基于深度网络训练的铝热轧轧制力预报
|
摘要
在铝热轧过程中,轧制力预报精度直接影响着成品的产量和质量。为了提高铝热连轧轧制力预报精度,提出一种基于深度学习方法的多层感知器(Multi-layerPerceptron,MLP)轧制力预报模型。模型利用MLP的函数逼近能力来回归轧制力。模型以小批量训练为基础,利用Batch Normalization方法稳定网络前向传播的输出分布,并使用Adam随机优化算法来完善梯度更新,以解决MLP模型难以训练的问题。仿真结果表明:模型使网络预测与实测数据的相对误差降低到3%以内,实现了轧制力的高精度预测。
In the aluminum hot rolling, the prediction accuracy of the rolling force directly affects the output and quality of the finished product. In view of the inherent defects of traditional rolling force model, a MLP rolling force prediction model based on deep learning method was proposed. The model uses MLP's function approximation ability to regress the rolling force. Based on the Mini-batch training, the model uses Batch Normalization method to stabilize the output distribution of the network forward propagation, and uses the Adam stochastic optimization algorithm to improve the gradient updating so as to solve the difficult training problem of the MLP model. The simulation results show that the model can reduce the relative error between the network prediction and the measured data to less than 3%. Compared with the traditional mathematical model, this method realizes the high precision prediction of the rolling force, and realizes a high-precision prediction of rolling force.
In the aluminum hot rolling, the prediction accuracy of the rolling force directly affects the output and quality of the finished product. In view of the inherent defects of traditional rolling force model, a MLP rolling force prediction model based on deep learning method was proposed. The model uses MLP's function approximation ability to regress the rolling force. Based on the Mini-batch training, the model uses Batch Normalization method to stabilize the output distribution of the network forward propagation, and uses the Adam stochastic optimization algorithm to improve the gradient updating so as to solve the difficult training problem of the MLP model. The simulation results show that the model can reduce the relative error between the network prediction and the measured data to less than 3%. Compared with the traditional mathematical model, this method realizes the high precision prediction of the rolling force, and realizes a high-precision prediction of rolling force.
引文
[1]LIU X,LIU X H,SONG M,SUN X K,LIU L Z.Theoretical analysis of minimum metal foil thickness achievable by asymmetric rolling with fixed identical roll diameters[J].Transactions of Nonferrous Metals Society of China,2016,26(2):501-507.
[2]ZUO Y B,FU X,CUI J Z,TANG X Y,MAO L,LI L,ZHUQing-feng.Shear deformation and plate shape control of hot-rolled aluminium alloy thick plate prepared by asymmetric rolling process[J].Transactions of Nonferrous Metals Society of China,2014,24(7):2220-2225.
[3]燕猛,黄华贵,张彩云,杜凤山,张尚斌.板坯头/尾部平面形状对铝合金厚板粗轧头尾切除量的影响[J].中国有色金属学报,2017,27(6):1102-1108.YAN Meng,HUANG Hua-gui,ZHANG Cai-yun,DU Feng-shan,ZHANG Shang-bin.Effect of the shape of the head/tail of the slab on the head and tail cutting of aluminum alloy thick plate[J].Chinese Journal of Nonferrous Metals,2017,27(6):1102-1108.
[4]OROWAN E.The calculation of roll pressure in hot and cold flat rolling[J].Proceedings of the Institution of Mechanical Engineers,1943,150(1):140-167.
[5]MAHMOODKHANI Y,WELLS M A,SONG G.Prediction of roll force in skin pass rolling using numerical and artificial neural network methods[J].Ironmaking&Steelmaking,2016,44(4):281-286.
[6]BYON S M,NA D H,LEE Y S.Flow stress equation in range of intermediate strain rates and high temperatures to predict roll force in four-pass continuous rod rolling[J].Transactions of Nonferrous Metals Society of China,2013,23(3):742-748.
[7]CHEN Z M,LUO Z L.Rolling force prediction based on multiple support vector machines[C]//WATADA J.Proceedings of the 32nd Chinese Control Conference.New York:IEEE Press,2013:3306-3309.
[8]杨景明,顾佳琪,闫晓莹,车海军.基于改进遗传算法优化BP网络的轧制力预测研究[J].矿冶工程,2015,35(1):111-115.YANG Jing-ming,GU Jia-qi,YAN Xiao-ying,CHE Hai-jun.Prediction of rolling force based on improved genetic algorithm for optimizing BP neural network[J].Mining and Metallurgical Engineering,2015,35(1):111-115.
[9]赵文姣,闫洪伟,杨枕,温玉莲,孙祖乾.基于CA-CAMC网络的轧制力自学习预报模型[J].冶金自动化,2016,40(2):7-10.ZHAO Wen-biao,YAN Hon-wei,YANG Zhen,WEN Yu-lian,SUN Zu-qian.Rolling force self-learning prediction model based on CA-CAMC network[J].Metallurgical Industry Automation,2016,40(2):7-10.
[10]吕程,王国栋,刘相华,姜正义,朱洪涛,袁建光,解旗.基于神经网络的热连轧精轧机组轧制力高精度预报[J].钢铁,1998,33(3):33-35.LüCheng,WANG Guo-dong,LIU Xiang-hua,JIANG Zheng-yi,ZHU Hong-tao,YUAN Jian-guang,XIE Qi.High-precision prediction of rolling force in hot rolling finishing mill based on neural network[J].Iron and Steel,1998,33(3):33-35.
[11]HINTON G E,OSINDERO S,TEH Y W.A fast learning algorithm for deep belief nets[J].Neural Computation,2006,18(7):1527-1554
[12]毛勇华,桂小林,李前,贺兴时.深度学习应用技术研究[J].计算机应用研究,2016,33(11):3201-3205.MAO Yong-hua,GUI Xiao-lin,LI Qian,HE Xing-shi.Research on deep learning application technology[J].Journal of Computer Applications,2016,33(11):3201-3205.
[13]IOFFE S,SZEGEDY C.Batch normalization:Accelerating deep network training by reducing internal covariate shift[C]//BACHF.Proceedings of the 32nd International Conference on Machine Learning.Massachusetts:MIT Press,2015:448-456.
[14]DUCHI J,HAZAN E,SINGER Y.Adaptive subgradient methods for online learning and stochastic optimization[J].Journal of Machine Learning Research,2011,12(7):257-269.
[15]KINGMA D P,BA J L.Adam:A method for stochastic optimization[C]//KINGSBURY B.Proceedings of the the 3rd International conference on learning representations.San Diego:arXiv,2015:1412-6980.
[16]GLOROT X,BENGIO Y.Understanding the difficulty of training deep feedforward neural networks[J].Journal of Machine Learning Research,2010,9:249-256.
[17]GLOROT X,BORDES A,BENGIO Y.Deep sparse rectifier neural networks[C]//BACH F.Proceedings of the 14th International Conference on Artificial Intelligence and Statistics.Massachusetts:MIT Press,2011:315-323.
[18]杨景明,孙晓娜,车海军,刘畅.基于蚁群算法的神经网络冷连轧机轧制力预报[J].钢铁,2009,44(3):52-55.YANG Jing-ming,SUN Xia-ona,CHE Hai-jun,LIU Chang.Prediction of rolling force in neural network cold rolling mill based on ant colony algorithm[J].Iron and Steel,2009,44(3):52-55.
[2]ZUO Y B,FU X,CUI J Z,TANG X Y,MAO L,LI L,ZHUQing-feng.Shear deformation and plate shape control of hot-rolled aluminium alloy thick plate prepared by asymmetric rolling process[J].Transactions of Nonferrous Metals Society of China,2014,24(7):2220-2225.
[3]燕猛,黄华贵,张彩云,杜凤山,张尚斌.板坯头/尾部平面形状对铝合金厚板粗轧头尾切除量的影响[J].中国有色金属学报,2017,27(6):1102-1108.YAN Meng,HUANG Hua-gui,ZHANG Cai-yun,DU Feng-shan,ZHANG Shang-bin.Effect of the shape of the head/tail of the slab on the head and tail cutting of aluminum alloy thick plate[J].Chinese Journal of Nonferrous Metals,2017,27(6):1102-1108.
[4]OROWAN E.The calculation of roll pressure in hot and cold flat rolling[J].Proceedings of the Institution of Mechanical Engineers,1943,150(1):140-167.
[5]MAHMOODKHANI Y,WELLS M A,SONG G.Prediction of roll force in skin pass rolling using numerical and artificial neural network methods[J].Ironmaking&Steelmaking,2016,44(4):281-286.
[6]BYON S M,NA D H,LEE Y S.Flow stress equation in range of intermediate strain rates and high temperatures to predict roll force in four-pass continuous rod rolling[J].Transactions of Nonferrous Metals Society of China,2013,23(3):742-748.
[7]CHEN Z M,LUO Z L.Rolling force prediction based on multiple support vector machines[C]//WATADA J.Proceedings of the 32nd Chinese Control Conference.New York:IEEE Press,2013:3306-3309.
[8]杨景明,顾佳琪,闫晓莹,车海军.基于改进遗传算法优化BP网络的轧制力预测研究[J].矿冶工程,2015,35(1):111-115.YANG Jing-ming,GU Jia-qi,YAN Xiao-ying,CHE Hai-jun.Prediction of rolling force based on improved genetic algorithm for optimizing BP neural network[J].Mining and Metallurgical Engineering,2015,35(1):111-115.
[9]赵文姣,闫洪伟,杨枕,温玉莲,孙祖乾.基于CA-CAMC网络的轧制力自学习预报模型[J].冶金自动化,2016,40(2):7-10.ZHAO Wen-biao,YAN Hon-wei,YANG Zhen,WEN Yu-lian,SUN Zu-qian.Rolling force self-learning prediction model based on CA-CAMC network[J].Metallurgical Industry Automation,2016,40(2):7-10.
[10]吕程,王国栋,刘相华,姜正义,朱洪涛,袁建光,解旗.基于神经网络的热连轧精轧机组轧制力高精度预报[J].钢铁,1998,33(3):33-35.LüCheng,WANG Guo-dong,LIU Xiang-hua,JIANG Zheng-yi,ZHU Hong-tao,YUAN Jian-guang,XIE Qi.High-precision prediction of rolling force in hot rolling finishing mill based on neural network[J].Iron and Steel,1998,33(3):33-35.
[11]HINTON G E,OSINDERO S,TEH Y W.A fast learning algorithm for deep belief nets[J].Neural Computation,2006,18(7):1527-1554
[12]毛勇华,桂小林,李前,贺兴时.深度学习应用技术研究[J].计算机应用研究,2016,33(11):3201-3205.MAO Yong-hua,GUI Xiao-lin,LI Qian,HE Xing-shi.Research on deep learning application technology[J].Journal of Computer Applications,2016,33(11):3201-3205.
[13]IOFFE S,SZEGEDY C.Batch normalization:Accelerating deep network training by reducing internal covariate shift[C]//BACHF.Proceedings of the 32nd International Conference on Machine Learning.Massachusetts:MIT Press,2015:448-456.
[14]DUCHI J,HAZAN E,SINGER Y.Adaptive subgradient methods for online learning and stochastic optimization[J].Journal of Machine Learning Research,2011,12(7):257-269.
[15]KINGMA D P,BA J L.Adam:A method for stochastic optimization[C]//KINGSBURY B.Proceedings of the the 3rd International conference on learning representations.San Diego:arXiv,2015:1412-6980.
[16]GLOROT X,BENGIO Y.Understanding the difficulty of training deep feedforward neural networks[J].Journal of Machine Learning Research,2010,9:249-256.
[17]GLOROT X,BORDES A,BENGIO Y.Deep sparse rectifier neural networks[C]//BACH F.Proceedings of the 14th International Conference on Artificial Intelligence and Statistics.Massachusetts:MIT Press,2011:315-323.
[18]杨景明,孙晓娜,车海军,刘畅.基于蚁群算法的神经网络冷连轧机轧制力预报[J].钢铁,2009,44(3):52-55.YANG Jing-ming,SUN Xia-ona,CHE Hai-jun,LIU Chang.Prediction of rolling force in neural network cold rolling mill based on ant colony algorithm[J].Iron and Steel,2009,44(3):52-55.