Abstract
When a number of ground motions having different characteristics are considered as candidates of design input motions, it is difficult to select or synthesize a representative wave among those waves. In order to avoid this problem, we have presented a scheme for clustering ground motions considering their effects on nonlinear structural response values. Characteristics of ground motions are expressed using probability density functions (PDFs) of nonlinear response values of structural models and their dissimilarity is quantified by Jensen-Shannon divergence. As a clustering method, hierarchical clustering algorithm is applied. In this paper, for the purpose of verification, the proposed scheme is applied to a set of ground motions which are categorized into three types from the viewpoint of linear response spectra. When structural models are linear systems, the proposed scheme classifies those three types of ground motions. However, when nonlinear systems are used, it is found that clusters which are not in accord with their linear response spectra are formed.
