Broken rotor bar detection in variable frequency induction motor drives using wavelet energy based method

Abstract

The fault detection of an induction motor operated by variable frequency drives is a present industrial need as most of the line fed machines are replaced by power electronic drives, due to their improved speed regulation and fast dynamic response. This paper presents a technique based on discrete wavelet transform (DWT) for diagnosis of the rotor fault. According to the application, the speed and the load can change and therefore the amplitudes of the fault signals are modulated. A support vector machine is used in the detection to learn the complex relationship between the various frequency bands of wavelet features to obtain a model which is impervious to such variations. The sensitivity, specificity, accuracy and the ROC graphs indicate the effectiveness of the proposed method for detecting the fault.