Abstract: For random variable involving uncertain distribution parameter commonly existing in engineering, the uncertain distribution parameter is assumed as uniformly distributing in an interval, on which a new reliability sensitivity is defined and its solution is given by use of the directional sampling method. In the given directional sampling method for solving the new reliability sensitivity, Monte Carlo method is used to simulate the samples of the interval distribution parameter, and the directional sampling is employed to analyze the reliability sensitivity corresponding to the distribution parameter sample. Finally, the estimations of the reliability sensitivity can be calculated by integral on the whole distribution parameter space, the variances of the reliability sensitivity estimations by directional sampling method are derived approximately. Directional sampling based method is compared with double Monte Carlo based method by illustrations, and the results show that the directional sampling method is feasible and effective.

Key words: Directional sampling, Double Monte Carlo, Variance analysis, Reliability sensitivity