Bayesian Reliability Inference on Innovated Automotive Components

The need to assess the reliability of new automotive products in a timely manner compels manufacturers to make use of early failure warranty data. However, the narrow observation period and the moderate sizes of early warranty data sets result in reliability estimates that are not very accurate. Nevertheless, when a new product is not revolutionary but instead the result of making improvements to its predecessors, past failure data in conjunction with corporate technical knowledge are a valuable source of information which can be usefully exploited in a Bayesian estimation framework. To this end, a Bayesian procedure was developed which is based on a rigorous formalization of both objective information provided by observed failure data for past products as well as subjective information on the effectiveness of design or process modifications introduced into new products to improve their reliability. Information on modified working conditions is also formalized, and the effect of requested cost reductions for outsourced components is considered. The proposed procedure is then applied to a case study relating to a newly revised component that is assembled in a car model that was already on the market, and its ability to support reliability estimates and management decisions is addressed.