Modelling of melting dynamics of Direct Reduced Iron (DRI)

The global steel industry is undergoing a radical transition towards production methods with lower carbon intensity to meet the net zero emissions goal by 2050. In this context, the integration of direct reduced iron (DRI) into electric arc furnaces (EAF) and electric smelting furnaces (ESF) is central to the decarbonization process. Consequently, control ling process parameters to optimize the efficiency of these new technological pathways has become a primary focus at the forefront of scientific research. This work, developed in collaboration with the Danieli Research Center, introduces a hybrid computational framework for the rapid and accurate prediction of the melting dynamics of metallic spheres in a molten bath. The methodology integrates a physics based reduced-order model focusing on conduction and phase-change (discretized via FDM in a Python environment) with data-driven surrogate models, including artificial neural networks (ANN) and deterministic analytical expressions. Through the generation of an extensive database of 12,000 simulations, the study identifies a critical transition between a diluted and a non-diluted regime, the latter being dominated by thermal interference between particles. The original contribution of this work lies in the definition and analysis of melting intensity, defined as the melting rate per unit of bath volume. The results demonstrate that maximum furnace productivity is not achieved when the melting time of a single sphere is minimized; instead, optimal productivity is reached when the melting intensity is maximized. This occurs at the point where the increase in mass concentration within the bath effectively compensates for the kinetic deceleration caused by thermal interference. These findings serve as a foundation for advanced engineering tools, enabling optimized furnace design — specifically in determining the optimal number and positioning of feeding points — and providing a robust base for the development of online digital twins. Such systems allow for the real-time control of process parameters, directly aimed at maximizing production efficiency in next-generation steelmaking plants.