Signal Processing based SVM Classifier for Mixed Fault Detection in Induction Motor

Use of induction motor is widespread in industry. Early detection of faults is crucial for reliable and economical operation of induction motor in industries. In this study fault diagnosis of induction motor under mixed fault including stator inter turn and broken rotor bar is introduced. The base structure of study consists of current signature analysis, feature extraction, SVM and diagnosis algorithm. Motor current signals are recorded for healthy and faulty conditions of motor. To get sight of the effects of current signals that are in time domain is transformed into time frequency domain via discrete wavelet transform (DWT). Five types of wavelets are used for signal processing to demonstrate the superiority of Db4 over other standard wavelets for accurate fault classification. Features extracted using DWT are applied to SVM which is used as a fault classifier. Results obtained prove the suitability of proposed method.