Induction Motor Diagnosis Based on Machine Learning

This work includes comprehensive analysis of machine learning methods applicability in the electric drive field. Authors reviewed methods of induction motor failure identification. Machine learning methods were compared with application to this task with further highlighting of the most suitable ones. Additionally the paper goes into details about the preprocessing pipeline, which may be useful in context of results reproducibility. Out of all, the ensemble classifier proved itself to be the most promising one, with highest accuracy. The developed system is characterized with a low cost due to the use of only two measuring current sensors and high reliability, that is why the developed system can compete with existing products on the market.