Stochastic Response Surface Method with Advanced Sampling Approach for Reliability Analysis

摘要

The conventional stochastic response surface method (SRSM) has been demonstrated to be effective in reliability analysis.However, SRSM may produce large error on PoF estimation in the cases of low probability of failure (PoF). To mitigate this issue, a new form of SRSM with a advanced sampling strategy is proposed in this paper, in which the sample points are generated around the most probable point rather than the mean point.Since more sample points are located in the local region that is of great importance to the PoF estimation, the proposed new form of SRSM is supposed to yield more accurate PoF estimation than SRSM. The effectiveness and advantage of the new SRSM are demonstrated by comparative studies on three numerical examples.An application of it to the rocket design further demonstrates its effectiveness for PoF estimation.