Scholarlink

<p>This paper addresses the issue of inaccurate rolling force predictions in the finishing mill of the 1549 hot strip production line at a specific plant by proposing a fusion model that integrates data-driven and mechanistic approaches. This innovative model significantly improves the prediction accuracy of rolling force. First, the mechanistic model was analyzed, and its coefficients were optimized using a chaotic particle swarm optimization (CPSO), then its structure was improved by introducing a new adaptive coefficient to correct the model’s predicted values, thereby enhanced accuracy. Subsequently, an ELM-based data-driven model was developed using actual production data, effectively uncovering the intrinsic relationships within the dataset. Finally, the improved mechanistic model and the data-driven model were fused to create the final hybrid model. The performance of this fusion model demonstrated substantial improvements, with RMSE, MAPE, and PSET values improving by 517.876 kN, 2.73 %, and 37.85 %, respectively.</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

Data-mechanism fusion prediction of rolling force in hot rolling mill

  • 期刊论文

  • 发表日期：

    2026年1月13日

查看全文
  • Xiaomin Zhou,
  • Tianliang Xiong,
  • Huadong Qiu,
  • Xiaoke Hu,
  • Yuanlun Ma,
  • Bin Wang

摘要

This paper addresses the issue of inaccurate rolling force predictions in the finishing mill of the 1549 hot strip production line at a specific plant by proposing a fusion model that integrates data-driven and mechanistic approaches. This innovative model significantly improves the prediction accuracy of rolling force. First, the mechanistic model was analyzed, and its coefficients were optimized using a chaotic particle swarm optimization (CPSO), then its structure was improved by introducing a new adaptive coefficient to correct the model’s predicted values, thereby enhanced accuracy. Subsequently, an ELM-based data-driven model was developed using actual production data, effectively uncovering the intrinsic relationships within the dataset. Finally, the improved mechanistic model and the data-driven model were fused to create the final hybrid model. The performance of this fusion model demonstrated substantial improvements, with RMSE, MAPE, and PSET values improving by 517.876 kN, 2.73 %, and 37.85 %, respectively.