Scholarlink

<p>Driven by the needs of low-carbon transformation and cost reduction with efficiency improvement in the iron and steel industry, converter smelting with a high-scrap ratio faces the dual challenges of large deviations in energy consumption calculation and unstable control of flux addition. To address the operating condition issue of the 120-t converter, this paper clarifies the intrinsic correlation among molten pool heat demand, slag material consumption, and heat compensation measures under high-scrap-ratio (25–45&#xa0;pct) conditions by&#xa0;analyzing&#xa0;the “heat–material–oxygen” coupling relationship in the converter’s smelting heat and material balances. Based on key on-site production data from 1200 heats, after data preprocessing, a collaborative system for&#xa0;predicting the optimal&#xa0;flux addition amount, combining a mechanism-based model with an intelligent algorithm, was constructed. Pearson correlation analysis combined with the recursive feature elimination (RFE) algorithm was used to screen 8 key characteristic parameters, including scrap weight, hot metal Si content, and charging weight. Comparative analysis of BP, PSO-BP, and GA-BP models reveals that the GA-BP model performs best, achieving a Goodness of Fit (<i>R</i><sup>2</sup>) of 88.3&#xa0;pct, a Mean Absolute Percentage Error (MAPE) of 0.12&#xa0;pct, and an RMSE of 206.4&#xa0;kg. Industrial implementation in a 120-t converter demonstrates that model-guided prediction increased the comprehensive end-point hit rate from 82.4 to 91.5&#xa0;pct (a 9.1-pct absolute improvement). Statistical validation across 1000 heats confirms that the model effectively stabilizes flux additions by eliminating the “positive bias” (the tendency of over-adding lime/dolomite) found in manual operations, thereby reducing manual dependence, and lowering production costs, providing robust technical support for the green and efficient production of high-manganese steel for LNG tanks under high-scrap-ratio conditions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Development and Application of a Hybrid Mechanistic-AI Model for Flux Optimization in High-Scrap-Ratio Converter Steelmaking of High-Manganese Steel for LNG Tanks

  • 期刊论文

  • 发表日期：

    2026年5月19日

查看全文
  • Zhiqi Liu,
  • Ling Yan,
  • Xiao Han,
  • Xiangyu Qi,
  • Shiyu Shi,
  • Boyong Li

摘要

Driven by the needs of low-carbon transformation and cost reduction with efficiency improvement in the iron and steel industry, converter smelting with a high-scrap ratio faces the dual challenges of large deviations in energy consumption calculation and unstable control of flux addition. To address the operating condition issue of the 120-t converter, this paper clarifies the intrinsic correlation among molten pool heat demand, slag material consumption, and heat compensation measures under high-scrap-ratio (25–45 pct) conditions by analyzing the “heat–material–oxygen” coupling relationship in the converter’s smelting heat and material balances. Based on key on-site production data from 1200 heats, after data preprocessing, a collaborative system for predicting the optimal flux addition amount, combining a mechanism-based model with an intelligent algorithm, was constructed. Pearson correlation analysis combined with the recursive feature elimination (RFE) algorithm was used to screen 8 key characteristic parameters, including scrap weight, hot metal Si content, and charging weight. Comparative analysis of BP, PSO-BP, and GA-BP models reveals that the GA-BP model performs best, achieving a Goodness of Fit (R2) of 88.3 pct, a Mean Absolute Percentage Error (MAPE) of 0.12 pct, and an RMSE of 206.4 kg. Industrial implementation in a 120-t converter demonstrates that model-guided prediction increased the comprehensive end-point hit rate from 82.4 to 91.5 pct (a 9.1-pct absolute improvement). Statistical validation across 1000 heats confirms that the model effectively stabilizes flux additions by eliminating the “positive bias” (the tendency of over-adding lime/dolomite) found in manual operations, thereby reducing manual dependence, and lowering production costs, providing robust technical support for the green and efficient production of high-manganese steel for LNG tanks under high-scrap-ratio conditions.