An Intelligent Approach for Bearing Fault Diagnosis Based on Bayesian Networks and Alpha-Stable Distribution

In the field of fault diagnosis, neural networks have been widely applied in distinguishing the type of fault. However, neural networks require training with large number of sample data, and as the input of neural networks, fault pattern vectors does not guarantee that can completely represent faults. Bayesian networks (BN) is a powerful method widely used to solve the problem with uncertainty and incomplete information. Alpha-stable distribution model is effective to describe the statistical characteristics of machine fault signal with impulsive behaviors. Therefore, the combination of the alpha-stable distribution parameters and Bayesian networks for a new intelligent diagnosis method is proposed. The results prove that the proposed method is effective for bearing fault diagnosis.