Hybrid Modeling for Endpoint Carbon Content Prediction in EAF Steelmaking

Abstract

Considering the complicated and harsh conditions in the electric arc furnace (EAF) steelmaking process, the precise endpoint control technology is a crux that influences the product quality and production costs of the molten steel because precise endpoint control can control the endpoint carbon content and the endpoint oxidation. In this paper, a new hybrid prediction model was established to predict the endpoint carbon content in EAF steelmaking, which included the mechanism model based on the mass transfer process and the Extreme Learning Machine (ELM) optimized by the Evolving Membrane Algorithm (EMA). The mechanism model was calibrated with corrected parameters obtained from the ELM-EMA algorithm. As a result, the shortages that the mechanism model can’t work precisely and that the single mathematical algorithm model lacks the analysis of the metallurgy process were overcome by the hybrid prediction model. Meanwhile, modifying ELM algorithm by EMA algorithm can improve the generalization performance of single-hidden-layer feed-forward neural networks. The experiments on a 50t EAF demonstrated that the proposed model had a good generalization performance and good prediction accuracy.