Failure modes and effects analysis (FMEA) is a traditional safety analysis method that has a worldwide recognition across many industries. However, there are documented limitations to the technique, which make it difficult for direct conversion to the diagnosis models. This paper aims to contribute to the reuse of FMEA knowledge through a novel knowledge modeling approach. Taking the structure model into consideration, we define the failure modes as the inherent component properties at different hierarchical levels and propose a formal failure knowledge representation model based on the polychromatic sets theory. It is a structured modeling technique for representing FMEA causal knowledge in unified mathematical language, which in turn provides the framework for organizing the failure modes, the causes and effects more systematically and completely. Using the iterative search process operated on the reasoning matrices, the diagnosis reasoning for more than one top-level failure event can be realized automatically. The research shows this new approach has formed mathematical foundation which can make significant progress in FMEA knowledge formalization.
