Sunteți pe pagina 1din 7

Front. Archit. Civ. Eng. China 2007, 1(3): 267273 DOI 10.

1007/s11709-007-0034-y

RESEARCH ARTICLE

ZHU Xi, WANG Jianmin

Seismic performance of viaducts with probabilistic method

Higher Education Press and Springer-Verlag 2007

Abstract Due to the uncertainty of both ground motions and structural capacity, it is necessary to consider the seismic performance of viaduct structures using the probabilistic method. The risk is quantified by a procedure on the basis of a numerical determination of the fragility curves. A group of ground motions, Large Magnitude-Short Distance Bin (LMSR-N), selected specially due to its response spectra, is accorded well with the corresponding spectra of the Chinese code for seismic design. The characteristic values of the curvature ductility factors for the serviceability and the damage control limit states are obtained, and two equations for estimating the characteristic values of the curvature ductility factors are developed through regression analysis. Then, the serviceability and damage control limit states were proposed. Three damage states were constituted according the results of the experiment by Pacific Earthquake Engineering Research (PEER) Center. The analytical fragility curves were obtained specifically, using both Capacity Spectrum Method (CSM) (non-linear static) analysis and Ineremental Dynamic Method (IDM) (non-linear dynamic) analysis, respectively, in this paper. The structural fragility curves developed by CSM method can help make the structural analysis simple and quick, avoiding the implementation of the dynamic response history analysis (RHA). Although the dynamic RHA requires a lot of complicated analysis for the structure, the results from RHA are reliable and accurate. Fragility curves are powerful tools for use in performance-based seismic bridge design. Keywords viaduct structure, seismic design, probabilistic method, fragility, capacity spectrum method, nonlinear dynamic analysis

earthquakes. It is known that the seismic performance of transportation systems played a key role in past earthquake emergency management. Hence, it is necessary to evaluate both physical and functional aspects of bridges and viaducts. In the content of evaluating the seismic performance of bridges and viaducts in a specified hazardous environment, fragility curves are found to be useful tools for predicting the extent of probable damage [1]. A fragility curve defines the probability of structural damage due to earthquakes as a function of earthquake intensity. This study relates to method of developing analytical fragility curves based on the numerical consideration of both structural parameters and variation of input ground motions. For a Chinese viaduct type, the analytical fragility curves are obtained specifically, using both CSM (non-linear static) analysis and IDM (non-linear dynamic) analysis method.

Ground motions selection

Introduction

Bridges and viaducts are potentially among the most seismically vulnerable structures in the highway system during
Received March 8, 2007; accepted June 5, 2007 ZHU Xi ( ), WANG Jianmin School of Civil Engineering and Architecture, Beijing Jiaotong University, Beijing100044, China E-mail: xzhu@center.njtu.edu.cn

Considering the large variation in the intensity and frequency content of ground motions, there is no option but the use of statistical means to predict seismic demands. Therefore, the selection of representative ground motions is a critical aspect of any demand study. For this study a set of 40 recorded ordinary ground motions from Californian earthquakes with moment magnitude between 6.5 and 7 and closest distance to the fault rupture between 13 and 40 km is utilized, denoted as LMSR-N [2]. All ground motions were recorded on National Earthquake Hazards Reduction Program (NEHRP) site class D [3]. In order to compare, the acceleration response spectra of single ground motion, the average of the group and the defined in Code GB50011-2001 were convert to dynamical amplification factor spectra (damping ratio j = 5%) [4]. The objective acceleration response spectrum is considered as the standard response spectrum given by the Code under the third category of site and the first category of seismic design (characteristic period value Tg = 0.45 s). The dynamical amplification factor spectra of single ground motion, the average of the group and the object are shown in Fig. 1. It is illustrated that the average values of the acceleration response spectrum of inputs corresponded well with the objective response spectrum, and the level of fit is much better

268
5 4.5 4 Dynamical amplification factor b 3.5 3 2.5 2 1.5 1 0.5 0 Individual ground motion Median GB500112001(soil type III)

0.5

1.5

2.5

3.5

T /s
Fig. 1 Dynamical amplification factor spectra for the ground motions and GB50011-2001 spectrum

especially when period is less than 1.5 s. This results from the error weight which is developed in the short period stage at the program of selecting inputs. The reason to do so is that the acceleration response spectrum is raised artificially in the Code in the long period stage. Therefore, it is reasonable to control the relative error for response spectra at the short period stage because it is stricter than that in the long period stage.

14 000 1 750 1 000 9 600 1 500

Target viaduct and analytical model

Figure 2 shows a target viaduct consisting of a 5-span continuous deck supported by reinforced concrete columns and spread footing. Each span is 30 m long and 14 m wide. Because the geometric condition is nearly the same along the bridge axis, only column (P1) is analyzed here as shown in Fig. 3. The column is modeled as cantilevers with lumped masses. The column is assumed to have two segments: the first segment is a reinforced uniformly circular concrete flexural column LF, from the top of footing to the bottom of the cap beam, and the length of this flexural column LF is 12.2 m; the second one is assumed rigid LR, to model the cap beam and the deck as shown in Fig. 4. Three masses are considered in calculating the lumped mass (M): the mass of the superstructure (Md), the mass of the cap beam (Mcb), and
530 m=150 m 10 m

Fig. 3 Target viaduct (unit: mm)

M LR 2 200 75

LF
14.7 m 14.7 m

14.7 m

14.7 m

Section A-A

P1

P2

P3

P4

Fig. 2 Target viaduct

Fig. 4 Analytical model (unit: mm)

12 000

269 50% of the mass of the column (Mc/2). The length of the rigid segment, LR, is determined by finding the equivalent center of gravity of the above three masses. Table 1 shows the magnitude and location of the individual masses and the equivalent mass. The center of gravity of the equivalent mass is located at 2745 mm, LR, from the bottom of the cap beam, which is modeled as a rigid zone. Therefore, the total length of the column, L, including the rigid zone is 15.7 m.
Table 1 Lumped mass and length of the rigid segments
Mass Magnitude /kN Location /m Md 5 480.8 Mcb 994.5 Mc 1 136.2 M 7 043.4
400 350

Stress ss /MPa

300 250 200 150 100 50 0 0.00 0.02 0.04 0.06 0.08 0.10

Strain es

4.34

1.47

3.58

Fig. 6 Reinforcing steel constitutive relationship

The flexural segment, LF, is modeled in OpenSees using fiber nonlinear beam-column elements and a fiber element model using one element along the height of the flexural segment with five integration points. Column cross sections were discretized into 160 and 40 radial defined fibers in the core and cover concrete, respectively (Fig. 4). Theres a hint of the longitudinal steel bars in Fig. 4. For the properties of the materials used, the analyses were conducted using Rck = 30 MPa for concrete and fyk = 300 Mpa for the steel. Constitutive models used for concrete were based on the Kent-Scott-Park stress-strain relation [5], as modeled in OpenSees. For this study, the maximum confined concrete stress was determined from the Mander confined concrete model [6]. The confined and unconfined concrete stress-strain relationship is shown in Fig. 5.
30 25 Stress sc /MPa 20 15 10 5 0 0.000 0.005 0.010 0.015 Strain ec 0.020 0.025 Confined concrete Uncnofined concrete

Fig. 5 Confined and unconfined concrete constitutive relationship

terms of the uncertainty in the deformation capacity of the material and the geometric parameters are modeled as random variables. Using the Latin hypercube sampling (LHS) simulation, the probabilistic parameters of the limit states of dimensionless yielding, serviceability and damage control curvature are calculated for different axial load ratios, longitudinal reinforcement ratios and transverse reinforcement ratios. The results indicate that the dimensionless curvature limit state generally follows a normal distribution, that the axial load ratio is the main factor affecting the mean value of the dimensionless yielding and serviceability curvature limit states, and that the axial load ratio and transverse reinforcement ratio are the crucial factors associated with the dimensionless damage control curvature limit state. To implement performance-based earthquake engineering (PBEE), it is necessary to assess the probability of reaching multiple damage states in structural and nonstructural elements. To quantitatively implement PBEE for reinforced concrete columns, it is necessary to predict the likelihood of reaching important damage states at particular levels of the column. The characteristic values of the curvature ductility factors for serviceability and damage control limit states are respectively obtained. Two equations to estimate the curvature ductility factors are developed through the use of regression analysis as follows [8]. The curvature ductility factors for serviceability limit states are P P mS = 21.96x 16.37x + 4.85 f cp Ag f cp Ag The curvature ductility factors for damage control limit states P mDC = -3.0r +30.9rs 40.0 f cp Ag
2

The constitutive model used for the steel reinforcement was a simple bilinear model. The steel has initial stiffness E = 200 GPa, the post-yield hardening stiffness of 1.0% pre-yield stiffness, and the ultimate steel strain in tension was limited to 0.1. The steel stress-strain relationship is shown in Fig. 6.

2 s

4 Serviceability and damage control limit states by analysis


The curvature limit state and curvature ductility factor of circular reinforced concrete bridge piers are investigated in

P + 1.4rs2 16.3rs +10.0 +rs +8.6 f cp Ag

where r unit: 0.001.

270 The coefficient of serviceability and damage control limit states are 0.231 and 0.353, respectively. The first damage state is defined as the serviceability limit state (named dm1) and the third damage state as the damage control limit state (named dm3). to 20% of its peak value. Gardoni proposed the equations to predict the failure of RC column [9]. ln[DR failure ]= ln[df]+ cd +se 1 f y + fp) df= ( L (3) (4)

Damage states by experiment

A set of three damage states were defined in this study and they are concrete spalling, longitudinal bar buckling and column failure. For the second damage state (named as dm2) and for damage control limit state (named as dm4), the equations proposed by Eberhard were adopted [7]. The equations can be used to predict the mean (or median) Engineering Demand Parameter (EDP) at which the damage observation occurred. Such equations exist for spalling and bar buckling as shown in the Eq. 1. P L DR spall =1.6 1 1+0.1 D Ag f cp (1a)

where DRfailure is drift ratio at column failure, df is drift ratio f y and f y are which is estimated by deterministic model, the yield displacement and plastic displacement of column respectively, L is equivalent column cantilever length, cd, and s and e are unknown parameters. See Ref. [9] for more information describing the damage limit state. It is named as dm3 in this paper. The parameters for the lognormal distribution of the damage states defined above are given in Table 2.

6 Construction of the fragility curves through non-linear static analysis


The step-by-step procedure to construct the fragility curves through non-linear static analysis method is as follows. 1) Create a model of the structure and perform a pushover analysis. 2) Convert the capacity curve to the capacity spectrum. 3) For each damage state, determine the ductility demand and the elastic demand Sae from the target displacement demand Sd of selected damage state defined in Table 2 by the following equations m = SdLS / Sdy Sae = Say R (5) (6)

f y db P L DR b,b = 3.25 1 + ke rs 1 1+0.1 (1b) D f cp D Ag f cp where DRspall and DRb,b are drift ratio (percent) at spalling and bar buckling, respectively, P is axial load, Ag is gross cross sectional area, fcp is compressive strength of concrete, fy is yield stress of longitudinal reinforcement, L is equivalent column cantilever length, D is column diameter, ke = 150 constant for spiral reinforced concrete column, rs is volumetric transverse reinforcement ratio. For reinforced concrete (RC) circular columns, the coefficients of the variation of spalling and bar buckling damage states are 0.321 and 0.230, respectively. In order to make some simplifying assumptions, a lognormal distribution is required. The parameters necessary for the lognormal distribution can then be obtained from the following equations m DR m ln DR = ln 2 1 + dDR
2 sln DR = ln (1+dDR

Compute the yield reduction factor R by using Krawinkler and Nassars equation [10] R = [c(m 1)+1]1 / c c(Tn , a ) = T b + 1+Tna Tn
a n

(7) (8)

(2)

Following common practice, RC column failure is defined as the deformation capacity of a column as the displacement corresponding to a drop in the lateral force resistance equals
Table 2 Parameters for the lognormal distribution of the damage states
Damage states Parameters ln(DR|dmLS) sln(DR|dmLS) 4.328 0.127 3.717 0.313 Serviceability/dm1 Spalling/dm2

where SdLS is the spectral displacement demand and its value depend on the corresponding damage state, Sdy and Say are the yield spectral displacement and yield spectral acceleration of the capacity spectrum, respectively, and Tn is natural period of structure. The coefficients a and b depending on the hardening slope as follows: a = 1, b = 0.42 for a = 0%; a = 1, b = 0.37 for a = 2%; a = 0.8, b = 0.29 for a = 10%.

Damage control/dm3 3.599 0.274

Bar buckling/dm4 2.857 0.227

Failure/ dm5 2.495 0.383

271 (4) Determine the fragility functions. The conditional probability of being in, or exceeding, a particular damage state, dmLS, given the elastic spectral acceleration SaeLS is defined by 1 Sa F ( Sa ) = W ln LS bc Sae (9) dm5 Failure The deformation capacity of a column as the displacement corresponding to a drop in the lateral force resistance equal to 20% of its peak value.

8 Construction of the fragility curves where w i s the standard normal cumulative distribution through non-linear dynamic analysis
function; Sa is the spectral acceleration; SaeLS is the median of the expected value of spectral acceleration necessary to cause the damage level demanded; bc is the normalized and lognormal standard deviation that incorporates the aspects of uncertainty and casualness of the capacity and demand as mentioned above. The value of this latter parameter has been studied by different authors: the value proposed in [11] is bc = 0.6, while the one used in the Hazards US (HAZUS) system is bc = 0.4 [12]. In this study we select bc = 0.4. The result of the fragility through CSM method is presented in the Fig. 7.
1.0 dm1 0.8 dm2 dm3 0.4

Incremental dynamic analysis (IDA) is a parametric analysis method that has recently emerged in several different forms to estimate more thoroughly the structural performance under seismic loads. It involves subjecting a structural model to one (or more) ground motion record(s), each scaled to multiple levels of intensity, thus producing one (or more) curve(s) of response parameterized versus intensity level. Figure 8 shows the results of the IDA analysis of the structure. From IDA curves, we can predict the parameters of EDP conditioned on Intersity Measure (IM), assuming that IM is given, and EDP is well represented by a lognormal distribution. Figure 9 shows the histograms and the fitted lognormal distribution at spectral acceleration value of 0.005 g.
1.5

F ( Sa )

0.6

0.2

dm5

Sa /g
0.5
8 10 12 14

dm4

0.0 0 2 4 6

Sa / (ms2)
0

Fig. 7 Fragility curves of the viaducts (CSM method)

0.02

0.04

0.06

0.08

0.1

0.12

Drift ratio:DR
Fig. 8 IDA curves of the structure

Proposed performance levels


30 25 DR data Lognormal distribution fit

Density / %

A new system of performance objectives is proposed in this study as follows. The specific values are provided for illustration and are meant to be refined through future research and discussion. dm1 Serviceability Limit of concrete strain: 0.004, limit of longitude reinforce bar tensile strain: 0.015; Spalling Cover concrete to spall completely; Damage control A limit tensile strain equals to 60% of the strain is adopted to avoid buckling of compressive reinforce bar and low cycle fatigue; Bar buckling Longitude reinforce bar buckling;

20 15 10 5 0 0.01

dm dm3
2

0.02

0.03

0.04

0.05

0.06

0.07

0.08

Drift ratio:DR (Sa=1.005g)


Fig. 9 Histograms and lognormal distribution fitted (Sa = 1.005 g)

dm4

272 When we have the parameters of EDP at each IM, we can calculate the conditional probability of being in, or exceeding, a particular damage state, dmLS, given the IM, is defined by [13]
LS LS

1.0 Nolinear dynamic analysis method CSM method 0.8 dm2 dm1 0.4 dm3 dm5 0.2 dm4

edp | IM = im ] dedp

EDP = edp dP [ EDP < (10)

The first term is the CDF obtained from capacity simulation, often available for discrete Damage Measure (DM) values only. The second term is the derivative of Eq. (10), or simply the PDF of the demand model at the specified IM value. The resulting damage fragility curve is also a Cumulative Distribution Function (CDF) for a single damage limit state (dmLS) with intensity on the horizontal axis. Smooth fragility curves are obtained by arbitrarily fitting lognormal distribution functions to the simulation results. The simulation results and the fitting curves are shown as discrete points and smooth curves in Fig. 10.
1.0 dm1 0.8 dm2 dm3 dm4 dm5 0.2

F (S a )

P DMidm | IM = im = edp P DMidm |

0.6

0.0

10

12

14

Sa / (ms2)
Fig. 11 Comparison of fragility curves

F (Sa )

0.6

0.4

0.0 0 2 4 6 8 10 12 14

Sa / (ms2)
Fig. 10 Fragility curves of the viaducts (non-linear dynamic analysis)

9 Comparison between the fragility curves from two methods


Figure 11 shows the comparison of the structural fragility curves from nonlinear static analysis and nonlinear dynamic analysis. The difference between the results for the damage state of concrete covering layers spalling is not big. For the damage states of reinforced steel buckling and collapse, the fragility curve from the nonlinear dynamic analysis is on the left, which underestimates the probability of the specific damage state that the structure exceeds under given ground motion intensity.

10

Conclusion

the fragility curves of a viaduct structure are developed by the capacity spectra method and the nonlinear dynamic RHA method, respectively. LMSR-N of 40 ground motion records is adapted for the fragility analysis of the structure by nonlinear dynamic RHA method. In this paper, the LMSR-N earthquake records were compared with response spectra in the soil type III regulated in the Chinese Code for Seismic Design of Buildings. The analysis indicates that the median values of the acceleration response spectra from LMSR-N records are very compatible with the Chinese design response spectra in the code when the period is less than 1.5 s. The results from structural fragility curves computed by two methods show that there is no big difference between the two analysis methods for the serviceability limit state and the damage state of covering layer spalling, but a rightward excursion of the fragility curve developed during the nonlinear static analysis because of the damage states of the damage control limit state, longitudinal reinforcement buckling state and collapse state, which reveals that nonlinear static analysis underestimates the probability of the specific damage states that structure reaches under a given ground motion intensity. The structural fragility curves developed by CMS method can help make the structural analysis simple and quick, avoiding the implementation of the dynamic RHA. Although the dynamic RHA requires a lot of complicated analysis for the structure, the results from are reliable and accurate. It is of practical value to develop probability-based fragility curves for the estimation of seismic performance of a structure. We can lucubrate the exact information of the seismic performance of the structure provided for us by the fragility curves for the estimation of the existing structures. However, we will need to establish the relationship between the design parameters and fragility curve parameters of the structure when applying the fragility curves onto a performance based seismic design practice.
Acknowledgments This study was partly supported by the National
Natural Science Foundations of China (Grant No. 50578007).

The fragility curve is a powerful tool to analyze the performance of the structure under an earthquake. In this paper,

273

References
1. Shinozuka. Effect of Retrofit, Skew and Number of Spans on Bridge Fragility. Technical Report MCEER-04-0006. New York: Multidisciplinary Center for Earthquake Engineering, State University of New York, 2004 2. Medina R. Seismic demands for nondeteriorating frame structures and their dependence on ground motions. Dissertation for the Doctoral Degree. Stanford: Department of Civil and Environmental Engineering, Stanford University, 2002 3. Federal Emergency Management Agency. Prestandard and Commentary for the Seismic Rehabilitation of Buildings (Document 356). Washington: Federal Emergency Management Agency, 2000 4. GB50011-2001. Code for Seismic Design of Buildings. Beijing: China Architecture & Building Press, 2001 (in Chinese) 5. Kent D C, Park R. Flexural members with confined concrete. Journal of Structural Engineering, 1971, 97(ST7): 1 9691 990 6. Mander J B, Priestley M J N, Park R. Theoretical stress-strain model for confined concrete. Journal of the Structural Engineering, 1988, 114(ST8): 1 8041 826 7. Wang Jianmin, Zhu Xi. Probability analysis of the curvature limit state and ductility of circular RC Bridge piers. China Civil Engineering Journal, 2006, 39(12): 8894 (in Chinese)

8. Berry M P, Eberhard M O. Performance Models for Flexural Damage in Reinforced Concrete Columns. Peer Report 2003/18. Berkeley: Pacific Earthquake Engineering Research Center, University of California, 2003 9. Gardoni P. Probabilistic models and fragility estimates for structural components and systems. Dissertation for the Doctoral Degree. Berkeley: Department of Civil and Environmental Engineering, University of California, 2002 10. Krawinkler H, Nassar A A. Seismic design based on ductility and cumulative damage demands and capacities. In: Fajfar P, Krawinkler H, eds. Nonlinear Seismic Analysis and Design of Reinforced Concrete Buildings. New York: Elsevier Applied Science, 1992 11. Dutta A, Mander J B. Seismic fragility analysis of highway bridges. In: Proceedings of the Center-to-Center Project Workshop on Earthquake Engineering in Transportation Systems. Tokyo: [s. n.], 1999 12. HAZUS 99. Earthquake loss estimation methodology (Technical Manual). Washington: National Institute of Building Sciences for Federal Emergency Management Agency, 1999 13. Mackie K, Stojadinovic B. Fragility curves for reinforced concrete highway overpass bridges. In: Proceedings of 13th World Conference on Earthquake Engineering. Vancouver: [s. n.], CD-ROM No 1553

S-ar putea să vă placă și