Machine learning-based predictive maintenance: enhancing industrial reliability through data-driven approaches

Abstract

This study investigated the application of machine learning for predictive maintenance (PM) using synthetic data simulating industrial machinery failures. Different algorithms including random forest, support vector machine (SVM), artificial neural network (ANN), decision tree (DT), and logistic regression (LR) were evaluated in two test scenarios. Decision tree (DT) and logistic regression (LR) showed the best promise, despite challenges with data imbalance and data segmentation. However, these models are not yet suitable for industrial deployment due to the significant impact of misclassified faults. The results highlight the potential of machine learning to improve predictive maintenance (PM), while further improvements are needed before it can replace human supervision.