Fault diagnosis of ball bearings using Synchrosqueezed wavelet transforms and SVM

Bearing fault diagnosis is significant for the safe operation of machine devices. And the vibration signals can reflect the conditions of bearings, but they are non-stationary. In this study, a novel bearing fault diagnosis method is proposed by Synchrosqueezed wavelet transforms (SWT) and support vector machine (SVM). First of all, the vibration signal is decomposed into some intrinsic mode type functions (IMTFs) using SWT, and then the SWT energy entropy is computed, which shows the energy entropy varies with the fault types. Thus, the energy distribution of each IMTF is exacted as feature vectors. SVM is then used to classify the conditions of bearings. The experimental results show that this algorithm can achieve the accuracy of 100%. Besides, the effect of fault conditions on the Synchrosqueezed wavelet transforms is studied, and the results show that this method does not depend on fault severity.