Si content in hot metal is an important index for characterizing furnace temperature and economic benefit. It is of great value to establish a timely and accurate prediction model of Si content for ensuring the life and stable operation of blast furnace. In the process of blast furnace smelting, the factors affecting Si content are numerous and complex, with nonlinear, time-varying, and time series characteristics. To accurately predict the Si content in hot metal, this study proposes to use the Archimedes optimization algorithm (AOA) to optimize variational mode decomposition (VMD), which can help the model find the optimal combination of penalty factor, <InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="11837_2025_7625_Article_IEq1.gif" Format="GIF" Height="12" Rendition="HTML" Resolution="72" Type="Linedraw" Width="18" /> </InlineMediaObject> <EquationSource Format="TEX">$\alpha ,$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>α</mi> <mo>,</mo> </mrow> </math></EquationSource> </InlineEquation> and mode number, <InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="11837_2025_7625_Article_IEq2.gif" Format="GIF" Height="17" Rendition="HTML" Resolution="72" Type="Linedraw" Width="16" /> </InlineMediaObject> <EquationSource Format="TEX">$k,$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>k</mi> <mo>,</mo> </mrow> </math></EquationSource> </InlineEquation> in the decomposition parameters. For each decomposed subsequence, a bidirectional long short-term memory (BILSTM) network with bidirectional processing and long-term memory is established to thoroughly mine the temporal characteristics of Si content data. The model demonstrates strong predictive capability, with the high coefficient of determination (R2, 99.82%), low root means square error (RMSE, 0.0045), mean absolute error (MAE, 0.0036), and mean square error (MSE, 0.0020%) indicating good performance. This indicates that AOA-VMD-BILSTM has great potential in predicting Si content, which provides important theoretical support and practical guidance for blast furnace smelting production.

Prediction of Si Content in Hot Metal Based on BILSTM

期刊论文
发表日期：

2025年8月19日

查看全文

Jiahao Xi,
Xiangdong Xing,
Huhang Xie,
Baowen Zang,
Qi Xiao,
Penghui Guo,
Donghui Wei

摘要

Si content in hot metal is an important index for characterizing furnace temperature and economic benefit. It is of great value to establish a timely and accurate prediction model of Si content for ensuring the life and stable operation of blast furnace. In the process of blast furnace smelting, the factors affecting Si content are numerous and complex, with nonlinear, time-varying, and time series characteristics. To accurately predict the Si content in hot metal, this study proposes to use the Archimedes optimization algorithm (AOA) to optimize variational mode decomposition (VMD), which can help the model find the optimal combination of penalty factor, $\alpha ,$ $α,$ and mode number, $k,$ $k,$ in the decomposition parameters. For each decomposed subsequence, a bidirectional long short-term memory (BILSTM) network with bidirectional processing and long-term memory is established to thoroughly mine the temporal characteristics of Si content data. The model demonstrates strong predictive capability, with the high coefficient of determination (R², 99.82%), low root means square error (RMSE, 0.0045), mean absolute error (MAE, 0.0036), and mean square error (MSE, 0.0020%) indicating good performance. This indicates that AOA-VMD-BILSTM has great potential in predicting Si content, which provides important theoretical support and practical guidance for blast furnace smelting production.