Open AccessPredictive control of Si content in blast furnace smelting based on improved SA-BP
Author(s) -
Wenhan Liu,
Hongli Wang,
Liye Shi
Publication year - 2021
Publication title -
journal of mines, metals and fuels
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.125
H-Index - 11
ISSN - 0022-2755
DOI - 10.18311/jmmf/2021/28076
Subject(s) - blast furnace , smelting , metallurgy , ground granulated blast furnace slag , materials science , process engineering , engineering , cement
Blast furnace smelting process is a highly complex nonlinear dynamic process, its purpose is to refining the quality of liquid iron. From the point of view of chemical reaction kinetics, the main chemical reactions in the blast furnace are as many as 108 kinds, and the high complexity is obvious. From the point of view of hydrodynamics, there are three-phase mixed compressible viscous fluids in the blast furnace smelting process. The hydrodynamic equation is nonlinear, high-dimensional and high-coupling. In addition, the blast furnace smelting process has the characteristics of time-varying, high-dimensional, distributed parameters and other characteristics of the complex conditions and the failure of the operation under the conditions of the test, making the blast furnace smelting process automation and furnace temperature precision control to become metallurgical workers face the problem. In this study, from the data point of view, to explore the furnace temperature can be characterized by [Si] content prediction. Based on the data of 1000 furnace blast furnace, an accurate and reasonable prediction model of Si content in blast furnace is established. First of all, through the calculation of correlation coefficients of various factors and the time trend diagram, the general lag time between each parameter and Si content in the process of blast furnace production is obtained. Through correlation and analysis of the correlation between lag time, before and after the two largest furnace, with improved simulated annealing algorithm to determine the optimal initial weights, the content of Si, the N furnace of S content, the quantity of coal and air PML FL as input, the contents of Si in n+1 furnace as output to establish dynamic prediction model of improved SA-BP neural network based. Finally, the data used to detect the model is brought into the model to be tested, the error is analyzed, and the error local map is used to express the error visually. The model is tested.