Novel Fault Diagnosis for Roller Bearing by using Multi Scale Sample Entropy based Clustering

The rolling bearing vibration signals are nonlinear and non-stationary when a fault exists. A novel method of rolling bearing fault diagnosis based on multiscale sample entropy (MSE) and Gath-Geva (GG) clustering is proposed for feature extraction. Firstly, the MSE method is used to calculate the sample entropy value in different scales. Secondly, the principal component analysis model is chosen to reduce the dimension of the MSE eigenvector. Then the main components which include the primary fault information are regarded as the input of GG cluster algorithm. The experimental results show that the method proposed in this paper can effectively identify various rolling bearing faults with better performance than fuzzy c-means (FCM) and Gustafson-Kessel (GK) algorithms. In the meantime, the effect of the Gath - Geva algorithmis the best in the three cluster methods.