Smart Predictive Maintenance: AI-Driven Adaptation for Industrial Equipment

The machinery manufacturing sector has steadily raised its demands for suitable production technologies as social competitiveness has become more intense. This study proposes a machine learning-driven framework leveraging Deep Neural Networks (DNN) for fault detection and predictive maintenance using the MFP dataset. The data undergoes comprehensive preprocessing, including outlier detection, missing value handling, and normalization using robust scaling techniques. “Accuracy” (acc), “precision” (prec), “recall” (rec), “F1-score” (f-measure), and ROC-AUC (roc-auc) are vital performance measures that are used to train and assess the suggested DNN model. Results from experiments show that the DNN model reaches 93.21% accuracy which establishes strong performance over Deep Belief Networks (DBN), Random Forest (RF), and Isolation Forest. Moreover, the high AUC score of 0.978 validates that the model is sufficiently robust to discriminate between normal and faulty machinery conditions. Although the model has great predictive capability, it still faces computational costs and deploys in real-time problems.