Advanced surrogate-based time-dependent reliability analysis method by an effective strategy of reducing the candidate sample pool

For improving the efficiency of the time-dependent reliability analysis by the double-loop Kriging model combined with Monte Carlo simulation (DLK-MCS) and the single-loop one combined with MCS (SLK-MCS), this paper proposes a candidate sample pool (CSP) reduction strategy, and this strategy is nested to the DLK-MCS and SLK-MCS to respectively establish an advanced DLK-MCS (A-DLK-MCS) and an advanced SLK-MCS (A-SLK-MCS). In the proposed CSP reduction strategy, the samples with correctly recognized states by the current Kriging model are removed from the CSP gradually. Then in the process of updating the Kriging model, the samples with correctly recognized states can be avoided to be identified repeatedly, which results in the efficiency of the A-DLK-MCS and A-SLK-MCS which is higher than that of the DLK-MCS and SLK-MCS, respectively. Furthermore, the calibration operation is carried out for ensuring the accuracy of the convergent Kriging model obtained by the CSP reduction strategy. Three examples are used to verify the rationality of the proposed CSP reduction strategy. The results show that the proposed strategy can reduce the time of training Kriging model in both DLK-MCS and SLK-MCS while ensuring the accuracy.