A general discretization procedure for reliability computation in complex stress-strength models

摘要

This paper proposes, implements, and evaluates an original discretization method for estimating the reliability of systems for which stress and strength are defined as complex functions of continuous random variables, when reliability is not derivable through common analytic techniques. This method is compared to two other discretization approaches appeared in the literature, and subjected to a comparative closeness study comprising some engineering applications. In this study, both a normal and a non-normal distribution for the random variables involved are analyzed, focusing in the latter case on the Weibull distribution. The results show that the proposal is very effective in terms of the closeness of the estimates to the true (simulated) value of reliability. The method, due to its general applicability, is theoretically suitable for any parametric family and works with a small fraction of the calculation load needed for obtaining the true value by Monte Carlo simulation.