| This paper is mainly concern about motor fault diagnosis, which is based on the research of fault diagnosis domestic and abroad motor fault diagnosis, as well.It proposes a new diagnostic method based on Wavelet and Support Vector Machine (SVM). The experiment results have proved that the method of the combination of wavelet and SVM can effectively examine and diagnose motor fault.In view of the principle of the asynchronous motor, the motor's noise, vibration, and other common faults are analyzed at first. It has been elaborated that generate noise and vibration characteristics of the causes and failures. The motor's noise and vibration signal have been acquainted through the induction motor experimental systemAs electrical failure, the fault signal often contains a lot of composition which is time-varying, and breaking at short time. The Wavelet packet has time-frequency and multi-resolution feature. Therefore, not only the whole signal but the partial signal can be analyzed. The fault feature of non-stationary transient signal can be caught correctly. Hence, this paper uses method of wavelet packet of the acquainting signal to reduce the noising process.The feature energy of frequency-band has been got as the input feature vectors of motor fault diagnosis.Aiming at the characteristics of fault diagnosis with finite samples, the problem of over-learning and less-learning, which is evoked by artificial neural network algorithm, has been solved effectively by applying SVM to motor fault diagnosis system.The method of Least Squares Support Vector Machine (LSSVM), the betterment of SVM has been applied to fault diagnosis of motor. By comparing the RBF neural network and LSSVM, the result of experiment indicates that the diagnose algorithm of SVM is better than that of neural network. In conclusion, the methods of the combination of wavelet and SVM have been proved to be effective. |
