Machine Learning Based Predictive Maintenance: Comparative Evaluation of Fault Prediction Models

Abstract

This study investigates the effectiveness of machine learning techniques in predictive maintenance for reducing unexpected machine faults and optimizing maintenance strategies. A synthetic dataset was utilized to model Faults based on key operational parameters such as temperature, torque, rotational speed, and product quality. The study applies Logistic Regression, Decision Tree, Random Forest, and Support Vector Machine to classify machine Faults and evaluates their performance using accuracy, ROC-AUC, and confusion matrices. The results indicate that Random Forest achieved the highest accuracy, followed by Decision Tree, while Logistic Regression and Support Vector Machine demonstrated stable but slightly lower performance. Analysis of operational variables reveals that faults are most frequent in high-torque, low-speed conditions, and lower-quality products exhibit higher fault rates. Additionally, temperature variations play a significant role in Fault occurrence, particularly in cases related to Heat Dissipation fault.