Rolling Bearing Fault Classification Based on Envelope Spectrum and Support Vector Machine

Based upon Hilbert envelope spectrum and support vector machine (SVM), a method for the fault diagnosis of rolling bearing is proposed in this paper. Targeting the modulation characteristics of rolling bearing fault vibration signals, the Hilbert transform based envelope spectrum analysis is used to extract fault bearing features. In the envelope spectrum, character frequencies are quite clear and can be used as a reliable source of information for bearing diagnosis. Basic SVM is originally designed for two-class classification problem, while bearing fault diagnosis is multi-class case. A new bearing fault diagnosis system based on “one to others” SVM algorithm is presented to solve the multi-class recognition problems. Practical vibration signals measured from rolling bearings with ball fault, inner race fault and outer race fault are analyzed by the proposed method. The results show that the proposed method provides accurate diagnosis and good diagnostic resolution.