Machine Learning-Based Detection of Broken Bars in Dual Three-Phase Induction Motors: Methodology and Experimental Validation

Abstract

Dual three-phase induction motor drives are gaining attention in applications where reliability is a critical concern, such as automotive. Although the dual-stator winding reduces the impact of stator phase failures, these motors remain susceptible to broken rotor bar (BRB) faults. This paper presents a comprehensive methodology for their detection. The proposed approach encompasses all the key components of the detection algorithm, from motor modelling—used to generate a virtual dataset for training—to the selection of the most suitable variables for machine learning classification. Technical challenges, practical implementation insights, and experimental validation on real motors are discussed to provide a thorough understanding of the methodology and best design practices.