Reliability Optimization Design Based on Double-Criteria Kriging Model Considering Uncertainties of Distribution Parameters

Abstract

Uncertainty of distribution parameters has important influences on the system reliability, which will increase the complexity of reliability optimization design. Therefore, the reliability optimization design method based on the double-criteria Kriging model considering uncertainties of distribution parameters is proposed. The method consists of three key modules: the adaptive selection of sampling points, the approximation of failure probability function, and the reliability optimization solution. The double criteria adaptive Kriging model between uncertain distribution parameters and failure probability is constructed, the three-loop nesting problem of reliability optimization under uncertain distribution parameters is decoupled into the two-loop nesting problem of failure probability evaluation and deterministic optimization. The results of two engineering examples show that the proposed method can obtain accurate optimization results with fewer sampling points of uncertain distribution parameters than some existing methods, which can verify the effectiveness of the proposed reliability optimization method.