Reliability Assessment from Life Tests and Expert Testimony during the Development of a New Car Model

During the development of a new car model the manufacturer usually has to redesign subassemblies in order to attain higher reliability levels. Being new subassemblies mainly evolutionary instead of revolutionary, the Bayesian approach allows to combine corporate memory, expert testimony and experimental data in order to obtain realistic (and not optimistic) reliability assessment for the new model. Because of this specific advantage the Bayesian approach is now adopted in automotive industry, in spite of the difficulty in establishing prior and in managing the results of the Bayes theorem. Actually the sources of knowledge to be fused into the Bayes "melting pot" are: a) the corporate data bank, which provides the record of the number of failures occurred to the old version of the subassembly during the warranty period; b) the testimony of design engineer, who has a sound opinion about the expected reliability improvement due to his design revisions; c) the hard data, which are collected through life tests on the new version. In this paper, the authors present a Bayesian procedure to assess the reliability of a new car subassembly by means of a fully developed case study, which concerns the development of a revised version of a specific subassembly by means of a suitable quality improvement program. This methodology is useful to support the development of a new product: the reliability demonstration phase could be suitably arranged only after a Bayesian analysis has evidenced high credibility that the reliability target has been attained, saving test cost and developing time.