AK-PDEMi: A failure-informed enrichment algorithm for improving the AK-PDEM in reliability analysis

A failure-informed enrichment algorithm is devised to improve the performance of existing adaptive Kriging-probability density evolution method (AK-PDEM) for reliability analysis. This improved named AK-PDEMi. Contrary empirically prescribing sample size representative points in AK-PDEM, point set AK-PDEMi sequentially enriched by new sets generated a scheme, which aims making fine partitions key sub-regions where make critical contributions failure probability. In this regard, double-loop configuration devised: inner loop adaptively refines accuracy Kriging model reduce Kriging-induced error, and outer involves process alleviate PDEM-associated discretization error. The loops are complementary proceed until both their convergence criteria satisfied. Three numerical examples studied comprehensive comparisons made between proposed other conventional algorithms. Results show that shows remarkable advantage over AK-PDEM.