Evidence and Optimization Theory Based Multi-Source Information Fusion for Reliability Demonstration Test Plan Design

In both the development and use stages, reliability demonstration tests are crucial for assessing product reliability over time. However, obtaining product distribution data is often challenging due to limited testing. This paper proposes and compares four methods for determining distribution parameters by utilizing optimization with complex boundaries and D-S evidence theory. These methods reduce uncertainty in distribution parameters by integrating prior information such as various expert judgments. Specifically, D-S evidence theory fuses information at the distribution level, while the other three methods operate it at the feasible domain level. Additionally, risk calculation methods are derived. A simulation example is presented to evaluate the efficacy and practicality of these methods by calculating test time, failure numbers, and risks for different prior information scenarios. The results demonstrate that incorporating prior information significantly lowers risks in reliability demonstration test plans, and the optimization method yields lower risks compared to the D-S evidence method for the same test plan.