Pattern classification of bearing faults in PMSM based on time domain feature ensembles

This research aims to identify an effective feature-based pattern classification technique that uses vibration and current data to identify bearing conditions. The authors attempted non-conventional time-domain features to detect the bearing conditions in permanent magnet synchronous motors (PMSM). This work employs two case studies utilizing eight datasets from Paderborn University to identify the bearing conditions of three and twelve classes. In this work, support vector machine, k-nearest neighbor, random forest, decision tree, and naive Bayes classifiers with 10% holdout validation are applied to study 31 feature combinations. This study also examines the Henry gas solubility optimization technique for feature selection to identify the most discriminating features. The results indicate that four feature ensembles consisting of 2 to 5 features performed better with the support vector machine, k-nearest neighbor, and random forest classifiers. In contrast to previous relevant studies, the proposed features are useful in identifying PMSM-bearing conditions with the highest accuracy of 99.8% and 99% using current signals for 3 and 12 classes respectively for combined current operating conditions.