Bayesian graphical model based optimal decision-making for fault diagnosis of critical induction motors in industrial applications

In an effort to achieve an optimal availability time of induction motors via fault probabilities reduction and improved prediction or diagnostic tools responsiveness, a conditional probabilistic approach was used. So, a Bayesian network (BN) has been developed in this paper. The objective will be to prioritize predictive and corrective maintenance actions based on the definition of the most probable fault elements and to see how they serve as a foundation for the decision framework. We have explored the causes of faults for an induction motor. The influence of different power ranges and the criticality of the electric induction motor are also discussed. With regard to the problem of induction motor faults monitoring and diagnostics, each technique developed in the literature concerns one or two faults. The model developed, through its unique structure, is valid for all faults and all situations. Application of the proposed approach to some machines shows promising results on the practical side. The model developed uses factual information (causes and effects) that is easy to identify, since it is best known to the operator. After that comes an investigation into the causal links and the definition of the a priori probabilities. The presented application of Bayesian networks is the first of its kind to predict faults of induction motors. Following the results of the inference obtained, prioritizations of the actions can be carried out.