An efficient stratified-based ground motion selection for cloud analysis

Abstract

In the quantitative seismic risk assessment of structures, cloud analysis has been widely used due to its simplicity to obtain the conditional probability of structural response exceeding a certain level, conditioned on the ground motion intensity. The accuracy of this analysis relies on the selected ground motion records in terms of seismic hazard levels and the number of records. This paper presents an adaptive ground motion selection approach with a stratified sampling scheme to reduce the number of required analysis and to accurately capture structural response with a desired level of confidence. The stratified sampling scheme is used to obtain enough data points from each hazard level in an iteration fashion, while the formulation of the seismic demand model of interest is determined using a Gaussian mixed model clustering algorithm at each iteration. The proposed ground motion selection approach is applied to obtain seismic demand hazards of a non-linear single-degree-of-freedom system and the results are compared to a site-consistent model. The results show that the proposed selection method is efficient, particularly at near collapse limit states.