A Novel Hybrid Method for Fault Diagnosis of Industrial Equipment Based on Vibration Signals

Industrial equipment reliability is a critical issue that smart manufacturing systems must address. Implementation of online fault diagnosis for industrial equipment is a promising technology that can improve equipment reliability. Vibration analysis is a popular technique for equipment condition monitoring. For the non-linear and non-stationary features of industrial vibration signals, a hybrid model based on wavelet transform (WT), support vector machine (SVM), and grey wolf optimization algorithm (GWO) is proposed. The time-frequency domain features of fault signals are extracted by wavelet transform to solve the non-linearity of signals, and the obtained feature vectors are input into SVM for classification. Aiming at the problem of setting penalty coefficient and kernel function parameters, GWO is introduced to seek hyperparameters by using the hunting principle of wolves. In the experimental part, firstly, the comparison of the GWO-WT-SVM, GWO-SVM and SVM models was carried out. Secondly, Comparison with classical classification algorithms is made. Apparently, the GWO-WT-SVM model achieves higher accuracy than other models in bearing fault diagnosis.