Factor mapping method for grouped input variables and its application to seismic damage analysis

摘要

The study considers the characterization of global seismic damage indices of randomly parametered inelastic structures with a view to estimate the relative importance of underlying random variables with respect to variability in these damage indices. We build on an existing simulation-based approach, namely, the factor mapping method, to tackle the problem. Here, an ensemble of input variables is classified into two categories, depending upon whether a sample vector produces a global response damage measure in a pre-defined interval or not. We propose that a measure of distance between samples in these two categories be established in terms of an estimate of the Bhattacharyya distance. This distance is taken to measure the relative importance of individual/groups of input variables with respect to the damage index of interest. Illustrations include results of incremental dynamic analysis of a randomly parametered four-storied inelastic frame under a suite of scaled recorded ground motions.