One-Class Classifiers for Detecting Faults In Induction Motors

This paper deals with the problem of broken rotor bar detection in induction motors. In this technique current signal monitoring of an induction motor is applied to characterize its operation condition, and identify the normal state from broken rotor bar situation. To this aim, a fault detection scheme is proposed, which makes use of a low computational cost preprocessing method along with the six state-of-the-art one-class classifiers. Moreover, a feature selection approach is employed to form a suitable feature subset that is highly correlated with the operating condition of the machine. The achieved experimental results show the effectiveness of the proposed detection technique.