Local reliability based sensitivity analysis with the moving particles method

Local reliability sensitivity methods aim at determining the partial derivatives of the failure probability or the reliability index with respect to model parameters. For efficient local reliability based sensitivity analysis, it is important to avoid repeated evaluations of the performance function. To this end, an extension of the moving particles method to local reliability based sensitivity analysis is presented that is completely based on the already evaluated samples for the reliability estimate and thus avoids repeated evaluations of the performance function. In order to further reduce the variance of the estimator and to increase the efficiency, a multilevel variant of the estimator is proposed. The method is discussed in detail and illustrated by means of examples.

Automated summary

The local reliability sensitivity methods aim to determine the partial derivatives of failure probability or reliability index with respect to model parameters. To avoid repeated evaluations of the performance function, an extension of the moving particles method is proposed for efficient local reliability sensitivity analysis. Additionally, a multilevel variant of the estimator is suggested to further reduce the variance and increase efficiency.