seismic response prediction of bridges using ...

10NCEE

Tenth U.S. National Conference on Earthquake Engineering Frontiers of Earthquake Engineering July 21-25, 2014 Anchorage, Alaska

SEISMIC RESPONSE PREDICTION OF BRIDGES USING INCREMENTAL DYNAMIC ANALYSIS WITH SUBDUCTION ZONE AND CRUSTAL GROUND MOTION RECORDS Payam Tehrani1 , Katsuichiro Goda2, Denis Mitchell3, Gail M. Atkinson4 and Luc E. Chouinard3 ABSTRACT Typically ground motion records from crustal earthquakes have been used in practice for the seismic performance assessment of structures. For some sites, such as Vancouver and Seattle, subduction earthquakes (i.e., interface and inslab events) with very different characteristics (e.g., spectral content and duration) can occur. The effects of using ground motion records from three different earthquake types on the seismic response predictions for a continuous 4-span reinforced concrete bridge located in Vancouver are investigated. The bridge is designed according to the current Canadian seismic design provisions. The seismic response of the bridge is investigated using Incremental Dynamic Analysis (IDA). IDA is carried out separately for records selected from three different earthquake sources, including shallow crustal events, interface (megathrust) and deep inslab subduction earthquakes. The median structural capacities, in terms of spectral acceleration, are predicted for different damage states of the columns including, yielding, cover spalling, bar buckling and collapse for the three different earthquake types. The sensitivity of the IDA results to the record selection procedures, including the conditional mean spectrum (CMS)based and epsilon-based methods, is also studied. The CMS is developed using the seismic deaggregation results for Vancouver.

1

Post-doctoral Fellow, Dept. of Civil Engineering, McGill University, Montreal, Quebec H3A 0C3, Canada Senior Lecturer, Department of Civil Engineering, University of Bristol, Bristol BS8 1TR, United Kingdom 3 Professor, Department of Civil Engineering, McGill University, Montreal, Quebec H3A 0C3, Canada 4 Professor, Department of Earth Sciences, Western University, London, Ontario N6A 5B7, Canada 2

Tehrani P, Goda K, Mitchell D, Atkinson GM, Chouinard LE. Seismic response prediction of bridges using incremental dynamic analysis with subduction zone and crustal ground motion records. Proceedings of the 10th National Conference on Earthquake Engineering, Earthquake Engineering Research Institute, Anchorage, AK, 2014.

10NCEE

Tenth U.S. National Conference on Earthquake Engineering Frontiers of Earthquake Engineering July 21-25, 2014 Anchorage, Alaska

Seismic Response Prediction of Bridges Using Incremental Dynamic Analysis with Subduction Zone and Crustal Ground Motion Records Payam Tehrani1, Katsuichiro Goda2, Denis Mitchell3, Gail M. Atkinson4 and Luc E. Chouinard3

ABSTRACT Typically ground motion records from crustal earthquakes have been used in practice for the seismic performance assessment of structures. For some sites, such as Vancouver and Seattle, subduction earthquakes (i.e., interface and inslab events) with very different characteristics (e.g., spectral content and duration) can occur. The effects of using ground motion records from three different earthquake types on the seismic response predictions for a continuous 4-span reinforced concrete bridge located in Vancouver are investigated. The bridge is designed according to the current Canadian seismic design provisions. The seismic response of the bridge is investigated using Incremental Dynamic Analysis (IDA). IDA is carried out separately for records selected from three different earthquake sources, including shallow crustal events, interface (megathrust) and deep inslab subduction earthquakes. The median structural capacities, in terms of spectral acceleration, are predicted for different damage states of the columns including, yielding, cover spalling, bar buckling and collapse for the three different earthquake types. The sensitivity of the IDA results to the record selection procedures, including the conditional mean spectrum (CMS)based and epsilon-based methods, is also studied. The CMS is developed using the seismic deaggregation results for Vancouver.

Introduction Incremental Dynamic Analysis (IDA) [1] can be used for the probabilistic seismic performance assessment of structures [2]. One of the important issues in the IDA is the selection of input ground motion records, since spectral shape of the selected records can significantly influence the IDA results. To include the spectral shape effects in record selection the conditional mean spectrum (CMS) -based [3] and epsilon-based record selection methods [4] are used in this research for three different earthquake types. For the CMS-based method the records are selected to match a CMS (conditioned on the fundamental period of the structure, T1) [3]. Typically, only records from crustal earthquakes are used for design and evaluation of structures in current seismic practice [2], while for some sites, such as Vancouver and Seattle three different earthquake types (i.e., crustal, subduction interface, and inslab events) having very different 1

Post-doctoral Fellow, Dept. of Civil Engineering, McGill University, Montreal, Quebec H3A 0C3, Canada Senior Lecturer, Department of Civil Engineering, University of Bristol, Bristol BS8 1TR, United Kingdom 3 Professor, Department of Civil Engineering, McGill University, Montreal, Quebec H3A 0C3, Canada 4 Professor, Department of Earth Sciences, Western University, London, Ontario N6A 5B7, Canada 2

Tehrani P, Goda K, Mitchell D, Atkinson GM, Chouinard LE. Seismic response prediction of bridges using incremental dynamic analysis with subduction zone and crustal ground motion records. Proceedings of the 10th National Conference on Earthquake Engineering, Earthquake Engineering Research Institute, Anchorage, AK, 2014.

physical characteristics (e.g., spectral content and duration) can occur. The effect of considering different earthquake types in IDA for the seismic performance evaluation of a bridge located in south-western British Columbia is investigated in this research. Record database 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 minimum limits on the magnitude of 6.0, peak ground acceleration (PGA) of 0.1 g and peak ground velocity (PGV) of 10 cm/sec are imposed to consider strong ground motions available in the databases. 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 including 111 interface and 78 inslab pairs of records) are described by Goda and Atkinson [5]. More detailed information regarding the records selected for this research is provided there. To perform the IDA, 44 horizontal components of the records (i.e., 22 pairs of records) are chosen using different record selection methods. In Fig. 1 the mean and logarithmic standard deviation (i.e., record-to-record variability) are presented for different event types. The record-to-record variability is typically larger for the subduction events. 1.4

0.9

1.2

Sa (T) (g)

1

Logarithmic standard deviation

a) Crustal (as-recorded) Interface (as-recorded) Inslab (as-recorded)

0.8 0.6 0.4 0.2 0 0

1

2

Period (T)

3

4

5

b)

0.8 0.7 0.6 0.5

Crustal (as-recorded) interface (as-recorded) Inslab (as-recorded)

0.4 0.3 0

1

2

Period (T)

3

4

5

Fig. 1. Record properties: a) mean spectrum; b) logarithmic standard deviation Bridge properties and modeling To investigate the impact of the record selection methods and the type of records on the IDA results, a regular 4-span bridge, as shown in Fig. 2, is considered in this study. The bridge is designed according to the 2006 CHBDC [6] with an importance factor of I =1.5 (i.e., emergencyroute bridge). The failure of shear keys is treated indirectly by considering two cases of restrained and unrestrained transverse movements at the abutments. For the restrained case (shear keys effective), the first period of the structure in the transverse direction is around T1=0.7 sec. For the case of failed shear keys, the first period of the bridge is T1=1.3 sec. For straight bridges the responses in the orthogonal directions are likely to be essentially independent [7]. The responses of the bridge in the transverse direction are chosen to assess the significance of the different ground motion types and different record selection methods. Several damage states are considered in the seismic evaluation of the bridge under study including yielding, cover spalling, bar buckling and dynamic instability (regarded as collapse). The damage states in this study are predicted using empirical equations developed by Berry and

Eberhard [8]. The serviceability limit state is predicted based on the criteria given by Priestley et al. [7]. More detail regarding the determination of the damage states for the bridge under study is given by Tehrani [9]. The modified Takeda hysteresis model [10] is used in this study to model the behavior of the ductile RC columns using Ruaumoko software [11]. The structural modeling considered in this study is similar to that used by Priestley et al. [7], except that the backbone curve including the post peak response is used for the IDA [9]. The bridge under study is designed and detailed to meet the code requirements for ductile response, including capacity design concepts and adequate support lengths at the abutments, hence is likely to fail by flexural yielding, rather than brittle failure mechanisms and unseating failures. For this continuous bridge, with all other failure modes avoided, sidesway collapse is the governing collapse mechanism. The collapse prediction is based on dynamic instability of the structure [1]. More detail on design and modeling of the bridge is given by Tehrani [9].

Fig. 2. Bridge properties Conditional mean spectrum The CMS provides the expected response spectrum, conditioned on occurrence of a target spectral acceleration value at the period of interest. The CMS can be used as an appropriate target response spectrum for selecting ground motions as input for dynamic analyses [3]. In the development of a CMS, some important aspects of the records including magnitude, M, distance, R, and epsilon, ε, are considered from the deaggregation of seismic hazard. Baker [3] proposed a method for calculating the CMS, and this approach is used in this study. For the development of a CMS, the mean values of spectral accelerations at different periods are computed using an appropriate ground motion prediction equation (GMPE). These mean values will then be modified considering the inter-period correlations, standard deviations of spectral accelerations and mean epsilon values at different periods. A complicated aspect in constructing a CMS for a site in south-western British Columbia, compared 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”, “CMSInterface”, and “CMS-Inslab” [5]. The seismic deaggregation analysis is carried out based on the updated seismic hazard data for Vancouver (site class C) provided by Atkinson and Goda [12]. Seismic deaggregation can be carried out using either the matching or exceeding methods [13]. Seismic deaggregation results for different periods using different GMPEs and matching/exceeding methods are given by Tehrani [9]. In this research the deaggregation analysis based on an “approximately equal criterion” (i.e., matching method) [13] is adopted, where seismic events reaching a seismic intensity level between 90% and 110% of the target

Sa(T) value are used to produce deaggregation results. For the purpose of this study the GMPE by Boore and Atkinson (BA08 GMPE) [14] is used for crustal events. For subduction events (interface and inslab), the GMPE by Zhao et al. (Z06 GMPE) [15], and Atkinson and Boore (AB03 GMPE) [16] are considered separately (i.e., values are not combined) to predict the CMS. These GMPEs are the main GMPEs used in seismic hazard analysis by Atkinson and Goda [12]. The CMS for T1=0.7 sec obtained using the matching method at 2% and 0.5% probability of exceedance in 50 years are illustrated in Figs. 3a and 3b, respectively. The CMS-Interface has rich spectral content in the long period range, which is found to be the most critical case for the bridge structure studied, while the CMS-Inslab has rich spectral content in the short period range. The normalized CMS for interface and inslab events using the Z06 and AB03 GMPEs are shown in Fig. 4a and 4b, respectively. For the case of the interface events the use of the Z06 GMPE results in higher predictions for all period ranges. However, for the case of the inslab events (Fig. 4b), the use of the Z06 GMPE results in higher predictions for shorter period ranges and lower predictions for the longer period range compared to the spectral values obtained using the AB03 GMPE. In Fig. 5a the CMS curves computed using the exceeding method are presented. The spectral values of the predicted CMS curves for this case exceed those from the UHS at the target period. In addition, in Fig. 5b the normalized CMS curves obtained using the matching and exeeding methods are compared. The normalized spectral values of the CMS obtained using the exceeding method are smaller in both long and short period ranges. This is because the exceeding method takes into account seismic events with seismic intensities exceeding the target Sa(T) value alone; thus higher epsilon values are predicted, resulting in more peaked spectral shapes. Therefore, the use of the exceeding method may result in an overestimatation of the spectral shape effects. a)

2.5

CMS-Crustal CMS-Interface

2

Spectral acceleration (g)


2.5

CMS-Inslab

1.5

UHS (2% in 50 years)

1 0.5

b)

2

CMS-Crustal CMS-Interface CMS-Inslab

1.5

UHS (0.5% in 50 years)

1 0.5 0

0 0.1

0.7

1.3 1.9 Period, T (Sec)

0.1

2.5

0.7


2.5

Fig. 3. UHS and CMS for different events at T1=0.7 sec: a) 2% in 50 years; b) 0.5% in 50 years 2

Sa(T) / Sa(0.7)

2

a)

1.5

CMS-Interface (AB03)

1

b)

1.5

CMS-Inslab (AB03)

1

CMS-Interface (Z06)

0.5

CMS-Inslab (Z06)

0.5

0 0

1

Period, T (sec)

2

3

0 0

1

Period, T (sec)

2

Fig. 4. Normalized CMS at T1=0.7 sec for a) interface events; b) inslab events

3

As shown in Figs. 6a and 6b, the CMS-Average, which is the weighted average of CMS-Crustal, CMS-Interface, and CMS-Inslab by considering the relative influences of the individual earthquake types, is close to the CMS-Crustal for different periods and different probabilities of exceedance. These two CMS become even more similar, when the Z06 GMPE is used for the subduction events. This may indicate that the CMS-Crustal can be substituted for the CMSAverage. However, no general conclusions can be made, as this may be specific to the cases considered in this research.

a)

2

2

CMS-Crustal (Exceeding method) Interfacel (Exceeding method) Inslabl (Exceeding method)

1.5

Sa(T) / Sa(0.7)


2.5

UHS (0.5% in 50 years)

1 0.5

Crustal (exceeding method) Crustal (matching method) Interface (exceeding method) Interface (matching method) Inslab (exceeding method) Inslab (matching method)

b)

1.5 1 0.5

0

0 0.1

0.7

1.3 1.9 Period, T (sec)

2.5

0.1

0.7

1.3 1.9 Period, T (sec)

2.5

Fig. 5. a) CMS using exceeding method; b) normalized CMS using different methods 1.5

CMS-Crustal (0.5% in 50 years) CMS-Average (Z06; 0.5% in 50 years) CMS-Average (AB03; 0.5% in 50 years) CMS-Crustal (2% in 50 years) CMS-Average (Z06; 2% in 50 years) CMS-Average (AB03; 2% in 50 years)

1.5

b)

1

1

CMS-Crustal (0.5% in 50 years) CMS-Average (Z06; 0.5% in 50 years) CMS-Average (AB03; 0.5% in 50 years) CMS-Crustal (2% in 50 years) CMS-Average (Z06; 2% in 50 years) CMS-Average (AB03; 2% in 50 years)

0.5

0.5

p

Sa(T) (g)

(g)

a)

0

0.1

0.7


2.5

0

0.1

0.7


2.5

Fig. 6. Comparison of CMS-Average and CMS-Crustal at a) T1=0.7 sec; b) T1=1.3 sec Spectral shapes of the records and epsilon values Epsilon, ε, is computed by subtracting the mean predicted lnSa(T) from the record’s lnSa(T), and dividing by the logarithmic standard deviation as predicted by the GMPE). It has been shown that ε(T1) (i.e., epsilon value of the ground motion record at the fundamental period) is a proxy of the spectral shape [4]. As the epsilon value increases, the spectral shape tends to be more peaked. Rare ground motions that may cause modern structures to collapse have peaked spectral shapes that are much different than a standard uniform hazard spectral shape; accounting for this shape has been shown to increase the predicted collapse capacity significantly [4]. The most direct approach to account for spectral shape in structural analysis is to select ground motions with ε(T1) values similar to the mean ε values obtained from seismic hazard deaggregation analysis for the site and hazard level of interest (referred to as the epsilon-based method). In this

Sa(T) / Sa(T=1.0)

0.2 0.1

Period, T (Sec)

1

ε=0 ε = 0.5 ε=1 ε=1.5

b) Interface Sa(T) / Sa(T=1.0)

a) Crustal

2

2

ε=0 ε = 0.5 ε=1 ε= 1.5

Sa(T) / Sa(T=1.0)

research the epsilon-based method as described by Baker and Cornell [4] is adopted. In Fig. 7 the influence of the ε values on the average normalized spectra of the records is illustrated. The average geometric mean response spectra of 50 ground motion records (22 records for inslab events) with the average ε(T1) values being similar to the target epsilon values are presented. As the ε value increases, the resulting average response spectrum of the records becomes more peaked (i.e., spectrum is more peaked at the fundamental period). For the crustal and interface events, this is observed for both long and short period ranges. However, for the case of the inslab events, the spectral shape in long period ranges is relatively insensitive to the epsilon values (Fig. 7c). This may be primarily due to weak frequency content of such events in long period ranges. c) Inslab

5

ε=0 ε = 0.5 ε=1 ε= 1.5

0.5

0.05

0.2 0.1

Period, T (Sec)

1

0.1

Period, T (Sec)

1

Fig. 7. Normalized mean spectra of records at T=1 sec: a) crustal; b) interface; c) inslab events IDA results for the bridge with restrained abutment conditions The IDA results obtained using the crustal records are presented for the case of the CMS-based and epsilon-based methods in Figs. 8a and 8b and Tables 1a and 1b, respectively. The predicted median capacities at bar buckling and collapse obtained using the CMS-based method are about 30% smaller than those obtained using the epsilon-based method. These differences increase for lower IDA percentiles. For the cover spalling damage state, the difefrence between the methods is about 20%, while for the yielding damage state the predictions are almost equal. IDA is also carried out using the interface records. In Figs. 9a and b and Tables 2a and b, the IDA results obtained for the cases of the CMS-based method using the Z06 and AB03 GMPEs are presented. As shown in Fig. 4a, the spectral values computed using the Z06 GMPE are higher than those obtained using the AB03 GMPE. The collapse capacity obtained using the Z06 GMPE is about 35% smaller than that obtained using the AB03 GMPE. However, the differences are much smaller for the cover spalling and yielding damage states. For the case of the CMS using the AB03 GMPE, the variability of the IDA results is larger than that obtained using the Z06 GMPE. This is because the records used in this study are more compatible with the Z06 GMPE than the AB03 GMPE and a better spectral matching is possible for the former case. The IDA results obtained using the epsilon-based record selection for the interface event are shown in Fig. 9c and Table 2c. The epsilon values are predicted using the Z06 GMPE. The median collapse capacity obtained using the CMS-based method (Fig. 9a) is about 20% smaller than that predicted using the epsilon-based method (Fig. 9c), when the Z06 GMPE is used in both methods. However, the predictions for the spalling and yielding damage states are similar. Similarly, the differences between the two methods increase, when lower IDA percentiles (e.g., 16% percentiles) are compared.

The IDA results obtained using the inslab records are shown in Fig. 10 for the cases of the CMSbased method using the Z06 GMPE, CMS-based method using the AB03 GMPE, and epsilonbased method. As presented in Fig. 4b, for the inslab events the spectral values computed using the AB03 GMPE are higher than those obtained using the Z06 GMPE for periods longer than the target period. Therefore, the use of the AB03 GMPE results in lower capacity predictions for the inslab events. For the bridge studied, the median collapse capacity obtained based on the CMSbased record selection using the AB03 GMPE is about 15% smaller than that obtained using the Z06 GMPE (see Table 3). Similar to the case of the interface events, the discrepancy caused by different GMPEs are smaller for the yielding and spalling damage states than the severer damage states. The IDA results obtained using the epsilon-based record selection for the case of the inslab events are listed in Table 3c, where the epsilon values are predicted using the Z06 GMPE. The median collapse capacity obtained using the CMS-based method (Table 3a) is similar to that predicted using the epsilon-based method (Table 3c), when the Z06 GMPE is used in both methods. Typically the use of the epsilon-based record selection method results in higher collapse capacity predictions, as demonstrated for the crustal and interface events. However, for the inslab events the spectral shapes of the records are relatively insensitive to the epsilon values (see Fig. 7c). As a result, the predictions from the CMS-based and epsilon-based methods are similar. The variability for the case of the inslab events is greater than that for the other events. 3.5 3

50% Percetile

b) Epsilon-based (Crustal)

Records

3

16% Percetile

2.5

Sa ( 0.7 ) (g)

3.5

a) CMS-based (Crustal)

Records

16% Percetile 50% Percetile

2.5

84% Percetile

84% Percetile 2

2

1.5

1.5

1

1

0.5

0.5 0

0 0

0.02

0.04 0.06 Maximum column drift

0.08

0.1

0

0.02


0.08

0.1

Fig. 8. IDA results for crustal events at T1=0.7 sec: a) CMS-based; b) epsilon-based method Table 1. IDA results for crustal events at T1=0.7 sec: a) CMS-based; b) epsilon-based method a) Damage states

CMS-based (Crustal) IDA percentiles (g) 50%

84%

16%

St Dev

b) Damage states

Epsilon-based (Crustal) IDA percentiles (g) 50%

84%

16%

St Dev

Yielding

0.15 0.14 0.15

0.08

Yielding

0.14 0.13 0.14

0.09

Serviceability

0.27 0.25 0.33

0.12

Serviceability

0.28 0.25 0.32

0.14

Cover Spalling

0.52 0.41 0.68 1.27 0.89 1.80 1.31 0.95 1.93

0.25 0.41 0.41

Cover Spalling

0.66 0.47 0.81 1.73 1.10 2.41 1.84 1.12 2.62

0.28 0.41 0.42

Bar buckling Collapse


A comparison of the IDA results obtained using three event types for the case of the bridge with restrained abutments indicates that the use of the interface events results in the lowest predictions of structural capacity. For the case of the epsilon-based method, the collapse predictions obtained using the interface records are about 47% smaller than those obtained using the crustal records. The use of the inslab records is not critical and results in the highest capacity predictions.

3

3

2.5

1.5

50% Percetile

1.5

1.5

1

1

0.5

0.5

0.5

0

0

0.02

0.04 0.06 Maximum column drift (DM)

0.08

0.1

0

84% Percetile

2

84% Percetile

1

0

16% Percetile

50% Percetile 2

84% Percetile

c) Epsilon-based (Interface)

Records 2.5

16% Percetile

50% Percetile

(IM)

3

b) CMS-Interface (AB03 GMPE)

Records

2.5

16% Percetile

2

Sa (0.7)

(g)

a) CMS-Interface (Z06 GMPE)

Records

0.02

0.04

0.06

0.08

0.1

0 0

Maximum column drift (DM)

0.02

0.04 0.06 Maximum column drift (DM)

0.08

Fig. 9.

IDA results for interface events at T1=0.7 sec: a) CMS-based method (Z06 GMPE); b) CMS-based method (AB03 GMPE); c) epsilon-based method ( Z06 GMPE)

Table 2.

IDA results for interface events at T1=0.7 sec: a) CMS-based method (Z06 GMPE); b) CMS-based method (AB03 GMPE); c) epsilon-based method (Z06 GMPE) CMS-based (Z06) IDA percentiles (g)

a)

St Dev

Damage states

St Dev

Damage states

0.13 0.11 0.15

0.27

Yielding

0.13 0.08 0.15

0.36

Yielding

0.13 0.06 0.15

0.60

Serviceability

0.27 0.22 0.32

0.24

Serviceability

0.24 0.17 0.28

0.30

Serviceability

0.27 0.13 0.29

0.57

Cover Spalling

0.51 0.38 0.77 1.11 0.72 1.88 1.17 0.76 1.97

0.34 0.47 0.49

Cover Spalling

0.42 0.32 0.53 0.77 0.59 1.05 0.77 0.61 1.12

0.27 0.32 0.33

Cover Spalling

0.43 0.31 0.63 0.92 0.55 1.78 0.97 0.56 1.83

0.54 0.57 0.56


50%

84%

16%


6

50%

6

a) CMS-Inslab (Z06 GMPE)

Records 16% Percetile

5


4

16% Percetile

4

84% Percetile

3

3

2

2

2

1

1

1

0

0

0.02

0.04

0.06

0.08

0.1

0

0.02

St Dev

0.04

0.06

0.08

0.1

84% Percetile

0 0

0.02

Maximum column drift

Maximum column Drift

16%

50% Percetile

3

0

84%

c) Epsilon-based (Inslab)

Records 5

50% Percetile

84% Percetile

50%

6

16% Percetile

50% Percetile 4

16%

b) CMS-Inslab (AB03 GMPE)

Records

5

84%

Epsilon-based (Z06) IDA percentiles (g)

c)

Yielding

Damage states

Sa ( 0.7 ) (g)

CMS-based (AB03) IDA percentiles (g)

b)


0.08

Fig. 10.

IDA results for inslab events at T1=0.7 sec: a) CMS-based method (Z06 GMPE); b) CMS-based method (AB03 GMPE); c) epsilon-based method (Z06 GMPE)

Table 3.

IDA results for inslab events at T1=0.7 sec: a) CMS-based method (Z06 GMPE); b) CMS-based method (AB03 GMPE); c) epsilon-based method (Z06 GMPE) CMS-based (Z06) IDA percentiles (g)

a) Damage states

50%

84%

16%

b)

CMS-based (AB03) IDA percentiles (g)

St Dev Damage states 50%

84%

16%

0.14 0.08 0.37

0.69

Yielding

0.15 0.09 0.24

0.58

Serviceability

0.27 0.16 0.57

0.61

Serviceability

0.27 0.16 0.48

Cover Spalling

0.66 0.32 1.13 1.97 0.71 4.09 2.08 0.76 4.09

0.66 0.74 0.72

Cover Spalling

0.59 0.31 1.04 1.63 0.60 4.08 1.76 0.62 4.09

Collapse


c)

Epsilon-based (Z06) IDA percentiles (g)


Yielding

Bar buckling

0.1

84%

16%

St Dev

Yielding

0.13 0.09 0.21

0.67

0.52

Serviceability

0.27 0.17 0.38

0.62

0.60 0.78 0.78

Cover Spalling

0.69 0.32 1.00 1.97 0.72 4.34 2.06 0.73 4.77

0.70 0.76 0.76


0.1

IDA results for the bridge with free abutments The IDA is carried out for the case of the bridge with free abutment conditions. The IDA results obtained for the case of the crustal events using the CMS-based method are presented in Table 4a. The median collapse capacity predicted using the CMS-based method is about 32% smaller than that obtained using the epsilon-based method for this case [9]. For the case of the interface and inslab events, the use of the CMS-based and epsilon-based method results in similar predictions. The IDA results obtained using the CMS-based record selection are presented in Tables 4b and 4c for the interface and inslab events, respectively. A comparison of the IDA results obtained using three different earthquake types shows that the predictions obtained using the interface events are more critical than those obtained using the crustal and inslab events. The predicted collapse capacity for this case obtained using the interface records is about 35% smaller than that obtained using the crustal reccords. Table 4. Results for CMS-based method at T=1.3 sec for: a) crustal; b) interface; c) inslab events a) Damage states

CMS-based (Crustal) IDA percentiles (g) 50%

84%

16%

Yielding

0.10 0.09 0.12

Serviceability

0.19 0.17 0.23

Cover Spalling

0.39 0.29 0.45 0.72 0.61 0.97 0.73 0.63 0.98


CMS-based (Interface, Z06) IDA percentiles (g)

b)


84%

16%

0.16

Yielding

0.09 0.07 0.12

0.15

Serviceability

0.17 0.13 0.22

0.22 0.28 0.28

Cover Spalling

0.28 0.19 0.35 0.47 0.31 0.88 0.47 0.32 0.90


c)

CMS-based (Inslab, Z06) IDA percentiles (g)


84%

16%

St Dev

0.24

Yielding

0.11 0.07 0.24

0.67

0.24

Serviceability

0.19 0.12 0.44

0.64

0.33 0.47 0.57

Cover Spalling

0.37 0.18 0.86 0.88 0.37 2.73 0.89 0.37 2.81

0.71 0.93 0.93


Conclusions Conclusions from this study are summarized as follows: 1- IDA results are sensitive to the type of records used. The use of the records from interface events results in the lowest collapse capacity predictions for the bridge studied. This is attributed to the rich frequency content and long duration of such ground motion records. 2- The use of the epsilon-based method results in higher capacity predictions with larger variability compared to the CMS-based method for most cases studied, except for the case where inslab records are used. The differences between different record selection methods are larger at severe/ultimate damage states (i.e., bar buckling and collapse) than those at yielding and spalling states. Also the differences are greater for lower IDA percentiles. 3- The use of different GMPEs for the CMS-based method results in 35% and 15% differences of collapse capacity for the case of the interface and inslab events, respectively, at collapse. For yielding and cover spalling the differences are much smaller. The use of the Z06 GMPE and AB03 GMPE results in lower capacity predictions for the case of the interface and inslab events, respectively. 4- The use of the exceeding method for seismic deaggregations results in an overestimation of the spectral shape effects. Therefore, the use of the matching method is recommended.

5- The average weighted CMS predicted using three different event types is similar to the CMSCrustal for the two different periods and different hazard levels studied in this research. 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 were obtained from the K-NET, KiK-NET and PEER-NGA databases. References 1.

Vamvatsikos D, Cornell CA. Incremental dynamic analysis. Earthquake Engineering and Structural Dynamics 2002; 31 (3): 491-514.

2.

ATC-63. Quantification of Building Seismic Performance Factors, ATC-63 Project Report, Applied Technology Council: Redwood City, California, 2008.

3.

Baker JW. The conditional mean spectrum: A tool for ground motion selection. Journal of Structural Engineering 2011; 137 (3), 322-331.

4.

Baker JW, Cornell CA. Spectral shape, epsilon and record selection. Earthquake Engineering & Structural Dynamics 2006; 34 (10): 1193-1217.

5.

Goda K, Atkinson GM. Seismic performance of wood-frame houses in south western British Columbia. Earthquake Engineering and Structural Dynamics 2011; 40 (8): 903-924.

6.

Canadian Standards Association, CAN/CSA-S6-06. Canadian Highway Bridge Design Code and Commentary. CSA: Mississauga, ON, 2006.

7.

Priestley MJN, Calvi GM, Kowalsky MJ. Direct Displacement Based Design of Structures. IUSS Press: Pavia, Italy, 2007.

8.

Berry MP, Eberhard MO. Performance Modeling Strategies for Modern Reinforced Concrete Bridge Columns. PEER-2007/07, Pacific Earthquake Eng. Research Center: University of California- Berkeley, Berkeley, California, 2007.

9.

Tehrani P. Seismic analysis and behaviour of continuous reinforced concrete bridges, PhD thesis, Department of Civil Engineering, McGill University: Montreal, Canada, 2012.

10. Otani S. Hysteresis model of reinforced concrete for earthquake response analysis. Journal of Fac. of Eng., Univ. of Tokyo 1981, Series B XXXVI-11 (2): 407-441. 11. Carr A. RUAUMOKO: A Computer Program for Inelastic Dynamic Analysis, Department of Civil Engineering, University of Canterbury: New Zealand, 2009. 12. Atkinson GM, Goda K. Effects of seismicity models and new ground-motion prediction equations on seismic hazard assessment for four Canadian cities. Bull. Seism. Soc. Am. 2011; 101 (1): 176-189. 13. Hong HP, Goda K. A comparison of seismic-hazard and risk deaggregation. Bulletin of the Seismological Society of America 2006; 96 (6), 2021-2039. 14. Boore DM, Atkinson GM. 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 2008; 24 (1): 99138. 15. Zhao JX, Zhang J, Asano A, Ohno Y, Oouchi T, Takahashi T, Ogawa H, Irikura K, Thio HK, Somerville PG, Fukushima Y. Attenuation relations of strong ground motion in Japan using site classification based on predominant period. Bulletin of the Seismological Society of America 2006; 96 (3): 898-913. 16. Atkinson GM, Boore DM. Empirical ground-motion relations for subduction zone earthquakes and their application to Cascadia and other regions. Bull. Seism. Soc. Am. 2003; 93 (4): 1703-1729.