In system reliability analyses, computation of rare (failure) event probability is a main task. In most cases, the numerical model of a rare event is nonlinear and the resulting failure domain is often multimodal.
One strategy for estimating failure probability is the importance sampling (IS) method. The efficiency of IS depends on the choice of IS density.
With the focus on the sequential importance sampling, Kim Sang Rim, a researcher at the Faculty of Applied Mathematics, has investigated the improvement of its initial density.
First, he reviewed the sequential importance sampling (SIS). Then, he newly found the approximate optimal IS density using the scaling density. On this basis, he proposed a technique for estimating target failure probability by IS.
The numerical simulation showed that the proposed method seems to be appropriate to the problems for estimating probability lower than 10-6.