Structural seismic damage and loss assessments using a multi-conditioning ground motion selection approach based on an efficient sampling technique

Abstract

The application of Latin Hypercube and Monte Carlo (MC) sampling techniques for ground motion selection purposes is investigated. Latin Hypercube Sampling (LHS) works by first stratifying a probability distribution domain into multiple equally spaced and non-overlapping stripes and then by permutationally drawing samples from those stripes. To examine the efficiency of these two distinct sampling methods, a set of conditional multivariate distributions was fit to an intensity measure vector based on a single, two, or average of more-than-two (average) conditioning intensity measure. LHS was then utilized for sampling purposes from the conditional multivariate distributions, which in turn demonstrated superiority over MC given the same number of realization samples. Accordingly, it was utilized as an underlying peace of a broader ground motion selection framework to facilitite the selection of a number of ground motion suites based on different methods of conditioning. Using the selected suites, response history and subsequent damage/loss analyses were conducted on a generic 4-story non-ductile reinforced concrete building. The outcomes of these latter studies demonstrated that the ground motion suite selected based on an average-intensity-measure conditioning criterion performed better than those selected through single- and two-intensity-measure conditioning criteria.