A Machine Learning Framework for Bearing Fault Detection in Three-Phase Induction Motors

Three-phase induction motors are widely employed in industry due to their rugged performance and easy maintenance. Bearing faults in three phase induction motors are responsible for 40%-50% of unplanned shutdowns in industrial settings. Therefore, early detection of bearing faults is essential to implement preventive measures and enhance planning of maintenance strategies. This paper thus proposes a machine learning (ML) framework that consistently monitors acceleration and temperature of bearing to detect bearing faults. The results show that the ML framework using k-nearest neighbor (k-NN) and support vector machine (SVM) approaches is better than the variation-based thresholding approach, where the former method is able to detect faulty conditions with more than 99% accuracy.