Data-Based Fault Diagnosis Model Using a Bayesian Causal Analysis Framework

Abstract

This paper provides a comprehensive data-driven diagnosis approach applicable to complex manufacturing industries. The proposed approach is based on the Bayesian network paradigm. Both the implementation of the Bayesian model (the structure and parameters of the network) and the use of the resulting model for diagnosis are presented. The construction of the structure taking into account the issue related to the explosion in the number of variables and the determination of the network’s parameters are addressed. A diagnosis procedure using the developed Bayesian framework is proposed. In order to provide the structured data required for the construction and the usage of the diagnosis model, a unitary traceability data model is proposed and its use for forward and backward traceability is explained. Finally, an industrial benchmark — the Tennessee Eastman process — is utilized to show the ability of the developed framework to make an accurate diagnosis.