2020 Volume 76 Issue 1 Pages 75-86
It is well known that probabilistic estimation of the residual seismic performance of existing bridges is important for their maintenance and it is hence desired to build an accurate and efficient structural reliability method. In this study, a surrogate modeling method, namely AK-MCMC, based on the adaptive Kriging and Markov chain Monte Carlo (MCMC) is introduced. The adaptive Kriging allows to automatically select important samples for constructing the surrogate model and MCMC searches intermediate failure regions, which will converge to the failure region, step by step. In order to extend the method to dynamic nonlinear problems, a method for calculating the failure probability based on Subset simulation using the obtained Kriging surrogate model is proposed. The applicability to the seismic performance estimation of an aging seismic-isolated bridge is examined. The results show that the proposed method is computationally very efficient and applicable to the seismic performance estimation of both the health and deteriorated conditions with different failure probability.