Mahine Learning-Based Fault Detection of Electric Motor Bearings Using Vibration Analysis

Abstract

Bearing faults are a leading cause of electric motor failure, often resulting in costly downtime and maintenance. This research proposes a machine learning-based framework for early fault detection using vibration signal analysis. Vibration data from a bearing test rig under five operating conditions were collected and transformed using Fast Fourier Transform (FFT) to extract key frequency-domain features, while Principal Component Analysis (PCA) was used to reduce data complexity. Five machine learning models were evaluated, with the Medium Gaussian SVM achieving the highest accuracy of 96.8%. Although PCA-based models like Ensemble Subspace KNN had slightly lower accuracy (94.0%), they significantly reduced training time. The findings highlight a trade-off between classification accuracy and computational efficiency, with FFT offering superior fault-specific pattern retention and PCA enhancing speed. The proposed approach offers a scalable solution for real-time, predictive maintenance, and future work may explore hybrid feature extraction or deep learning to further improve performance.