A BAYES-COMPETING RISK MODEL FOR THE USE OF EXPERT JUDGMENT IN RELIABILITY ESTIMATION

In an analysis of the reliability of heat exchangers, a number of causes of failure and a number of potential failure patterns were identified. To allow a simplification of the analysis, five independent groups of failure causes were defined and formed the basis of a competing risks model. Further, lack of data and poor quality data required the experience of 'experts' to be used in a quantitative way through Bayesian methods. Lastly, to simplify the elicitation of the prior densities, a shape, scale and location parameter model for the failure time distributions was adopted.