A Novel Concept for Reliability Evaluation Using Multiple Deterministic Analyses

Abstract

A novel reliability analysis concept for large Structural/Mechanical systems represented by finite elements using multiple deterministic analyses is presented in this paper. The intent is to extract reliability information by conducing only tens instead of millions of deterministic analyses as an alternative to the classical Monte Carlo Simulation. It is particularly applicable when a deterministic analysis requires a considerable amount of computational time to satisfy the underlying physics. It is developed by integrating the second-order reliability method and an improved response surface method by removing its deficiencies. The efficiency of the integrated scheme is further improved by using advanced statistical and factorial schemes producing compounding beneficial effect. The concept is elaborated using two different illustrative examples. To validate the procedure, the underlying reliabilities are estimated by using the basic Monte Carlo simulation technique to develop the reference or benchmark value. Then, the accuracy and efficiency of the method are compared and verified.