Development of Feature Extraction and Classification for Bearing Fault Analysis of Induction Motor

Faults in induction machines are preventable and treatable upon early detection. Among various induction machine faults, failure of the bearing is considered as the leading cause of the shutdown of the industries and accounted approximately 41%. It is characterized by persisting of vibration signal which is usually very low in amplitude and associated with frequencies. An investigation has been carried out on enhancing the diagnostic relevance of vibration measurement using the identification of representative features and classification of damage, the location of damage and severity of damage. To fulfill the purpose, a hybrid feature pool has been prepared using calculated features from time, frequency, and time-frequency domain. In feature pool, there are total forty-six features which have been calculated and the optimal features have been selected using principal component analysis (PCA) and minimum Redundancy Maximum Relevance (mRMR) from feature pool. The representative features are input to the Radial Basis Function Neural Network (RBFNN), Support Vector Machine (SVM), and Random Forest (RF) Classifier. The performance of classifier has been compared.