Modified Time-Domain Features for Permanent Magnet Synchronous Motor Bearing Fault Detection Using Current Signal

A new modified time domain features are proposed namely, average energy, absolute value of summation of square root, log detector, modified mean absolute value, and absolute value of summation of exponential root to classify the bearing fault through phase current signal as well as vibration signal using support vector machine (SVM), k-nearest neighbor (k-NN), and linear discriminant analysis classifiers for four different operating conditions in eight derived datasets. In this work, the proposed modified time-domain features are experimentally investigated using Paderborn University bearing database with healthy, outer race and inner race faults. Finally, the paper concludes, the current signal based fault classification accuracies performs better in comparison with vibration signal under individual operating conditions for all possible 31 feature ensembles. The classification accuracies of modified time-domain feature yields up to 100% on five different bearing conditions with SVM and k-NN classifiers using motor current signal under operating condition-1. The proposed features shows that not only good accuracy using motor current signal, but also have less computation time.