Motor fault identification based on least squares support vector machine
Jianwen Zhang · Power System Protection and Control · 2012
In view of the structure of asynchronous motor and the characteristics of asynchronous motor faults, the combination of signal-processing technology and fault diagnosis with support vector machine (SVM) is proposed based on statistical learning theory. With the study of squirrel-cage motor fault, the rotor-fault-experiment system is set up and the fault signals are collected. The fault classification is realized with the use of Least Square Support Vector Machine (LS-SVM). Then two different ways of signal-processing technique: FFT and wavelet packet transform are used. The collected stator current signals and motor vibration signals are analyzed respectively and the fault-characteristic-vectors are collected. Then, with the use of LS-SVM classification technology, effective classification of the rotor fault is achieved. Finally, it’s indicated by the experimental results that the wavelet-LS-SVM way of classification is of high accuracy.