A Framework to Support Failure Cause Identification in Manufacturing Systems through Generalization of Past FMEAs

This study proposes a framework for inferring the causes of failures occurring in manufacturing systems from past Failure Mode and Effect Analyses (FMEAs) conducted on other systems to assist in inspecting and maintaining the systems. Among various manufacturing systems, a framework to search past FMEAs and the corresponding causes of the failure requires solving the following problems. First, the difference in products, equipment, and wording to represent them make it difficult to search the similar failure phenomenon from FMEAs. Secondly, the causes of failure highly depend on the process flow of the system until the failure occurs. Therefore, it is also hard to find appropriate failure causes from FMEAs without reflecting on the process. The framework solves the first issue by generalizing descriptions in past FMEAs based on structured concepts of manufacturing systems in an ontology before inference of causes to address. Furthermore, the framework analyzes the correspondence of the process flows between the target manufacturing system and past FMEAs using a process order model generated by SysML diagrams to solve the second issue. The comparison between the causes inferred by the proposed framework and by skilled experts for three typical failures in the manufacturing system and the interview with them about the plausibility of the inference results showed that more than 73 % of them were valid.