Scholarlink

<p>Optimizing the coal blending ratio (CBR) is an effective approach for reducing energy consumption and carbon emissions in copper smelting; however, its stability is strongly influenced by the complexity and variability of furnace feed materials. In this study, an interpretable machine learning (ML) framework was developed to predict and optimize CBR in a copper bottom-blown smelting process. An industrial CBR database was constructed using feed composition and key operational parameters. Multiple ML models, including support vector machines, random forests, CatBoost, LightGBM, XGBoost, and ExtraTrees, were systematically evaluated. The ExtraTrees model exhibited the best predictive performance, achieving <i>R</i><sup>2</sup> values of 0.9999 and 0.9253 for the training and test datasets, respectively, with most prediction errors within ±&#xa0;0.1 pct. SHapley Additive exPlanation (SHAP) was applied to enhance model interpretability and quantify feature contributions. The results indicate that smelting oxygen volume is the most influential variable affecting CBR, followed by the slag concentrates ratio and the contents of S, Cu, and Zn. Higher oxygen volume, blast supply, and sulfide content are generally associated with lower coal consumption. The proposed framework provides practical insights for CBR optimization and shows potential for integration into intelligent process control in industrial copper smelting.</p> Graphical abstract <p></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

Prediction of Coal Blending Ratio in Copper Bottom-Blowing Smelting using ExtraTrees Based on SHAP Interpretation

  • 期刊论文

  • 发表日期：

    2026年6月16日

查看全文
  • Haoli Yan,
  • Bo Li,
  • Jinyue Liu,
  • Yonggang Wei,
  • Hua Wang

摘要

Optimizing the coal blending ratio (CBR) is an effective approach for reducing energy consumption and carbon emissions in copper smelting; however, its stability is strongly influenced by the complexity and variability of furnace feed materials. In this study, an interpretable machine learning (ML) framework was developed to predict and optimize CBR in a copper bottom-blown smelting process. An industrial CBR database was constructed using feed composition and key operational parameters. Multiple ML models, including support vector machines, random forests, CatBoost, LightGBM, XGBoost, and ExtraTrees, were systematically evaluated. The ExtraTrees model exhibited the best predictive performance, achieving R2 values of 0.9999 and 0.9253 for the training and test datasets, respectively, with most prediction errors within ± 0.1 pct. SHapley Additive exPlanation (SHAP) was applied to enhance model interpretability and quantify feature contributions. The results indicate that smelting oxygen volume is the most influential variable affecting CBR, followed by the slag concentrates ratio and the contents of S, Cu, and Zn. Higher oxygen volume, blast supply, and sulfide content are generally associated with lower coal consumption. The proposed framework provides practical insights for CBR optimization and shows potential for integration into intelligent process control in industrial copper smelting.

Graphical abstract