Quantitative FMEA and Functional Safety Metrics Evaluation in Bayesian Networks

For technical product development, especially with high demands concerning reliability and safety, Failure Modes and Effect Analysis (FMEA) and Fault Tree Analysis (FTA) typically are being applied to support an evaluation and verification of a system design. The method concepts generally are based on relating failure causes with their consequential failure effects on the system operation. Whereas FMEA provides a holistic overview on possible single failure modes, FTA can provide a quantitative appraisal for one system-level effect in each tree model. For this reason, the cumbersome elaboration of both may be beneficial to reveal as many potential deficiencies in the system design as possible, even though they refer to the same basis of system properties. Still, blind spots remain, such as mainly a lack of a holistic failure causal model that could provide an overall quantitative probabilistic evaluation. To strive for providing such a holistic quantitative methodological approach, this contribution demonstrates a modelling concept based on the more recent probabilistic FMEA (probFMEA) approach implemented in Bayesian Networks. The probFMEA approach is being revisited and subsequently its application demonstrated on an exemplary system. Therewith, its usability for reliability and safety analysis with a full system perspective is carved out. Especially the correspondence with safety related mechanisms and metrics according to the ISO 26262 standard for automotive functional safety is being referred to.