Reliability estimation following a field intervention

Comparison of life distributions is a common problem in product reliability applications. These comparisons are used to assess new materials, evaluate alternative designs, and make competitive product analysis. A standard comparison involves independent samples of life data from multiple populations. The case study application of this paper deals with a newly designed brush for wind power generators. The level of stress exerted on some of the in-service brushes was deemed too high and was reduced through field intervention. It is desired to assess the effect of stress adjustment on brush life. The standard methodology for comparing the before and after field adjustment samples could not be applied because the samples were not independent. We use the cumulative damage framework to develop a statistical model for brush life. The application of the model to the brush problem provided maximum likelihood estimates of life distributions before and after the change in the stress level. Additionally, the analyses revealed important insights in support of the engineering investigation into the underlying failure modes. We discuss possible extensions of the proposed approach.