A Comparative Study of Hybrid AI Models for Predictive Maintenance in Machine Processes

Predictive Maintenance (PdM) plays an integral role in modern manufacturing by minimizing downtime, enhancing efficiency, and reducing costs. However, many existing approaches struggle with noisy data, limited interpretability, and inadequate scalability. This research introduces a hybrid Artificial Intelligence (AI) model for analyzing multivariate time-series data by integrating Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and Random Forest (RF) models fused through a Linear Regression (LR)-based meta-model. The ensemble captures spatial, temporal, and nonlinear patterns, while LR enables interpretable aggregation. The model achieves a precision of 0.88, recall of 0.65, F1 score of 0.75, accuracy of 98% and a Receiver Operating Characteristic Curve (ROC-AUC) score of 0.96. Furthermore, Shapley Additive Explanations (SHAP) are employed to assess feature importance and quantify the impact of base models which enhances model transparency. These findings demonstrate the model's robustness and practical potential for real-time PdM in industrial environments.