A smart modelling for the casting temperature prediction in an electric arc furnace

The efficient and reliable control of an electric arc furnace (EAF) is a challenging problem, due to the strong intercorrelation among process variables, the large dimension of the input and output space, the strong interaction among process variables, a large time delay, and a highly nonlinear behaviour. This paper presents a model that allows us to optimize the control and, therefore, the electric power consumption in an EAF. The data used for this study were collected from Bizkaia Steel Mill (Arcelor Company). Neural network and fuzzy logic techniques have been applied on these data in order to get an improved model of the casting temperature inside the furnace's hearth. First, we developed some neural network models with different topologies and input variables. Then we used the best model obtained in the previous step to combine it with a fuzzy logic technique to get the final model. Comparison with experimental data and other models is carried out to validate the proposed model. Finally, the conclusions and future studies are exposed.