Classification and Prediction of Blast Furnace Temperature Based on Representation Learning Technology

Tianxin Chen¹, Shihua Luo¹

^{1Jiangxi Science and Technology Normal University, China}

To further identify the key factors affecting the silicon content in molten iron, this paper employs mutual information maximization as the backbone network to classify and predict the silicon content in molten iron. By controlling the numerical value of the silicon content, the blast furnace temperature can be indirectly regulated. In the actual blast furnace ironmaking process, excessively high furnace temperatures can have serious impacts on both the furnace itself and the quality of molten iron. The objective of this study is to determine a reasonable critical point and classify the silicon content in molten iron into two categories: “high temperature” and “non-high temperature” based on this threshold. This classification enables the use of representation learning technology with mutual information maximization for classification and prediction. Through this research, the intrinsic patterns and control range of blast furnace temperature can be identified, ultimately achieving smarter control of blast furnace temperature. These findings hold significant practical value in the actual blast furnace ironmaking production process.

Keywords: Blast Furnace Temperature, Mutual Information Maximization, Representation Learning, Classification, Prediction