Você está na página 1de 10

Proceedings of the 8th U.S.

National Conference on Earthquake Engineering April 18-22, 2006, San Francisco, California, USA Paper No. 505

APPLICATION OF NONLINEAR STATIC ANALYSES TO PROBABILISTIC SEISMIC DEMAND ANALYSIS P. Tothong 1 and C. A. Cornell2

ABSTRACT Many approximate methods for Nonlinear Dynamic time history Analysis (NDA) have been recently proposed to estimate inelastic responses in multi-degree-offreedom (MDOF) structures. This paper will focus on two methods: (i) the modified modal pushover analysis (MMPA, Chopra et al. (2004)), and (ii) the method proposed by Mori (Mori et al. 2004), which incorporates inelastic shape functions into the approximation. The objective is to extend these two approximate methods to develop a structural demand hazard curve via a Probabilistic Seismic Demand Analysis. Unlike assessing the performance of structures for a given ground motion hazard level, the structural demand hazard curve provides multi-objective structural performance information, which has been integrated from all possible ground motion hazard levels. This paper will describe a methodology to approximate incremental dynamic analysis results from nonlinear static analyses, and then further to integrate such results with an inelastic spectral displacement ( Sdi ) ground motion hazard curve ( Sdi ) (Tothong and Cornell 2005a). These inelastic spectral ordinates provide an improvement over conventional elastic ones. The resulting demand hazard curves are then compared with one obtained from rigorous NDA. This comparison identifies apparent strengths and weaknesses of using these methods as estimates of NDA results.

Introduction One of the ultimate goals of Performance-Based Earthquake Engineering (PBEE) is to estimate the mean annual frequencies (MAFs) of an engineering demand parameter (EDP) exceeding a specified value, say, x at a designated site, as a function of x. This curve is denoted EDP ( x ) . The term EDP is the terminology used by the Pacific Earthquake Engineering Research (PEER) Center to define the structural response of interest, e.g., the peak (over time) maximum inter-story drift ratio ( max ). Note that max will be used as an EDP throughout this
1 2

Graduate Research Assistant, Dept. of Civil and Envi. Engineering, Stanford University, Stanford, CA, 94305 Professor Emeritus, Dept. of Civil and Envi. Engineering, Stanford University, Stanford, CA, 94305

paper. In PEERs PBEE methodology, one of the steps associated with estimating max ( x ) is to perform Incremental Dynamic Analysis (IDA) using Nonlinear Dynamic time history Analysis (NDA). The structural demand hazard curve for max , max , can provide a more direct comparison as to whether the structural performances satisfy the specified target performance objectives, e.g., whether the max ( xi ) values are less than the guidelines specified for a set of values of xi associated with various limit states (Collins and Stojadinovic 2000; Liu et al. 2005). The curve max ( x ) provides a multi-objective (i.e., the performances of a structure at various max levels) structural response hazard curve for engineers to assess the performance of structures. With the proposed concept, max can be obtained without performing NDA, which usually requires extensive computational time. The concept of approximating IDA results from nonlinear static analyses will be explained below. To demonstrate the proposed methodology, a generic frame model with nine stories and a fundamental period (T1) equal to 0.9 seconds was used. This structure has beam hinges with stiffness/strength degradation, a hardening stiffness ratio equal to 3%, a ductility capacity value equal to four, and post-capping stiffness value equal to 10% (in absolute value) of the elastic stiffness, which implies rather severe strength degradation for rotations in excess of four times the yield level. A description of the nomenclature is given in Ibarra et al. (2005). The structure has been modeled and analyzed by Ibarra and Krawinkler (2003). A suite of 40 historical earthquake ground motions was used for this analysis. The earthquake magnitude and distance range from 6.5 to 6.9 and from 13 to 40 km, respectively. The record set was aggregated by Medina and Krawinkler (2003). The disaggregation of the drift hazard (Baker et al. 2005) tells engineers which ground motion hazard level has the largest contribution to the probability of exceeding a particular drift level. Engineers can then disaggregate the ground motion hazard curve ( IM ) at that level to determine which ground motion characteristics (e.g., , Mw, distance) cause that ground motion intensity level; ultimately such information can guide seismologists and engineers in the selection of ground motion records. This detailed selection of ground motions can, however, be largely eliminated by using inelastic spectral displacement (Sdi) as an intensity measure (IM) because Sdi senses which records will tend to cause benign or aggressive inelastic responses on structures, in contrast to Sde (Tothong and Cornell 2005b). Further we find that this intensity measure is less prone to introducing record-scaling bias than Sde. The Sdi (as opposed to Sde) ground motion hazard curve ( Sdi , derived from a Sdi attenuation law) is currently available (Tothong and Cornell 2005a). Probabilistic Seismic Demand Analysis The Pacific Earthquake Engineering Research, PEER, Center has developed a methodology to evaluate the seismic performance of a structure (Cornell and Krawinkler 2000). The prediction of max requires both the structural analysis results from the IDA and the ground

motion hazard obtained from Probabilistic Seismic Hazard Analysis (PSHA, Cornell (1968)). With the mild assumption that the conditional distribution of max on im is lognormal, the max can be estimated by performing the integration shown in Eq. (1), e.g., Shome and Cornell (1999), and Luco and Cornell (2002). This approach has been extensively used in PBEE by PEER, e.g., Moehle and Deierlein (2004).

max

( x) =

Gmax |IM ( x | im ) d IM ( im )

(1)

all ims

where Gmax is the Gaussian complementary cumulative distribution function (ccdf) of the random variable max , which is defined as follows:

ln x lnmax |im Gmax |IM ( x | im ) = 1 ln |im max

(2)

where ( ) is the standard Gaussian cdf, and lnmax |im and lnmax |im are the conditional mean and dispersion (i.e., the standard deviation of the natural logarithm) of max given intensity level im.
d IM ( im ) is the negative differential of the hazard curve, or numerically speaking the absolute

value of the difference between the consecutive points on IM ( im ) . This MAF can be directly compared with a code specified tolerance limit, e.g., the MAF of a maximum interstory drift ratio of 3% needs to be less than, say, 4 104 . This fully probabilistic approach involves NDAs of a MDOF structure, which may require significant resources and time beyond that available to a practicing engineer. Hence, we introduce the simple notion of extending nonlinear static analyses to generate IDA results. This concept will be explained in the next section.
Nonlinear Static-Based Incremental Dynamic Analysis

Many methods built upon nonlinear static analyses have been proposed to approximate NDA. The conventional nonlinear static procedure has become a standard in engineering practice to evaluate structural performance (FEMA-356 2000). Past studies by Chopras group (e.g., Chopra et al. (2004)) and others (e.g., Mori et al. (2004), Dolsek and Fajfar (2004)) have led to the improvement in seismic demand estimation from nonlinear static analysis. In this paper, we chose to use the modified modal pushover analysis (MMPA, Chopra et al. (2004)) and Moris method (Mori et al. 2004) to illustrate the procedure. The difference between Moris procedure and MMPA is that Moris procedure incorporates the inelastic shape function into the first-mode response, while MMPA uses the elastic shape function of the structure. However, the differences in the results from the two procedures are comparatively small. The general procedures to perform MMPA and Moris method can be found in Chopra et al. (2004) and Mori et al. (2004), respectively. (Due to the length limitation, the details of these methods will not be repeated here.)

Steps to Generate Incremental Dynamic Analysis Results from Nonlinear Static Analysis

An IDA is a plot of a dynamic response measure (here max ) versus a ground motion intensity measure, often elastic spectral acceleration at the first-mode period (Vamvatsikos and Cornell 2002). As mentioned above, the intensity measure used here is inelastic spectral displacement, Sdi. This measure has two parameters, period T and yield displacement dy. In the example below, we adopted the following approaches to estimating these parameters. To estimate T, we used the first-mode period from an eigenvalue analysis. We used dy = 0.9 based on a procedure described by Tothong and Cornell (2005b). These are but two of several possible approaches, e.g., Fajfar (2000), and Goel and Chopra (2004). Once one has a representative single-degree-of-freedom system, one can prepare a seismic hazard analysis in terms of this Sdi and then choose a set of ground motion intensity levels to scale a sample of ground motion records to such that the levels cover the entire range of hazard levels of interest. For a given ground motion in the sample, one must next calculate the set of scale factors to apply to the as-recorded ground motion in order to match each specified intensity level chosen in the previous step. (It should be noted that an iterative search is needed when one uses Sdi as an IM.) Next, one uses this set of scaled ground motions to calculate the target roof displacement for the first-mode as well as the elastic response contributions from the higher modes. Lastly, one combines the peak modal responses from each mode using a simple Square-Root-of-Sum-ofSquares (SRSS) procedure (Chopra et al. 2004). This completes the full IDA from nonlinear static analysis for one record. These steps are repeated for each record in the sample. Note that if Sdi is used as an IM, the inelastic first-mode response obtained from pushing a structure to the target roof displacement (which is a step in the two static approximate methods) will essentially be the same for all ground motions (for a given ground motion hazard level). This is because the motions have all been scaled to a common value of Sdi. The differences in the responses among ground motions are essentially from higher mode contributions, i.e., depending on the frequency content of the ground motions (response spectra) and the vibration properties of the structure.

9 8 7 6

Sdi [in]

5 4 3 2 1 0

0.005

0.01

0.015

0.02

0.025

0.03

max

Figure 1. A set of 40 IDAs of the structure used in this study using NDA. The median and the one-standard-deviation bands are shown with solid and dotted lines, respectively. A dashed vertical line is shown to indicate an approximate yielding point of the structure, i.e., the onset of nonlinearity in the members, at max = 0.3%.
MMPA
9
9

Mori's procedure

Sdi [in]

0.005

0.01

0.015

0.02

0.025

0.03

0.005

0.01

0.015

0.02

0.025

0.03

max

max

(a) (b) Figure 2. The IDAs of the structure used in this study using the static approximations: (a) MMPA and (b) Moris procedure Figure 1 shows the IDAs of a suite of ground motions used in this study using NDA. Figure 2a and Figure 2b illustrate the IDAs generated by using MMPA and Moris procedure, respectively. Most notably, both approximate methods somewhat under-predict seismic demands for low ground motion intensity levels and significantly over-predict seismic demands for high ground motion intensity and ductility levels. A similar overestimation has been reported by, e.g.,

Chopra et al. (2004), and Goel and Chopra (2005). Examination of the figures above reveals that the two static procedures significantly underestimate the record-to-record variability of the responses. The authors believe that this underestimation has not been previously reported. This variability may be an important part of the estimation of max . The bias ratio of the conditional    median (i.e., MMPA | im NDA | im , where denotes the median value for max ) and the bias ratio of the record-to-record dispersion of max (i.e., lnmax |im, MMPA lnmax |im, NDA ) for the two static methods are plotted as a function of ground motion intensity level in Figure 3. Here bias ratio is defined as the median (or dispersion) of the response from an approximate method divided by the same statistic found from the exact responses, which in this case are the NDA results.
Median
1.8
1 MMPA Mori 0.9

Dispersion

1.6
0.8

1.4
0.7

Bias ratio

1.2

0.6

0.5

1
0.4

0.8
0.3

0.6

0.2

Sdi [in]

Sdi [in]

(a) (b) Figure 3. The bias ratios for the conditional (a) median and (b) dispersion for both methods. The bias-high of the max at a higher ground motion intensity level may be partially attributed to the assumption that higher modes remain elastic (Chopra et al. 2004). Note that the large bias (greater than 20%) occurs only after the system has experienced strength degradation ( max > 1% or Sdi > about 4 to 5). The static methods seem unable to capture the large increase in dispersion that occurs once one leaves the linear regime ( max > 0.3% or Sdi > about 1.5). The interesting chaotic behavior of the dynamic IDAs (Figure 1) is lost in the static approximations of the IDAs (Figure 2).
Demand Hazard Curve Results

We can use the above results in the integration with the ground motion hazard curve as per Eq. (1) and (2). The resulting product is a structural demand hazard curve, max .

10

NDA MMPA Mori

10
max

-1

10

-2

10

-3

10

-3

10

-2

max

Figure 4. max calculated using NDA, MMPA, and Moris procedure. From Figure 4, we can observe that the max curves obtained from the approximate methods yield about the same result as if one uses NDA up to a drift level of about 1%, which in this structure is a global ductility of about four (see Figure 1). However, for lower hazard levels (higher nonlinearity levels), the approximate methods overestimate structural responses relative to those obtained from NDA (for a given MAF level). The problem arises from predicting higher conditional medians of max at high Sdi and hence low hazard levels (Figure 3a). The under-estimation of the dispersion (Figure 3b) was insufficient to offset this effect. Both of these deficiencies should be improved if the methods are going to be used in assessing the performance of structures at high degrees of nonlinearity. If these biases are systematic and can be identified across the structural parameters (e.g., T, type of structure, number of stories, etc.), perhaps approximate correction factors can be developed. Clearly such approximate static methods - with bias corrected - can prove useful to estimate the performances of structures, especially at preliminary design stages of structures, where NDA is not an option. However this bias-high and bias-low for the median and dispersion, respectively, should be kept in mind. This may suggest that one still needs to use NDA to estimate seismic demand to obtain the structural demand hazard curve in order to ensure accuracy in the final proposed designs or rehabilitations. Note that this demonstration does not include the epistemic uncertainty in structural modeling and/or other parameters used in the analysis. Also, due to the limited sample size, the standard error of the biases and of max may be large. Therefore the bias ratios may be higher or lower than those shown in Figure 3. To estimate the standard error of the biases and max , the simple statistical method, bootstrap, can be used (Efron and Tibshirani 1993).

Discussion

Even though the max obtained using either of the approximate methods shown above is close to the one obtained from NDA for max as high as 1% (global ductility of about four), this is somewhat misleading. In this range, the biases in the conditional median and dispersion of max compensate each other. As can be seen in Figure 1, we can observe that, at the ground motion intensity level, Sdi of 4 to 5, the median max obtained from NDA is about 1%. In this range, the bias ratio for the median value of max is about 35% high, while the dispersion of max is only about 30% of that estimated by NDA. Because the two biases counteract one another, near max = 1% the max curves of the approximate methods are close to that of NDA (Figure 4). Researchers who try to improve the median prediction obtained from a nonlinear static analysis should recognize this problem with the dispersion as well. In order to fully adopt a nonlinear static analysis for predicting the structural response probabilistically, one should focus on improving both the median and dispersion estimates. It should be noted that it is an iterative procedure to design/retrofit a building to meet multi-level performance objectives, and it is computationally expensive/impossible to perform NDA after every proposed strengthening and/or stiffening of some part of the structure. An engineer may choose to use static approximations in preliminary stages and then to perform NDA in the final design to verify that the structure has met all target structural performance objectives (e.g., Intermediate Occupancy, Life Safety, and Collapse Prevention, (FEMA-356 2000)).
Conclusions

This paper illustrates how nonlinear static analyses of MDOF systems can be used to generate IDA results and, further, can be combined with a ground motion hazard curve to obtain a structural demand hazard curve. This curve provides a multi-objective performance measure for a structure (i.e., the performance of a structure at various max levels). Even though it may not be accurate at larger nonlinearity levels, a static-based estimate of max provides a tool for structural engineers to easily estimate how the structure will perform under all possible ground motion hazard levels, i.e., it provides the MAF (mean annual frequency) of max exceeding any specified value. The reader is cautioned that the quantitative conclusions in this study are based on a single example structure; the results shown here may or may not apply to other structures.
Acknowledgements

The authors wish to thank Dr. Luis Ibarra for providing the analysis results used in this study. The financial support for this research is provided by the National Science Foundation through the U.S.-Japan Cooperative Research in Urban Earthquake Disaster Mitigation Project (NSF 98-36) under grant number CMS-0200436. This support is gratefully acknowledged.

References
Baker, J., Cornell, C. A., and Tothong, P. (2005). "Disaggregation of Seismic Drift Hazard." Proceedings, 9th International Conference on Structural Safety and Reliability (ICOSSAR05), Rome, Italy, June 2005. Chopra, A. K., Goel, R. K., and Chintanapakdee, C. (2004). "Evaluation of a Modified MPA Procedure Assuming Higher Modes as Elastic to Estimate Seismic Demands." Earthquake Spectra, 20(3), 757-778. Collins, K. R., and Stojadinovic, B. (2000). "Limit states for performance-based design." 12th World Conference on Earthquake Engineering [Proceedings] [computer file]; 2000. Cornell, C. A. (1968). "Engineering Seismic Risk Analysis." Bulletin of the Seismological Society of America, 58, 5, 1968, pages 1583-1606. Cornell, C. A., and Krawinkler, H. (2000). "Progress and Challenges in Seismic Performance Assessment." PEER Center News, 2000. 3(2): 4p. (http://peer.berkeley.edu/news/2000spring/index.html) Dolsek, M., and Fajfar, P. (2004). "IN2 - A Simple Alternative for IDA." Proceedings, The 13th World Conference on Earthquake Engineering, Vancouver, BC Canada, August 2004, paper no. 3353. Efron, B., and Tibshirani, R. J. (1993). An Introduction to the Bootstrap, Chapman & Hall/CRC, New York. Fajfar, P. (2000). "A nonlinear analysis method for performance-based seismic design." Earthquake Spectra. Vol. 16, no. 3, pp. 573-592. Aug. 2000. FEMA-356. (2000). Prestandard and Commentary for the Seismic Rehabilitation of Buildings, SAC Joint Venture, Sacramento, California. Goel, R. K., and Chopra, A. K. (2004). "Evaluation of Modal and FEMA Pushover Analyses: SAC Buildings." Earthquake Spectra 20, 225 (2004). Goel, R. K., and Chopra, A. K. (2005). "Extension of Modal Pushover Analysis to Compute Member Forces." Earthquake Spectra, 21(1), 125-139. (http://link.aip.org/link/?EQS/21/125/1) Ibarra, L. F., and Krawinkler, H. (2003). "Global Collapse of Frame Structures under Seismic Excitations." Ph.D. Dissertation, Department of Civil and Environmental Engineering, Stanford University, John A. Blume Earthquake Engineering Center Report No. 152, Stanford, CA, 324 pp. Ibarra, L. F., Medina, R. A., and Krawinkler, H. (2005). "Hysteretic models that incorporate strength and stiffness deterioration." Earthquake Engineering & Structural Dynamics, 34(12), 1489-1511. (http://dx.doi.org/10.1002/eqe.495) Liu, M., Burns, S. A., and Wen, Y. K. (2005). "Multiobjective Optimization for Performance-Based Seismic Design of Steel Moment Frame Structures." Earthquake Engineering & Structural Dynamics, Volume 34, Issue 3, Pages 289 - 306. Luco, N., and Cornell, C. A. (2002). "Probabilistic Seismic Demand Analysis, SMRF Connection Fractures, and Near-Source Effects." Ph.D. Dissertation, Department of Civil and Environmental Engineering, Stanford University, Stanford, CA. (http://www.stanford.edu/group/rms/ (accessed 10/10/2005)) Medina, R. A., and Krawinkler, H. (2003). "Seismic Demands for Nondeteriorating Frame Structures and Their Dependence on Ground Motions." Ph.D. Dissertation, Department of Civil and Environmental Engineering, Stanford University, John A. Blume Earthquake Engineering Center Report No. 144, Stanford, CA 347 pp. Moehle, J. P., and Deierlein, G. G. (2004). "A Framework Methodology for Performance-Based Earthquake Engineering." Proceedings, The 13th World Conference on Earthquake Engineering, Vancouver, BC Canada, August 2004, paper no. 679. Mori, Y., Yamanaka, T., Luco, N., Nakashima, M., and Cornell, C. A. (2004). "Predictors of Seismic Demand of SMRF Buildings Considering Post-Elastic Mode Shape." Proceedings, The 13th

World Conference on Earthquake Engineering, Vancouver, BC Canada, August 2004. Shome, N., and Cornell, C. A. (1999). "Probabilistic Seismic Demand Analysis of Nonlinear Structures." Ph.D. Dissertation, Department of Civil and Environmental Engineering, Stanford University, Stanford, CA. (http://www.stanford.edu/group/rms/ (accessed 10/31/2005)) Tothong, P., and Cornell, C. A. (2005a). "An Empirical Attenuation Law for Inelastic Spectral Displacement." (In Preparation). Tothong, P., and Cornell, C. A. (2005b). "Improved Probabilistic Seismic Demand Analysis using Advanced Scalar Ground Motion Intensity Measures, Attenuation Relationships, and Near-Fault Effects." Ph.D. Dissertation, Department of Civil and Environmental Engineering, Stanford University, Stanford, CA, (In Preparation). Vamvatsikos, D., and Cornell, C. A. (2002). "Incremental Dynamic Analysis." Earthquake Engineering & Structural Dynamics. Vol. 31, no. 3, pp. 491-514. Mar. 2002.

Você também pode gostar