Use of heritage data and other partially applicable information in reliability prediction of systems under development

This paper presents a methodology for the reliability assessment of systems under development. The methodology employs Bayesian data analysis techniques to estimate reliability measures based on various types of data such as warranty data, test data, and engineering judgments regarding the impact of design changes on the system reliability. It also provides the possibility of incorporating evidence concerning previous revisions of the same system or even information on systems that are only similar to the one under development. The methodology is validated against observed failure data from a diagnostic medical system. An example of application in the context of the automotive industry is presented as well.