An Investigation on Early Fault Diagnosis Based on Naive Bayes Model

In rotating machinery, the risk of mechanical failures can be reduced by analyzing the occurrence of faults in bearings at an early stage. Therefore, it is essential to carry out early fault diagnosis of rolling bearings. This paper presents a method for early fault diagnosis of rolling bearings with the combination of wavelet analysis and Bayesian classification. The wavelet noise reduction method was selected for noise removal from the early fault signals of the bearings firstly, and then the energy features were extracted through wavelet packet decomposition. The naive Bayes classifier was utilized to discriminate early faults in bearings according to the features obtained. Experiments have been performed and the results proves the effectiveness and superiority of the method, where the proposed approach can realize the classification accuracy as high as 100%.