IM and Ground Motion Selection for EDP Hazard Assessment - 1312005

Project Title—ID Number IM and Ground Motion Selection for EDP Hazard Assessment - 1312005
Start/End Dates 10/1/05 – 9/30/06
Funding Source PEER-NSF
Project Leader (boldface) and Other Team Members Allin Cornell (Stanford/F), Jack Baker (Stanford/GS), Polsak Tothong (Stanford/GS), Iunio Iervolino (U. of Naples/O)
F=faculty; GS=graduate student; US=undergraduate student; PD=post-doc; I=industrial collaborator; O=other

Project goals and objectives

This project is developing recommended IMs for various site/building cases and recommended procedures for selection and processing of ground motion accelerograms for EDP hazard assessment. The recommendations will follow from development, consideration, study, and demonstration of alternatives, including the identification of their weaknesses and strengths. Included are both scalar and vector schemes for IMs. For example, for first-mode dominated buildings the preferred scalar may well be inelastic spectral displacement, as used in the Van Nuys test bed, whereas an alternative for important and/or taller structures may be a vector consisting of elastic Sa coupled with spectral ratios and/or epsilon. (Epsilon is the standardized deviation of Sa from its predicted (median) value.) Far and near-source situations are being considered; the latter require extra care and/or modified IMs. The recommended procedures include (1) record selection and record processing such as scaling and/or "shaping", (2) number and kinds of Nonlinear Time History (NTH) analyses, and (3) suggested post-processing of response output. Fundamental engineering insights can be derived by recognition of the highly inter-related issues of record selection, scaling and processing - as it relates to a specific site and structure. Structures are, in general, three-dimensional and with different natural periods in different directions. The project will address this problem as well.

Role of this project in supporting PEER's mission (vision)

PEER's vision includes NTH analysis. This project provides (1) guidance on the input to those analyses and (2) the connection to the site hazard analysis that permits accurate and efficient computation of structural response (EDP) annual frequencies.

Methodology employed

From data bases of the results on multiple NTH analyses of multiple frame structures, we study the efficiency (i.e., variability in response prediction given an IM level), the sufficiency (i.e., sensitivity of the results to the selection of the record set, e.g., to their magnitudes), and the EDP- bias induced by record-scaling in order to understand and evaluate the pros and cons of different IMs.

Brief Description of previous year's achievements, with emphasis on accomplishments during last year (Year 8)

The primary accomplishments of the project have been the thorough study of vector and advanced scalar IM's, including the role of epsilon in the effectiveness of Sa-based IMs and in the selection and scaling of records for NTH analysis. One specific accomplishment is the development of new approaches to record selection for non-linear analysis. These are based on our findings of the role of "epsilon" as a characterizer of structural performance via its role as a proxy for spectral shape. Two alternative selection recommendations are given, their implications explored and their positive effectiveness demonstrated. One suggests selecting records primarily based on epsilon (rather than M or R). This suggestion has been implemented in the PEER building benchmark study. We show that when used with simple Sa(T1) (i.e., first- mode period Sa) as an IM it is efficient (reduces EDP|IM sigma), sufficient (avoids potential bias due to ignoring epsilon role), and avoids potential record scaling bias. The second selection procedure suggests selecting records (regardless of M, R or epsilon) that best match (not the UHS shape which is current – e.g., testbed - practice) but the "conditional expected spectral shape given Sa(T1) = x", where x is the ground motion level associated with the mean return period of the "stripe" of records being selected. In particular at large return periods this shape will match the UHS only at T1, and will lie below it elsewhere – falling off as T differs from T1. This approach has all the same beneficial properties as the first. Both lead to lower extreme responses and higher collapse fragility curves than the schemes used today (e.g., the test bed approach) and they avoid the need for more computationally intensive vector IM's.

Figure 1 shows the median spectra of four sets of records: (1) denoted AR is for an arbitrarily- selected set of records; (2) denoted MR-BR is for a set of records that have approximately the mean magnitude and distance from the PSHA disaggregation at this site for an Sa (0.8 sec) level of 1.6g (i.e., this mimics current best practice and the PEER testbed); (3) denoted ∈-BR is for a set of records with approximately the same median ∈ from disaggregation (i.e., the first new method described above); and (4) denoted CMS-∈ is for a set of records selected to match the conditional mean spectrum given this level of Sa(0.8sec). 0.8 sec is the first natural period of an example MDOF frame and the Sa level of 1.6g is approximately the median level at which collapse occurs. Figure 2 shows the global collapse fragility curve results from multiple NTH analyses. The more accurate new methods suggest a more robust structure than the other two cases. This can be anticipated by the shape of their spectra, which lie every (except at 0.8 sec) below the other two. The accuracy of these less conservative results is confirmed by vector- valued analysis (using Sa and epsilon as arguments).

All this work is based on "ordinary" records and makes important use of the Luis Ibarra-Helmut Krawinkler data base of generic frame structure IDA's to demonstrate our conclusions over a wide range of stories, periods, and backbones. Additional work has been conducted for near- source records and new methods of analysis are under development for this more complicated case.

chart - Figure 1 - The CMS-∈ spectrum at Sa(0.8s)=1.6g (given M=6.4, R=11.5 km and ∈=2.1) and the mean response spectra of record sets selected using each of the four proposed methods.
chart - Figure 2 - Probability of collapse vs. Sa(T1) (i.e., the collapse

Describe any instances where you are aware that your results have been used in industry

We have been told informally that at least two firms have recognized now the issues induced by epsilon and are using our recommendations to avoid them when selecting and scaling records for their clients.

Expected milestones & deliverables

  1. Report on vector IM's and on the role of epsilon (Baker thesis).
  2. Report on improved scalar IM's (Tothong thesis). 3. Report of summary recommendations for practice.

Member company benefits

Better understanding and methods of ground motion selection and scaling for current and for future PBEE projects.

Back to Funded Project Archive main page