Decision-Based Approach for Reliability Design

Abstract

We propose a decision-based approach for reliability design when there is insufficient information for constructing probabilistic models. The approach enables a designer to perform reliability-cost trade-offs and to assess the importance of variability and epistemic uncertainty. A method for decision under epistemic uncertainty is first presented and justified by presenting axioms on a decision maker’s (DM’s) preferences and by assuming that the DM’s goal is to find the most immune act (in terms of having undesirable consequences) to deviations of the state of the world from an expected state. Thus, the philosophy of the method is similar to that of robust reliability (, 1996, Robust Reliability in the Mechanical Sciences, Springer-Verlag, Berlin). A new formulation of reliability design problems is proposed based on the above decision method and is compared to two reliability-based design optimization formulations that minimize cost given a maximum acceptable failure probability or maximize expected utility. The method is demonstrated on a decision where a designer has to choose between two materials for a structure.