Evaluation and Modeling of Scrap Utilization in the Steelmaking Process

Abstract

The study of scrap melting provides data for increasing the scrap utilization rate. Here, an evaluation model is established to analyze the effect of each factor on scrap melting using statistical methods for the first time. Subsequently, the quantitative relationship between the influencing factors and melting parameters is obtained. Back propagation (BP) neural networks and multiple regression are used for predictions. For scrap melting controlled by carbon mass transfer when the bath temperature range is 1573–1723 K, the relative contribution of each parameter was mixing power > bath temperature > specific surface area > carbon content. The predicted values of the BP neural network are more accurate than those of multiple regression. The relative errors of average melting rate, average mass melting speed, and mass transfer coefficient of training sets are 14.02%, 13.95%, and 7.19%, respectively, which decrease by 22.71%, 47.22%, and 69.46%, respectively, compared with those of the regression equations after outliers are removed.