Fault tree inference for one-shot devices using Bayes and empirical Bayes methods

Fault Trees are useful in structuring engineering judgement through explication of the system logic and of potential causes of failure. During design and development, systems may have few data on which to support quantification of the model. This results in heavy burdens being placed on the subjective quantitative assessments by the engineers or reliance upon generic data sources. Data might exist for relevant designs with common system heritage, but it is not clear how these data can be adapted to inform the assessment of the new design. Of particular interest is inference support on events that potentially could have occurred on heritage designs but have not been observed. We explore the use of Empirical Bayes methods, in conjunction with subjective Bayes methods to accomplish this goal in the context of reliability modelling of a one-shot device. The modelling approach and an industrial application are described.