Three-phase Squirrel Cage Induction Motor Rotor Fault Intelligent Diagnosis

Asynchronous induction motors are widely used in the industrial and other fields because of its simple structure, low price, high reliability and convenient maintenance. With the improvement of motor stator manufacturing process, the trend of stator windings failure is declining, squirrel cage motor rotor manufacturing process has not been changed for decades, hence the induction motor rotor faults are becoming major cause of motor failure.This paper begins with the circuit model of rotor fault and focus on analyzing multi-loop current principle which is applied in the faults principle analysis. Based on the multi-loop current method, the rotor fault circuit model is concluded and the rotor parameters change is qualitatively analyzed. According to the stator current anglicizing methods, the stator current is sampled, the fault spectrum is extracted from the sampling data through FFT, the fault would be observed.Because asynchronous motor stator current signal is a typically random signal, the characteristics frequency of fault is very close to the main frequency at the same time, besides its amplitude is smaller and easily drowned by the main fault. Wavelet transform technology has a good capacity of time– frequency and a stronger capacity of extracting feature. The asynchronous induction motor fault is identified by the wavelet packet transformation technology. The experimental data is decomposed by 8-layer wavelet packet. The energy of characteristic frequency is extracted, which is used as the basis so that fault signal is effectively identified.Support Vector Machine has certain advantages than other intelligent diagnostic technique in the field of induction motor fault diagnosis. The principle of SVM is expatiated and a multi-class classification model for squirrel cage induction motor rotor fault diagnosis is proposed in this paper. At first, the signals of the sample stator current are analyzed with FFT, and the normalized spectral characteristics are used as the inputs of the SVM, then the multi-class classifiers are constructed according to the one against one strategy, finally, the rotor types of induction motor are identified when the results of subordinate classifiers have been inosculated.