Bayes Inference from Failure Data Contaminated Due to Maintenance

In operating plants, records are routinely taken on component-failures, maintenance actions, and component-withdrawals. In some cases, these data are the only available information on the component reliability (proper life-tests being infeasible due to cost, duration or other considerations). For these data to be suitable for inference on the parameters of the underlying life distributions, one has to account for the homogeneity-constraints on the stopping rules and the effect of maintenance. We generalize the sampling plan proposed by Barlow & Proschan for coping with incomplete data obtained under non-homogeneous stopping-rules, by allowing components to be maintained. A Bayes model accounts for the effect of maintenance.