无头轧制非稳态阶段FET数据驱动预测及分析

doi:10.3901/JME.260183

摘要/Abstract

摘要： 精轧入口温度(Finishing entry temperature，FET)作为无头轧制产线的关键工艺参数，其控制精度直接影响终轧温度稳定性及产品尺寸性能指标。针对国内某钢厂无头轧制非稳态生产过程FET模型计算精度不足的问题，提出一种融合互信息特征选择与长短期记忆网络(Long short-term memory，LSTM)的数据驱动预测方法。首先基于互信息熵量化分析工艺变量与FET的非线性关联度，完成特征变量优选；继而构建具有深度序列特征提取能力的LSTM预测模型，并采用粒子群算法(Particle swarm optimization，PSO)对模型超参数进行优化。经数据试验验证，其预测性能指标平均绝对误差(Mean absolute error，MAE)为1.89 ℃、方均根误差(Root mean square error，RMSE)为3.07 ℃和相关系数(Correlation coefficient，R)为0.989 3显著优于文中其他模型；进一步采用Shapley加性解释模型(Shapley additive explanations，SHAP)可解释性分析方法分析了关键工艺参数对于FET的耦合作用机制。研究结果为提升无头轧制非稳态过程温度控制精度提供了新的方法体系，对实现薄规格高强热轧产品性能稳定性控制具有重要工程应用价值。

关键词: 精轧入口温度, 数据驱动建模, 互信息, 长短期记忆网络

Abstract: As a critical process parameter in endless rolling production line, the control accuracy of finishing entry temperature(FET) directly impacts the stability of final rolling temperature and product dimensional performance metrics. To address the insufficient computational precision of the FET control(FETC) model during the non-steady-state phase of endless rolling at a domestic steel plant, a data-driven prediction method integrating mutual information feature selection and long short-term memory(LSTM) is proposed. First, mutual information entropy is employed to quantitatively analyze the nonlinear correlation degree between process variables and FET, enabling optimal feature variable selection. Subsequently, an LSTM prediction model with deep sequence feature extraction capability is constructed, and particle swarm optimization is adopted to optimize the model’s hyperparameters. Experimental validation demonstrates that the proposed model achieves superior predictive performance, with evaluation metrics MAE (1.89 ℃), RMSE (3.07 ℃), and R (0.989 3) significantly outperforming other comparative models in the study. Furthermore, shapley additive explanations(SHAP) interpretability analysis is applied to elucidate the interaction mechanisms of key process parameters on FET. The precision of temperature control in the non-steady-state phase of endless rolling is enhanced by the proposed method, and substantial engineering value is offered for stabilizing the performance of thin-gauge high-strength hot-rolled products.

Key words: finishing entry temperature, data-driven modeling, mutual Information, LSTM

中图分类号:

TG335

董梓硕, 郭薇, 严乐明, 张敏, 于孟, 李继新. 无头轧制非稳态阶段FET数据驱动预测及分析[J]. 机械工程学报, 2025, 62(6): 163-173.

DONG Zishuo, GUO Wei, YAN Leming, ZHANG Min, YU Meng, LI Jixin. FET Prediction and Analysis with Data-Driven method for Non-steady Stage in Endless Rolling[J]. Journal of Mechanical Engineering, 2025, 62(6): 163-173.

导出引用管理器 EndNote|Reference Manager|ProCite|BibTeX|RefWorks

链接本文: http://www.cjmenet.com.cn/CN/10.3901/JME.260183

http://www.cjmenet.com.cn/CN/Y2025/V62/I6/163

参考文献

[1] 汪水泽，高军恒，吴桂林，等. 薄板坯连铸连轧技术发展现状及展望[J]. 工程科学学报，2022，44(4)：534-545. WANG Shuize，GAO Junheng，WU Guilin，et al. Thin slab casting and direct rolling technology：Current status and prospects[J]. Chinese Journal of Engineering，2022，44(4)：534-545.

[2] 康永林，朱国明，周建. 我国热轧宽带钢轧制技术的进步[J]. 轧钢，2024，41(5)：12-23. KANG Yonglin，ZHU Guoming，ZHOU Jian. Progress in rolling technology of hot rolled wide strip in China[J]. Steel Rolling，2024，41(5)：12-23.

[3] 张明生，徐永先. 基于多模式薄板坯连铸连轧产线700 MPa高强钢的开发与应用[J]. 金属材料与冶金工程，2023，51(3)：53-56. ZHANG Mingsheng，XU Yongxian. Development and application of 700 MPa high-strength steel based on multi-mode thin slab continuous casting and rolling production line[J]. Metal Materials and Metallurgy Engineering，2023，51(3)：53-56.

[4] 杜宁宇，李伟，范锦龙，等. 薄板坯连铸连轧ESP与MCCR技术分析及展望[J]. 轧钢，2023，40(5)：1-5. DU Ningyu，LI Wei，FAN Jinlong，et al. Analysis and prospect of ESP and MCCR technologies of thin slab casting and direct rolling[J]. Steel Rolling，2023，40(5)：1-5.

[5] 杨春政，袁天祥，刘延强，等. 新一代钢铁制造流程产品品牌化实践[J]. 中国冶金，2024，34(12)：24-35. YANG Chunzheng，YUAN Tianxiang，LIU Yanqiang，et al. Branding practice of new generation steel manufacturing process products[J]. China Metallurgy，2024，34(12)：24-35.

[6] 张殿华，彭文，丁敬国，等. 板带热连轧自动化系统的现状与展望[J]. 轧钢，2015，32(2)：6-12，26. ZHANG Dianhua，PENG Wen，DING Jingguo，et al. Automation system and its development for hot strip mill[J]. Steel Rolling，2015，32(2)：6-12，26.

[7] 李俊南，莫琳琳，李博. 基于数据驱动的热轧线精轧入口温度修正策略[J]. 冶金自动化，2024，48(1)：45-53，64. LI Junnan，MO Linlin，LI Bo. A data-driven entry temperature correction strategy for finishing rolling of hot rolling lines[J]. Metallurgical Industry Automation，2024，48(1)：45-53，64.

[8] 董占奎，王永和，王奎. 热轧带钢无头轧制技术的发展与支撑[J]. 河北冶金，2017(12)：43-46. DONG Zhankui，WANG Yonghe，WANG Kui. Development and support of endless rolling technology of hot-rolled strip steel[J]. Hebei Metallurgy，2017(12)：43-46.

[9] 张敏. 首钢京唐MCCR精轧入口温度控制系统介绍[J]. 中国金属通报，2018(3)：170-171. ZHANG Min. Introduction to the finishing mill entry temperature control system of Shougang Jingtang MCCR[J]. China Metal Bulletin，2018(3)：170-171.

[10] 王瑞丰. 无头轧制精轧入口温度补偿控制策略应用分析[J]. 冶金自动化，2021，45(1)：68-72，83. WANG Ruifeng. Application analysis of finisher entry temperature compensation control strategy in endless rolling[J]. Metallurgical Industry Automation，2021，45(1)：68-72，83.

[11] CUI Chunyuan，CAO Guangming，LI Xin，et al. The coupling machine learning for microstructural evolution and rolling force during hot strip rolling of steels[J]. Journal of Materials Processing Technology，2022，309：117736.

[12] DENG Lin，WANG Changhao，LUO Jinru，et al. Preparation and property optimization of FeCrAl-based ODS alloy by machine learning combined with wedge-shaped hot-rolling[J]. Materials Characterization，2022，188：111894.

[13] KATEB M，SAFARIAN S. Machine learning-driven predictive modeling of mechanical properties in diverse steels[J]. Machine Learning with Applications，2025，20：100634.

[14] DONG Zishuo，LI Xu，LUAN Feng，et al. Rolling theory-guided prediction of hot-rolled plate width based on parameter transfer strategy[J]. ISA Transactions，2024，146：352-365.

[15] 梁辉，童朝南，王恒. 基于数据驱动的热连轧厚度建模与控制方法[J]. 北京工业大学学报，2012，38(12)：1913-1920. LIANG Hui，TONG Chaonan，WANG Heng. Hot rolling thickness modeling and control method based on data driven[J]. Journal of Beijing University of Technology，2012，38(12)：1913-1920.

[16] LI Xu，HE Yaodong，DING Jingguo，et al. Predicting hot-strip finish rolling thickness using stochastic configuration networks[J]. Information Sciences，2022，611：677-689.

[17] LI Jingdong，SUN Youzhao，WANG Xiaochen，et al. Application of novel interpretable machine learning framework for strip flatness prediction during tandem cold rolling[J]. Measurement，2025，244：116516.

[18] LU Jia，WANG Pengfei，HUANG Huagui，et al. Novel online prediction model for thermal convexity of work rolls during hot steel rolling based on machine learning algorithms[J]. Expert Systems with Applications，2024，254：124384.

[19] MENG Lingming，DING Jingguo，LI Xiaojian，et al. Novel shape control system of hot-rolled strip based on machine learning fused mechanism model[J]. Expert Systems with Applications，2024，255：124789.

[20] ZHANG Lei，HE Guowei. Mutual-information-based dimensional learning：Objective algorithms for identification of relevant dimensionless quantities[J]. Computer Methods in Applied Mechanics and Engineering，2025，440：117922.

[21] CHEN Zehua，ZHANG Yongan，MA Minglong，et al. Study on hot processing map with ANN modification and DRX behaviors of a novel Mg-Ga-Sn anode for an air battery[J]. Materials Chemistry and Physics，2024，317：129146.

[22] TANG Wenyuan，HE Liang，HUI Xinyu，et al. A real-time optimization method for thermo-chemical coupled curing process of composites with LSTM network[J]. Journal of Manufacturing Processes，2025，138：90-99.

[23] ZHU Ting，CHEN Zhen，ZHOU Di，et al. Adaptive staged remaining useful life prediction of roller in a hot strip mill based on multi-scale LSTM with multi-head attention[J]. Reliability Engineering & System Safety，2024，248：110161.

[24] 赵延玉，赵晓永，王磊，等. 可解释人工智能研究综述[J]. 计算机工程与应用，2023，59(14)：1-14. ZHAO Yanyu，ZHAO Xiaoyong，WANG Lei，et al. Review of explainable artificial intelligence[J]. Computer Engineering and Applications，2023，59(14)：1-14.

[25] 纪守领，李进锋，杜天宇，等. 机器学习模型可解释性方法、应用与安全研究综述[J]. 计算机研究与发展，2019，56(10)：2071-2096. JI Shouling，LI Jinfeng，DU Tianyu，et al. Survey on techniques，applications and security of machine learning interpretability[J]. Journal of Computer Research and Development，2019，56(10)：2071-2096.

[26] TONG Piao，ZHANG Zhipeng，LIU Qiao，et al. Interpretable prediction model for decoupling hot rough rolling camber-process parameters[J]. Expert Systems with Applications，2024，256：124872.

[27] DING Chengyan，YE Juncheng，LEI Jiawei，et al. An interpretable framework for high-precision flatness prediction in strip cold rolling[J]. Journal of Materials Processing Technology，2024，329：118452.

[28] 任鹏帆，王振华，贾燚，等. 机理和数据融合的304不锈钢极薄带轧制力模型[J]. 钢铁，2024，59(10)：64-76. REN Pengfan，WANG Zhenhua，JIA Yi，et al. Rolling force model for 304 stainless steel ultra-thin strip based on mechanism and data fusion[J]. Iron and Steel，2024，59(10)：64-76.

[29] WANG Zhenhua，LIU Yunfei，NIU Lengyuan，et al. Intelligent prediction model of mechanical properties of ultrathin niobium strips based on XGBoost ensemble learning algorithm[J]. Computational Materials Science，2024，231：112579.