Real-Time Hot-Rolled Coil Placement Recommendation System with Data-Driven Model

Abstract

Hot-rolled coils (HRCs) are essential in various industries, including automotive, construction, and machinery. However, the cooling process of HRCs in the yard tends to be nonuniform because of complex thermal interactions between adjacent coils and varying environmental conditions, which affect the mechanical properties and steel quality. In this study, we used simplified heat transfer models based on the finite element method (FEM) to generate realistic simulation data. We developed a novel management system that integrates two trained artificial neural networks with deep and wide networks using hyperparameter tuning to improve prediction speed, a known limitation of FEM. The system predicts temperature variations at multiple points on the coil, enabling strategic placement that minimizes temperature deviations and enhances cooling uniformity. This real-time computational approach eliminates the necessity for additional cooling equipment and ensures high product quality. The system's efficacy was validated through case studies, revealing dynamic adjustments and optimized placements. The proposed system achieved a mean absolute error of 3.44 and a mean absolute percentage error of 0.24%, outperforming conventional regression techniques. These results demonstrated the effectiveness of the system in simulating real-world cooling scenarios and its feasibility for real-time cooling optimization in steel manufacturing.