Application of Hybrid Wavelet-SVM Algorithm to Detect Broken Rotor Bars in Induction Motors

Induction motors are essential components in the modern manufacturing settings and can function continuously for hours without minor issues due to their reliability and construction. However, any fault in these types of machinery may result in extended downtime, costly maintenance, and safety issues. Consequently, advanced diagnostic methods that prevent possible failure by recognizing the early signs of deficiency can increase motors' reliability. In the past few years, researchers conducted many experiments to identify Broken Rotor Bars by integrating Motor Current Signal Analysis and Artificial Intelligence solutions. However, the motor may be subjected to load fluctuation, then oscillation related signatures exhibit similar behavior that of broken bar which leads misleading signatures. This overlapping frequencies, induced by Broken Rotor Bars and the Low-frequency Torque Oscillations (LTOs), can generate False Positive alarms and frustrates the diagnostic system. In this work, we propose a diagnostic algorithm to distinguish and disentangle the Broken Rotor Bar's frequencies from misleading LTOs, and detects Broken Rotor Bars by applying a Hybrid Wavelet-Support Vector Machines algorithm. The results verify the efficiency and reliability of our proposed algorithm compared to the existing methods.