Abstract: Based on the regenerative adaptive subset simulation (RASS) method, a modified regenerative adaptive subset simulation (MRASS) method is proposed for reliability and reliability sensitivity analysis of structural system. The method not only inherits the advantages of the four main mechanisms by the RASS method, the regeneration algorithm, the delayed rejection algorithm, the adaptive Markov chain process and thecomponent-wise sampler, but also improves the adaptive Markov chain process. By comparing the current acceptance ratio with the proposed ratio, MRASS method is able to take a better proposal probability density function variance value. Thus, the appropriate variance of the proposal distribution can be found with a clear target, which can improve the efficiency of sampling. Several numerical and engineering examples are used to illustrate the advantages of the presented method. The results show that MRASS method is more efficient, precise and robust than RASS method and the traditional subset simulation (SS) method in dealing with structure system with highly-dimensional random variables, small failure probability and highly nonlinear limit state equations.

Key words: Markov processes, reliability, sensitivity analysis, probability density function, distribution functions, modified subset simulation method