2016/10/23 by Gero Walter, Walter, Gero, Frank P. A. Coolen +1
Decision Sciences · Engineering · Mathematics · #62 #FOS: Computer and information sciences #Methodology (stat.ME) #Probabilistic and Robust Engineering Design #Reliability and Maintenance Optimization #Statistical Distribution Estimation and Applications
paper · pdf · doi:10.48550/arxiv.1610.07222
openalex publication_date 2016/10/23 · openalex created_date 2025/10/24 · openalex updated_date 2026/07/28
In reliability engineering, data about failure events is often scarce. To\narrive at meaningful estimates for the reliability of a system, it is therefore\noften necessary to also include expert information in the analysis, which is\nstraightforward in the Bayesian approach by using an informative prior\ndistribution. A problem called prior-data conflict then can arise: observed\ndata seem very surprising from the viewpoint of the prior, i.e., information\nfrom data is in conflict with prior assumptions. Models based on conjugate\npriors can be insensitive to prior-data conflict, in the sense that the spread\nof the posterior distribution does not increase in case of such a conflict,\nthus conveying a false sense of certainty. An approach to mitigate this issue\nis presented, by considering sets of prior distributions to model limited\nknowledge on Weibull distributed component lifetimes, treating systems with\narbitrary layout using the survival signature. This approach can be seen as a\nrobust Bayesian procedure or imprecise probability method that reflects\nsurprisingly early or late component failures by wider system reliability\nbounds.\n