Modelling unexpected failures with a hierarchical Bayesian model

Abstract

Systems, especially those in the design and development phase, frequently suffer from unexpected failures, which are caused by insufficient knowledge of the system failure processes. In this paper, we develop a hierarchical Bayesian reliability model that account for unexpected failures. For this, the sample space of all failures is broken down into subspaces of expected failures and unexpected failures. The overall reliability is, then, derived based on the total probability theorem. A Bayesian network model is developed to explicitly compute the probability of unexpected failures. A single board computer reliability analysis is considered and the results show that neglecting unexpected failures overestimates the reliability.