Enhancing the applicability of a neural network model for predicting the final phosphorus content of steel from a scrap-based electric arc furnace

The steel industry plays a crucial role in the transition to a sustainable economy, especially with the emergence of electric arc furnaces (EAFs). Electric arc furnaces melt scrap metal using an electric arc generated between the electrodes and the bath of liquid steel. This process allows for efficient recycling of steel scrap and significantly reduces carbon emissions. Although it is a sustainable operation that is expected to account for 50% of steel production by 2050, it faces challenges such as reducing phosphorus levels in the steel produced, as phosphorus negatively affects the mechanical properties of the steel. The behavior of phosphorus in an electric arc furnace is complex, influenced by the interaction of thermodynamic and kinetic phenomena, which are controlled by the initial conditions and process parameters. This problem is exacerbated by the diverse nature of scrap materials.

We developed an artificial neural network (ANN) model in collaboration with our industrial partner to predict the ultimate phosphorus content of steel using crucial process input parameters in a scrap-based electric furnace. The model demonstrated excellent predictive ability by capturing the complex relationships between the parameters of the electric arc furnace and the dephosphorization process.

As a next step, we aim to develop a user-friendly interface for the ANN model, allowing engineers and operators to test the effect of process parameters on the final phosphorus content of the steel. This will guide the optimization of process parameters to achieve the target phosphorus content in the steel. In addition, in the industrial context, process conditions are continuously subject to change, requiring maintenance and updating of the ANN model. Therefore, we aim to extend the current model with new historical data from the plant to improve its predictive ability.