Capitalisation of rex data for use in predictive reliability databases for the distribution of failure modes

The failure mode catalogue is a common data set used in reliability studies. It lists the failures and symptoms of failure for each type of component and suggests occurrence ratios. These lists are particularly useful for conducting FMECA or fault trees, for electronic card designers to make designs more robust against the most predictable faults, or for diagnostic in case of failure observation. However, the main limitation of these catalogues is their static nature. Failure modes and their occurrence distributions are highly dependent on the context in which the component is used, i.e. its life profile and the environmental stresses it is subjected to. As a result, the predominant failure mechanisms and modes vary. For a more accurate assessment, the FIDES methodology already proposes evaluating the failure rate by integrating environmental stresses via a detailed mission profile for each component family. In this article, we propose a method for capitalising on test data or data from operational feedback for use in predictive reliability databases. This method is based on two main approaches: The use of FMMEA (Failure Modes, Mechanisms, and Effects Analysis) based on the physics of failure and the design of Bayesian networks to update predictive reliability data. This statistical technique refines initial estimates by integrating new observed data. This methodology is then illustrated by an application to ceramic capacitors, where accelerated test data is used to update an existing predictive reliability study.

Keywords: Failure mode, PoF, Predictive reliability, FIDES, FMMEA.