摘要

A new sampling strategy is proposed to improve the robustness and efficiency of the Kriging-based response surface method. In the initial stage, the number-theoretic method is utilized to generate deterministic initial design samples, which are further optimized by an interpolation method. In the sequential iterative process, a new sampling criterion called reliability infill sampling criterion is proposed. Besides, a more reasonable two-process strategy is suggested where the iterative process is divided into two sub-processes using different sampling criteria. The proposed method ensures that the approximation can be accurately and efficiently fitted near the limit state. The numerical examples and the applications in structural reliability indicate the improved accuracy and efficiency of the proposed method.