Relevance vector machine based bearing fault diagnosis

This paper introduces a new bearing fault detection and diagnosis scheme based on Relevance Vector Machine (RVM) of vibration signals, i.e. two relevance vector machines are viewed as observer and classifier respectively. The observer is applied to identify and estimate various faults of bearing to gain fault state residual sequence while the classifier is used to classify multiple fault modes of bearings. Also, the algorithms constructing observer and classifier are discussed and reasoned. From the experimental results, we can see that estimation and classification based on RVM perform well in bearing fault diagnosis compared with neural networks approach, which indicates that this fault diagnosis method is valid and has promising application.