An ontology-driven bayesian network approach to fault diagnosis and correction in manufacturing

Efficient Fault Diagnosis and Correction (FDC) is crucial for maintaining high availability in manufacturing systems. This paper presents a novel ontology-driven approach for creating Bayesian Networks (BNs) to support decision-making in FDC. An ontology is developed to represent fault knowledge from domain-specific data sources, including Failure Mode, Effects, and Criticality Analysis (FMECA) data. This ontology is used as a template for creating the BN structure. The probability parameters of the BN are identified heuristically via FMECA criticality information, deterministic relationships, and expert knowledge. A case-based evaluation using an assembly line for solenoid valves demonstrates the BN’s accurate FDC prediction performance.