Application of machine learning for production optimization and predictive maintenance in an iron processing plant

Abstract

The modern metallurgical industry requires advanced solutions for process optimization, cost reduction, and predictive maintenance. This paper proposes a unified simulation-based framework using machine learning (ML) to jointly address production optimization and maintenance prediction in a virtual iron processing environment. Several ML models, including random forest (RF), extreme gradient boosting (XGBoost), light gradient boosting machine (LightGBM), support vector machine (SVM), and k-nearest neighbors (k-NN), were evaluated on synthetic datasets representing production, maintenance, and transport processes. A reproducible methodology was adopted, including preprocessing, time-aware data splitting, and cross-validation to prevent information leakage. Model performance was assessed using F1-score, area under the receiver operating characteristic curve (AUC), and regression metrics. Tree-based models achieved near-perfect classification performance (AUC ≈ 1, precision and recall > 0.99), while light gradient boosting machine (LightGBM) and CatBoost provided the best regression accuracy. Feature importance analysis using SHapley Additive exPlanations (SHAP) identified vibration and temperature as key maintenance indicators. Although based on simulation, the framework is designed for integration with supervisory control and data acquisition (SCADA) and the Industrial Internet of Things (IIoT), supporting real-time industrial deployment and alignment with operational key performance indicators.