Integration of Machine Learning Algorithms for Predictive Maintenance in Electrical Machines

Abstract

Electrical machines are critical assets in laboratories and industries but are often susceptible to failures such as insulation breakdown, overheating, vibration anomalies, and electrical imbalances. Traditional maintenance practices have remained reactive and schedule-based, resulting in costly downtime and safety hazards. The aim of this study is to develop a predictive maintenance framework that integrates machine learning (ML) algorithms with fault detection and diagnosis (FDD) to enhance the reliability of electrical machines. The objectives are to (i) simulate machine faults using MATLAB/Simulink and utilize public datasets, (ii) train and compare ML classifiers including Artificial Neural Networks (ANN), Support Vector Machines (SVM), and Decision Trees, and (iii) evaluate their performance in predicting fault occurrence. Anchored on the Reliability-Centered Maintenance (RCM) theory, the study applies feature extraction techniques such as FFT and wavelet transforms, followed by supervised learning classification. The findings indicate that ANN achieved the highest accuracy (96.5%), precision (95.8%), and recall (97.1%), outperforming both SVM and Decision Tree models. The study concludes that integrating ML into predictive maintenance improves diagnostic accuracy, minimizes downtime, reduces costs, and enhances safety, thereby offering practical applications for both academic and industrial environments.

Keywords

References