AI-Enabled Predictive Maintenance for Industrial Equipment: Reliability, Probability Modeling, and Applications

In this work, a predictive maintenance model for gas turbines is proposed, which combines machine learning classifiers and reliability models. Random forest (RF) and support vector machine (SVM) models were employed to classify fault states on a real-world dataset comprising 1,386 records, featuring nine sensor features and binary fault labels. Reliability analysis was conducted using the two-parameter Weibull distribution, which was fitted to the inter-failure intervals. Evidence indicates that SVM performed pretty well in detecting faults (39.6%), whereas RF experienced a significant drop in recall. The Weibull fit yielded β = 1.2416 and η = 2.7957 (arbitrary time units), consistent with early wear-out behavior. A combination of classification outputs with reliability functions (r(t), f(t), h(t)) yields risk-based maintenance planning. Such results indicate the significance of considering data imbalance, feature engineering, and the combination of AI and probabilistic models towards sound predictive maintenance.