Analysis of Blast Furnace Permeability Regulation Strategy Based on Machine Learning

Abstract

The permeability index of a blast furnace is an important parameter to characterize the reasonable countercurrent movement between the charge and the gas flow. The prediction modeling and regulation of the permeability index are of great significance for energy saving and emission reduction in the ironmaking process. Herein, predictive modeling of the permeability index after one hour (PI‐1h) is carried out by selecting appropriate machine learning models for the different clusters separately based on six machine learning methods and fuzzy‐C‐means. In addition, The SHapley Additive exPlanations (SHAP) method is used to gain insight into the relevance of each parameter to the PI‐1h. The simulation results show that within the allowable error range, the prediction accuracy of the support vector regression and Gaussian process regression models reaches 95.2% and 99.5%, respectively. Based on the data used in this article and the parameter importance analysis of SHAP, permeability index, pressure difference, wind velocity, and hourly coal injection rate are the main parameters to regulate PI‐1h.