On efficient time-dependent reliability analysis method through most probable point-oriented Kriging model combined with importance sampling

Time-dependent reliability analysis (TRA) has attracted widespread attention due to its viability in evaluating the reliability of structures during the entire service life. However, TRA for complicated structures often leads to extremely high computational cost. To alleviate computational burden, this study develops a most probable point (MPP)-oriented Kriging model combined with the importance sampling (MPP-KIS) method for TRA. The strategy is that a comprehensive Kriging modeling method based on MPP is developed to construct the surrogate models for instantaneous performance functions discretized from the time-dependent reliability problem. A new learning function and a precise stopping criterion that take into account of the accuracy of the Kriging around MPP are contrived for updating the surrogate models. An adaptive screening strategy is introduced to identify the safe time trajectories to spare calculating the responses. The importance sampling method is integrated with the adaptive screening strategy for efficient computation of the time-dependent probability of failure. Two numerical examples and two engineering cases are exemplified to demonstrate the effectiveness and proficiency of the proposed method. The results show that the proposed MPP-KIS method achieves reliable results with substantially improved computational efficiency.