目的 为解决不均匀数据分布对轧制力模型精度的影响,提出理论模型与智能算法相结合的建模新方案。方法 首先构建了基于自助法的BP神经网络模型,进而采用该模型对已有轧制力理论模型进行修正,从而构建了整合模型。具体来说,首先对轧制生产数据进行预处理和相关性分析,将理论模型计算值作为BP神经网络的输入变量之一进行建模,以此提高了模型的预测精度。其次,对数据聚类后进行分类训练,针对因样本分布不平衡而引起的模型精度不足的问题,采用自助法进行数据扩充,并结合生产数据进行预测对比。结果 采用乘法补偿法对理论模型进行修正,构建了具有较高精度的轧制力整合模型。同时结合自助法实现了轧制数据集的均衡化,提高了模型的精度和稳定性。该模型误差波动较小且最大平均绝对百分比误差（MAPE）仅为4.61%。结论 采用的自助法为解决不均匀数据集问题下的轧制力高精度建模提供了一种新的解决方案,而所得的整合模型可为实际生产过程的精确控制提供了科学依据。
Abstract
To address the adverse impacts of non-uniform data distribution on the predictive accuracy of rolling force models, the work aims to propose a novel modeling paradigm integrating theoretical frameworks with intelligent algorithmic methodologies. A bootstrap-based BP neural network model was initially constructed and subsequently employed to modify existing theoretical rolling force models, thereby forming an integrated model. Specifically, preprocessing and correlation analysis were conducted on rolling production data, where calculated values from theoretical models were incorporated as input variables in the BP neural network architecture to improve predictive performance. The data were clustered and subject to classified training to address accuracy limitations caused by imbalanced sample distributions. Bootstrap-based data augmentation was implemented for predictive comparison with production data. A high-precision integrated rolling force prediction model was developed through multiplicative compensation-based correction of theoretical models. Dataset balancing was achieved via bootstrap integration, resulting in enhanced model accuracy and stability with minimal error fluctuations, where the maximum mean absolute percentage error (MAPE) was reduced to 4.61%. The bootstrap method is demonstrated to provide a novel solution for high-precision rolling force modeling under imbalanced datasets. The established integrated model serves as a scientific foundation for achieving precise control in practical manufacturing processes.
关键词
轧制力 /
自助法 /
整合模型 /
BP神经网络 /
聚类
Key words
rolling force /
bootstrap method /
integrated model /
BP neural network /
clustering
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 赵文姣, 闫洪伟, 杨枕, 等. 提高轧制力设定精度的一种离线优化方法[J]. 中国冶金, 2017, 27(8): 19-22.
ZHAO W J, YAN H W, YANG Z, et al.An Off-Line Optimization Method for Improving Precision of Rolling Force Setting[J]. China Metallurgy, 2017, 27(8): 19-22.
[2] 李维刚, 邓肯, 赵云涛, 等. 基于连续曲面的轧制模型自学习方法[J]. 钢铁, 2017, 52(12): 61-66.
LI W G, DENG K, ZHAO Y T, et al.Self-Learning Method for Rolling Model Based on Continuous Surface[J]. Iron and Steel, 2017, 52(12): 61-66.
[3] SIMS R B.The Calculation of Roll Force and Torque in Hot Rolling Mills[J]. Proceedings of the Institution of Mechanical Engineers, 1954, 168(1): 191-200.
[4] WAN L W, ZHANG S H, YIN Z Q, et al.Adaptive Modeling of Rolling Force for Hot Rolled Plate Based on Industrial Data[J]. Journal of Manufacturing Processes, 2024, 129: 253-260.
[5] LI X D.Rolling Force Prediction of Hot Strip Based on Combined Friction[J]. IOP Conference Series: Materials Science and Engineering, 2017, 269(1): 012053.
[6] CHE L Z, ZHANG S H, TIAN W H, et al.A New Model for Thermal-Mechanical Coupled of Gradient Temperature Rolling Force Based on Geometrical Unified Yield Criterion[J]. Journal of Manufacturing Processes, 2023, 101: 904-915.
[7] JACOBS L J M, ATZEMA E H, MOERMAN J, et al. Quantification of the Influence of Anisotropic Plastic Yielding on Cold Rolling Force[J]. Journal of Materials Processing Technology, 2023, 319: 118055.
[8] DENG J F, SUN J, PENG W, et al.Application of Neural Networks for Predicting Hot-Rolled Strip Crown[J]. Applied Soft Computing, 2019, 78: 119-131.
[9] ZHANG F, ZHAO Y T, SHAO J.Rolling Force Prediction in Heavy Plate Rolling Based on Uniform Differential Neural Network[J]. Journal of Control Science and Engineering, 2016, 2016(1): 6473137.
[10] ZHENG G, GE L H, SHI Y Q, et al.Dynamic Rolling Force Prediction of Reversible Cold Rolling Mill Based on BP Neural Network with Improved PSO[C]// 2018 Chinese Automation Congress (CAC). Xi'an, China. IEEE, 2018: 2710-2714.
[11] SONG C N, CAO J G, XIAO J, et al.Control Strategy of Multi-Stand Work Roll Bending and Shifting on the Crown for UVC Hot Rolling Mill Based on MOGPR Approach[J]. Journal of Manufacturing Processes, 2023, 85: 832-843.
[12] ZHANG S X, LIU Z T, AN T, et al.Hot Rolled Prognostic Approach Based on Hybrid Bayesian Progressive Layered Extraction Multi-Task Learning[J]. Expert Systems with Applications, 2024, 249: 123763.
[13] 王孟超, 武川, 孟亚飞, 等. 基于机器学习模具钢大型锻件力学性能预测[J]. 精密成形工程, 2024, 16(12): 198-208.
WANG M C, WU C, MENG Y F, et al.Prediction of Mechanical Properties of Large Forgings in Mold Steel Based on Machine Learning[J]. Journal of Netshape Forming Engineering, 2024, 16(12): 198-208.
[14] BAGHERIPOOR M, BISADI H.Application of Artificial Neural Networks for the Prediction of Roll Force and Roll Torque in Hot Strip Rolling Process[J]. Applied Mathematical Modelling, 2013, 37(7): 4593-4607.
[15] ZHANG S H, CHE L Z, LIU X Y.Modelling of Deformation Resistance with Big Data and Its Application in the Prediction of Rolling Force of Thick Plate[J]. Mathematical Problems in Engineering, 2021, 2021(1): 2500636.
[16] HUANG C G, HUANG H Z, LI Y F, et al.A Novel Deep Convolutional Neural Network-Bootstrap Integrated Method for RUL Prediction of Rolling Bearing[J]. Journal of Manufacturing Systems, 2021, 61: 757-772.
[17] 王秀梅, 吕程, 王国栋, 等. 轧制力预报中的神经网络和数学模型[J]. 东北大学学报, 1999, 20(3): 319-321.
WANG X M, LYU C, WANG G D, et al.Neural Networks and Mathematical Models in the Prediction of Rolling Load of the Finisher[J]. Journal of Northeastern University (Natural Science), 1999, 20(3): 319-321.
[18] RATH S, SENGUPTA P P, SINGH A P, et al.mathematical-Artificial Neural Network Hybrid Model to Predict Roll Force during Hot Rolling of Steel[J]. International Journal of Computational Materials Science and Engineering, 2013, 2(1): 1350004.
[19] CAO L, LI X, LI X H, et al.Variable Speed Rolling Force Prediction with Theoretical and Data-Driven Models[J]. International Journal of Mechanical Sciences, 2024, 264: 108833.
[20] ZHANG S H, DENG L, CHE L Z.An Integrated Model of Rolling Force for Extra-Thick Plate by Combining Theoretical Model and Neural Network Model[J]. Journal of Manufacturing Processes, 2022, 75: 100-109.
[21] 邓继飞. 基于数据驱动的热连轧板凸度预测与故障分类[D]. 沈阳: 东北大学, 2020: 40-41.
DENG J F.Data-Driven Strip Crown Prediction and Fault Classification for Hot-rolled Strips[D]. Shenyang: Northeastern University, 2020: 40-41.
[22] 胡贤磊, 王昭东, 于解民, 等. 结合模型自学习的BP神经元网络的轧制力预报[J]. 东北大学学报, 2002, 23(11): 1089-1092.
HU X L, WANG Z D, YU J M, et al.Prediction of Rolling Load by BP Neural Networks Integrating with Self-adaption of Traditional Model[J]. Journal of Northeastern University, 2002, 23(11): 1089-1092.
[23] ZHANG S H, TIAN W H, DENG L.A Novel Yield Criterion and Its Application to Calculate the Rolling Force of a Thick Plate during Hot Rolling[J]. Journal of the Brazilian Society of Mechanical Sciences and Engineering, 2021, 43(1): 12.
[24] HARRIS J N.Mechanical Working of Metals: Theory and Practice[M]. Oxford: Pergamon Press, 1983: 107-159.
[25] FISHER R A.Breakthroughs in Statistics: Statistical Methods for Research Workers[M]. New York: Springer Nature, 1992, 66-67.
[26] 张立明. 人工神经网络的模型及其应用[M]. 上海: 复旦大学出版社, 1993: 32-37.
ZHANG L M.Models and Applications of Artificial Neural Networks[M]. Shanghai: Fudan Press, 1993: 32-37.
基金
国家自然科学基金（U1960105,52074187,52274388）
PDF(7007 KB)
渝公网安备 50010702501719号 渝ICP备15012534号-4
