锻压技术

基于比例损失去噪自编码器的冷连轧轧制力预测分析

英文标题：Rolling force prediction analysis of tandem cold rolling based on proportional loss denoising autoencoder

作者：张海霞1 李灿2

分类号：TG335

出版年,卷(期):页码：2022,47(4):190-194

摘要：

为了更加精确地预测冷连轧轧制力，设计了一种通过分层提取和目标相关特征来实现的比例损失堆叠去噪自编码器。首先，通过堆叠去噪自编码器(SDAE)构建深度网络并在SDAE顶层中加入输出层；然后，通过部分有标签样本实现网络权重变量的调节；最后，按照设定目标参数调节深度网络变量，从而降低网络预测值和目标值的偏差。本方法通过在训练过程加入目标值信息实现了特征提取有效性的显著提升，具有很好的预测稳定性。通过试验测试本算法，其预测结果在±3%误差内，可以满足实际生产控制要求。本算法能够从输入层内找到和目标值关联的特征，在预训练阶段完成目标值的整合。相比其他预测算法，本算法获得了很小的预测误差，能够更快完成收敛，表现出了更优的预测精度和效率。

In order to predict the rolling force of tandem cold rolling more accurately, a proportional loss stack denoising autoencoder (SDAE) was designed by the hierarchical extraction and target correlation features. Firstly, the output layer was added to the top layer of SDAE by the stack denoising autoencoder. And then, the weight variables of network were adjusted through some labeled samples. Finally, the depth network variables were adjusted according to the set target parameters, so as to reduce the deviation between network predicted value and target value. Furthermore, by the method of adding target value information in the training process, the effectiveness of feature extraction was improved significantly and had a good predictive stability. The experimental results show that the prediction result of this algorithm is within ±3% error through test, which can meet the actual production control requirements. The algorithm can find the features associated with the target values from the input layer and complete the integration of target values in the pre-training stage. Compared with other prediction algorithms, this algorithm has a small prediction error and can complete convergence faster which shows better prediction accuracy and efficiency.

基金项目：

河南省软科学研究计划项目（152400410203）；河南省科技攻关项目（192102210134）

作者简介：张海霞（1980-），女，硕士，讲师 E-mail：zhxhngm@163.com

参考文献：

[1]朱永波, 张飞, 张勇军, 等. 基于粒子群优化的带钢凸度神经网络预测模型研究[J].冶金自动化, 2019, 43(2):11-15，28.

Zhu Y B, Zhang F, Zhang Y J, et al. Prediction model of strip crown based on particle swarm optimization neural network [J]. Metallurgical Automation, 2019, 43(2):11-15，28.

[2]王付杰, 毛飞龙, 双远华, 等. 管材斜连轧过程的运动学分析及实验研究[J]. 锻压技术, 2021, 46(4): 215-222.

Wang F J, Mao F L, Shuang Y H, et al. Kinematics analysis and experimental study of pipe tanque rolling process [J]. Forging & Stamping Technology, 2021, 46(4): 215-222.

[3]孙孟乾, 孙建亮, 韩辉, 等. 大型铝合金筒节轧制过程圆度控制及影响因素分析[J]. 燕山大学学报, 2021, 45(2): 108-115.

Sun M G, Sun J L, Han H, et al.Roundness control and analysis of influencing factors in rolling process of large aluminum alloy barrel [J]. Journal of Yanshan University, 2021, 45(2): 108-115.

[4]李飞飞, 宋勇, 刘超, 等. 热轧带钢力学性能预报模型的误差分布建模研究[J].冶金自动化, 2019, 43(6):28-33.

Li F F, Song Y, Liu C, et al. Error distribution modeling of prediction model for mechanical properties of hot rolled strip [J]. Metallurgical Automation, 2019, 43(6):28-33.

[5]阎昱, 李嘉欣. 不同加载路径下的AZ31B镁合金的成形极限[J]. 锻压技术, 2021, 46(2): 40-46.

Yan Y, Li J X. Forming limit of AZ31B magnesium alloy under different loading paths [J]. Forging & Stamping Technology, 2021, 46(2): 40-46.

[6]乔及森, 王兵. 深冷轧制制备超细晶纯铝的微观组织与热稳定性研究[J]. 塑性工程学报, 2021, 28(2): 102-108.

Qiao J S, Wang B. Microstructure and thermal stability of ultrafine crystalline pure aluminum prepared by deep cold rolling [J]. Journal of Plasticity Engineering, 2021, 28(2): 102-108.

[7]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.

[8]魏立新，张宇，孙浩，等.基于改进OS-ELM的冷连轧在线轧制力预测[J].计量学报，2019，40(1): 111-116.

Wei L X，Zhang Y，Sun H，et al.On-line rolling force prediction of cold continuous rolling based on improved OS-ELM[J].Acta Metrology Sinica，2019，40(1): 111-116.

[9]赵文姣，闫洪伟，杨枕，等.基于CA-CAMC网络的轧制力自学习预测算法[J].冶金自动化，2016，40(2): 7-10.

Zhao W J, Yan H W, Yang Z, et al. Rolling force self-learning prediction algorithm based on CA-CAMC network [J]. Metallurgical Automation, 2016, 40(2): 7-10.

[10]Yao L，Ge Z Q.Deep learning of semi-supervised process data with hierarchical extreme learning machine and soft sensor application [J].IEEE Transactions on Industrial Electronics，2017，25(6): 1490-1498.

[11]曹建国, 江军, 赵秋芳, 等. 基于数据挖掘的宽厚板板凸度控制[J].中南大学学报:自然科学版, 2019, 50(11):2743-2752.

Cao J G, Jiang J, Zhao Q F, et al. Control of plate crown for wide and thick plates based on data mining [J]. Journal of Central South University: Science and Technology, 2019, 50(11):2743-2752.

[12]王强，吕政，王霖青，等.基于深度去噪核映射的长期预测算法[J].控制与决策，2019，34(5): 989-996.

Wang Q, Lyu Z, Wang L Q, et al. Long term prediction algorithm based on deep denoising kernel mapping [J].Control and Decision, 2019, 34(5): 989-996.

[13]Bengio Y，Lamblin P，Popvici D，et al.Greedy layer-wise training of deep networks [J].Advances in Neural Information Processing Systems，2007，19: 153-160.

[14]魏立新，魏新宇，孙浩，等.基于深度网络训练的铝热轧轧制力预测[J].中国有色金属学报，2018，28(10): 2070-2076.

Wei L X, Wei X Y, Sun H, et al. Rolling force prediction of aluminum hot rolling based on deep network training [J]. Chinese Journal of Nonferrous Metals, 2018, 28(10): 2070-2076.

[15]马湧, 王晓鹏, 马莎莎.基于Keras深度学习框架下BP神经网络的热轧带钢力学性能预测[J].冶金自动化, 2019, 43(2):6-10.

Ma Y, Wang X P, Ma S S. Prediction of mechanical properties of hot rolled strip based on BP neural network based on Keras deep learning framework [J]. Metallurgical Automation, 2019, 43(2):6-10.

[16]陈丹, 邵健, 殷实, 等. 基于大数据平台的冷连轧轧制力自学习模型优化[J].冶金自动化, 2020, 44(6)：25-29, 61.

Chen D, Shao J, Yin S, et al. Optimization of rolling force self-learning model for cold tandem rolling based on big data platform [J]. Metallurgical Automation, 2020, 44(6): 25-29, 61.

[17]曹建国, 江军, 邱澜, 等. 新一代高技术宽带钢冷轧机全机组一体化板形控制[J].中南大学学报:自然科学版, 2019, 50(7):1584-1591.

Cao J G, Jiang J, Qiu L, et al. Integrated shape control of new generation high technology wide strip cold rolling mill [J]. Journal of Central South University: Science and Technology, 2019, 50(7):1584-1591.

服务与反馈：

【文章下载】【加入收藏】