Dominant failure mode analysis using representative samples obtained by multiple response surfaces method

Abstract

The conventional methods for failure mode analysis usually fail to identify dominant failure modes efficiently for large structures. To overcome this issue, an approach based on representative samples is proposed, which combines the MRS (multiple response surfaces) method with iterative strategies. The main steps are: (1) use MRS method to approximate the system failure function piecewise and to search the multiple design points; (2) perform deterministic structural analysis to identify failure sequences for samples in the important domain rather than the total domain; (3) solve representative samples with iterative strategies to obtain a converged solution based on a visualization plot. A key merit of this approach is that it can identify dominant failure modes efficiently due to the utilization of samples with most contributions to system failure, such as design points, etc. Numerical examples show that the proposed method can be used well to search dominant failure modes for structures.