The use of indirect evidence for Bayesian reliability analysis

摘要

This paper presents an approach that has been used to incorporate indirect evidence in Bayesian reliability analysis for automobile products during the design phase. Since no field performance data is available during the design stage of the product life cycle, reliability prediction has to rely on information that is not completely applicable to the product population of interest or the ultimate customer usage environment. The use of this indirect evidence requires special treatment to compensate for the differences in reliability performance due to differences in the actual design, operating environment, etc. When all available and relevant information is considered and the indirect evidence is subject to Bayesian analysis, a much more realistic estimate of the product reliability performance is provided. This reliability analysis method facilitates not only the exploration and documentation of relevant information but also the development of consensus on the evaluation and application of expert information. In addition, it permits the development of a reliability prediction for a new design before any testing has been done on the new product; this benefits the manufacturer in such areas as resource allocation during the development phase.