IM) versus a
damage measure (DM) of interest. IM is a non-negative scalable scalar, which is a
function of the unscaled accelerogram, and is monotonically increased with a scale factor
(Vamvatsikos and Cornell, 2002). Many quantities are available to characterize the
intensity of a ground motion record; however such quantities are often treated as nonscalable parameters. Examples of such quantities include the moment magnitude,
duration, or Modified Mercalli Intensity. Examples of scalable IMs which are commonly
used in the IDA include the Peak Ground Acceleration (PGA), Peak Ground Velocity,
and the 5% damped Spectral Acceleration at the structures first-mode period (Sa
(T1,5%)). The use of the Sa(T1,5%) quantity as the IM is often recommended and used in
different research studies and guidelines [e.g., ATC-63 (2008), and Vamvatsikos and
Cornell (2002)].

The damage measure, DM, is defined as a non-negative scalar quantity that characterizes
the response of the structure to seismic excitations and can be deduced from the output of
the nonlinear dynamic analysis. A wide range of such quantities are available and can be
selected as a DM. Selecting an appropriate DM depends on the objective of the analyses
and the characteristics of the structure. Some of the common choices include the
maximum base shear, maximum columns ductility demands, various proposed damage
indices (e.g., global cumulative hysteretic energy), and maximum column drift ratio). For
bridges the maximum drift ratio of the critical columns or the maximum ductility demand
is typically used for the DM parameter.

2.2.3

IDA Curves

The results of an IDA for a structure is often presented in the form of one or more IDA
curves each representing the IDA results performed for each ground motion record. An
IDA curve is the plot of the damage measure (DM) variable versus one or more intensity
measure (IM) parameter(s). These curves demonstrate the state of the DM parameters at

30

different intensity levels of the input records. This enables the study of the seismic
demand parameters from low seismic intensity levels prior to yielding of the structure up
to ultimate performance levels such as dynamic instability of the structure (i.e., collapse),
buckling or fracture of the steel bars, shear failure, etc. Each non-linear dynamic analysis
performed at each seismic intensity level is presented by a point in the IDA curve. This
point demonstrates the maximum seismic demand (in terms of the DM parameter)
obtained at the corresponding intensity level of the ground motion record. The full IDA
curve then can be developed by fitting a curve through the points computed for different
intensity levels (IM). The interpolation between the points is often used for this purpose.
An example of an IDA curve is shown in Fig. 2.1 in which the spectral acceleration at the
fundamental period, Sa(T1), and the maximum drift ratio of columns are used as IM and
DM parameters, respectively.

The various types of IDA responses that may be observed are discussed by Vamvatsikos
and Cornell (2002). Such responses may include hardening behaviour, softening
behaviour, waving behaviour, and structural resurrection. Increasing the ground motion
intensity in IDA does not always necessarily result in higher damage predictions. In fact
the details of the ground motion records will also influence of the IDA results. For
example a higher intensity of ground motion records may result in an earlier yielding of
the structure. This in turn can change the effective period of the structure and the amount
of dissipated energy due to inelastic deformations. So that the resulting maximum
displacement of the structure may be even smaller than that obtained for a lower
intensity.

Once the nonlinear dynamic analyses performed at certain ground motion intensities and
the corresponding DM values are obtained, it is possible to use interpolation to
approximate the entire IDA curve without performing additional analyses. The
interpolations are typically performed either using basic piecewise linear approximation,
or the spline interpolation (Vamvatsikos and Cornell, 2004). However to use interpolation
the nonlinear dynamic analyses should be performed at a sufficient number of intensity
levels to make sure that this approximation is acceptable. The use of spline interpolation

31

has the advantage that it is more precise and the interpolations can be performed using
smaller number of points (see Vamvatsikos and Cornell (2004) for more details).
However the use of the linear interpolation is more straightforward for practical purposes,
provided that a sufficient number of points is available at closely spaced IM values to
approximate the full IDA curves.

Other methods to approximate the IDA curves and to determine the statistics of the IDA
curves include the parametric methods such as the two-parameter, power-law model
max= [Sa(T1)] introduced by Shome and Cornell (1999). Such parametric methods
often provide a simple yet powerful description of the curves, although they lack the
flexibility to accurately capture each IDA curve (Vamvatsikos and Cornell, 2002).

6.0

Sa(T1) (g) (IM)

5.0
4.0
3.0
2.0
1.0
0.0
0

0.02

0.04
0.06
Maximum drift ratio (DM)

0.08

Fig. 2.1. An example of an IDA curve developed using Sa(T1) as IM and maximum drift
ratio as DM. The results of nonlinear dynamic analysis for each IM are shown by dots
and linear interpolation is used to fit a curve though these points. The last point of the
IDA curve corresponds to an infinite drift ratio (i.e., dynamic instability).

32

2.2.4

Summarizing IDA results

Different ground motion records in IDA often result in different response predictions
which are quite dissimilar, due to a wide range of behaviour and large record-to- record
variability. Hence it is difficult to choose one particular response prediction to represent
the behaviour of the structure. Therefore it is essential to summarize such large amount of
data obtained through all IDA curves to quantify the randomness introduced by the
records. Appropriate summarization techniques should be used to reduce this data to the
distribution of DM given IM and to the probability of exceeding any specific limit-state
given the IM level. There are several methods to summarize the IDA curves, but the use
of percentiles (fractiles) are the most appropriate approach, since at the point of dynamic
instability, introduced by flatlines in the IDA curves, the DM values are infinite and
therefore the use of mean values or similar parameters is not possible (Vamvatsikos and
Cornell, 2002).

In order to evaluate the IDA results often the median responses are determined along
with the predicted dispersion of the results from different ground motion records. The
IDA results thus can be summarized in percentiles, including median (50% percentile),
16% and 84% percentiles. With the assumption of a lognormal distribution of maximum
drift ratio as a function of Sa(T1), the median (i.e., 50% percentile) is the natural central
value and the 84%, 16% percentiles correspond to the median times edispersion, where
dispersion is the standard deviation of the logarithms of the values (Jalayer and Cornell,
2003). These percentile curves are much smoother than the individual IDA curves and
can better represent the overall behaviour of a structure.

For example if 44 records are used in IDA, for each arbitrary IM value 44 values of DM
are obtained (i.e., DM given IM values). By summarizing the DM values into their 16%,
50%, and 84% percentiles the percentile values of DM given IM are obtained for the
arbitrary IM value considered. This procedure can then be repeated for a number of
different IM values (preferably spaced at equal intervals) and the corresponding
percentiles of DM values can be obtained. By interpolating the points obtained for each

33

DM fractile, the percentile IDA curves can be generated and the IDA results can be
summarized into 16%, 50%, and 84% percentile IDA curves, as shown in Fig. 2.2.

A summary of the IDA curves can be expressed, either by the values of DM given IM or
the values of IM given DM can be used. The first method provides the distribution of
demand (i.e., DM) in the structure at a given intensity level, IM, while the second
approach provides the distribution of intensities, IM, that cause a given level of damage,
DM, in the structure. It has been shown that when percentiles are used to summarize the
IDA results, the 16%, 50%, and 84% percentiles given IM (e.g., Sa(T1)) almost perfectly
match the 84%, 50%, and 16% percentiles, respectively, given DM (e.g., maximum drift
ratio). In another word the line connecting the x% percentiles of DM given IM is the
same as the one connecting the (100 x)% percentiles of IM given DM (Vamvatsikos and
Cornell, 2004).

Spectral acceleration at T=0.7 sec (g)

84% IM|DM (16% DM|IM)

2.5
2

50% IM|DM ( 50% DM|IM)

1.5
1
0.5

16% IM|DM (84% DM|IM)

0
0

0.02

0.04
0.06
Maximum column drift

0.08

0.1

Fig. 2.2. An example of IDA curves developed using 44 crustal records for a 4-span
bridge (percentile 16%, 50% and 84% IDA curves are shown by heavy lines) (Adapted
from Tehrani et al. (2012))

Typically the median IDA curves exhibit similar trends and they include distinct
segments which are influenced by the capacity boundary curves (i.e., the back-bone
response curve) of the structural components. Based on such observations the capacity
34

boundary of the structures may be used to predict the median IDA curves using simpler
analyses. The idea is to obtain the capacity boundary curve of the structure by means of a
simple pushover analysis and then use this curve as a back-bone curve of a single degree
of freedom system for nonlinear dynamic analyses (e.g., see Vamvatsikos and Cornell,
(2006)).

2.2.5

Characteristics of median IDA curves

Individual IDA curves for single ground motion records are typically very sensitive to
dynamic interaction between the properties of the system and the characteristics of the
ground motion such as the frequency content and the duration of the records. Fractile
IDA curves (16th, 50th and 84th percentiles), however, are much more stable and provide
better information on the central value (i.e., median) and variability (i.e., dispersion) of
the response. In general, median IDA curves exhibit the following characteristics as
shown in Fig. 2.3 (ATC-62, 2008):
x An initial linear segment is observed which corresponds to linear-elastic behaviour of
the structural model prior to yielding in which the lateral deformation is proportional
to the ground motion intensity. However in some systems, the initial linear segment
may be extended beyond the yielding point into the inelastic deformation range as
shown in Fig. 2.3. In this pseudo-linear segment, lateral deformation demand is
approximately proportional to ground motion intensity, which is consistent with the
familiar equal-displacement approximation for estimating inelastic displacements.
The range of lateral deformation demands over which the equal-displacement
approximation is applicable depends on the characteristics of the force-displacement
capacity boundary and the period of vibration of the structure.
x A second segment on the median IDA curve corresponds to inelastic behaviour of the
structure in which lateral deformation (DM) is no longer proportional to ground
motion intensity (IM). As intensity increases, lateral deformation demands increase at
a faster rate. This segment corresponds to softening of the system, or reduction in
35

stiffness of the structure (reduction in the slope of the IDA curve). In this segment,
the system transitions from linear behaviour to eventual dynamic instability.
Although a curvilinear segment is always present, in some cases the transition can be
relatively long and gradual, while in other cases it can be very short and abrupt.
Typically structures with higher redundancy exhibit a longer transition segment, due
to redistribution of the inelastic deformation demands.
x A final linear segment that is horizontal, or nearly horizontal, in which infinitely large
lateral deformation demands occur at small increments in ground motion intensity.
This segment corresponds to the point at which a system becomes unstable (lateral
dynamic instability) and corresponds to the collapse capacity of the structure.

84% fractile (IM|DM)


Sa (T1 ) (g)

Instability
Softening

50% fractile (IM|DM)


16% fractile (IM|DM)

Pseudo
linear
Fc
Fy
Linear

Back-bone curve of
structure

y

collapse Drift ratio, 

cap

Fig. 2.3. Characteristics of median IDA curves (adapted from ATC-62, 2008)

2.2.6

Defining limit-States on IDA curves

For performance-based seismic assessments, the limit-states should be defined on the


IDA curves. Such limit states are often defined in terms of engineering demand
parameters such as drift ratios, plastic rotations, ductility demands, etc. Generally,
exceeding a structural limit-sate on the IDA curves can be identified by means of either

36

DM-based or IM-based rules (Vamvatsikos and Cornell, 2002). The DM-based rule states
that if the DM value exceeds a certain limit (e.g., the drift capacity at different limit states
such as serviceability, damage control, etc.) then the limit state is exceeded. This rule is
on the basis of the fact that DM is a measure of structural damage and when it exceeds a
certain limit the structure fails to satisfy the corresponding limit-state (i.e., the expected
structural damage is higher than that acceptable for the limit-state). Such limits on DM
values for different limit-states can be obtained through experiments, theory, or
judgment. An example of such limits is the drift limits recommended by Dutta and
Mander (1998) for bridge columns at different damage states which are mainly based on
engineering experience and judgement. Such capacity limits may not be deterministic, but
be defined using a probability distribution function which are typically reported in terms
of the central values and dispersions. An example of this case is the empirical equations
developed by Berry and Eberhard (2007) to predict the median and standard deviation of
the drift capacities of bridge columns at different damage states. When the DM-based
rule is used to define the limit states, in some cases multiple points on the IDA curve may
be found that satisfy a limit state. Such cases mainly occur, when the IDA curve exhibits
the weaving or hardening behaviour. In these cases often the point with the lowest IM
value that satisfies the limit sate is selected as the capacity point (e.g., see Fig. 2.4). The
use of the DM-based rules is simple and is recommended for the limit states other than
collapse. In the case of collapse, where the IDA curves flatten and the DM values become
infinite, the application of the DM-based approach is not often possible. This is because
in this case a wide range of DM values may correspond to only a small range of IM
values.

By generating the IDA curve for each record and defining the limit state capacities for
each IDA curve, the IM given DM value (set as the DM value at the limit state capacity)
can be obtained for each curve. The limit-state capacities can then be summarized into
central values (e.g., the mean or the median) and a measure of dispersion (e.g., the
standard deviation, or the difference between two fractiles).

37

The IM-based rule is used mainly for the prediction of the collapse capacity. For this
purpose the minimum IM value at which dynamic instability occurs is determined from
the IDA curves and is defined as the collapse capacity of the structure. However in this
case each IDA curve should be treated individually, since each IDA curve exhibit a
distinct collapse capacity .This is contrary to the DM-based rule that a single drift limit
(or any other DM parameter) is defined for all IDA curves at each limit-state.

Other limit states have also been used to determine the state of collapse. An example is
the FEMA-350 (2000), 20% tangent slope approach in which the last point on the curve
with a tangent slope equal to 20% of the elastic slope is determined to be the capacity
point. The idea is that the flattening of the IDA curve is an indicator of impending
dynamic instability with the DM increasing at higher rates and approaching infinity
(Vamvatsikos and Cornell, 2002). This limit state is also determined using an IM-based
rule. An example of defining different limit-state points (e.g., serviceability, damage
control, and collapse) on the IDA curves is shown in Fig. 2.4. The predicted drift
capacities are only specific for the case considered here.

Spectral acceleration at T1 (g) (IM)

Sa collapse = 2.78 (g)

2.5
IM values
corresponding to
damage control
limit-state for
different IDA curves

2
Sa collapse = 1.45 (g)

1.5

IM values
corresponding
to servicability
limit-state

1
0.5

Sa collapse = 0.66 (g)


Servicability drift capacity=0.09%
Damage control drift capacity=3.9%

0
0

0.02

0.04
0.06
0.08
Maximum column drift ratio (DM)

Fig. 2.4. Prediction of different limit states on IDA curves

38

0.1

2.3 Different IDA procedures and algorithms


Depending on the purpose of structural analyses several algorithms can be used to
perform IDA. Since the application of IDA often involves numerous inelastic dynamic
analyses for each record, such analyses are often time-consuming and can take several
hours to several days depending on the structure under analysis and the available
computational resources. Therefore the use of time-efficient algorithms to perform IDA
can be very beneficial and significantly decrease the computation time. The following
algorithms are often used to run IDA analyses:

2.3.1

Regular IDA algorithms

The simplest way to perform the IDA is to scale the records incrementally at uniform
steps. For this purpose the analysis for each record starts with a small IM value and ends
when the maximum IM value of interest is reached. For each step the IM value will be
increased by a fixed value, IM. The value of IM should be chosen such that a sufficient
number of points can be obtained for the developments of the IDA curves. If the IM is
large, the analysis can be performed faster, but the IDA curves may not be interpolated
with acceptable precisions. An optimum IM should be chosen to have a balance between
the required precisions in the prediction of the structural responses and the computation
time. Another important issue is the estimation of the maximum IM values which can
significantly affect the computation time. For example, if the objective of the IDA is the
prediction of the collapse capacity of structure for each record, a large value should be
assigned to the maximum IM to make sure that the ultimate collapse capacity for each
record can be determined. As a result, for all ground motion records the analyses should
be carried out up to the maximum IM value, while for many of them the structural
collapse occurs in much smaller IM values. This results in many redundant and
unnecessary analyses which increase the computation time.

39

2.3.2

Time-efficient algorithms

Time-efficient algorithms can be developed to significantly reduce the computation time


and to increase the precisions in the prediction of the seismic responses. For example the
analyses can start with larger increments of IM and the size of the increments can be
adjusted in the subsequent steps based on the results obtained in the previous steps. The
IM increments, IM, can be significantly decreased at the IM values near collapse.
Accordingly the analyses for each record can stop when the structural collapse occurs or
when the ultimate damage measures of interest are determined. The use of such advanced
algorithms, such as the hunt & fill algorithm (Vamvatsikos and Cornell 2002), ensures
that the record scaling levels are appropriately selected to minimize the number of
required runs and the required precisions in prediction of the seismic responses are
attained at the lowest computation costs. The extra efforts required in developing the
more complex time-efficient algorithm will be paid back with significantly lower
computation time, especially when a large number of records (especially when the
duration of records are also long) are used in IDA, or when a large number of structures
needs to be evaluated using IDA.

2.3.3

ATC-63 procedure

In the ATC-63 provisions only the median collapse capacity of the structure should be
determined and thus the computation of the full IDA curves is not necessary. For this
purpose first the records are normalized by their respective peak ground velocities (to
remove unwarranted variability due to different characteristics of records such as
magnitude, distance, and soil type) and then all the records are collectively scaled up or
down until 50% of the records cause structural collapse. The spectral acceleration, at
which 50% of the records cause collapse, is the median collapse capacity of the structure
and then will be used to compute the collapse margin ratios for the seismic performance
assessment of the structure.

40

In the ATC-63 procedure the collapse capacity of the structure for each of the 44 records
is unknown (i.e., at least 50% of the records causing dynamic instability) and only the
median collapse capacity of the record set is determined. Thus a direct evaluation of the
variability in collapse prediction is not possible using this method. In fact, the ATC-63
provisions use a presumed value of the logarithmic standard deviation of 0.4 to reflect the
record-to-record variability. It must be noted that often the record-to-record variability is
simply estimated using the elastic response spectra of the records (e.g., ATC63
provisions). Although this may provide a simple and rough estimate of the actual
variability in the prediction of the collapse capacities, it does not take into account
problems associated with the duration of the records, strength and stiffness degradation
and associated period elongation, and other important aspects that can only be considered
through performing nonlinear dynamic analyses.

2.3.4

Fast IDA procedures

Other algorithms to perform IDA can also be developed, based on the objectives of the
structural analyses. Often the most important outcome of the IDA is the prediction of the
median collapse capacity which will be used in the probabilistic performance based
assessment of the structures. Therefore for prediction of collapse capacities the structural
analyses can be started at higher intensities in which structural collapse is expected. The
initial estimates can be predicted using a pushover analysis or can be chosen based on
judgement. A good prediction of the initial estimate of the median collapse capacity will
minimize the computation time of the IDA. Such algorithms are especially helpful when
the records with long durations (e.g., subduction interface and inslab ground motion
records) or a large number of records are used in the IDA. However for such fast IDA
algorithms the full IDA curves are not available, since the analyses are performed only at
high ground motion intensities that can cause structural collapse. If the prediction of the
standard deviation of the results is also required, the collapse capacity of the structure for
each record are predicted using the fast IDA algorithm and the median and the dispersion
of the results can be estimated.

41

Such fast algorithms can be performed even faster if the prediction of the dispersion of
the results is not required. An example is the method recommended by the ATC-63
provisions which only requires the prediction of the median collapse capacity, while a
presumed value of 0.4 is used as the logarithmic standard deviation of the results due to
record-to-record variability. For such cases the use of all records in all steps will not be
required. For example if a record cause structural collapse at a certain intensity level, it
can be assumed that it will cause structural collapse at a higher intensity level as well
(i.e., it is assumed that structural resurrection or other bizarre cases in IDA is not
applicable). Similarly if a record does not cause structural collapse at a certain intensity
level, it will not cause collapse at lower intensity levels. Therefore such records can be
removed from the IDA in the subsequent steps and IDA can be performed much faster.

2.4 Structural modelling for IDA


In IDA the intensity of the ground motion records are increased until they cause structural
instability which is referred to as structural collapse. Therefore the structural models in
the IDA should be capable of simulating structural collapse and should directly or
indirectly consider all significant deterioration modes that contribute to collapse
including the stiffness, strength, and inelastic deformations under reversed cyclic loading.
The most important structural parameters that influence the IDA predictions include the
plastic deformation capacity, cap, and the post-capping rotation capacity, pc (Ibarra et
al., 2005). These parameters are used to define a component backbone curve, as shown in
Fig. 2.5.

42

Moment (M)

My

Mc

Ks

pc
Ke
Kc

y

cap
Chord rotation ()

Fig. 2.5. Backbone curve parameters (adapted from ATC-63 (2008))

Fig. 2.6. Differences between cyclic and in-cycle strength degradation (from FEMA
P440A (ATC-62), 2008).
In combination with stiffness degradation, structural components and systems may
experience in-cycle strength degradation (Fig. 2.6). In-cycle strength degradation is
characterized by a loss of strength within the same cycle in which yielding occurs. As
additional lateral deformation is imposed, a smaller lateral resistance is developed. This
results in a negative post-yield stiffness within a given cycle. In-cycle strength
degradation can occur as a result of geometric nonlinearities (P-delta effects), material

43

nonlinearities, or a combination of these. In reinforced concrete components, material


nonlinearities that can lead to in-cycle strength degradation include concrete crushing,
shear failure, buckling or fracture of longitudinal reinforcement, and splice failures
(FEMA P440A (ATC-62), 2008).

The distinction between cyclic and in-cycle degradation, as shown in Fig. 2.6, is
important because the consequences of each are vastly different. Dynamic response of
systems with cyclic strength degradation is generally stable, while in-cycle strength
degradation can lead to lateral dynamic instability (i.e., collapse). In modelling for IDA
the in-cycle strength degradation is very important for prediction of the collapse capacity
and the global dynamic instability of the structures and is often considered by the
negative post capping stiffness defined in the component back-bone curves. (e.g., see Fig.
2.5).

Modern bridges are designed and detailed to meet the seismic code requirements for
ductile response, including capacity design concepts and adequate support lengths at the
abutments. The ductile columns contain code-compliant spiral reinforcement to confine
the concrete, avoid shear failure and to control buckling of the vertical reinforcing bars.
For continuous bridges, with all other failure modes avoided, the flexural response
governs the response of the bridge and sidesway collapse is the governing collapse
mechanism. While the columns will undergo a ductile inelastic response during a major
event, the bridge superstructure is designed to remain elastic and is typically modelled
using elastic elements.

When structures with poor detailing are studied or when other collapse modes are
probable, such degrading collapse modes should be either considered directly in the
structural models or be treated indirectly through non-simulated component limit state
criteria. Such degrading modes may include shear failure or failure due to degradation of
shear capacity at high ductility levels (e.g., interaction of shear and flexure which is
referred as to shear-flexure or ductile shear failure). Including code-conforming
confinement reinforcement in the plastic hinge region of bridge columns will prevent

44

such undesirable failure modes. Where such collapse modes cannot be included directly
in the models, often they are considered indirectly by imposing some drift limits
corresponding to such collapse modes. That is, after performing the IDA, the collapse
capacities can be modified for other collapse modes such that the corresponding IM
values to the non-simulated collapse modes (in terms of DM) will be considered as the
collapse capacity of the structure. However this simplification in modelling through
application of the non-simulated component limit state criteria should be accounted for
by increasing the modelling uncertainty in the overall seismic performance assessment of
the structure. The model proposed by Elwood (2004) is an example of the models that
can predict the drift corresponding to shear failure and axial failure of columns with
insufficient transverse reinforcement.

The influence of the abutments on the seismic response of the bridges can be included in
structural modelling. However the effects of abutments may be conservatively neglected,
if they do not significantly influence the seismic response. An alternative approach may
be analyzing the bridges for two cases of restrained and unrestrained movements at the
abutments to evaluate the seismic response of the structure and to recognize the
controlling case. For example in the transverse response often shear keys are used to
restrain the transverse movements of the superstructure at the abutments. The shear keys
may either be designed based on the capacity design concepts so that they will survive
under high ground motion intensities or they can be designed as fuses to fail at low
ground motion intensity levels to prevent damage in the abutments and piles. Therefore
according to the design philosophy for such elements different modelling approaches may
be accepted. The most direct approach would be to include the degrading behaviour and
failure of shear keys in the structural models. The shear key models by Megally et al.
(2003) can provide the required information for this purpose. The spring abutment model
developed by Aviram et al. (2008) can be used when the effects of abutments are directly
considered in modelling. In this abutment model the effects of all important abutment
parameters such as strength and stiffness due to back-fill soil, abutment wing wall and
back wall, piles, bearing pads, shear keys and gaps are considered in the response.

45

The effects of expansion joints can be considered using gap elements in the structural
models. For example for the case of the continuous bridges, the expansion joints are
typically situated at the end abutments and can be considered by means of gap elements
in the abutment models. If cap beams are present then the cap beams and beam-column
joints are designed as capacity protected elements (design to transmit the probable
resistance of the columns) and therefore are modelled as elastic elements. However this is
not applicable for the case of the single column bridges studied in this research which are
assumed to be hinged at the top and fixed at the base.

2.5 Scaling of records in IDA


Due to the limited number of strong records available that can cause significant damage
(or collapse) in modern structures, the use of scaled records in IDA and nonlinear
dynamic analyses of structures is often inevitable and is widely used in practice and
research. However, concern is often expressed about the validity of structural analysis
results obtained using scaled ground motion records.

The use of an excessive scaling factor (e.g., factor of 10 or greater) could induce
significant bias in the predicted structural responses (Luco and Bazzurro, 2007). There is
a large growing body of literature concerning questions related to scaling of ground
motions, so that it is difficult to provide a simple answer in this regard. However it can be
concluded that in general the legitimacy of the scaling procedure depends on the
structure, the choice of the DM, and the choice of the IM (Vamvatsikos and Cornell,
2002).

The similarity of the response spectral shape of a record to the target response spectrum
has been shown to be an important criterion in selecting records which result in unbiased
response predictions. For moderate period structures with maximum drift ratio taken as
the DM and Sa(T1) as the IM and for a general class of records (moderate to large
magnitudes, M) the use of scaling typically results in similar predictions with those
obtained using un-scaled records. On the other hand where the response of the structure is

46

dominated by the first mode, the use of PGA as the IM can result in biased structural
response predictions, since the average ratio of Sa(T1) to PGA changes with magnitude
and therefore the spectral shape effects cannot be well captured when PGA is used as the
IM. In fact if the IM has been chosen such that the regression of DM jointly on IM, M
and R is found to be effectively independent of magnitude and distance (in the range of
interest), scaling of records will provide good estimates of the distribution of DM given
IM (Vamvatsikos and Cornell,2002). It has been demonstrated that the use of scaling in
cases where the records are selected based on methodologies that account for the spectral
shape effects (such as the epsilon-based method and CMS-based method), result in
unbiased predictions similar to those obtained using un-scaled records (Baker and
Cornell, 2005 and 2006a).

2.6 Prediction of damage states


In order to determine the capacity of a structure (e.g., in terms of IM parameters) at
different limit states, some criteria are needed to determine the state of damage (i.e., onset
of spalling, bar buckling, etc.) at different DM values. Such criteria may not be
determined deterministically, but be reported as random variables (i.e., the median drift
capacity and a standard deviation). Such models are also referred as the capacity models
which predict the probability of exceeding a damage state given DM. For example if the
maximum drift ratio of columns is used as DM, some criteria on drift limits should be
determined to predict the state of damage in the structural elements at various drift ratios
(i.e., drift capacity corresponding to cover-spalling, or drift capacity corresponding to bar
buckling, etc.). Such models can be developed either empirically (e.g., experiments,
experience, etc.) or theoretically (e.g., using cross-sectional analysis). Such data are
usually provided in terms of engineering demand parameters (such as tensile and
compressive strain limits, ductility capacity, drift capacity, etc.) for different damage
states (e.g., yielding, cover spalling, bar buckling, bar fracture, and collapse). Such
information is also required in structural modelling to define the back-bone curves of the
response of structural elements (e.g., to determine deformation capacity, cap, post-

47

capping rotation capacity, pc ,etc.) as shown in Fig. 2.5. Such information is available
from several experimental and theoretical studies as discussed below.

2.6.1 Study by Dutta and Mander (1998)


Dutta and Mander (1998) recommended some simple drift limits for bridge columns at
different damage states consistent with those defined in HAZUS (1999). These limits are
given for two cases of seismically and non-seismically designed bridges. Such estimates
of the structural damage at different drift limits provide good information for the
preliminary evaluation of bridges. However the use of such constant drift limits for a
large variety of bridge columns with different geometric and mechanical characteristics
may not be appropriate where more precise seismic evaluations are carried out. The
ultimate drift capacity of columns in such cases should be defined explicitly as a function
of important parameters that influence the deformation capacity. The recommended drift
limits along with the expected damage at each damage state are summarized in Table 2.1.

48

Table 2.1. The recommended drift limits by Dutta and Mander (1998) at different damage
states

2.6.2

Damage State

Description of Bridge Damage


States (HAZUS 99)

Drift Limits
(non-seismically
designed)

Drift Limits
(seismically
designed)

No

No damage to a bridge

0.005

0.008

Minor/ Slight

Minor cracking and spalling to


abutments, hinges, columns or
minor cracking to the deck

0.007

0.01

Moderate

Any column experiencing moderate


cracking and spalling (column
structurally still sound), any
connection having cracked shear
keys or bent bolts, or moderate
settlement of the approach

0.015

0.025

Major/Extensive

Any column degrading without


collapse (column structurally
unsafe), any connection losing some
bearing support, or major settlement
of the approach

0.025

0.05

Complete
Collapse

Any column collapsing and


connection losing all bearing
support, which may lead to
imminent deck collapse

0.05

0.075

Theoretical method by Priestley et al. (2007)

Structural damage can be defined as a function of the maximum compression strain in


the confined core concrete and the maximum tension strain in the reinforcing bars. The
strain limits then can be converted for simplicity to displacements, drifts, and rotations,
which are widely used by the engineers, using appropriate relationships between strain,
curvature and displacement assuming an appropriate plastic hinge length in yielding
elements.

Serviceability limit state is usually exceeded when the concrete cover outside the
confined core starts spalling or crack widths are larger than allowable limits. The spalling
49

state is determined according to the maximum compression strain in unconfined concrete


cover which is usually limited to a value in a range of 0.003 to 0.004. To control the
crack widths, the maximum strains in the reinforcing bars should not exceed an
appropriate limit. In ordinary environments a crack width of 1.0 mm is typically
appropriate and the corresponding maximum tension strain to this crack width is around
0.015 for column or wall elements (Priestley et al., 2007).
The ultimate compression strain in the confined core is mainly a function of degree of
confinement and the quality of detailing in the plastic hinge regions. Several empirical
equations are available in the literature to estimate the ultimate compression strain in the
confined concrete core, among them is the widely used Mander equation (Mander et al.,
1988). The Mander et al. equation is based on equal energy principles in which the
ultimate compression strain corresponds to the fracture of hoops. Beyond this level,
crushing of the concrete, buckling, and fracture of the steel bars are unavoidable and the
structural elements must be usually replaced, since repairing is not economical.
Experimental studies however indicate that he ultimate strains predicted by Mander
equation are conservative in which the actual ultimate strains exceed the predicted values
by a factor of about 1.3 to 1.6. This conservatism in the Mander et al. equation is mainly
due to ignoring the combined effect of axial compression and flexure while the original
Mander et al. model was derived based on the assumption of pure axial compression of
the members such as foundations or columns. To account for this, Priestley et al. (2007)
recommend using the ultimate strains based on this equation for damage control limit
state for which this degree of conservatism is deemed to be appropriate. However for the
Life Safety (LS) limit state the predicted values from Mander et al. equation may be
increased by 50% to correspond to the experimental results.

For the damage control tension strain limit, the ultimate strain in steel bars, su, from
the monotonic tests cannot be used directly for moment curvature analysis and the
calculation of the ultimate curvature. Due to the possibility of buckling of the bars when
subjected to reversed loading and considering low-cycle fatigue of the reinforcing bars in
addition to slip between the reinforcing steel and the concrete, the ultimate tension strain

50

from the monotonic tests must be modified. The level of this strain will depend on the
volumetric ratio and longitudinal spacing of the transverse reinforcement. Usually a strain
level of 0.6 to 0.7 times the ultimate strain of the steel bars in monotonic tests is
recommended for calculating the ultimate curvature of the section which should not be
taken larger than 0.05 (Priestly et al., 1996 and 2007). In order to attain this level of
strain, the spacing of the transverse reinforcement must be code conforming. For the life
safety performance level the value of 0.9 times the ultimate strain of steel bars, 0.9 su, is
recommended by Priestley et al. and this value should not be larger than 0.08 (Priestly et
al., 2007). The strain limits for different damage states are summarized in Table 2.2.

Table 2.2. Maximum strain in concrete and steel at different damage states (Priestley et
al., 2007)

2.6.3

Damage state

Rebar tension strain ( s)

Concrete comp. strain ( c)

Yielding

0.002

0.002

Serviceability

0.015

0.004

Damage Control

0.6su (0.05)

c< Mander eq.

Life Safety

0.9su(0.08)

c< 1.5*Mander eq.

Equations by Berry and Eberhard (2007)

A study by Berry and Eberhard (2007), based on experimental data from the PEER
Structural Performance Database (Berry et al., 2004), provides some empirical equations
to estimate the engineering demand parameters (EDP) including drift ratio, plastic
rotation, and strain in the longitudinal bars for circular bridge columns based on the
properties of columns including the longitudinal and transverse steel ratio, the axial load
ratio, and the geometry. Such equations are provided for different damage states
including spalling, reinforcing bar buckling, and bar fracture. The equations given for
different damage states as a function of different EDP are summarized in Table 2.3. The
bar buckling state can represent the point at which strength degradation starts to define

51

capacity of the columns. Few test results are available to calibrate the post-capping
stiffness of the columns. Mackie and Stojadinovic (2007) provide an equation for
estimation of the drift at failure, using the same test data used by Berry and Eberhard.
The equations by Berry and Eberhard for spalling, bar buckling, and bar fracture damage
states along with the equation by Makie and Stojadinovic (2007) to predict the drift at
!eff= fys s,trans / fc, is the effective
#$''$!s,trans is the volumetric transverse reinforcement ratio, fys is the yield
stress of the transverse reinforcement fc is the concrete compressive strength, db is the
diameter of the longitudinal reinforcement, D is the column diameter, P is the axial load,
Ag is the gross area of the cross section, and L is the distance from the column base to the
point of contraflexure.

Table 2.3. Summary of the equations by Berry and Eberhard (2007) for cover-spalling,
bar buckling, and bar failure, and equation by Mackie and Stojadinovic (2007) to estimate
drift at failure
Measured/Calculated
coefficient of
mean
variation

Damage State

EDP

Equation

Cover
Spalling

sp+>?@\

1.6 (1-P/Agf 'c ) (1+L/10D)

1.07

0.35

p-sp
sp

1.2
0.008

0.98
0.99

0.34
0.45

bb+>?@\

3.25 (1+150eff db / D)(1 - P / Agf'c )(1 + L / 10D)

1.01

0.25

Bar Buckling

Bar Fracture

Failure

'

p-bb

2.75 (1+150eff db / D)(1 - P / Agf c )(1 + L / 10D)

1.01

0.24

bb

0.045+0.25eff 

1.00

0.24

bf+>?@\

3.5 (1+150eff db / D)(1 - P / Agf'c )(1 + L / 10D)

0.97

0.20

p-bf

3.0 (1+150eff db / D)(1 - P / Agf'c )(1 + L / 10D)

0.97

0.20

bf

0.045+0.30eff 

0.96

0.21

0.35

ff+>?@\

16 + 0.71 L - 15D-3.8e - 5f'c - 2.8e-5fy + 1.7s,long


`{}s,trans - 12P/Agf'c

52

2.6.4

Study by Fardis and Biskinis (2003)

Panagiotakos and Fardis (2001) have provided some empirical equations to estimate the
yield and ultimate deformations of reinforced concrete members with rectangular crosssections. Fardis and Biskinis (2003) then provided a more complete set of equations for
different structural elements such as columns (with rectangular and circular sections) and
walls. Such equations are given for the case of different collapse modes such as flexure
and shear-flexure failure modes and for different loading types including monotonic and
reversed cyclic loadings. The ultimate deformation in these equations was considered as
the point with 20% strength degradation. Therefore these equations may be more
appropriate for the case of conventional seismic evaluations based on the seismic codes
where such definitions for the ultimate displacements have been used. However, for the
purpose of structural modelling in IDA, where the prediction of the cap point (i.e., point
where strength degradation begins) is required, some back calculations may be required.
Such calculations may not be straightforward since the post-capping stiffness is not given
in these studies. Nevertheless such equations provide an excellent source of information
for structural modelling and evaluation.

2.6.5

Study by Haselton et al. (2007)

Haselton et al. (2007) provide a comprehensive set of equations for the purpose of
modelling and evaluation of the reinforced concrete structural members. This study
provides equations to define the important modelling parameters in the backbone curves
(e.g., the capping deformation and post capping stiffness, etc.) as a function of the
geometric and mechanical characteristics of the structural members. These equations
were primarily provided for the purpose of structural modelling in IDA and were also
included in the ATC-63 provisions. Furthermore, equations are given to predict the
hysteresis parameters based on a sophisticated hysteresis model by Ibarra et al. (2005).
This study provides all the necessary information for structural modelling in IDA.
However the application of such equations for the case of circular bridge columns may

53

not be appropriate, since only the experimental data from the members with rectangular
cross-sections were used to derive the equations.

2.7 Treatment of uncertainties in IDA

2.7.1

Types of variability

Each source of variability in general can be classified into aleatory (randomness) and
epistemic (uncertainty) types of variability. The random nature of ground motions
(known as record-to-record variability) is considered as aleatory variability. The
epistemic variability in the prediction of collapse capacity is primarily due to lack of
knowledge about the ground motion hazard model and the structural model. Accurate
estimation of the epistemic uncertainties (including the modelling uncertainty) is often
time-consuming and involves extensive application of Monte Carlo simulations or the
FOSM (First Order Second Moment) method (Liel et al., 2009, and Zareian and
Krawinkler, 2007). The ATC-63 provisions provide a simplified judgmental method to
estimate total uncertainty in prediction of the collapse capacity which may be used when
such information is not available.

To de-convolve the effects of epistemic and aleatory uncertainties on the collapse


probability, it is usually assumed that the median collapse capacity is a random variable
and the dispersion due to record-to-record variability is independent of the dispersion due
to epistemic uncertainty in the structural model. Based on this assumption, collapse
capacity predictions can be carried out using the ground motion records applied to the
structural mathematical model with the properties of elements set to their median values.
Using such an approach provides the collapse fragility curve with median value of the
collapse capacity, c, and the dispersion due to record to record variability, RTR
(randomness in collapse capacity). In order to incorporate the effect of epistemic
uncertainty, it is assumed that the median of the collapse capacity, c, is a random

54

variable which is lognormally distributed with median (median estimate of median of


collapse capacity) K c and dispersion of epistemic (Zareian and Krawinkler, 2007).

Combining the effects of different sources of uncertainty can be carried out using
different approaches, including the confidence interval approach and the mean estimate
approach. The use of the confidence interval method results in predictions that are highly
dependent on the confidence interval chosen. Therefore the use of the mean estimate
approach is preferred, since it provides a more practical and consistent approach for this
purpose (see Liel et al., (2009), and Zareian and Krawinkler (2007) for more details).
When aleatory (record-to-record) uncertainties only are considered, the structural
response is well-described by a lognormal distribution (Cornell et al., 2002), with
median, c, and logarithmic standard deviation (RTR). In the mean estimates approach, it
is assumed that the epistemic (e.g., modeling) uncertainty describes uncertainty in the
median collapse capacity (c), and that this random variable is also lognormally
distributed with median value of K c and lognormal standard deviation of epistemic. The
random variables associated with epistemic and aleatory uncertainty are assumed to be
independent. It can be shown that when these two distributions are combined the
resulting distribution is also lognormal with the median value of K c and a logarithmic
variance that is the sum of the two logarithmic variances (i.e. ETOT

2
2
E RTR
 Eepistemic
)

(Benjamin and Cornell, 1970). Thus, when the mean estimates approach is used, the
median is unchanged when modeling uncertainties are incorporated, but the variance
increases. The results derived from this approach are not sensitive to whether individual
uncertainties are classified as aleatory or epistemic, which is helpful when the
classification of a particular uncertainty is not obvious.

2.7.2

Sources of uncertainty and combining uncertainties

Many sources of uncertainty contribute to total variability in collapse capacities predicted


using IDA. The main sources of uncertainty include (ATC-63, 2008):

55

1-Record-to-Record (RTR) Uncertainty. This uncertainty is due to variablity in the


sesmic response of structures to different ground motion records. The variability in
response is mainly due to different frequency content and dynamic characterstics of the
records, variablity in the hazard characterization of the ground motions records, and
different duration of records used in the IDA.

2- Design Requirements-Related (DR) Uncertainty. Such uncertainty is associated with


the quality of the design requirements.

3- Test Data-Related (TD) Uncertainty. This type of uncertainty is related to the quality
of the test data used to define the structural system.

4- Modelling (MDL) Uncertainty. This uncertainty is associated with the quality and
robustness of the nonlinear structural models and the ability to accurately simulate the
nonlinear behaviour of the structural elements subjected to seismic excitations.

The total uncertainty can be predicted by combining RTR, DR, TD, and MDL
uncertainties. The collapse fragility of the structure can be estimated by multiplying the
predicted median collapse capacity from IDA, to a random lognormal variable such as
~TOT with a median value of unity and a lognormal standard deviation of TOT. This
lognormal random variable, ~TOT, in turn consists of four independent lognormal random
variables such that, ~TOT= ~RTR ~DR ~TD ~MDL, where ~RTR, ~DR, ~TD, and ~MDL are
lognormal random variables with a median value of unity and a lognormal standard
deviation of RTR, DR, TD, and MDL respectively. Since these random variables are
assumed to be statistically independent, the total uncertainty in prediction of the collapse
capacity, TOT, described in terms of lognormal standard deviation can be computed using
Eq.[2.1].

[2.1] ET OT

2
E R2T R  E DR
 ET2D  E M2 DL

56

Based on the ATC-63 provisions the uncertainty due to record-to-record variability is


taken as RTR = 0.40 in all cases. However this is based on the period range of building
structures and considering only crustal earthquakes. For bridges or other structures RTR
can be directly estimated from the IDA results. Quality ratings for design requirements,
test data, and nonlinear modeling are translated into quantitative values of uncertainty
based on the following scale: (A) Superior,  = 0.20; (B) Good,  = 0.30; (C) Fair,  =
0.45; and (D) Poor,  = 0.65 (ATC-63, 2008). More details concerning the criteria to
determine the quality ratings for different sources of uncertainty are available in ATC-63
(2008). Other studies have also used similar values for uncertainty. For example in a
study by Mander et al. (2007) the uncertainty due to modelling and predicting the
capacity of the bridge columns were considered as 0.25 and 0.2 respectively.

2.8 Development of fragility curves using IDA results


IDA results are used for the development of the fragility curves of structures for different
limit-states. The fragility curve gives the conditional probability that a certain limit-state
be exceeded (i.e., probability of failure) at a given IM value.
Fractile IDA curves can be used to draw fragility curves. For example if the fragility
curves for the collapse state needs to be developed, at each level of IM the number of
IDA curves with flat-lines (i.e., indicator of the global dynamic instability) can be
counted and the corresponding probability of collapse can be simply calculated as the
ratio of the number of IDA curves with flat-lines to the total number of IDA curves (i.e.,
the number of records caused collapse at intensities lower than chosen IM divided by the
total number of records used in IDA).

For example in Fig. 2.7 the corresponding probability of failure at different IM levels
(i.e., Sa(T1) is used as IM) are derived from the IDA results shown in Fig. 2.2. The
fragility curves were developed for collapse and cover-spalling damage states. The coverspalling state was defined as exceeding the drift ratio of 1.85% determined using the
equations by Berry and Eberhard (2007). The fragility curves developed using IDA data
are shown by dots, in which each dot represents the number of records that caused failure

57

to the total number of records at different IM levels. In Fig. 2.7 the cumulative lognormal
distribution curves developed using the median and standard deviation of the IDA results
at collapse and cover-spalling limit states are also shown. As can be seen the fragility
curves using IDA data can be well estimated using such curves.

1
0.9

Probability of failure

0.8
0.7
0.6
0.5
0.4
Collapse fragility curve (IDA data)
Spalling fragility curve (IDA data)
Collapse fragility curve (Idealized lognormal)
Spalling fragility curve (Idealized lognormal)

0.3
0.2
0.1
0
0

0.5

1.5
Sa(T1) (g)

2.5

Fig. 2.7. Development of fragility curves for different limit states using IDA results

Therefore assuming that the data is lognormally distributed, it is possible to develop the
fragility curves at collapse (or any other limit-state) by computing only the median
collapse capacity and logarithmic standard deviation of the IDA results at collapse. The
fragility curves then can be analytically computed using Eq.[2.2]:

[2.2] P (failure | S a

x)

C
ln( x )  ln(Sa50%
)
)

E RTR

where (.) is the cumulative normal distribution function, Sa

50%

is the median capacity

determined from IDA, and RTR is the record to record variability in the IDA results.
In the IDA results only the uncertainty due to record-to-record variability has been
considered. Therefore it is necessary to modify the fragility curves developed using the
58

IDA results to include the uncertainty due to other sources of uncertainty as discussed
before (i.e, DR, TD, and MDL). Hence it is possible to inflate the fragility curves using
the mean estimate approach (as discussed above) to include the total uncertainty in
predictions. The modified fragility curve then can be computed using the following
equation:

[2.3]

P ( failure | S a

x)

C
ln( x )  ln(Sa50%
)
)

ET OT

where TOT is the total uncertainty as defined by Eq. [2.1]. Uncertainty influences the
shape of a fragility curve plotted from the results of IDA. Fig. 2.8 shows collapse fragility
curves reflecting different levels of uncertainty (i.e., TOT = 0.4, 0.65, and 0.9). As
indicated in the figure, additional uncertainty has the effect of flattening the curve.
While the median collapse intensity is unchanged, additional uncertainty causes a large
increase in the probability of collapse at the lower IM levels such as the IM level
corresponding to the maximum considered earthquake (MCE).

1
=0.4

Probability of failure

0.9
=0.65

0.8
0.7
0.6

=0.9

0.5
0.4
0.3
0.2

Median capacity =1.31 g

0.1
0
0

0.5

1.5

Sa (T1) (g)

Fig. 2.8. Effect of uncertainty on the shape of fragility curves

59

2.5

2.9 Seismic risk assessment


Seismic risk can be expressed as the potential economic, social and environmental
consequences of seismic events that may occur in a specified period of time. In order to
quantify the seismic risk associated with a certain structure, one needs to combine two
important elements of seismic hazard and seismic fragility analyses. Seismic hazard is in
fact a description of the severity of the environment, in terms of the probability that a
certain limit of seismic intensity measure, IM, is exceeded in a certain period of time, T.
The determination of the seismic hazard is accomplished through using conventional
probabilistic seismic hazard analysis (PSHA). An example of a seismic hazard curve in
terms of mean annual frequency of exceedance developed for Vancouver (site class C) at
T= 1 sec. using the updated seismic hazard data provided by Goda et al. (2010) is shown
in Fig. 2.9.

Frequency of exeedance

0.01

1E-3

1E-4

1E-5

0.1

Sa(T=1 sec) (g)

Fig. 2.9. Hazard curve for Vancouver at T=1 sec. using the updated seismic hazard data
by Goda et al. (2010)

In addition to seismic hazard, the influence of the occurrence of seismic events with
certain intensities on the structures should also be known. Fragility is in fact a conditional
probability of exceeding a limit state, given a level of seismic hazard. The fragility
analysis can be carried out either empirically using experimental studies or field

60

observations, or analytically through nonlinear dynamic analysis (such as IDA) of the


structural models. The result of the fragility analysis is often presented using fragility
curves which present the probability of exceeding a certain a limit state (such as collapse,
bar buckling, fracture of hoops, shear failure, etc.) as a function of the seismic intensity,
IM (e.g., Sa(T1,5%)).
Once quantified, the hazard and fragility functions are combined to produce the
probability of failure in a period of time, T. This is done by using the Total Probability
Theorem (Benjamin and Cornell 1970), which states that, if {B1,..., Bn} is a set of
mutually exclusive and collectively exhaustive events and A is any other event, then the
probability of A can be calculated using Eq. [2.4].

[2.4] P [A ]

P [A | B ] P [B ]
i

i 1

To use this result for seismic risk assessments, the seismic intensity, IM, is discretized
into n distinct levels, say I1, ..., In and the events A and Bi in Eq. [2.4] are taken as A
=The structure fails at least once in T (i.e., The structure exceeds the limit state at
least once in T) and Bi = IM= Ii. Exceeding a certain limit state (which is in turn
expressed as exceeding a certain damage measure, DM) is referred to as failure here. It is
often reasonable to assume that the structure of interest survives in T if it does not fail
under the highest seismic intensity, IM, experienced during that period of time, T
(Veneziano, 2005). Under such assumption, the probability of failure in T is obtained
from Eq. [2.4] as:

[2.5] P [at least one failure inT ]

P [failure | IM

I i ] P [IM

Ii ]

i 1

Eq. [2.5] shows how the hazard (probabilities P[IM= Ii]) and the fragility (the
probabilities P[failure| IM= Ii]) are combined in the assessment of seismic risk. The IDA

61

results can be used to develop the fragility curves which in turn will be used for seismic
risk assessment of the structures.

In some cases, one is not interested in the physical damage to a facility, but in the
consequences that such damage might have on the exposed population or the
environment. The risk should in this case be measured through the probability that a
consequence C (e.g., number of fatalities, repair cost, or down time) exceeds a certain
level c* in T years (Veneziano, 2005). The probability that C > c* at least once in T years
then can be evaluated through a second application of the total probability theorem, as
follows :

[2.6] P [C ! c ]
*

P [C ! c

|D

d j ] P [D

dj]

j 1

where d1,..., dm are m discretized levels of damage D. The probabilities P[D = dj] are
evaluated through repeated application of Eq. [2.5], each time defining failure as the
event D = dj, this will result in Eq. [2.7].

[2.7] P [C ! c * ]

P [C ! c

|D

d j ] P [D

d j | IM

I i ] P [IM

Ii ]

j 1 i 1

Whereas the probabilities P[C > c*|D = dj] are the result of a consequence model. This is
also known as the loss or performance model and the parameter C is also known as the
decision variable, DV, in the framing equation adopted by the Pacific Earthquake
Engineering Center (Cornell and Krawinkler, 2000) as will be discussed later.

In case the level of damage, dj, is expressed in terms of drift ratios, the (annual)
probability of exceeding a drift level dj, ~ (dj), can be calculated using Eq. [2.8]. ~ (dj) is
also known as the drift hazard curve (or demand hazard curves in general).

62

[2.8] O (d j )

P [D ! d

| IM

I i ] P [IM

Ii ]

i 1

In continuous, integral form Eq. [2.8] is expressed as :

[2.9] O (d )

IM f

P [D ! d | IM ] d O (IM )

IM 0

To compute PPL , the (annual) probability of the performance level not being met (e.g.,
the annual probability of collapse), P[C > c*|D = dj] in Eq. [2.7] will be the probability
that the drift capacity at limit state, Dc, is less than some level of D, which can be
c

expressed by P[DD |D=dj]. The term P[D  D | D = dj] can to a first approximation be
assumed to be independent of the information about the drift level itself (Cornell et al.
c

2002), so that this term can be written as P[D  D |D = dj]= P[dj  D ]. Therefore to
determine PPL, Eq. [2.7] can be rewritten as Eq.[2.10]:

[2.10]

PPL

P [d

t D C ] P [D

d j | IM

I i ] P [IM

Ii ]

j 1 i 1

From Eq. [2.8] we have P [ D

d j | IM

I i ] P [ IM

Ii ]

therefore:

[2.11] PPL

P [d

t D C ] O (d j )  O (d j 1 )

j 1 i 1

In continuous, integral form Eq. [2.11] is expressed as:

[2.12 ] PPL

d f

d 0

P [d t D C ] d O (d )

63

O (d j )  O (d j 1 ) and

2.10 Framework for performance-based earthquake


engineering (PBEE)
The power of IDA as an analysis method is put to use well in a probabilistic framework,
where the estimation of the annual likelihood of the event that the demand exceeds the
limit-state or capacity is required. This is the likelihood of exceeding a certain limit-state,
or of failing a performance level, within a given period of time. The concepts and
equations presented before can be generalized in a framing equation adopted by the
Pacific Earthquake Engineering Center (Cornell and Krawinkler, 2000), given in Eq.
[2.13].

[2.13]

O (DV )

G (DV

| DM ) |dG (DM | IM ) | | d O (IM ) |

Where in this equation IM, DM and DV are vectors of intensity measures, damage
measures and decision variables (such as the annual earthquake loss and/or the
exceedance of one or more limit states), respectively. G(DV|DM) is the probability that
the (vector of) decision variable(s) exceed specified values given that the engineering
damage measures are equal to particular values. This term is in fact equivalent to the term
P[C > c*|D = dj] in Eq. [2.7]. ~ (IM) is the conventional hazard curve (i.e. the mean
annual frequency of IM exceeding a certain value) and d~ (IM) is equivalent to P[IM= Ii]
in Eq. [2.7] presented before. Further, G(DM|IM) is the probability that the Damage
Measure(s) exceed certain values (corresponding to certain limit states) given that the
Intensity Measure(s) equal particular values .Therefore dG(DM|IM) is the differential of
the complementary cumulative distribution function (CCDF) of DM given IM. This term
is equivalent to the term P[D=dj| IM= Ii] in Eq. [2.7].
Examples of G(DV|DM) when the DV is a binary damage state indicator variable (i.e.,
the decision variable is simply a scalar indicator variable: DV=1 if the limit-state is
exceeded and zero otherwise) are various fragility curves and ~(DV) is the expected
annual loss in this case .

64

The term |dG(DM|IM) | is in fact obtained from the statistical characterization of the IDA
curves. Therefore the IDA produces precisely the information needed both for PBEE
demand characterization and for global collapse capacity characterization (Vamvatsikos
and Cornell, 2002).

When DV is an indicator variable, ~(DV) in Eq. [2.13] is simply the mean annual
frequency of exceeding a limit state (i.e., annual loss). Generally the prediction of ~(DV)
can be carried out using either a DM-based approach or IM-based approach. In the DMbased approach an intermediate variable DM is used (e.g., in Eq. [2.13]) so that the total
annual loss can be de-coupled into different damage state terms (Jalayer and Cornell,
2003). It is convenient to express the extent of damage and the associated financial loss in
terms of DM, which makes the use of DM as an intermediate variable in Eq. [2.13] very
useful. For example, where the predictions of the total financial loss are required, such
predictions can be provided for different damage states separately and the total financial
risk associated to the structure can then be estimated by summing the financial losses for
all damage states. For instance, the recommended loss ratios by Dhakal and Mander
(2006) for different damage states (DS) as defined by HAZUS (1999) are presented in
Table 2.4. The loss ratio is defined as the ratio of the cost of repair or
retrofit/strengthening needed to restore the functionality of the bridge to the total
replacement cost.

Table 2.4. Loss ratio for different damage states (Dhakal and Mander, 2006)
Damage

DS1

DS2

DS3

DS4

DS5

Assumed

0.1

0.3

Range

0.05-0.15

0.2-0.4

1.0-1.2

state

If the intermediate variable DM is integrated out, Eq. [2.13] can be rewritten as Eq. [2.14]
in an IM-based format.

[2.14]

O (DV )

G (DV

| IM ) | d O (IM ) |
65

For example in the case that DV is an indicator variable, the mean annual frequency of
exceeding a limit state
LS can be computed from Eq. [2.15] as:

[2.15]

OLS

O (DV

1)

G (IM

! IM C | IM ) | d O (IM ) |

where DV=1 when the demand IM exceed capacity in terms of IM, IMC. The first term in
Eq. [2.15] is in fact the probability that the capacity IMC is less than some level of IM,
which is equal to CDF of IMC (i.e., F(IMC|IM)). Eq. [2.15] can then be rewritten as:

[2.16] OLS

IM f

IM 0

F ( IM C | IM )

d O (IM )
dIM
dIM

F(IMC|IM) is in fact the fragility curve developed for the capacity of the structure in
terms of IM corresponding to a certain limit-state. For example in the case of the collapse
limit state, the IMC for each IDA curve is the collapse capacity, which is defined as the
IM at which the curves become flat (i.e., global dynamic instability of the structure).
Therefore F(IMC|IM) function can be predicted using IDA results, which is exactly the
percentile IDA curves developed in terms of IM. An example of the fragility curves
developed for collapse and spalling damage states are presented in Fig. 2.7.

Using integration by parts it can be shown that Eq. [2.16] can also be rewritten as Eq.
[2.17]:

[2.17]

OLS

IM

f

IM 0

f ( IM C | IM ) O ( IM )dIM

where f(IMC|IM) is the probability density function of capacity in terms of IM and ~(IM)
is the conventional hazard curve. This can be numerically computed by Eq. [2.18] using
the fragility and hazard curves. Where in Eq. [2.18], FR is the fragility function developed

66

for a damage state (e.g., cover-spalling, or collapse) and spectral acceleration, Sa, is used
as intensity measure, IM.

[2.18]

OLS

[F

All S ai

(S ai )  FR (S ai 1 )]O (S ai )

On the other hand, if some reasonable approximations can be made, Eq. [2.16] can be
analytically integrated (Shome and Cornell 1999, Cornell et al. 2002). It is only needed to
assume that the IM-values of capacity are lognormally distributed and that the IM-hazard
curve can be approximated by fitting a straight line in the log-log space, O(IM ) = k0IM-k,
either by a global regression, same for all limit-states, or by a local fit at the median IMcapacity for each limit-state (Vamvatsikos and Cornell 2004). Based on these
assumptions the integration in Eq. [2.16] will result in Eq. [2.19]:
[2.19]

OLS

C
).exp ( k S ln IM ) 2
O (IM 50%
2
C

where SlnIMc = (ln IMC50% - ln IMC16%) is (approximately) the standard deviation of the
natural logarithm of the IM-capacity and IMCx% is the x% percentile of the capacity (in
terms of IM) derived from the fractile IDA curves.

2.11 Ground motion prediction equations


A ground motion prediction equation (GMPE) (also known as the attenuation function) is
referred to a statistical model which estimates the mean and variance of ground shaking
with distance from an earthquake source. Such equations are often developed separately
for different tectonic regions (e.g., shallow crustal regions, subduction zones, etc.).

Different GMPEs may require different parameter to estimate the ground motion intensity
at a site of interest. Such parameters basically include earthquake magnitude, distance
and site condition (e.g., rock, soft soil). The soil type is considered more explicitly using

67

the average shear wave velocity, Vs30, in modern GMPEs. More sophisticated GMPEs
require more complex and detailed parameters including the fault type and mechanisms,
different measures for distance (such as Joyner-Boore distance ,Rjb, closest distance from
the site to the ruptured area, Rrup, etc.), aftershock or main shock, site on hanging wall or
foot wall, fault rupture width and length, hypocentral depth, etc. For crustal earthquakes
such information for different records are available from the PEER-NGA database at
http://peer.berkeley.edu/nga/flatfile.html. Such data can be used to estimate the mean and
variance of the ground motion intensity of the records at a site with a specific distance
and soil condition using the selected GMPEs. These mean and variance values are also
used to predict the epsilon, , parameter. Epsilon is an important characteristic of a
ground motion which is used in the record selection and prediction of the spectral shapes.

The list of GMPEs used by the Geological Survey of Canada (GSC) and the updated
model by Goda et al.(2010) are given by Atkinson and Goda (2011) which is
summarized in Table 2.5. More details about the seismic hazard results and the program
developed to extract the deaggregation of the seismic hazard are discussed in Appendix
C.

68

Table 2.5. Summary of the adopted ground motion prediction equations (GMPE) (Table
adapted from Atkinson and Goda, 2011)
Earthquake type

Shallow crustal
earthquakes in eastern
Canada

Source

Variable

Seismic hazard model implementation:


Weight

AB95 : Atkinson and Boore (1995)

M, r hypo, NEHRP
site class A

GSC 1995 model: [AB95 (1.0)]

SGD02 : Silva et al . (2002)

M, r jb , NEHRP

A08 : Atkinson (2008)

M, r rup, NEHRP
site class A
M, r rup, V s30,
stress drop
M, r jb, V s30

BJF97 : Boore et al . (1997)

M, r jb, V s30

A05 : Atkinson (2005)


HG07 : Hong and Goda (2007)

M, r rup, V s30
M, r jb, V s30

BA08 : Boore and Atkinson (2008)

M, r jb, V s30

YCSH97 : Youngs et al . (1997)

M, r rup, H ,
NEHRP site class

AB03 : Atkinson and Boore (2003)


Z06 : Zhao et al . (2006)

M, r rup, H , V s30
M, r rup, H , V s30

GA09 : Goda and Atkinson (2009)

M, r rup, H , V s30

YCSH97 : Youngs et al . (1997)

M, r rup, H ,
NEHRP site class

C03 : Campbell (2003)


AB06 : Atkinson and Boore (2006)

Shallow crustal
earthquakes in western
Canada

Inslab earthquakes in the


Cascadia subduction zone

Interface earthquakes in
the Cascadia subduction
zone

AB03 : Atkinson and Boore (2003)


Z06 : Zhao et al . (2006)

M, r rup, NEHRP
site class C or D
M, r rup, H , V s30
M, r rup, H , V s30

AM09 : Atkinson and Macias (2009)

M, r rup, V s30

GSWY02 : Gregor et al . (2002)

Updated model: [SDG02 (0.2) ,


C03 (0.3), AB06 (0.4) , A08 (0.1)]

GSC 1995 model: [BJF97 (1.0)]


Updated model: [A05 (0.25),
HG07 (0.25), BA08 (0.5)]
GSC 1995 model: [YCSH97 (1.0)]
Updated model: [AB03 (0.6) ,
Z06 (0.2), GA09 (0.2)]
GSC 1995 model: [YCSH97 (1.0)]

Updated model: [GSWY02 (0.25),


AB03 (0.25) , Z06 (0.25),
AM09 (0.25)]

2.12 Record selection for IDA

2.12.1 Spectral shapes of the records and epsilon values


The spectral shapes of the records are especially important for collapse assessments
which can change the calculated collapse capacity of the structures by 70% (Baker and
Cornell 2006a, Haselton and Baker 2006, Zareian 2006). Therefore neglecting the

69

influence of the spectral shapes in collapse assessment of the structures using IDA can
result in very conservative and biased predictions.

It has been shown that (T1) (i.e., epsilon value of the ground motion record at the
fundamental period) is a proxy of spectral shape (Baker and Cornell, 2006a). As the
epsilon value increases, the spectral shapes tend to be more peaked (e.g., see Fig. 2.11).
Epsilon, , is computed by subtracting the mean predicted ln Sa(T), from the records ln
Sa(T), and dividing by the logarithmic standard deviation (as estimated by the ground
motion prediction equation). This is given by Eq. [2.20]:

[2.20]

H (T )

ln S a (T )  Pln S a ( M , R ,T )

V ln S (T )
a

where
ln Sa (M,R,T) and ln Sa (T) are the predicted mean and standard deviation,
respectively, of ln Sa at given period and ln Sa(T) is the log of the spectral acceleration of
interest. The first two parameters are computed using ground motion prediction equations
(GMPE) which are also known as attenuation models. An example of the epsilon values
computed for the case of the Imperial Valley (1979) record is shown in Fig. 2.10.

Rare ground motions that can cause modern structures to collapse have peaked spectral
shapes that can be represented by positive epsilon values at the fundamental period ,
(T1). This peaked spectral shape is much different than a standard uniform hazard
spectral shape and accounting for this has been shown to increase the computed collapse
capacity significantly. The values of spectral acceleration at the longer periods are
important, because the effective period of the structure increases as the structure
undergoes inelastic deformation. This is even more important for the modern structures
that have high ductility capacities and are designed for high inelastic demands. Similarly
the values of spectral acceleration at shorter periods are also important especially for
structures with significant contributions of higher modes in the seismic response.

70

2.5

Sa(T) g (or) (T)

1.5
BA08 Prediction : Median
BA08 Prediction : Median+ 1 Sigma
BA08 Prediction : Median - 1 Sigma
BA08 Prediction : Median + 2 Sigma
BA08 Prediction : Median - 2 Sigma
"Imperial Valley (1979) response spectrum"

(T)

0.5

0
0.1

0.6

1.1

1.6
Period (T) Sec.

2.1

2.6

Fig. 2.10. An example of  values computed for the case of Imperial Valley (1979) record
using BA08 GMPE

The most direct approach to account for spectral shape in structural analysis is to select
ground motions with (T1) values similar to  values obtained from seismic hazard
deaggregation analysis for the site and hazard level of interest. An alternative approach is
to use an intensity measure (IM) other than Sa(T1) that accounts for spectral shape, if
epsilon values are not considered in the record selection and analysis. Such intensity
measures include inelastic spectral displacement or Sa values averaged over a period
range (Backer and Cornell 2006 and 2008). However the use of these intensity measures
are not common, while Sa(T1) intensity measure is widely used to describe the seismic
hazard.

71

=0
=1
=1.5

0.2

0.02

0.2
0.1

a)

 = 0.5

Sa(T) / Sa(T=1.0)

Sa(T) / Sa (T=0.3)

=0
 = 0.5
=1
= 1.5

Period (T) (Sec)

0.1

b)

Period (T) (Sec)

Fig. 2.11. The normalized average response spectra of crustal ground motion records for
different  values at a) T=0.3 sec and b) T=1 sec.
=0

=0

 = 0.5

 = 0.5

=1

=1
= 1.5

= 1.5
Sa(T) / Sa(T=1.0)

Sa(T) / Sa(T=0.3)

0.2

0.2

0.02

a)

0.1

Period (T) (Sec)

0.1

b)

Period (T) (Sec)

Fig. 2.12. The normalized average response spectra of interface ground motion records
for different  values at a) T=0.3 sec and b) T=1 sec.

5
=0

=0
5

=1

= 1.5

= 1.5

0.5

0.5

0.05

0.05

a)

 = 0.5

=1
Sa(T)/Sa(T=1.0)

Sa(T)/Sa(T=0.3)

 = 0.5

0.1

Period (T) (Sec)

b)

0.1

Period (T) (Sec)

Fig. 2.13. The normalized average response spectra of inslab ground motion records for
different  values at a) T=0.3 sec and b) T=1 sec.

72

In Fig. 2.11 to 2.14 the influence of the  values on the average spectral shape of the
records are demonstrated. The average geometric mean response spectra of 50 ground
motion records (22 records for the case of inslab events) with average (T1) values similar
to the target epsilon values were used to develop these graphs. Such curves are presented
for two different periods of T=0.3 sec and T=1 sec. As can be seen, as the  values
increase the resulting average response spectra of the records become more peaked. For
crustal and interface events this issue is observed for both cases of T=0.3 sec and T=1
sec. However in the case of inslab events it was found that in longer periods the spectral
shapes are insensitive to the epsilon values, as can be seen in Fig. 2.13b. This may be
primarily due to weak frequency content of such events in longer periods.

2.12.2 Record selection strategies


Record selection for inelastic time history analysis is a critical issue and many different
methods have been developed and used for this purpose. This subject is even more
important for the case of Incremental Dynamic Analysis (IDA) and collapse capacity
evaluation of structures. It has been demonstrated that the record selection methodology
has an important influence on the IDA results (Tehrani et al., 2012).

A realistic record selection strategy should take into account important characteristics of
the ground motion records which influence the structural response. Such important
aspects of records mainly include the magnitude, M, distance, R, , and site conditions.
Due to the limited number of natural records available, selecting ground motions with
similar M, R, and  values to match target values obtained through seismic deaggregation
is very difficult, if not impossible. Therefore usually one or two of these parameters are
chosen as key parameters for record selection. Accordingly different potential record
selection methodologies may be utilized.

1-UHS-Based approach:
The simplest method for record selection is to choose the records such that their response
spectra match a target uniform hazard spectrum (UHS) over a range of periods. The seed

73

records used in this method can be either artificial, natural, or manipulated natural
records. Artificial records can be generated using some available software to match a
UHS. Alternatively, natural records can be manipulated by changing the frequency
content, so that the response spectra of the records match a target UHS. Scaled natural
records that are matched to a target UHS over a range of periods are often recommended
by some codes. This range must include the important modes of the structure as well as
the effects of period elongation due to inelastic deformation of the structure. A period
range of 0.2T1 to 2T1 (where T1 is the fundamental vibration period of a structure) is
usually recommended and used for this purpose. This range is similar to the 0.2T1 to
1.5T1 range specified by ASCE 7-05 (ASCE, 2005), but statistical studies suggest that
nonlinear structures are often sensitive to response spectra at periods longer than 1.5T1
(Baker and Cornell 2008, Haselton and Baker 2006, and Vamvatsikos and Cornell 2005).

Although conceptually simple, these methods for record selection are deemed to be
conservative, particularly when ultimate performance limits of a structure (such as bar
buckling, fracture, collapse, etc.) are the primary objective of the analysis. This is
because the response spectra of the records are forced to match a target UHS, such as that
specified in the National Building Code of Canada (NBCC, 2010). The UHS is a
composite of predicted responses at different periods, and may not be representative of
individual ground motion spectra. Furthermore, the UHS tends to be dominated at any
individual period by motions above the median, whereas individual spectra are unlikely
to be equally above-median at all periods. That is, often no natural record can be found to
match the UHS over a wide period range and hence the analysis results using UHSmatched records may be conservatively unrealistic and biased. To solve these problems,
the use of the Conditional Mean Spectrum (CMS) in lieu of the UHS has been
recommended in recent years (Baker, 2011).

2-MR-based record selection:


In this method the records with the closest magnitude, M, and distance, R, to the mean
magnitude and mean distance obtained from the seismic deaggregations are chosen. A
difference of one unit in magnitude may be treated as equivalent to a difference of 40 km

74

in distance (Baker and Cornell, 2006a). For the records selected based on magnitude and
distance, it has been shown that inclusion of the parameter  in a vector IM results in
reduced estimates of structural response relative to the estimates obtained with the scalar
IM Sa(T1).This indicates that response predictions obtained using the MR-based record
selection approach are biased when Sa(T1) is used as IM.
3-Epsilon-based record selection:
The record property  at the fundamental period of the structure, (T1), is an important
property to match when selecting ground motions for analysis, where the ground motion
intensity is measured using Sa(T1). The value of (T1) is in fact a proxy for spectral shape,
and it has been demonstrated that selecting records with similar epsilon values to mean
epsilon values obtained from the seismic hazard deaggregation can result in a more
realistic prediction of seismic responses, especially at ultimate performance levels
including collapse where the spectral shape effects are more prominent.
4-CMS-based record selection:
The CMS-based record selection strategy is the only method that can take into account
the influence of all three ground motion parameters (i.e., M, R, and ) in record selection.
The idea is to predict the expected spectral shape of the records with magnitude, distance
and  values equal to the target values obtained from the seismic deaggregation. After
predicting the target spectral shape, the records with similar spectral shape over a range
of period of interest (usually 0.2T1 to 2T1) will be selected, while no strict criteria on M,
R and  are imposed. In another words, the records with spectral shapes matching the
CMS for a given M, R and  will be accurate predictors of structural response, regardless
of their actual M, R, and  values (Baker and Cornell, 2006a). This method will be
discussed more in the next section.

A study by Baker and Cornell (2005) suggests that the records selected based on  or
CMS can also be scaled without resulting in biased predictions, unlike other methods
such as UHS-based or M-R-based record selection. These observations can be explained
by the idea that spectral shape (i.e., spectral acceleration values at other periods, given Sa

75

at T1) is the record property that directly affects seismic response of structures, whereas
magnitude, distance and  are proxies for spectral shape (and  is a more important
proxy) (Backer and Cornell, 2006).

2.12.3 Conditional mean spectrum (CMS)


The CMS provides the expected response spectrum, conditioned on occurrence of a target
spectral acceleration value at the period of interest. It has been found that the CMS can be
used as an appropriate target response spectrum for selecting ground motions as input for
dynamic analysis (Baker, 2011). In the development of a CMS, some important aspect of
records including magnitude (M), distance (R), and epsilon (epsilon, , is defined as the
number of logarithmic standard deviations of a target ground motion from a median
ground motion prediction equation (GMPE) for a given M and R) will be considered
from the deaggregation of the seismic hazard. Baker (2011) proposed a method for
calculating the CMS, given by Eq. [2.21]:

[2.21]

Pln S

a (Ti )|ln S a (Tn )

Pln S ( M , R, Ti )  U (Ti , Tn )H (Tn )V ln S (Ti )


a

where an appropriate ground motion prediction equation (GMPE) must be used to


evaluate the mean and standard deviation of the natural logarithm of the spectral
acceleration at the vibration period Ti denoted by Pln Sa ( M , R, Ti ) and V ln Sa (Ti ) ,
respectively, in which M , R , and H (Tn ) are the mean magnitude, mean distance and
mean value of epsilon at the considered period Tn, respectively. These mean values are
calculated from the seismic hazard deaggregation and (Ti,Tn) is the inter-period
correlation of spectral accelerations at vibration periods Ti and Tn. The equation proposed
by Baker and Cornell (2006b), Eq.[2.22], is used to calculate this inter-period correlation.

76

[2.22]

U T n 1 ,T n 2 1  cos(  0.359  0.163 IT


2

min0.189

ln

Tmin Tmax
ln
0.189 Tmin

Where Tmax and Tmin are the larger and the smaller of Tn1 and Tn2, respectively, and

ITmin0.189 is an indicator function that equals one if Tmin is less than 0.189 sec and equals
zero otherwise.

For measuring the match with the target spectrum sum of squared errors (SSE) between
the logarithms of the ground motions spectrum and the target spectrum (Eq. [2.23]) is
used as recommended by Baker (2011).

[2.23] SSE

(ln Sa(T

)  ln SaCMS (T j ))2

j 1

Where ln Sa(Tj) is the log spectral acceleration of the ground motion at period Tj and ln
SaCMS(Tj) is the log CMS value at period Tj . Therefore the values of SSE will be
computed for all records considered and the records with the smallest SSE values will be
selected. This approach is more effective if ground motion scaling is also allowed. To
scale the ground motions, Baker (2011) recommends scaling each ground motion so that
its spectral acceleration at the considered period, Sa(T1), matches the target spectral
acceleration from the CMS, SaCMS(T1). Although the concept of scaling the records is
sometimes questioned, it has been observed that ground motions selected and scaled to
match the CMS produce displacements that are comparable to displacements produced by
unscaled ground motions, unlike scaling procedures using other methods (Baker and
Cornell, 2005 and Luco and Bazzurro, 2007). Thus the scaling procedure outlined by
Baker (2011) is not expected to adversely impact the resulting structural responses.

A complicated aspect in constructing a CMS for a site in south-western British Columbia,


in comparison with a site in California, is that three earthquake types, having distinctly
different characteristics, contribute to the overall seismic hazard. Therefore, three CMS
must be constructed for record selection (Goda and Atkinson, 2011), CMS-Crustal,
77

CMS-Interface, and CMS-Inslab. In order to account for these events, two sets of
CMS can be considered as proposed by Goda and Atkinson (2011). The first approach is
the CMS-Event-based procedure which is based on the weighted average of the CMS that
are computed by using applicable GMPEs and the corresponding scenarios for the three
earthquake types, in proportion to the relative influences of the scenarios (i.e., the number
of records from each event type is proportional to the percentage of contribution of that
event type in seismic hazard). The second approach is the CMS-All-based which is the
weighted average of CMS-Crustal, CMS-Interface, and CMS-Inslab by
considering relative influences of the individual earthquake types without considering
different earthquake type records in the selection procedure. In fact CMS-All-based is
the simplified version of CMS-Event-based in which different event types are considered
in the construction of the CMS, but in selecting the records the relative influences of the
scenarios are not considered. This results in variability of the seismic response based on
the CMS-All-based approach being smaller than that based on the CMS-Event-based
approach. However, it must be noted that for each earthquake type, the variability of the
selected records in general tends to be underestimated for CMS-based or UHS-based
approaches, since a close match to a target response spectrum is imposed (Baker, 2011).
Nevertheless, the CMS-Event-based approach can account for the variability of the
CMS among different event types more reasonably than does the CMS-All-based
approach.

2.12.4 Simplified method to include the spectral shape effects


The direct use of the epsilon-based method to account for the spectral shape of the
ground motion records is often time-consuming and complicated in practice. A simplified
method has been developed by Haselton et al. (2011) to modify the collapse capacity
predictions for the effects of epsilon and spectral shapes. For this purpose, a fixed set of
records can be used to perform the incremental dynamic analyses. The predicted
capacities obtained using the IDA will then be modified for the effects of spectral shapes
using a spectral shape factor (SSF). To compute the SSF a regression analysis needs to
be carried out to derive a relationship between the natural logarithm of the collapse

78

capacities versus the epsilon (T1) values for each record. The result of such a regression
analysis can be presented in the form of ln[SC] = 0 + 1. Where the SC is the collapse
capacity (at the fundamental period) , 0 is the average collapse capacity when  = 0, and
1 indicates how sensitive the collapse capacity (SC) is to changes in the  value. The SSF
is then calculated using Eq. [2.24].

[2.24]

SSF

exp[E1 (H 0 (T )  H (T )records )]





where H (T )records is the mean epsilon of the records and H 0 (T ) is the mean epsilon value
from the seismic deaggregation. The probability of exceedance of 0.5% in 50 years is
recommended by ATC-63 to calculate mean epsilon values for ductile structures. As an
example the 1 factor was computed as 0.317 for a case study using a linear regression
analysis as shown in Fig. 2.14.

1.5

ln[SC]

ln[SC] = 0.317 + 0.046


1

0.5
0
-1.5

-1

-0.5

0.5

1.5

2.5

 (T1)

-0.5
-1

Fig. 2.14. An example of the regression analysis to determine the 1factor

2.13 Evaluation of the seismic performance


The IDA results can be used to evaluate the seismic performance of structures. In the
ATC-63 provisions the probability of collapse at the maximum considered earthquake
79

level (MCE) is used for this purpose. The probability of collapse at MCE level, which
usually corresponds to 2% probability of exceedance in 50 years, should be reasonably
low. The probability of collapse at this seismic excitation level is typically limited to 10%
(and to 20% for a few cases) (ATC, 2008). These limits are currently based on judgement
and accordingly different limits on the maximum probability of collapse may be accepted
for different types of structures and the importance of the structure.

Eq.[2.3] can be used to estimate the probability of collapse at MCE level. Replacing x
in Eq.[2.3] by the spectral acceleration at the fundamental period of the structure at the
MCE level, SaMCE, the probability of collapse can be estimated using Eq. [2.25].

[2.25]

P (failure | Sa

SaMCE )

ln(SaMCE )  ln(Sa
)
ETOT

C
50%

Sa50%

ln(

SaMCE
)

ETOT

The ratio of the median collapse capacity of the structure (in terms of spectral
C
acceleration at the fundamental period of the structure), S a50%
, to the spectral

acceleration at MCE level (i.e., Sa (T1) for 2% probability of exceedance in 50 years


given by seismic design codes), SaMCE , is defined as the collapse margin ratio (CMR).
Therefore using the values of CMR and TOT, which are determined from the IDA results,
the probability of collapse at MCE level can be estimated using Eq.[2.26]. The seismic
performance of a structure is satisfactory if the estimated probability of collapse is
acceptable (e.g., typically less than about 10%).

[2.26]

P(collapse at MCE level )

 ln(CMR )
)

ETOT

80

(CM R

C
Sa50%
)
SaMCE

The fragility curves derived from the IDA results can also be combined with the
seismic hazard analysis to predict the seismic risk and the total probability of collapse
(i.e., mean annual rate of collapse) as discussed before.

References:
Please see the last section

81

3 Effects of Column and Superstructure Stiffness on the


Seismic Response of Bridges in the Transverse Direction
Payam Tehrani1 and Denis Mitchell1

3.1 Preface
The significance and problems regarding bridges with different column heights and
stiffnesses are discussed in Chapter 1 and Appendix A. This chapter describes a
parametric study on a wide range of bridges with different column and superstructure
properties to investigate the safety of bridges designed based on the CHBDC provisions
and to investigate whether the use of linear analysis methods is appropriate for the
prediction of the seismic response of bridges with different column stiffnesses.
Verification of the computer models and preliminary results and other modelling details
are presented in Appendix A. The computer program developed for the parametric studies
is discussed in Appendix B.

This chapter was summarized into a manuscript: Tehrani, P. and Mitchell, D. Effects of
Column and Superstructure Stiffness on the Seismic Response of Bridges in the
Transverse Direction, Canadian Journal of Civil Engineering, Manuscript 2011-0516,
accepted in March 2012.

Some parts of the manuscript including the introduction, literature review and the
definition of some regularity indices were also presented in Chapter 1. However these
parts were not removed from the manuscript to maintain its integrity and consistency.

McGill University, Department of Civil Engineering, 817 Sherbrooke St. West, Montreal QC H3A 2K6

82

According to the NRC Research Press policies, the authors retain the right to reuse all or
part of their manuscript in other works created by them for non-commercial purposes.
(http://www.nrcresearchpress.com/page/authors/information/rights)

83

3.2 Abstract
The transverse seismic responses of continuous 4-span bridges designed based on the
2006 Canadian Highway Bridge Design Code were studied using inelastic time history
analyses. A total of 648 bridge configurations were considered in which the column
heights, column diameters, superstructure stiffness and mass as well as abutment restraint
conditions were studied. The maximum ductility demands obtained using elastic and
inelastic analyses were compared to study the influence of the degree of irregularity. The
effects of column stiffness ratios and superstructure to substructure stiffness ratios on the
maximum ductility demands and concentration of ductility demands were investigated. A
number of different regularity indices were compared to determine the suitability of these
different indices in predicting the influence of irregularity. This study demonstrates the
conservative nature of the 2006 Canadian Highway Bridge Design Code and provides
some guidance on factors for determining the degree of irregularity and suitable
regularity indices when carrying out non-linear dynamic analyses of bridges.

Key words: Irregular bridges, Inelastic time history analysis, Regularity index, CHBDC
2006, NBCC 2010, Seismic response

3.3 Introduction
The seismic response of bridges becomes more complex, when geometrical properties
such as the number of spans, the ratio of adjacent span lengths, subtended angle for
curved bridges, and the span-to-span bent or pier stiffness ratio exceed criteria typically
defined by seismic design provisions. For such cases, simple linear analysis methods
typically fail to predict the seismic response and therefore the use of more sophisticated
analysis methods is necessary. Bridges with complex seismic behaviour are classified as
irregular bridges in seismic design codes. This study focuses on the effects of different
column and superstructure stiffnesses on the seismic response of straight bridges. A
bridge becomes more irregular as the difference between the column stiffnesses
84

increases, leading to greater differences between the results of elastic and inelastic
analyses, and concentration of ductility demands in a few stiffer elements.

For bridges with regular configurations (i.e., having columns with similar stiffnesses),
subjected to significant ground motions, the inelastic energy dissipation does not tend to
concentrate in a few ductile elements and it is expected that all of the plastic hinges will
contribute to the seismic response of the structure. However it is not possible to always
have regular configurations due to the need for different column heights for valley
crossings and ramps. Experience indicates that a bridge is more likely to be vulnerable if:
(1) excessive deformation demands occur in a few elements, (2) the structural
configuration is complex, or (3) a bridge lacks redundancy (Chen and Duan, 2000).

Bridges with significantly different column heights result in considerable


concentration of shears and moments in the stiffer shorter columns. In some cases, the
deformation demands on the short columns can cause failure before the longer, more
flexible adjacent columns can fully participate. In addition, the sequential yielding of the
ductile members may result in substantial deviations of the nonlinear response predictions
from linear analyses performed with the assumption of a global force reduction factor, R.
This difference is due to the formation of plastic hinges which appear first in the stiffer
columns and may lead to a concentration of unacceptably high ductility demands in these
hinges. Following the formation of the first plastic hinges, the distribution of stiffness and
hence of forces may change from that predicted by linear analysis. This may lead to a
substantial change in the assumed sequence of plastic hinges (CEN, 2005). One may try
to solve the problem by reducing the value of the R factor, in which case it is certainly
possible to reduce the ductility demand in the stiffer piers, but the resulting overall design
may be very costly. Calvi et al., (1994) proposed that different modification factors could
be used as a function of bridge geometry.

Examples of earthquake damage to irregular bridges with different column heights as


well as poorly designed and detailed bridges have been reported (Broderick and Elnashai,
1995; Mitchell et al., 1995; Chen and Duan, 2000). Some studies have been conducted on
85

irregular bridges designed using the Eurocode provisions (Calvi and Pinto, 1996; Pinto et
al., 1996; Fischinger et al., 1997; Isacovic and Fischinger, 2006 and Isacovic et al., 2008).
There is little research on this subject especially for bridges designed using American and
Canadian provisions. Currently the use of elastic analysis is permitted in these codes for
bridges with significant irregularity features, while the use of elastic analysis may not be
appropriate for some irregular bridges. Due to the vulnerability of irregular bridges, more
research is needed to investigate the seismic behaviour and suitable methods of analysis
for such bridges.

3.4 Regularity and irregularity indices


The purpose of a regularity index is to determine the degree of regularity or
irregularity of a bridge as a function of key structural parameters. The first attempt to
provide a regularity index was made by Calvi et al., (1994 and 1996). They proposed an
index of regularity, IR, which is a measure of the difference between the mode shape of
the whole bridge with deck and columns and the mode shape of the deck without
columns, given by Eq. 3.1.

[3.1]

IR

()

B
i

M ) Dj ) 2

j 1

[3.2]

IMR

1

[(1  G

ij

) ) iB M ) Dj ]

i 1 j 1

Following this study a new index, IMR (Eq. 3.2), was then proposed by subtracting the
norm of the products of the off-diagonal terms to increase the sensitivity of the index.
Where Bi , Dj and M are the mode shapes of the bridge and the mode shapes of the
deck alone and the mass matrix, respectively and n is the number of modes considered. In
Eq. 3.2, ij is the Kronecker delta which is equal to 1 if i=j and 0 otherwise. In these
equations Di is a measure of the inelastic mode shapes by neglecting the stiffness of
columns. When Bi and Dj are similar, the predicted response of the structure using the
elastic and inelastic analysis methods are generally in good agreement.

86

A number of other indices for assessing regularity have been proposed, including:

1- Use of Modal Assurance Criterion-MAC (Ewins, 2000):

[3.3-a]

MACi ( D, B)

^ `^ `
^) ` ^) ` ^) ` ^) `
)iD

D
i

)iB

D
i

B
i

[3.3-b]

(MAC ( D, B))

i 1

MAC ( D, B, n)

B
i

2- Use of Modal Scale Factor-MSF (Ewins, 2000):

^) ` ^) `
^) ` ^) `
D

[3.4-a]

MSFi

[3.4-b]

D T

(MSF ( D, B))

MSF ( D, B, n)

i 1

3-Difference Ratio of Mode Shapes-DRMS (Fischinger et al., 2003 and Maalek et al.,
2009):
n

[3.5-a]

DRMSi ( D, B)

)D  )B

SD  SB

)D

SD

[3.5-b]

( DRMS ( D, B))

DRMS ( D, B, n)

i 1

In Eq. 3.5-a SD and SB are the area under the mode shapes D and B, respectively. It
must be noted that contrary to the other indices described above, the DRMS index is 0 for
regular bridges and it increases with the degree of irregularity. A summary of different
indices for regularity are given by Maalek et al., (2009). These indices were derived for
bridges with transverse restraint at the abutments.

3.5 Code Provisions

3.5.1

Caltrans (2006) and AASHTO Guidelines (2009)

The requirements in Caltrans (2006) and the AASHTO Guidelines (2009) compare
the ductility demands with the ductility capacities in the energy dissipating members
87

(displacement-based design) instead of using an overall modification factor for design


(force-based design). Although the problem of different ductility demands of columns in
irregular bridges can be better addressed using this method, the ductility demands which
are based on the linear analysis methods such as response spectrum analysis (which is
permitted for design of irregular bridges) usually underestimate the ductility demands in
critical members and can lead to unsafe designs. Some conservative limitations are
recommended in Caltrans and the AASHTO Guidelines for the ratio of the column
stiffnesses in a frame or bent to balance the bent or pier stiffness along the bridge. For
any two bents or any two columns within a bent this stiffness ratio is recommended be
greater than 0.5. For adjacent bents or adjacent columns within a bent this ratio is
recommended to be less than 0.75. Some of the consequences of not meeting these
relative stiffness indicators include increased damage in the stiffer elements and an
unbalanced distribution of inelastic response throughout the structure.
If project constraints make it impractical to satisfy the stiffness requirements, a careful
evaluation of the local ductility demands and capacities shall be required for bridges in
regions of high seismicity.

3.5.2

Eurocode

In the Eurocode (CEN, 2005) a force-based design approach is used for the design of
bridges. The reduction factor, q (similar to the modification factor R), for reinforced
concrete columns depends on the shear span to depth ratio and on whether the member is
inclined or vertical. The q factor will then be modified in the cases of high axial loads and
irregular seismic behaviour of the bridge. In this regard the local force reduction factor ri
associated with member i is defined as:

[3.6-a]

ri = q (MEd,i / MRd,i)

[3.6-b]

 = rmax / rmin  o

[3.6-c]

qr = q (o / )  1.0

88

where MEd,i and MRd,i are the maximum values of design moment at the intended plastic
hinge location of ductile member i from the seismic analysis and the design flexural
resistance of the section with the actual reinforcement, respectively. The bridge is
considered to have regular seismic behaviour when Eq. 3.6-b is satisfied. Where rmax and
rmin are the maximum and minimum values of ri, respectively and o is a limit value
selected so as to ensure that sequential yielding of the ductile members will not cause
unacceptably high ductility demands on the member. The recommended value for o is
2.0. The force reduction factor, is then reduced to qr for the irregular seismic behaviour
according to Eq. 3.6-c. To capture the actual seismic behaviour of an irregular bridge
where the ductility demands concentrate in a few elements and the distribution of the
forces deviate from that predicted by the linear analysis, a combination of an equivalent
linear analysis with a non-linear static analysis is recommended.

3.5.3

CSA-S6-06 and AASHTO-04

The seismic design provisions in the Canadian Highway Bridge Design Code
(CHBDC) (CSA, 2006) are based on the AASHTO Specifications (AASHTO, 2004). S6
uses a force modification factor, R, and an importance factor, I, of 1.0, 1.5 and 3.0 for
other, emergency-route and lifeline bridges, respectively. The AASHTO
Specifications combine the force modification factor for ductility with the importance
factor to give one force modification factor for other, essential and critical bridges.
Nevertheless, the resulting design forces are similar. The CHBDC requires that the MM
(multi-mode spectral method and also referred as response spectrum analysis in this
paper) be used for the analysis of irregular bridges, even in seismic performance zone 4.
However for irregular lifeline bridges, time-history analysis is required in zones 3 and 4.
Concerning the column stiffness ratios for regular bridges, the 2004 AASHTO and the
2006 CHBDC specifications indicate that the maximum bent or pier stiffness ratio from
span to span should not exceed 4.0 for bridges up to 4 spans, 3.0 for 5 spans and 2.0 for
6 span bridges. These stiffness limits are for bridges with a continuous superstructure or
multiple simple spans with longitudinal restrainers and transverse restraint at each
support or a continuous deck slab, otherwise this ratio shall not exceed 1.25.
89

The modification factors, R, are taken conservatively and lower than the expected
displacement ductility capacities, since the procedure is intended to apply to a wide
variety of bridge geometries. Where possible, pier stiffnesses should be adjusted to
attempt to achieve uniform yield displacements and ductility demands on individual
piers. In cases where attempts to regularize the structure are impractical, suitable
analyses need to be developed to account for localized, rather than simultaneous, yielding
of piers. In some cases, it may be possible to use stiff piers with energy dissipating
bearings to alleviate the problem as discussed in the CHBDC Commentary (CSA, 2006).

3.6 Methods to improve the seismic behaviour of irregular bridges


Several methods could be used to improve the seismic performance of bridges with
varying column heights. These methods include the use of foundation sleeves for piers
with appropriate depths to equalize the effective length and stiffness of the columns (i.e.,
lower footings and isolation casing), the use of isolation devices (such as elastomeric and
sliding bearings) with appropriate stiffness to adjust the stiffness distribution and to
improve the damping level (Priestley et al., 1996) and the use of in-span hinges and
abutments with sacrificial shear keys (Saiidi et al., 2001).

Caltrans (2006) recommends some techniques to alleviate problems associated with


stiffness irregularities including using oversized pile shafts, using modified end fixities,
reducing and/or redistributing superstructure mass, varying the column cross section and
longitudinal reinforcement ratios, adding or relocating columns and modifying the
hinge/expansion joint layout.

A bridge memo for designers (Caltrans, 2011), gives some recommendations for the
design of columns that may be useful for the case of irregular bridges with different
column heights. These include pinning the columns at footings in multi-column bents,
pinning the base of the column adjacent to abutments in single-column bents, the use of
broader sections for taller columns and the use of pile shafts in lieu of footings.

90

3.7 Modelling and evaluation of the bridges

3.7.1

Limit states and corresponding strain limits

Emergency-route bridges must be repairable after the occurrence of a design


earthquake with 10% probability of exceedance in 50 years (CSA, 2006). This
performance level requires damage control, in which the damage in the structural
elements should not exceed some acceptable limits which are usually defined as a
maximum compression strain in the confined core concrete and a maximum tensile strain
in the reinforcing bars. After occurrence of a maximum considered earthquake (MCE)
with 2% probability of exceedance in 50 years, significant damage is expected but
collapse should be prevented. Several approaches are available to estimate the ultimate
compression strain in the confined concrete core (Mander et al., 1988; Paultre and
Lgeron, 2008). The Mander et al. equation is based on equal energy principles in which
the ultimate compression strain corresponds to the fracture of the hoops. Beyond this
level, crushing of the concrete, buckling and fracture of steel bars can occur.
Experimental studies indicate that the ultimate strains predicted by the Mander equation
are conservative with the actual ultimate strains exceeding the predicted values by a
factor of about 1.3 to 1.6 (Priestley et al., 2007). This conservatism is mainly due to the
presence of combined axial compression and flexure while the original Mander model
was derived for pure axial compression and it also does not take into account the
additional confinement provided by connecting members such as foundations or cap
beams. To account for this, Priestley et al., (2007) recommended using conservative
ultimate strains based on the Mander equation for damage control. However for the Life
Safety (LS) limit state (i.e., no collapse), they recommended that the predicted values
from the Mander equation be increased by 50% as suggested by the experimental results.

Due to the possibility of buckling of the bars when subjected to reversed cyclic
loading and considering low-cycle fatigue of the reinforcing bars in addition to slip
between the reinforcing steel and the concrete, the ultimate tensile strain, su, from the
monotonic tests must be modified to predict the ultimate curvature. The level of this
91

strain will depend on the volumetric ratio and spacing of the transverse reinforcement.
For the damage control limit state, a strain level of 0.6 to 0.7 times the ultimate strain of
steel bars in monotonic tests is recommended for calculating the ultimate curvature of the
section which should not be taken larger than 0.05 (Priestley et al., 1996 and 2007). In
order to attain this level of strain, the spacing of the transverse reinforcement must be
code conforming. For the life safety performance level a strain of 0.9 times the ultimate
strain of steel bars, 0.9 su, is recommended and this value should not be larger than 0.08
(Priestley et al., 2007).

Based on the degree of confinement either steel strains or concrete strains may control
the ultimate curvature of a section. However, considering the relatively high percentage
of transverse reinforcement based on current seismic design codes, the ultimate strain in
the steel often controls the ultimate curvature of the section. A study by Berry and
Eberhard (2007) provides some empirical equations to estimate the drift ratio, plastic
rotation, and longitudinal strain for circular bridge columns based on the longitudinal and
transverse steel ratio, axial load ratio, and geometric properties of the column. For
example as shown in Table 3.1-a the predicted ultimate drift at damage control for the
circular bridge column described is 3.9%, while the equations proposed by Berry and
Eberhard (2007) result in a predicted ultimate drift capacity of 5.3% corresponding to the
bar buckling damage state (see Table 3.1-b). This indicates that the prediction for the drift
limit for the damage control state is about 35% lower (i.e., more conservative) than that
predicted by the experimentally derived equations for this case. The predicted ultimate
drift capacity at the life safety limit state of 5.69% is in good agreement with that
predicted using the empirical equations (i.e., 5.30% for bar buckling and 5.74% for bar
fracture).

For the emergency-route bridges studied, the ultimate tensile strain in the steel bars
corresponding to bar buckling given by Berry and Eberhard (2007) was used to compute
the ultimate ductility of the columns.

92

Table 3.1. Deformations at different damage states for a column with D=1.5m, H=5m,
s=2.0%, and v =1.2% a) Using moment-curvature analyses (Priestley et al., 2007); b)
Using experimentally derived equations (Berry and Eberhard, 2007)
a)

b)

Damage state

Drift

Yielding

0.59%

Serviceability

0.93%

Damage Control

3.91%

Life Safety

5.69%

3.7.2

Strain

c=0.004 or s=0.015
c< Mander eq.or
s=0.6su < 0.05
c< 1.5*Mander eq.
or s=0.9su< 0.08

Curvature

Disp.

ductility

ductility

Damage state

Drift

Strain

Cover Spalling

1.85%

0.008

2.7

1.6

Bar Buckling

5.30%

0.075

15.9

6.2

Bar Fracture

5.74%

0.081

25.6

9.6

Calculating ductility capacities and modelling assumptions

Moment-curvature

analyses

were

performed

to

determine

the

curvatures

corresponding to the strain levels in the steel and the concrete for different damage states.
The corresponding displacement, drift, curvature ductility, and displacement ductility can
be computed for each performance level. A computer program was developed (Tehrani,
2012) to design the columns and carry out the moment-curvature analyses to compute the
curvature at different steel and concrete strains. The confinement effects in the concrete
core were considered using the Mander equation (Mander et al., 1988) in the momentcurvature analysis assuming confinement reinforcement in accordance with the
provisions of the CHBDC (CSA, 2006).

The effective curvature at yield can be either estimated from available experimental
data or from moment-curvature analysis. For circular sections Eq. 3.7 proposed by
Priestley et al., (2007) provides a good estimate for the yield curvature, y, given by :
[3.7]

y=2.25 y / D
93

In which y is the yield strain of the flexural reinforcement and D is the diameter of the
column cross section. In this study the yield curvature was computed for each crosssection using moment-curvature analysis, although the results were generally close to the
estimates using Eq. 3.7. The bilinear idealization of the moment-curvature responses
were carried out using the procedure describe by Priestley et al., (2007). For this purpose
the first yield is defined as the point on the moment-curvature response when the extreme
tension reinforcement first attains the yield strain, or when the extreme concrete
compression fibre reaches a strain of 0.002, whichever occurs first. The moment and
curvature at first yield are denoted by My and y respectively. The line defining the
elastic stiffness is extended up to the nominal moment capacity, Mn, and corresponding
curvature, y. Mn is the moment resistance corresponding to an extreme fibre
compression strain of 0.004 in the concrete or an extreme tension strain of 0.015 in the
steel bars, whichever occurs first. The corresponding curvature is defined as the nominal
yield curvature y for the idealized response. The plastic branch then can be defined by
connecting the nominal yield point to the ultimate point of the curve. The elastic stiffness
of the section (also known as the effective stiffness) then can be calculated using Eq. 3.8.

[3.8]

EIeff = My / y = Mn / y

It is well known that the curvature distribution cannot simply be integrated along the
column height for estimating the displacements since this ignores some deformations
including shear deformation and strain penetration due to bar anchorage. A simplified
approach is often used which involves the concept of a plastic hinge of length Lp over
which the curvature is assumed to be constant and equal to the maximum value at the
column base. For a column fixed at the base and pinned at the top, the curvature
distribution is considered to be linear at other parts of the column. The plastic hinge
length includes a strain penetration length, Lsp, to account for anchorage deformation due
to strain penetration over a length equal to development length of the steel bars. The
strain penetration length may be computed using Eq. 3.9 proposed by Priestley et al.,
(1996 and 2007):

94

[3.9]

Lsp=0.022 fy dbl (fy in MPa)

where fy and dbl are the yield strength and diameter of the longitudinal reinforcement,
respectively. The plastic hinge length, Lp, can then be calculated using Eq. 3.10.
[3.10]

Lp = k Lc + Lsp  2 Lsp

where Lc is length from the critical section to the point of contraflexure in the member
and the k factor is mainly a function of the ratio of the ultimate tensile strength to yield
strength of the flexural reinforcement (k = 0.08 was used in this study). The force versus
displacement relationship can then be calculated using the moment-curvature response.
For a cantilever column of height H, the displacement at yield, y, and displacement at
ultimate, u, can be calculated using Eq. 3.11 and 3.12.
[3.11]

y= y (H + Lsp)2 / 3

[3.12]

u=y + ( u - y) LpH

The modified Takeda hysteresis model (Otani, 1981) was used in this study to model
the behaviour of the RC columns using Ruaumoko software (Carr, 2009). This model has
two main parameters, alpha and beta, which control the unloading and the reloading
stiffness, respectively (see Fig. 3.1). Alpha is usually in the range of 0 to 0.5 and beta
varies between 0 and 0.6. Increasing the alpha parameter decreases the unloading
stiffness and increasing the beta parameter increases the reloading stiffness. For bridge
columns Priestley et al., (1996 and 2007) recommend using an alpha value of 0.5 and a
beta value of zero which was used in this study.

95

dp

Fy

E dp

dy D
K u K ( )D
y

K0

dm
m

d
dy

Ku

dm

Fig. 3.1. Modified Takeda hysteresis loop (adapted from Carr (2009))
3.7.3

Range of parameters studied

The seismic responses in the transverse direction of 4-span continuous bridge


structures were studied. According to Priestley et al., (2007), the longitudinal and
transverse behaviour can be considered independently for straight bridges, such as those
considered in this study. The bridge structures shown in Fig. 3.2 were designed according
to the Canadian Highway Bridge Design Code (CSA, 2006) using a force modification
factor, R, of 3 and an importance factor of 1.5 (i.e., I = 1.5 for emergency-route bridges).
The bridges were designed for Vancouver assuming a site class C (i.e., 360  Vs30 760
m/sec).

To investigate the effects of parameters such as column heights and diameters,


superstructure stiffness and mass, a parametric study was carried out. Column heights
were considered as 7, 14, and 21 m and column diameters were 1.5, 2.0, and 2.5 m. It
must be noted that in each configuration the three columns all have the same diameters
(see Fig. 3.2). The minimum and maximum longitudinal reinforcement ratios were 0.8%
and 6.0%, respectively. The transverse steel ratios were determined based on the CHBDC
2006 provisions to satisfy the requirements for maximum factored shear forces,
confinement in the plastic hinge regions, and the capacity design philosophy. However in
most cases the spiral confinement reinforcement ratio of 1.2% controlled.
96

50 m

50 m

50 m

50 m

H1= 7 - 21 m
D=1.5, 2.0, and 2.5 m

H2= 7 - 21 m
D=1.5, 2.0, and 2.5 m

H3= 7 - 21 m
D=1.5, 2.0, and 2.5 m

D
m4

Is = 40- 120
mm
m = 200 - 300 KN/m
mm

Fig. 3.2. Bridge configurations studied with different column heights

The superstructure lateral stiffness is also an important parameter in the seismic


response which is proportional to the moment of inertia of the superstructure crosssection relative to the vertical axis. The superstructure moment of inertia of box-girder
bridges usually varies between 40 to 150 m4 (Dwary and Kowalski, 2006), hence values
of 40, 80 and 120 m4 were chosen for the parametric study. The superstructure mass was
considered as a uniformly distributed load of 200 kN/m and 300 kN/m. The concrete
compressive strength and the yield stress of the reinforcing bars were taken as 40 and 400
MPa, respectively. These combinations of column heights (i.e., 3 different heights for
three columns), diameters (i.e., 3 different diameters), superstructure stiffness (i.e., 3
different values of Is), and masses (i.e., 2 different values) resulted in 324 bridges with
different configurations. In addition, the study was carried out for two cases of restrained
and unrestrained movements in the transverse direction at the abutments, so that a total of
648 bridges were studied.

For inelastic time history analyses 7 spectrum matched records were used. These
artificial records were generated using the SIMQKE software (Vanmarcke and Gasparini,
1979; Carr, 2009). For the nonlinear analyses of the bridges, the records were matched to
the design spectra given by the National Building Code of Canada (NBCC) (2010). These
spectra correspond to 2% probability of exceedance in 50 years, while the design
97

spectrum given in the CHBDC (CSA, 2006) corresponds to 10% probability of


exceedance in 50 years. The 2010 NBCC spectrum with the hazard level of 2% in 50
years, which will be used in the next edition of the CHBDC, was used to evaluate the
seismic behaviour of the bridges. The effects of irregularity on the seismic response are
expected to be more pronounced at higher seismic intensity levels. The ductility demands
in the columns should not exceed the ductility capacities computed based on the life
safety limit state (i.e., no collapse). The resulting average response spectrum of the seven
records used along with design spectra for Vancouver based on the CHBDC (2006) and
NBCC (2010) are shown in Fig. 3.3. The average displacement for each bridge was
determined using the averaging procedure proposed by Priestley et al., (2007) which
considered the maximum positive and maximum negative displacements predicted for
each input record.

1
Average Spectrum

Spectral acceleration (g)

0.9
0.8

NBCC spectrum

0.7
0.6

CHBDC Spectrum

0.5
0.4
0.3
0.2
0.1
0
0

0.5

1.5

2.5

3.5

Period (Sec)
Fig. 3.3. Average response spectrum of 7 records used for inelastic time history analysis
matching the 2010 NBCC spectrum (2% in 50 years)

98

3.8 Predicted response of bridges with restrained abutments


In this section the results obtained for the case of bridges with restrained transverse
movements at the abutments are presented.

3.8.1

Effect of column stiffness ratios

Fig. 3.4 illustrates the influence of the maximum column stiffness ratio (stiffness of
the stiffest column divided by the stiffness of the most flexible column) on the ductility
demand. As can be seen the maximum ductility demands obtained are less than 3.
However when the stiffness ratios are more than about 8 the maximum ductility demands
exceed 2.5 with minimum ductility demands of around 1.7. For more regular bridges
(small column stiffness ratio) the minimum range of the ductility demands are smaller
(e.g., about 0.5 for stiffness ratio around 1). However it must be noted that in general it
cannot be concluded that ductility demands in more irregular structures are always higher
than those of regular structures. Some other parameters such as strength of columns and
the overall stiffness of the structure play an important role in controlling the maximum
ductility demand on columns. For example, regular bridges often require lower column
strengths, compared to irregular structures, due to the uniform distribution of seismic
demands in regular bridges. This may increase the ductility demands on a regular bridge,
while an irregular bridge with larger column strengths may have even lower seismic
demands at the design earthquake level. Similarly, stiffer substructures attract more
seismic forces which result in higher ductility demands. Thus a stiff regular structure may
have higher ductility demands than a more flexible irregular structure. While the
maximum ductility demand may not be an appropriate parameter to identify the degree of
irregularity of a bridge, the maximum to minimum ductility demand ratio provides a
better parameter for this purpose.

99

3.5

Maximum ductility

3.0
2.5
2.0
1.5
1.0
0.5
0.0
1

10
Maximum column stiffness ratio

100

Fig. 3.4. Effect of column stiffness ratio on maximum ductility demand for bridges with
various configurations

Max/Min ductility

12
10
8
6
4
2
0
1

10
Maximum column stiffness ratio

100

Inelastic / Elastic displacement

Fig. 3.5. Effect of column stiffness ratio on maximum to minimum ductility ratio

2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
1

10
Maximum column stiffness ratio

100

Fig. 3.6. Effect of column stiffness ratio on the ratio of the displacements obtained using
elastic and inelastic analysis
100

Fig. 3.5 shows that the column stiffness ratios influence the maximum to minimum
ductility demand ratios. This ratio indicates how structural components participate in the
seismic behaviour. A large value of the max/min ductility ratio indicates that the seismic
demands are concentrated in a few elements. This figure also shows that when the
maximum column stiffness ratios are more than about 10, the max/min ductility ratios
increase significantly with values as high as 10 being predicted. Maintaining a maximum
column stiffness ratio of about 4 (regularity limit according to CHBDC for 4 span
bridges) limited the max/min ductility to about 3.0. Fig. 3.6 indicates that the ratio of
displacements obtained from inelastic analysis to those obtained from elastic analysis can
be more than 1.4, when the maximum column stiffness ratios are more than about 5. For
some irregular configurations, there are significant deviations of the displacement
envelopes determined using elastic and inelastic analysis. An example of such a
configuration is the irregular bridge with column heights of 7, 14 and 21 m, column
diameters of 2.5 m and a flexible superstructure with Is = 40 m4. As shown in Fig. 3.7, the
displacement envelopes determined from elastic and inelastic analysis are significantly
different for this configuration. For the irregular bridge the displacement of the critical
central short column is underestimated using the elastic response spectrum analysis
(multi-mode spectral method). This can lead to an unsafe design. However, due to the
conservative design approach in the 2006 CHBDC, with an R factor of 3.0 and an
importance factor of 1.5, the maximum ductility demands in such critical cases did not
exceed the ductility capacity of these ductile columns. The 2006 CHBDC permits the use
of elastic multi-modal analysis for irregular bridges and therefore would not provide
realistic designs for some irregular configurations.

101

100
90
80
70
60
50
40
30
20
10
0

Displacement (mm)

a)

Inelastic time history analysis


Elastic multi-mode spectral method
0

25

50

75
100
125
Bridge length (m)

150

175

200

175

200

120

b)

Displacement (mm)

100
80
60
40

Inelastic time history analysis

20

Elastic multi-mode spectral method

0
0

25

50

75
100
125
Bridge length (m)

150

Fig. 3.7. Comparison of the displacement envelopes obtained using elastic and inelastic
analysis for a) a regular bridge and b) an irregular bridge
3.8.2

Effect of superstructure to substructure stiffness ratio

Another parameter which is known to be very important in the seismic response of a


bridge in the transverse direction is the superstructure to substructure stiffness ratio, RS.
This ratio was calculated using Eq. 3.13.

[3.13]

RS

Ks
Kc

32I s
5L3s
3I c
hc3

102

where Ic is the moment of inertia of the stiffest column with cracked section
properties, hc is the column height, Is is the gross moment of inertia of the deck, and Ls is
the deck total length. It is noted that in this study the substructure stiffness, Kc, was taken
as the stiffness of the stiffest column, since it was found that this provided a better
prediction parameter than the sum of the stiffness of all columns.

As can be seen in Fig. 3.8, when the superstructure to substructure stiffness ratio is
less than 1.0, higher ductility demands are observed and more importantly the maximum
to minimum ductility demands can be drastically increased. In addition the same trend is
observed in the ratio of displacements obtained from elastic response spectrum analysis
and inelastic time history analysis. When the superstructure to substructure stiffness is
less than 0.3, the displacement ratio from elastic and inelastic analysis can exceed 1.4 as

Maximum ductility

can be seen in Fig. 3.9.

3.5
3.0
a)
2.5
2.0
1.5
1.0
0.5
0.0
0.01
0.10
1.00
10.00
100.00
Superstructure/Substructure stiffness ratio

Max/Min ductility

12
10

b)

8
6
4
2
0
0.01

0.10
1.00
10.00
100.00
Superstructure/Substructure stiffness ratio

Fig. 3.8. Effect of superstructure to substructure stiffness on a) maximum ductility


demand and b) maximum to minimum ductility demand
103

Inelastic / Elastic displacement

2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
0.01

0.10
1.00
10.00
Superstructure/Substructure stiffness ratio

100.00

Fig. 3.9. Effect of superstructure/substructure stiffness on the results from elastic and
inelastic analysis
3.8.3

Prediction of regularity indices

Several regularity indices have been proposed in the past that are used to estimate the
degree of irregularity of a bridge structure and determine if linear analysis can be used to
predict the maximum displacements of a structure. In Figs. 3.10 to 3.12 different
regularity indices are compared using several response parameters such as maximum
ductility demands, max/min ductility ratios, and the deviation of results from elastic and
inelastic analysis. This enables a comparison of the suitability of these indices. The
results from IR and MSF regularity indices were similar and thus only the results
obtained for the IR index are presented.

104

2
1
IMR

0
0.0

Maximum Ductility

Maximum Ductility

0.5

1.0

4
3
2
1
DRMS
0
0.001

0.010

0.100

4
3
2
1
IR
0
0.80
4

Maximum Ductility

Maximum Ductility

1.000

0.90

1.00

3
2
1
MAC
0
0.70

0.80

0.90

1.00

Max/Min Ductility

12
10
8
6
4
2
0

Max/Min Ductility

0.0

12
10
8
6
4
2
0
0.001

0.5
IMR

1.0

12
10
8
6
4
2
0
0.80

Max/Min Ductility

Max / Min Ductility

Fig. 3.10. Different regularity indices versus maximum ductility demand

0.010
0.100
DRMS

1.000

12
10
8
6
4
2
0
0.70

0.90
IR

1.00

0.80
0.90
MAC

1.00

Fig. 3.11. Different regularity indices versus maximum to minimum ductility demand
ratio
Fig. 3.10 shows the regularity indices versus the maximum displacement ductility
demands obtained from the inelastic time history analysis. It is clear that the sensitivity of
the IR and MAC regularity indices are quite low. For example, for the variety of the
configurations considered the value of IR and MAC indices obtained were in the range of
0.8 to 1.0 and 0.7 to 1.0, respectively. In addition there is no obvious trend between the
value of these indices and the predicted maximum ductility demands. However, it can be
seen that the structures with IR index of less than 0.99 and MAC index of less than 0.98
105

have higher ductility demands in general. The sensitivity of the IMR index is better than
those of IR and MAC and it can be seen that the bridges with IMR index of less than 0.9
generally have a higher ductility demand. However there are cases with small IMR index
values that have lower ductility demands than those with higher IMR index values. The
DRMS index seems to provide a better indicator for maximum ductility demands and a
trend can be seen that with increases in the DRMS index the ductility demands increase
in general. The bridge configurations studied had DRMS values between 0 and 0.4.

In Fig. 3.11, the maximum to minimum ductility demands, which are an indication of
the concentration of demand in a few elements, are shown versus different regularity
indices. As can be seen the IR, IMR, and MAC indices show no clear trend in predicting
this parameter. On the other hand, the DRMS index predicts the trend of this parameter
and as can be seen the bridges with the DRMS index of less than 0.025 have lower
concentration of ductility demands and thus this value may be used as a criterion to avoid

1.5
1.0
0.5
0.0
0.0

0.5
IMR

1.0

Inelastic/ Elastic Disp.

2.0
1.5
1.0
0.5
0.0
0.001

0.010
0.100
DRMS

1.000

Inelastic / Elastic Disp

Inelastic / Elastic Disp

2.0

Inelastic / Elastic Disp

problems with the concentration of ductility demands.

2.0
1.5
1.0
0.5
0.0
0.80

0.85

0.90
IR

0.95

1.00

2.0
1.5
1.0
0.5
0.0
0.70

0.80

0.90

1.00

MAC

Fig. 3.12. Different regularity indices versus the ratio of the displacements obtained using
inelastic and elastic analysis
The IR, IMR, and MAC indices are primarily aimed at identifying structures that
would give rise to poor correlation between their seismic response using elastic and
inelastic analyses. Fig. 3.12, illustrates that these indices are relatively successful in
106

providing such an indicator. For example, based on the results obtained the response of
structures with the IMR index of more than 0.85, IR index of more than about 0.95, or
MAC index of more than 0.9 can be predicted using elastic analysis otherwise care must
be taken to adopt an appropriate method of analysis. The DRMS index provides a good
indicator, with DRMS index values of less than 0.08 resulting in good agreement between
elastic and inelastic analysis.

A comparison of the regularity indices in Figs. 3.10 to 3.12, indicates that the DRMS
index is the most useful indicator of regularity and therefore the DRMS index is
recommended for determining an appropriate method of analysis.

3.9 Predictions of responses of Bridges with unrestrained abutment


conditions
In this section the results obtained for the case of bridges with free abutment
conditions at the ends (i.e., unrestrained transverse movements) are presented.

3.9.1

Effect of maximum column stiffness ratios

A similar trend is observed in the case of unrestrained abutment conditions. As can be


seen in Fig. 3.13, when the maximum column stiffness ratio exceeds 5.0, the ductility
demands concentrate in a few elements so that only a few columns contribute to the
seismic response. For cases having a maximum column stiffness ratio exceeding 8.0, the
results from the elastic analysis may not be representative of the actual response.

107

Inelastic / Elastic Displacement

Max/Min Ductility

10
9
8
7
6
5
4
3
2
1
0

a)

10
Maximum column stiffness ratio

100

2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0

b)

10
Maximum column stiffness ratio

100

Fig. 3.13. Influence of column stiffness ratios on a) maximum to minimum ductility


demands and b) results obtained from elastic and inelastic analyses (Bridges with
unrestrained abutment conditions)

3.9.2

Effect of superstructure to substructure stiffness ratios

Fig. 3.14 illustrates the importance of the superstructure to substructure stiffness ratio
on the seismic response of the bridge structures in the transverse direction. When this
value is less than 0.8, the variations in the column stiffnesses can result in uneven
distribution of seismic demands and can increase the maximum ductility demands. In
addition, when this stiffness ratio is less than 0.8, the results from elastic analysis may
substantially deviate from those obtained using inelastic analyses.

No general trend was found between the modal participation mass ratio of the first
mode, M1, and the parameters studied. However it was seen that typically structures with
M1 of more than 96% had lower maximum ductility demands and in addition the results
from elastic and inelastic analysis in such structures were in good agreement.

108

3.5

a)
Max/Min ductility

Maximum ductility

3.0
2.5
2.0
1.5
1.0
0.5

0.10
1.00
10.00
100.
Superstructure/Substructure stiffness ratio

Inelastic / Elastic Displacement

0.0
0.01

2.0
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
0.01

10.0
9.0
8.0
7.0
6.0
5.0
4.0
3.0
2.0
1.0
0.0
0.01

b)

0.10
1.00
10.00
100.
Superstructure/Substructure stiffness ratio

c)

0.10
1.00
10.00
100
Superstructure/Substructure stiffness ratio

Fig. 3.14. Effect of supersructure to substructure stiffness ratio on a) maximum


ductility demands, b) maximum to minimum ductility demands ratio, and c) results
obtained from elastic and inelastic analyses

3.10 Displacements from elastic versus inelastic analysis


The inelastic-to-elastic displacement ratio for each bridge was determined as the
maximum of the ratios of the predicted inelastic to elastic displacements for all bridge
columns. The difference between the inelastic and elastic responses is due to:
1- Inelastic analysis accounts for the concentration of inelasticity in critical elements.
2- Research has shown that considering the stiffness or strength degradation in the
hysteretic models typically results in higher predictions of the maximum displacements
than elastic analysis relying on the equal displacement concept (Priestley et al., 2007).
3- In this study conservative values were used for the hysteretic parameters controlling
the reloading and unloading stiffness in RC columns based on the recommendations by
Priestley et al., (1996 and 2007) that can result in slightly higher displacement predictions
than those obtained from elastic analysis.
109

4- There are two different load paths in the transverse direction (one via elastic
deformation of superstructure and one via inelastic deformation of the columns) which
results in different predictions of displacements when compared with the displacements
from the elastic analysis.

3.11 Demand to capacity ratios


To evaluate the safety of the bridges the maximum demand to capacity ratios were
computed for the 648 cases studied. In each case the capacities of the columns were
computed using a moment curvature analysis considering the confinement effects from
the spiral reinforcement. The displacement ductility capacity is a function of several
factors including the longitudinal and transverse reinforcement ratios, the column heights
and the column diameters, resulting in different ductility capacities for each design
configuration.

As shown in Fig. 3.15 the ductility demand to ductility capacity ratios are typically less
than 0.5 for both cases of restrained and unrestrained abutment conditions. For bridges
with restrained abutment conditions this ratio varies from 0.1 to 0.45 while for bridges
with unrestrained abutment conditions this ratio varies between about 0.2 and 0.5. The
results indicate that the safety margins are not uniform for bridges with different
configurations and structures designed using the force-based CHBDC (2006) provisions
have different levels of safety. This non uniform level of safety is expected when the
force-based design philosophy is used. Displacement-based design approaches are often
recommended to resolve this problem (Priestley et al., 2007).

It is noted that the longitudinal responses of the bridges were also studied (Tehrani, 2012)
indicating similar conservatism in the design using the CHBDC (2006).

110

Max (Demand / Capacity) ratio


(Free abutments)

Max (Demand / Capacity) ratio


(Restrained abutments)

0.6

a)

0.5
0.4
0.3
0.2
0.1
0.0
0

2
Period (sec)

0.6

b)

0.5
0.4
0.3
0.2
0.1
0.0

3
4
Period (sec)

Fig. 3.15. The maximum capacity to demand ratios for a) restrained abutments and b)
unrestrained abutments

3.12 Summary and Conclusions


Six hundred and forty-eight bridges with different configurations and with restrained
and unrestrained abutment conditions were designed based on the 2006 Canadian
Highway Bridge Design Code (CHBDC). Non-linear time history analyses were used to
predict the responses of these bridges using 7 spectrum-matched records. The conclusions
from this study on four-span reinforced concrete bridges located in Vancouver are
summarized as follows:
(1) The force-based design procedures of the 2006 CHBDC gives overly conservative
designs due to the combined use of low R factors and the use of importance
factors, even for some irregular bridges. In addition, the records used for the
predicting the responses matched the 2% in 50 year spectrum of the 2010 National
Building Code of Canada, whereas the CHBDC design spectrum is based on
accelerations with a probability of exceedance of 10% in 50 years.
(2) The different approaches to predict damage states for different performance levels
used in this study provide useful guidance for the next edition of the CHBDC that
will be based on performance requirements.

111

(3) The maximum ductility demands obtained were typically less than 3.0. Based on
current seismic design practice and the high percentage of confinement
reinforcement in the columns, much higher ductility capacities are expected. The
ductility capacities in the range of 6.5 to 8.5 were obtained in the life safety limit
state (i.e., no collapse) for the range of columns with different properties (i.e.,
height, diameter, and reinforcement ratios) studied.
(4) It is noted that conservative hysteretic parameters were used in this study along
with spectrum-matched records which are known to be conservative (since they
have high frequency content over a wide period range which is not observed in
natural records). Therefore, considering less conservative approaches can result in
predictions of even lower seismic demands.
(5) The transverse responses of the bridges are reported in this paper, because the
behaviour of irregular bridges in the transverse direction is more prone to
irregularities than in the longitudinal direction. This is because the response in the
transverse direction is governed by a multi-degree of freedom system
(displacements are different at different columns), while in the longitudinal
direction the response is typically governed by a single degree of freedom system
(displacements are the same for all columns). In addition, in the transverse
direction, contrary to that in the longitudinal direction, the seismic behaviour is
also affected by the position of columns along the length of the bridge and the
superstructure stiffness properties. Such parameters are not currently considered
in the definition of irregularity in the seismic codes. A study of these bridges in
the longitudinal direction (Tehrani and Mitchell, 2012c) indicated similar
conservatism in the design using the CHBDC (2006).
(6) It was shown that exceeding a maximum column stiffness ratio of about 5 to 8 can
cause significant concentration of ductility demand as well as deviation of the
predicted non-linear response from the predicted elastic response.

112

(7) A superstructure to substructure stiffness ratio, RS, of less than 0.3 in the case of
restrained movements at the abutments, and 0.8 in the case of free movements at
the abutments may result in incorrect response predictions using elastic multimode analysis. Bridges with RS ratios greater than 1.0 typically had more uniform
ductility demands in the columns, while bridges with RS values less than 1.0 may
exhibit concentrations of seismic ductility demand in a few elements.
(8) The effects of column stiffness ratio and superstructure to substructure stiffness
ratio were important for both cases of restrained and unrestrained movements at
the abutments. The concentration of ductility demands were slightly larger in the
case of restrained abutment condition, while the maximum ductility demands
obtained were larger on average for the case of bridges with unrestrained
abutments. In the case of unrestrained abutments, the difference between elastic
and inelastic analyses was more sensitive to the superstructure to substructure
stiffness ratio.
(9) While the column stiffness ratios and the superstructure to substructure stiffness
ratios provide simple, practical means of identifying regularity, the DRMS index
can also be used to choose an appropriate method of analysis. For bridges with
restrained abutment conditions and with DRMS index values of less than 0.08 the
results from the elastic and inelastic analyses were in good agreement.
(10)

Based on the results obtained, for some configurations, the displacement

envelopes were significantly different from the elastic and inelastic analysis;
however due to high conservatism in the current CHBDC this did not lead to an
unsafe design at least for the cases considered in this study. This problem may be
more important if less conservative design approaches are used and when higher
earthquake levels are considered. It is recommended that the CHBDC be changed
to exclude the use of elastic analysis for irregular emergency-route and lifeline
bridges that do not satisfy the criteria obtained in this study.

113

Acknowledgements

The financial support provided by the Natural Sciences and Engineering Research
Council of Canada for the Canadian Seismic Research Network is gratefully
acknowledged.

References
Please see the last section.

114

4 Effects of Column Stiffness Irregularity on the Seismic


Response of Bridges in the Longitudinal Direction
Payam Tehrani2 and Denis Mitchell1

4.1 Preface
The effects of different column stiffness irregularity and other important parameters
which influence the seismic response of bridges in the transverse direction are presented
in Chapter 3. It was concluded that for some cases the use of elastic multi-mode method
underestimates the maximum seismic demands in the bridge columns. This chapter
describes a study that investigates if the use of elastic analysis is appropriate for
predicting the seismic response of irregular bridges in the longitudinal direction.

This chapter focuses on the effects of column stiffness irregularity in the longitudinal
direction. Contrary to the transverse direction, the superstructure stiffness in the
longitudinal direction is very large and the response of the continuous bridges in the
longitudinal direction is similar to a single degree of freedom system. Therefore the main
parameters influencing the seismic response in the longitudinal direction include the
stiffness of columns and the abutment properties. The influences of the abutments are
also studied in the longitudinal direction of the bridges. The results obtained in Chapter 3
for the transverse direction can be combined with the results obtained in the longitudinal
direction of bridges to estimate the overall demand to capacity ratios. Similar modelling
assumptions, as discussed in Chapter 3 and Appendix A, were used for modelling of

McGill University, Department of Civil Engineering, 817 Sherbrooke St. West, Montreal QC H3A 2K6

115

bridges in Chapter 4. The computer program developed for the parametric studies is
discussed in Appendix B
This chapter was included into a manuscript: Tehrani, P. and Mitchell, D. Effects of
Column Stiffness Irregularity on the Seismic Response of Bridges in the Longitudinal
Direction, Canadian Journal of Civil Engineering, Manuscript 2012-0091, submitted
March, 2012.

According to the NRC Research Press policies, the authors retain the right to reuse all or
part of their manuscript in other works created by them for noncommercial purposes.
(http://www.nrcresearchpress.com/page/authors/information/rights)

116

4.2 Abstract
The longitudinal seismic responses of 4-span continuous bridges designed based on the
2006 Canadian Highway Bridge Design Code were studied using elastic response
spectrum and inelastic time-history analyses. The seismic response of more than 2900
bridges were studied to determine the effects of different design and modelling
parameters including the effects of different column heights, column diameters, and
superstructure mass as well as different abutment stiffnesses. The bridges were designed
using two different force modification factors of 3 and 5. The effects of column stiffness
ratios on the elastic and inelastic analysis results, maximum ductility demands,
concentration of ductility demands, and demand to capacity ratios were investigated. The
results indicate that the seismic response and maximum ductility demands in the
longitudinal direction are influenced by important parameters such as the total stiffness of
the substructure, the column stiffness ratio and the aspect ratio of the columns.

Key words: Irregular bridges, Column stiffness, Bridge abutment, Inelastic time history
analysis, CHBDC 2006, NBCC 2010, Seismic response.

4.3 Introduction
Bridges with significantly different column heights result in considerable
concentration of seismic demands in the stiffer, shorter columns. In some cases, the
deformation demands on the short columns can cause failure before the longer, more
flexible columns can fully participate. In addition, the sequential yielding of the ductile
members may result in substantial deviations of the nonlinear response predictions from
the linear response predictions made with the assumption of a global force reduction
factor, R. This difference is due to the fact that plastic hinges, which appear first in the
stiffer columns, may lead to concentrations of unacceptably high ductility demands in
these columns. Where possible, the stiffness of piers should be adjusted to attempt to
achieve uniform yield displacements and ductility demands on individual piers. A
117

summary of the methods to improve the seismic performance of such bridges is discussed
by Tehrani and Mitchell (2012a).

Examples of earthquake damage to irregular bridges with different column heights


(e.g., failure of the shorter columns), has been reported (Broderick and Elnashai, 1995;
Mitchell et al., 1995; Chen and Duan, 2000). The transverse response of bridges with
different column heights and different superstructure stiffnesses was studied by Tehrani
and Mitchell (2012a) which demonstrated that irregularities due to different column
stiffnesses have significant effects on the seismic behaviour of bridges. Research is
needed to investigate if these effects are also important in the longitudinal direction and
to evaluate the seismic safety of bridges with column stiffness irregularities. In this paper,
the effects of different column stiffnesses and stiffness ratios on the longitudinal
responses of bridges are presented. In addition, the influence of the abutments on the
seismic response of bridges is investigated. The main parameters in this study include the
column heights and diameters, different force modification factors, abutment stiffness
and strength, and the hysteresis stiffness degradation parameters used in the nonlinear
analyses. The influence of such parameters on the maximum ductility demands,
maximum drift ratios, concentrations of ductility demands, ductility demand to capacity
ratios, as well as comparisons of predictions from the elastic and inelastic analyses, are
presented.

4.4 Modelling and analysis of the bridges


A computer program was developed (Tehrani, 2012) to design the columns and to
carry out moment-curvature analyses to compute the curvatures corresponding to
different steel and concrete strains used as damage indicators. The displacements, drifts,
curvature ductilities, and displacement ductility capacities then can be computed for
different performance levels. The confinement effects in the concrete core were
considered using the Mander equation (Mander et al., 1988) in the moment-curvature
analysis assuming that spiral confinement reinforcement was provided in accordance with
the provisions of the Canadian Highway Bridge Design Code (CHBDC) (CSA, 2006).

118

The bilinear idealization of the moment curvature curves and the determination of the
effective curvature at yield, plastic hinge lengths and strain penetration depths for the
vertical bars were included in the analyses based on the recommendations by Priestley et
al., (1996 and 2007). More details are given by Tehrani and Mitchell (2012a).

The modified Takeda hysteresis model (Otani, 1981) was used in this study to
represent the behaviour of the RC columns using Ruaumoko software (Carr, 2009). This
model has two main parameters,  and , which control the unloading and the reloading
stiffness, respectively (see Fig. 4.1). The parameter  is usually in the range of 0 to 0.5
and  varies between 0 and 0.6. Increasing the parameter  decreases the unloading
stiffness and increasing the parameter  increases the reloading stiffness. For bridge
columns, Priestley et al., (1996 and 2007) recommend using the conservative values of
= 0.5 and  =0.

dp

Fy

E dp

dy D
K u K ( )D
y

K0

dm
m

d
dy

Ku

dm

Fig. 4.1. Modified Takeda hysteresis loop (adapted from Carr (2009))

4.5 Range of parameters studied


The seismic responses in the longitudinal direction of 4-span continuous straight
bridge structures were studied (see Fig. 4.2). According to Priestley et al., (2007), the
longitudinal and transverse behaviour may be studied independently for straight bridges,
such as those considered in this study.

119

H2= 7 to 28 m
D=1.5, 2.0 and 2.5 m

H1= 7 to 28 m
D=1.5, 2.0 and 2.5 m

Deck cross section

H3= 7 to 28 m
D=1.5, 2.0 and 2.5 m

Column cross section

D
mm

mm

Fig. 4.2. Bridge properties

The bridge structures were designed according to the CHBDC (CSA, 2006) using a
force modification factor, R, of 3 and 5 in the longitudinal direction and an importance
factor of 1.5 (i.e., I = 1.5 for emergency-route bridges). The bridges were designed for
Vancouver assuming a site class C (i.e., 360  Vs30 760 m/sec).
To investigate the effects of different column heights, different diameters and varying
column stiffness ratios on the seismic response, a parametric study was carried out.
Column heights were varied from 7 to 28 m with increments of 3.5 m (i.e., 7 different
heights for each column). The column diameters were 1.5, 2.0, and 2.5 m and in each
configuration the three columns all had the same diameters (see Fig. 4.2). These
combinations of column heights and diameters resulted in 252 bridges with different
column arrangements (i.e., it was assumed that the position of columns along the bridge
has no effects on the seismic response in the longitudinal direction). The superstructure
mass was considered as a uniformly distributed load of 200 kN/m. The effect of
increasing the superstructure mass to 300 kN/m on the seismic response was also
investigated for some cases. The bridges were designed and evaluated for two different R
values of 3 and 5. Further, different hysteresis parameters,  and , were used in the
120

structural modelling. In addition, the effects of abutment stiffness and capacity on the
seismic response of bridges were also considered for different number of piles and
different gap lengths between the superstructure and the abutment. The significance of
the P-Delta effects was also studied. Considering all of the parameters, more than 2900
bridge structures with different geometries, designs and modelling parameters were
studied. A schematic view of the structural models is shown in Fig. 4.3.

Abutment model

Nodes (lumped mass)

Rigid links

Superstructure (elastic element)

1/2 Superstructure depth

H2
Column

H1

H3

( lumped plasticity model)

Fig. 4.3. Structural modelling of the bridges

The minimum and maximum longitudinal reinforcement ratios in bridge columns were
0.8% and 6.0%, respectively. The transverse steel ratios were determined based on the
CHBDC 2006 provisions to satisfy the requirements for confinement in the plastic hinge
regions and to provide factored shear resistances corresponding to the capacity design
philosophy. However in most cases the spiral confinement reinforcement ratio of 1.2%
controlled. The concrete compressive strength and the yield stress of the reinforcing bars
were taken as 40 and 400 MPa, respectively.

For the inelastic time history analyses, 7 spectrum matched records were used. These
artificial records were generated using the SIMQKE software (Vanmarcke and Gasparini,
1979 and Carr, 2009). For the nonlinear analyses of the bridges, the records were
matched to the design spectra given in the National Building Code of Canada (NBCC)
(NRCC, 2010). These spectra correspond to 2% probability of exceedance in 50 years,
while the design spectrum given in the CHBDC (CSA, 2006) corresponds to 10%
probability of exceedance in 50 years. The 2010 NBCC spectrum with the hazard level of
2% in 50 years, which will be used in the next edition of the CHBDC, was used to

121

evaluate the seismic behaviour of the bridges. For this more appropriate probability of
occurrence the effects of irregularity on the seismic response are expected to be more
pronounced. The bridges should not collapse at this seismic hazard level.

The resulting average response spectrum of the seven records used along with the
design spectra for Vancouver based on the CHBDC (CSA, 2006) and the NBCC (NRCC,
2010) are shown in Fig. 4.4. The average displacement for each bridge was determined
using the averaging procedure proposed by Priestley et al. (2007) which considered the
maximum positive and maximum negative displacements predicted for each input record.

1
0.9

Average Spectrum

Spectral acceleration (g)

0.8

NBCC spectrum

0.7
CHBDC Spectrum

0.6
0.5
0.4
0.3
0.2
0.1
0
0

0.5

1.5

2
Period (Sec)

2.5

3.5

Fig. 4.4. Average response spectrum for 7 records used for inelastic time history analysis
matching the 2010 NBCC spectrum (2% in 50 years)

The capacity of the columns were determined assuming that the maximum strains in
steel bars attained the bar buckling strain limits given by Berry and Eberhard (2007) and
the maximum concrete compression strain predicted by the Mander equation modified
based on the recommendations by Priestley et al., (2007) for the life safety limit state.
More details concerning the evaluation of ductility capacity of columns using different
methods are available in Tehrani and Mitchell (2012a).

122

4.6 Modelling the abutments


Fig. 4.5 shows the abutment details, with a gap between the backwall and the
superstructure which is supported on bearing pads. The simplified abutment model
developed by Mackie and Stojadinovic (2003) and Aviram et al. (2008), as shown in Fig.
4.6, was used to study the influence of the abutments on the seismic response of the
bridges in the longitudinal direction. The longitudinal response is a function of the system
response including the elastomeric bearing pads, the gap, the abutment backwall, the
abutment piles, and the soil backfill material. Prior to impact due to gap closure, the
superstructure forces are transmitted through the elastomeric bearing pads to the
abutment, and subsequently to the piles and backfill, in a series system. After gap closure,
the superstructure bears directly on the abutment backwall and mobilizes the full passive
backfill pressure (Aviram et al., 2008). In the simplified model used the effects of the
bearing pads on the responses are ignored. However, it has been shown that the results
from the simplified abutment models in the longitudinal direction are in good agreement
with those obtained using more detailed models (Aviram et al., 2008).
Wing wall

wbw
W
bw

Shear key

Shear key
Stem wall
Back wall

hhbw
bw

Stem wall

Gap
Gap

Superstructure
Bridge deck

Bridge deck

Back wall
Back wall

Piles

Bearing
Bearingpad

Fig. 4.5. Schematic view of the seat-type abutment and its components

123

End releases:
Shear and moment

Gap element

Spring with backbone


curve for longitudinal
response (from Caltrans)

Elastic
superstructure
Joint restra ints : Only
longitudina l tra nsla tion
a llowed

Rigid elements

Fig. 4.6. Simplified abutment model for the longitudinal response

The abutment response in the longitudinal direction is accounted for by two zerolength elements at the extreme locations of rigid elements connected to the
superstructure, as shown in Fig. 4.6. The abutment stiffness, Kabt, and its ultimate
strength, Pbw, are obtained from Eq. [4.1] and [4.2] from Caltrans (2006) which are based
on a study by Maroney and Chai (1994).

[4.1] Kabt=Ki wbw (hbw/1.7)


[4.2] Pbw= p Ae (hbw/1.7)
[4.3] Ae= hbw wbw
where Ki is the initial embankment fill stiffness and is taken as 11500 kN/m/m, p is the
passive soil pressure taken as 239 kPa and Ae is the effective abutment area, defined as
the area which is effective for mobilizing the backfill. The effective backwall area is
given by Eq. [4.3] (Caltrans, 2006), with hbw and wbw defined in Fig. 4.5. In this study hbw
and wbw were taken as 2.0 m and 9.0 m, respectively based on the superstructure
geometry.

124

To model the abutments the bilinear hysteresis loop model with slackness (Carr,
2009), as shown in Fig. 4.7a, was used in the RUAUMOKO software (Carr, 2009). This
hysteresis model includes a gap and a spring in series which can be used to model the
initial gap and the resulting stiffness and strength associated with the abutments. An
example of the hysteretic behaviour of such an element obtained in the analyses is
demonstrated in Fig. 4.7b. As shown, the abutment elements only resist compression
forces.

500

F
Fy+

Displacement (m)
0

-0.3

Ko

-0.2

-0.1

0.2

0.3

-500

Ko

Gap-

-1000

d
-1500

Gap+
Ko

0.1

Ko

Force (KN)

a)

b)

-2000

Fy-

-2500

-3000

Fig. 4.7. a) Hysteresis model with gap and nonlinear spring used to model abutment response (Carr,
2009) b) typical nonlinear response from analysis

Different abutment stiffnesses and strengths were considered in the structural


modelling to study the seismic response of bridges in the longitudinal direction including
cases with (see Fig. 4.5) and without piles. To estimate the stiffness and strength of the
piles the empirical pile resistant equations given by Goel and Chopra (1997) (see Eqs. []
and []) were used. These equations provide an ultimate strength that is assumed to occur
at 1 in. (25 mm) displacement. The maximum displacement of the piles was taken as 2.4
in. (60 mm). The stiffness of the piles was conservatively neglected when deformations
exceeded this value, since the abutments and back walls are expected to be damaged at
deformations higher than this level (Goel and Chopra, 1997). The combined response of
the backfill and piles is determined by the combined response predicted by Eqs. [4.1] to
[4.5].

125

[4.4] Rpile =40 kips / pile = 178 KN/pile

[4.5] Kpile =40 kips /in. = 7000 KN/m

Four different gap lengths (i.e., 25, 50, 75, and 150 mm) along with three different pile
configurations (i.e., 0, 10, and 20 piles) were considered for modelling the influence of
the abutments. In addition, to study the influence of the abutments on different bridge
configurations, the column heights were varied from 7 to 28 m for each column with
increments of 7 m. This resulted in more than 720 bridge models.

The addition of the backfill contribution without piles provided a significant decrease
in the ductility demand. The seismic responses were only slightly improved when piles
were added to the abutments, however the effects of increasing the number of piles were
more pronounced for the case of smaller gap lengths and for more flexible bridges.

The reductions of ductility demand in the bridge columns are shown in Fig. 4.8b as a
function of the ratio of the total stiffness of the columns, Kcols, and the abutment effective
stiffness, Keff-abt. The abutment effective stiffness, Keff-abt, accounting for the gap closure is
defined in Fig. 4.8a. As this ratio, Kcols / Keff-abt, decreases the influence of the abutments
in the seismic response becomes more pronounced with up to 80% reduction in the
ductility demands. This indicates that the seismic response of bridges with flexible
columns will be affected more significantly by including the abutments in the structural
modelling.

126

a)

Pbw

KKeff-abt
eff-abt
Kabt

Gap

Deflection

100%
b)
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
0.01

% Reduction of ductility demand

Force

y = -0.131ln(x) + 0.196
R = 0.85

0.1

1
Kcols/Keff-abt

10

Fig. 4.8. Influence of abutment stiffness: a) effective abutment stiffness (Caltrans,


2006); b) the influence of the ratio of the total stiffness of columns to effective abutment
stiffness on the maximum ductility demands

4.7 Evaluation of maximum ductility demands


An important outcome of the nonlinear dynamic analyses is the maximum column
ductility demands which will be used to assess the seismic performance. The effects of
column stiffness irregularities on the maximum displacement ductility demands were
investigated. For each bridge the ductility demand in the most critical column is reported
as the maximum ductility demand in the bridge. The results are based on the average
responses obtained by means of inelastic time history analysis using 7 spectrum matched
records.

The effects of the total stiffness of the columns on the maximum ductility demands
obtained from nonlinear dynamic analyses are presented in Fig. 4.9. As expected the
stiffer structures typically attract higher ductility demands in the columns. Another
important parameter which will influence the maximum column ductility demand is the
maximum stiffness ratio of the columns (i.e., maximum column stiffness divided by
minimum column stiffness). Larger column stiffness ratios typically result in a
concentration of ductility demand in the stiffest column which in turn imposes higher
ductility demands on this element. The impact of the maximum column stiffness ratio on
the maximum ductility demand of columns is depicted in Fig. 4.10. Since both the total
127

100

stiffness of the columns and the maximum stiffness ratio of columns affect the maximum
ductility demands, another parameter has been defined as the product of these two
variables (i.e., total stiffness of columns times maximum stiffness ratio of columns). As
presented in Fig. 4.11, this new parameter shows an improved correlation with the
maximum ductility demands in the columns. The results are presented for two different
force modification factors of 3 and 5 used in design. As the force modification factor, R,
increases, the scatter in the maximum predicted ductility demands increases and the
maximum ductility demands become more sensitive to the product of the total stiffness of
the columns and the maximum stiffness ratio, as indicated by the slope of the regression
lines in Fig. 4.9 to Fig. 4.11.

3.5
3.0
2.5

4.0

a)

Maximum ductility demand

Maximum ductility demand

4.0

y = 0.19ln(x) - 0.10
R = 0.49

2.0
1.5
1.0
0.5
0.0
1000

10000
100000
Total stiffness of columns (kN/m)

1000000

3.5
3.0

b)

y = 0.33ln(x) - 1.07
R = 0.39

2.5
2.0
1.5
1.0
0.5
0.0
1000

10000
100000
Total stiffness of columns (kN/m)

Fig. 4.9. Effects of total stiffness of columns on the maximum displacement ductility
demands: a) R=3, =0.5 and =0; b) R=5, =0.5 and =0

128

1000000

4.0

a)

3.5

Maximum ductility demand

Maximum ductility demand

4.0

3.0
2.5
2.0
1.5

y = 0.18ln(x) + 1.47
R = 0.58

1.0
0.5
0.0

b)

3.5
3.0
2.5
2.0
1.5

y = 0.38 ln(x) + 1.50


R = 0.62

1.0
0.5
0.0

10
100
Maximum column stiffness ratio

1000

10
100
Maximum column stiffness ratio

1000

Fig. 4.10. Effects of maximum column stiffness ratio on the maximum displacement
ductility demands: a) R=3, =0.5 and =0; b) R=5, =0.5 and =0.
3.5

4.0

a)

Maximum ductility demand

Maximum ductility demand

4.0
3.0
2.5
2.0

y = 0.118ln(x) + 0.37
R = 0.69

1.5
1.0
0.5
0.0
1.0E+02 1.0E+03 1.0E+04 1.0E+05 1.0E+06 1.0E+07 1.0E+08
Total stiffness of columns X Max / Min stiffness (kN/m)

3.5
3.0

b)

y = 0.236ln(x) - 0.62
R = 0.66

2.5
2.0
1.5
1.0
0.5
0.0
1.0E+02 1.0E+03 1.0E+04 1.0E+05 1.0E+06 1.0E+07 1.0E+08
Total stiffness of columns X Max / Min stiffness (kN/m)

Fig. 4.11. Effects of columns total stiffness times max stiffness ratio on the maximum
displacement ductility demands: a) R=3, =0.5 and =0; b) R=5, =0.5 and =0.
Maximum ductility demand

3.5
3.0
2.5

y = 0.3 ln(x) - 2.1


R = 0.84

2.0
1.5
1.0
0.5
0.0
1.0E+02 1.0E+03 1.0E+04 1.0E+05 1.0E+06 1.0E+07 1.0E+08
Total stiffness of columns x Max / Min stiffness (kN/m)

Fig. 4.12. Effects of columns total stiffness times max stiffness ratio on the maximum
displacement ductility demands considering the influence of abutments (Gap=50 mm,
R=5, =0.5 and =0)

129

The correlation between the maximum ductility demand and the product of the total
stiffness of columns and maximum stiffness ratio is even better, as shown in Fig. 4.12,
when the influence of the abutments was included. This is probably due to the fact that
the inclusion of the abutments in the structural modelling significantly reduces the
nonlinear geometric effects due to P-Delta effects.

The results presented in Figs. Fig. 4.9 to Fig. 4.11 were derived assuming the
unloading and reloading hysteresis parameters of =0.5 and =0 in the modified Takeda
hysteresis model. The analysis results using hysteresis parameters of =0 and =0.6 (i.e.,
lower bound values) and also =0.3 and =0.3 (i.e., close to the average values from
tests) are demonstrated in Figs. Fig. 4.13a and Fig. 4.13b, respectively. The maximum
ductility demands for these cases are about 3.0 and 3.2, respectively, while in the case of
=0.5 and =0 (i.e., upper bound values) the maximum ductility demands were about 3.5.
The influence of the hysteresis parameters was larger for the bridges with higher values
of the parameter total stiffness of columns times the max stiffness ratio. However, in
general, the influence of the hysteresis parameters on the seismic response was not very
significant. These effects were even smaller when a force modification factor of R=3 was

4.0
3.5 a)
y = 0.186ln(x) - 0.23
3.0
R = 0.67
2.5
2.0
1.5
1.0
0.5
0.0
1E+2 1E+3 1E+4 1E+5 1E+6 1E+7 1E+8
Total stiffness of columns x Max / Min stiffness (kN/m)

Maximum ductility demand

Maximum ductility demand

used in design, due to the smaller nonlinear deformations in the columns.

4.0
3.5 b)
y = 0.203ln(x) - 0.34
3.0
R = 0.66
2.5
2.0
1.5
1.0
0.5
0.0
1E+2 1E+3 1E+4 1E+5 1E+6 1E+7 1E+8
Total stiffness of columns x Max / Min stiffness (kN/m)

Fig. 4.13. Effects of hysteresis parameters,  and , on the maximum displacement


ductility demands for R=5 and: a) =0 and =0.6; b) =0.3 and =0.3

130

4.7.1

P-Delta effects

To investigate the importance of P-Delta effects, the bridges previously designed with
these effects were also designed neglecting P-Delta effects (i.e., using a first-order
analysis) and an R factor of 5. The maximum ductility demands versus the average aspect
ratio (i.e., Have/D) of the bridge columns for this case are shown in Fig. 4.14. Where the
average aspect ratios were more than about 9 (corresponding to an average slenderness
ratio of 72), the P-Delta effects were more important and the bridges were susceptible to
instability due to P-Delta effects when subjected to ground motions corresponding to 2%
probability of exceedance in 50 years. This slenderness limit may be smaller for bridges
with larger dead loads. As shown in Fig. 4.14, the extremely large ductility demands (i.e.,
larger than 10) obtained for average slenderness ratios greater than 72 are due to
structural instability. It is noted that, for these slender cases, the inclusion of the stiffness
and strength of the bearing pads and abutments resulted in stable seismic responses.

Maximum ductility demand

40.0
35.0
30.0
25.0
20.0
15.0
10.0
5.0
0.0
0

6
8
10
12
14
16
Average aspect ratio of columns (H/D)

18

20

Fig. 4.14. Possible instability due to P-Delta effects for bridges that were designed
neglecting P-Delta effects.

4.8 Predictions from inelastic versus elastic analysis


A study by Tehrani and Mitchell (2012a) demonstrated that column stiffness
irregularities can result in significant deviations of the elastic multi-mode analysis results
from the inelastic dynamic analysis when the transverse response of bridges were studied.
131

As shown in Fig. 4.15 the differences in the maximum displacement predictions using the
elastic and inelastic analyses were typically small for response in the longitudinal
direction. The differences were generally less than 20% with higher differences when the
total stiffness of columns was quite low. This is probably due to P-Delta effects and the
higher dispersions of the response spectra of the ground motion records in the longer
period range. However, for the bridges with lower total column stiffness the maximum
ductility demands are typically small (e.g., see Fig. 4.9). The ratio of the displacements
obtained from the inelastic and elastic analyses were not significantly affected by the
maximum stiffness ratio of the columns. The use of R=3 in design also led to similar
results obtained for the case of R=5.

1.4
Inelastic / Elastic displacement

Inelastic / Elastic displacement

1.4

a)

1.2
1.0
0.8
0.6
0.4
0.2
0.0
1.0E+04

1.0E+05
Total stiffness of columns (kN/m)

1.0E+06

Inelastic / Elastic displacement

Inelastic / Elastic displacement

1.0
0.8
0.6
0.4
0.2
0.0
1.0E+04

1.0E+05
Total stiffness of columns (kN/m)

1.0E+06

1.2

1.4

c)

1.2
1.0
0.8
0.6
0.4
0.2
0.0
1.0E+04

b)

1.2

1.0E+05
Total stiffness of columns (kN/m)

1.0E+06

d)

1.0
0.8
0.6
0.4
0.2
0.0
1.0E+04

1.0E+05
Total stiffness of columns (kN/m)

1.0E+06

Fig. 4.15. Effects of total stiffness of columns on the ratio of the displacements obtained
using inelastic and elastic analysis for R=5 and: a) =0 and =0.6; b) =0.5 and =0; c)
=0.3 and =0.3; d) considering the influence of the abutments on seismic response (no
piles, gap=50 mm, =0.5 and =0)

132

The ratios of the maximum displacement demands obtained using the inelastic and
elastic analyses are presented in Fig. 4.15a-c for different hysteresis loop parameters. The
Takeda hysteresis loop model with =0 and =0.6 represents no unloading stiffness
degradation and small reloading stiffness degradation (i.e., lower bound values), while
the choice of =0.5 and =0 overestimates the unloading and reloading stiffness
degradation (i.e., upper bound values). Nevertheless, the effects of using different
hysteresis parameters are not significant and the differences between the inelastic and
elastic results are typically small, as shown in Fig. 4.15a -c. It should be noted that the
equal displacement concept is based on bi-linear hysteresis models with no stiffness
degradation. When stiffness degradation is considered in the nonlinear response,
somewhat different predictions may be obtained.
For the cases where the effects of the abutments are included in the nonlinear analysis
the resulting displacements for the flexible structures will be much smaller (see Fig.
4.15d) than those predicted with the assumption of free longitudinal movements at the
ends (see Fig. 4.15b). The elastic responses were computed assuming free movement at
the abutments in the longitudinal direction (i.e., roller supports), an assumption typically
made in practice for design and analysis. Such an assumption is conservative, as is
evident by comparing the results in Fig. 4.15b with those shown in Fig. 4.15d.

4.9 Concentration of seismic demands


In the longitudinal direction of continuous bridges all of the columns have almost
equal displacement demands. On the other hand, in the transverse direction the columns
will have different displacements depending on: a) stiffness and position of the columns;
b) the superstructure transverse stiffness and c) the abutment restraint conditions.
The influence of the maximum stiffness ratio of columns on the maximum to
minimum (max/min) ductility demands for response in the longitudinal direction is
shown in Fig. 4.16. Increasing the maximum stiffness ratio of columns leads to higher
concentrations of ductility demands in a few columns.

133

Max/Min ductility demand

16
14
12
10

=1

8
6
4

= 2.5

2
0
1

10

100

Maximum stiffness ratio of columns

Fig. 4.16. Effects of maximum column stiffness ratio on the Max/Min ductility ratio
(R=5, =0.5 and =0) and predictions using Eq.[4.11].

The relationship between the maximum column stiffness ratio and the maximum to
minimum ductility demands can be derived, assuming that the displacement demand, d,
is equal for all columns in the longitudinal direction. The maximum and minimum
ductility demands can be computed using Eq. [4.6a-b], where y is the displacement at
yield.

[4.6a] Pmax

'd
and [4.6b]
( ' y ) min

Pmin

'd
(' y ) max

where max and min are the maximum and minimum displacement ductility demands
in the columns.

The strain penetration depth, Lsp, is much smaller than the column height and can be
ignored. The displacement at general yielding, y, can be approximated using Eq.[ 4.7ab] for the cantilever columns.

[4.7a] (' y )max

2
) y H max

& [4.7b]

134

(' y )min

2
) y H min

The yielding curvature, y , is mainly a function of the column diameter and the yield
strain of the reinforcing bars and can be estimated using Eq. [4.8] for circular columns
(Priestley et al., 2007).
2.25H y

[4.8] ) y

Since all of the columns in this study have the same diameters for each configuration,
it can be assumed that the yielding curvature of the columns are almost equal and thus the
maximum to minimum ductility demand ratio can be estimated using Eq. [4.9].
[4.9]

(' y ) max

Pmax
Pmin

(' y )min

H max 2
)
H min

For the cantilever columns considered in this study the maximum column stiffness
ratio can be calculated using Eq. [4.10]:

[4.10]

K max
K min

J(

H max 3
)
H min

where Kmax and Kmin are the maximum and minimum stiffness of the columns, is the
ratio of the effective moment of inertia (i.e., moment at general yielding of column
divided by yield curvature) of the stiffest column to that of the most flexible column.
Although the column diameters are equal in this study, due to different flexural strengths
of the columns the effective moment of inertia of the stiffer columns which often require
higher reinforcement ratios is typically higher than that of the more flexible columns
assuming almost equal axial loads in the columns (since the span lengths are equal). From
Eqs. [4.9] and [4.10], the max/min ductility ratio can be estimated using Eq. [4.11].

[4.11]

Pmax
Pmin

J 0.67 (

135

K max 0.67
)
K min

This expression can be adjusted for cases with different column diameters by
multiplying the right hand side of Eq. [4.11] by the ratio of the column diameter of the
shortest column to that of the longest column. The predicted maximum to minimum
ductility demands using Eq. [4.11] are compared with the results from nonlinear analysis
in Fig. 4.16. In this comparison Eq. [4.11] has been plotted with two values of (1 and
2.5) which represents the range of the ratios of the effective moments of inertia for the
bridge columns in this study. As can be seen in Fig. 4.16, Eq. [4.11] provides a fairly
accurate representation of the maximum to minimum ductility ratio, bounding the data for
the range of columns studied. Eq. [4.11] can be used by designers to minimize the
concentration of ductility demands by adjusting Kmax/Kmin.

4.10 Drift ratios


The maximum drift ratios of columns obtained using nonlinear dynamic analyses are
shown in Fig. 4.17a. These maximum drift ratios are generally around 0.5% to 3.5% and
typically decrease with increasing values of total stiffness of the columns. The drift ratios
were relatively similar for different R values, as shown by comparing Fig. 4.17a and b.
The addition of the abutments in the structural modelling decreased the maximum drift
ratio by about 1.5%, as shown in Fig. 4.18, especially in the case of flexible
substructures.
4.0%

4.0%

a)

3.0%

2.5%

2.5%

2.0%

2.0%

1.5%

1.5%

1.0%

1.0%

0.5%

0.5%

0.0%
1000

10000
100000
Total stiffness of columns (kN/m)

b)

3.5%

3.0%
Drift

Drift

3.5%

1000000

0.0%
1000

10000
100000
Total stiffness of columns (kN/m)

Fig. 4.17. Maximum drift ratio of columns for: a) R=5 and Takeda hysteresis model with
=0.5 and =0; b) R=3 and Takeda hysteresis model with =0.5 and =0
136

1000000

2.5%

Drift

2.0%
1.5%
1.0%
0.5%
0.0%
1000

10000
100000
Total stiffness of columns (kN/m)

1000000

Fig. 4.18. Maximum drift ratio of columns considering the influence of the abutments in
nonlinear response (no piles, gap=50 mm, R=5, =0.5 and =0)
4%

0.007

b)

0.006

3%

(max D)/Hmin2

Maximum drift ratio

a)

2%
1%

0.005
0.004
0.003
0.002
0.001

0%

0.000
0

2
3
Maximum ductility demand

2
3
Maximum ductility demand

Fig. 4.19. Maximum drift ratio versus maximum ductility demand using R=5, =0.5 and
2

=0 for: a) maximum drift ratio versus maximum ductility demand; b) ( D) / H versus
maximum ductility demand

A comparison of the maximum drift ratio versus the maximum ductility demand of the
columns is shown in Fig. 4.19a. Drift ratios are widely used in practice (e.g., in codes) for
the assessment of structural performance. However the drift ratios may not be a good
indicator of structural damage. As indicated before, the ductility was found to be
proportional to ( D) / H2 (see Eq. [4.10]). In Fig. 4.19b another damage indicator is
defined as (max D) /Hmin2 (i.e., max drift ratio divided by the aspect ratio). This new
parameter shows a significantly improved correlation with the column ductility demands,

137

as shown Fig. 4.19b, compared to Fig. 4.19a. Therefore this parameter is a better
indicator of damage, since the ductility demands and the corresponding structural damage
are proportional to D/H2 rather than 1/H.

4.11 Ductility demand versus ductility capacity


The maximum ductility demands obtained from the nonlinear dynamic analyses were
compared to the ductility capacities of the columns to evaluate the safety margin of the
columns. In Fig. 4.20 the ratio of the maximum ductility demands to the ductility
capacities are shown versus the maximum stiffness ratio of the columns for various
hysteresis parameters and R factors. The ductility capacity for each column is a function
of the geometric properties and details of reinforcement. These ductility capacities were
computed for the life safety performance level (i.e., no collapse).

The ductility demand to ductility capacity ratios from the analyses in the longitudinal
direction are in the range of 0.2 to about 0.5 when a force modification factor of R=5 was
used (e.g., see Fig. 4.20a). In the case of R=3 these ratios were around 0.2 to 0.4 as
shown in Fig. 4.20b. Similar results were obtained when the transverse response of
similar bridges were studied (Tehrani and Mitchell, 2012a). Using a lower force
modification factor of R=3 did not significantly increase the safety margins as shown in
Fig. 4.20c and d.

As the maximum stiffness ratio of the columns increases, the range of the maximum
demand to capacity ratios obtained increases as well. For example in Fig. 4.20a when the
maximum stiffness ratios are less than about 5.0 the maximum demand to capacity ratios
are less than around 0.3. For maximum stiffness ratios between 5.0 to 10.0 the maximum
demand to capacity ratio is around 0.4 and exceeding the stiffness ratio of 10.0 can lead
to demand to capacity ratios of about 0.5.

When lower stiffness degradations were considered in the modelling (i.e., =0.3 and
=0.3), the maximum demand to capacity ratios decreased to around 0.4 as demonstrated

138

in Fig. 4.20c. The influence of hysteresis parameters was more pronounced for bridges
with higher column stiffness ratios, possibly due to the higher nonlinear deformations and
the concentration of nonlinear demands on the columns. The same trends were observed
for the case of R=3 as shown in Fig. 4.20d. When the superstructure mass was increased
to 300 kN/m the effects of the column stiffness ratios became even more important as
shown in Fig. 4.21a and b. Including the influence of the abutments in the structural
modelling reduced the maximum demand to capacity ratios to around 0.35 as shown in
Fig. 4.22.

0.5

a)

Max (Demand / Capacity) ratio

Max (Demand / Capacity) ratio

0.5
0.4
0.3
0.2
0.1
0.0
1.0E+00

1.0E+01
Max / Min stiffness

1.0E+02

0.3
0.2
0.1
0.0
1.0E+00

1.0E+01

1.0E+02

Max / Min stiffness

0.5

0.5

c)

Max (Demand / Capacity) ratio

Max (Demand / Capacity) ratio

b)
0.4

0.4
0.3
0.2
0.1
0.0
1.0E+00

1.0E+01
Max / Min stiffness

1.0E+02

d)
0.4
0.3
0.2
0.1
0.0
1.0E+00

1.0E+01
Max / Min stiffness

1.0E+02

Fig. 4.20. Maximum ductility demand to ductility capacity ratios obtained for: a) R=5,
=0.5 and =0; b) R=5, =0.3 and =0.3; c) R=3, =0.5 and =0; d) R=3, =0.3 and
=0.3

139

Max (Demand / Capacity) ratio

0.7
0.6

a)

b)

0.5
0.4
0.3
0.2
0.1
0.0
1.0E+00

1.0E+01
Max / Min stiffness

1.0E+02

Fig. 4.21. Effects of increasing the superstructure mass to 300 KN/m on the maximum
demand to capacity ratios obtained for: a) R=5, =0.5 and =0; b) R=3, =0.5 and =0

Max (Demand / Capacity) ratio

0.5
0.4
0.3
0.2
0.1
0.0
1.0E+00

1.0E+01

1.0E+02

Max / Min stiffness

Fig. 4.22. Maximum ductility demand to ductility capacity ratios obtained when abutment
effects were considered in modelling (Gap=50 mm, R=5, =0.5 and =0)

4.12 Normalized ductility demand


To improve the correlation between the results presented a new parameter is defined
as the normalized ductility demand which is simply calculated by dividing the maximum
ductility demands by the minimum aspect ratio of the columns (i.e., Hmin/D). It was
observed that the use of the average stiffness ratio of all of the columns in lieu of the

140

maximum stiffness ratios can also slightly reduce the scatter in the results. The
normalized ductility demand and normalized demand to capacity ratios are presented in
Fig. 4.23a and b. As can be seen by introducing these parameters the correlation has been
significantly improved compared to the results presented in Fig. 4.9 to Fig. 4.11. Hence,
the overall seismic response and the maximum ductility demands in the longitudinal
direction are controlled by at least three important parameters including the total stiffness
of the substructure, the stiffness ratio of the columns, and the minimum aspect ratio of the
columns.

1.00

a)

Max (Demand / Capacity) / (Hmin/D)

(Max ductility demand ) / (H min /D)

5.0

y = 0.152ln(x) - 1.27
R = 0.92
0.5

0.1
1.0E+03

1.0E+04

1.0E+05

1.0E+06

1.0E+07

Total stiffness of columns x average stiffness ratio (kN/m)

b)
y = 0.017 ln(x) - 0.138
R = 0.93
0.10

0.01
1.0E+03
1.0E+04
1.0E+05
1.0E+06
1.0E+07
Total stiffness of columns x average stiffness ratio (kN/m)

Fig. 4.23. Analysis results using R=5 and Takeda hysteresis model with =0.5 and =0
for: a) normalized ductility demands; b) normalized demand to capacity ratios

4.13 Demand to capacity ratios considering transverse and longitudinal


responses
A combination of orthogonal seismic displacement demands are often used to
approximately account for the directional uncertainty of earthquake motions and the
simultaneous occurrence of earthquake effects in the two perpendicular horizontal
directions. Based on the AASHTO guide specifications (AASHTO, 2009) the seismic
displacements resulting from analyses in the two perpendicular directions can be
combined. The seismic demand displacements can be obtained by adding 100% of the
seismic displacements resulting from the analysis in one direction to 30% of the seismic
141

displacements resulting from the analysis in the perpendicular direction and vice-versa to
form two independent cases (AASHTO, 2009).

The transverse responses of similar bridges were studied by Tehrani and Mitchell
(2012a). To estimate the resulting displacement ductility demands due to bidirectional
ground motions the resulting displacements from the analysis in the longitudinal and
transverse directions were combined using the 100% / 30% rule stated above.
The resulting displacement ductility demand to capacity ratios are shown in Fig. 4.24.
The ductility capacities of the columns were computed based on the Life Safety
performance level (i.e., collapse prevention) according to the recommendations of
Priestley et al., (2007). The beneficial effects from the abutments in the longitudinal
direction were conservatively neglected. The resulting demand to capacity ratios are
generally less than 0.7 considering the combination of maximum ductility demands from
transverse and longitudinal directions. The use of the SRSS combination rule also
resulted in similar predictions with ductility demands being about 5% larger on average.
As can be seen in Fig. 4.24, the demand to capacity ratios are less than 0.5 for the
majority of cases. However as the maximum stiffness ratio of columns exceeds about 8.0
the demand to capacity ratios are increased by 40%.

0.7

0.8

a)

Max (Demand / Capacity) ratio

Max (Demand / Capacity) ratio

0.8

0.6
0.5
0.4
0.3
0.2
0.1
0.0
1.0E+00

1.0E+01
Max / Min stiffness

1.0E+02

0.7

b)

0.6
0.5
0.4
0.3
0.2
0.1
0.0
1.0E+00

1.0E+01
Max / Min stiffness

Fig. 4.24. Ductility demand to ductility capacity ratios for different bridge
configurations considering transverse and longitudinal responses based on the 100%/30%
rule for: a) Bridges with restrained transverse movements; b) Bridges with unrestrained
transverse movements

142

1.0E+02

4.14 Conclusions
Bridges with different configurations were designed based on the 2006 Canadian
Highway Bridge Design Code (CHBDC). Non-linear time history analyses were used to
predict the longitudinal seismic responses of these bridges using 7 spectrum-matched
records. The conclusions from this study on four-span bridges located in Vancouver are
summarized as follows:
(1)

The seismic response and the maximum ductility demands in the

longitudinal direction are controlled by the total stiffness of the substructure, the
stiffness ratio of the columns, and the minimum aspect ratio of the columns. Seismic
ductility demands in the longitudinal direction were correlated with the product of the
total stiffness of the columns and the maximum stiffness ratio of the columns. This
indicates that the ductility demands in bridge columns increase as the structural
stiffness and stiffness irregularity increases.

(2)

It was demonstrated that the concentration of ductility demands increases

significantly with an increase in the column stiffness ratio, Kmax/Kmin. An equation was
developed to provide a means of estimating the ratio of the maximum to minimum
ductility demands. This equation is useful for designers in determining the influence of
the column stiffness irregularity on the concentration of the ductility demands.

(3)

The influence of the abutments on the longitudinal seismic responses of

bridges was studied. Up to an 80% decrease in seismic ductility demands were


observed when the abutments were considered in the structural models. The reduction
of ductility demand correlates with the ratio of the total stiffness of the columns and
the effective stiffness of the abutments. The influence of the abutments was more
pronounced for the bridges with more flexible columns and stiffer abutments.

(4)

Unlike the transverse responses, there was relatively good agreement

between the displacements obtained from elastic and inelastic analyses and the

143

longitudinal displacements were not significantly affected by the stiffness ratio of the
columns for bridges with different column heights.

(5)

A dimensionless parameter was defined as ( D) / H2 (i.e., drift ratio

divided by column aspect ratio) which provided an improved indicator of the structural
damage compared to the conventional drift ratio (i.e.,

 / H ). It was also

demonstrated that normalizing the maximum ductility demands by the minimum


aspect ratio of the columns significantly reduced the dispersions in the results.

(6)

The predicted maximum ductility demands were generally less than 3.5 in

the longitudinal direction, when a modification factor of R=5 was used in design. The
predictions were even lower when the influence of the abutments was considered in
the seismic response. It should be noted that the predictions are expected to be
somewhat conservative, since spectrum matched records were used for the analyses.

(7)

The seismic ductility demands to ductility capacity ratios were estimated

for the combination of the seismic responses in the longitudinal and transverse
directions. It was observed that the demand to capacity ratios were lower than 0.7 with
the majority of the cases having values less than 0.5. These ratios decreased, when the
influence of the abutments were considered in the seismic response. The range of
demand to capacity ratios was quite high which indicate uneven safety margins for
different bridges. Exceeding the maximum stiffness ratio of about 5.0 to 8.0 resulted in
much larger demand to capacity ratios.

(8)

CSA S6-06 requires elastic dynamic analysis for an emergency-route

bridge in seismic performance zones 2 and higher if the bridge is irregular. This study
indicates that the elastic dynamic analysis is appropriate for irregular bridges in the
longitudinal direction. However, nonlinear dynamic analysis would be required for
such irregular bridges in the transverse direction in order to accurately predict the
displacement envelope and the ductility demands (Tehrani and Mitchell, 2012a).

144

Acknowledgements

The financial support provided by the Natural Sciences and Engineering Research
Council of Canada for the Canadian Seismic Research Network is gratefully
acknowledged.

References

Please see the last section.

145

5 Effects of Different Record Selection Methods and


Earthquake Types on the Transverse Response of Bridges
Payam Tehrani1, Katsuichiro Goda2, Denis Mitchell1, Gail M. Atkinson3,
and Luc E. Chouinard1
1

Department of Civil Engineering & Applied Mech., McGill University, Montreal H3A 2K6,

Canada
2

Department of Civil Engineering, University of Bristol, Bristol BS8 1TR, United Kingdom

Department of Earth Sciences, University of Western Ontario, London, Ontario N6A 5B7,

Canada

5.1 Preface
The concepts of the IDA method for seismic evaluation of structures, different record
selection methods, different ground motion prediction equations (GMPE) and prediction
of different damage states using different methods were presented in Chapter 2.
In this chapter the IDA method has been used to evaluate the seismic response of a
continuous bridge designed based on the 2006 CHBDC provisions. For this purpose a
back-bone curve was defined based on different bridge specific studies for the purpose of
the structural modelling for IDA. The results from the experimental and theoretical
studies to define this back-bone curve were also compared.

Record selection in IDA is an important issue which can significantly affect the results.
There is no research available to investigate the influence of different record selection
methodologies for seismic evaluation of bridges using IDA. In this chapter the influence
of using different record selection methods including the UHS-based, CMS-based and
epsilon-based on the IDA results are studied. These record selection methods were
explained in Chapter 2.

146

The bridge under study is located in Vancouver, where the seismic hazard is influenced
by three different earthquake types with distinct characteristics. Little research is
available to investigate the influence of considering the subduction earthquakes in the
seismic assessments of bridges and currently only the records from the crustal
earthquakes (e.g., the records from the PEER-NGA database) are used for seismic
evaluations. Recent earthquakes in Chile (i.e., 2010 Chile earthquake) and Japan (i.e.,
2011 Tohoku earthquake) demonstrated the significance of the large subduction zone
earthquakes, indicating that more research is needed to focus in this subject. The records
from the 2011 Tohoku earthquakes were also considered in this study.

The seismic deaggregation results for Vancouver are used to compute the CMS for
different earthquake types and accordingly different CMS-based approaches are used
concerning different earthquake types as will be discussed in Chapter 5.
The importance of the epsilon and spectral shapes, as explained in Chapter 2, were also
included in the IDA results and the use of a simplified method to modify the results for
these effects was investigated for the bridge studied. Different GMPEs were used to
predict the CMS and the influence of using different GMPEs on the IDA results was
investigated.

The IDA results were evaluated at different damage states and the effects of different
record selection methods on the prediction of different damage states were investigated.
More details regarding the prediction of different damage states using different methods
and determining the these limits on the IDA curves are explained in Chapter 2.

More details regarding the automated program used for the large number of nonlinear
dynamic analyses in the IDA method is available in Appendix B.

More details regarding the program developed to extract the seismic deaggregation
results are available in Appendix C.

147

More details regarding the program developed for developing the CMS and selecting the
records using different methods considering different earthquake types and different
GMPEs are discussed in Appendix D.

More explanations regarding different earthquake types are given in Appendix F.

A pushover analysis of the bridge studied is also presented in Appendix H.

The detailed IDA results for the bridge studied in this chapter obtained using a large set
of records (i.e., including 234 records) are also presented in Appendix I.
Chapter 5 was summarized into a manuscript: Tehrani, P., Goda, K., Mitchell, D.,
Atkinson, G.M. and Chouinard, L.E. Effects of Different Record Selection Methods and
Earthquake Types on the Transverse Response of Bridges, Journal of Earthquake
Engineering and Structural Dynamics, Manuscript EQE-11-0079 (revised version),
submitted in December 2011.

According to Wiley publication policy, the reuse of the contents of the journals is
permitted for the authors (http://onlinelibrary.wiley.com/journal/10.1002/(ISSN)10969845/homepage/Permissions.html).

148

5.2 Abstract
The seismic response of a continuous 4-span bridge designed according to the current
Canadian seismic provisions is investigated using Incremental Dynamic Analysis (IDA).
The bridge is evaluated for two cases of restrained and unrestrained transverse
movements at the abutments. Different earthquake types, including shallow crustal
events, interface Cascadia subduction and deep inslab subduction, are considered in the
seismic analyses. The median collapse capacities calculated using different record
selection methods including CMS-based, UHS-based, and epsilon-based methods are
compared. In the CMS-based method, the conditional mean spectra (CMS) for three
different event types are constructed based on seismic deaggregation results. The median
collapse capacities obtained considering crustal events were generally close to those
obtained considering different event types. However, considering different event types in
the IDA resulted in increased record-to-record variability which should be incorporated in
probabilistic seismic performance assessments. The use of the epsilon-based method
generally resulted in the highest collapse capacity predictions. Although the use of the
CMS-based methods resulted in somewhat conservative predictions compared to those
obtained using the epsilon-based method, this method was less sensitive to the number of
records considered in the IDA.

KEY WORDS:

Record selection, Incremental dynamic analysis, Bridges,


Conditional mean spectrum, Seismic performance.

5.3 Introduction
Incremental Dynamic Analysis (IDA) (Vamvatsikos and Cornell, 2002) can be used
for seismic performance assessment of structures (ATC-63, 2008). One of the most
important issues in the IDA is record selection which can significantly affect the results.
This study investigates the effects of different record selection methods, including
different earthquake types on the seismic response of a bridge.

149

One of the methods that has been widely used for record selection is spectrum
matching. The use of scaled natural or artificial records that are matched to a target
uniform hazard spectrum (UHS) over a range of periods (referred to as the UHS-based
method) is recommended by some codes. This range must include the important modes of
the structure as well as the effects of period elongation due to inelastic deformation of the
structure. A period range from 0.2T1 to 2T1 (where T1 is the fundamental vibration period
of a structure) is usually recommended and used for this purpose. This range is similar to
the 0.2T1 to 1.5T1 range specified by ASCE 7-05 (ASCE, 2005), but statistical studies
suggest that nonlinear structures are often sensitive to response spectra at periods longer
than 1.5T1 (Baker and Cornell, 2008; Haselton and Baker, 2006; Vamvatsikos and
Cornell, 2005).

Although conceptually simple, these methods for record selection are deemed to be
conservative, particularly when the estimation of collapse capacity is the primary
objective of the analysis. This is because the response spectra of the records are forced to
match a target UHS. The UHS is a composite of predicted responses at different periods,
and may not be representative of individual ground motion spectra. Furthermore, the
UHS tends to be dominated at any individual period by motions above the median,
whereas individual spectra are unlikely to be above-median at all periods simultaneously.
No natural record can be found to match the UHS over a wide period range and hence the
analysis results using UHS-matched records are often unrealistic and conservatively
biased. To solve these problems, the use of the Conditional Mean Spectrum (CMS) in
lieu of the UHS has been recommended in recent years (Baker, 2011). In this study
selecting the records to match a CMS (conditioned on the fundamental period of the
structure, T1) is referred to as the CMS-based record selection method. This method is
discussed in more detail in section 2.1.

Another method which has been used for record selection is the epsilon-based method
(Baker and Cornell, 2006a). Epsilon, , is defined as the number of logarithmic standard
deviations of a target ground motion from a median ground motion prediction equation
(GMPE) for a given magnitude, M, and distance, R.

150

Rare ground motions (large

earthquake) typically have a peaked spectral shape that is much different than a standard
uniform hazard spectral shape and accounting for this has been shown to increase the
predicted collapse capacity significantly (increase of up to 70% (Baker and Cornell,
2006a; Haselton and Baker, 2006) . Thus neglecting the spectral shape of the records
selected (e.g., in UHS-based method) for collapse assessment of the structures can result
in very conservative and unrealistic predictions. The most direct approach to account for
spectral shape in structural analysis is to select ground motions with epsilon (T1) values
similar to epsilon values obtained from seismic hazard deaggregation analysis for the site
and hazard level of interest.

Usually only the records from crustal earthquakes have been used in seismic
performance assessments (e.g., ATC-63 provisions), while for some sites such as
Vancouver or Seattle subduction earthquakes with very different physical characteristics
(e.g., spectral content and duration) can occur. Another important issue which is
investigated in this research is the effect of considering different earthquake types (i.e.,
crustal, subduction interface, and inslab events) on the seismic performance evaluation of
bridges located in south-western British Columbia.

5.4 Seismic hazard analysis and conditional mean spectrum

5.4.1

Conditional mean spectrum (CMS)

The CMS provides the expected response spectrum, conditioned on occurrence of a


target spectral acceleration value at the period of interest. It has been found that the CMS
can be used as an appropriate target response spectrum for selecting ground motions as
input for dynamic analyses (Baker, 2011). In the development of a CMS, some important
aspects of the records including magnitude, M, distance, R, and epsilon are considered
from the deaggregation of the seismic hazard. Baker (Baker, 2011) proposed a method for
calculating the CMS, and this approach was used in this study. For the development of a
CMS, the mean values of spectral accelerations at different periods are computed using

151

an appropriate GMPE. These mean values will then be modified considering the interperiod correlations, standard deviations of spectral accelerations and mean epsilon values
at different periods. The resulting CMS has a peak value at the fundamental period of the
structure which is equal to the corresponding spectral acceleration from the UHS.

A complicated aspect in constructing a CMS for a site in south-western British


Columbia, in comparison with a site in California, is that three earthquake types, having
distinctly different characteristics, contribute to the overall seismic hazard. Therefore,
three CMS must be constructed for record selection, CMS-Crustal, CMS-Interface,
and CMS-Inslab (Goda and Atkinson, 2011). In order to account for these events, two
sets of CMS are considered in this study, similar to Goda and Atkinson (2011). The first
approach is the CMS-Event-based procedure which is based on the weighted average of
the CMS that are computed by using applicable GMPEs and the corresponding scenarios
for the three earthquake types, in proportion to the relative influences of the scenarios
(i.e., the number of records from each event type is proportional to the percentage of
contribution of that event type in the overall seismic hazard). The second approach is the
CMS-All-based which is the weighted average of CMS-Crustal, CMS-Interface,
and CMS-Inslab by considering relative influences of the individual earthquake types
without considering different earthquake type records in the selection procedure. In fact
CMS-All-based is the simplified version of CMS-Event-based. This simplification in
the CMS-All-based approach results in variability of the seismic response being
smaller than that based on the CMS-Event-based approach. However, it must be noted
that for each earthquake type, the variability of the selected records in general tends to be
underestimated for CMS-based or UHS-based approaches, since a tight match to a target
response spectrum is imposed (Baker, 2011). Nevertheless, the CMS-Event-based
approach can account for the variability of the CMS among different event types more
reasonably than does the CMS-All-based approach.
5.4.2

Preliminary record selection

Ground motion records that are used for the IDA are selected from two extensive
databases, the PEER-NGA database and the K-NET/KiK-NET database. Some
152

limitations were imposed on the minimum magnitude (M), peak ground acceleration
(PGA), and peak ground velocity (PGV) of the records to consider strong ground motions
available in the database. The characteristics of the records from the PEER-NGA
database (179 records from 28 earthquakes) and the K-NET/KiK-NET database (189
records from 23 earthquakes) are described by Goda and Atkinson (2009). More detailed
information regarding the records selected for this research and their criteria is available
elsewhere (Goda and Atkinson, 2009 and 2011).
5.4.3

Seismic hazard analysis and target scenarios

Probabilistic seismic hazard analysis (PSHA) is a standard procedure for seismic


hazard assessment. The results of such analysis usually include UHS and scenario events
that are associated with the UHS and identified by seismic hazard deaggregation. For
Canadian cities, the fourth generation of national seismic hazard maps of Canada are
available (Adams and Halchuk, 2003). In this study, an updated seismic hazard model for
western Canada, developed by Goda et al., (2010) will be used to take advantage of new
seismic information and seismological models to improve several aspects of the current
Geological Survey of Canada model.

PSHA for Vancouver was carried out using the updated seismic hazard model based
on Monte Carlo simulation (Goda et al., 2010). The obtained UHS for site class C (Vs30 =
555 m/sec, the average shear wave velocity in the top 30 m) are shown in Fig. 5.1. The
assessment is based on the simulated seismic activities during 5 million years, and the
considered annual non-exceedance probabilities range from 0.996 to 0.9999 (i.e., 10% to
0.5% probability of exceedance in 50 years). The UHS shown in Fig. 5.1b are mean
UHS, rather than median UHS. To further investigate the characteristics of seismic events
that contribute to a selected probability level, seismic hazard deaggregation for Sa(0.7)
and Sa(1.3) was carried out. These two periods correspond to the fundamental periods for
the bridge being considered for the restrained and unrestrained transverse movement at
the abutments (restrained abutments and free abutments), respectively. Deaggregation
results at different hazard levels in terms of mean magnitude M, mean distance R, and 
along with percentages of contributions of different seismic event types to the seismic
153

hazard for different probability levels are shown in Tables 5.1 and 5.2. The deaggregation
analysis is based on an approximately equal criterion (i.e., matching method) (Hong
and Goda, 2006), where seismic events reaching a seismic intensity level between 90%
and 110% of the target Sa(Tn) value are used to produce deaggregation results. It was
observed that the use of the exceeding method in seismic deaggregation lead to
predictions with higher mean epsilon values. This can be important especially when the
epsilon-based record selection method is used, potentially resulting in overestimation of
the capacity of the structures.

The constructed CMS at 2% probability of exceedance in 50 years are illustrated in


Figs. 5.2 and 5.3 for two vibration periods Tn = 0.7 sec and Tn = 1.3 sec, respectively. As
expected, all CMS are approximately equal to the spectral acceleration values of the UHS
at the target vibration periods considered (i.e.,Tn=0.7 and Tn=1.3). For the considered
cases, the CMS-Crustal and CMS-All are similar. The CMS-Interface has a rich spectral
content in the long vibration period range, which was found to be the most critical case
for the bridge structure studied, while the CMS-Inslab has a rich spectral content in the
short period range.

Table 5.1. Seismic deaggregation results for T= 0.7 sec (epsilon values using the BA08
and Z06 GMPEs)
Probability of
exceedance

Percentage of contribution:
[Crustal, Interface, Inslab]

Crustal events
[M,R,]

Interface events
[M,R,]

Inslab events
[M,R,]

10% in 50 yr

[48%, 17%, 35%]

[6.61, 37.2, 1.27]

[8.51, 142.4, 0.20]

[6.65, 76.1, 1.19]

2% in 50 yr

[40%, 27%, 33%]

[6.78, 18.3, 1.51]

[8.63, 142.8, 1.04]

[6.82, 64.6, 1.64]

0.5% in 50 yr

[38%, 30%, 32%]

[7.03, 10.3, 1.56]

[8.69, 141.4, 1.63]

[7.00, 55.9, 1.83]

154

Table 5.2. Seismic deaggregation results for T= 1.3 sec (epsilon values using the BA08
and Z06 GMPEs)
Probability of
exceedance

Percentage of contribution:
[Crustal, Interface, Inslab]

Crustal events

Interface events

Inslab events

[M,R,]

[M,R,]

[M,R,]

10% in 50 yr

[46%, 20%, 34%]

[6.74, 45.5, 1.30]

[8.49, 142.6, -0.1]

[6.78, 76.2, 1.28]

2% in 50 yr

[35%, 37%, 28%]

[6.86, 21.7, 1.51]

[8.63, 142.6, 0.79]

[6.94, 63.6, 1.77]

0.5% in 50 yr

[34%, 39%, 27%]

[7.13, 12.8, 1.67]

[8.69, 141.1, 1.34]

[7.09, 59.9, 2.10]

Meanepsilonvalue
Crustal
events
CrustalEvents

a)

Interface
events
InterfaceEvents

-3
3

-2
2

-1
1

00

11

22

33

Inslab
events
InslabEvents

0.06
0.04
0.02

9
8

0
0

Relativefrequency
Relativefrequency
Relativ e frequenc y

0.08

0.06

0.04

0.02

99
88

0
0
0

77

50

7
50

50
100

6
150
200

Distance(KM)

Magnitude

Magnitude

55

150

150

Distance (Km)

66

100

100

200
200

Distance (Km)

Distance(KM)

2.5
10%probabilityofexceedancein50years

b)

Spectralacceleration(g)

Relative frequency

Relativefrequency

0.08

2
2%probabilityofexceedancein50years
1.5
0.5%probabilityofexceedancein50years
1
0.5
0
0

0.5

1
1.5
2
NaturalvibrationperiodTn (sec)

2.5

Fig. 5.1. PSHA results for Vancouver: a) Seismic hazard deaggregation for 2%
probability of exceedance in 50 years at T=0.7 sec b) Uniform hazard spectra

155

Magnitude
Magnitude

UHS

Vancouver
Vs30=555 (m/sec)
2% probability of exceedance in 50 years
GMPEs: Z06 & BA08

a)

CMS-INSLAB
CMS-CRUSTAL
CMS-INTERFACE

0.5

UHS
CMS-INSLAB
CMS-CRUSTAL
CMS-INTERFACE

Spectral acceleration (g)

CMS-ALL

Spectral acceleration (g)

Vancouver
Vs30=555 (m/sec)
2% probability of exceedance in 50 years
GMPEs:AB03 & BA08

b)

CMS-ALL

0.5

0.05

0.05
0.2

Natural vibration period, Tn (Sec)

0.2

Natural vibration period, Tn (Sec)

Fig. 5.2. CMS for T=0.7 sec: a) for BA08 and Z06 GMPEs, b) for BA08 and AB03
GMPEs

Vancouver
Vs30=555 (m/sec)
2% probability of exceedance in 50 years
GMPEs: Z06 & BA08

CMS-INSLAB

Vancouver
Vs30=555 (m/sec)
2% probability of exceedance in 50 years
GMPEs:AB03 & BA08

b)

CMS-CRUSTAL
CMS-INTERFACE
CMS-ALL

Spectral acceleration (g)

Spectral acceleration (g)

a)

UHS

0.5

0.05

UHS
CMS-INSLAB
CMS-CRUSTAL
CMS-INTERFACE
CMS-ALL

0.5

0.05
0.2

Natural vibration period, Tn (Sec)

0.2

Natural vibration period, Tn (Sec)

Fig. 5.3. CMS for T=1.3 sec: a) for BA08 and Z06 GMPEs, b) for BA08 and AB03
GMPEs

The adopted ground-motion prediction equations and the corresponding weighting


factors used for seismic hazard analysis are provided by Atkinson and Goda (2011). For
constructing a CMS however using all of the GMPEs used in PSHA may not be
appropriate for practical reasons. It can be argued that contrary to conventional design
using a uniform hazard spectrum where the absolute values of spectral accelerations at
different periods are important, in CMS-based record selection strategy, the important
aspect is the spectral shape and as long as this shape can be predicted appropriately by a
limited number of GMPEs, the use of all the GMPEs may not be necessary. For the

156

purpose of this study the newly developed PEER-NGA relationship BA08 (Boore and
Atkinson, 2008) was used for crustal events. For the case of subduction events (interface
and inslab) the GMPE by Zhao et al., (2006) (Z06 GMPE) and Atkinson and Boore
(2003) (AB03 GMPE) were considered separately. A comparison of the constructed CMS
in Figs. 5.2 and 5.3 shows that in most cases, CMS-All which is the average of all the
CMS according to the percentage of contribution of each event type is very close to
CMS-Crustal. This may indicate that CMS-Crustal can be a replacement for CMS-All .
However this may be specific to the cases considered here and general conclusions
should not be made withouth further investigations.

5.5 Bridge properties, design and modelling assumptions

5.5.1

Bridge properties

To investigate the impact of the type of records on the structural performance, a


regular 4-span bridge is considered in this study. IDA (Vamvatsikos and Cornell, 2002) is
used to evaluate the seismic response of the bridge at different damage states. In typical
IDA analyses, records are scaled up or down until different performance levels, including
collapse, are reached. At high seismic levels, bridge shear keys at the abutments are
expected to fail, so that the bridge structure will be unrestrained at the ends. According to
the Caltrans recommendations, the shear keys must be designed to fail in low seismic
intensities to prevent damage to elements such as piles and abutments that are more
expensive and difficult to inspect and repair (Caltrans, 2006). However the Canadian
Highway Bridge Design Code (CHBDC) (CSA, 2006) does not recommend such a
design strategy and as a result, one may not be certain whether failure of the shear keys
will occur prior to the collapse of columns. Thus to consider the critical case in the
seismic performance assessment, two different cases including bridges with restrained
and unrestrained transverse movement at the abutments are evaluated.

157

The structure is a regular 4-span bridge supported by single columns with the
transverse movements restrained at the abutments. The column diameter is 1.5 meters, the
column height is 5 meters, the span length is 50 meters, and the superstructure consists of
a box girder with a uniform dead load of 200 kN/m. (see Fig. 5.4.) The bridge was
designed according to the 2006 CHBDC (CSA, 2006) with an importance factor of I =1.5
(i.e., Emergency-Route bridge). For the restrained case (effective shear keys), the first
period of the structure in the transverse direction is around T1=0.7 sec. In the case of
failed shear keys, the first period of the bridge is T1=1.3 sec. In this case, however, the
second mode, T2=0.7, has a significant contribution. The fundamental periods are
computed using the effective stiffness of the columns (i.e., Ke =My/ y where My and y
are the moment and rotation at yield, respectively). In straight bridges the responses in
the orthogonal directions are likely to be essentially independent (Priestley et al., 2007).
The responses of the bridge in the transverse direction were chosen to assess the
significance of the different ground motion types and different record selection methods.

Fig. 5.4. Bridge properties


5.5.2

Incremental dynamic analysis

IDA is usually used for performance assessment and collapse evaluation of structural
systems. This method involves numerous nonlinear time history analyses using a number
of records in which each record is scaled incrementally until a desired performance level
is reached.

158

One of the advantages of IDA is that one can evaluate the response of a structure at
multiple performance levels, including serviceability, damage control, and life safety,
according to the corresponding damage states that can be defined using appropriate
engineering demand parameters such as drift, rotation, and strain. The IDA results relate
statistics of a damage measure (DM) (i.e., inelastic seismic demand) with an intensity
measure (IM) (usually Sa(T1) with 5% damping ratio), and can be used in the
probabilistic seismic performance assessment and fragility analysis of the structure.

Figure 5.5 shows the predicted moment versus chord rotation response for the central
column of the bridge. Points A, B and C represent the damage states of reinforcement
yielding, cover spalling and bar buckling, respectively. The cover spalling stage is used
as a serviceability limit indicator based on experimental results on bridge columns (Berry
and Eberhard, 2007). Point C is used as a damage control indicator beyond which the
column is not repairable and represents the maximum capacity of the column (i.e., large
strains, bar buckling and concrete crushing). After the maximum capacity has been
reached at point C, there is strength degradation and failure of the structure will be
predicted when dynamic instability has been reached at point D which occurs with a
small incremental deformation beyond point C. The strength degradation slope (C-D) was
determined using a conservative approach suggested by Haselton et al. (2007), resulting
in a post-capping chord rotation of 10% (see Figure 5.5).

159

A - Yielding
B - Spalling
C - Bar buckling
D - Dynamic instability

l= 2.04%
s= 1.2%
H = 5000 mm
D = 1500 mm

Mc=13425 kN.m
My=12095 kN.m

C
B

Moment (M)

pc
Ke

y= 0.59% spalling= 1.8%

cap= 5.3%

collapse= 15.3%

Chord rotation ()


Fig. 5.5. Backbone curve parameters

5.5.3

Damage states and performance indicators

Several damage states were considered in the seismic evaluation of the bridge under
study including yielding, spalling, and collapse. ATC-63 provisions provide some
equations to compute the parameters of the backbone curve as shown in Fig. 5.5 for
building-type elements with rectangular cross-sections. To apply this methodology to
bridge structures similar back-bone curves were defined for the circular bridge columns
considered in this research using the results obtained from bridge-specific studies.
Generally two types of approaches (i.e., experimental and theoretical methods) can be
used for this purpose.

The Mander et al. equation (Mander et al., 1988) is typically used for the theoretical
prediction of the ultimate strain of concrete in the confined core. Experimental results
however indicate that the ultimate strains predicted by this equation are conservative in
comparison with actual ultimate strains exceeding predicted values by a factor of about

160

1.3 to 1.6. Usually a strain level of 0.6 to 0.7 times the ultimate strain of steel bars in
monotonic tests is recommended for calculating the ultimate curvature of the section with
an upper strain limit of 0.05 (Priestley et al., 1996 and 2007) for damage control
assessment. In order to attain this level of strain, the amount and spacing of the transverse
confinement reinforcement must be code-conforming. However, for life safety
assessment, the value of 0.9 times the ultimate strain of steel bars (0.9 su) is
recommended by Priestley et al. and this value should not be taken greater than 0.08
(Priestley et al., 2007).

A study by Berry and Eberhard (2007) provides some empirical equations to estimate
the engineering demand parameters including drift ratio, plastic rotation, and strain in the
longitudinal bars for circular bridge columns based on the properties of columns
including longitudinal and transverse steel ratio, axial load ratio, and geometry. A
comparison between these approaches shows that the ultimate drifts calculated using
these methods are in good agreement (e.g., the drift ratio of 5.69% corresponding to the
life safety limit state using the theoretical approach is close to that obtained for the bar
buckling and bar fracture damage states using the experimental equations as shown in
Table 5.3a and 5.3b, respectively). In this research, the ultimate tensile strain in the steel
bars corresponding to bar buckling from experimental equations is used to compute the
ultimate curvature of the columns and the corresponding drift and curvature ductility is
defined as the point at which strength degradation begins (i.e., cap in Fig. 5.5).

161

Table 5.3. Deformations at different damage states for the central column of the
bridge: a) Using the theoretical approach by Priestley et al., (2007), b) Using the
experimental equations by Berry and Eberhard (2007).
Damage state

Drift

Yielding (nominal)

0.59%

Serviceability

0.93%

Damage Control

3.91%

Life Safety

5.69%

a)

Strain

c =0.004 or s =0.015
c < Mander eq.or
s =0.6su < 0.05
c < 1.5*Mander eq.
or s =0.9su< 0.08

Curvature Displacement
b)
ductility
ductility

Damage state

Drift

Rotation Strain

Cover Spalling

1.85%

0.012

0.008

2.65

1.58

Bar Buckling

5.30%

0.0451

0.075

15.9

6.2

Bar Fracture

5.74%

0.0492

0.081

25.6

9.6

Few test results are available to calibrate the post-capping stiffness of the columns. In
ATC-63 a conservative upper limit of 0.1 is recommended for the post-capping chord
rotation of columns which is controlling for most beams and columns designed according
to current seismic design practice. In addition, P- effects are considered directly in the
analysis.

5.5.4

Bridge modelling

The modified Takeda hysteresis model (Otani, 1981) was used in this study to model
the behavior of the RC columns using Ruaumoko software (Carr, 2009). This model has
two main parameters, alpha and beta, which control the unloading and the reloading
stiffness, respectively. Alpha is usually in the range of 0.0 to 0.5 and beta varies between
0.0 and 0.6. Increasing the alpha parameter decreases the unloading stiffness and
increasing the beta parameter increases the reloading stiffness. Some researchers
computed the mean values of alpha and beta for bridge columns based on experiments, as
0.26 and 0.49, respectively (Mechakhchekh, 2008). By considering the recommended
values of these parameters (Priestley et al., 1996 and 2007), the values of alpha=0.3 and
beta=0.3 were adopted to comply with both the recommended values in practice and the
mean values from the tests. The lower value of beta was chosen to be more conservative
and to account for possible defects and pinching effects in the hysteresis loops. It has

162

been demonstrated that for a structure with ductile columns and a period in the mediumto-long range, the maximum response of the structure is relatively insensitive to the
hysteretic parameters (Otani, 1981; ATC-62, 2008). It was also found that the median
collapse capacity of the structure is insensitive to the stiffness degradation parameters in
the Takeda model in which the differences in median collapse capacities were negligible
when significantly different values of alpha and beta were considered (see Figure 5.6(b)).
The structural modelling considered in this study is similar to that used by Priestley et al.,
(Priestley et al., 1996 and 2007), except that the backbone curve shown in Figure 5.5,
including the post peak response was used.

dp

Fy

E dp

IDA result percentiles


(Sa @T1=0.7 sec)

dy D
K u K ( )D
y

K0

dm
m

d
Ku

dy

dm

b)

a)

Parameters:

Median

16%

84%

Alpha=0.5 & Beta=0

1.26

0.80

2.27

Alpha=0.3 & Beta=0.3

1.31

0.82

2.30

Alpha=0.2 & Beta=0.5

1.34

0.84

2.30

Fig. 5.6. Effects of hysteresis parameters on the collapse capacity: a) Modified Takeda
hysteresis loops; b) The collapse capacity predictions using the ATC-63 record set for
different Takeda hysteresis loop parameters (alpha and beta).

5.5.5

Collapse modes and prediction of probability of collapse

The bridge under study was designed and detailed to meet the code requirements for
ductile response, including capacity design concepts and adequate support lengths at the
abutments, hence promoting flexural yielding and eliminating brittle failure mechanisms.
The ductile columns contain code-compliant spiral reinforcement to confine the concrete,
avoid shear failure and to control buckling of the vertical reinforcing bars. For this
continuous bridge, with all other failure modes avoided, the flexural response governs the
response of the bridge and sidesway collapse is the governing collapse mechanism. The

163

collapse prediction is based on dynamic instability of the structure (Vamvatsikos and


Cornell, 2002; ATC-63, 2008). The failure of shear keys is treated indirectly by
considering two cases of restrained and unrestrained transverse movements at the
abutments.

Because this research focuses on the influence of different record selection methods
and earthquake types on the seismic response of bridges, a ductile bridge structure is
considered with non-ductile collapse modes avoided. It is noted that poorly designed and
detailed bridges would have undesirable, brittle collapse modes which would give rise to
excessive variability in the predicted responses.

To evaluate the seismic performance of a structure at collapse, the median collapse


capacity of the structure is computed using IDA. The probability of collapse at the
maximum considered earthquake (MCE) level, which usually corresponds to 2%
probability of exceedance in 50 years, should be reasonably low. The probability of
collapse at this seismic excitation level is typically limited to 10% (ATC-63, 2008). To
estimate the probability of collapse at the MCE level based on the computed median
collapse capacity a simplified assumption is made that the cumulative probability
function follows a lognormal distribution.

The collapse margin ratio, CMR, which is the ratio of the median 5%-damped spectral
acceleration of the collapse level ground motions to the 5%-damped spectral acceleration
of the MCE ground motions, can then be computed at the fundamental period of the
structure (ATC-63, 2008).

To succinctly describe the uncertainty regarding the collapse capacity of the bridges,
the standard deviation of the collapse capacities, expressed in terms of spectral
acceleration, can be used. These quantities are reported in the results for individual
analysis cases.

164

5.6 Seismic evaluation using different record selection methods


The response spectra shapes for different record selection methods provide input for
the seismic evaluation of structures using IDA. The response spectra shapes can be
represented by the normalized average spectral shapes; those for different record
selection methods are shown in Fig. 5.7. A lower spectral value in the long period range
typically results in a higher collapse capacity prediction. The spectral accelerations in the
short period range are also important, if the response of the structure is influenced by
higher-mode effects. The comparison between the spectral shapes obtained using the
considered methodologies indicate that generally the results using CMS-All-based and
CMS-Event-based methods are close and the use of the epsilon-based method usually
results in the highest collapse capacity prediction.

Normalized average response spectra

Normalized average response spectra

a)

(Sa@T) / (Sa@T1=1.3)

(Sa@T) / (Sa@T1=0.7)

CMS-EVENT-BASED
CMS-ALL-BASED
EPSILON-BASED
UHS-BASED

0.25
0.14

Natural period (T) Sec

b)

2.5

2.5

1.4

0.25
0.26

CMS-EVENT-BASED
CMS-ALL-BASED
EPSILON-BASED
UHS-BASED

Natural period (T) Sec

Fig. 5.7. Normalized average spectra from different methods at a) T=0.7 sec, b) T=1.3 sec

However, based on the results obtained here, one can see that the difference between
the CMS-based methods and epsilon-based method is greater for the case of T1=1.3 sec
(e.g., first fundamental period of the bridge with failed shear keys) compared to the case
of T1=0.7 sec. This is because in the case of T1=1.3 sec the beneficial effects from
spectral shapes using CMS-based methods were small, while at T1=0.7 sec the use of the
CMS-based methods can result in collapse capacities similar to those obtained using the
epsilon-based method. It was also found that the predicted collapse capacities using the
165

2.6

CMS-based methods were not as sensitive to the annual probability of exceedance as the
epsilon-based methods. In the epsilon-based method, the epsilon values are computed
using the BA08 GMPE for crustal events and Z06 GMPE for subduction events.

5.6.1

Seismic evaluation using crustal records

The performance of the structure is evaluated considering only crustal events in the
epsilon-based method and compared to the case of using the ATC-63 fixed set of farfield
records.

5.6.1.1 Bridge with restrained abutment condition (T1=0.7 sec)


The results obtained using the epsilon-based record selection may be sensitive to the
record sets used for analysis and sometimes different sets of records with similar mean
epsilon values could give different results. As an example, in Fig. 5.8 the collapse
capacity is about 1.84 g for records selected based on mean epsilon value of 1.56 from
seismic deaggregation for 0.5% probability of exceedance in 50 years, while the collapse
capacity calculated using another set of records with a similar mean epsilon was about
2.07g. The standard deviations reported in the tables reflect the logarithmic standard
deviations of the collapse predictions considering all of the records studied and do not
include other sources of uncertainty.

166

Spectral acceleration at T=0.7 sec (g)


(IM)

Epsilon-Based (Crustal events)


3.5
3
2.5

EPSILON-Based

(CRUST AL) 0.5% IN 50 YRS


1.5

Damage states

0.02

0.04

0.06

0.08

Median 16%

84%

St Dev

Yielding

0.14

0.13

0.14

0.09

Cover Spalling (Exp.)

0.66

0.47

0.81

0.28

Collapse

1.84

1.12

2.62

0.42

0.5
0

IDA result percentiles


(Sa @T1=0.7)

0.1

Maximum column Drift (DM)

Fig. 5.8. IDA results using the epsilon-based method (Crustal event,T1=0.7 sec,
CMR=3.5)

Spectral acceleration at T=0.7 sec (g)


(IM)

ATC-63 records
3
2.5
2

ATC-63

1.5
1

Damage states

0.5
0
0

0.02

0.04

0.06

0.08

0.1

IDA result percentiles


(Sa @T1=0.7)
Median 16%

84%

St Dev

Yielding

0.15

0.14

0.15

0.05

Cover Spalling (Exp.)

0.55

0.41

0.78

0.31

Collapse

1.31

0.82

2.32

0.50

Maximum column Drift (DM)

Fig. 5.9. IDA results using the ATC-63 record set at T1=0.7 sec (CMR=2.531.26=3.2)
The IDA results based on a fixed record set recommended by the ATC-63 provisions,
shown in Fig. 5.9, should be modified for the effects of epsilon which is a proxy measure
for the spectral shape effects (Baker and Cornell, 2006a; ATC-63, 2008). The probability
of exceedance of 0.5% in 50 years is recommended to calculate the mean epsilon values
for ductile structures at collapse level (ATC-63, 2008). The regression analysis is carried
out for the epsilon values of each record at the fundamental period of the structure versus
the corresponding collapse capacity of each record. The result of such a regression
analysis is given in a form of ln(Sacollpase) = 1  + 0. The 1 factor represents the
sensitivity of the collapse capacity to a change in the epsilon value (Haselton et al.,
2011). The 1 factor which is used to modify the collapse capacity was 0.317 based on
the BA08 GMPE. The 1 value using the AS97 GMPE (Abrahamson and Silva, 1997)
was found to be 0.309 showing that this value is insensitive to the GMPE used to
167

calculate the epsilon values. This is in agreement with the values reported in the ATC-63
for ductile moment frames ( i.e., 0.311). The simplified spectral shape factor, SSF, is
calculated as follows (Haselton et al., 2011):

[5.1] SSF

exp[E1 (H 0 (T )  H (T )records )]

where the mean epsilon of the records at T1=0.7 sec, H (T )records , is found to be 0.82 using
the BA08 GMPE and the mean epsilon value from seismic deaggregation for 0.5%
probability of exceedance in 50 years for crustal events, H 0 (T ) , is 1.56. The SSF factor
then can be calculated as 1.26 (i.e. SSF=exp[0.317(1.56-0.82)]=1.26). Thus the
modified collapse capacity can be calculated as 1.261.31 g=1.65 g. The median collapse
capacity of 1.65 g is comparable to the value of 1.84g calculated based on the epsilonbased record selection method and yet more conservative.

5.6.1.2 Bridge with free abutments (T1=1.3)


The IDA results using the epsilon-based method for this case is shown in Fig. 5.10.
The use of the epsilon-based method resulted in a collapse capacity of 1.08g and a
standard deviation of 0.45.

Spectral acceleration at T=1.3 sec (g)


(IM)

Epsilon-Based (Crustal events)


2.5

Epsilon-Based

1.5

(Crustal records)
1

Damage states

0.5

0
0

0.02

0.04

0.06

0.08

IDA result percentiles


(Sa @T1)
Median 16%

84%

St Dev

Yielding

0.12

0.08

0.13

0.28

Cover Spalling (Exp.)

0.42

0.30

0.56

0.34

Collapse

1.08

0.67

1.66

0.45

0.1

Maximum column Drift (DM)

Fig. 5.10. IDA results using the epsilon-based record selection method at T1=1.3 sec
(CMR=3.8)

168

Spectral acceleration at T=1.3 sec (g)


(IM)

ATC-63 records
2
1.8
1.6
1.4

ATC-63 fixed record


set

1.2
1

IDA result percentiles


(Sa @T1)

0.8

Damage states

0.6
0.4

0.02

0.04

0.06

0.08

84%

St Dev

Yielding

0.09

0.05

0.11

0.35

Cover Spalling (Exp.)

0.32

0.24

0.41

0.28

Collapse

0.79

0.42

1.21

0.44

0.2
0

Median 16%

0.1

Maximum column Drift (DM)

Fig. 5.11. IDA results using the ATC-63 record setat T1=1.3 sec (CMR= 1.36 * 2.8 = 3.8)
The IDA results using the fixed record set in ATC-63, shown in Fig. 5.11, need to be
modified for the effects of spectral shape. The 1 factor in this case was equal to 0.3
based on the BA08 GMPE. The simplified spectral shape factor, SSF (Eq. [5.1]), is
calculated as 1.36 and the corresponding modified collapse capacity can then be
calculated as 1.360.79 g=1.07g. This is similar to the value of 1.08g calculated based on
the epsilon-based record selection method demonstrating that the simplified method can
predict collapse capacities that are in good agreement with those obtained using the
epsilon-based record selection method. However a question still remains whether the
epsilon-based record selection method is reliable enough for seismic evaluations. More
investigations should be carried out to verify this problem.

5.6.1.3

Bridge with free abutments (T2=0.7 Sec)

The modified collapse capacity in this case was computed as 1.351.36 g=1.84 g.
Again the modified collapse capacity is similar to that obtained using the epsilon-based
record selection method which predicts a value of 1.82g as the collapse capacity of the
structure.

169

5.6.1.4 Longitudinal response

The response of the bridge in the longitudinal direction was also studied considering
the effects of abutment strength and stiffness. In practice however the abutment restraints
are typically neglected which leads to a conservative design. To model the abutments the
spring abutment model by Aviram et al., (2008) was used. The analyses were carried out
for 4 cases including neglecting the abutment stiffness and strength, and considering
abutments with no piles, 10 piles and 20 piles. The stiffness and strength of the
abutments, soil and piles were computed based on the recommendations by Caltrans
(2006).
The gaps between the abutment and bridge deck at the ends were also considered in
the abutment model. As shown in Fig. 5.12, the CMR values are higher in the
longitudinal direction than in the transverse direction (see Fig. 5.11) when the influence
of abutments are included in the structural model. This indicates that the transverse
response of the bridge is controlling. Increasing the number of piles had minor effects on
the collapse capacity.

Spectral acceleration at T=0.85 sec (g)


(IM)

Longitudinal direction- ATC-63 records


3
2.5

IDA results
Sa@T1 =0.85 Sec (g)

Longitudinal direction ATC-63 record sets

Case

1.5
1

Median 16%

84%

Modified
median

Modified
CMR

No abutments

0.92

0.58

1.60

1.15

2.72

0.5

Abutments with no piles

1.34

0.86

1.92

1.68

3.96

Abutments with 10 piles

1.44

0.94

1.94

1.80

4.26

Abutments with 20 piles

1.48

0.98

1.96

1.85

4.37

0.02

0.04
0.06
Maximum column Drift (DM)

0.08

0.1

Fig. 5.12. IDA results in the longitudinal direction using 44 crustal records

5.6.1.5

Three dimensional versus two dimensional analysis

Priestley et al., (2007) concluded that the seismic responses in the orthogonal
directions for straight bridges and symmetrical buildings are essentially independent and
the current state-of-the-art is to model the ductility effects independently. However, a
170

comparison of 2D and 3D nonlinear analyses was made in order to investigate the


differences.

Because ground motions records are applied in pairs in three-dimensional nonlinear


dynamic analyses, the resulting behaviour from each ground motion component is
coupled. It has been found that the median collapse capacity resulting from three
dimensional analyses is on average about 20% less than the median resulting from twodimensional analyses (ATC-63, 2008). The application and scaling of pairs of ground
motion records in IDA for three-dimensional analyses introduces a conservative bias that
is not present in two-dimensional analyses (ATC-63, 2008).

It should also be noted that the comparison of the record selection methods including
CMS-based and epsilon-based methods, which is the main focus of this study, requires a
fundamental period to be defined for a structure so that the CMS and epsilon values at
this period can be computed. The use of the 3D analysis increases the number of
variables and makes the application and comparison of such methods complicated (e.g.,
the period to be used to compute a CMS or compute epsilon values). More research on
such subjects is required.

Table 5.4. Results of IDA using a 3D analysis for different abutment stiffness and
strength using 22 pairs of records applied twice at different principal directions (44 cases)
IDA result percentiles at collapse (g)

3D analysis results
Case
Abutments with no piles
(Fixed record set)
Abutments with no piles
( Epsilon-based method)

Median 16%

84%

Modified median
Modified CMR for Modified CMR
St Dev
for spectral shapes
spectral shapes
for 3D analysis

0.78

0.52

1.40

0.99

0.41

2.11

2.54

1.06

0.64

1.58

1.06

0.52

2.27

2.72

A series of 3D analyses was carried out in which the twenty-two pairs of records were
applied twice to the model, once with the ground motion records oriented along one
principal direction, and then again with the records rotated 90 degrees. The predicted
median collapse and percentiles using 3D analysis are reported in Table 5.4 in terms of

171

the geometric means of the pair of ground motions at the fundamental periods (i.e., T=0.7
sec and T=0.85 sec in the transverse and longitudinal directions, respectively). While the
collapse capacity of 0.99(g) was predicted using 3D analysis, the corresponding
geometric mean of collapse capacities using 2D analysis for the transverse and
longitudinal directions was 1.66(g) (i.e., 1.65(g) in the transverse direction and 1.68(g) in
the longitudinal direction ). This indicates that the predicted collapse capacity from 3D
analysis is about 40% less than that obtained using 2D analysis. Applying the records in
pairs may induce some conservative bias in the prediction of the collapse capacities and
in the ATC-63 provisions this conservatism is accounted for by increasing the predicted
median collapse capacity by 20% (i.e., the average differences). In the case of the bridge
under study even higher differences (i.e., about 40%) were observed. Therefore care
should be taken when treating the IDA results obtained using a 2D analysis, however the
IDA results obtained from 3D analyses may be conservatively biased (ATC-63, 2008).
While more research is required to study the differences between 2D and 3D analyses,
this research focuses on the 2D nonlinear analysis in the transverse direction to compare
the effects of different earthquake types and record selection methods.

5.6.2

Seismic evaluation considering all event types

Record sets used for the IDA should include all the probable earthquake types for the
site under consideration. Contrary to the last section where crustal events only were
considered in the IDA, in this section records from other event types (i.e., subduction
interface and inslab events) are also considered. Several record selection methodologies
are investigated for the seismic evaluation of the bridge under study using IDA. These
methods include the epsilon-based and CMS-based record selection methods. The CMSbased method can be carried out in several ways such as CMS-Event-based and CMSAll-based (Goda and Atkinson, 2011). For the CMS-Event-based and epsilon-based
methods, the number of records considered in the IDA from each event type is
proportional to the percentage of contribution of that event type from seismic
deaggregation.

172

5.6.2.1

Bridge with fixed abutments (T1=0.7 Sec)

The IDA results using the CMS-based methods and considering different GMPEs are
shown in Fig. 5.13 to Fig. 5.15 and Table 5.5. For the CMS-Event-based and CMS-Allbased record selection methods, results are generally very similar, even using different
GMPEs considered in this study (i.e., AB03 and Z06). This may be partially due to the
fact that the Z06 GMPE is more conservative for interface events and less conservative
for inslab events, while for the AB03 GMPE the opposite is true. Thus the considered
GMPEs have counteracting effects for interface and inslab events, so that the average
CMS in both cases are close to the CMS-Crustal using the BA08 GMPE. Although the
median values using the CMS-Event-based and the CMS-All-based methods are similar,
the standard deviations were higher in the case of the CMS-Event-based method than
those obtained based on the CMS-All-based method as expected.

To investigate the sensitivity of the results from the CMS-based method to the number
of the records used, the IDA was carried out using 78 records as shown in Fig. 5.15. This
resulted in similar predictions using 44 records which indicates that the CMS-based
methods are relatively insensitive to the number of records used in the IDA.

The IDA results using the epsilon-based method are summarized in Table 5.6 for 0.5%
and 2% probability of exceedance in 50 years. It can be seen that although the median
collapse capacity prediction for 0.5% probability of exceedance in 50 years is a little
lower than that for 2% probability of exceedance in 50 years, the 16% and 84%
percentiles are higher. This may indicate that the use of 44 records is not sufficient in this
case to capture the median collapse capacity accurately. The use of 78 records in the IDA
increased the median prediction by about 25% as shown in Fig. 5.16. This demonstrates
that the epsilon-based method can be sensitive to the number of records used in the
analysis.The median collapse capacity of 1.82 g obtained in this case is similar to that
obtained before using only crustal records (e.g., see Fig. 5.8).
A comparison of the UHS-based method, shown in Fig. 5.17, with the CMS or epsilonbased methods at T=0.7 sec indicates that accounting for spectral shape and epsilon can

173

increase the predicted collapse capacity by 40% to 75%. Similar results for the case of
buildings, have been reported (Baker and Cornell, 2006a).

Spectral acceleration at T=0.7 sec (g)


(IM)

CMS-Event-Based (Z06 GMPE)


3.5
3
2.5

CMS-EVENT-BASED IDA result percentiles


(Z06 GMPE)
(Sa @T 1)

2
1.5
1
0.5
0
0

0.02

0.04

0.06

0.08

Damage states

Median

16%

84%

St Dev

Yielding

0.14

0.12

0.15

0.27

Cover Spalling (Exp.)

0.57

0.36

0.77

0.38

Collapse

1.38

0.74

2.67

0.58

0.1

Maximum column Drift (DM)

Fig. 5.13. IDA results (CMS-Event based method, T1=0.7) for Z06 GMPE (CMR=2.7)

Spectral acceleration at T=0.7 sec (g)


(IM)

CMS-Event-Based (AB03 GMPE)


2.5

CMS-EVENT-BSD
(AB03 GMPE)

1.5

IDA result percentiles


(Sa @T1)

Damage states
0.5

0
0

0.02

0.04

0.06

0.08

Median 16%

84%

St Dev

Yielding

0.13

0.12

0.15

0.43

Cover Spalling (Exp.)

0.52

0.38

0.76

0.41

Collapse

1.42

0.72

2.15

0.52

0.1

Maximum column Drift (DM)

Fig. 5.14. IDA results ( CMS-Event-based method, T1=0.7) for AB03 GMPE (CMR=2.8)

Increasing the ductility capacity of the structure and the resulting larger period
elongation can lead to even higher collapse capacities due to higher beneficial effects of
spectral shapes in longer periods. Comparisons of the results obtained for yielding and
spalling damage states indicate that different record selection methods had minor effects
on these limit states. This is expected, as the effect of spectral shapes becomes more
important at higher ductility levels. Figs. 5.13 to 5.17 indicate that typically the standard
deviations are lower for the yielding and spalling states than for the collapse state. Based
on the results presented it is important to note that the variability of the results are

174

typically higher for the epsilon-based and the CMS-Event-based methods than the CMSAll-based and UHS-based methods .

Spectral acceleration at T=0.7 sec (g) (IM)

CMS-Event-Based (Z06 GMPE) using 78 records


4
3.5
3

IDA result percentiles


(Sa @T1)

CMS-EVENT-BSD
(Z06 GMPE)

2.5
2
1.5

Damage states

Median 16%

84%

St Dev

Yielding

0.13

0.11

0.15

0.43

Cover Spalling (Exp.)

0.51

0.38

0.69

0.44

Collapse

1.38

0.74

2.54

0.58

0.5
0
0

0.02

0.04
0.06
Maximum column Drift (DM)

0.08

0.1

Fig. 5.15. IDA results (CMS-Event based method, T1=0.7) for Z06 GMPE using 78
records

Table 5.5. IDA results (CMS-All-based method, T1=0.7) for a) Z06 GMPE
b) AB03 GMPE
IDA result percentiles
(Sa @T1)

CMS-ALL-BASED
(Z06 GMPE)
Damage states

a)

Median 16%

IDA result percentiles


(Sa @T1)

CMS-ALL-BASED
(AB03 GMPE)

84%

St Dev

Damage states

Median 16%

84%

St Dev

Yielding

0.14

0.13

0.15

0.35

Yielding

0.15

0.14

0.16

0.28

Cover Spalling (Exp.)

0.57

0.45

0.69

0.34

Cover Spalling (Exp.)

0.57

0.46

0.78

0.31

Collapse

1.42

1.06

2.25

0.44

Collapse

1.45

0.99

2.28

0.44

b)

Table 5.6. Epsilon-based method for probability of exceedance in 50 years of a) 2% and


b) 0.5% probability of exceedance in 50 years
EPSILON-BASED
2% IN 50 YRS
Damage states

IDA result percentiles


(Sa @T1)
Median 16%

EPSILON-BASED
0.5% IN 50 YRS

84% St Dev

Damage states

IDA result percentiles


(Sa @T1)
Median 16%

84%

St Dev

Yielding

0.13

0.11

0.15

0.41

Yielding

0.14

0.12

0.16

0.39

Cover Spalling (Exp.)

0.56

0.32

0.77

0.42

Cover Spalling (Exp.)

0.59

0.41

0.90

0.47

Collapse

1.47

0.72

2.22

0.52

Collapse

1.34

0.91

3.02

0.60

175

Epsilon-Based method using 78 records (0.5% in 50 yrs)


Spectral acceleration at T=0.7 sec (g) (IM)

4
3.5
3

EPSILON-BASED
0.5% IN 50 YRS

2.5
2

Damage states

1.5
1

IDA result percentiles


(Sa @T1)
Median 16%

84%

St Dev

Yielding

0.18

0.14

0.23

0.43

Cover Spalling (Exp.)

0.71

0.46

1.25

0.52

Collapse

1.82

0.93

3.09

0.62

0.5
0
0

0.02

0.04
0.06
Maximum column Drift (DM)

0.08

0.1

Fig. 5.16. IDA results for Epsilon-based method using 78 records (CMR=3.5)

Spectral acceleration at T=0.7 sec (g)


(IM)

UHS-Based (All event types)


2.5

UHS-BASED

1.5

Damage states

0.5

0
0

0.02

0.04

0.06

0.08

IDA result percentiles


(Sa @T1)
Median 16%

84%

St Dev

Yielding

0.13

0.08

0.16

0.44

Cover Spalling (Exp.)

0.47

0.31

0.64

0.38

Collapse

1.04

0.62

1.53

0.47

0.1

Maximum column Drift (DM)

Fig. 5.17. IDA results based on the UHS-based record selection at T1=0.7 sec
(CMR=2.0)

5.6.2.2 Bridge with free abutments (T1=1.3 Sec)


The IDA results obtained using different record selection methods for the case of
bridge with unrestrained abutments at the period of T1=1.3 sec are shown in Figs. 5.18 to
5.21 and Table 5.7.

176

Spectral acceleration at T=1.3 sec (g)


(IM)

UHS-Based (All event types)


1.4
1.2

UHS-BASED

IDA result percentiles


(Sa @T1)

Damage states

Median 16%

1
0.8
0.6

84% St Dev

0.4

Yielding

0.09

0.05

0.13

0.54

Cover Spalling (Exp.)

0.29

0.18

0.40

0.49

Collapse

0.50

0.32

0.90

0.67

0.2
0
0

0.02

0.04

0.06

0.08

0.1

Maximum column Drift (DM)

Fig. 5.18. IDA results using the UHS-based record selection method at T1=1.3 sec
(CMR=1.8)

While the use of the CMS-Event-based and CMS-All-based methods again resulted in
similar collapse capacity predictions, the use of the epsilon-based record selection
resulted in a higher collapse capacity prediction. Increasing the number of records to 78
decreased the predicted median collapse capacity by 30% in the epsilon-based method, as
shown in Fig. 5.20, while this increase in the number of records had a minor impact on
the IDA results using the CMS-based method as shown in Figure 5.21 and Table 5.7.
This again indicates that the IDA results using the epsilon-based record selection method
can be sensitive to the number of records.

Spectral acceleration at T=1.3 sec (g)


(IM)

Epsilon-Based (All events)


3
2.5
2
1.5

Epsilon-Based
0.5% in 50 yrs

IDA result percentiles


(Sa @T1)

Damage states

Median 16%

0.02

0.04

0.06

0.08

St Dev

Yielding

0.12

0.07

0.18

0.51

Cover Spalling (Exp.)

0.39

0.22

0.69

0.59

Collapse

1.07

0.34

2.25

0.83

0.5
0

84%

0.1

Maximum column Drift (DM)

Fig. 5.19. IDA results using epsilon-based record selection method at T1=1.3 sec
(CMR=3.8)

177

Epsilon-Based 0.5% in 50 yrs (All events)


Spectral acceleration at T=1.3 sec (g) (IM)

2.5

EPSILON-BASED
0.5% IN 50 YRS

1.5

Damage states

0.02

0.04
0.06
Maximum column Drift (DM)

0.08

Median 16%

84%

St Dev

Yielding

0.11

0.07

0.13

0.56

Cover Spalling (Exp.)

0.37

0.21

0.58

0.57

Collapse

0.74

0.34

1.83

0.80

0.5

IDA result percentiles


(Sa @T1)

0.1

Fig. 5.20. IDA results for Epsilon-based method using 78 records (CMR=2.62)

The use of the CMS-based record selection in this case increased the collapse capacity
prediction by about 25% compared to the results obtained based on the UHS-based
method. On the other hand the use of the epsilon-based record selection method increased
the predicted collapse capacity by 50%. A comparison of the results indicates that the
16% percentiles from different record selection methods are in good agreement. However
the difference between the results becomes more significant for higher percentiles (see
Figures 18 to 21 and Table 5.7).

Table 5.7. IDA results at T1=1.3 sec using 44 records for a) CMS-All-based method
(CMR=2.2) and b) CMS-Event-based method (CMR=2.14)
CMS-ALL-BASED
Damage states

a)

IDA result percentiles


(Sa @T1)
Median

CMS-EVENTBASED

IDA result percentiles


(Sa @T1)
Median 16%

16%

84%

St Dev

Damage states

84%

St Dev

0.10

0.34

Yielding

0.09

0.08

0.12

0.30

Yielding

Cover Spalling (Exp.)

0.33

0.26

0.42

0.29

Spalling (Exprimental)

0.28

0.19

0.42

0.38

Collapse

0.62

0.44

0.88

0.39

Collapse

0.60

0.34

0.92

0.51

b)

178

0.09

0.08

CMS-EVENTBASED

IDA result percentiles


(Sa @T1)

Damage states

Median

16%

84%

St Dev

Yielding

0.10

0.08

0.12

0.38

Cover Spalling (Exp.)

0.32

0.23

0.44

0.41

Collapse

0.57

0.35

0.97

0.60

Fig. 5.21. IDA results using CMS-Event-based method by using 78 records (CMR=2.02)

5.6.2.3

Bridge with free abutments (T2= 0.7 Sec)

For the case of free abutments with a period of T2=0.7 sec corresponding to the second
important mode, using CMS-Event-based and CMS-All-based methods again resulted in
similar collapse capacity predictions of 1.17g and 1.22g respectively. However, as noted
before, the variation of the IDA results was larger for the case of CMS-Event-based (i.e.,
standard deviation of 0.54 versus 0.7). The collapse capacity obtained based on the
epsilon-based method (i.e., 1.43g) is once again higher than those obtained using the
CMS-based methods. By comparing the results obtained using the UHS-based record
selection it was found that the use of the CMS-based methods can increase the predicted
collapse capacity by 30% (e.g., increase from 0.9g to 1.2 g). On the other hand the
increase in the collapse capacity prediction using the epsilon-based record selection
method is around 60% (i.e., increase from 0.9g to 1.43g).

5.6.3

Sensitivity of the results to the number of records

To provide more rigorous results regarding the effects of the number of records and to
investigate the impact of different earthquake types on the collapse capacity individually,
larger record sets were considered. For this purpose 78 fixed records with the highest
PGA and PGV values were selected for each earthquake type (i.e., a total of 3*78= 234

179

records) (Tehrani, 2012). The unmodified and modified median collapse capacities for
the spectral shape effects are presented in Table 5.8. The results obtained for combining
all records in proportion to the event contributions are also presented which confirmed
the results obtained in sections 4.1 to 4.2. The results in Table 5.8 also indicate that the
interface events can result in the lowest collapse capacity predictions for the bridge
studied in this research.

Table 5.8. IDA results using 78 fixed records for each event type for a)T=0.7 sec and
b)T=1.3 sec
Records type

Unmodified Modified Modified


median (g) median (g)
CMR

Records type

Unmodified Modified Modified


median (g) median (g) CMR

78 Crustal records

1.37

1.63

3.20

78 Crustal records

0.75

0.93

3.24

78 Interface records

1.35

1.43

2.80

78 Interface records

0.56

0.60

2.09

78 Inslab records

1.69

2.78

5.46

78 Inslab records

0.77

1.48

5.16

Combined results

1.44

1.84

3.61

Combined results

0.71

0.88

3.07

a)

5.7

b)

Summary and conclusions

The seismic behavior of a 4-span bridge located in Vancouver was investigated using
incremental dynamic analysis (IDA). First, the response of the bridge was evaluated by
considering crustal events only. The seismic evaluation was then extended to consider all
three earthquake types (i.e., shallow crustal, interface Cascadia subduction and deep
inslab subduction events) that contribute significantly to overall seismic hazard in southwestern British Columbia. A large pool of ground motion records was used for this
purpose, compiled by combining the PEER-NGA database and the K-NET/KiK-NET
database. The conclusions pertaining to the cases studied are summarized below:

(1) The median values of collapse capacities obtained in this study by considering
crustal events alone were not very different from those obtained by considering the other
event types. This was due to the fact that the constructed CMS-All was close to the CMSCrustal and thus combining the effects of different seismic event types led to similar
results as considering crustal events alone. Making general conclusions regarding this
180

matter needs more investigation, as this may be specific to the site condition and the
period ranges considered here. However the uncertainty due to record to record
variability was higher when different event types were considered. Thus accounting for
different event types is important and should be implemented in seismic performance
evaluation of bridges.

(2) The use of the UHS-based method resulted in very conservative predictions of
collapse capacities which were significantly lower than those predicted using the epsilonbased method.

(3) The use of the CMS- or UHS-based methods generally underestimated the recordto-record variability, while the use of the epsilon-based record selection method resulted
in larger predictions for record-to-record variability.

(4) It was found that the use of interface records in IDA resulted in the lowest
capacity predictions for the bridge studied in this research. This can be attributed to the
longer duration and different frequency content of the interface events compared to
crustal events.

(5) It was demonstrated that the IDA results can be sensitive to the choice of the
record selection method. The epsilon-based method in all of the cases resulted in the
highest prediction for the capacity of the structure. However more research is required to
verify the reliability of this method for seismic evaluations, as it was found that the
results can be sensitive to the number of records used. Furthermore, different sets of
records with comparable mean epsilon values could result in somewhat different collapse
capacity predictions. Increasing the number of records used in the IDA did not influence
the results obtained using the CMS-based record selection methods. The use of different
GMPEs for record selection did not significantly affect the results in IDA. The influence
of the record selection methods was greater on the collapse capacity prediction than on
the other damage states such as yielding and spalling.

181

(6) The CMS and the mean epsilon values were predicted using the matching method
in seismic deaggregation. The use of the exceeding method for seismic deaggregation
tended to overestimate the mean epsilon values and thus the collapse capacity of the
structure.

Acknowledgements

The financial support provided by the Natural Sciences and Engineering Research
Council of Canada for the Canadian Seismic Research Network is gratefully
acknowledged. Ground motion data for Japanese earthquakes were obtained from the KNET and KiK-NET databases at www.k-net.bosai.go.jp and www.kik.bosai.go.jp and
other crustal ground motion data obtained from the PEER-NGA database available at
http://peer.berkeley.edu. The authors are grateful for the constructive and insightful
comments provided by the reviewers.

References

Please see the last section.

182

6 Seismic Response of Bridges Subjected to Different


Earthquake Types Using IDA
Payam Tehrani1 and Denis Mitchell1

Department of Civil Engineering & Applied Mech., McGill University, Montreal H3A 2K6,

Canada

6.1 Preface
In Chapter 5 the IDA method was applied for the seismic evaluation of a bridge structure.
The results indicated that including different earthquake types in the seismic evaluations
results in a larger record to record variability, while the median collapse capacity
predictions are similar to the case where only crustal events are considered. To further
investigate the results obtained in Chapter 5, in Chapter 6 the IDA method is used to
predict the collapse capacity of a number of bridges with different configurations and
different period ranges. The bridges were also designed using different force modification
factors, R.

In Chapter 5, it was also shown that when different earthquake types are considered, the
IDA needs to be performed using a larger number of records, due to increased record-torecord variability in the results. For this purpose a large set of records were selected to
perform the IDA. Three sets of records were selected for three different earthquake types.
Each set includes 78 horizontal components (i.e., 39 pairs of ground motions). The details
of the three sets of ground motion for different earthquake types are given in Appendix
G.

In Chapter 5 it was demonstrated that a simplified method can be used to modify the IDA
results for the effects of epsilon and spectral shapes. The results from the simplified
183

method were in good agreement with those obtained using the direct use of the epsilonbased method. Therefore in Chapter 6 this method is used to modify the IDA results
obtained using the large fixed record sets for the spectral shape effects.

Since in Chapter 6 the IDA method is applied to a number of bridges using a large
number of records (i.e., 234 records for each bridge), the computation time of the
analyses could be extremely long. Further, the durations of the subduction earthquake
records are much longer than the crustal records, which result in the computation time of
the IDA being much longer, when such earthquake types are considered in the analyses.
To remove this problem, a fast IDA algorithm was developed, as discussed in Chapter 2
and will be discussed in more details in Chapter 6. In the fast IDA algorithm only the
collapse capacity of the bridges will be predicted, since only the median collapse capacity
and the variability in the results are needed for the seismic evaluations, as discussed in
Chapter 2. It must be noted that even using the fast IDA algorithm the analyses were
computationally intensive (see Appendix B for the details of the computer program).
More details regarding the program developed to extract the seismic deaggregation
results are available in Appendix C.

More information regarding different earthquake types are given in Appendix F.


More details of the partial results obtained are also presented in Appendix I.
A main part of Chapter 6 was included into a manuscript: Tehrani, P. and Mitchell, D.
Seismic Response of Bridges Subjected to Different Earthquake Types using IDA,
Journal of earthquake Engineering, Manuscript UEQE-2012-1345, submitted in January
2012.

Based on the Taylor and Francis publication policy: the authors retain the right to
include an article in a thesis or dissertation that is not to be published commercially
(http://journalauthors.tandf.co.uk/preparation/copyright.asp).

184

6.2 Abstract
Incremental Dynamic Analysis (IDA) was used to evaluate the seismic response of
straight, continuous 4-span bridges with different sub-structure configurations. Three
different record sets were chosen to represent three different earthquake types which can
occur for a site such as Vancouver (i.e., crustal, subduction interface, and subduction
inslab earthquakes). Seventy eight records were considered in each set (i.e., a total of 234
records) and the capacities of the bridges were evaluated using a fast IDA algorithm. A
simplified method to account for the effects of spectral shapes was used. Different
subsets of the records with specific characteristics were also used in the IDA. The bridges
were designed and evaluated for two different design force modification factors and
bridges with different degrees of irregularity were studied. Comparisons of the IDA
results obtained indicated that in most of the cases the interface record sets resulted in
lower median collapse capacities and hence were the most critical of the ground motions
studied.

Keywords Incremental Dynamic Analysis (IDA); Subduction Earthquakes; Crustal


Earthquakes; Spectral Shape; Epsilon; Bridges; Regularity

6.3 Introduction
Incremental dynamic analysis (IDA) (Vamvatsikos and Cornell, 2002) is a useful tool for
seismic performance assessment of structures (e.g., ATC-63 provisions (ATC-63, 2008)).
The IDA results which relate statistics of a damage measure (DM) (i.e., inelastic seismic
demand) with an intensity measure (IM) (usually Sa(T1)) can be used in probabilistic
seismic performance assessment and fragility analysis of structures.

For a realistic seismic performance assessment using IDA, appropriate records should be
chosen based on the properties and location of the structure. This could result in the
selection of different earthquake records for different structures. In the seismic
performance assessment of a number of structures the use of different records for each
185

structure can be very time-consuming and hence for practical reasons simplified methods
can be used.

The spectral shape of the records used for seismic evaluations using IDA can
significantly affect the seismic response predictions (e.g., (Baker and Cornell, 2006a).
The ATC-63 provisions propose a simplified method in which a fixed set of records are
used in the IDA without considering the spectral shape in the record selection. The IDA
results obtained using this fixed set of records are then modified for the effects of spectral
shapes based on the fundamental period and ductility capacity of each structure. This can
reduce the complexity of the seismic performance evaluations. Research has shown that
the results obtained from the simplified method are in good agreement with those
obtained using more precise methods that directly consider the spectral shape in the
record selection (e.g., (Haselton et al., 2011) and (Tehrani, 2012)). For the seismic
assessment of a number of bridges with different configurations, having different degrees
of irregularity, this simplified method will be used.

Usually only the records from the crustal earthquakes have been used in seismic
performance assessments (e.g., ATC-63 provisions (2008)), while for some sites such as
Vancouver or Seattle subduction earthquakes with very different characteristics can
occur. The effects of considering different earthquake types (i.e., crustal, subduction
interface, and inslab events) on the seismic performance evaluation of bridges for a site
such as south-western British Columbia is the main objective of this study. The
transverse responses of the bridges were chosen in this study, rather than 3D responses, to
better demonstrate the influence of different earthquake types on the seismic response
and to limit the variables.

6.4 Epsilon-based method and spectral shape issue


Rare ground motions (large earthquakes) typically have a peaked spectral shape that is
much different than a standard uniform hazard spectral shape and accounting for this
shape has been shown to increase the predicted collapse capacity significantly (increase
186

of up to 70% (Haselton et al, 2011; Baker and Cornell, 2006; Tehrani, 2012). Thus
neglecting the spectral shape of the records selected for collapse assessment of a structure
can result in conservative and unrealistic predictions.

Epsilon, , is defined as the number of logarithmic standard deviations of a target ground


motion from a median ground motion prediction equation (GMPE) for a given
magnitude, M, and distance, R. The most direct approach in accounting for spectral shape
in structural analysis is to select ground motions with (T1) values (i.e., the epsilon values
at the fundamental period of the structure) similar to the epsilon values obtained from
seismic hazard deaggregation analysis for the site and hazard level of interest. It has been
found that the (T1) values are in fact a measure of the spectral shape of the records
(Baker and Cornell, 2006). This approach is known as the epsilon-based record selection
method. Because the direct use of this approach is often time-consuming and complicated
in practice, a simplified method has been developed to modify the collapse capacity
predictions for the effects of epsilon and spectral shapes (Zareian, 2006; Haselton et al.,
2011). For this purpose, a fixed set of records can be used to perform the incremental
dynamic analyses. The predicted capacities obtained using the IDA will then be modified
for the effects of spectral shapes using a spectral shape factor (SSF). To compute the SSF
a regression analysis needs to be carried out to derive a relationship between the natural
logarithm of the collapse capacities versus the epsilon (T1) values for each record. The
result of such a regression analysis can be presented in the form of ln[SC(T1)] = 0 + 1.
Where the SC is the collapse capacity, 0 is the average collapse capacity when  = 0, and
1 indicates how sensitive the collapse capacity (SC(T1)) is to changes in the  value. The
SSF is then calculated using Eq. [6.1]

[6.1] SSF

exp[ E1 (H 0 (T )  H (T )records )]

where H (T ) records is the mean epsilon of the records and H 0 (T ) is the mean epsilon value
from the seismic deaggregation. For the prediction of the epsilon values of the records,
the BA08 (Boore and Atkinson, 2008) ground motion prediction equation (GMPE) was

187

used for the crustal earthquakes and the Z06 GMPE (Zhao et al., 2006) was used for the
subduction zone earthquakes (i.e., interface and inslab).

The most important issue in the simplified method which will affect the results is the
prediction of the 1 factor. In the ATC-63 provisions simplified equations are given to
estimate the 1 factor based on the ductility capacities of building structures. To develop
these simplified equations to estimate the 1 factor, a large record set including 78 crustal
records was selected from the PEER-NGA database (Haselton et al., 2007) for the
analyses.

The equations given in the ATC-63 provisions are specific to building structures rather
than bridges and in addition, these equations may not be valid for other earthquake types
(i.e., interface and inslab events). Therefore, in this study the 1 factors are calculated
directly for each bridge structure using a regression analysis. The evaluation of the 1
factors have been carried out using the whole set of data and also using some subsets of
data with specific characteristics such as positive epsilon values, lower scale factors,
removal of outliers and as well as a combination of these factors. This enables an
evaluation of the sensitivity of the 1 factors and the collapse capacity estimates as a
function of the characteristics of the records in the set.

The seismic evaluations based on the simplified method often involve using a fixed set of
records including 22 crustal records (i.e., 44 record components) (ATC63, 2008). Since
one of the important steps is the prediction of the 1 factors for each structure, the use of
44 record components may not be adequate and more records is recommended to be used
for a more precise prediction of this factor. Further some studies indicate that the use of
the epsilon-based method may be sensitive to the number of records used in the IDA
(Tehrani, 2012). Thus to address these issues 78 record components (39 pairs) are used
for each of the three event types to evaluate the seismic response of the bridges (i.e., a
total of 3*78=234 records).

188

6.5 Fast-IDA analysis


In a typical IDA, the analyses start with very low spectral acceleration levels and then the
records are scaled up until they cause dynamic instability of the structure which indicates
structural collapse. The resulting wide variation of the scale factors from small values to
large values allows the determination of different performance levels (e.g., from
serviceability states to structural collapse) and enables the development of the IDA
curves.

For the seismic evaluations in this study, the prediction of the collapse capacities of the
bridges is the primary objective and thus the full IDA curves are not required. Only the
median value of spectral acceleration of the records at the fundamental period of the
bridge that causes dynamic instability (collapse capacity of the structure) needs to be
determined. This will significantly reduce the total number of required inelastic dynamic
analyses and the computation time. It must be noted that the duration of the interface and
inslab records are significantly longer than those of the crustal events and thus
computation time for IDA analyses using subduction records will be significantly longer
(about 4 to 6 times) than that using crustal records. In this research the variability was
computed based on the predicted collapse capacities for all of the records in each set.
The probability of collapse at the maximum considered earthquake (MCE) level, which
usually corresponds to 2% probability of exceedance in 50 years, should be reasonably
low. To estimate the probability of collapse at the MCE level based on the computed
median collapse capacity a simplified assumption is often made that the cumulative
probability function follows a lognormal distribution.

The ATC-63 provisions provide a simplified judgmental method to estimate the total
uncertainty in the prediction of the collapse capacity. For the bridges considered in this
study, the total uncertainty, expressed by TOT of 0.6 is deemed appropriate (i.e., the
modelling quality and quality of test data used for the nonlinear hysteretic models were
considered as good, and the quality of code conforming ductile design requirements
was considered as superior).

189

The collapse margin ratio, CMR, which is the ratio of the median 5%-damped spectral
acceleration of the collapse level ground motions to the 5%-damped spectral acceleration
of the MCE ground motions, can be then computed at the fundamental period of the
structure (ATC-63, 2008). The probability of collapse at the MCE level can be predicted
based on the lognormal distribution model.

It should be noted that the computed median collapse capacities and CMR values are
determined based on the seismic excitations applied only in the transverse direction of the
bridges and thus the results do not include the effects of longitudinal seismic response.
This assumption is deemed appropriate for the sake of this research, since the relative
influence of considering different earthquake types on the seismic response was the main
focus of this study. This simplification was made for the following reasons: to exclude
further variables and uncertainty associated with bi-directional responses; problems with
scaling of bi-directional excitations (Priestley et al., 2007; Baker and Cornell, 2006c); the
conservative biases introduced due to the application and scaling of pairs of ground
motion records in IDA (ATC-63, 2008); and the difficulty in choosing an appropriate
period and scaling procedure in the epsilon-based method due to the different periods in
the longitudinal and transverse directions. More research is required on these subjects.

Although a general conclusion cannot be made, studies of the longitudinal responses of


similar bridges, including the influence of the abutments in the seismic response,
demonstrated that the transverse response was controlling (Tehrani, 2012). The current
state-of-the-art is to model the ductility effects independently and Priestley et al. (2007)
concluded that the seismic responses in the orthogonal directions for straight bridges and
symmetrical buildings are essentially independent. Nevertheless it is recommended that
the effects of bi-directional excitations be considered when the seismic performance
evaluations of the structures are carried out. More research is required on this subject.

In this research different event types and different sets of records are used and the recordto-record variability and the collapse capacity must be computed for each record set. The
collapse capacities obtained for each record should be known to compute the 1 factors

190

and the spectral shape factors. For this purpose a time efficient algorithm was developed
to compute only the spectral accelerations (at the fundamental period of the structure) that
cause structural collapse for each record. For this purpose an initial estimate of collapse
capacity was computed based on the expected acceptable collapse margin ratio of the
structure assuming 10% probability of collapse at the maximum considered earthquake
(MCE) level and that the cumulative density function follows a lognormal distribution
rule (see Eq. [6.2]).

[6.2] Sa1(T1)collapse= Sa(T1)MCE*CMRacceptable


[6.3]

CMRacceptable =1/ exp ( -1(pcollapse)*TOT)

Where, Sa1(T1)collapse is the initial estimate of the collapse capacity, Sa(T1)MCE is the
spectral acceleration at the fundamental period of the structure from the maximum
considered earthquake spectrum, CMRacceptable is the acceptable collapse margin ratio, -1
is the inverse cumulative normal distribution function, pcollapse is the acceptable
probability of collapse at the MCE level (usually taken as 10%), and TOT is the total
uncertainty in predicting the collapse capacity of the structure (i.e., including
uncertainties due to record-to-record variability, modelling, test data , and design
requirements). Eq. [6.2] is only used to compute an initial estimate of the collapse
capacity which then will be modified in several subsequent steps to compute the collapse
capacity of the structure. In this study the precision in the prediction of the collapse
capacities was considered as 0.02 g.

6.6 Record selection for IDA


Three fixed sets of records including 39 records (78 components) for each earthquake
type (i.e., 234 records) were chosen to be used in the IDA. This large number of records
was chosen to compute the 1 values for each event type more precisely, to predict the
variability in the seismic response, and to improve the capacity predictions. The records
chosen in the sets are among the strongest natural records available and are deemed

191

appropriate for the seismic performance evaluations. The criteria used for record
selections are briefly discussed below.

a) Crustal record set:


For crustal earthquakes, the basic far-field records used by (Haselton et al., 2007) were
used in this study. This set includes 39 crustal records from the PEER-NGA database.
Some minimum limits on the magnitude of the events, peak ground acceleration and peak
ground velocity were imposed in the record selection to be representative of strong
ground motion. A limit on the maximum number of records from a single seismic event
was imposed to make sure that the predictions are not biased, however a sufficient
number of records must also be selected. The criteria imposed for the record selection is
presented in Table 6.1.

b) Interface record set:


Since the number of available records due to subduction events is relatively small, the
selection criteria used for the crustal events need to be somewhat loosened so that enough
number of records can be selected. The selected records represent the strongest records
available from each event in the database and an attempt was made to include at least a
few records from each earthquake that roughly meets the criteria. In addition, 10 records
from the destructive 2011 Tohoku earthquake were included in the set to take advantage
of the latest data available for interface earthquakes.

c) Inslab record set:


Only a limited number of inslab earthquake records are available in the database, and
thus the selection criteria needed to be loosened even more so that sufficient number of
records can be selected. The selected records represent the strongest records available
from the inslab events in the database. In addition to the inslab events from Japan, 5
records from the 2001 Nisqually earthquake in North America were included in the set. It
must be noted that the shear wave velocity data, VS30, for the Nisqually earthquake was
not originally available by the COSMOS and USGS databases. In this study the data
192

concerning the shear wave velocities were obtained from Caki and Walsh (2011) and
Wong et al. (2011). The VS30 values are necessary to compute the Sa values using the
corresponding GMPEs, compute the epsilon values, and to determine the soil type.
Further details concerning the records used for the analyses including the list of records is
available by Tehrani (2012).

Table 6.1. Criteria used for record selection for different earthquake types

Earthquake
Magnitude
type

Distance
Focal
(km)
depth (km)

VS30
(m/sec)

PGA PGV
Maximim number of
(g) (cm/sec) records from each event

Databases

Crustal

6.5

10

180

0.2

15

PEER-NGA

Interface

6.7

<50

180

0.17

13

10

K-NET and KiKNET

Inslab

6.0

50

180

0.11

11

12

K-NET and KiKNET and


COSMOS

6.7 Bridge properties


The seismic behaviour of different bridge configurations were studied using Incremental
Dynamic Analysis (IDA) considering three different earthquake types by using the 78
records for each event type. Continuous 4-span straight bridges were studied and the
seismic evaluations were carried out in the transverse direction of the bridges to assess
the importance of different earthquake types. It is known that the seismic response of the
bridges in the transverse direction is governed by a multi-degree of freedom system
which is more complex than the response in the longitudinal direction. For straight
bridges the responses in the orthogonal directions are likely to be essentially independent
(Preistley et al., 2007).

As shown in Fig. 6.1, five different bridge configurations were studied with different
arrangements of column heights. For each configuration 3 different column diameters are
considered (i.e., a total of 5 x 3=15 bridges). The 2006 Canadian Highway Bridge Design

193

Code (CHBDC) (CSA, 2006) classifies the importance of bridges in accordance with
their performance requirements. Emergency-route bridges must be open to emergency
vehicles immediately after the design earthquake. The CHBDC uses a response
modification factor, R together with an importance factor, I. For single ductile reinforced
concrete columns R is 3.0. For emergency-route bridges I is 1.5 and for other bridges I is
1.0. Based on the results obtained from the inelastic time history analyses of more than
600 bridge configurations designed based on the CHBDC, the use of a response
modification factor, R, of 3 along with an importance factor, I, of 1.5 was found to be
conservative for current seismic design practice (Tehrani, 2012). In order to reduce the
conservatism by accounting for a larger R factor for the ductile bridge columns, the
bridges in this study were also designed with R=5 and two cases were considered, one
with I=1.5 (R/I=3.33) for emergency-route bridges and another case where I is taken as
1.0 (R/I = 5).

The design spectrum defined in the 2010 National Building Code of Canada (NBCC)
(NRCC, 2010) rather than that of the CHBDC were used for the design of the bridges. It
must be noted that the design spectra defined in NBCC correspond to 2% probability of
exceedance in 50 years, while those defined in CHBDC correspond to 10% probability of
exceedance in 50 years. It is expected that the design spectra defined in the NBCC be
adopted for the next edition of the CHBDC. The bridges were designed for Vancouver
assuming a site class C.

A computer program was developed to automatically generate the designs for the
columns and the input files for the RUAUMOKO software (Carr, 2009) for different
bridge configurations. The columns of the bridges were designed using the load
combinations defined in the CHBDC based on the multi-mode elastic response analyses
and the design spectra from the NBCC. For this purpose and also for the seismic
performance evaluations a moment curvature analysis was carried out for each column
and also the moment axial force interaction curves were determined. The columns were
designed using capacity design principles and detailed for ductile response (CSA 2006).

194

In order to predict the ductility capacities, the confinement effects in the concrete core
were considered using the confinement model by Mander et al. (1988).
The configurations and properties of the bridge structures studied are summarised in
Tables 6.2 to 6.4. Where H1, H2 and H3 are the heights of columns, D is the diameter of
the columns, and R is the response modification factor. For the classification of regularity
the provisions of the CHBDC (CSA 2006) were used except for configuration 4 which is
actually regular for the transverse response. It is noted that the minimum amount of
longitudinal reinforcement in a column is 0.8% which controls the amount of flexural
reinforcement in columns with low seismic demands. The spiral confinement
reinforcement ratio of 1.2% controlled the amount of transverse reinforcement in the
plastic hinge regions of the columns. The spiral reinforcement consisted of 20 mm
diameter bars spaced at 70 and 50 mm for column diameters of 1.5, 2.0 m, respectively
and 25 mm diameter bars spaced at 70 mm for column diameter of 2.5 m.

Table 6.2. Summary of bridges with different configurations studied (a total of 30


bridges)
Bridge
configuration

Regularity

H1 (m)

H2 (m)

H3 (m)

D (m)

R/I

Regular - stiff
columns

1.5, 2.0 and


2.5

3.3 and 5

Regular - more
flexible columns

14

14

14

1.5, 2.0 and


2.5

3.3 and 5

Irregular bridge
("ramp")

14

21

1.5, 2.0 and


2.5

3.3 and 5

Regular bridge
with flexible central
column

14

1.5, 2.0 and


2.5

3.3 and 5

Highly irregular
bridge

14

21

1.5, 2.0 and


2.5

3.3 and 5

195

Table 6.3. Periods (sec) and (%M at the fundamental period) for different bridge
configurations studied in the case of a) R/I=3.33 and b) R/I=5
a)

b)

Conf. D= 1.5 m D= 2 m D= 2.5 m

Conf. D= 1.5 m D= 2 m D= 2.5 m

1
2
3
4
5

1.09
(86%)
2.17
(86%)
1.65
(84%)
1.33
(84%)
1.28
(88%)

0.75
(86%)
1.75
(86%)
1.32
(79%)
0.96
(82%)
0.90
(90%)

0.57
(86%)
1.36
(87%)
1.13
(74%)
0.75
(78%)
0.72
(83%)

1
2
3
4
5

1.17
(86%)
2.17
(86%)
1.75
(85%)
1.45
(85%)
1.41
(88%)

0.79
(86%)
1.75
(86%)
1.38
(81%)
1.03
(83%)
1.01
(90%)

0.57
(86%)
1.36
(87%)
1.16
(76%)
0.77
(79%)
0.79
(89%)

Table 6.4. Percentage of longitudinal reinforcement in the bridge columns C1,C2 and C3
for different bridge configurations
R/I= 3.33

R/I= 5

Conf.

D= 1.5 m
(C1 , C2 , C3 )

D= 2.0 m
(C1 , C2, C3)

D= 2.5 m
(C1 , C2 , C3)

D= 1.5 m
(C1 , C2 , C3 )

D= 2.0 m
(C1 , C2 , C3 )

D= 2.5 m
(C1 , C2, C3)

1.3%,1.5%,1.3%

1.0%,1.17%,1.0%

0.8%,0.8%,0.8%

0.8%,0.8%,0.8%

0.8%,0.8%,0.8%

0.8%,0.8%,0.8%

0.8%,0.8%,0.8%

0.8%,0.8%,0.8%

0.8%,0.8%,0.8%

0.8%,0.8%,0.8%

0.8%,0.8%,0.8%

0.8%,0.8%,0.8%

2.71%,0.8%,0.8%

2.04%,0.8%,0.8%

1.6%,0.8%,0.8%

1.33%,0.8%,0.8%

1.07%,0.8%,0.8%

0.88%,0.8%,0.8%

1.93%,0.8%,1.93%

1.3%,0.8%,1.3%

1.0%,0.8%,1.0%

0.8%,0.8%,0.8%

0.8%,0.8%,0.8%

0.8%,0.8%,0.8%

0.8%,2.92%,0.8%

0.8%,2.2%,0.8%

0.8%,1.71%,0.8%

0.8%,1.52%,0.8%

0.8%,1.22%,0.8%

0.8%,0.98%,0.8%

196

D
mm

mm

Fig. 6.1. Bridge configurations studied

197

6.8 Modelling of the bridges for IDA


In using IDA, the post-peak (i.e., cap point) response of structural elements should be
included in modelling. The most important factors in structural modelling for IDA are the
plastic rotation capacity, capp, and the post-capping rotation capacity, pc (ATC-63,
2008). These parameters are used to define a component backbone curve, as shown in
Fig. 6.2.

Moment (M)

My

Mc

Ks

pc
Ke
Kc

y

cap
Chord rotation ()

Fig. 6.2. Backbone curve parameters (adapted from ATC-63(2008))

A study by Berry and Eberhard (2007) provides some empirical equations to estimate the
engineering demand parameters. These parameters include drift ratio, plastic rotation and
strain in the longitudinal bars for circular bridge columns based on the properties of the
columns including longitudinal and transverse steel ratio, axial load ratio, and geometry.
In this research, the ultimate tensile strain in the steel bars corresponding to the bar
buckling damage state from the empirical equations is used to compute the ultimate
curvature of the columns and the corresponding drift and curvature ductility is defined as
the point at which strength degradation begins (i.e., cap in Fig. 6.2). Mackie and
Stojadinovic (2007) provide an equation for estimating the drift beyond the peak load
when the column reaches zero strength which was used in this research to define the postcapping stiffness, Kc. The modified Takeda hysteresis model (Otani, 1981) was used in
this study to model the hysteretic behaviour of the RC columns using the Ruaumoko

198

software (Carr, 2009). More details about the modelling parameters are available
elsewhere (Tehrani, 2012).

The bridges under study were designed and detailed to meet the code requirements for
ductile response, including capacity design concepts and adequate support lengths at the
abutments. The ductile columns contained code-compliant spiral reinforcement to confine
the concrete, avoid shear failure and to control buckling of the vertical reinforcing bars.
For this continuous bridge, with all other failure modes avoided, the flexural response
governs the response of the bridge and sidesway collapse is the governing collapse
mechanism. The collapse prediction in this study is based on dynamic instability of the
structures (Vamvatsikos and Cornell, 2002; ATC63, 2008).

6.9 Seismic hazard analysis


An updated seismic hazard model for western Canada, developed by Goda et al. (2010),
was used to take advantage of new seismic data and seismological models to improve
several aspects of the current Geological Survey of Canada model. Probabilistic seismic
hazard analysis (PSHA ) for Vancouver was carried out using the updated seismic hazard
model based on Monte Carlo simulation by Goda et al. (2010). The assessment was based
on the simulated seismic activities for 5 million years, and the considered annual nonexceedance probabilities range from 0.996 to 0.9999 (i.e., 10% to 0.5% probability of
exceedance in 50 years). The deaggregation analysis is based on an approximately equal
criterion (i.e., matching method) (Hong and Goda, 2006), where seismic events reaching
a seismic intensity level between 90% and 110% of the target Sa(Tn) value are used to
produce deaggregation results. The deaggregation results for the mean epsilon values and
the contributions of different event types are shown in Table 6.5. The results are based on
the BA08 and Z06 GMPE which were used to compute the epsilon values in this study
for the crustal and subduction earthquakes, respectively. The mean epsilon values from
the seismic deaggregation are used in the epsilon-based record selection method and also
in the prediction of the spectral shape factors (SSF, see Eq. [6.1]) based on the
fundamental period of each bridge configuration. The percentage of contributions of each
199

event type in the overall seismic hazard is used to determine the required number of
records from each event type, when the results from different events are combined to
estimate the overall collapse capacity due to all earthquake types.

Table 6.5. The deaggregation results for mean epsilon and percentage of contribution of
different event types at 0.5% probability of exceedance in 50 years (using BA08 and Z06
GMPE).
Percentage of contribution
Crustal Interface
PGA
31%
4%
T=0.2 sec 26%
7%
T=0.5 sec 39%
27%
T=1 sec
36%
34%
T=2 sec
29%
48%
T=3 sec
36%
54%

Inslab
65%
67%
34%
30%
23%
10%

Mean epsilon values


Crustal Interface Inslab
1.40
2.31
1.76
1.28
2.23
1.87
1.49
1.79
1.82
1.66
1.38
1.85
1.71
1.24
2.67
1.72
1.15
2.17

6.10 Results obtained using different record sets and subsets


A fast IDA algorithm was used to predict the collapse capacity of the bridges. Fig. 6.3(a)
shows an example of a full IDA curve carried out for one of the cases studied and the
corresponding unmodified spectral accelerations obtained at different damage states.
Sa(T1) is used as the intensity measure (IM) and the maximum drift ratio of columns is
used as the damage measure (DM). The heavy lines in Fig. 3(a) show the IDA percentiles
(median, 16% and 84%).The fast-IDA results for this case are also presented in Fig.
6.3(b) (see last row in table). The results obtained using the fast-IDA algorithm are the
same as those obtained using a full IDA analysis at the collapse state. The minor
differences are due to the precision of 0.02g accepted in the prediction of collapse
capacity.

200

2.5
Individual curves
84% percentile

Sa (T=1.1 sec) (g) - (IM)

50% percentile
16% percentile

1.5

0.5

0
0

0.02

0.04

0.06

0.08

0.1

0.12

Maximum column drift - (DM)

a)

IDA result percentiles


(Sa (T1))

78 Crustal records
Damage states

b)

Median 16%

84%

St Dev

Yielding

0.10

0.09

0.12

0.11

Cover Spalling

0.36

0.29

0.45

0.22

Bar buckling

1.03

0.68

1.37

0.37

Collapse

1.37

0.83

1.87

0.41

Collapse (Fast-IDA)

1.36

0.84

1.85

0.40

Fig. 6.3. IDA results: a) full IDA curves for the bridge with configuration 1, D=1.5 m and
R/I=3.3 using 78 crustal records (unmodified results) b) statistics of the IDA results for
different damage states

Three different subsets of records were considered to include the three different
earthquake types. Since a large number of records are available in each set (i.e., 78
records), it is possible to consider some different subsets of records. In this study 8
different subsets of records were considered. In each subset only the records with specific
properties were considered. This involved selecting records with positive epsilon values,
considering records requiring lower scale factors, removal of outliers and a combination
of these parameters. To compute the modified collapse capacities, record-to-record
variability, spectral shape factors, etc., for each subset only the records in that subset
were considered. The details of different subsets considered are presented in Table 6.6.

201

Table 6.6. Different record subsets considered to compute the predicted median collapse
capacity
Subset
No.

Criteria

All
Records with
Records
Records with lower
Records with lower
records
positive
All records
Records with lower
All
with
scale factors and
scale factors and
with
epsilons and
with lower
scale factors and
records positive
positive epsilon
positive epsilon values
outliers
outliers
scale factors
outliers removed
epsilon(T1)
values
and outliers removed
removed
removed

A summary of the results obtained using different record sets and subsets are shown in
Tables 6.7 to 6.11. The results are given in terms of the modified median collapse
capacities and the ratio of the obtained collapse margin ratio (CMR) to the acceptable
collapse margin ratio (CMRacceptable) assuming a 10% probability of collapse at the
maximum considered earthquake level (i.e., 2% in 50 years). Direct comparison of the
modified collapse capacities obtained for different configurations is not appropriate, since
the fundamental periods of the structures in each configuration are different and
consequently the seismic demands will be different for each configuration. To better
compare the results from different configurations, the CMR/ CMRacceptable ratios are
compared. The CMR includes the effects of different seismic demands and CMRacceptable
includes the effects of different variability in the predictions of the collapse capacities
using different sets and subsets.

In addition to these eight subsets, another subset was considered in which only the
records with similar epsilon values to those obtained from the seismic deaggregations
were considered (i.e., epsilon-based method). The results obtained using this subset is
also shown in the tables which can be used to assess the predictions obtained using the
spectral shape factors. For the case of some subsets for the inslab events there was an
insufficient number of records, especially when the records with required lower scale
factors were considered.

202

Table 6.7. Comparison of the median collapse capacity (g) and (CMR/ CMRacceptable)
ratios using different subsets for configuration 1with R/I = 3.33 and a) D = 1.5 m and b)
D = 2.5 m
a)

b)

Modified median &


(CMR/CMRacceptable)

Crustal
records

Subset 1

1.60
(2.34)

1.06
(1.14)

4.00
(3.59)

Subset 2

1.62
(2.38)

1.01
(1.05)

Subset 3

1.61
(2.41)

Subset 4
Subset 5
Subset 6
Subset 7
Subset 8
Epsilon-Based
method

Modified median &


(CMR/CMRacceptable)

Crustal
records

1.61
(1.57)

Subset 1

2.75
(2.11)

2.93
(1.65)

7.11
(3.64)

3.45
(1.81)

3.35
(3.43)

1.63
(1.64)

Subset 2

2.75
(2.12)

2.82
(1.55)

6.02
(3.40)

3.46
(1.96)

1.05
(1.16)

4.44
(4.27)

1.64
(1.59)

Subset 3

2.77
(2.17)

2.85
(1.73)

7.13
(3.90)

3.40
(1.89)

1.65
(2.48)

0.90
(0.97)

3.44
(3.88)

1.67
(1.74)

Subset 4

2.78
(2.19)

2.75
(1.63)

6.11
(3.64)

3.39
(2.06)

1.64
(2.40)

0.59
(0.75)

2.45
(2.85)

1.55
(1.74)

Subset 5

2.71
(2.09)

2.22
(1.48)

4.49
(2.69)

3.17
(2.13)

1.62
(2.38)
1.66
(2.48)
1.65
(2.48)

0.64
(0.81)
0.55
(0.74)
0.66
(0.90)

2.54
(3.17)
2.76
(3.24)
2.54
(3.17)

1.55
(1.77)
1.56
(1.70)
1.58
(1.81)

2.74
(2.14)
2.73
(2.15)
2.76
(2.19)

2.10
(1.39)
2.04
(1.40)
2.04
(1.39)

4.67
(2.94)
4.49
(2.69)
4.67
(2.94)

3.17
(2.17)
3.26
(2.21)
3.18
(2.21)

1.49
(2.34)

0.70
(0.73)

2.73
(3.38)

1.42
(1.58)

2.99
(2.36)

2.38
(1.24)

5.68
(3.74)

2.70
(1.77)

Interface Inslab
All
records records events

Subset 6
Subset 7
Subset 8
Epsilon-Based
method

Interface Inslab
All
records records events

Table 6.8. Comparison of the median collapse capacity (g) and (CMR/ CMRacceptable)
ratios using different subsets and methods for configuration 2 (R/I = 3.3) and a) D = 1.5
m and b) D = 2.5 m
a)

b)

Modified median &


(CMR/CMRacceptable)

Crustal
records

Subset 1

1.37
(4.31)

1.01
(1.98)

3.74
(7.51)

Subset 2

1.43
(4.49)

0.96
(1.76)

Subset 3

1.35
(4.32)

Subset 4
Subset 5
Subset 6
Subset 7
Subset 8
Epsilon-Based method

Modified median &


(CMR/CMRacceptable)

Crustal
records

1.30
(2.39)

Subset 1

2.29
(3.69)

1.56
(1.97)

6.37
(6.59)

2.47
(2.53)

2.99
(5.99)

1.36
(2.59)

Subset 2

2.56
(4.15)

1.34
(1.57)

7.75
(8.32)

2.64
(2.69)

0.89
(2.11)

3.52
(7.69)

1.20
(2.26)

Subset 3

2.29
(3.75)

1.38
(1.87)

6.54
(7.52)

2.41
(2.55)

1.39
(4.46)

0.85
(2.06)

3.04
(6.58)

1.25
(2.50)

Subset 4

2.60
(4.27)

1.26
(1.62)

7.29
(8.93)

2.64
(2.84)

1.32
(4.25)

0.47
(1.15)

2.72
(5.74)

1.05
(1.79)

Subset 5

2.22
(3.63)

0.90
(1.29)

1.32
(4.23)
1.32
(4.27)
1.31
(4.24)

0.49
(1.26)
0.47
(1.21)
0.47
(1.25)

2.22
(4.65)
2.77
(5.98)
2.22
(4.65)

1.00
(1.83)
1.04
(1.81)
0.98
(1.81)

2.31
(3.83)
2.24
(3.70)
2.20
(3.71)

0.93
(1.35)
0.78
(1.23)
0.82
(1.38)

1.38
(4.43)

0.75
(1.77)

1.84
(3.61)

1.03
(2.49)

2.66
(4.76)

1.30
(1.66)

Interface Inslab
All
records records events

Subset 6
Subset 7
Subset 8
Epsilon-Based method

203

Interface Inslab
All
records records events

5.64
(6.68)

2.04
(2.51)

Table 6.9. Comparison of the median collapse capacity (g) and (CMR/ CMRacceptable)
ratios using different subsets and methods for configuration 3 (R/I = 3.3) and a) D = 1.5
m and b) D = 2.5 m
a)

b)

Modified median & Crustal


(CMR/CMRacceptable) records

Interface Inslab
All
records records events

Modified median &


(CMR/CMRacceptable)

Crustal
records

Interface Inslab
All
records records events

Subset 1

1.06
(2.21)

0.78
(1.21)

3.37
(4.36)

1.15
(1.49)

Subset 1

1.86
(2.80)

1.19
(1.36)

4.38
(4.14)

1.86
(1.86)

Subset 2

1.02
(2.21)

0.80
(1.19)

3.07
(4.72)

1.11
(1.46)

Subset 2

1.87
(2.85)

1.19
(1.31)

3.84
(3.98)

1.88
(1.95)

Subset 3

1.04
(2.20)

0.74
(1.28)

3.24
(4.60)

1.11
(1.49)

Subset 3

1.91
(2.94)

1.17
(1.39)

4.78
(4.63)

1.94
(1.91)

Subset 4

1.02
(2.20)

0.77
(1.31)

2.16
(3.82)

1.10
(1.73)

Subset 4

1.92
(3.01)

1.18
(1.33)

3.68
(4.04)

1.89
(2.02)

Subset 5

1.05
(2.19)

0.50
(0.88)

1.57
(2.53)

0.89
(1.36)

Subset 5

1.87
(2.87)

0.73
(0.95)

1.57
(1.92)

1.59
(2.09)

1.01
(2.18)
1.05
(2.19)
1.00
(2.16)

0.49
(0.91)
0.50
(0.89)
0.49
(0.91)

Subset 6

1.57
(2.53)

0.89
(1.36)

Subset 7

Subset 8

1.90
(2.95)
1.89
(2.93)
1.93
(3.03)

0.77
(0.99)
0.68
(0.91)
0.77
(1.07)

2.10
(2.65)
1.35
(1.73)
2.10
(2.65)

1.82
(2.26)
1.55
(2.06)
1.81
(2.25)

0.90
(2.02)

0.77
(1.15)

2.15
(2.90)

0.78
(1.33)

Epsilon-Based method

1.88
(3.00)

0.77
(0.85)

4.06
(4.66)

1.44
(1.71)

Subset 6
Subset 7
Subset 8
Epsilon-Based
method

Table 6.10. Comparison of the median collapse capacity (g) and (CMR/ CMRacceptable)
ratios using different subsets and methods for configuration 4 (R/I = 3.3) and a) D = 1.5
m and b) D = 2.5 m
a)

b)
Modified median &
(CMR/CMRacceptable)

Crustal
records

1.69
(1.78)

Subset 1

2.69
(2.41)

2.34
(1.62)

4.80
(3.53)

3.07
(2.27)

5.00
(5.57)

1.85
(1.96)

Subset 2

2.74
(2.50)

2.13
(1.39)

4.91
(4.04)

3.06
(2.34)

1.02
(1.36)

4.67
(5.13)

1.68
(1.77)

Subset 3

2.65
(2.45)

2.32
(1.67)

5.14
(3.85)

3.09
(2.29)

1.76
(2.83)

0.89
(1.13)

4.62
(5.73)

1.81
(2.00)

Subset 4

2.72
(2.56)

2.14
(1.46)

4.91
(4.04)

3.06
(2.40)

1.61
(2.51)

0.66
(0.95)

1.16
(1.91)

1.40
(1.97)

Subset 5

2.67
(2.47)

1.29
(1.08)

3.43
(2.31)

2.83
(2.30)

1.74
(2.74)
1.64
(2.61)
1.63
(2.62)

0.65
(0.97)
0.61
(0.93)
0.62
(1.01)

Subset 6

1.16
(1.91)

1.40
(1.96)

Subset 7

Subset 8

2.73
(2.58)
2.71
(2.53)
2.63
(2.55)

1.32
(1.09)
1.25
(1.10)
1.28
(1.15)

2.82
(2.21)
3.43
(2.31)
2.82
(2.21)

2.74
(2.29)
2.83
(2.29)
2.66
(2.26)

1.60
(2.71)

0.86
(1.09)

3.17
(4.02)

1.39
(1.73)

Epsilon-Based
method

2.84
(2.71)

2.04
(1.30)

4.20
(3.62)

2.79
(2.11)

Modified median &


(CMR/CMRacceptable)

Crustal
records

Subset 1

1.56
(2.45)

1.08
(1.36)

4.51
(4.75)

Subset 2

1.77
(2.80)

0.90
(1.05)

Subset 3

1.50
(2.42)

Subset 4
Subset 5
Subset 6
Subset 7
Subset 8
Epsilon-Based
method

Interface Inslab
All
records records events

204

Interface Inslab
All
records records events

Table 6.11. Comparison of the median collapse capacity (g) and (CMR/ CMRacceptable)
ratios using different subsets and methods for configuration 5 (R/I = 3.3) and a) D = 1.5
m and b) D = 2.5 m
a)

b)
Modified median &
(CMR/CMRacceptable)

Crustal
records

1.20
(1.52)

Subset 1

2.25
(1.98)

1.93
(1.38)

3.49
(2.51)

2.49
(1.86)

1.78
(2.58)

1.31
(1.71)

Subset 2

2.24
(1.98)

1.81
(1.24)

3.09
(2.58)

2.44
(1.91)

0.78
(1.03)

2.27
(2.96)

1.23
(1.53)

Subset 3

2.19
(2.01)

1.82
(1.37)

3.81
(2.86)

2.47
(1.87)

1.35
(2.15)

0.60
(0.75)

1.71
(2.60)

1.31
(1.81)

Subset 4

2.24
(2.03)

1.81
(1.32)

2.85
(2.62)

2.35
(1.98)

1.25
(1.94)

0.57
(0.79)

1.55
(2.04)

1.11
(1.42)

Subset 5

2.29
(2.02)

1.40
(1.11)

3.11
(2.05)

2.28
(1.75)

1.30
(2.05)
1.26
(2.00)
1.35
(2.15)

0.53
(0.72)
0.45
(0.66)
0.52
(0.71)

1.73
(2.61)
1.57
(2.09)
1.65
(2.66)

1.13
(1.53)
1.06
(1.34)
1.13
(1.54)

Subset 6

2.25
(1.99)
2.22
(2.02)
2.24
(2.03)

1.28
(1.01)
1.23
(1.02)
1.24
(1.00)

2.96
(2.19)
3.11
(2.05)
3.10
(2.44)

2.23
(1.72)
2.21
(1.70)
2.19
(1.72)

2.31
(2.09)

1.64
(1.10)

2.91
(2.53)

2.23
(1.78)

Modified median &


(CMR/CMRacceptable)

Crustal
records

Subset 1

1.23
(1.91)

0.80
(1.00)

2.07
(2.64)

Subset 2

1.30
(2.05)

0.68
(0.79)

Subset 3

1.23
(1.94)

Subset 4
Subset 5
Subset 6
Subset 7
Subset 8
Epsilon-Based
method

1.30
(2.18)

Interface Inslab
All
records records events

0.65
(0.83)

1.46
(2.36)

Subset 7
Subset 8
Epsilon-Based
method

1.03
(1.27)

Interface Inslab
All
records records events

As can be seen from the tables, the use of the simplified method for the crustal events for
all the subsets of records studied resulted in similar predictions. This indicates that for
this case the modified median collapse capacities were not sensitive to the subsets of the
records used. On the other hand the predicted modified median collapse capacities for the
subduction interface and the inslab events were similar within the first four subsets
(subset 1 to 4) and within the second four subsets (subsets 5 to 8). However the results
from subset 1 to 4 were somewhat different from those obtained using subset 5 to 8 in
which only the records with required smaller scale factors were considered. This
emphasizes the need to impose a limitation on the maximum allowable scale factors to
avoid the results being biased and unrealistic. For these cases imposing this limitation
decreased the predicted collapse capacity. This problem was found to be more important
for the subduction earthquake records, while the collapse capacities obtained using the
crustal records were less sensitive to the limitation on the scale factor. This may be due to
the fact that a larger number of strong crustal records are available in the databases and
hence smaller scale factors are required.

205

A comparison of the results obtained using the epsilon-based record selection method
with those obtained using the simplified method (modifying the results using SSF factor)
for the case of crustal events shows a very good agreement between the results even using
different record subsets. As an example, for the case of configuration 1 with column
diameters of 1.5 m the predicted median collapse capacity for crustal records using the
simplified method varied between 1.60 g to 1.66 g for 8 subsets of records studied, the
direct use of the epsilon-based method for this case resulted in a median collapse capacity
of 1.49 g. For other cases shown in the tables even better agreement of the results are
observed. In the epsilon-based method only the results from the records with similar
epsilon values to the mean epsilon values obtained at the fundamental period of the
structure from the seismic deaggregation were considered.

In the case of the subduction interface events the predicted median collapse capacity
using the simplified method was between 0.9 g to 1.06 g using the subsets 1 to 4 of the
records and for the subset 5 to 8 the predicted median varied between 0.59 g to 0.66 g.
The direct use of the epsilon-based record selection method however results in a median
collapse capacity of 0.70 g which is closer to the results obtained using subset 8 for this
case. The results obtained for various configurations studied indicate that typically the
use of subset 4 or 8 resulted in the predictions which were in better agreement with those
obtained using the epsilon-based method. This is mainly because in these subsets only the
records with positive epsilon values are considered which somewhat account for the
spectral shapes effects.

It is interesting to note that considering the records which required lower scaling factors
also reduced the record-to-record variability in the collapse capacity predictions. For
example in the case of interface events the record-to-record variability was around 0.7 in
subsets 1 to 4, while this variability reduced to around 0.5 to 0.6 in subsets 5 to 8. It must
be noted that since both the median collapse capacity and record-to-record variability are
important in the seismic performance evaluations, the use of the CMR/ CMRacceptable ratio
(i.e. the ratio of the collapse margin ratio to the acceptable collapse margin ratio) can
provide a better parameter for the purpose of comparing the results.

206

The predicted median collapse capacity for the inslab events varied between 3.35 g to
4.44 g in the subsets 1 to 4, and between 2.45 g to 2.76 g in subsets 5 to 8. The CMR/
CMRacceptable ratio varied between 3.43 to 4.27 in the subsets 1 to 4, and between 2.85 to
3.38 in the subsets 5 to 8. Similar high values of CMR and CMR/ CMRacceptable ratios
obtained for other configurations studied. These high values of CMR/ CMRacceptable ratio
indicate that the inslab events are the least critical case and the bridges are predicted to
have high collapse margin ratios for inslab earthquakes.

Typically the interface record sets resulted in the lowest collapse capacity predictions for
the bridges studied. However the difference between the results obtained using records
from different earthquake types is different for each configuration studied. For some
configurations the interface and crustal record sets resulted in similar collapse capacity
predictions (e.g., configuration 1 with D=2.5 m) and for most of the configurations this
difference was larger (e.g., configuration 3 with D=2.5 m). The results from the inslab
record sets were not found to be critical for the cases considered in this study. Similar
trends were observed for other configurations studied as can be seen in the tables.

6.11 Combining the results from different earthquake types


The collapse capacities for different event types (for all sets) were computed separately.
In order to predict an average collapse capacity, the results from different events should
be combined in proportion to the contribution of each earthquake type to the overall
seismic hazard obtained from the seismic hazard deaggregation. This means that for
example if many records with positive epsilon values are available in the crustal record
set but fewer records with positive epsilon values are available in the interface and inslab
record sets, then fewer crustal records were selected to maintain the ratio of the number
of records proportional to the percentages of contribution of the event types. In such
cases the priority in record selection was to the records with the closest epsilon values (at
the fundamental period of the structure) to the mean epsilon values obtained from the
seismic hazard deaggregation. In addition to the average median collapse capacity,
logarithmic standard deviation of the collapse capacity can be directly computed in this
207

method which represents the record-to-record variability in the collapse capacity


predictions. It must be noted that the IDA results are combined for different subsets of
records (i.e., 8 subsets were considered).

In order to combine the results from different events, the predicted collapse capacity of
each record was modified by applying the SSF factor to the unmodified collapse capacity
of the records. To do this, instead of using the average epsilon of records in Eq. [6.1], the
epsilon value of each record at the fundamental period of the structure was used to predict
the SSF for each record individually. It must be noted that for each earthquake type the
GMPEs specific to that earthquake type were used to compute the epsilon values of the
records and to predict the 1 factors.
It is interesting to note that typically the median collapse capacities obtained by
combining the effects of three earthquake types are close to the results obtained
considering the crustal records alone. Although the median collapse capacities were
similar, the record-to-record variability considering all the earthquake types was higher
than that obtained considering only crustal events. This may indicate that the median
collapse capacity of the structure could be predicted by considering only the crustal
events. To account for the influence of different earthquake types, the record-to-record
variability could be increased in the prediction of the acceptable collapse margin ratios
and probability of collapse.

6.12 1 and SSF values using different sets and subsets of records
The use of the simplified method to modify the predicted collapse capacities for the
effects of spectral shapes requires the 1 factors to be computed. The 1 factor shows how
sensitive the collapse capacity predictions are to changes in  (T1) values. Larger values
of 1 indicate that the predictions are more sensitive to the spectral shape which is
represented by the epsilon values at the fundamental period of the structure (i.e.,  (T1)).
The regression analysis method to compute the 1 factors was described in Section 6.4.

208

For example in Fig. 6.4 the 1 value from the regression analysis was calculated as 0.263
using 78 crustal records. The BA08 GMPE was used to compute  (T1).
2

ln[SC(T1)]

1.5
1
ln[SC(T1)] = 0.263  + 0.402

0.5
0

-0.5
-2

-1

 (T1)

Fig. 6.4. Regression analysis to compute 1 for the case of configuration 1 with D = 2 m
using BA08 GMPE

The results obtained for the bridges studied indicate that although the calculated values of
1 and SSF are different using different subsets of records, the predicted modified
collapse capacities computed by applying the spectral shape factor (SSF) to the median
collapse capacity of each subset, are similar. An example of the calculated 1 and SSF
factors for different subset of records and different earthquake types are shown in Table
6.12.

Results in this table also show the percentiles and standard deviations of the collapse
capacities obtained using different record sets and subsets. As can be seen the variability
of the results for the interface and inslab events were larger than that for the crustal
events especially in subsets 1 to 4. However when only records with small scale factors
were selected in subsets 5 to 8, the variability of the results from interface events became
lower, but at the same time the median predicted collapse capacities also decreased.

209

Regression analyses were performed to obtain a simplified relationship between the


1factors and the period of the structures studied. The records in subset 3 (i.e., all the
records with outliers removed) were used for this purpose. It must be noted that
considering different subsets of records resulted in different 1factors.
Table 6.12. An example of Statistics of unmodified collapse capacity predictions for
different earthquake types and for different subsets of records in each
subset(Configuration 4, D = 2.5 m, and R/I = 3.33).
IDA Percentiles (g)
(Unmodified for the
spectral shape effects)

Subset 1

Subset 2

Subset 3

Subset 4

Subset 5

Subset 6

Subset 7

Subset 8

50%

16%

84%

St
Dev

Num.
Mean 
of Rec. (Records)

SSF

Crustal

2.26

1.61

4.20

0.47

78

0.84

0.24

1.19

Interface

2.09

0.96

4.43

0.72

78

0.86

0.15

1.12

Inslab

2.73

1.34

5.13

0.67

78

0.33

0.38

1.76

Crustal

2.40

1.64

4.27

0.45

68

1.03

0.24

1.14

Interface

1.73

0.91

4.78

0.78

66

1.13

0.45

1.23

Inslab

3.50

1.78

5.30

0.56

48

0.85

0.35

1.40

Crustal

2.22

1.59

3.98

0.44

75

0.85

0.25

1.20

Interface

2.08

0.95

4.20

0.69

77

0.86

0.15

1.11

Inslab

2.71

1.32

4.89

0.65

76

0.36

0.43

1.90

Crustal

2.34

1.63

4.00

0.42

65

1.04

0.29

1.16

Interface

1.73

0.91

4.68

0.74

65

1.13

0.47

1.24

Inslab

3.50

1.78

5.30

0.56

48

0.85

0.35

1.40

Crustal

2.18

1.54

3.89

0.44

71

0.91

0.30

1.22

Interface

1.37

0.89

2.67

0.54

56

0.83

-0.08

0.94

Inslab

1.81

0.85

4.29

0.75

17

0.62

0.52

1.89

Crustal

2.32

1.63

3.94

0.41

64

1.05

0.31

1.18

Interface

1.28

0.84

2.69

0.55

50

1.00

0.06

1.03

Inslab

2.15

1.31

4.72

0.61

13

1.05

0.35

1.31

Crustal

2.17

1.51

3.50

0.42

69

0.93

0.34

1.24

Interface

1.36

0.87

2.61

0.49

54

0.83

-0.10

0.92

Inslab

1.81

0.85

4.29

0.75

17

0.62

0.52

1.89

Crustal

2.26

1.63

3.50

0.38

62

1.04

0.29

1.17

Interface

1.21

0.84

2.49

0.47

47

1.01

0.10

1.06

Inslab

2.15

1.31

4.72

0.61

13

1.05

0.35

1.31

210

0.35

0.25

a)

0.30

b)

0.20
0.15

0.20
1

1

0.25

0.15
0.10

1 = -0.109 T1 + 0.22

0.10
0.05

1 = -0.095 T1 + 0.32

0.05
0.00

0.00
0

0.5

1
1.5
Period (T1) (Sec)

2.5

0.5

1.5

-0.05
Period (T1) (Sec)

0.60

c)

0.50

1

0.40
0.30

1 = 0.0836 T1 + 0.3439

0.20
0.10
0.00
0

0.5

1
Period (T1) (Sec)

1.5

Fig. 6.5. Regression analysis for 1and period (T1) considering all records with removed
outliers (i.e., subset 3) for a) crustal records, b) interface records and c) inslab records

As can be seen from Fig. 6.5(a) and (b), an increase in the period of a bridge results in a
decrease in the 1 factor. Each point in these figures represents the IDA results obtained
for a specific bridge configuration subjected to 78 records. For the inslab events the
1factors increased by increasing the period of the structure (see Fig. 6.5 (c)).

6.13 Effect of number of records (44 records vs 78 records)


For each earthquake type 78 records were used to perform the IDA and to compute the
collapse capacities of the bridges. A comparison was made of the results obtained using
78 and 44 records for the bridge configurations designed using R=5 and I=1.5. Only the
results computed using subset 3 (all records with outliers removed) will be studied here to
facilitate the comparisons.

211

2.5

The results indicate that the differences in the unmodified collapse capacities are around
5% on average (except for inslab records which are not critical). A similar trend was
observed in the modified collapse capacities and CMR/ CMRacceptable ratios. The
modified median collapse capacities are given in Table 6.13.

Table 6.13. Comparison of the modified median collapse capacities obtained using 44
records and 78 records (in the parentheses) for different earthquake types and combined
earthquake types
Modified median collapse
capacity (g) using 44
records & (78 records)
Configuration1 (D=1.5 m)
Configuration1 (D=2.0 m)
Configuration1 (D=2.5 m)
Configuration2 (D=1.5 m)
Configuration2 (D=2.0 m)
Configuration2 (D=2.5 m)
Configuration 3 (D=1.5 m)
Configuration 3 (D=2.0 m)
Configuration 3 (D=2.5 m)
Configuration 4 (D=1.5 m)
Configuration 4 (D=2.0 m)
Configuration 4 (D=2.5 m)
Configuration 5 (D=1.5 m)
Configuration 5 (D=2.0 m)
Configuration 5 (D=2.5 m)

Crustal
records

Interface
records

Inslab
records

Combined

1.56 (1.61)

0.98 (1.05)

4.35 (4.44)

1.74 (1.64)

2.09 (2.12)

1.93 (1.83)

5.90 (6.33)

2.54 (2.51)

2.89 (2.77)

2.44 (2.85)

7.23 (7.13)

3.33 (3.40)

1.45 (1.35)

0.86 (0.89)

3.04 (3.52)

1.22 (1.20)

1.63 (1.71)

1.19 (1.21)

5.90 (4.99)

1.66 (1.61)

2.26 (2.29)

1.23 (1.38)

7.85 (6.54)

2.39 (2.41)

1.04 (1.04)

0.70 (0.74)

4.23 (3.24)

1.07 (1.11)

1.61 (1.45)

0.73 (0.98)

4.63 (4.40)

1.49 (1.49)

1.62 (1.91)

1.14 (1.17)

4.65 (4.78)

1.83 (1.94)

1.76 (1.50)

0.87 (1.02)

4.55 (4.67)

1.71 (1.68)

2.03 (2.08)

1.38 (1.51)

4.42 (3.91)

2.26 (2.21)

2.77 (2.65)

2.35 (2.32)

5.51 (5.14)

3.25 (3.09)

1.34 (1.23)

0.61 (0.78)

2.30 (2.27)

1.23 (1.23)

1.72 (1.76)

1.28 (1.19)

2.62 (2.84)

1.78 (1.74)

2.28 (2.19)

1.81 (1.82)

4.17 (3.81)

2.67 (2.47)

The differences in the case of crustal records were in the range of 5%, while somewhat
higher differences were observed in the case of subduction events. Nevertheless on
average the differences were typically less than 15%. Results indicate that the interface
and inslab events are more sensitive to the number of records used which is due to higher
variability of responses due to such events.

212

6.14 Comparison of the bridge responses for different configurations


To investigate the influence of the regularity of the bridge structures on the overall
seismic response and safety, the collapse margin rations were compared for the different
bridges studied. The results obtained for the regular bridges and irregular bridges are
compared (see Table 6.14). The results are based on subset 3 of the records. Because the
inslab events were not critical only the results obtained from the crustal and interface
events are presented.

Table 6.14. Modified and unmodified median collapse capacities and corresponding
collapse margin ratios (CMR) and (CMR /CMRacceptable) using subset 3 of records. (D
= 2 m, and R/I = 3.33)

Crustal
Configuration 1
Interface
(D=2.0 m)
Average
Crustal
Configuration 2
Interface
(D=2.0 m)
Average
Crustal
Configuration 3
Interface
(D=2.0 m)
Average
Crustal
Configuration 4
Interface
(D=2.0 m)
Average
Crustal
Configuration 5
Interface
(D=2.0 m)
Average

Median

CMR

1.71
1.68
1.69
1.45
1.20
1.32
1.21
0.93
1.07
1.79
1.33
1.56
1.52
1.12
1.32

3.53
3.46
3.49
6.86
5.66
6.26
4.34
3.34
3.84
5.05
3.75
4.40
3.88
2.85
3.37

Modified Modified
median
CMR
2.12
1.83
1.98
1.71
1.21
1.46
1.45
0.98
1.21
2.08
1.51
1.79
1.76
1.19
1.47

4.37
3.78
4.07
8.06
5.73
6.90
5.19
3.51
4.35
5.85
4.25
5.05
4.49
3.02
3.76

TOT
0.61
0.81
0.71
0.57
0.75
0.66
0.60
0.76
0.68
0.58
0.86
0.72
0.57
0.79
0.68

CMRacceptable CMR/CMRacceptable
2.18
2.81
2.50
2.08
2.60
2.34
2.15
2.64
2.40
2.10
3.01
2.55
2.08
2.74
2.41

2.00
1.34
1.67
3.88
2.21
3.04
2.41
1.33
1.87
2.79
1.41
2.10
2.16
1.10
1.63

As can be seen in the table the modified collapse margin ratio is 3.76 for configuration 5,
which is the most irregular bridge. This predicted value is smaller than those for the other
configurations. Regular configurations 2 and 4 have average modified collapse margins
of 6.9 and 5.05 respectively which are the largest CMR values obtained. For
configuration 3, which represents a bridge ramp, the average CMR is 4.35 which is lower
than those obtained for the regular configurations. It is noted that regular configuration 1
with stiff columns has relatively low collapse margin ratios. This is due to the increased

213

demand in the middle column. This illustrates that in addition to the relative stiffness
ratio of the columns along the bridge that the deflected shape is also important and should
be addressed in the design of bridge structures. Hence the columns at the locations of
maximum expected displacement demands (i.e., center of bridge for bridges with
restrained abutment conditions and near the ends for bridges with free movements at the
abutments) could result in lower safety margins. It should be noted however that for
configuration 1, the design reinforcement ratio is around 1.17% in the central column,
while for configuration 5 this ratio is around 2.2%. This resulted in the nominal moment
strength of the central column in configuration 5 being about 50% greater than that of the
column in configuration 1. However, due to the influence of irregularity on the seismic
response, the collapse margin ratios are smaller for configuration 5.

Table 6.15. Modified and unmodified median collapse capacities and corresponding
CMR and CMR /CMRacceptable using subset 3 of records. (D = 2.5 m, and R/I = 5)

Crustal
Configuration 1
Interface
(D=2.5 m)
Average
Crustal
Configuration 2
Interface
(D=2.5 m)
Average
Crustal
Configuration 3
Interface
(D=2.5 m)
Average
Crustal
Configuration 4
Interface
(D=2.5 m)
Average
Crustal
Configuration 5
Interface
(D=2.5 m)
Average

Median

CMR

2.42
2.52
2.47
1.80
1.32
1.56
1.38
0.98
1.18
2.05
1.88
1.96
1.51
1.31
1.41

4.06
4.22
4.14
6.60
4.85
5.72
4.53
3.22
3.87
4.33
3.98
4.16
3.29
2.85
3.07

Modified Modified
median
CMR
2.83
4.74
2.83
4.74
2.83
4.74
2.29
8.41
1.38
5.08
1.84
6.74
1.62
5.33
1.05
3.46
1.34
4.40
2.52
5.33
2.12
4.48
2.32
4.91
1.81
3.94
1.44
3.12
1.62
3.53

TOT
0.59
0.79
0.69
0.63
0.78
0.70
0.60
0.78
0.69
0.63
0.82
0.73
0.59
0.78
0.69

CMRacceptable CMR/CMRacceptable
2.13
2.76
2.45
2.24
2.71
2.48
2.17
2.71
2.44
2.26
2.88
2.57
2.14
2.72
2.43

2.22
1.72
1.97
3.75
1.87
2.81
2.46
1.28
1.87
2.36
1.56
1.96
1.84
1.15
1.49

The results obtained for the case of different bridge configurations with D=2.5 and R/I=5
are shown in Table 6.15. The results in this case are interesting, since the minimum
reinforcement requirements were controlling for most of the configurations so that the
bridge columns had almost similar strengths. A comparison between the CMR values
obtained for the regular and irregular bridges in this case can better present the effects of

214

irregularity on the seismic behaviour of the bridges. For example, the CMR values
obtained for the highly irregular bridge (i.e., configuration 5) is significantly lower than
those obtained for the regular configurations 2, 4 and 1 which highlights the significance
of regularity on the seismic response.

6.15 Comparing design using R/I=3.33 and R/I=5


The bridge structures were designed using two different R/I values of 3.33 and 5. While
higher R/I values give lower seismic demands, for R/I=5 the minimum reinforcement
limit of 0.8% was controlling in most cases and this resulted in strengths somewhat above
the demand strength. In addition, the use of a higher R/I value in design will result in a
more flexible bridge with lower seismic demands. As can be seen in Table 6.16, the use
of R/I of 5 in design did not significantly reduce the predicted collapse margin ratios.

Table 6.16. Modified median collapse capacities and corresponding collapse margin
ratios (CMR) using subset 3 of records for R/I=3.33 and R/I=5. (D = 2.0 m)
R/I = 3.33

Crustal
Configuration 1
Interface
(D=2.0 m)
Average
Crustal

Configuration 2
(D=2.0 m) Interface
Average
Crustal
Configuration 3
Interface
(D=2.0 m)
Average
Crustal
Configuration 4
Interface
(D=2.0 m)
Average
Crustal
Configuration 5
Interface
(D=2.0 m)
Average

Modified
median
2.12
1.83

R/I = 5

Modified
CMR/CMRacceptable
CMR
4.37
2.00
3.78

1.34

Modified
median
1.97

Modified
CMR
4.23

1.53

3.28

CMR/CMRacceptable
1.95
1.22

1.98

4.07

1.67

1.75

3.75

1.58

1.71

8.06

3.88

1.71

8.06

3.88

1.21

5.73

2.21

1.21

5.73

2.21

1.46
1.45
0.98
1.21
2.08
1.51
1.79
1.76
1.19
1.47

6.90
5.19
3.51
4.35
5.85
4.25
5.05
4.49
3.02
3.76

3.04
2.41
1.33
1.87
2.79
1.41
2.10
2.16
1.10
1.63

1.46
1.26
0.80
1.03
1.79
1.33
1.56
1.31
0.97
1.14

6.90
4.67
2.96
3.81
5.51
4.10
4.81
3.99
2.94
3.47

3.04
2.21
1.13
1.67
2.66
1.37
2.01
1.94
1.00
1.47

215

Table 6.16 compares the results for bridges designed with R/I values of 3.33 and 5. The
results obtained for bridges with D=2.0 m is presented, while similar results observed for
other cases. For configuration 2, because the minimum reinforcement requirement
controlled the design for all of the columns for both cases of R/I equal to 3.33 and 5, the
predicted results are similar. Based on the results obtained for all cases, a reduction of
about 10 % to 15% in the collapse margin ratios was typically observed for the case of
R/I= 5.

6.16 Conclusions
The seismic responses of bridges with 5 different configurations and having different
column stiffnesses were studied using Incremental Dynamic Analysis (IDA). The
influence of different earthquake types on the response was considered by using three
different sets of records in IDA representing three different earthquake types. The
conclusions from this study of bridges located in Vancouver are summarized below:

(1) The median collapse capacities where typically lower for the bridges when the
subduction interface records were used in the IDA. This is likely due to different
frequency contents and much longer duration of the interface events compared to the
crustal events. It was found that the inslab events were not critical and resulted in higher
median collapse capacities and collapse margin predictions. This may be due to the
relatively long period of the 4-span bridge structures, while inslab events are usually
more critical for structures with shorter periods.

(2) When the records were combined in proportional to the percentage of contribution
of each of the three event types, it resulted in the average median collapse capacities
being similar to those determined using the crustal records only. However the variability
of the collapse capacity predictions was higher when the three different event types were
considered. This may indicate that the crustal records could be used along with a
modified record-to-record variability to include the effects of different event types in
216

seismic performance assessments. However, more research is however required to verify


this.

(3) The use of the simplified method to account for the effects of spectral shapes
typically resulted in predictions that were in good agreement with those obtained using
the more complex epsilon-based record selection method which directly accounts for the
spectral shapes.

(4) Considering only the records with lower scale factors at collapse reduced the
variability in the collapse capacity predictions. The median collapse capacities obtained
considering the lower scaling factors were generally lower than those obtained using
higher scaling factors, particularly for interface and inslab events. Care must be taken to
make sure that the results obtained are not biased due to the application of large scaling
factors.

(5) The values of 1 and SSF factors were computed using different subsets of records
and for different earthquake types and simplified equations were derived to predict the 1
values. It was found that although the predicted values of 1 and SSF factors using
different subsets of records were different, the application of these factors to the
unmodified median collapse capacity of the corresponding subsets resulted in relatively
similar modified median collapse capacities.

(6) The results from IDA for the different bridge configurations indicate that for
irregular bridges the collapse margin ratios were lower than for regular bridges and the
effects of irregularity were more pronounced as the force modification factor, R,
increased.

217

Acknowledgements
The financial support provided by the Natural Sciences and Engineering Research
Council of Canada for the Canadian Seismic Research Network is gratefully
acknowledged. Ground motion data for Japanese earthquakes were obtained from the KNET and KiK-NET databases at www.k-net.bosai.go.jp and www.kik.bosai.go.jp and
other crustal ground motion data obtained from the PEER-NGA database available at
http://peer.berkeley.edu. The Nisqually earthquake records were obtained through
COSMOS virtual data center available at http://db.cosmos-eq.org. Data on Japanese
earthquake records was provided by Katsuichiro Goda.

References
Please see the last section.

218

7 Conclusions

7.1 Summary and main conclusions


The conclusions from each manuscript were stated at the end of the chapters. The main
conclusions from this thesis, based on studies of four-span reinforced concrete bridges
located in Vancouver, are summarized as follows:

7.1.1

Conclusions regarding seismic response and evaluation of bridges with


column stiffness irregularities

The force-based design procedures of the 2006 CHBDC gives conservative


designs due to the combined use of low R factors and the use of importance
factors, even for irregular bridges.

In the transverse direction, contrary to that in the longitudinal direction, the


seismic behaviour is also affected by the position of columns along the length of
the bridge and the superstructure stiffness properties. Such parameters are not
currently considered in the definition of irregularity in the seismic codes.

It was shown that exceeding a maximum column stiffness ratio of about 5 to 8 can
cause significant concentration of ductility demand as well as deviation of the
predicted non-linear response from the predicted elastic response.

A superstructure to substructure stiffness ratio, RS, of less than 0.3 in the case of
restrained movements at the abutments, and 0.8 in the case of free movements at
the abutments may result in incorrect response predictions using elastic multimode analysis. Bridges with RS ratios greater than 1.0 typically had more uniform

219

ductility demands in the columns, while bridges with RS values less than 1.0 may
exhibit concentrations of seismic ductility demand in a few elements.
x

While the column stiffness ratios and the superstructure to substructure stiffness
ratios provide simple, practical means of identifying regularity, the DRMS index
can also be used to choose an appropriate method of analysis. For bridges with
restrained abutment conditions and with DRMS index values of less than 0.08 the
results from the elastic and inelastic analyses were in good agreement.

Based on the results obtained, for some irregular bridge configurations, the
displacement envelopes were significantly different from the elastic and inelastic
analysis; however due to high conservatism in the current CHBDC this did not
lead to an unsafe design at least for the cases considered in this study. This
problem may be more important if less conservative design approaches are used
and when higher earthquake levels are considered.

The seismic response and the maximum ductility demands in the longitudinal
direction are controlled by the total stiffness of the substructure, the stiffness ratio
of the columns, and the minimum aspect ratio of the columns. Seismic ductility
demands in the longitudinal direction were correlated with the product of the total
stiffness of the columns and the maximum stiffness ratio of the columns. This
indicates that the ductility demands in bridge columns increase as the structural
stiffness and stiffness irregularity increases.

Analyses in the longitudinal direction of bridges demonstrated that the


concentration of ductility demands increases significantly with an increase in the
column stiffness ratio, Kmax/Kmin. An equation was developed to provide a means
of estimating the ratio of the maximum to minimum ductility demands. This
equation is useful for designers in determining the influence of the column
stiffness irregularity on the concentration of the ductility demands.

220

The influence of the abutments on the longitudinal seismic responses of bridges


was studied. Up to an 80% decrease in seismic ductility demands were observed
when the abutments were considered in the structural models. The reduction of
ductility demand correlates with the ratio of the total stiffness of the columns and
the effective stiffness of the abutments. The influence of the abutments was more
pronounced for the bridges with more flexible columns and stiffer abutments.

A dimensionless parameter was defined as ( D) / H2 (i.e., drift ratio divided by


column aspect ratio) which provided an improved indicator of the structural
damage compared to the conventional drift ratio (i.e.,  / H ). It was also
demonstrated that normalizing the maximum ductility demands by the minimum
aspect ratio of the columns significantly reduced the dispersions in the results.

The seismic ductility demands to ductility capacity ratios were estimated for the
combination of the seismic responses in the longitudinal and transverse directions.
It was observed that the demand to capacity ratios were lower than 0.7 with the
majority of the cases having values less than 0.5. These ratios decreased, when the
influence of the abutments were considered in the seismic response. The range of
demand to capacity ratios was quite high which indicate uneven safety margins
for different bridges. Exceeding the maximum stiffness ratio of about 5.0 to 8.0
resulted in much larger demand to capacity ratios.

CSA S6-06 requires elastic dynamic analysis for an emergency-route bridge in


seismic performance zones 2 and higher if the bridge is irregular. This study
indicates that elastic dynamic analysis is appropriate for irregular bridges in the
longitudinal direction. However, nonlinear dynamic analysis would be required
for such irregular bridges in the transverse direction in order to accurately predict
the displacement envelope and the ductility demands.

221

7.1.2

Conclusions regarding the use of incremental dynamic analysis (IDA) for


seismic analysis and evaluation of bridges

The use of the UHS-based method resulted in very conservative predictions of


collapse capacities which were significantly lower than those predicted using the
epsilon-based method. The use of the CMS- or UHS-based methods generally
underestimated the record-to-record variability compared to the epsilon-based
record selection method.

It was demonstrated that the IDA results can be sensitive to the choice of the
record selection method. The epsilon-based method in all of the cases resulted in
the highest prediction for the capacity of the structure. However more research is
required to verify the reliability of this method for seismic evaluations, as it was
found that the results can be sensitive to the number of records used. Furthermore,
different sets of records with comparable mean epsilon values could result in
somewhat different collapse capacity predictions. Increasing the number of
records used in the IDA did not influence the results obtained using the CMSbased record selection methods. The use of different GMPEs for record selection
did not significantly affect the results in IDA. The influence of the record
selection methods was greater on the collapse capacity prediction than on the
other damage states such as yielding and spalling.

The CMS and the mean epsilon values were predicted using the matching method
in seismic deaggregation. The use of the exceeding method for seismic
deaggregation tended to overestimate the mean epsilon values and thus the
collapse capacity of the structure.

The median collapse capacities where typically lower for the bridges when the
subduction interface records were used in the IDA. This is likely due to different
frequency contents and much longer duration of the interface events compared to
the crustal events. It was found that the inslab events were not critical and resulted

222

in higher median collapse capacities and collapse margin predictions. This may be
due to the relatively long period of the 4-span bridge structures, while inslab
events are usually more critical for structures with shorter periods.
x

When the records were combined in proportional to the percentage of contribution


of each of the three event types, it resulted in the average median collapse
capacities being similar to those determined using the crustal records only. It was
also demonstrated that the average conditional mean spectrum for all earthquake
types (i.e., CMS-All) was close to the CMS-Crustal and thus combining the
effects of different seismic event types led to similar results as considering crustal
events alone. However the variability of the collapse capacity predictions was
higher when the three different event types were considered. This may indicate
that the crustal records could be used along with a modified record-to-record
variability to include the effects of different event types in the seismic
performance assessments. More research is however required to verify this.
The increased record-to-record variability due to considering different earthquake
types has at least two outcomes: first, the IDA needs to be carried out using a
larger number of records for a more precise prediction of the collapse capacity of
the structure and second, the increased variability will increase the probability of
collapse at the MCE level.

The use of the simplified method to account for the effects of spectral shapes
typically resulted in predictions that were in good agreement with those obtained
using the more complex epsilon-based record selection method which directly
accounts for the spectral shapes. The values of 1 and spectral shape factors (SSF)
were computed using different subsets of records and for different earthquake
types and simplified equations were derived to predict the 1 values. It was found
that although the predicted values of 1 and SSF factors using different subsets of
records were different, the application of these factors to the unmodified median
collapse capacity of the corresponding subsets resulted in similar modified
median collapse capacities.

223

The results from IDA for different bridge configurations indicate that for irregular
bridges the collapse margin ratios were lower than for regular bridges and the
effects of irregularity were more pronounced as the force modification factor, R,
increased.

7.2 Future research


x

This research focused on the irregularity of bridges due to different stiffnesses of

columns. Similar studies can be conducted to study the influence of different span
lengths, skewness and curvature of bridges accompanied with the irregular distribution of
column stiffnesses.
x

The behaviour of irregular bridges with different column stiffnesses in the near-

field regions subjected to high velocity pulses needs to be studied.


x

The analyses presented in Chapters 3 and 4 can be carried out using natural

records matched to the conditional mean spectrum (CMS), rather than the UHS, for a
more precise evaluation of the seismic performance of bridges.
x

The seismic behaviour of irregular bridges was studied accounting for the

maximum considered earthquake level (i.e., 2% probability of exceedance in 50 years).


More studies can be carried out to study the seismic response of irregular bridges at
different hazard levels such as that corresponding to a serviceability limit state.
x

The effects of bidirectional excitation of ground motions could be studied directly

by applying the pairs of ground motions simultaneously.


x

A similar study needs to be carried out for case of older existing bridges with poor

detailing. The effects of column stiffness irregularity are even more important for older
bridges. This is because due to poor detailing, the concentration of ductility demands in
the shortest column will result in degradation of the shear strength of the columns. On the
other hand high shear forces are attracted in the shortest columns due to larger stiffness of
columns. This may lead to brittle shear failure and catastrophic collapse of older bridges.

224

In this study only columns with aspect ratios larger than 2.5 were considered (i.e.,

flexure dominant columns). Some studies could be carried out for columns with lower
aspect ratios. The response of such columns is controlled by shear and more detailed
nonlinear shear models should be incorporated in structural modelling. The shear-flexure
interactions (especially in the case of older bridges) can also be studied.
x

More research is needed on the applications of other methods of analysis

including the single-mode-pushover, modal pushover analysis (as used in Appendix A)


and adaptive pushover analyses in prediction of the seismic response of irregular bridges.
Appropriate criteria for application of such analyses methods to irregular bridges are
required.
x

The use of direct displacement-based design approach (as discussed in Chapter 1)

for the design of a wide range of irregular bridges can be studied and the results be
compared with those obtained in this study using the force-based design approach. It is
expected that the seismic performance of irregular bridges will be improved, when
displacement-based design concepts are used.

225

7.3 Statement of original contributions


The original contributions in this thesis include:
x

The seismic behaviour and safety margin of bridges with different stiffness
ratio and with different modelling and design properties for a wide range of
bridges with different column and superstructure stiffnesses was studied.
Important aspects which influence the seismic behaviour of bridges were
investigated and recommendations were made on suitable ranges of these
parameters to improve the seismic performance of bridges.

The results of elastic versus inelastic analysis were compared for bridges with
different configurations and different parameters. The use of different
regularity indices, proposed in the past, to predict the response of bridges were
also evaluated and compared for a wide range of bridges. Recommendations
were made for the improvement of the current CHBDC provisions to address
the effects of irregularity in seismic analysis and design of bridges.

The use of different record selection methodologies for the seismic


performance evaluation of bridges using incremental dynamic analysis was
studied. In addition, the effects of including the records from three different
earthquake sources including subduction interface events, inslab events and
crustal events, were investigated.

The effects of epsilon and spectral shapes on the seismic performance


assessment of bridges were studied for different damage states and for
different earthquake types. The use of the Conditional Mean Spectrum (CMS)
on the seismic evaluation of bridges was studied and the results were
compared with the case of the conventional UHS-based method. The use of
CMS for different event types and the average CMS for all event types in the
seismic evaluation of bridges was also studied. The use of different ground

226

motion prediction equations on the seismic performance assessment of bridges


using conditional mean spectrum (CMS) was also investigated
x

A large set of ground motion records for subduction earthquakes (i.e.,


interface and inslab) and crustal earthquakes were used for the seismic
evaluation of different bridge configurations and the effects of using different
subsets of records with specific properties in the seismic performance
assessments were also studied. The use of three different earthquake types
based on the seismic deaggregation results in seismic evaluation of different
bridge configurations were studied and the results were compared with the
case that only crustal events are considered (i.e., current design practice).
Recommendations were made to include the effects of different event types in
the seismic assessment of bridges.

The ATC-63 (ATC, 2008) methodology was applied to bridge structures. The
evaluation of damage states using different theoretical and experimental
approaches were carried out and a backbone curve for the evaluation of the
bridges using the IDA approach was defined.

A comprehensive program was developed, as discussed in Appendix B, which


is capable of design, modelling and evaluating the bridges for large parametric
studies using a large number of ground motion records and by means of
different methods of analysis. A large pool of ground motions (both artificial
and natural records) were incorporated in the program. Furthermore, other
programs were also developed as discussed in Appendices C to E for record
selections using different methods, computing the CMS and the epsilon values
and extracting the seismic deaggregation results.

227

Appendix A: Verification of the models, preliminary studies and other


modelling details
The modelling details for bridges studied in this research are discussed in Chapters 3 and
5. This appendix provides further details concerning the modelling and verification of the
results. In addition, some preliminary results obtained regarding the seismic response of
irregular bridges with different column heights are presented.
A.1. Verifications of computer models

The lumped plasticity models, as discussed in Chapter 3, have been widely used in
research and practice to predict the seismic response of structures. The application of
these models along with appropriate hysteresis loop models can successfully predict the
maximum displacement obtained during the seismic excitations which are in good
agreement with the experimental results (e.g., Saiidi and Sozen, 1979; Otani, 1981; Calvi
et al. 1994; Fischinger and Isacovic, 2003; Priestley et al., 1996 and 2007 and ATC-62,
2008). The recommendations by Priestley et al. (1996 and 2007) were used in this
research for the prediction of the modelling parameters including plastic hinge lengths,
strain penetration length, moment curvature idealization, etc.

Although the use of such models for prediction of the seismic response of structures are
widely verified and used and are recommended by many seismic design codes (e.g.,
Caltrans (2006) and AASHTO Guidelines (2009)) for analysis and design of structures,
an attempt is also made in this study to compare the numerical results with those obtained
from experiments.

228

a) Cyclic tests of columns

The modified Takeda hysteresis model (Otani, 1981) was used in this study to model the
behaviour of the RC columns using Ruaumoko software (Carr, 2009).This model has two
main parameters,  and , which control the unloading and the reloading stiffness,
respectively. The parameter  is usually in the range of 0 to 0.5 and  varies between 0
and 0.6. Increasing the parameter  decreases the unloading stiffness and increasing the 
parameter increases the reloading stiffness. Some researchers computed the mean values
of  and  based on experiments, as 0.26 (Stdev=0.13) and 0.49 (Stdev=0.15),
respectively (Mechakhchekh, 2008). Priestley et al. (2007) recommend using =0.3 and
=0.6 for the case of well-detailed beams with no axial force (i.e., Fat Takeda). For the
building and bridge columns the values of  =0.5 and  =0 is recommended (i.e., Thin
Takeda).
Some reinforced concrete columns were tested by Benzoni and Priestley (1996). The
experimental data from that study is available in the PEER structural database (Berry et
al., 2004). The lumped plasticity models along with the modified Takdea hysteresis loop
(Otani, 1981) were used to compare the experimental and numerical results. As shown in
Fig. A.1a, the use of =0.5 conservatively underestimate the unloading stiffness while the
use of =0.3, as shown in Fig. A.1b, can predict the unloading stiffness very well. It
should be noted that the column tested by Benzoni and Priestley (1996) had a small
percentage of longitudinal reinforcement (i.e., 0.5%) and small axial load ratio (i.e., 6%).
This resulted in high pinching effects in the hysteresis behaviour. The effects of pinching
on the overall response of structures with medium to long period ranges (e.g., the range
of period of the bridges studied in this research) have been shown to be minimal (e.g.,
ATC-62, 2008). Where the effects of pinching are controlling, the SINA hysteresis loop
model (Saiidi and Sozen, 1971) may be used to include the pinching effects (e.g., see Fig.
A.1.c).

229

500

500
TEST
Numerical

400

Test
Numerical

Force (KN)

300

Force (KN)

-15

200

100

100

-5-100

15

-25

25

-15

Disp. (cm)

-200

a)

300

200
0
-25

400

0
-5
-100

-300

-400

-400

Takeda model ( =0.5 and =0)

15

25

Disp. (cm)

-200

-300
-500

Takeda model ( =0.3 and =0)

-500

b)
500

Force (KN)

400

Test

300

Numerical

200
100
0

-25

-15

-5

-100

5
Disp. (cm)

15

25

-200
-300

SINA hysteresis model

-400
-500

c)

Fig. A.1. The use of lumped plasticity model along with Takeda hysteresis model to predict the
experimental results by Benzoni and Priestley (1996).

b) Bridge tested in the European Laboratory for Structural Assessment (ELSA)

In the framework of an integrated European programme of pre-normative research in


support of Eurocode 8 (CEN, 2002), some bridge prototypes, representative of typical
multi-span continuous deck motorway bridges, have been designed (Pinto et al., 1996)
with different procedures for a PGA of 0.35g, in medium soil conditions (soil type B),
applying the EC8 provisions. Corresponding large-scale (1:2.5) bridge models have then
been constructed and tested in pseudo-dynamic fashion at the Joint Research Centre at
Ispra (Italy) (Pinho, 2007).

230

The tested bridge model labelled as B213C in the experimental study by Pinto et al.
(1996) will be considered to evaluate the predictions from the numerical models. The
experimental data from this test was provided by the European Laboratory for Structural
Assessment (ELSA). This bridge consists of three piers 5.6, 2.8 and 8.4 m high and a
continuous deck with four identical 20 m spans. The superstructure is restrained against
the transverse movements at the ends, as shown in Fig. A.2, by means of the abutment
shear keys. The deck-pier connections are assumed to be hinged (no transmission of
moments). The piers have rectangular hollow sections with 160 mm wall thickness (Fig.
A.2). The reinforcement layout of the pier models are shown in Fig. A.3. The cross
sections of piers 1 and 3 in Fig. A.2, are shown in Fig. A.3 as section type 4 and the cross
section of pier 2 is shown as section type 1. The mechanical characteristics of materials,
pier and superstructure details and other required data are available by Pinho (2007) and
Pinto et al. (1996).

Fig. A.2. Bridge configuration and member cross sections of 1:2.5 scale model (from
Pinho (2007))

231

Fig. A.3. Reinforcement layout for columns (from Pinho (2007))

Fig. A.4. Acceleration history of the design earthquake (from Pinto et al. (1996))

232

Fig. A.5. Design earthquake, corresponding response spectrum and EC8-Soil B spectrum
(5% damping) (from Pinto et al. (1996))
10
8

Acceleration (m/s 2)

6
4
2
0
-2

0.5

1.5

2.5

3.5

-4
-6
-8
-10

Time (Sec)

Fig. A.6. Acceleration history of the design earthquake applied to the (1:2.5) scale model

The details of the ground motion record used along with its response spectrum are shown
in Fig. A.4 and Fig. A.5, respectively. The same record was used in this study to compare
the numerical and experimental results. Fig. A.6 demonstrates the ground motion record
which is applied to the bridge model. The application of the ground motion record in Fig.
A.6 to the 1:2.5 scale model has the same effect as the application of the ground motion
record in Fig. A.4 to the original bridge.

233

To further investigate the validity of the use of lumped plasticity models, an attempt is
made to model this bridge in the RUAUMOKO program. First, the middle column was
modelled and subjected to the reversed cyclic displacement pattern shown in Fig. A.7.
The hysteresis loops from the numerical analysis are compared to those obtained from the
test. As shown in Fig. A.8, the hysteresis loops can be predicted with reasonable
approximation, when appropriate hysteresis parameters are adopted. The area of the
hysteretic loops was computed to estimate the dissipated energy obtained from the test
and the numerical results. The use of unloading stiffness parameter, , of 0.0, 0.3, and 0.4
in the modified Takeda hysteresis model resulted in 0.91, 0.65, and 0.57 MN.m energy
dissipation, respectively. The measured dissipated energy from the test was computed as
0.58 MN.m. This indicates that the use of =0.4 is appropriate to estimate the total
amount of energy dissipated through cyclic testing. Increasing the parameter  decreases
the unloading stiffness and thus the area under the hysteresis loops.
0.1

Displacement (m)

0.08
0.06
0.04
0.02
0
-0.02
-0.04
-0.06
-0.08

Fig. A.7. Displacement history used for reversed cyclic testing

234

Short column (modified Takeda model, =0 , =0)

Short column (modified Takeda model, =0.3 , =0)

a)

b)

1.5

1.5

Force (MN)

Force (MN)
1

0.5

0.5

0
-0.1

-0.08

-0.06

-0.04

-0.02

0.02

0.04

0.06

0.08

0.1-0.1

-0.08

-0.06

-0.04

-0.02

-0.5

0.02

0.04

0.06

0.08

0.1

Displacement (m)

-0.5

Displacement (m)
-1

-1

-1.5

-1.5

-2
Test

-2

Numerical

Test

Numerical

Short column (modified Takeda model, =0.4 , =0)


2

c)

1.5

Force (MN)

0.5

0
-0.1

-0.08

-0.06

-0.04

-0.02

0.02

-0.5

0.04

0.06

0.08

0.1

Displacement (m)

-1

-1.5

-2
Test

Numerical

Fig. A.8. Comparison of the test results with the numerical predictions using the modified
Takeda model with: a) =0 and =0; b) =0.3 and =0; c) =0.4 and =0
The bridge structure, shown in Fig. A.2, was modelled in the RUAUMOKO program.
The detailed properties of the superstructure and substructure elements are described by
Pinto et al. (1996) and Pinho (2007). The superstructure was modelled using elastic beam
elements and the superstructure mass was lumped at the nodal points as shown in Fig.
A.9. In addition, the mass of the columns were considered in the corresponding nodal
points. Rigid elements were defined to model the superstructure depth. The heights of
235

these elements were considered as half of the superstructure depth (see Fig. A.9). The
columns were modelled using inelastic beam elements. The plastic hinge lengths and the
strain penetration depths were computed using the equations given by Priestley et al.
(2007) which are accepted for the bridges studied in this research. To consider the
deformations due to bond slippage the strain penetration depth, Lsp, was considered in
modelling the column heights as recommended by Priestley et al. (1996 and 2007). A
Takeda hysteresis model with =0.5 and =0 (i.e., thin Takeda model) was used based on
the recommendations by Priestley et al. (2007) for bridge columns.

Nodes (lumped mass)

Rigid links

Super structure (elastic element)

1/2 * Superstructure depth

H2+Lsp

H1 +Lsp

H3+Lsp
Columns
( lumped plasticity model)

Fig. A.9. Schematic presentation of the structural modelling of the bridge in the
RUAUMOKO program
The bridge was subjected to the design earthquake ground motion (as shown in Fig. A.6)
and was also subjected to 1.2 times the design earthquake. The resulting displacements
and shear forces of the columns are compared to those obtained in the tests as shown in
Fig. A.10 to Fig. A.13. The results indicate that the computer models can predict the
maximum displacements and shear forces reasonably well. The analytical models can be
refined even more by incorporating the stiffness of the foundations of the columns and by
refining the hysteresis parameters. Such refined models are out of the scope of this
research. The recommendations made by Priestley et al. (2007) for modelling bridge
structures provided sufficient accuracy in predicting the maximum displacement
demands. Therefore these modelling assumptions were adopted for the seismic analysis
and evaluation of the bridges in this research.

236

Short Column Displacement


25
20
Experimental
15

Numerical

Disp. (mm)

10
5
0
-5

0.5

1.5

-10

2.5

3.5

Time (Sec)

-15
-20
-25

Medium Column Displacement


20
Experimental
Numerical

15
10

Disp. (mm)

5
0
0

0.5

1.5

2.5

3.5

3.5

-5
-10
-15
-20
Time (Sec)

Tall Column Displacement


40
Experimental
Numerical

30

Disp. (mm)

20
10
0
0

0.5

1.5

2.5

-10
-20
-30
Time (Sec)

Fig. A.10. Comparison of the column displacements obtained using experimental data
and numerical analysis for the design earthquake
237

Short Column Force


1500
Experimental
Numerical

Shear (KN)

1000
500
0
0

0.5

1.5

2.5

3.5

-500

-1000
-1500
Time (Sec)

Medium Column Force


800
Experimental
Numerical

Shear (KN)

600
400
200
0

-200

0.5

1.5

2.5

3.5

-400
-600
-800
Time (Sec)

Tall Column Force


400
Experimental
Numerical

300

Shear (KN)

200
100
0
-100

0.5

1.5

2.5

3.5

-200
-300
-400
Time (Sec)

Fig. A.11. Comparison of the column shear forces obtained using experimental data and
numerical analysis for the design earthquake

238

Short Column Displacement


80
Experimental
Numerical

60

Disp. (mm)

40
20
0
0

0.5

1.5

-20

2.5

3.5

3.5

Time (Sec)

-40
-60
-80

Medium Column Displacement


40
Experimental
Numerical

30
20

Disp. (mm)

10
0
-10

0.5

1.5

-20

2.5

Time (Sec)

-30
-40
-50

Tall Column Displacement


80
60
Experimental

Disp. (mm)

40

Numerical

20
0
-20
-40

0.5

1.5

2.5

3.5

Time (Sec)

-60
-80

Fig. A.12. Comparison of the column displacements obtained using experimental data
and numerical analysis for maximum earthquake (i.e., 1.2 * design earthquake)

239

Short Column Force


1500

Shear (KN)

1000
500
0
0

0.5

1.5

2.5

3.5

2.5

3.5

2.5

3.5

-500
-1000
-1500
Time (Sec)

Medium Column Force

Shear (KN)

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

0.5

1.5

Time (Sec)

Tall Column Force


600

Shear (KN)

400
200
0
-200

0.5

1.5

-400
-600
Time (Sec)

Fig. A.13. Comparison of the column shear forces obtained using experimental data and
numerical analysis for maximum earthquake (i.e., 1.2 * design earthquake)

240

A.2. Shear deformations

The aspect ratios of columns considered in this research are larger than 2.8. When this
ratio is more than 2.5 the column behaviour is governed by flexure and shear
deformations are typically small. In addition, bridge columns were designed based on the
new seismic design philosophy including capacity design. Therefore the significant
amounts of confinement reinforcement required in columns will ensure a ductile flexural
collapse mode. This means that the shear capacity of the columns are always higher than
the corresponding shear developed in the columns due to formation of the plastic hinges
in the columns.

To further investigate this issue and also the effects of strength degradation of concrete
due to high flexural ductility demands in the plastic hinge regions, a study was carried out
using Response-2000 software (Bentz, 2001). For this purpose a column with an aspect
ratio of 2.5 was considered. This aspect ratio is equivalent to that of a 10 m long column
with diameter of 2.0 m which is fixed at the ends. The column contained 3% longitudinal
reinforcement and the corresponding required shear reinforcement (e.g., around 2%).

Fig.A.14. Cross section and properties of the column modelled in Response-2000


241

Fig. A.15. Example of the response of the column modelled in Response-2000

Fig. A.14 shows a sample model and Fig. A.15 demonstrates an example of the analysis
results obtained in Response-2000. As can be seen from Fig. A.15, the Response-2000
program gives a relationship for the shear force-total displacement and it does not directly
provide the shear force-shear displacement results. However the plots of the shear strain
over the length of the column are given by the program. Therefore shear strain could be
integrated over the length of the column to obtain the shear displacement corresponding
to each load stage. There are several load stages and hence, the diagrams should be
integrated for each load stage, in order to obtain the shear force shear displacement
relationship (Patwardhan, 2005). These integrations were performed for different load
stages and the results are shown in Fig. A.16. To include the effects of strength
degradation of concrete due to high flexural ductility demands, the tension stiffening
factor, ft, in Response-2000 program was varied between 1 to 0. The tension stiffening
effect represents the capacity of the intact concrete between cracks to continue to carry
tensile stresses and offer stiffness. The results indicate that no shear failure occurs and
that shear deformations are almost linear, before and after shear cracking, up to the
242

ultimate deformation. The results indicate that in the case of this column with a small
aspect ratio of 2.5, the shear deformations are less than 15% of the total deformation. The
contribution of shear deformations is even smaller in columns having larger aspect ratios.

7000
ft=1

6000

ft=0.5

5000
Shear force (KN)

ft=0.2
4000
3000

ft=0

2000
1000
0
0

6
8
10
Shear deformation (mm)

12

14

16

Fig. A.16. Shear force versus shear deformation for different tenstion stiffening factors,
ft.
Where column aspect ratios are smaller than 2.5, the shear deformations are important
and should be included in structural modelling. AASHTO-Guidelines (2009)
recommends using the effective shear area for ductile elements with aspect ratios lower
than 2.5 (i.e., pier walls) according to Eq. [A.1].

[A.1]

(GA )eff

Gc Av

E c I eff
Ec I g

Where (GA)eff is the effective shear stiffness, Ec is the modulus of elasticity of concrete,
Gc is the shear modulus of concrete, Av is the shear area, Ig is the gross moment of inertia,
and Ieff is the effective moment of inertia of the reinforced concrete cross section (i.e.,
EcIeff = Mn/ y).

243

However the use of this equation is not mandatory for the columns with larger aspect
ratios. This is because the shear deformations are not significant and are often negligible
for bridge columns with aspect ratios larger than 2.5. In this research Eq. [A.7.1] was
used to compute the shear stiffness of the columns. All bridge columns considered in this
research had aspect ratios larger than 2.5. Deformations due to bond slippage at the
column base were considered by incorporating the strain penetration length in plastic
hinge length and length of columns, based on the recommendations by Priestley et al.,
(1996 and 2007).

A.3. Effect of irregularity due to different column heights on the seismic response of
bridges (preliminary studies)
For the preliminary evaluations in this study and to demonstrate the research significance
of this subject, the seismic responses of an irregular and a regular bridge were studied.
The bridges are straight continuous four-span bridges with span lengths of 50 m as shown
in Fig. A.17.

The bridges were modelled in SAP2000 (Computers and Structures, 1999) as well as in
RUAUMOKO (Carr, 2009). The bridges were designed using the 2006 CHBDC
provisions in SAP2000 for the maximum zonal acceleration ratio of A=0.4 in the
CHBDC provisions. This maximum zonal acceleration was chosen to carry out the
preliminary studies. The importance factor of the bridges was considered as 1.0 and the
site coefficient of S=1.2 for soil profile type II was used. The super structure mass was
computed as 300 KN/m.
The seismic responses of bridges were evaluated using different methods of analysis
including elastic response spectrum analysis, pushover analysis with uniform load
pattern, modal pushover analysis and inelastic time history analysis. The bridges were
studied for two cases of free and restrained transverse movements at the abutments.
The elastic response spectrum analysis and the pushover analyses were performed using
SAP2000 software. The inelastic time history analysis was carried out using the

244

RUAUMOKO software. For the time-history analyses, 7 spectrum matched records were
used. The records were matched to the design spectrum from the 2006 CHBDC
provisions.

a)

50 m

50 m

50 m

2.4 m

2.4 m

50 m

2.4 m

b)

50 m

50 m

50 m

2.4 m

2.4 m

c)

50 m

2.4 m

d)

D= 2.4 m
Superstructure cross section

Fig. A.17. The properties of bridges studied: a) regular bridge, b) irregular bridge, c)
superstructure cross section and d) columns cross section

For the modal pushover analyses (MPA) the response of the bridges in each mode was
determined using the capacity-spectrum method as defined by the ATC-40 provisions
(ATC-40, 1996). The capacity spectum method concept is shown in Fig. A.18 and an
example of this method for the prediction of the seismic demands for two different modes
are shown in Fig. A.19. The maximum drift ratio in the most critical bridge column was
used to determine the responses using MPA as recommended by Kappos et al. (2004) and
Paraskeva et al., (2006).

MPA is based on two principal simplifications. First, The coupling among modes is
essentially neglected and second, it is assumed that the SRSS or CQC combination rules
are valid.
245

MPA can be performed in following steps :


Calculating the natural periods, Tn and modes, n
Performing separate pushover analyses for each mode using force distribution, sn
=mn (where m is the mass matrix)
Predicting the earthquake displacement demands for each mode (e.g. using the
capacity spectrum method )
Combining the peak modal responses using the SRSS or the CQC combination
rule.

Fig. A.18. Application of the capacity spectrum method to determine the seismic
demands (Paraskeva et al. (2006)).

a)

b)

Fig. A.19. Examples of capacity-spectrum method used in SAP2000 to perform modal


pushover analyses for: a) first mode (inelastic deformations); b) second mode (controlled
by elastic deformations)
246

The seismic response predictions for the regular bridge with restrained transverse
movements at the ends are shown in Fig. A.20 for different methods of analyses. As can
be seen, the results from the elastic response spectrum analysis, inelastic time history
analysis and the modal pushover analysis are in good agreement for this case. The results
indicate that the pushover analysis using the uniform load pattern underestimates the
maximum displacement demands. The similar results were observed for the case of the
regular bridge with free movements at the abutments as shown in Fig. A.21.
For the irregular bridge two different superstructure stiffnesses were considered in the
analyses. The superstructure transverse moment of inertia was considered as 140 m4 (i.e.,
rigid superstructure) and 70 m4 (i.e., flexible superstructure). The seismic response
predictions for the two cases of rigid and flexible superstructures are shown in Fig. A.22
and Fig. A.23, respectively. In the case of a rigid superstructure (Fig. A.22) the
predictions using different methods of analyses are in relatively good agreement,
although the overall displacement envelope patterns are somewhat different from the
different methods of analysis. On the other hand, in the case of a more flexible
superstructure (Fig. A.23) the predictions from the inelastic and elastic analyses are
significantly different. The displacement demands at the central column, which is the
most critical column that controls the bridge response for this case, is significantly
underestimated using the elastic response spectrum analysis method. Therefore using the
response spectrum analysis, which is permitted in the current seismic codes for the design
of such irregular bridges, can lead to unsafe designs. The predictions using the modal
pushover analysis in this case are in good agreement with those obtained using the
inelastic time history analysis.

The mode shapes of the irregular bridge with a flexible superstructure are compared with
the mode shape of the bridge deck in Fig. A.24. As can be seen the mode shapes of the
bridge resemble the displacement pattern obtained using the elastic analysis, while the
mode shape of the bridge deck somewhat resembles the displacement envelopes obtained
using the inelastic time history analysis. As the difference between the mode shapes of

247

the bridge and mode shapes of the deck increases, the results from the elastic analyses
differ more from the actual response of the bridge. That is why in most of the regularity
indices the difference between the mode shapes of the bridge and deck are used to
measure the degree of irregularity.

Similar results are observed for the case of the irregular bridge with free transverse
movements at the abutments as shown in Fig. A.25. The use of the elastic response
spectrum analysis underestimated the maximum displacement demand in the central
critical column. Even the use of the modal pushover analysis did not significantly
improve the predictions for this case.

Based on the preliminary results obtained, more research is needed to investigate the
effects of different column heights, superstructure stiffness, and position of the columns
on the seismic response of bridges. The comparisons of the inelastic and elastic responses
for a wide range of bridge structures with different configurations can provide useful
information to recognize the limitations of the elastic analyses and to improve the seismic
design codes to better address such problems for analysis and design of bridges.

Lateral Displacement (m)

Displacement shapes
0.35
0.3
0.25
0.2
0.15
0.1
0.05

Bridge length (m)

0
0

50
Uniform load

100
MPA

150
ITHA

200

ELASTIC

Fig. A.20. Seismic response predictions using different methods of analysis for the
regular bridge with restrained transverse movements at the abutments.

248

Lateral Displacement (m)

Displacement shapes
0.6
0.5
0.4
0.3
0.2
0.1
0

Bridge length (m)


0
50
Uniform load

100
MPA

ITHA

150
200
ELASTIC

Fig. A.21. Seismic response predictions using different methods of analysis for the
regular bridge with free transverse movements at the abutments.

Lateral Displacement (m)

Displacement shapes
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0
0

50
Uniform load

100
MPA X (m)
ITHA

150
ELASTIC

200

Fig. A.22. Seismic response predictions using different methods of analysis for the
irregular bridge with a rigid superstructure and with restrained transverse movements at
the abutments.

249

Lateral Displacement (m)

Displacement shapes
0.16
0.14
0.12
0.1
0.08
0.06
0.04
0.02
0

Bridge length (m)


0

50
Uniform load

100
MPA

ITHA

150
ELASTIC

200

Fig. A.23. Seismic response predictions using different methods of analysis for the
irregular bridge with a flexible superstructure and with restrained transverse movements
at the abutments.

Fig. A.24. Comparison of the mode shapes of an irregular bridge with the mode shapes of
the bridge deck for the first 3 modes.

250

Lateral Displacement (m)

Displacement shapes
0.8
0.6
0.4
0.2
0
0

50

100

Uniform Load

MPA

150
ITH

ELASTIC

Fig. A.25. Seismic response predictions using different methods of analysis for the
irregular bridge with a flexible superstructure and with free transverse movements at the
abutments.

251

200

Appendix B: Computer program for designing, modelling, running the


analyses, extracting and post processing the results
A comprehensive computer program was developed which is capable of designing and
evaluating the bridges by means of elastic dynamic and inelastic dynamic analysis. The
program is capable of automatically generating the input files and extracting all the
required information for the seismic evaluations (according to the research objectives)
using the RAUAMOKO program. Some of the main capabilities of the program are
briefly as follows:

Main outputs and options available in the program:


x

Computes the moment curvature and moment-axial force interaction diagrams

considering the confinement effect due to the transverse reinforcement for the core
concrete and considering unconfined concrete for the cover concrete.
x

Designs the bridge columns for flexure and shear according to the 2006 CHBDC

provisions by means of response spectrum analysis in the RAUAMOKO program and


updates the design by updating the effective stiffnesses in several trials.
x

Shear design is carried out based on the 2006 CHBDC provisions. The transverse

steel ratios are determined to satisfy all requirements of the 2006 CHBDC provisions
including the requirements for maximum factored shear forces, confinement in the plastic
hinge regions, and the capacity design philosophy.
x

Calculates the nodal points, elements, loads, masses, elastic and inelastic

properties of the columns and ductility capacities at different damage states for modelling
in the RAUAMOKO program automatically.
x

Extracts the modal properties (see Figs. B.1 and B.2) and calculates several

regularity indices to be used in the seismic evaluations. The program automatically


recognizes the mode number and the type of the mode (i.e., torsional, transverse,
longitudinal or vertical).

252

Compares the displacements, drifts, curvatures and ductility demands and also the

displacement envelopes from the elastic and inelastic time history analyses. The
displacement envelopes from the elastic and inelastic analyses will be plotted
automatically as shown in Fig. B.3. Displacements, drifts, ductility demands, and forces
are extracted for all nodal points and column elements. The moment and shear forces in
the superstructure elements are also extracted as shown in Figs.B4 and B.5.
x

Calculates the mean or median responses for the desired number of records

corresponding to two different levels of earthquakes for seismic performance evaluations.


x

The mean or median values can be computed using either the maximum absolute

values or the negative and positive displacement envelopes (see Priestley et al. (2007) for
more details) to compute the average response of the structures subjected to a number of
ground motion records.
x

Calculates and compares the displacements and curvature demands to the

corresponding ductility capacities calculated based on the maximum allowable strain in


concrete or steel bar for different levels of damage and performance objectives
corresponding to the earthquake level considered. The program also calculates the
demand to capacity ratios to evaluate the available safety margins. The ductility
capacities can be computed based on either theoretical or empirical methods in the
program as discussed in Chapter 2.
x

Performs incremental dynamic analysis (IDA) and evaluates the bridges based on

the ATC-63 procedure. IDA can be performed using several algorithms including, regular
IDA, time-effective IDA, fast IDA and also the method used by the ATC63 provisions.
Such algorithms are discussed in Chapters 2 and 6.
x

Summarises the results and develops of the IDA curves using linear interpolation.

Computes the percentiles (i.e., 16%, 50%, and 84% percentiles) and predicts the collapse
capacity of the structure for each record. Computes the capacity of the structure at
different limit states or damage states including yielding, serviceability, cover spalling,
bar buckling, bar fracture, and collapse. Different methods can be chosen in the program
to predict the engineering demand parameters at different damage states (e.g., the method
by Priestley et al, (2007) or the method by Berry and Eberhard (2007)). The median,
percentiles and standard deviation at different damage states are computed. The IDA

253

cures are developed for different engineering demand parameters such as displacements,
drifts, displacement ductility, curvature ductility, ductility demand to ductility capacity
(i.e., D/C ratios) and inelastic rotation. Such IDA curves are developed for each bridge
column separately and then the maximum drift ratio of all columns will be used to
compute the IDA curves to represent the overall response of the bridge including the
maximum response of all bridge columns.
x

Computes the fragility curves based on the IDA results. The fragility curves are

developed for various damage states (e.g., yielding, cover spalling, bar buckling and
collapse).
x

A large pool of records is incorporated in the program including the NGA databse

records, K-KIK net records for subduction events, Atkinson records compatible with the
NBCC spectra (Atkinson, 2009) and simulated records to match target spectra for
different hazard levels for the cities of Vancouver and Montreal. The program is capable
of scaling the records based on the method described by Atkinson (2009) to select the
best records for the structural analysis (i.e., the records with better matching to the target
spectrum and records that requires smaller scale factors) based on the fundamental period
of the structure. The number of records from different record sets can be selected based
on the seismic deaggregation results.
x

The structural analyses can be carried out for the transverse or longitudinal bridge

separately or the pairs of records can be applied in a 3D structural analysis. In the latter
case, the program is also able to obtain the maximum radial displacements during the
excitation.
x

Options are available for the number of spans, number of nodes for each span (to

capture the displacement of the superstructure), column heights for different columns,
span lengths, abutment conditions, dead loads (i.e., structural mass), material properties,
maximum and minimum reinforcement for design, concrete cover, bar diameters, strength
factors s, and c for design, and etc.
x

Options are available for force modification factors, R, in the longitudinal and

transverse direction separately.


x

Options are available to design the bridges based on either the NBCC or CHBDC

design spectra and for Vancouver or Montreal.

254

Options are available to compute the effective moment of inertia of columns

based on different methods. This includes the iterative approach in which the bridge is
designed in several trials. The column properties are updated in each trial and the
structural analyses are performed using the updated column properties and new design
forces are obtained until the required reinforcements are similar to those obtained in the
last step.
x

The abutment model by Aviram et al. (2008) is incorporated in the program to

compute the stiffness and strength properties of the abutment components for modelling
in the RUAUMOKO program.
x

The program carries out pushover analysis for the bridges in RUAUMOKO and

computes the yielding and ultimate displacements (based on the guidelines of the
ASCE/SEI 41-06 provisions (ASCE, 2006)) for the obtained pushover curves for
different columns and for different nodes as monitoring points for pushover. The ductility
capacity of the structures is then determined. The pushover curves are converted to the
Sa-Sd format as indicated in the ATC-40 (ATC, 1996) provisions.
x

The program computes the collapse capacity of the bridges using different sets of

records (e.g., three sets of records for crustal, interface and inslab earthquakes) and also
using different subset of records (e.g., 8 different subsets as described in the thesis). The
results can also be computed for a set of records with similar epsilon values to those from
the seismic deaggregation (i.e., epsilon-based method). The program computes the mean
epsilon values and percentages of contribution of different earthquake types at the
selected hazard level (which can be set to 10%, 2% and 0.5% probability of exceedance
in 50 years). The program performs regression analysis to compute the 1 factor and the
spectral shape factor (SSF). The IDA results then will be modified for the computed SSF
values. The regression analyses are performed for three different earthquake types and
different subsets of records separately. After computing the percentiles and the standard
deviation of the IDA results, the results from different earthquake types are combined
based on the seismic deaggregation results to compute the overall collapse capacity of the
structure. The collapse margin ratios, acceptable collapse margin rations, mean epsilon
values of the records are also computed by the program for this purpose.

255

Subroutines are incorporated in the program to automatically perform large

parametric studies using a large pool of records for inelastic time history analyses and for
incremental dynamic analyses as demonstrated in the papers presented in this thesis
which involved a large number of bridge structures and a large number of inelastic
dynamic analyses.
x

All important parameters from all of the analyses are extracted and stored in a file

for comparison between different cases. All the results obtained for each structure is also
stored in a file.

Mode shapes
1.5

Mode shape

0.5
Mode1
Mode2

0
0

25

50

75

100

125

150

175

200

Mode3
Mode4

-0.5

-1

-1.5

Length

Fig.B.1. Extraction of the mode shapes of the bridges

256

Displaced shapes
0.35
0.3

Displacement

0.25
0.2

Mode1
Mode2

0.15

Mode3
Mode4

0.1

SRSS

0.05
0
0

25

50

75

-0.05

100

125

150

175

200

Length

Fig.B.2. Modal displacements of the bridge

Displacement envelopes
300
ITHA (NBCC)
Response spectrum analysis (I=1.5)

Displacement (mm)

250

Response spectrum analysis (I=1.0)

200

150

100

50

0
0

25

50

75

100
Length (m)

125

150

175

200

Fig.B.3. Comparison of the displacement envelopes predicted using elastic response


spectrum and inelastic time history analyses.

257

Moment-Envelopes
180000
160000

Moment (KN.m)

140000
120000
100000
Moment (ITHA)
80000

Moment (RS)
Moment (RS)-Reduced

60000
40000
20000
0
0

25

50

75

100
125
Length (m)

150

175

200

Fig.B.4. Comparison of the superstructure moment forces predicted using elastic


response spectrum and inelastic time history analyses.

Shear-Envelopes
3500

3000

Shear (KN)

2500

2000
Shear (ITHA)
Shear (RS)

1500

Shear (RS)-Reduced
1000

500

0
0

25

50

75

100
125
Length (m)

150

175

200

Fig.B.5. Comparison of the superstructure shear forces predicted using elastic


response spectrum and inelastic time history analyses.

258

Appendix C: Probabilistic seismic hazard analysis and the program


developed for seismic deaggregation
The main objective of the probabilistic seismic hazard analysis (PSHA) is to estimate the
mean frequency per unit time or, alternatively, the probability in a given future time
period that a specified level of some ground motion parameter will be exceeded at a site
of interest. More formally, the PSHA methodology allows computation of the mean
annual frequency of exceedance, , of a ground motion with amplitude, y, based on the
aggregated hazard from N sources located at different distances and capable of generating
events of different magnitudes (Bazzurro and Cornell, 1999). The standard formulation of
probabilistic seismic hazard to calculate a frequency of exceedence, , of a ground motion
amplitude, y, is given in Eq. [C.1] (McGuire 1995):

[C.1]

J (y )

Q f
i

(m ) f R (r ) P [Y ! y | m , r ]dm dr

where i is the activity rate for source i (e.g., the mean annual rate of occurrence of
earthquakes generated by source i with magnitude greater than some specified lower
bound), fM(m) is the magnitude probability density function which describes the
frequency of occurrence of each event with magnitude m, and function fR(r) denotes the
probability density function of distances, from the site, of the locations of earthquakes,
given an earthquake occurring in a seismic source. P[Y>y| m, r] is the conditional
probability that the ground motion amplitude exceeds the value of y given the magnitude,
m, and distance, r. Assuming that ground motion amplitude y has a log-normal density
function, ln(y) has a normal distribution function. Therefore the probability density
function (pdf) of y, p(y) can be obtained using Eq. [C.2] and the cumulative distribution
function (CDF), P(y), can be obtained using Eq. [C.3].

259

2
1
e  (ln( y ) g ( m ,r )) /2V
V 2S

[C.2]

p(y )

[C.3]

P (Y ! y | m , r ) [1  )(

ln( y )  g (m , r )

)]

where g(m, r) and V are the mean and standard deviation of ln(y) respectively, given by
the attenuation relationship (i.e., ground motion prediction equation (GMPE) ) .

The probability in the integrand of Eq. [C.1] can be expressed explicitly as a function of
the ground-motion randomness , given (McGuire, 1995):

[C.4]

J (y )

Q f
i

(m ) f R (r )f H (H ) P [Y ! y | m , r , H ]dm dr d H

[C.5]

ln( y )  g (m , r )

where  is defined as the number of logarithmic standard deviations by which the


logarithmic ground motion deviates from the median (Eq. [C.5]) and f() =
(1/(2S))exp(-2/2) represents the standardized Normal distribution. In this formulation,
the probability P[Y>y| m, r, ] is simply an indicator function which is zero if ln g(m, r, )
from the attenuation function (i.e., GMPE) is less than ln( y), and 1 otherwise.

The physical image of earthquake scenarios in terms of magnitude and distance is lost in
the highly integrated framework of the PSHA. For physical interpretation of the results
from PSHA it is desirable to have the contribution of each earthquake scenarios with
magnitude, M and distance, R to the overall seismic hazard obtained through the PSHA
method. This can be achieved through the deaggregation of the probabilistic seismic
hazard. The deaggregation of the hazard results by magnitude M, distance R, and groundmotion deviation  is achieved by accumulating (by magnitude, distance, and ) the
annual frequencies of exceedence of the target ground-motion amplitude for each period,

260

T, separately. Dividing these annual frequencies by the total hazard (the total annual
frequency) gives the probability that, given an exceedence of that amplitude, it has been
caused by a certain combination of M, R, and  (McGuire, 1995).
For seismic deaggregations Eq. [C.6] is used instead of the indicator function (as
discussed before), in order to derive the contributions to (y) by M, R, and :

[C.6]

P [Y ! y | m , r , H ] G [lnY (m , r , H )  ln y ]

where is the Dirac delta function, which gives a probability of 1 at In Y(m, r, ) = In(y)
and zero otherwise ( is used because the determination of M-R-  sets that equal the
target ground motion, not that exceed it, is sought). The formulations in Eqs. [C4] and
[C.6] are used for deaggregation to derive the contributions to (y) by m, r, and 
(McGuire,1995).

Program developed for the seismic deaggregation (Deaggregation calculator)

The probabilistic seismic hazard analyses (PSHA) for Vancouver and Montreal was
carried out based on the updated seismic hazard model using the Monte Carlo simulations
by Goda et al., (2010). For the purpose of this study, a program was developed to process
the raw seismic hazard data provided by Goda et al., (2010).

The computer program was developed in visual basic to process the raw data and to
compute the required information for deaggregation of the seismic hazard, mean
magnitude, distance, and epsilon, contribution of different event types (i.e., crustal,
interface, and inslab events).

261

The program is capable of performing the following tasks:

Inputs:
x

Probability of exceedance: Any probability of exceedance, up to 0.5% in 50 years,

can be considered.
x

Bin sizes: The bin sizes for magnitude, M, Distance, R, and epsilon, , are given.

The range of minimum to maximum desired magnitude, distance, and epsilon for

deaggregations
x

Site of interest: e.g., Vancouver or Montreal

Data type: Data using an updated model by Goda et al. (2010) or data from the

Geology Survey of Canada (GSC)) model can be used for the seismic deaggregation.
x

Deaggregation type: Either matching method or exceeding method (Hong and

Goda, 2006) can be used for the seismic deaggregation. If the matching method is chosen,
the percentages above and below of the target value to match should be defined (e.g.,
10% was considered for this study).
x

Type of distance: Either hypocentral or extended distances (the distance type used

in the GMPE (e.g., RRUP, RJB, etc.)) can be used.


x

Vibration periods: The deaggregation can be carried out for spectral accelerations

at different periods including PGA, Sa(0.2), Sa(0.5), Sa(1.0), Sa(2.0), and Sa(3.0).
x

Event types: Four options in this case are available. The user can choose either

any earthquake type (i.e., crustal, interface, and inslab) or all earthquake types. If
individual earthquake type is chosen then the seismic deaggregation will be carried out
only for the selected earthquake type. For example if the interface events are chosen, the
mean magnitude, distance and epsilon only for this event type will be computed. If all
earthquake types are chosen then the average mean magnitude, distance and epsilon for
all earthquake types will be computed. Further, in the later case the percentage of
contribution of each event type to the overall seismic hazard at different vibration periods
can be computed.

262

Selected GMPE: An option is available to choose the desired ground motion

prediction equations (GMPE) to carry out the seismic deaggregation. Any number of
GMPEs can be selected or alternatively all GMPEs used in the seismic hazard analyses
can be used. For example the epsilon values from different GMPEs are somewhat
different and specific GMPEs can be selected to derive epsilon values to be used in the
epsilon-based record selection or prediction of the conditional mean spectrum.

Out puts:
The main output of the program is:
x

Computing the seismic deaggregation for magnitude, distance and epsilon (i.e.,

the relative frequencies of each bin with specific M,R, and ).
x

Prediction of the mean values: The mean values of magnitude, M, distance, R,

and epsilon, , for the input variables discussed before will be computed. In addition, the
mode values of magnitude, distance and epsilon can be computed. The computation of
mode values can be based on either the marginal probability functions of M, R, and 
separately, or they can be computed for the joint probabilities of M-R or M-R-.
However the mode values were not required for the purpose of this study and were not
reported.
x

Prediction of the uniform hazard spectrum (UHS) : The UHS values for different

periods and different probability of exceedances can be computed.


x

Prediction of the percentages of contribution of each earthquake type on the

overall seismic hazard


x

Auto computation of the seismic deaggregation for different periods and different

hazard levels: A subroutine was added to compute the deaggregations for a full range of
periods and for different hazard levels and to print the results in the tables automatically.
x

Development of the conditional mean spectrum (CMS): The CMS at any desired

period for different earthquake types and different period ranges can be computed. For
periods other than those used in the seismic hazard analysis (i.e., PGA, T=0.2, 0.5, 1, 2,
3) seismic deaggregation will be carried out two times for the lower and upper periods

263

and a linear interpolation is used to predict the deaggregation results for the period of
interest. The computation of the CMS is carried out by computing a CMS for each bin
and considering the relative frequency of each bin. The overall CMS will be computed by
aggregating the CMS of each bin weighted by the relative frequency of that bin from the
seismic deaggregation results. For the purpose of this study the CMS was calculated
using certain GMPEs (i.e., BA08, Z06 and AB03). Although the other GMPEs are also
incorporated in the program and the weighting factors can be assigned to compute the
average CMS computed for different GMPEs, for the purpose of this study only the
BA08, Z06 and AB03 GMPEs were fully verified and used.

Fig. C.1. A view of the program developed for deaggregation of the seismic hazard data

264

Mean
D
69.15
72.43
66.40
67.76
66.16
65.74

1.29
1.68
1.64
2.14
1.94
2.51

Contribution %
Crustal Interface Inslab Sa (UHS)
33.6% 6.9% 59.5%
0.24
12.3% 3.7% 84.0%
0.35
32.0% 5.4% 62.6%
0.49
8.1%
1.9% 90.0%
0.68
30.3% 4.2% 65.5%
0.79
4.4%
0.6% 95.0%
1.07

10% in Updated
50
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC

M
7.10
6.62
7.33
6.70
7.43
6.71

Mean
D
74.77
77.38
73.50
71.76
70.20
68.13

epsilon
1.28
1.54
1.63
2.00
1.90
2.41

Contribution %
Crustal Interface Inslab Sa (UHS)
43.5% 20.9% 35.5%
0.34
16.9% 7.8% 75.3%
0.47
42.7% 24.8% 32.6%
0.75
14.5% 5.7% 79.9%
0.96
48.7% 23.8% 27.5%
1.31
12.2% 3.0% 84.8%
1.55

M
7.31
6.79
7.56
6.88
7.64
6.91

Mean
D
83.91
81.52
83.70
73.55
80.78
65.46

epsilon
1.21
1.50
1.58
1.93
1.91
2.24

Contribution %
Crustal Interface Inslab Sa (UHS)
40.4% 27.0% 32.6%
0.18
23.0% 10.4% 66.7%
0.23
37.9% 34.2% 28.0%
0.43
22.7% 9.5% 67.8%
0.45
44.5% 34.5% 21.0%
0.75
24.2% 6.6% 69.3%
0.73

M
7.52
6.86
7.80
6.99
7.96
7.07

Mean
D
94.37
83.66
96.85
74.02
98.48
66.41

epsilon
1.13
1.57
1.52
1.94
1.84
2.22

Contribution %
Crustal Interface Inslab Sa (UHS)
35.7% 35.3% 29.1%
0.09
37.2% 10.2% 52.5%
0.11
30.0% 46.5% 23.4%
0.23
32.8% 12.2% 55.0%
0.22
30.7% 51.3% 18.0%
0.41
34.3% 10.0% 55.7%
0.36

M
7.62
6.77
7.95
6.93
8.05
6.97

Mean
D
99.23
98.53
102.60
94.77
100.17
91.48

epsilon
1.07
1.57
1.46
2.06
1.76
2.46

Contribution %
Crustal Interface Inslab Sa (UHS)
37.2% 39.6% 23.3%
0.05
0.0%
16.0% 84.0%
0.04
30.5% 53.8% 15.6%
0.13
0.0%
18.2% 81.8%
0.09
36.7% 55.3%
8.0%
0.25
0.0%
16.8% 83.2%
0.17

Sa(0.5)
10% in Updated
GSC
50
Updated
2% in 50
GSC
0.5% in Updated
50
GSC

Sa(1)
10% in Updated
GSC
50
Updated
2% in 50
GSC
0.5% in Updated
50
GSC

Sa(2)
10% in Updated
GSC
50
Updated
2% in 50
GSC
0.5% in Updated
GSC
50

Sa(3)
10% in Updated
50
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC

265

Table C.1. Deaggregation results for all earthquake types using all GMPEs using exceeding method

M
6.78
6.29
6.96
6.34
7.06
6.38

PGA

10% in Updated
50
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC

D
70.88
75.66
67.61
68.41
64.51
68.32

1.07
1.38
1.40
1.85
1.66
2.27

Crustal Interface
36.4%
7.7%
16.0%
4.7%
33.3%
6.5%
10.9%
2.4%
30.8%
4.0%
6.7%
0.9%

M
6.89
6.54
7.21
6.68
7.43
6.73

Mean
D
75.09
82.04
73.80
75.79
77.80
71.25

epsilon
1.07
1.28
1.45
1.73
1.68
2.11

Contribution %
Crustal Interface Inslab Sa (UHS)
48.1% 15.8% 36.1%
0.34
20.4%
8.2%
71.4%
0.47
40.9% 23.0% 36.1%
0.75
13.8%
8.5%
77.7%
0.96
38.9% 27.4% 33.7%
1.31
13.5%
5.3%
81.3%
1.55

M
7.08
6.71
7.47
6.86
7.63
6.90

Mean
D
84.03
86.29
83.90
77.42
85.07
72.26

epsilon
1.03
1.24
1.33
1.63
1.64
2.01

Contribution %
Crustal Interface Inslab Sa (UHS)
47.2% 19.4% 33.5%
0.18
25.9% 10.0% 64.1%
0.23
37.9% 33.2% 28.9%
0.43
23.4% 10.9% 65.7%
0.45
36.1% 34.3% 29.6%
0.75
21.3%
9.0%
69.8%
0.73

M
7.23
6.77
7.73
6.91
7.86
7.04

Mean
D
91.93
89.38
98.23
79.10
96.96
71.20

epsilon
0.93
1.33
1.30
1.71
1.62
2.03

Contribution %
Crustal Interface Inslab Sa (UHS)
44.8% 22.9% 32.3%
0.09
41.2%
9.2%
49.6%
0.11
29.1% 45.4% 25.5%
0.23
34.3% 10.5% 55.2%
0.22
28.6% 48.4% 23.0%
0.41
36.7% 12.2% 51.1%
0.36

M
7.27
6.64
7.85
6.86
7.95
6.93

Mean
D
94.92
101.24
104.58
97.15
99.00
93.62

epsilon
0.92
1.25
1.26
1.79
1.55
2.21

Sa(0.5)
10% in Updated
GSC
50
Updated
2% in 50
GSC
0.5% in Updated
GSC
50

Sa(1)
10% in Updated
50
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC

Sa(2)
10% in Updated
50
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC

Sa(3)
10% in
50

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
GSC
50

Contribution %

M
6.64
6.24
6.88
6.29
6.99
6.37

Inslab Sa (UHS)
55.8%
0.24
79.3%
0.35
60.2%
0.49
86.6%
0.68
65.2%
0.79
92.4%
1.07

Contribution %
Crustal Interface Inslab Sa (UHS)
47.2%
23.1%
29.6%
0.05
0.0%
13.3%
86.7%
0.04
29.4%
50.2%
20.4%
0.13
0.0%
17.3%
82.7%
0.09
35.6%
54.2%
10.3%
0.25
0.0%
18.4%
81.6%
0.17

266

Table C.2. Deaggregation results for all earthquake types using all GMPEs using matching method

Mean

PGA

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
GSC
50
10% in
50

Sa(0.5)
Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC
10% in
50

Sa(1)
10% in
50

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC
Sa(2)
10% in
50

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC
Sa(3)
10% in
50

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC

Mean
M
D
6.41
16.04
6.43
25.77
6.74
8.32
6.59
15.52
6.96
4.67
7.03
13.41

Mean 
All GMPEs BA08 GMPE
1.22
1.34
1.40
1.37
1.48
1.71
1.58
1.74
1.89
-

Mean
M
D
6.58
24.57
6.75
34.70
6.83
12.95
6.90
20.38
7.04
8.11
7.01
11.76

Mean 
All GMPEs BA08 GMPE
1.26
1.35
1.24
1.38
1.48
1.37
1.54
1.64
1.52
-

Mean
M
D
6.76
30.71
6.96
51.17
6.95
14.86
7.07
33.32
7.10
8.96
7.11
19.65

Mean 
All GMPEs BA08 GMPE
1.31
1.41
1.31
1.43
1.56
1.49
1.65
1.79
1.59
-

Mean
Mean 
M
D
All GMPEs BA08 GMPE
6.90
35.96
1.32
1.47
6.92
63.45
1.39
7.07
15.71
1.42
1.59
7.03
39.29
1.57
7.19
7.66
1.55
1.80
7.14
29.33
1.79
Mean
M
D
6.97
40.55
7.14
18.38
7.24
9.86
-

Mean 
All GMPEs BA08 GMPE
1.31
1.39
1.39
1.57
1.45
1.47
-

267

Table C.3. Deaggregation results for only crustal events using all GMPEs and BA08 GMPE using exceeding method

PGA

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
GSC
50
10% in
50

Sa(0.5)
Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC
10% in
50

Sa(1)
10% in
50

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC
Sa(2)
10% in
50

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC
Sa(3)
10% in
50

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC

Mean
M
D
6.23
21.80
6.34
34.42
6.59
12.00
6.48
17.70
6.82
5.57
6.82
14.55

Mean 
All GMPEs BA08 GMPE
1.12
1.22
1.29
1.27
1.40
1.58
1.24
1.40
1.74
-

Mean
M
D
6.43
33.59
6.66
48.19
6.69
19.31
6.79
24.66
6.92
10.50
6.92
15.55

Mean 
All GMPEs BA08 GMPE
1.19
1.25
1.24
1.38
1.53
1.30
1.30
1.49
1.29
-

Mean
M
D
6.65
43.91
6.87
64.27
6.85
21.58
6.99
40.30
7.09
13.26
7.12
30.11

Mean 
All GMPEs BA08 GMPE
1.26
1.31
1.21
1.30
1.47
1.31
1.41
1.66
1.47
-

Mean
Mean 
M
D
All GMPEs BA08 GMPE
6.78
50.18
1.26
1.29
6.83
76.57
1.23
6.98
24.94
1.41
1.60
6.97
50.70
1.45
7.17
11.60
1.33
1.71
7.09
38.88
1.72
Mean
M
D
6.87
56.10
7.09
31.48
7.10
14.17
-

Mean 
All GMPEs BA08 GMPE
1.24
1.26
1.44
1.51
1.44
1.72
-

268

Table C.4. Deaggregation results for only crustal events using all GMPEs and BA08 GMPE using matching method

PGA

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
GSC
50
10% in
50

Sa(0.5)
Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC
10% in
50

Sa(1)
10% in
50

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC
Sa(2)
10% in
50

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC
Sa(3)
10% in
50

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC

Mean
Mean 
M
D
All GMPEs AB03 GMPE Z06 GMPE
8.58
141.42
1.55
2.21
1.40
8.15
142.50
1.77
8.64
141.09
2.25
3.15
2.18
8.15
142.50
2.51
8.76
140.12
2.71
2.67
8.15
142.50
3.05
-

Mean
Mean 
M
D
All GMPEs AB03 GMPE Z06 GMPE
8.56
141.84
1.11
1.11
0.99
8.15
142.50
1.39
8.63
142.23
1.83
2.10
1.66
8.15
142.50
2.15
8.64
141.11
2.40
2.81
2.18
8.15
142.50
2.74
-

Mean
Mean 
M
D
All GMPEs AB03 GMPE Z06 GMPE
8.56
141.97
0.94
0.94
0.72
8.15
142.50
1.29
8.62
141.32
1.60
1.76
1.31
8.15
142.50
1.98
8.67
140.62
2.08
2.33
1.79
8.15
142.50
2.53
Mean
Mean 
M
D
All GMPEs AB03 GMPE Z06 GMPE
8.55
142.02
0.80
0.79
0.59
8.15
142.50
1.35
8.61
141.99
1.47
1.59
1.17
8.15
142.50
2.04
8.67
140.81
1.90
2.21
1.63
8.15
142.50
2.54
Mean
Mean 
M
D
All GMPEs AB03 GMPE Z06 GMPE
8.55
142.07
0.68
1.11
0.46
8.63
141.68
1.41
2.02
1.06
8.70
141.47
1.89
2.76
1.56
-

269

Table C.5. Deaggregation results for only interface events using all GMPEs, AB03 and Z06 GMPE using exceeding method

PGA

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
GSC
50
10% in
50

Sa(0.5)
Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC
10% in
50

Sa(1)
10% in
50

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC
Sa(2)
10% in
50

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC
Sa(3)
10% in
50

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC

Mean
Mean 
M
D
All GMPEs AB03 GMPE Z06 GMPE
8.56
142.42
1.17
1.77
0.87
8.15
142.50
1.37
8.61
140.15
1.93
3.15
1.77
8.15
142.50
2.24
8.73
142.88
2.41
2.31
8.15
142.50
2.88
-

Mean
Mean 
M
D
All GMPEs AB03 GMPE Z06 GMPE
8.53
141.98
0.55
0.42
0.33
8.15
142.50
0.90
8.59
142.62
1.37
1.47
1.18
8.15
142.50
1.84
8.62
141.89
2.00
2.26
1.79
8.15
142.50
2.41
-

Mean
Mean 
M
D
All GMPEs AB03 GMPE Z06 GMPE
8.52
142.66
0.33
0.20
0.01
8.15
142.50
0.81
8.59
142.08
1.21
1.19
0.82
8.15
142.50
1.63
8.64
141.65
1.71
1.94
1.38
8.15
142.50
2.25
Mean
Mean 
M
D
All GMPEs AB03 GMPE Z06 GMPE
8.49
142.12
0.10
-0.03
-0.25
8.15
142.50
0.91
8.56
142.34
1.13
1.10
0.71
8.15
142.50
1.62
8.61
141.43
1.62
1.84
1.24
8.15
142.50
2.29
Mean
Mean 
M
D
All GMPEs AB03 GMPE Z06 GMPE
8.47
142.49
-0.01
0.48
-0.26
8.15
142.50
0.65
8.57
142.11
1.01
1.60
0.58
8.15
142.50
1.44
8.64
142.61
1.53
2.33
1.15
8.15
142.50
1.97
-

270

Table C.6. Deaggregation results for only interface events using all GMPEs, AB03 and Z06 GMPE using matching method

PGA

10% in Updated
GSC
50
Updated
2% in 50
GSC
0.5% in Updated
GSC
50

Mean
M
D
6.77
68.29
6.18
75.59
6.93
61.71
6.28
70.39
7.00
57.23
6.34
67.43

Mean 
All GMPEs AB03 GMPE Z06 GMPE
1.29
1.20
1.32
1.71
1.72
1.74
1.67
2.17
2.06
2.21
1.96
2.54
-

Mean

Mean 
All GMPEs AB03 GMPE Z06 GMPE
1.39
1.28
1.43
1.63
1.79
1.76
1.78
2.11
2.09
2.20
2.03
2.53
-

Sa(0.5)
Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
GSC
50
10% in
50

M
6.86
6.43
6.99
6.56
7.09
6.61

Mean

Sa(1)
10% in
50

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
GSC
50

M
6.95
6.53
7.08
6.64
7.12
6.72

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
GSC
50

M
7.03
6.57
7.14
6.70
7.23
6.83

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC

D
67.61
85.00
57.93
77.72
54.67
73.38
Mean

Sa(3)
10% in
50

D
63.76
81.59
55.51
76.08
49.93
72.34

Mean

Sa(2)
10% in
50

D
63.38
79.51
55.24
75.21
51.23
72.78

M
7.07
6.51
7.20
6.66
7.28
6.73

D
69.29
90.13
57.59
84.14
55.63
81.21

Mean 
All GMPEs AB03 GMPE Z06 GMPE
1.30
1.14
1.49
1.60
1.75
1.62
1.97
2.07
2.17
1.98
2.45
2.44
Mean 
All GMPEs AB03 GMPE Z06 GMPE
1.29
1.18
1.59
1.75
1.75
1.68
2.06
2.13
2.16
2.05
2.99
2.43
Mean 
All GMPEs AB03 GMPE Z06 GMPE
1.36
1.22
1.67
1.64
1.76
1.65
2.06
2.12
2.30
2.24
2.47
2.52
-

271

Table C.7. Deaggregation results for only inslab events using all GMPEs, AB03 and Z06 GMPE using exceeding method

PGA

10% in Updated
GSC
50
Updated
2% in 50
GSC
0.5% in Updated
GSC
50

Mean
M
D
6.64
73.47
6.11
79.37
6.86
65.67
6.22
72.01
6.96
60.47
6.32
71.05

Mean 
All GMPEs AB03 GMPE Z06 GMPE
1.03
0.88
1.12
1.40
1.41
1.34
1.41
1.88
1.81
1.88
1.76
2.30
-

Mean

Mean 
All GMPEs AB03 GMPE Z06 GMPE
1.13
1.01
1.17
1.33
1.59
1.57
1.55
1.80
1.85
1.83
1.82
2.23
-

Sa(0.5)
10% in
50

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC

M
6.78
6.32
6.91
6.50
7.04
6.60

Mean

Sa(1)
10% in
50

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
GSC
50

M
6.84
6.42
7.00
6.60
7.12
6.67

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
GSC
50

M
6.95
6.46
7.10
6.64
7.13
6.73

Updated
GSC
Updated
2% in 50
GSC
0.5% in Updated
50
GSC

D
75.42
88.82
61.72
83.12
56.29
75.10
Mean

Sa(3)
10% in
50

D
70.17
85.63
57.61
78.71
53.12
74.77

Mean

Sa(2)
10% in
50

D
70.60
84.01
58.36
76.81
52.94
74.95

M
6.98
7.15
7.20
-

D
75.55
63.71
51.35
-

Mean 
All GMPEs AB03 GMPE Z06 GMPE
1.10
0.95
1.23
1.32
1.53
1.34
1.77
1.74
1.85
1.82
1.85
2.15
Mean 
All GMPEs AB03 GMPE Z06 GMPE
1.06
0.91
1.40
1.48
1.45
1.34
1.78
1.89
1.99
1.79
2.69
2.19
Mean 
All GMPEs AB03 GMPE Z06 GMPE
1.13
1.00
1.41
1.60
1.47
2.06
2.02
1.90
2.17
-

272

Table C.8. Deaggregation results for only inslab events using all GMPEs, AB03 and Z06 GMPE using matching method

PGA

Appendix D: Computer program for different record selection methods


and predictions of CMS
A computer program was developed which is capable of selecting ground motion records
using several record selection methodologies. The information for a large pool of records,
compiled by combining the records from different sources including PEER-NGA, KNET, KiK-NET, and COSMOS databases, was incorporated in the program. The records
from three different earthquake types (i.e., crustal, interface, and inslab) are included.

The program inputs and options:


x

Target period: The target period, T1, is typically the fundamental period of the

structure.
x

Period range to match: If UHS-based or CMS-based record selection methods are

selected, the minimum and maximum range of periods for spectrum matching should be
assigned. This is typically taken as 0.2T1 to 2T1.
x

Number of records: The number of required records to be selected are given

Record type: Four options for record types are available including crustal,

interface, inslab, and all types. For example if inslab is chosen, the program only
selects the inslab records in the database and if all is selected, the program search all
the database without any limitations on the earthquake type.
x

Scaling option: The records can be selected with or without scaling. If scaling is

enabled, a better matching will typically obtained and if scaling is disabled only original
records will be used to match the target spectrum.
x

Response matching option: An option is available to choose whether the

geometric means of the pairs of the records spectra or the individual record spectra are
used for spectrum matching.
x

Database option: The data base can also be selected to determine whether the

records should be searched within a specific database or all databases should be searched
for record selection.
273

Ranges of record parameters: Limitations on the minimum and maximum range

of different record parameters including record magnitude, distance, shear wave velocity
(Vs30), peak ground acceleration (PGA) and peak ground velocity (PGV) can be imposed.
x

Based on the record selection method to be used, the results from the seismic

deaggregation (which is computed using the deaggregation calculator program described


in Appendix C) are given to the program. Such results include the UHS or CMS
(including CMS-crustal, CMS-interface, CMS-inslab, and CMS-All), the contribution of
each event types to the overall seismic hazard, the mean magnitude, distance and epsilon,
for different event types.
The main functions of the program are summarized as follows:

Record selection methodologies:


The program can choose the desired number of records based on different methodologies
including:

a) UHS-based and CMS-based record selection:

The uniform hazard spectrum (UHS) and conditional mean spectrum (CMS) are
computed using the deaggreagtion calculator program (Appendix C) and the information
will be given to this program. The records that most closely match the defined spectra
over a range of periods (which is defined for the program) will be selected. The
information of these records in addition to the response spectra and the average response
spectra of the records will be determined.

b) Epsilon-based record selection:


The program can compute the epsilon values of the records at different periods using
different GMPEs which will be used for the epsilon-based record selection.

c) MR-based record selection:


In this method the records with the closest magnitude and distance to the mean magnitude
and distance from the seismic deaggregation will be selected.
274

d) All-based and event-based methods:

For record selection methods discussed, the selection can either be carried out using Allbased procedure or Event-based procedure. For the Event-based procedure the
number of records from each earthquake type will be proportional to the percentage of
contribution of that event type to the overall seismic hazard and thus the record selection
will be carried out for each event types separately and the required number of records will
be selected for each earthquake types. In All-based procedure no limitation on the type
of record is imposed and the records from all earthquake types can be selected.

Other outputs of the program:

a) The list of selected records and detailed information of records:


A list of the required number of records that match the criteria will be determined along
with all required record data. The average response spectrum of the records will be
calculated and the response spectrum of individual records will also be printed. If the
scale factors are allowed the required scale factor for each record will be printed. In the
epsilon-based record method the epsilon values of the records at the fundamental period
and the average epsilon value of the record sets will be computed. The sum of the square
errors (SSE) which denotes the error in matching will be computed for each record and
the records will be sorted in the order of lowest to largest SSE.

b) Response spectrum of records:


A subroutine is incorporated which computes the elastic response spectra of the records
for the desired period ranges and damping ratios.

c) Epsilon values of the records:


The epsilon values will be computed for each record at different vibration periods. The
ground motion prediction equations (GMPE) specific for each event type are used to
compute the epsilon values.

275

d) Changing the format of the records:


Some subroutines were incorporated to read the earthquake record files with different
formats and convert it to the desirable formats or the formats compatible with the
RUAUMOKO program. The format of records here is referred to the way the
acceleration histories and time are written in the ground motion record file (e.g., Number
of columns in each row, whether time steps are included or not, etc.)

e) Output for the RUAUMOKO program:


The selected records data will also be printed out in a format compatible with the
RUAUMOKO program which will be used in the main program (Appendix B) for the
structural analyses.

276

Appendix E: Dynaplot helper


After performing the nonlinear analyses using the RUAUMOKO program, it is very
difficult to view the details of the analyses such as hysteresis loops of elements,
displacement and force histories. To view such results from RUAUMOKO program one
should use a program called Dynaplot (Carr, 2009). To do this the information of the
structures and elements should be given to the program one by one which is very time
consuming. Moreover, if any mistake occurs all the information should be given from the
beginning. The use of such methods to control the accuracy of the structural analyses and
hysteresis loops of elements can be extremely time consuming.
To facilitate the inspection of various parameters from the structural analyses, a computer
code was developed which creates all the required input files for the Dynaplot software,
execute the program, extract the results and show the results along with the graphs in
excel sheets. Therefore the required analysis details can be viewed only by pressing the
execute bottom in the program.

Fig. F.1. A view of the Dynaplot-helper program

277

Appendix F: Different earthquake types


In common practice for seismic assessment of the structures crustal earthquake records,
such as those available in PEER-NGA database, are mainly used and little attention has
been given to other earthquake types that can play an important role in seismic
performance assessment of the structures located near subduction zones such as the
Cascadia subduction zone. The Cascadia subduction zone extends from northern
Vancouver island to northern California and separates the Juan de Fuca tectonic plate and
the North America plates (see Fig. F.1). Major cities affected by this subduction zone
include Vancouver and Victoria, British Columbia, Seattle, Washington, Portland,
Oregon, and Sacramento California.

Seismic hazard for the Strait of Georgia region of British Columbia (including
Vancouver, Victoria and a substantial fraction of B.C.s population) comes from three
sources of earthquakes as shown in Fig. F.2. These sources include the crustal seismicity
in the North American plate, great earthquakes of the Cascadia subduction zone on the
interface between the North American and subducting Juan de Fuca plate(i.e., subduction
interface or subduction thrust earthquakes) and deep earthquakes within the subducting
slab (i.e., inslab earthquakes). Therefore, in addition to the common crustal earthquakes,
for a site located in Vancouver at least two other earthquake types resulting from the
Cascadia subduction zone (i.e., interface and inslab earthquakes) should be considered in
seismic hazard evaluations, structural analyses, and risk assessments.

278

Fig. F.1. Juan de Fuca plate and Cascadia subduction zone (from Hyndman and Rogers,
2010)

Fig. F.2. Three earthquake types of the Cascadia subduction zone (Hyndman and Rogers,
2010)

279

Interface earthquakes (also known as Megathrust and interplate earthquakes)

Interface (megathrust) earthquakes occur at subduction zones at destructive plate


boundaries where one tectonic plate is forced under (subducts) another (at the interface of
the tectonic plates) as shown in Fig. F.2 and Fig. F.3. The interface earthquakes are
among the largest earthquakes possible in the world. The moment magnitudes of such
earthquakes can exceed 9.0. All of the past earthquakes in the globe with magnitude of
9.0 and greater have been interface earthquakes and no other known tectonic activity can
produce earthquakes with such large magnitudes.

It is recognized that the coast of British Columbia and the US pacific northwest has
substantial earthquake hazard however until the mid 1980s the risk of great subduction
interface (mega thrust) events was not appreciated, because no great earthquakes of this
type was available in the written historical record. The evidence that there is indeed
active subduction under thrusting was summarized by Riddihough and Hyndman (1976)
and the case for great earthquakes was made more than 20 years ago (e.g, Heaton and
Kanamori 1984; Heaton and Hartzell 1987; Rogers 1988). Although supporting data were
limited, the accumulated evidence has left little doubt that great earthquakes and
associated tsunamis have occurred many times in the past and that they will occur and
produce damage in the future (Hyndman and Rogers, 2010).

Very few earthquakes of any size have been detected on the Cascadia subduction thrust.
In general the lack of any subduction interface (thrust) earthquakes is not common. Most
of the worlds great earthquakes with magnitude of M>8 have occurred on subduction
zone thrust faults, and most subduction zones have experienced historical great
earthquakes. However some subduction zones that have had the largest earthquakes (e.g.,
M=9 Sumatra in 2004, Alaska in 1964, Chile in 1960, and Kamchatka in 1952) also have
had very long time intervals between the events, while just a few small interface
earthquakes occurred between these intervals.

280

For Cascadia, the written historical record is short, with only a little more than 200 years.
This limited written history is in clear contrast to the detailed Japanese record of great
subduction zone earthquakes and tsunami waves that extends back to the 7th century.

Fig. F.3. Schematic cross-section of the Cascadia subduction and megathrust (interface)
earthquake zone and the epicentres of some larger historical earthquakes (Hyndman and
Rogers, 2010).

There are three possible explanations for the lack of historical Cascadia great
earthquakes:

(1) the Juan de Fuca plate is no longer converging and underthrusting North America
(2) Underthrusting is continuing, but it is accommodated by smooth stable sliding
(3) The thrust fault is completely locked with not enough motion to generate even
small earthquakes.

281

Unlike the first two options, the third option implies that there is a potential for very large
and damaging earthquakes that has not been included in hazard estimates until recently.
Riddihough and Hyndman (1976) compiled a variety of evidence against the fist option
(i.e., no convergence) indicating that there is ongoing convergence and underthrusting
and since then the evidence has become conclusive. The debate over the second
possibility, smooth aseismic underthrusting, continued until recently. Again the contrary
evidence is now strong, especially from paleoseismicity, the traces of past great
earthquakes preserved in the coastal and deep sea geological record and from
measurements of present elastic strain building up in the continent near the coast. The
observed deformation corresponds to that expected for a locked thrust fault.
Thus based on the available variety of evidence the third option has a strong possibility in
which great earthquakes do occur, but the last one was more than 200 years ago, prior to
the historical written record (Hyndman and Rogers, 2010).

The evidence based on the repeated sediment deposits on the floor of the Cascadia deep
sea basin indicate that the most recent major event was about 300 years ago. The intervals
for the last 13 events range mostly from about 300 to 900 years. This very long intervals
between Cascadia events compared with most subduction zones means that there is large
elastic strain build up and thus very large earthquake slip and large magnitude.
For the Cascadia subduction zone, the rate of convergence between the Juan de Fuca
plate and North American plate is about 40 mm/year (Riddihough, 1984). This
convergence represents an average rupture displacement of about 20m if the event
interval is 500 years on average.

282

Fig.F.4. Simplified model of great earthquake elastic strain build up and release
(Hyndman and Rogers, 2010)

In the simple subduction earthquake model, ongoing convergence drags down the
seaward nose of the continent and causes an upward flexural bulge (Fig. F.4). There also
is a region of crustal shortening (Fig. F.4). At the time of the earthquake, the edge of the
continent springs back seaward and the bulge collapses downward. The abrupt coseismic
uplift of the outer continental shelf and slope and subsidence near the coast are the main
cause of the great tsunamis. The collapse of the flexural bulge causes the sudden coastal
subsidence recorded in buried intertidal marshes (Hyndman and Rogers, 2010).

Inslab earthquakes (also known as intraplate, intraslab and Wadati-Benioff slab


earthquakes)

There are two regions of observed deep seismicity in the Cascadia system: one just north
of the Mendocino triple junction and another under the Puget lowlands/ Georgia Strait.
Large intraslab earthquakes, by virtue of their frequency of occurrence, locations directly

283

beneath population centers and source characteristics, represent a major earthquake


hazard in the Cascadia subduction zone. Investigations to date of intraslab earthquake
sources worldwide show that they tend to be enriched in high-frequency energy
compared to nearby interplate thrust earthquakes (i.e., interface earthquakes) of
comparable scalar seismic moment. Subduction in some particular regions, such as
Mexico, Peru and Chile has produced many large (M>7.08.0) and very destructive slab
earthquakes probably in the subducted slab mantle. These events are sometimes
exceptionally enriched in high-frequency energy, a factor that contributes to their
destructiveness. These unusual regions may involve down dip changes in the sign of slab
curvature (implied by their flat-slab geometry that occurs in some of those localities). The
normal-faulting focal mechanisms of these events are consistent with reverse-curvature
flexure in the slab mantle but the physical mechanisms for faulting in such settings are
not known (Kirby et al., 2002).

Inslab earthquakes (earthquakes within the subducting Juan de Fuca plate) also make a
major contribution to seismic hazard for the Strait of Georgia region of British Columbia.
Key knowledge of in-slab earthquakes needed to improve seismic hazard estimates for
southwestern British Columbia includes the constraints on the spatial distribution, rate
and maximum size of the earthquakes, the ground motions to be expected, the nature of
the earthquake sources and the structure and properties of the lithosphere through which
the waves propagate (Adams and Halchuk, 2002). Relatively little knowledge is available
about this type of earthquake.

In Canadas fourth generation seismic hazard model (see Adams et al., 1999a, 1999b and
2000), these earthquakes dominate the hazard despite their greater depth, firstly because
they occur at a rate up to five-fold higher per unit area than the shallower crustal
earthquakes, and secondly, because their predicted shaking is stronger than crustal events
of the same size (Adams and Halchuk, 2002).

The largest historical inslab event occurred in 1949, of moment magnitude about 6.9.
Compared with recent earthquakes, almost nothing is known about the rupture parameters

284

of this earthquake, such as its depth extent, fault length or stress drop. Some geophysical
constraints such as temperature in the slab are believed to limit the thickness of brittle
rock thus restricting fault width; larger earthquakes therefore require greater fault lengths
or greater slip (or both). The GSC model currently allows an upper bound magnitude of
7.3 for Puget Sound (with an uncertainty range of 7.17.6), presuming that a future large
earthquake could extend deeper into the slab, or have larger displacement, or rupture a
longer fault. In 1997, the USGS adopted an upper bound magnitude of 7.0. Because of
the high rate for these large earthquakes, their contribution to the total seismic hazard is
not trivial (for the GSCs results, about 1424% of the seismic hazard, dependent on
model, comes from earthquakes larger than the 1949 one) (Adams and Halchuk, 2002).

Different assumptions were adopted by the USGS in 1997 and the GSC in 19941999
and resulted in different estimates of seismic hazard for the U.S. and Canadian territory
overlying these inslab earthquakes. This indicates that the knowledge about this
earthquake type needs to be improved.

Fig. F.5. Seismic hazard deaggregations of 0.2 second spectral acceleration values at
2%/50 years for Bellingham show the GSC results are dominated by the contribution
from in-slab earthquakes, unlike the 1997 USGS results (Adams and Halchuk (2002)).

The seismic hazard analyses by Goda et al., (2010) using updated seismic data indicates
that the significance of the interface earthquakes are underestimated by the current model

285

of the Geological Survey of Canada (e.g., see the deaggregation results given in
Appendix C) and this event type contributes significantly to the overall seismic hazard at
medium to long period ranges.

286

Appendix G: Detailed information of the ground motion records used

NO.

Eq.Name

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39

Northridge
Northridge
Duzce,Turkey
HectorMine
ImperialValley
ImperialValley
Kobe,Japan
Kobe,Japan
Kocaeli,Turkey
Kocaeli,Turkey
Landers
Landers
LomaPrieta
LomaPrieta
Manjil,Iran
SuperstitionHills
SuperstitionHills
CapeMendocino
Chi Chi,Taiwan
Chi Chi,Taiwan
SanFernando
Friuli,Italy
Northridge
Northridge
Northridge
ImperialValley
ImperialValley
Kobe,Japan
Kobe,Japan
Landers
LomaPrieta
LomaPrieta
LomaPrieta
LomaPrieta
SuperstitionHills
Chi Chi,Taiwan
Chi Chi,Taiwan
Chi Chi,Taiwan
Chi Chi,Taiwan

RecordID EventID
953
960
1602
1787
169
174
1111
1116
1158
1148
900
848
752
767
1633
721
725
829
1244
1485
68
125
1003
1077
952
162
189
1107
1106
864
783
776
777
778
728
1524
1506
1595
1182

127
127
138
158
50
50
129
129
136
136
125
125
118
118
144
116
116
123
137
137
30
40
127
127
127
50
50
129
129
125
118
118
118
118
116
137
137
137
137

Type
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal
Crustal

Mag. Epicentral
(M) Dist.(Km)

Focal
depth

Vs30 PGA PGV

6.69
6.69
7.14
7.13
6.53
6.53
6.9
6.9
7.51
7.51
7.28
7.28
6.93
6.93
7.37
6.54
6.54
7.01
7.62
7.62
6.61
6.5
6.69
6.69
6.69
6.53
6.53
6.9
6.9
7.28
6.93
6.93
6.93
6.93
6.54
7.62
7.62
7.62
7.62

17.5
17.5
10
5
9.96
9.96
17.9
17.9
15
15
7
7
17.48
17.48
19
9
9
9.6
6.76
6.76
13
5.1
17.5
17.5
17.5
9.96
9.96
17.9
17.9
7
17.48
17.48
17.48
17.48
9
6.76
6.76
6.76
6.76

356
309
326
685
275
196
609
256
276
523
354
271
289
350
724
192
207
312
259
705
316
425
309
336
546
231
339
312
312
379
249
371
199
216
194
447
401
259
438

17.15
12.44
12.04
11.66
22.03
12.45
7.08
19.15
15.37
13.49
23.62
19.74
15.23
12.82
12.56
18.2
11.16
14.33
9.96
26
22.77
15.82
27.01
26.45
18.36
10.45
9.64
22.5
0.96
11.03
74.26
27.93
27.6
24.82
13.03
45.18
19.02
9.96
9.77

0.46
0.44
0.72
0.31
0.28
0.37
0.49
0.23
0.32
0.17
0.21
0.35
0.48
0.47
0.52
0.29
0.37
0.44
0.39
0.47
0.20
0.34
0.45
0.58
0.51
0.24
0.38
0.27
0.71
0.26
0.28
0.28
0.23
0.28
0.21
0.53
0.21
0.39
0.36

Table. G.1. Details of the 39 pairs of ground motions from crustal earthquakes

287

54
45
59
34
28
37
36
34
54
28
38
32
34
42
47
43
32
45
91
39
18
26
35
32
34
19
23
22
78
35
42
45
42
40
27
56
56
72
52

Databas
e
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA
NGA

No.

RecordID

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

27538
27548
27539
27451
27557
27545
9815
9697
9813
9839
9837
9841
9831
6301
6494
6306
6510
6267
20480
19650

EventID Type

Recordfilename

Datab
ase

Inslab
Inslab
Inslab
Inslab
Inslab
Inslab
Inslab
Inslab
Inslab
Inslab
Inslab
Inslab
Inslab
Inslab
Inslab
Inslab
Inslab
Inslab
Inslab
Inslab

6.8
6.8
6.8
6.8
6.8
6.8
7
7
7
7
7
7
7
6.8
6.8
6.8
6.8
6.8
6
6.2

111.88
117.20
108.12
114.01
109.86
112.09
68.78
63.77
117.21
53.87
58.16
103.06
79.59
69.64
45.57
58.31
72.32
46.89
52.26
79.79

120
120
120
120
120
120
71
71
71
71
71
71
71
51
51
51
51
51
52
50

390
368
733
392
521
967
456
392
390
934
670
305
521
469
362
247
487
268
282
272

0.85
0.71
0.52
0.48
0.47
0.48
0.74
0.80
0.75
0.74
0.72
0.61
0.58
0.60
0.52
0.41
0.46
0.39
0.15
0.14

23
21
14
16
17
12
31
19
19
13
15
25
20
24
15
33
16
25
13
10

IWTH020807240026.NS(EW)
IWTH120807240026.NS(EW)
IWTH030807240026.NS(EW)
IWT0070807240026.NS(EW)
IWTH210807240026.NS(EW)
IWTH090807240026.NS(EW)
IWTH040305261824.NS(EW)
IWT0070305261824.NS(EW)
IWTH020305261824.NS(EW)
MYGH030305261824.NS(EW)
IWTH270305261824.NS(EW)
MYGH050305261824.NS(EW)
IWTH210305261824.NS(EW)
HRS0090103241528.NS(EW)
EHMH050103241528.NS(EW)
HRS0140103241528.NS(EW)
HRSH030103241528.NS(EW)
EHM0030103241528.NS(EW)
CHB0100504110722.NS(EW)
HKD0680501182309.NS(EW)

21

WA:Olympia,
Nisqually Inslab
WSDOTTestLab

6.8

55.50

52.4

187

0.24

19

20010228_1.corrected.0730a_
cosmos
a.smc(0730c)

22

Seattle,WA
Nisqually Inslab
SeatacAirportFS

6.8

69.27

52.4

347

0.16

15

20010228_1.corrected.1421a_
cosmos
a.smc(1421c)

6.8
6.8
6.8
6.8
6.8
7
7
7
7
7
6.8
6.8
6.8
6.8

108.95
109.02
108.19
119.26
108.08
64.68
63.91
86.50
63.95
78.84
38.10
43.35
59.69
46.06

120
120
120
120
120
71
71
71
71
71
51
51
51
51

816
892
367
305
439
850
429
371
923
892
258
211
280
254

0.40
0.39
0.35
0.35
0.36
0.59
0.55
0.51
0.48
0.46
0.40
0.37
0.29
0.33

13
14
21
15
14
16
15
20
14
13
20
21
30
21

6.8

74.82

52.4

328

0.15

6.8

69.33

52.4

463

0.18

6.8

56.19

52.4

416

0.11

23
24
25
26
27
28
29
30
31
32
33
34
35
36

368
368
368
368
368
368
184
184
184
184
184
184
184
141
141
141
141
141
294
285

Mag. Epicentral Focal


V
PGA PGV
(M) Dist.(Km) depth s30

27550
368
Inslab
27554
368
Inslab
27463
368
Inslab
27544
368
Inslab
27460
368
Inslab
9840
184
Inslab
9816
184
Inslab
9836
184
Inslab
9833
184
Inslab
9828
184
Inslab
6270
141
Inslab
6311
141
Inslab
6278
141
Inslab
6493
141
Inslab
WA:WestSeattle;
37
Nisqually Inslab
FS29
WA:Bremerton;
38
Nisqually Inslab
FireStation
WA:GigHarbor
39
Nisqually Inslab
FireStation

IWTH140807240026.NS(EW)
IWTH180807240026.NS(EW)
IWT0190807240026.NS(EW)
IWTH080807240026.NS(EW)
IWT0160807240026.NS(EW)
MYGH040305261824.NS(EW)
IWTH050305261824.NS(EW)
IWTH260305261824.NS(EW)
IWTH230305261824.NS(EW)
IWTH180305261824.NS(EW)
EHM0070103241528.NS(EW)
HRS0190103241528.NS(EW)
EHM0150103241528.NS(EW)
EHMH040103241528.NS(EW)
20010228_1.corrected.1416a_
15
a.smc(1416c)
20010228_1.corrected.1422a_
12
a.smc(1422c)
20010228_1.corrected.0725a_
12
a.smc(0725c)

Table. G.2. Details of the 39 pairs of ground motions from subduction inslab earthquakes

288

K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK

K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
K KIK
cosmos
cosmos
cosmos

NO. RecordID EventID


1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39

19085
19004
19002
11026
11025
11020
11019
19266
19262
19192
21598
21597
21471
552
556
4409
757
169
175
237
304
128
18995
19083
19089
19084
19081
11075
11028
11024
10796
19199
21468
21473
370
323
168
167
232

276
276
276
194
194
194
194
278
278
278
301
301
301
12
12
99
17
Tohoku
Tohoku
Tohoku
Tohoku
Tohoku
276
276
276
276
276
194
194
194
194
278
301
301
Tohoku
Tohoku
Tohoku
Tohoku
Tohoku

Type
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface
Interface

Mag. Epicentral
(M) Dist.(Km)
7
7
7
7.9
7.9
7.9
7.9
6.7
6.7
6.7
7.1
7.1
7.1
6.7
6.7
6.2
6.7
9
9
9
9
9
7
7
7
7
7
7.9
7.9
7.9
7.9
6.7
7.1
7.1
9
9
9
9
9

79.98
93.02
73.08
119.95
62.65
110.04
73.17
58.62
87.85
59.53
97.14
115.52
115.21
59.03
77.94
21.01
39.99
83.70
71.00
69.14
93.91
52.32
49.15
66.74
50.80
92.82
94.44
57.68
148.64
79.56
73.16
80.28
96.84
117.12
74.46
62.49
66.35
56.45
78.41

Focal
depth

Vs30

PGA

PGV

48
48
48
42
42
42
42
46
46
46
42
42
42
39
39
15
35
24
24
24
24
24
48
48
48
48
48
42
42
42
42
46
42
42
24
24
24
24
24

326
311
243
213
230
250
219
213
326
282
859
348
284
217
348
258
217
463
436
418
487
399
282
189
213
389
219
353
315
204
437
243
303
240
574
450
494
505
492

0.665
0.342
0.296
0.568
0.387
0.501
0.396
0.390
0.386
0.320
0.380
0.358
0.260
0.207
0.181
0.188
0.171
1.748
0.960
0.906
0.869
0.778
0.343
0.276
0.322
0.331
0.254
0.465
0.493
0.420
0.359
0.266
0.238
0.206
0.687
0.672
0.625
0.651
0.382

24.60
20.18
24.19
36.60
60.15
29.72
45.67
19.30
16.03
14.27
13.28
13.31
14.21
14.82
14.32
14.72
15.04
70.90
44.43
56.84
35.62
30.49
14.33
27.10
17.89
15.82
26.41
29.97
23.34
35.90
44.94
16.58
15.56
14.74
46.01
27.09
28.47
23.94
58.94

Recordfilename

KSRH060411290332.NS(.EW) K KIK
HKD0800411290332.NS(.EW) K KIK
HKD0780411290332.NS(.EW) K KIK
KSRH100309260450.NS(.EW) K KIK
KSRH090309260450.NS(.EW) K KIK
KSRH030309260450.NS(.EW) K KIK
KSRH020309260450.NS(.EW) K KIK
KSRH100412062315.NS(.EW) K KIK
KSRH060412062315.NS(.EW) K KIK
HKD0710412062315.NS(.EW) K KIK
MYGH110508161146.NS(.EW) K KIK
MYGH100508161146.NS(.EW) K KIK
MYG0130508161146.NS(.EW) K KIK
MYZ0139610192344.NS(.EW) K KIK
MYZ0179610192344.NS(.EW) K KIK
TKY0100007011602.NS(.EW) K KIK
MYZ0139612030718.NS(.EW) K KIK
MYG0041103111446.NS(.EW) K KIK
MYG0121103111446.NS(.EW) K KIK
TCG0141103111446.NS(.EW) K KIK
FKSH101103111446.NS2(.EW) K KIK
IBR0061103111446.NS(.EW) K KIK
HKD0710411290332.NS(.EW) K KIK
KSRH040411290332.NS(.EW) K KIK
KSRH100411290332.NS(.EW) K KIK
KSRH050411290332.NS(.EW) K KIK
KSRH020411290332.NS(.EW) K KIK
TKCH080309260450.NS(.EW) K KIK
NMRH020309260450.NS(.EW) K KIK
KSRH070309260450.NS(.EW) K KIK
HKD0840309260450.NS(.EW) K KIK
HKD0780412062315.NS(.EW) K KIK
MYG0100508161146.NS(.EW) K KIK
MYG0150508161146.NS(.EW) K KIK
TCGH131103111446.NS2(.EW) K KIK
IBRH151103111446.NS2(.EW) K KIK
MYG0031103111446.NS(.EW) K KIK
MYG0021103111446.NS(.EW) K KIK
TCG0061103111446.NS(.EW) K KIK

Table. G.2. Details of the 39 pairs of ground motions from subduction interface
earthquakes

289

Data
base

a)

10

Crustal set (78 records)

Sa (T) (g)

0.1

0.01

0.001
0

2
3
Vibration Period, T (Sec)

b)

Fig. G.1. Response spectra for the crustal record set and the average response spectrum
for a) 78 records b) 44 records

290

a)

100

Interface set (78 records)

Sa (T) (g)

10

0.1

0.01

0.001
0

b)

0.5

1.5

100

2
2.5
3
Vibration Period, T (Sec)

3.5

4.5

Interface set (44 records)

Sa (T) (g)

10

0.1

0.01

0.001
0

2
3
Vibration Period, T (Sec)

Fig. G.2. Response spectra for the interface record set and the average response spectrum
for a) 78 records b) 44 records

291

a)

10

Inslab set (78 records)

Sa (T) (g)

0.1

0.01

0.001
0

b)

0.5

10

1.5

2
2.5
3
Vibration Period, T (Sec)

3.5

4.5

3.5

4.5

Inslab (44 records)

Sa (T) (g)

0.1

0.01

0.001
0

0.5

1.5

2
2.5
3
Vibration Period, T (Sec)

Fig. G.3. Response spectra for the inslab record set and the average response spectrum for
a) 78 records b) 44 records

292

Crustal set (78 records)


3

 values for individual records


Mean  value of all records

(T)

2
1
0
0

-1
-2
-3

Vibration Period, T (Sec)

Crustal set (44 records)


3

 values for individual records


Mean  value of all records

(T)

2
1
0
0

-1
-2
-3

Vibration Period, T (Sec)

Fig. G.5. (T) for the crustal record set and mean (T) for 78 records (top) and 44 records
(bottom) ( values computed using the BA08 GMPE).

293

Interface set (78 records)


3

 values for individual records


Mean  value of all records

(T)

2
1
0
0

-1
-2
-3

Vibration Period, T (Sec)

Interface set (44 records)


3

 values for individual records


Mean  value of all records

(T)

2
1
0
0

-1
-2
-3

Vibration Period, T (Sec)

Fig. G.5. (T) for the interface record set and mean (T) for 78 records (top) and 44
records (bottom) ( values computed using the Z06 GMPE).

294

Inslab set (78 records)


3

 values for individual records


Mean  value of all records

(T)

2
1
0
0

-1
-2
-3

Vibration Period, T (Sec)

Inslab set (44 records)


3

 values for individual records


Mean  value of all records

(T)

2
1
0
0

-1
-2
-3

Vibration Period, T (Sec)

Fig. G.6. (T) for the inslab record set and mean (T) for 78 records (top) and 44 records
(bottom) ( values computed using the Z06 GMPE).

295

Appendix H: Pushover analysis of the bridge in Chapter 5


Nonlinear static analysis (pushover) can provide some general information about the
seismic capacity of structures and is widely used for the case of buildings. However for
bridges pushover analysis must be carried out with caution especially when higher mode
effects are important or the structure is highly irregular. For such cases, the use of modal
pushover (Chopra and Goel, 2002) and adaptive pushover analysis seems to provide a
better prediction of the seismic response of the structure (Isakovic and Fischinger, 2006).
For the bridge structure under study, an adaptive pushover analysis was performed to
provide general information of the seismic behaviour including overall ductility capacity
and over-strength factor. The adaptive pushover analysis uses an updated stiffness
calculated for each load step. The initial load pattern was based on the fundamental mode
and during the analysis this pattern was updated in each step based on the stiffness and
mode shape in that step. The pushover curve shows the shear force of columns versus the
drift ratio of the central column (i.e., the critical column). The total elastic shear force of
columns using response spectrum analysis was 13000 kN, and by applying a response
modification factor of 3 in design according to the 2006 CHBDC, the total design shear
force is reduced to 4333 kN for the columns. The maximum shear force from pushover
analysis (Vmax) was found 6140 kN as shown in Fig. H.1. Thus the over-strength factor
defined as a ratio of the maximum shear force to the design shear force, is 6140/4333
=1.42.

To evaluate the structural collapse ductility, the procedure defined by ASCE/SEI 41-06
(ASCE, 2006), which was used by the ATC-63 provisions, was utilized here. For this
purpose, the effective yield displacement (y) and the ultimate displacement (ult),
defined at the point of 20% strength loss, was calculated as 33(mm) and 420 (mm),
respectively from the pushover analysis and the displacement at which strength
degradation starts was 260 mm. The structural collapse ductility, c, is ult/y =420/33 =
12.8. The ductility at the point of strength degradation is about 8.0.

296

Total base shear in columns (KN)

7000
6000
5000
4000
3000
2000
1000
0
0

0.02

0.04

0.06

Drift ratio (central column)

Figure H.1. Pushover analysis of the bridge

297

0.08

Appendix I: Some of the results from Chapter 6


In Chapter 6 the IDA method was applied (using a large record set) to evaluate the
seismic response of 30 bridges with different configurations designed using different
force modification factors. The detailed results from all 30 cases are available in tables.
However due to space limitations only the results obtained for the configurations 4 and 5
with column diameters of D=2.0 m, designed using a force modification factor of R=5 are
presented as examples.

Also the bridge structure studied in Chapter 5 was subjected to the large record sets
considered in Chapter 6 and the detailed results for this bridge for the case of the
restrained transverse movements at the abutments are also presented in this appendix.

The details of the ground motion record sets are given in Appendix G for 39 pairs of
ground motions (i.e., 78 horizontal components) for each earthquake type. The record
numbers in the tables are consistent with those given in Appendix G, considering that 39
pairs of records have 78 horizontal components. In the tables, CMR is the collapse
margin ratio and ACMR is the required collapse margin ratio for 10% probability of
collapse in MCE level (i.e., acceptable collapse margin ratio). More details concerning
the computation of CMR are available in Chapters 2, 5 and 6.
The standard deviation of the IDA results given in the tables represents the record-torecord variability, RTR. The total uncertainty, TOT, reported in the tables includes all
sources of uncertainty as discussed in more details in Chapter 2.
The criteria for the subsets of records (i.e., case 1 to 8 in the tables) are given in Chapter
6.

298

CollapseCapacities(g)

CollapseCapacities(g)

Record
Crustal
Interface
Inslab
Record Crustal
Interface
Inslab
number recordsset recordssets recordsset number recordsset recordssets recordsset
1
1.98
0.44
3.28
41
1.40
1.29
8.65
2
1.72
0.49
2.89
42
1.23
1.72
5.96
3
1.05
0.51
1.51
43
2.30
1.23
1.43
4
2.92
0.66
2.55
44
3.25
1.98
3.15
5
1.80
0.83
2.47
45
2.87
1.82
4.32
6
4.70
1.13
3.67
46
2.19
2.09
3.14
7
2.74
0.47
3.10
47
0.99
0.64
1.93
8
1.61
0.44
2.14
48
1.56
0.88
1.11
9
0.87
0.79
4.73
49
1.65
0.80
2.54
10
1.30
0.70
5.84
50
2.17
0.57
1.87
11
0.82
0.59
1.65
51
2.65
0.47
1.24
12
1.17
0.52
1.74
52
1.29
0.45
2.60
13
1.19
0.59
0.59
53
2.02
1.19
0.81
14
1.15
0.51
0.99
54
1.43
1.79
0.89
15
1.45
0.76
7.31
55
1.30
0.56
4.10
16
1.41
0.55
9.57
56
1.67
0.53
3.08
17
1.08
0.64
0.82
57
2.96
0.64
0.53
18
1.09
0.45
0.98
58
3.84
0.65
0.70
19
1.60
4.86
1.77
59
1.88
0.63
2.57
20
0.69
2.88
1.26
60
1.32
0.65
1.23
21
1.26
0.34
0.43
61
1.97
1.92
5.46
22
1.58
0.41
0.45
62
2.13
1.51
3.92
23
2.61
0.63
3.68
63
2.10
1.18
1.30
24
2.25
0.53
4.36
64
1.99
1.08
1.09
25
1.98
1.83
1.72
65
1.74
2.38
5.17
26
2.30
1.62
0.81
66
1.94
4.04
4.56
27
2.05
5.06
3.18
67
0.95
3.57
3.02
28
0.85
3.06
2.56
68
2.06
4.69
5.17
29
1.22
1.76
0.56
69
1.51
2.06
4.23
30
1.09
2.73
0.61
70
2.10
1.97
3.76
31
1.13
1.86
3.07
71
1.85
1.32
0.64
32
0.57
4.62
2.47
72
1.56
2.09
0.51
33
1.74
4.05
6.21
73
1.14
1.18
10.05
34
1.28
4.63
3.56
74
0.99
1.85
5.35
35
2.12
1.21
2.46
75
0.94
0.90
8.07
36
3.87
0.87
2.12
76
1.22
1.46
4.30
37
1.04
3.16
5.36
77
2.48
1.22
1.06
38
1.25
3.67
7.88
78
0.95
1.63
0.76
39
1.70
1.66
10.92
Median
1.61
1.18
2.57
40
2.36
4.08
12.54
Stdev
0.41
0.74
0.85

Table I.1. Collapse capacity of the bridge (configuration 4, D=2.0 m and R/I= 5) for
78x3=234 records from three earthquake types.

299

Percentiles

Case 1

Case2

Case3

Case4

Case5

Case6

Case7

Case8

Mean 
(Records)

SSF

78

0.78

0.157

1.148

0.74

78

0.59

0.146

1.121

50%

16%

84%

St Dev Num.

Crustal

1.61

1.09

2.30

0.41

Interface

1.18

0.53

2.83

Inslab

2.57

0.84

5.36

0.85

78

0.22

0.440

2.064

Crustal

1.58

1.08

2.26

0.41

69

0.95

0.268

1.208

Interface

0.98

0.51

3.52

0.81

58

1.00

0.465

1.192

Inslab

3.15

1.60

5.96

0.69

49

0.76

0.260

1.337

Crustal

1.59

1.09

2.21

0.35

74

0.78

0.138

1.129

Interface

1.18

0.53

2.67

0.73

77

0.60

0.159

1.131

Inslab

2.56

0.83

5.33

0.84

77

0.25

0.494

2.230

Crustal

1.56

1.08

2.16

0.34

65

0.96

0.259

1.198

Interface

0.88

0.51

3.08

0.78

56

1.03

0.602

1.234

Inslab

3.16

1.69

6.02

0.66

48

0.77

0.199

1.245

Crustal

1.57

1.07

2.21

0.40

74

0.83

0.215

1.194

Interface

0.79

0.51

1.79

0.59

57

0.46

-0.106

0.908

Inslab

0.70

0.51

2.38

0.72

15

0.17

0.669

3.109

Crustal

1.57

1.07

2.20

0.40

68

0.95

0.266

1.207

Interface

0.65

0.48

1.58

0.61

41

0.89

0.206

1.106

Inslab

1.99

1.02

3.00

0.58

0.97

0.805

2.068

Crustal

1.56

1.08

2.13

0.34

70

0.84

0.199

1.178

Interface

0.73

0.51

1.64

0.51

54

0.46

-0.110

0.904

Inslab

0.70

0.51

2.38

0.72

15

0.17

0.669

3.109

Crustal

1.56

1.08

2.13

0.33

64

0.96

0.257

1.196

Interface

0.64

0.47

1.22

0.46

38

0.92

0.395

1.200

Inslab

1.99

1.02

3.00

0.58

0.97

0.805

2.068

Table I.2. Percentiles of unmodified collapse capacity, RTR, 1 factor and spectral shape
factor (SSF) predicted for different earthquake types and for different subsets of records
in each set (Configuration 4, D = 2.0 m and R/I = 5).

300

Case 1

Case 2

Case 3

Case 4

Case 5

Case 6

Case 7

Case 8

Median

CMR

Modified
median

Modified
CMR

TOT

ACMR

CMR/ACMR
2.61

Crustal

1.61

4.94

1.84

5.67

0.61

2.17

Interface

1.18

3.64

1.33

4.08

0.87

3.04

1.34

Inslab

2.57

7.90

5.30

16.30

0.96

3.42

4.76

Average

1.74

5.35

2.68

8.24

0.81

2.84

2.90

Crustal

1.58

4.84

1.90

5.85

0.60

2.17

2.70

Interface

0.98

3.01

1.17

3.59

0.92

3.26

1.10

Inslab

3.15

9.69

4.21

12.95

0.83

2.88

4.49

Average

1.83

5.63

2.33

7.15

0.79

2.76

2.59

Crustal

1.59

4.88

1.79

5.51

0.57

2.07

2.66

Interface

1.18

3.62

1.33

4.10

0.86

3.00

1.37

Inslab

2.56

7.88

5.71

17.56

0.96

3.40

5.16

Average

1.73

5.32

2.78

8.56

0.80

2.79

3.07

Crustal

1.56

4.81

1.87

5.76

0.56

2.06

2.80

Interface

0.88

2.70

1.08

3.33

0.90

3.17

1.05

Inslab

3.16

9.73

3.94

12.11

0.80

2.77

4.37

Average

1.80

5.52

2.21

6.78

0.76

2.66

2.55
2.67

Crustal

1.57

4.83

1.87

5.76

0.60

2.16

Interface

0.79

2.44

0.72

2.21

0.74

2.59

0.85

Inslab

0.70

2.16

2.19

6.73

0.85

2.98

2.26

Average

1.05

3.21

1.57

4.81

0.73

2.55

1.89

Crustal

1.57

4.83

1.89

5.82

0.60

2.16

2.70

Interface

0.65

2.01

0.72

2.22

0.76

2.64

0.84

Inslab

1.99

6.13

4.12

12.68

0.73

2.56

4.95

Average

1.38

4.23

2.14

6.58

0.70

2.44

2.69
2.75

Crustal

1.56

4.80

1.84

5.66

0.56

2.05

Interface

0.73

2.25

0.66

2.03

0.68

2.39

0.85

Inslab

0.70

2.16

2.19

6.73

0.85

2.98

2.26

Average

1.02

3.14

1.53

4.71

0.70

2.44

1.93

Crustal

1.56

4.80

1.87

5.74

0.56

2.04

2.81

Interface

0.64

1.97

0.77

2.36

0.64

2.28

1.04

Inslab

1.99

6.13

4.12

12.68

0.73

2.56

4.95

Average

1.37

4.21

2.15

6.61

0.64

2.28

2.90

Table I.3. Modified and unmodified median collapse capacities and corresponding
collapse margin ratios (CMR) and acceptable CMR (ACMR) (Configuration 4, D = 2.0 m
and R/I = 5).

301

Unmodified

Modified

16%
84%
16%
84%
St Dev. Median
St Dev.
Percentile Percentile
Percentile Percentile

Median
Subset 1

1.72

0.76

3.67

0.74

1.98

1.04

4.78

0.83

Subset 2

1.80

0.82

3.90

0.76

1.96

1.08

4.81

0.76

Subset 3

1.70

0.78

3.38

0.73

1.95

1.01

4.46

0.84

Subset 4

1.75

0.81

3.67

0.75

1.91

1.12

4.35

0.72

Subset 5

1.25

0.56

2.16

0.65

1.92

0.65

3.51

0.73

Subset 6

1.61

0.64

2.40

0.61

1.90

0.65

3.20

0.72

Subset 7

1.25

0.56

2.16

0.63

1.90

0.65

3.31

0.72

Subset 8

1.61

0.64

2.40

0.61

1.90

0.65

3.20

0.71

Table I.4. Unmodified and modified collapse capacities for the combined records from 3
event types (configuration 4, D = 2.0 m and R/I = 5).
Number of records
Crustal Interface Inslab

Beta
total

CMR

Modified Acceptable CMR/A


CMR
CMR
CMR

Subset 1

78

75

64

0.94

5.29

6.10

3.35

1.82

Subset 2

59

57

49

0.88

5.52

6.04

3.09

1.95

Subset 3

74

71

60

0.95

5.21

6.00

3.40

1.77

Subset 4

57

56

47

0.85

5.37

5.87

2.96

1.98

Subset 5

19

18

16

0.85

3.85

5.90

2.98

1.98

Subset 6

11

10

0.85

4.96

5.83

2.97

1.96

Subset 7

19

18

16

0.85

3.85

5.85

2.96

1.98

Subset 8

11

10

0.84

4.96

5.83

2.95

1.98

Table I.5. Number of records in each subset and the computed CMR, ACMR, and total
uncertainty (configuration 4, D = 2.0 m and R/I = 5)
Unmodified
Median

Modified

16%
84%
16%
84%
St Dev. Median
St Dev.
Percentile Percentile
Percentile Percentile

Subset 1

1.45

0.64

2.45

0.62

1.69

0.73

2.89

0.61

Subset 2

1.51

0.64

2.80

0.65

1.65

0.81

3.06

0.61

Subset 3

1.43

0.65

2.30

0.60

1.67

0.75

2.62

0.59

Subset 4

1.40

0.64

2.39

0.63

1.74

0.80

2.70

0.55

Subset 5

1.28

0.63

2.10

0.57

1.40

0.56

2.29

0.63

Subset 6

1.22

0.55

2.18

0.63

1.43

0.63

2.29

0.62

Subset 7

1.25

0.60

1.99

0.53

1.36

0.53

2.05

0.59

Subset 8

1.22

0.54

1.99

0.57

1.41

0.65

1.98

0.52

Table I.6. Unmodified and modified collapse capacities for the combined records from 2
event types (excluding inslab events) (configuration 4, D = 2.0 m and R/I = 5)

302

Number of records
Crustal Interface Inslab

 TOT

CMR

Modified
CMR

ACMR

CMR/A
CMR

Subset 1

78

75

0.76

4.46

5.21

2.65

1.97

Subset 2

59

58

0.75

4.63

5.09

2.63

1.94

Subset 3

74

71

0.74

4.41

5.13

2.58

1.99

Subset 4

57

56

0.71

4.31

5.34

2.49

2.15

Subset 5

58

57

0.77

3.94

4.29

2.68

1.60

Subset 6

42

41

0.76

3.76

4.41

2.66

1.66

Subset 7

55

54

0.74

3.85

4.17

2.59

1.61

Subset 8

39

38

0.69

3.74

4.34

2.41

1.80

Table I.7. Number of records in each subset and the computed CMR, ACMR, and total
uncertainty for the case inslab events are excluded (configuration 4, D = 2.0 m and R/I =
5)

1.90 1.39 2.66

0.32

28

0.55

5.83

2.03

2.88 1.42

INTERFACE

1.16 0.54 3.08

0.76

32

0.88

3.56

3.10

1.15

INSLAB

3.15 2.48 5.73

0.54

15

0.70

9.69

2.46

3.93

ALLEVENTS

1.64 0.64 3.57

0.74

18

17

15

0.87

5.05

3.04

1.66 1.52 1.34 1.35

ALLEVENTS
1.70 0.64 2.81
(NoInslab)

0.65

28

27

0.79

5.21

2.75

1.90 1.42 1.20

CMR/
ACMR

Inslab

Inslab

CRUSTAL

St Dev

1.25

1.35

Crustal

Interface

TOT CMR ACMR

Earthquake
50% 16% 84%
type

Interface

Meanepsilon
ofrecords

Num.ofrecords
Crustal

Percentiles

Table I.8. Results from the epsilon-based method including the collapse capacity statistics
and number of records from each set (configuration 4, D = 2.0 m and R/I = 5)

303

Modified median &


(CMR/ACMR)

Combined
events

Combined
(no inslab records)

Subset 1
1.84 (2.61) 1.33 (1.34) 5.30 (4.76)
(simplified method)

1.98 (1.82)

1.69 (1.97)

Subset 2
1.90 (2.70) 1.17 (1.10) 4.21 (4.49)
(simplified method)

1.96 (1.95)

1.65 (1.94)

Subset 3
1.79 (2.66) 1.33 (1.37) 5.71 (5.16)
(simplified method)

1.95 (1.77)

1.67 (1.99)

1.87 (2.80) 1.08 (1.05) 3.94 (4.37)

1.91 (1.98)

1.74 (2.15)

1.87 (2.67) 0.72 (0.85) 2.19 (2.26)

1.92 (1.98)

1.40 (1.60)

1.89 (2.70) 0.72 (0.84) 4.12 (4.95)

1.90 (1.96)

1.43 (1.66)

1.84 (2.75) 0.66 (0.85) 2.19 (2.26)

1.90 (1.98)

1.36 (1.61)

1.87 (2.81) 0.77 (1.04) 4.12 (4.95)

1.90 (1.98)

1.41 (1.80)

1.90 (2.88) 1.16 (1.15) 3.15 (3.93)

1.64 (1.66)

1.70 (1.90)

Subset 4
(simplified method)
Subset 5
(simplified method)
Subset 6
(simplified method)
Subset 7
(simplified method)
Subset 8
(simplified method)
Epsilon-Based
method

Crustal
records

Interface
records

Inslab
records

Table I.9. Comparison of the median collapse capacity and CMR/ACMR ratios using
different subsets and methods (configuration 4, D = 2.0 m and R/I = 5)

304

CollapseCapacities(g)

CollapseCapacities(g)

Record
Crustal
Interface
Inslab
Record Crustal
Interface
Inslab
number recordsset recordssets recordsset number recordsset recordssets recordsset
1
1.54
0.31
2.97
41
0.99
0.86
4.63
2
1.22
0.37
2.19
42
1.19
1.53
4.07
3
0.70
0.37
1.08
43
1.64
0.91
1.04
4
2.38
0.51
1.58
44
2.13
1.87
2.26
5
1.35
0.64
1.83
45
1.99
1.29
4.15
6
2.76
0.93
2.63
46
1.60
1.71
2.19
7
1.85
0.35
1.53
47
0.61
0.46
1.65
8
0.95
0.32
1.84
48
1.16
0.68
0.84
9
0.75
0.58
4.46
49
1.19
0.50
1.60
10
0.98
0.54
4.55
50
1.43
0.47
0.99
11
0.63
0.42
1.23
51
1.86
0.36
0.99
12
0.79
0.39
1.26
52
0.89
0.36
1.92
13
0.91
0.42
0.41
53
1.55
0.89
0.63
14
0.91
0.36
0.71
54
1.09
1.16
0.74
15
1.01
0.58
9.34
55
0.91
0.50
3.46
16
1.05
0.39
7.42
56
1.12
0.42
2.77
17
0.86
0.55
0.66
57
2.16
0.42
0.41
18
0.79
0.42
1.08
58
2.70
0.46
0.55
19
1.21
3.87
1.27
59
1.29
0.43
1.86
20
0.58
2.08
1.05
60
0.95
0.51
0.95
21
0.81
0.24
0.42
61
1.55
0.95
4.47
22
1.08
0.31
0.41
62
1.51
1.11
3.21
23
2.14
0.42
3.10
63
1.59
0.87
1.25
24
1.52
0.37
3.12
64
1.47
0.76
0.82
25
1.36
1.46
1.35
65
1.33
1.47
4.15
26
1.12
1.19
0.75
66
1.17
2.68
3.63
27
1.32
3.27
6.59
67
0.72
2.70
2.02
28
0.66
2.09
2.41
68
1.64
3.93
2.72
29
0.98
1.30
0.48
69
1.13
1.93
2.96
30
0.85
1.98
0.54
70
1.33
1.48
2.23
31
0.79
1.54
2.01
71
1.45
1.08
0.54
32
0.42
3.30
1.52
72
1.14
1.70
0.45
33
1.21
2.90
5.45
73
0.85
0.70
6.16
34
1.00
3.33
5.19
74
0.79
1.34
4.40
35
1.49
1.07
1.85
75
0.79
0.72
4.52
36
1.86
0.65
1.51
76
1.04
1.13
3.17
37
0.85
2.17
3.73
77
1.88
0.77
0.78
38
1.10
2.42
5.31
78
0.74
0.99
0.55
39
1.35
0.99
8.35
Median
1.15
0.86
1.86
40
1.56
1.75
9.13
Stdev
0.37
0.73
0.84

Table I.10. Collapse capacity of the bridge (configuration 5, D=2.0 m and R/I= 5) for
78x3=234 records from three earthquake types.

305

Percentiles

Case 1

Case2

Case3

Case4

Case5

Case6

Case7

Case8

50%

16%

84%

St Dev Num.

Mean 
(Records)

SSF

Crustal

1.15

0.79

1.63

0.37

78

0.79

0.156

1.145

Interface

0.86

0.40

1.96

0.73

78

0.60

0.140

1.115

Inslab

1.86

0.72

4.47

0.84

78

0.23

0.347

1.755

Crustal

1.14

0.79

1.60

0.37

69

0.96

0.240

1.181

Interface

0.72

0.38

2.16

0.78

58

1.00

0.471

1.195

Inslab

2.10

1.02

4.54

0.73

48

0.78

0.353

1.462

Crustal

1.14

0.79

1.59

0.34

75

0.81

0.166

1.151

Interface

0.86

0.39

1.92

0.72

77

0.61

0.154

1.126

Inslab

1.85

0.71

4.45

0.83

77

0.25

0.411

1.931
1.164

Crustal

1.14

0.80

1.60

0.34

67

0.97

0.222

Interface

0.68

0.37

2.09

0.76

57

1.02

0.557

1.223

Inslab

2.10

1.02

4.54

0.73

48

0.78

0.353

1.462

Crustal

1.14

0.79

1.60

0.37

76

0.83

0.198

1.177

Interface

0.64

0.37

1.49

0.65

63

0.55

0.071

1.061

Inslab

0.76

0.45

2.01

0.64

26

0.23

0.539

2.399

Crustal

1.14

0.79

1.60

0.37

69

0.96

0.240

1.181

Interface

0.54

0.37

1.30

0.67

45

1.01

0.528

1.218

Inslab

1.26

0.69

2.11

0.57

17

0.79

0.579

1.852

Crustal

1.13

0.79

1.58

0.34

73

0.86

0.209

1.182

Interface

0.58

0.37

1.39

0.60

61

0.49

-0.038

0.967

Inslab

0.74

0.44

1.88

0.61

25

0.21

0.518

2.345

Crustal

1.14

0.80

1.60

0.34

67

0.97

0.222

1.164

Interface

0.51

0.37

1.08

0.61

42

1.04

0.650

1.246

Inslab

1.26

0.69

2.11

0.57

17

0.79

0.579

1.852

Table I.11. Percentiles of unmodified collapse capacity, RTR, 1 factor and spectral shape
factor (SSF) predicted for different earthquake types and for different subsets of records
in each set. (Configuration 5, D = 2.0 m and R/I = 5)

306

Case 1

Case 2

Case 3

Case 4

Case 5

Case 6

Case 7

Case 8

Median

CMR

Modified
median

Modified
CMR

TOT

ACMR

CMR/ACMR
1.90

Crustal

1.15

3.49

1.31

4.00

0.58

2.11

Interface

0.86

2.63

0.96

2.94

0.86

3.00

0.98

Inslab

1.86

5.66

3.26

9.93

0.95

3.39

2.93

Average

1.26

3.83

1.77

5.38

0.80

2.79

1.93
1.94

Crustal

1.14

3.47

1.35

4.10

0.58

2.11

Interface

0.72

2.19

0.86

2.62

0.90

3.16

0.83

Inslab

2.10

6.40

3.07

9.35

0.86

3.00

3.12

Average

1.28

3.89

1.69

5.14

0.78

2.73

1.88

Crustal

1.14

3.47

1.31

3.99

0.56

2.06

1.94

Interface

0.86

2.61

0.97

2.94

0.84

2.95

1.00

Inslab

1.85

5.64

3.58

10.89

0.94

3.33

3.27

Average

1.25

3.82

1.86

5.67

0.79

2.74

2.07

Crustal

1.14

3.47

1.33

4.04

0.56

2.05

1.97

Interface

0.68

2.08

0.84

2.54

0.88

3.09

0.82
3.12

Inslab

2.10

6.40

3.07

9.35

0.86

3.00

Average

1.27

3.85

1.67

5.09

0.77

2.69

1.89

Crustal

1.14

3.46

1.34

4.07

0.58

2.11

1.93

Interface

0.64

1.96

0.68

2.08

0.79

2.76

0.75

Inslab

0.76

2.31

1.82

5.53

0.78

2.71

2.04

Average

0.86

2.60

1.25

3.82

0.72

2.51

1.52

Crustal

1.14

3.47

1.35

4.10

0.58

2.11

1.94

Interface

0.54

1.64

0.66

2.00

0.80

2.80

0.71

Inslab

1.26

3.82

2.32

7.08

0.72

2.53

2.80

Average

0.97

2.94

1.40

4.25

0.71

2.47

1.72
1.99

Crustal

1.13

3.45

1.34

4.08

0.56

2.06

Interface

0.58

1.76

0.56

1.71

0.75

2.62

0.65

Inslab

0.74

2.25

1.73

5.27

0.76

2.63

2.00

Average

0.83

2.51

1.19

3.61

0.69

2.42

1.49

Crustal

1.14

3.47

1.33

4.04

0.56

2.05

1.97

Interface

0.51

1.56

0.64

1.94

0.76

2.65

0.73

Inslab

1.26

3.82

2.32

7.08

0.72

2.53

2.80

Average

0.96

2.91

1.38

4.21

0.68

2.40

1.75

Table I.12. Modified and unmodified median collapse capacities and corresponding
collapse margin ratios (CMR) and acceptable CMR (ACMR). (Configuration 5, D = 2.0
m and R/I = 5)

307

Unmodified
Median

Modified

16%
84%
16%
84%
St Dev. Median
St Dev.
Percentile Percentile
Percentile Percentile

Subset 1

1.21

0.57

2.69

0.73

1.45

0.77

3.02

0.80

Subset 2

1.24

0.60

2.71

0.74

1.44

0.82

3.22

0.76

Subset 3

1.19

0.58

2.43

0.72

1.45

0.80

3.16

0.81

Subset 4

1.21

0.58

2.69

0.74

1.41

0.84

3.16

0.74

Subset 5

0.98

0.42

1.69

0.62

1.33

0.50

2.38

0.67

Subset 6

1.04

0.42

1.86

0.62

1.25

0.49

2.43

0.68

Subset 7

0.98

0.42

1.63

0.61

1.35

0.47

2.31

0.67

Subset 8

1.04

0.42

1.86

0.62

1.25

0.51

2.42

0.67

Table I.13. Unmodified and modified collapse capacities for the combined records from 3
event types (configuration 5, D = 2.0 m and R/I = 5)
Number of records
Crustal Interface Inslab

 TOT

CMR

Modified
CMR

ACMR

CMR/A
CMR

Subset 1

78

75

64

0.92

3.68

4.41

3.24

1.36

Subset 2

58

56

48

0.88

3.78

4.37

3.10

1.41

Subset 3

75

71

61

0.93

3.64

4.43

3.28

1.35

Subset 4

58

56

48

0.87

3.70

4.31

3.04

1.41

Subset 5

31

30

26

0.81

2.98

4.04

2.82

1.43

Subset 6

20

19

17

0.81

3.18

3.81

2.84

1.34

Subset 7

30

29

25

0.81

2.98

4.10

2.81

1.46

Subset 8

20

19

17

0.81

3.18

3.80

2.81

1.35

Table I.14. Number of records in each subset and the computed CMR, ACMR, and total
uncertainty (configuration 5, D = 2.0 m and R/I = 5)
Unmodified
Median

Modified

16%
84%
16%
84%
St Dev. Median
St Dev.
Percentile Percentile
Percentile Percentile

Subset 1

1.06

0.50

1.86

0.60

1.18

0.54

2.09

0.59

Subset 2

1.06

0.47

1.91

0.62

1.19

0.63

2.00

0.58

Subset 3

1.05

0.50

1.74

0.59

1.18

0.54

2.01

0.57

Subset 4

1.03

0.46

1.81

0.61

1.21

0.63

1.94

0.54

Subset 5

0.98

0.44

1.58

0.58

1.12

0.49

1.78

0.58

Subset 6

0.98

0.42

1.64

0.62

1.11

0.56

1.85

0.54

Subset 7

0.96

0.45

1.54

0.55

1.10

0.42

1.64

0.59

Subset 8

0.95

0.42

1.55

0.60

1.10

0.57

1.51

0.48

Table I.15. Unmodified and modified collapse capacities for the combined records from
2 event types (excluding inslab events) (configuration 5, D = 2.0 m and R/I = 5)

308

Number of records
Crustal Interface Inslab

 TOT

CMR

Modified
CMR

ACMR

CMR/A
CMR

Subset 1

78

74

0.74

3.23

3.59

2.60

1.38

Subset 2

60

58

0.73

3.22

3.63

2.54

1.43

Subset 3

75

71

0.73

3.19

3.59

2.54

1.41

Subset 4

59

57

0.70

3.12

3.68

2.46

1.50

Subset 5

65

63

0.73

2.98

3.41

2.56

1.33

Subset 6

47

45

0.70

2.98

3.39

2.46

1.38

Subset 7

63

61

0.74

2.94

3.34

2.57

1.30

Subset 8

43

42

0.65

2.89

3.34

2.31

1.44

Table I.16. Number of records in each subset and the computed CMR, ACMR, and total
uncertainty for the case inslab events are excluded (configuration 5, D = 2.0 m and R/I =
5)

Inslab

Inslab

CRUSTAL

1.35 0.98 1.88

0.31

26

0.55 4.11

2.01

2.04 1.42

INTERFACE

0.67 0.37 2.10

0.76

32

0.88 2.06

3.10

0.66

INSLAB

2.23 1.54 4.51

0.59

15

0.74 6.79

2.58

2.64

ALLEVENTS 1.11 0.51 2.70

0.69

18

17

15

0.82 3.37

2.86

1.18 1.56 1.34 1.35

ALLEVENTS
1.15 0.45 2.09
(NoInslab)

0.66

26

24

0.80 3.50

2.77

1.26 1.42 1.23

CMR/
ACMR

1.20

1.35

Crustal

Interface

TOT CMR ACMR

Earthquake
50% 16% 84% St Dev
type

Interface

Meanepsilon
ofrecords

Num.ofrecords
Crustal

Percentiles

Table I.17. Results from epsilon-based method including collapse capacity statistics and
number of records from each set (configuration 5, D = 2.0 m and R/I = 5)

309

Modified median &


(CMR/ACMR)

Combined
events

Combined
(no inslab records)

Subset 1
1.31 (1.90) 0.96 (0.98) 3.26 (2.93)
(simplified method)

1.45 (1.36)

1.18 (1.38)

Subset 2
1.35 (1.94) 0.86 (0.83) 3.07 (3.12)
(simplified method)

1.44 (1.41)

1.19 (1.43)

Subset 3
1.31 (1.94) 0.97 (1.00) 3.58 (3.27)
(simplified method)

1.45 (1.35)

1.18 (1.41)

1.33 (1.97) 0.84 (0.82) 3.07 (3.12)

1.41 (1.41)

1.21 (1.50)

1.34 (1.93) 0.68 (0.75) 1.82 (2.04)

1.33 (1.43)

1.12 (1.33)

1.35 (1.94) 0.66 (0.71) 2.32 (2.80)

1.25 (1.34)

1.11 (1.38)

1.34 (1.99) 0.56 (0.65) 1.73 (2.00)

1.35 (1.46)

1.10 (1.30)

1.33 (1.97) 0.64 (0.73) 2.32 (2.80)

1.25 (1.35)

1.10 (1.44)

1.35 (2.04) 0.67 (0.66) 2.23 (2.64)

1.11 (1.18)

1.15 (1.26)

Subset 4
(simplified method)
Subset 5
(simplified method)
Subset 6
(simplified method)
Subset 7
(simplified method)
Subset 8
(simplified method)
Epsilon-Based
method

Crustal
records

Interface
records

Inslab
records

Table I.18. Comparison of the median collapse capacity and CMR/ACMR ratios
using different subsets and methods (configuration 5, D = 2.0 m and R/I = 5)

310

CollapseCapacities(g)

CollapseCapacities(g)

Record Crustal
Interface
Inslab
Record Crustal
Interface
Inslab
number recordsset recordssets recordsset number recordsset recordssets recordsset
1
1.00
0.55
2.17
41
0.67
1.46
3.02
2
0.88
0.77
3.57
42
1.26
2.06
3.00
3
1.34
0.52
1.63
43
1.50
1.80
1.83
4
3.37
1.15
1.30
44
3.27
1.86
3.43
5
1.25
1.69
2.23
45
2.31
1.44
2.71
6
1.09
1.66
2.91
46
2.15
1.55
1.25
7
1.05
0.88
1.20
47
1.07
1.40
1.15
8
0.67
0.82
1.39
48
1.57
1.73
1.06
9
1.84
0.84
5.39
49
0.90
1.12
2.33
10
1.14
0.64
2.56
50
1.97
0.64
1.43
11
0.84
0.53
0.86
51
2.95
0.39
1.43
12
1.49
0.68
1.10
52
0.94
0.59
1.36
13
2.31
0.75
0.62
53
1.82
0.96
1.78
14
3.55
0.77
0.48
54
2.25
1.26
1.67
15
1.57
0.82
6.27
55
1.35
0.94
2.89
16
2.58
0.49
3.29
56
0.84
0.87
2.89
17
1.13
0.45
0.88
57
2.26
0.77
0.47
18
0.84
0.62
1.16
58
2.36
0.74
0.44
19
1.35
9.94
0.68
59
2.11
0.74
2.34
20
1.00
5.29
0.61
60
2.04
0.66
1.50
21
1.07
0.55
0.52
61
1.32
2.01
1.40
22
0.79
0.53
0.75
62
1.54
1.96
1.64
23
3.30
0.49
2.65
63
1.11
2.06
0.50
24
4.50
0.59
2.19
64
1.39
1.71
0.35
25
1.55
1.67
1.99
65
1.35
1.30
4.10
26
4.50
2.85
1.44
66
1.00
1.18
2.85
27
2.62
4.29
2.00
67
1.70
3.54
3.15
28
0.77
3.33
1.98
68
1.58
4.99
2.82
29
2.29
2.65
0.59
69
1.03
3.21
6.55
30
0.83
3.27
0.53
70
1.03
3.74
3.34
31
1.09
1.66
3.40
71
1.39
2.22
0.83
32
0.66
1.60
2.90
72
1.26
1.55
0.63
33
1.36
3.00
2.55
73
1.74
1.01
1.20
34
1.31
3.87
1.76
74
1.25
1.59
1.71
35
1.17
1.01
2.02
75
1.48
1.80
1.29
36
1.41
1.50
2.91
76
0.86
1.94
0.95
37
1.41
1.22
2.44
77
1.84
0.77
1.13
38
0.73
1.61
0.89
78
1.72
0.75
0.92
39
2.06
2.44
2.55
Median
1.37
1.35
1.69
40
1.98
2.42
2.03
Stdev
0.52
0.75
0.65

Table I.19. Collapse capacity of the bridge in Chapter 5 (H=5m and D=1.5m) for
78x3=234 records from three earthquake types.

311

Percentiles

Case 1

Case2

Case3

Case4

Case5

Case6

Case7

Case8

50%

16%

84%

St Dev Num.

Mean 
(Records)

SSF

Crustal

1.37

0.91

2.28

0.45

78

0.89

0.257

1.189

Interface

1.35

0.64

2.58

0.68

78

0.96

0.087

1.059

Inslab

1.69

0.78

2.91

0.66

78

0.35

0.335

1.644

Crustal

1.41

1.00

2.31

0.44

71

1.02

0.258

1.148

Interface

1.15

0.61

2.91

0.72

67

1.21

0.339

1.149

Inslab

2.02

1.20

3.00

0.50

51

0.82

0.126

1.135

Crustal

1.35

0.90

2.13

0.40

74

0.88

0.249

1.184

Interface

1.28

0.64

2.42

0.62

76

0.95

0.070

1.048

Inslab

1.67

0.82

2.90

0.62

75

0.37

0.410

1.822

Crustal

1.39

0.95

2.20

0.39

67

1.03

0.266

1.152

Interface

1.13

0.60

2.77

0.68

66

1.21

0.343

1.151

Inslab

2.03

1.27

2.95

0.41

48

0.80

0.037

1.039

Crustal

1.37

0.91

2.28

0.45

78

0.89

0.257

1.189

Interface

1.20

0.63

2.04

0.59

72

0.91

0.019

1.013

Inslab

1.36

0.59

2.96

0.72

47

0.51

0.497

1.930

Crustal

1.41

1.00

2.31

0.44

71

1.02

0.258

1.148

Interface

0.96

0.59

2.06

0.62

61

1.18

0.268

1.124

Inslab

1.90

1.10

3.16

0.57

32

1.05

0.237

1.202

Crustal

1.35

0.90

2.13

0.40

74

0.88

0.249

1.184

Interface

1.18

0.62

2.00

0.57

71

0.89

-0.024

0.983

Inslab

1.39

0.61

2.92

0.71

44

0.48

0.557

2.124

Crustal

1.39

0.95

2.20

0.39

67

1.03

0.266

1.152

Interface

0.95

0.59

2.04

0.61

60

1.19

0.301

1.137

Inslab

1.98

1.18

3.19

0.52

31

1.06

0.213

1.178

Table I.20. Percentiles of unmodified collapse capacity, RTR, 1 factor and spectral shape
factor (SSF) predicted for different earthquake types and for different subsets of records
in each set (bridge in Chapter 5 (H=5m and D=1.5m)).

312

Case 1

Case 2

Case 3

Case 4

Case 5

Case 6

Case 7

Case 8

Median

CMR

Modified
median

Modified
CMR

TOT

ACMR

CMR/ACMR
1.43

Crustal

1.37

2.69

1.63

3.20

0.63

2.25

Interface

1.35

2.65

1.43

2.80

0.81

2.83

0.99

Inslab

1.69

3.32

2.78

5.46

0.80

2.78

1.97

Average

1.47

2.88

1.94

3.80

0.74

2.59

1.47

Crustal

1.41

2.77

1.62

3.18

0.63

2.23

1.42

Interface

1.15

2.26

1.32

2.59

0.85

2.97

0.87

Inslab

2.02

3.96

2.29

4.50

0.67

2.36

1.91

Average

1.53

3.00

1.75

3.42

0.71

2.50

1.37
1.46

Crustal

1.35

2.65

1.60

3.14

0.60

2.15

Interface

1.28

2.50

1.34

2.62

0.77

2.67

0.98

Inslab

1.67

3.28

3.04

5.97

0.77

2.67

2.24

Average

1.43

2.81

1.98

3.89

0.71

2.48

1.57

Crustal

1.39

2.73

1.60

3.15

0.59

2.14

1.47

Interface

1.13

2.22

1.30

2.56

0.81

2.84

0.90
1.90

Inslab

2.03

3.97

2.10

4.13

0.60

2.17

Average

1.52

2.97

1.67

3.28

0.67

2.36

1.39

Crustal

1.37

2.69

1.63

3.20

0.63

2.25

1.43

Interface

1.20

2.36

1.22

2.39

0.74

2.57

0.93

Inslab

1.36

2.67

2.62

5.15

0.85

2.97

1.73

Average

1.32

2.58

1.82

3.58

0.74

2.58

1.39
1.42

Crustal

1.41

2.77

1.62

3.18

0.63

2.23

Interface

0.96

1.88

1.08

2.11

0.76

2.65

0.80

Inslab

1.90

3.73

2.29

4.49

0.72

2.53

1.77

Average

1.43

2.81

1.67

3.27

0.70

2.46

1.33

Crustal

1.35

2.65

1.60

3.14

0.60

2.15

1.46

Interface

1.18

2.31

1.16

2.27

0.72

2.53

0.90

Inslab

1.39

2.74

2.96

5.81

0.84

2.92

1.99

Average

1.31

2.58

1.90

3.73

0.72

2.51

1.48
1.47

Crustal

1.39

2.73

1.60

3.15

0.59

2.14

Interface

0.95

1.86

1.08

2.11

0.75

2.62

0.81

Inslab

1.98

3.88

2.33

4.57

0.69

2.42

1.89

Average

1.45

2.83

1.68

3.29

0.68

2.38

1.38

Table I.21. Modified and unmodified median collapse capacities and corresponding
collapse margin ratios (CMR) and acceptable CMR (ACMR) (bridge in Chapter 5 (H=5m
and D=1.5m)) .

313

Unmodified
Median

Modified

16%
84%
16%
84%
St Dev. Median
St Dev.
Percentile Percentile
Percentile Percentile

Subset 1

1.44

0.79

2.89

0.60

1.84

0.93

3.52

0.63

Subset 2

1.60

0.85

2.90

0.59

1.83

0.96

3.22

0.58

Subset 3

1.43

0.82

2.65

0.56

1.84

0.94

3.46

0.62

Subset 4

1.58

0.87

2.83

0.53

1.76

0.96

2.94

0.52

Subset 5

1.35

0.69

2.55

0.61

1.76

0.88

3.06

0.62

Subset 6

1.66

0.83

2.94

0.58

1.83

0.89

3.05

0.59

Subset 7

1.33

0.71

2.32

0.59

1.78

0.91

3.04

0.61

Subset 8

1.63

0.88

2.89

0.54

1.79

0.94

2.87

0.56

Table I.22. Unmodified and modified collapse capacities for the combined records from 3
event types (bridge in Chapter 5 (H=5m and D=1.5m))
Number of records
Crustal Interface Inslab

 TOT

CMR

Modified
CMR

ACMR

CMR/A
CMR

Subset 1

78

63

66

0.78

2.83

3.61

2.70

1.34

Subset 2

60

48

51

0.73

3.15

3.59

2.56

1.40

Subset 3

74

59

62

0.77

2.81

3.60

2.68

1.35

Subset 4

56

45

48

0.68

3.09

3.45

2.40

1.44

Subset 5

55

44

47

0.76

2.64

3.45

2.66

1.30

Subset 6

37

30

32

0.74

3.26

3.58

2.57

1.39

Subset 7

51

41

44

0.76

2.61

3.50

2.65

1.32

Subset 8

36

29

31

0.71

3.20

3.51

2.49

1.41

Table I.23. Number of records in each subset and the computed CMR, ACMR, and total
uncertainty (bridge in Chapter 5 (H=5m and D=1.5m))
Unmodified
Median

Modified

16%
84%
16%
84%
St Dev. Median
St Dev.
Percentile Percentile
Percentile Percentile

Subset 1

1.35

0.77

2.44

0.59

1.62

0.84

2.69

0.58

Subset 2

1.40

0.77

2.61

0.60

1.70

0.92

2.81

0.58

Subset 3

1.31

0.75

2.29

0.54

1.56

0.81

2.53

0.53

Subset 4

1.35

0.77

2.31

0.55

1.61

0.88

2.55

0.52

Subset 5

1.32

0.75

2.26

0.54

1.55

0.77

2.50

0.55

Subset 6

1.33

0.76

2.30

0.55

1.55

0.87

2.52

0.53

Subset 7

1.26

0.75

2.06

0.51

1.51

0.76

2.32

0.51

Subset 8

1.32

0.77

2.15

0.52

1.55

0.89

2.37

0.49

Table I.24. Unmodified and modified collapse capacities for the combined records from 2
event types (excluding inslab events) (bridge in Chapter 5 (H=5m and D=1.5m))

314

Number of records
Crustal Interface Inslab

Beta
total

CMR

Modified Acceptable CMR/A


CMR
CMR
CMR

Subset 1

78

62

0.73

2.64

3.18

2.55

1.25

Subset 2

71

57

0.73

2.75

3.34

2.55

1.31

Subset 3

74

59

0.69

2.58

3.05

2.43

1.26

Subset 4

67

53

0.69

2.65

3.15

2.42

1.30

Subset 5

78

62

0.71

2.58

3.04

2.47

1.23

Subset 6

71

57

0.69

2.61

3.04

2.42

1.25

Subset 7

74

59

0.68

2.47

2.96

2.40

1.24

Subset 8

67

53

0.66

2.58

3.03

2.33

1.30

Table I.25. Number of records in each subset and the computed CMR, ACMR, and total
uncertainty for the case inslab events are excluded (bridge in Chapter 5 (H=5m and
D=1.5m))

Inslab

Inslab

CRUSTAL

1.74 1.09 2.34

0.42

41

0.61 3.40

2.20

1.55 1.37

INTERFACE

1.18 0.62 2.99

0.74

45

0.86 2.31

3.03

0.76

INSLAB

2.26 1.38 3.36

0.47

18

0.65 4.43

2.29

1.94

ALLEVENTS 1.71 1.09 3.54

0.58

21

17

18

0.73 3.35

2.56

1.31 1.45 1.59 1.56

ALLEVENTS
1.60 0.83 2.92
(NoInslab)

0.58

41

32

0.73 3.15

2.56

1.23 1.37 1.59

CMR/
ACMR

1.43

1.56

Crustal

Interface

TOT CMR ACMR

Earthquake
50% 16% 84% St Dev
type

Interface

Meanepsilon
ofrecords

Num.ofrecords
Crustal

Percentiles

Table I.26. Results from epsilon-based method including collapse capacity statistics and
number of records from each set (studied bridge (H=5m and D=1.5m))

315

Modified median &


(CMR/ACMR)

Combined
events

Combined
(no inslab records)

Subset 1
1.63 (1.43) 1.43 (0.99) 2.78 (1.97)
(simplified method)

1.84 (1.34)

1.62 (1.25)

Subset 2
1.62 (1.42) 1.32 (0.87) 2.29 (1.91)
(simplified method)

1.83 (1.40)

1.70 (1.31)

Subset 3
1.60 (1.46) 1.34 (0.98) 3.04 (2.24)
(simplified method)

1.84 (1.35)

1.56 (1.26)

1.60 (1.47) 1.30 (0.90) 2.10 (1.90)

1.76 (1.44)

1.61 (1.30)

1.63 (1.43) 1.22 (0.93) 2.62 (1.73)

1.76 (1.30)

1.55 (1.23)

1.62 (1.42) 1.08 (0.80) 2.29 (1.77)

1.83 (1.39)

1.55 (1.25)

1.60 (1.46) 1.16 (0.90) 2.96 (1.99)

1.78 (1.32)

1.51 (1.24)

1.60 (1.47) 1.08 (0.81) 2.33 (1.89)

1.79 (1.41)

1.55 (1.30)

1.74 (1.55) 1.18 (0.76) 2.26 (1.94)

1.71 (1.31)

1.60 (1.23)

Subset 4
(simplified method)
Subset 5
(simplified method)
Subset 6
(simplified method)
Subset 7
(simplified method)
Subset 8
(simplified method)
Epsilon-Based
method

Crustal
records

Interface
records

Inslab
records

Table I.27. Comparison of the median collapse capacity and CMR/ACMR ratios
using different subsets and methods (studied bridge (H=5m and D=1.5m))

316

References
AASHTO, Guide specifications for LRFD seismic bridge design. 2009., Subcommittee
T-3 for Seismic Effects on Bridges, American Association of State Highway and
Transportation Officials, Washington D.C.

AASHTO. 2004. AASHTO LRFD Bridge Design Specifications American Association


of State Highway and Transportation Officials, Washington, D.C.

Abrahamson, N.A., and Silva, W.J. 1997. Empirical response spectral attenuation
relations for shallow crustal earthquake. Seismological Research Letters, 68 (1): 94-126.

Adams, J., and Halchuk, S. 2003. Fourth generation seismic hazard maps of Canada:
values for over 650 Canadian localities intended for the 2005 National Building Code of
Canada. Open-File 4459, Geological Survey of Canada, Ottawa, Canada.

Adams, J., and Halchuk,S. 2002., Knowledge of inslab earthquakes needed to improve
seismic hazard estimates for southwestern British Columbia, The Cascadia subduction
zone and related subduction systems. U.S. Geological Survey Open-File Report 02328
Geological Survey of Canada Open File 4350.

Adams, J., Halchuk, S., and Weichert, D.H. 2000., Lower probability hazard, better
performance? Understanding the shape of the hazard curves from Canadas Fourth
Generation seismic hazard results, Paper 1555, 12th World Conference on Earthquake
Engineering, Auckland, 30th January4th February 2000, 8 pp.

Adams, J., Weichert, D.H., and Halchuk, S. 1999a. Trial seismic hazard maps of Canada
1999: 2%/50 year values for selected Canadian cities, 107 pp., Geological Survey of
Canada Open File 3724.

317

Adams, J., Weichert, D.H., and Halchuk, S. 1999b. Lowering the probability level
fourth generation seismic hazard results for Canada at the 2% in 50 year probability level,
in Proceedings 8th Canadian Conference on Earthquake Engineering, Vancouver 1316th
June 1999, pp. 8388.

Antoniou, S., Rovithakis, A., and Pinho, R. 2002. Development and verification of a fully
adaptive pushover procedure. Paper No. 822. Proceedings of the 12th European
Conference on Earthquake Engineering, London, U.K.

ASCE 7-05. 2005. Minimum design loads for buildings and other structures. American
Society of Civil Engineers (ASCE), Reston, Virginia.

ASCE/SEI 41-06. 2006. Seismic rehabilitation of existing buildings. American Society of


Civil Engineers (ASCE), Reston, Virginia.

ATC-40 .1996. Seismic evaluation and retrofit of concrete buildings. Applied


Technology Council (ATC). Redwood City, CA.

ATC-62. 2008. Effects of strength and stiffness degradation on seismic response, ATC62/FEMA P440A, prepared by the Applied Technology Council for the Federal
Emergency Management Agency, Applied Technology Council, Redwood City,
California.

ATC-63. 2008. Quantification of building seismic performance factors, ATC-63 Project


Report - 90% Draft, prepared by the Applied Technology Council for the Federal
Emergency Management Agency, Applied Technology Council, Redwood City,
California.

Atkinson, G.M., and Boore, D.M. 2003. Empirical ground-motion relations for
subduction-zone earthquakes and their application to Cascadia and other regions. Bull.
Seism. Soc. Am., 93: 1703-1729.

318

Atkinson, G.M., and Goda, K. 2011. Effects of seismicity models and new groundmotion prediction equations on seismic hazard assessment for four Canadian cities.
Bulletin of the Seismological Society of America, 101 (1): 176-189.

Aviram, A., Mackie, K.R., and Stojadinovic, B. 2008. Effect of abutment modeling on
the seismic response of bridge structures. Earthquake Eng & Eng Vibration, 7(4): 395402.

Aydinoglu, M.N. 2004. An improved pushover procedure for engineering practice:


incremental response spectrum analysis IRSA. Proceedings of the International
Workshop: Performance-Based Seismic Design Concepts and Implementation, Bled,
Slovenia, 345356.

Baker, J.W. 2011. The conditional mean spectrum: A tool for ground motion selection,
Journal of Structural Engineering, 137 (3): 322-331.

Baker, J.W., and Cornell, C.A. 2005, Vector-valued ground motion intensity measures for
probabilistic seismic demand analysis, Report #150, John A. Blume Earthquake
Engineering Center, Stanford, CA .

Baker, J.W., and Cornell, C.A. 2006a. Spectral shape, epsilon and record selection.
Earthquake Engineering& Structural Dynamics, 34(10):1193-1217.

Baker, J.W., and Cornell, C.A. 2006b. Correlation of response spectral values for
multicomponent ground motions. Bulletin of the Seismological Society of America, 96
(1): 215-227.

Baker, J.W., and Cornell, C.A. 2006c. Which Spectral Acceleration Are You Using?
Earthquake Spectra, 22(2): 293312.

319

Baker, J.W., and Cornell, C.A. 2008. Vector-valued intensity measures incorporating
spectral shape for prediction of structural response. Journal of Earthquake Engineering,
12 (4): 534-554.

Bazzurro, P., and Cornell, C.A. 1999. Disaggregation of seismic hazard. Bulletin of the
Seismological Society of America, 89(2): 501-520.

Benjamin, J.R., and Cornell, C.A. 1970. Probability, statistics and decision for civil
engineers. New York: McGraw-Hill.

Bentz, E. 2001. Response-2000, Shell-2000, Triax-2000, Membrane-2000: User Manual,


Department of Civil Engineering, University of Toronto, Toronto, Ontario, Canada.

Berry, M., Parrish, M., and Eberhard, M. 2004. PEER Structural Performance Database
Users Manual. (www.ce.washington.edu/~peera1), Pacific Earthquake Engineering
Research Center, University of California, Berkeley, Calif.

Berry, M.P., and Eberhard, M.O. 2007. Performance modeling strategies for modern
reinforced concrete bridge columns. PEER-2007/07, Pacific Earthquake Engineering
Research Center, University of California- Berkeley, Berkeley, Calif.

Boore, D.M., and Atkinson, G.M. 2008. Ground-motion prediction equations for the
average horizontal component of PGA, PGV, and 5%-damped PSA at spectral periods
between 0.01 s and 10.0 s, Earthquake Spectra, 24 (1): 99-138.

Botero, J.C.A. 2004. Displacement-based design of continuous concrete bridges under


transverse seismic excitation. Master thesis, European School for Advanced Studies in
Reduction of Seismic Risk (ROSE School), Pavia, Italy.

320

Bracci, J.M., Kunnath, S.K., and Reinhorn, A.M. 1997. Seismic performance and retrofit
evaluation for reinforced concrete structures. Journal of Structural Engineering (ASCE),
123(1):310.

Broderick, B.M., and Elnashai, A.S. 1995. Analysis of the failure of Interstate 10 freeway
ramp during the Northridge earthquake of 17 January 1994, Earthquake Engineering &
Structural Dynamics, 24(2): 189-209.

Caltrans. 2006. Seismic Design Criteria. California Department of Transportation,


California.

Caltrans. 2011. Bridge memo to designers. California Department of Transportation.


Available from: http://www.dot.ca.gov/hq/esc/techpubs/manual/bridgemanuals/bridgememo-to-designer/bmd.html [accessed Sep. 2011].

Calvi, G.M., Elnashai, S.A., and Pavese, A. 1994. Influence of regularity on the response
of RC bridges, Second International Workshop on Seismic Design of Bridges,
Queenstown, New Zealand, pp. 80-89.

Calvi, G.M., and Kingsley, G.R. 1995. Displacement-based seismic design of multidegree of freedom bridge structures. Earthquake Engineering and Structural Dynamics,
24: 1247-1266.

Calvi, G.M., and Pavese, A. 1997. Conceptual design of isolation systems for bridge
structures. Journal of Earthquake Engineering, 1(1): 193-218.

Calvi, G.M., and Pinto, P.E. 1996. Experimental and numerical investigations on the
seismic response of bridges and recommendations for code provisions. ECOESJIPREC8
Report No 4, LNEC. Lisbon.

321

Carr, A. 2009. RUAUMOKO, a computer program for Inelastic Dynamic Analysis.


Department of Civil Engineering, University of Canterbury, New Zealand.

CEN. 2005. Eurocode 8: Design of structures for earthquake resistance, part2: Bridges,
EN 1998-2-2005, Comit Europen de Normalization, Brussels, Belgium.

CSA. 2006. CAN/CSA-S6-06 Canadian Highway Bridge Design Code (CHBDC) and
commentary. Canadian Standards Association (CSA), Mississauga, ON.

Chen, W.F. and Duan, L. 2000. Bridge Engineering Handbook, CRC Press LLC.

Chopra, A.K., and Goel, R.K. 2002. A modal pushover analysis procedure for estimating
seismic demands for buildings. Earthquake Engineering and Structural Dynamics, 31(3):
561582.

Chopra, A.K., and Goel, R.K. 2004. A modal pushover analysis procedure to estimate
seismic demands for unsymmetric-plan buildings. Earthquake Engineering and Structural
Dynamics, 33(8):903927.

Computers and Structures Inc. 1999. SAP2000: Three dimensional static and dynamic
finite element analysis and design of structures. Computers and Structures Inc. Berkeley,
CA.

Cornell, C.A., and Krawinkler, H. 2000. Progress and challenges in seismic performance
assessment.

PEER

Center

News;

3(2):12

URL

http://peer.berkeley.edu/news/2000spring/performance.html (accessed: Jan, 2012).

Cornell, C.A., Jalayer, F., Hamburger, R.O., and Foutch, D.A. 2002. The probabilistic
basis for the 2000 SAC/FEMA steel moment frame guidelines. ASCE Journal of
Structural Engineering, 128 (4): 526533.

322

Dhakal, R.P., and Mander J.B. 2006. Financial risk assessment methodology for natural
hazards. Bulletin of the New Zealand Society of Earthquake Engineering, 39(2): 91-105.

Dutta, A., and Mander, J.B. 1998. Seismic fragility analysis of highway bridges,
Proceedings of the INCEDE-MCEER Center-to-Center Workshop on Earthquake
Engineering Frontiers in Transportation Systems, Tokyo, Japan, pp. 311-325.

Dwairi, H., and Kowalsky, M. 2006. Implementation of inelastic displacement patterns in


direct displacement-based design of continuous bridge structures. Earthquake Spectra,
22(3): 631-662.

Elwood, K. 2004. Modeling failures in existing reinforced concrete columns. Canadian


Journal of Civil Engineering, 31:846859

Ewins, D.J. 2000. Modal testing: theory, practice and application, 2nd Ed., Research
Studies Press, Baldock, UK.

Fardis, M.N., and Biskini, D.E. 2003. Deformation capacity of RC members, as


controlled by flexure or shear. Otani Symposium. 511- 530.

FEMA-356. 2000. Prestandard and commentary for the seismic rehabilitation of


buildings, American Society of Civil Engineers (ASCE). Washington, DC.

FEMA-350. 2000. Recommended seismic design criteria for new steel moment-frame
buildings, SAC Joint Venture, Federal Emergency Management Agency, Washington,
DC.

Fischinger, M., Beg D., Isakovic, T., Tomazevic, M., Zarnic, R. 2004. Performance based
assessment from general methodologies to specific implementations. International
Workshop on PBSD, Bled, Slovenia; 293308 (published in PEER Report 2004-05 (UC
Berkeley)).

323

Fischinger, M., Isakovic, T., and Fajfar, P. 1997, Seismic analysis of viaduct structuresWhich method to choose?, Seismic design methodologies for the next generation of
codes, Fajfar and Krawinkler (eds), Balkema, Rotterdam.

Fischinger, M., and Isakovic,T., 2003. Inelastic Seismic analysis of reinforced concrete
viaducts, Structural Engineering International, 13(2): 111-118.

Goda, K., and Atkinson, G.M. 2011. Seismic performance of wood-frame houses in
south-western British Columbia. Earthquake Engineering and Structural Dynamics, 40
(8): 903-924.

Goda, K., and Atkinson, G.M. 2009. Probabilistic characterization of spatially-correlated


response spectra for earthquakes in Japan. Bull.Seism. Soc. of Am., 99 (5): 3003- 3020.

Goda, K., Hong, H.P., and Atkinson, G.M. 2010. Impact of using updated seismic
information on seismic hazard in western Canada. Canadian Journal of Civil Engineering,
37 (4): 562-575.

Goel, R., and Chopra, A.K. 2004. Extension of modal pushover analysis to compute
member forces. Earthquake Spectra, 21(1):125139.

Goel, R.K., and Chopra, A. 1997. Evaluation of bridge abutment capacity and stiffness
during earthquakes. Earthquake Spectra, 13(1): 1-23.

Gulkan, P., and Sozen, M. 1974. Inelastic response of reinforced concrete structures to
earthquake motions, ACI Journal, 71: 604-610.

Gupta, B., and Kunnath, S.K. 2000. Adaptive spectra-based pushover procedure for
seismic evaluation of structures. Earthquake Spectra, 16(2):367392.

324

Haselton, C., Liel, A., Lange, S.T., and Deierlein, G.G. 2007. Beam- column element
model calibrated for predicting flexural response leading to global collapse of RC frame
buildings. Pacific Earthquake Engineering Research Center 2007/03, University of
California, Berkeley.

Haselton, C.B., and Baker, J.W. 2006. Ground motion intensity measures for collapse
capacity prediction: Choice of optimal spectral period and effect of spectral shape. 8th
National Conference on Earthquake Engineering, San Francisco, California.

Haselton, C.B., and Deierlein, G.G. 2007. Assessing seismic collapse safety of modern
reinforced concrete moment-frame buildings. PEER Report 2007/08, Pacific Engineering
Research Center, University of California, Berkeley, California.

Haselton, C.B., Baker, J.W., Liel, A.B., and Deierlein, G.G. 2011. Accounting for
ground-motion spectral shape characteristics in structural collapse assessment through an
adjustment for epsilon, ASCE J. Struct. Eng., 137 (3), 332.

HAZUS. 1999. Earthquake loss estimation methodology, Technical Manual, National


Institute of Building Sciences for the Federal Emergency Management Agency.

Heaton, T.H., and Hartzell, S.H. 1987. Earthquake hazards on the Cascadia subduction
zone. Science, 236: 162168.

Heaton, T.H., and Kanamori, H. 1984. Seismic potential associated with subduction in
the northwestern United States. Bulletin of the Seismological Society of America, 74:
933944.

Hong, H.P., and Goda, K. 2006. A comparison of seismic-hazard and risk deaggregation.
Bulletin of the Seismological Society of America, 96 (6): 2021-2039.

325

Hyndman, R.D., and Rogers, G.C. 2010. Great earthquakes on Canadas west coast: a
review, Can. J. Earth Sci., 47(5): 801820 .

Ibarra, L.F., Medina, R.A., and Krawinkler, H. 2005. Hysteretic models that incorporate
strength and stiffness deterioration. Earthquake Engineering and Structural Dynamics, 34
(11):1489-1511.

Isakovic, T., and Fischinger, M. 2006. Higher modes in simplified inelastic seismic
analysis of single column bent viaducts. Earthquake Engineering and Structural
Dynamics, 35(1): 95-114.

Isakovic, T., Pompeyo, M., Lazaro, N., and Fischinger, M. 2008. Applicability of
pushover methods for the seismic analysis of single-column bent viaducts. Earthquake
Eng. Struct. Dynamics, 37(8):11851202.

Jalayer, F., and Cornell, C.A. 2003. A technical framework for probability-based demand
and capacity factor (DCFD) seismic formats. PEER-2003/08, Pacific Earthquake
Engineering Research Center, University of California- Berkeley, Berkeley, CA.

Jones, M.H., Salama, A.E., and Dreher, S.T. 2001. BART SFO extension aerial
guideways: Using nonlinear analysis to control and minimize seismic damage, Journal of
Bridge Engineering, 6 (6): 451-460.

Kappos, A., Paraskeva, T., and Sextos, A. 2004. Seismic assessment of a major bridge
using modal pushover analysis and dynamic time-history analysis. Advances in
Computational and Experimental Engineering and Science; 673680.

Kappos, A.J., Manolis, G.D., and Moschonas, I.F. 2002. Seismic assessment and design
of R/C bridges with irregular configuration, including SSI effects, Engineering
Structures, 24 (10):13371348.

326

Kirby, S., Wang, K., and Dunlop, S. 2002. The Cascadia subduction zone and related
subduction systems, U.S. Geological Survey Open-File Report 02328 Geological
Survey of Canada Open File 4350.

Kowalsky, M.J. 1997. Direct displacement-based design: A seismic design methodology


and its application to concrete bridges, PhD thesis, University of California, San Diego.

Kowalsky, M.J. 2002. A displacement-based design approach for the seismic design of
continuous concrete bridges, Earthquake Enginieering and Structural Dynamics, Vol.
31, pp. 719-747.

Liel, A.B., Haselton, C.B., and Baker, J.W. 2009. Incorporating modelling uncertainties
in the assessment of seismic collapse risk of buildings. Structural Safety, 31 (2): 197-211.

Luco, N., and Bazzurro, P. 2007. Does amplitude scaling of ground motion records result
in biased nonlinear structural drift responses?. Earthquake Engineering & Structural
Dynamics, 36(13): 1813-1835.

Maalek, S., Akbari, R., and Maheri, M.R. 2009. The effect of higher modes on the
regularity of single-column-bent highway viaducts. Bridge Structures, 5(1): 2943.

Mackie, K.R. and Stojadinovic, B. 2003. Seismic demands for performance-based design
of bridges. PEER-2003/16, Pacific Earthquake Engineering Research Center, University
of California- Berkeley, Berkeley, Calif.

Mackie, K.R. and Stojadinovic, B. 2007. R-factor parameterized bridge damage fragility
curves. Journal of Bridge Engineering, 12(4): 500510.

Mander, J.B, Priestley, M.J.N., and Park, R. 1988. Theoretical stress-strain model for
confined concrete. ASCE Journal of Structural Engineering, 114(8): 1804-1826.

327

Mander, J.B., Dhakal, R., Mashiko, N., and Solberg, K. 2007. Incremental dynamic
analysis applied to seismic financial risk assessment of bridges. Engineering
Structures, 29(10): 2662-2672.

Maroney, B.H., and Chai, Y.H. 1994. Bridge abutment stiffness and strength under
earthquake loadings. Proceedings, 2nd international workshop on the seismic design of
bridges, Queenstown, New Zealand.

McGuire, R.K. 1995. Probabilistic seismic hazard analysis and design earthquakes:
Closing the loop. Bull. Seism. Soc. Am., 85: 1275-1284.

Mechakhchekh, A. 2008. Seismic risk analysis of bridge structural systems. PhD thesis,
City University of New York, NY.

Megally, S.H., Seible, F., Bozorgzadeh, A., Restrepo, J., and Silva, P.F. 2003. Response
of sacrificial shear keys in bridge abutments to seismic loading. Proceedings of the FIB
Symposium on Concrete Structures in Seismic Regions, Athens, Greece.

Mitchell, D., Bruneau, M., Williams, M., Anderson, D., Saatcioglu, M., and Sexsmith, R.
1995. Performance of bridges in the 1994 Northridge earthquake, Can. J. Civ. Eng.,
22(2): 415-427.

National Research Council of Canada (NRCC). 2010. National Building Code of Canada
(NBCC). Ottawa, Ontario.

Otani, S. 1981. Hysteresis model of reinforced concrete for earthquake response analysis.
Journal of Fac. of Eng., Univ. of Tokyo, Series B, XXXVI-11(2): 407-441.

Pangiotakos, T.B., and Fardis, M.N. 2001. Deformations of reinforced concrete members
at yielding and ultimate, ACI Structural Journal, 98(2): 135-148.

328

Paraskeva ,T. S., Kappos, A. J., and Sextos, A. G. 2006. Extension of modal pushover
analysis to seismic assessment of bridges, Earthquake Engng Struct. Dyn, 35:12691293.

Patwardhan,C. 2005. Shear strength and deformation modelling of reinforced concrete


columns. Master thesis, The Ohio State University.

Paultre, P., and Lgeron, F. 2008. Confinement reinforcement design for reinforced
concrete columns. ASCE, Journal of Structural Engineering, 134(5): 738-749.

Pinho, R. 2007. Nonlinear dynamic analysis of structures subjected to seismic action.


Advanced Earthquake Engineering Analysis, Pecker, A. ed., Springer, NY.

Pinho, R., Antoniou, S., Casarotti, C., and Lopez, M. 2005. A displacement-based
adaptive pushover for assessment of buildings and bridges. NATO International
Workshop on Advances in Earthquake Engineering for Urban Risk Reduction, Istanbul,
Turkey.

Pinho, R., Casarotti, C., Antoniou S. 2007. A comparison of single-run pushover


analysis techniques for seismic assessment of bridges. Earthquake Engineering and
Structural Dynamics, 36(10): 1347-1362.

Pinto, A.V., Verzeletti, G., Pegon, P., Magonette, G., Negro, P., and Guedes, J., 1996.
Pseudo- Dynamic Testing of Large- Scale R/ C Bridges, Report EUR 16378, Ispra ( VA),
Italy.

Priestley, M.J.N. 1993. Myths and fallacies in earthquake engineering Conflicts


between design and reality. Bulletin NZ National Society for Earthquake Engineering, 26
(3): 328-341.

Priestley, M.J.N., and Benzoni, G. 1996. Seismic performance of circular columns with
low longitudinal reinforcement ratios, ACI Structural Journal, 93(4): 474-485.

329

Priestley, M.J.N., Calvi, G.M. and Kowalsky, M.J. 2007. Direct displacement based
design of structures, IUSS Press, Pavia, Italy.

Priestley, M.J.N., Seible, F., and Calvi, G.M. 1996. Seismic design and retrofit of
bridges, John Wiley and Sons, New York.

Restrepo, J.C.O. 2006. Displacement-Based Design of Continuous Concrete Bridge under


Transverse Seismic Excitation. . Master thesis, European School for Advanced Studies in
Reduction of Seismic Risk (ROSE School), Pavia, Italy.

Riddihough, R.P. 1984. Recent movements of the Juan de Fuca plate system. Journal of
Geophysical Research, 89: 69806994.

Riddihough, R.P., and Hyndman, R.D. 1976. Canadas active western margin - the case
for subduction. Geoscience Canada, 3: 269278.

Rogers, G.C. 1988. An assessment of the megathrust earthquake potential of the Cascadia
subduction zone. Canadian Journal of Earth Sciences, 25: 844852.

Saiidi, M., and Sozen, M.A. 1979. Simple and complex models for nonlinear seismic
response of reinforced concrete structures. Rep. No. UILU-ENG-79-2013, Struct. Res.
Series No. 465, Univ. of Illinois, Urbana, Ill.

Saiidi, S., Moore, R.T., and Itani, A. 2001. Seismic performance of reinforced concrete
bridges with unconventional configuration, ACI Structural Journal, 98(5): 717-126.

Sasaki, K.K., Freeman, S.A., Paret, T.F. 1998. Multimode pushover procedure (MMP)
a method to identify the effects of higher modes in a pushover analysis. Proceedings of
the 6th US National Conference on Earthquake Engineering, Seattle; Paper No. 271.

330

Shome, N. and Cornell, C.A. 1999. Probabilistic seismic demand analysis of non-linear
structures. Report No.RMS-35, RMS Program, Stanford University, Stanford.

Tehrani, P. 2012. Seismic analysis and behaviour of continuous reinforced concrete


bridges. PhD thesis, Department of Civil Engineering, McGill University, Montreal.

Tehrani, P., and Mitchell, D. 2012a. Effects of column and superstructure stiffness on the
seismic response of bridges in the transverse direction. Canadian. Journal of Civil
Engineering (in press).

Tehrani, P., and Mitchell, D. 2012b. Seismic response of bridges subjected to different
earthquake types using IDA. Submitted to Journal of Earthquake Engineering.

Tehrani, P., and Mitchell, D. 2012c. Effects of Column Stiffness Irregularity on the
Seismic Response of Bridges in the Longitudinal Direction, submitted to Canadian.
Journal of Civil Engineering.

Tehrani, P., and Mitchell, D. 2012d. Incremental dynamic analysis applied to seismic
performance and risk assessment of RC bridges, Submitted to Seismic Risk Analysis and
Management of Civil Infrastructure Systems, Editors: Tesfamariam, S., and Goda, K.
Woodhead Publishing Limited.

Tehrani, P., Goda, K., Mitchell, D., Atkinson, G.M., and Chouinard, L. E. 2012. Effects
of different record selection methods and earthquake types on the transverse response of
bridges. Submitted to Journal of Earthquake Engineering and Structural Dynamics.

Vamvatsikos, D., and Cornell, C.A. 2004. Applied incremental dynamic analysis.
Earthquake Spectra, 20(2): 523553.

Vamvatsikos, D., and Cornell, C.A. 2002. Incremental dynamic analysis, Earthquake
Engineering and Structural Dynamics, 31 (3): 491-514.

331

Vamvatsikos, D., and Cornell, C.A. 2005. Developing efficient scalar and vector intensity
measures for IDA capacity estimation by incorporating elastic spectral shape information.
Earthquake Engineering and Structural Dynamics, 34 (13): 1573-1600.

Vamvatsikos, D., and Cornell, C.A. 2006, Direct estimation of the seismic demand and
capacity of oscillators with multi-linear static pushovers through Incremental Dynamic
Analysis. Earthquake Engineering and Structural Dynamics, 35(9): 1097-1117.

Vanmarcke, E. H. and Gasparini, D.A. 1976. Simulated earthquake motions compatible


with prescribed response spectra, Research report R76-4, Dept. of Civil Engineering,
Massachusetts Inst. of Technology, Cambridge.

Veneziano, D. 2005. Probability and statistics in engineering, Lecture notes, MIT Course
Number: 1.151, available online at http://ocw.mit.edu.

Zareian, F. 2006. Simplified performance-based earthquake engineering, PhD


Dissertation, Department of Civil and Environmental Engineering, Stanford University.

Zareian, F., and Krawinkler, H. 2007. Assessment of probability of collapse and design
for collapse safety. Earthquake Engineering and Structural Dynamics, 36 (13): 19011914.

Zhao, J.X., Zhang, J., Asano, A., Ohno, Y., Oouchi, T., Takahashi, T., Ogawa, H.,
Irikura, K., Thio, H.K., Somerville, P.G., and Fukushima, Y. 2006 Attenuation relations
of strong ground motion in Japan using site classification based on predominant period.
Bulletin of the Seismological Society of America 96(3): 898-913.

332

You might also like