An efficient algorithm for analyzing multimode structure system reliability by a new learning function of most reducing average probability of misjudging system state

Abstract

• A new MRF learning function for multimode structure system is proposed. • A contribution function is established and explicitly derived. • A reasonable convergence criterion is employed. • The accuracy and efficiency of the proposed method are verified by five examples. For efficiently analyzing reliability of multimode structure system by adaptive Kriging model combined with Monte Carlo simulation, a new most reduction learning function (MRF) is proposed in this paper. Firstly, a contribution function is established to quantify the effect of candidate sample point (CSP) on reducing average probability of misjudging single mode structure state in the differential domain adjacent to CSP. Secondly, the contribution function is explicitly derived after selecting a reasonable size of the differential domain. Thirdly, by the maximum expected contribution function of single mode structure and the logical relationship among modes in multimode structure system, the new learning function named MRF is obtained. The main novelty of this work is proposing a new MRF based strategy for multimode structure system reliability analysis, in which it considers the logical relationship among modes in multimode structure system, probability of misjudging single mode structure state, probability density function and the absolute value of Kriging prediction of single mode structure at CSP. Compared with the up-to-date methods, the proposed strategy for multimode structure system reliability analysis is easier and more efficient to be executed, and its superiority is verified by numerical and engineering examples.