An efficient method for calculating system non-probabilistic reliability index

Collecting enough samples is difficult in real applications. Several interval-based non-probabilistic reliability methods have been reported. The key of these to estimate system index. In this paper, a new method proposed calculate Kriging model used replace time-consuming simulations, and the efficient global optimization determine training samples. A refinement learning function best component (or performance function) during iterative process. has considered two important factors: (1) contributions components nonprobabilistic index, (2) accuracy at current iteration. Two stopping criteria are given terminate algorithm. index finally calculated based on Monte Carlo simulation. numerical examples show applicability method.