A Novel Bayesian Melding Approach for Reliability Estimation Subjected to Inconsistent Priors and Heterogeneous Data Sets

Abstract

Motivated by practical engineering requirement, we propose a Bayesian-based approach mainly addressing the reliability estimation of the system under the imperfect situation. In particular, a novel Bayesian melding approach is developed in the presence of inconsistent priors. Limitations that embedded in traditional melding approach have been eliminated by setting the weighing factor as a hyper parameter. We implement our proposal via a modified sampling importance resampling algorithm, which is developed to balance the contributions of two different probability distributions in a flexible way. Both numerical case and practical engineering application example are demonstrated for validation and benefit illustration.

Keywords

• System reliability
• inconsistent priors
• Bayesian melding
• multi-source data fusion