SIMULATION OF STRONG GROUND MOTION DISTRIBUTIONS OF HORIZONTAL TWO COMPONENTS BASED ON SINGULAR VALUE DECOMPOSITION ANALYSIS

Abstract

 This study proposes a method for simultaneously simulating strong ground motion distributions of horizontal two components (Fault-Normal and Fault-Parallel components) while preserving the correlated spatial variation structure in two components. First, singular value decomposition (SVD) analysis is applied to a cross variance matrix consisting of strong ground motion distributions of horizontal two components. The SVD analysis provides the modal forms representing spatial characteristics with strong correlation in two components and the principal component scores that characterize each strong ground motion distribution. Second, the principal component scores are replaced by vectors preserving the covariance structure of two components, continued by synthesizing vectors and modal forms to simultaneously simulate the strong ground motion distribution of two components. The study results show that both the method of Cholesky decomposition applied correlation coefficient between principal component scores of two components and the method of generating difference in principal component scores of two components by random simulation are able to simulate strong ground motion distributions that reproduce the correlation between horizontal two components.