Report - PEER - University of California, Berkeley

shows that the scalar IM proposed by Luco discussed above not only strongly reducesdispersion (vis-a-vis S a ), but also shows apparent sufficiency with respect to situationswhere near-source directivity effects may be important. Improved IM’s, whilereducing the requisite number of nonlinear structural analyses, may come with aprice. Luco’s predictor, for example, requires the development of a new attenuationlaw. Vector IM’s require new attenuation information (e.g., correlations) and/orPSHA computer code modifications.LA9 , SH+NH+RC simulated earthquake recordsLA9 , SH+NH+RC simulated earthquake recordsIM 1E∝ S a( T 1=2.23s ) [rad]10 −1 3 410 −28210 −310 −3 10 −2 10 −1Max. Peak Story Drift Angle, θ max[rad]83384444333333434343342232a = 1.50σ = 0.44(n = 30)IM 1I&2E[rad]10 −1 310 −28210 −310 −3 10 −2 10 −1Max. Peak Story Drift Angle, θ [rad] max843384443333 43334433324 322a = 1.21σ = 0.17(n = 30)Figure 2. (a) Regression of MIDR versus (in effect) Sa; (b) regression versus theLuco IM. The records are geophysical synthetic accelerograms for a site nearthe Hayward Fault simulating 30 repetitions of an event rupturing all threesegments of that fault. (Luco, 2002).3.1 The Basics3. PROBABILISTIC ASSESSMENTProbabilistic assessment of structures for PBSD has been both researched and appliedin various fields in various degrees. For example, the U.S. nuclear power industry hasused Seismic PRA’s (Probabilistic Risk Assessments) for two decades (Kennedy,1980)) applying it to virtually all the plants in the country under the IPEEE program.Much more recently, the U.S. building industry has produced guidelines forprobabilistic assessment of steel moment-resisting frames (FEMA-SAC (2000),Cornell(2002)). Both of these approaches are based on integration over the product ofthe hazard curve times some representation of the probability of a specified limit stategiven the IM level. Both procedures call for the explicit quantification of both thealeatory (“random”) and epistemic (“knowledge”) sources of uncertainty. The formermethod is based on fragility curves (P[C|IM]) and provides a mean estimate of thelimit state annual frequency. The latter document is based on a non-lineardisplacement-based “load and resistance factor” (LRFD-like) scheme derived from adistributions of displacement demand and displacement capacity; it sets criteria interms such as a 90% confidence that the annual frequency of collapse is less than1/2500 (i.e., explicitly separating epistemic and aleatory uncertainty).46