A Bayes Optimized SVM and Frequency Estimation for Diagnosis of Eccentricity Faults

Abstract

Induction motors are the most used electromechanical energy conversion device in today's industry. Although these motors are generally robust, they are exposed to faults in different components. The most common component that faults occur is the bearing related faults. Especially in motors fed by the drive, it is more difficult to determine the faults caused by the operating conditions at different speeds. The main reason for this is that the location of the frequency component related to the fault is not fixed and extra sensors are required for its calculation. In this study, a new method is proposed for determining eccentricity faults in motors that are powered by the drive and have wide operating conditions. The proposed method automatically estimates the location of the frequency component and supply frequency associated with the fault, without the need for extra sensor information. By evaluating the amplitudes of the frequency components of three phase currents with bayes optimized support vector machines, both fault type and its severity can be determined. The accuracy of the proposed method has been confirmed by experimentally obtained data.