In this paper, an efficient methodology has been presented for solving the RBO problems. It combines the weighted approach and a sequential approximation optimization. The ‘weighted approach’ has been used to decouple the RBO problem by explicitly approximating the reliability constraints. It can obtain the FPF through a single reliability analysis. The framework of sequential approximation optimization is adopted to ensure the accuracy of the RBO problem efficiently. That is, after the reliability analysis and optimization loop are decoupled, they can be performed sequentially until the optimization results converge. The proposed method does not depend on approximate reliability analysis method and allow simulation method to use, leading to higher accuracy and actual feasibility.
Numerical results reported in this contribution indicate that the most desirable feature is that the effort for constructing approximation of the FPF only need one traditional reliability analysis, i.e., importance sampling, thus the proposed approach can be competitive in terms of efficiency and accuracy when compared with other methods (e.g. that of using an predefined exponential function). In addition to these advantages, a complete decoupling of the required optimization and reliability calculations has been obtained. Hence, any algorithm that can solve the deterministic optimization problem can be used.
Further numerical studies are needed for researching the following possible subjects include (a) extensive application to high dimensional problem; (b) application to large number of design variables that can be handled with the proposed framework.