Broken bar detection on IM using ROCOF and decision tree

Nowadays, instrumentation plays a leading role in new tendencies like Industry 4.0, the Internet of Things, or cloud computing. Instrumentation applied to induction motors for fault detection seeks to streamline scheduled maintenance. Maintenance tasks are imperative to ensure the correct operation of the induction motors. Therefore, fault detection is a topic of interest for researchers. Several techniques for electric machine monitoring have been developed to solve problems inherent to induction motors. Some are commonly based on spectral analysis that requires identifying the characteristic fault components. This paper proposes a novel technique non-based on identifying fault components using the Rate Of Change of Frequency of current signals and a decision tree for broken squirrel-cage rotor bar detection under three load conditions: unloaded, 50%, and 75% of load. Provided results reported an accuracy of 100% for the three loading conditions and 2% of false positives for the half-loaded motor.