锻压技术

基于迁移学习和多方面特征提取的冷轧轧制力预测轧制

英文标题：Prediction on cold rolling force based on transfer learning and multi-aspect feature extraction

单位：1.河北工程大学 2.河北省冀南新区现代装备制造协同创新中心 3.邯郸市永年区职业技术教育中心

分类号：TF345

出版年,卷(期):页码：2024,49(6):141-148

摘要：

针对冷轧机各机架轧制工况复杂，实际采集的工艺数据存在分布差异，导致小数量样本下轧制力预测模型对不同机架预测时出现精度降低的问题，提出一种基于迁移学习和多方面特征提取的轧制力预测模型。该方法通过迁移学习将建立的轧制力预测基准模型参数迁移至目标机架，从而实现小数量样本下不同机架的轧制力快速预测。预测时，利用源域数据建立了Inception-LSTM-Attention轧制力预测基准模型，基准模型结合了Inception网络模块的空间特征提取能力和LSTM的时序特征预测能力，同时引入注意力机制Attention调节神经网络学习到的信息向量权重，基准模型预测精度高达98.8%。然后将基准模型部分网络模块冻结迁移至目标域模型，并对模型参数进行微调，建立了最终的轧制力预测模型。实验结果表明，小数量样本下采用迁移学习方法建立的模型在预测精度方面高于传统神经网络预测模型。

Aiming at the problem of reduced prediction accuracy of rolling force prediction model under small number of samples when predicting the rolling force of different frames,which is caused by the complex rolling condition of each frame for cold rolling mill to make the distribution of actual collected process data different, a rolling force prediction model based on transfer learning and multi-aspect feature extraction was proposed. This method could transfer the parameters of established rolling force prediction benchmark model to the target frame by transfer learning to realize the rapid prediction of rolling force in different frames with small number of samples. When predicting, Inception-LSTM-Attention rolling force prediction benchmark model was established by using the source domain data, which combined the spatial feature extraction ability of Inception network module and the temporal feature prediction ability of LSTM. At the same time, the attention mechanism Attention was introduced to adjust the weight of information vector learned by the neural network, and the prediction accuracy of the benchmark model was as high as 98.8%. Then, some network modules of the benchmark model were frozen and transfered to the target domain model, and the model parameters were fine-tuned to establish the final rolling force prediction model. The experimental results show that the prediction accuracy of the model established based on the transfer learning method under small number of samples is higher than that of the traditional neural network prediction model.

基金项目：

国家自然科学基金青年项目（52005148）

作者简介：张明（1988-），男，博士，副教授,E-mail：zhangming@hebeu.edu.cn

参考文献：

[1]魏立新,王恒,孙浩，等.基于改进深度信念网络训练的冷轧轧制力预报[J].计量学报,2021,42 (7):906-912.

Wei L X, Wang H, Sun H, et al. Research on cold rolling force prediction model based on improved deep belief network[J]. Acta Metrologica Sinica，2021,42(7):906-912.

[2]Hu Z Y,Wei Z H,Sun H, et al. Optimization of metal rolling control using soft computing approaches: A review[J].Archives of Computational Methods in Engineering,2019,28(2): 405-421.

[3]Shen S H, Guye D, Ma X P, et al. Multistep networks for roll force prediction in hot strip rolling mill[J].Machine Learning with Applications,2021,7(4):100245.

[4]郭金涛,王龙,余建波,等.基于深度学习的宽厚板热轧轧制力预测[J].锻压技术,2022,47(7):167-174.

Guo J T, Wang L, Yu J B, et al. Prediction on rolling force in hot rolling of wide and thick plate based on deep learning [J]. Forging ＆ Stamping Technology,2022,47(7):167-174.

[5]崔桂梅,刘伟,张帅，等.基于差分进化支持向量机的轧制力预测[J].中国测试,2021,47(8):83-88.

Cui G M, Liu W, Zhang S, et al. Rolling force prediction based on differential evolution support vector machine[J]. China Measurement & Test,2021,47(8):83-88.

[6]吴爽,闫奕,李爽，等.冷连轧轧制力深度神经网络模型泛化能力并行优化[J].机械设计与制造,2023,(8):171-174.

Wu S, Yan Y, Li S, et al. Parallel optimization of generalization capability of rolling force deep neural network model in tandem cold rolling mill[J]. Machinery Design & Manufacture,2023,(8):171-174.

[7]陈树宗,白芸松,侯佳琦，等.基于GA-FELM算法的冷轧轧制力预测模型[J].燕山大学学报,2022,46(3):224-229.

Chen S Z, Bai Y S, Hou J Q, et al. Rolling force prediction model for cold rolling based on GA-FELM[J]. Journal of Yanshan University,2022,46(3):224-229.

[8]孙浩,赵明达,李静，等.基于LSTM-JITRVM的冷轧轧制力建模方法研究[J].计量学报,2023,44(9):1409-1416.

Sun H,Zhao M D, Li J, et al. Research on modeling method of cold rolling force based on LSTM-JITRVM[J]. Acta Metrologica Sinica, 2023,44(9):1409-1416.

[9]魏立新,翟博豪,赵志伟，等.基于半监督深度网络的冷连轧轧制力预报[J].塑性工程学报,2020,27(11):70-76.

Wei L X, Zhai B H, Zhao Z W, et al. Prediction of cold continuous rolling force based on semi-supervised deep network[J]. Journal of Plasticity Engineering,2020,27(11):70-76.

[10]Liu J Y,Liu X X,Le T B.Rolling force prediction of hot rolling based on GA-MELM[J].Complexity,2019,(2019):1-11.

[11]Wang C L,Zhang M.Research on dynamic rolling force prediction model based on CNN-BN-LSTM[J].Journal of Advanced Mechanical Design, Systems and Manufacturing,2022,16(3):1-14.

[12]李维刚,刘玮汲,谢璐，等.基于图卷积网络的热轧带钢轧制力预测[J].钢铁,2023,58(3):89-96，127.

Li W G, Liu W J, Xie L, et al. Rolling force prediction of hot rolled strip by graph convolutional networks[J]. Iron and Steel,2023,58(3):89-96，127.

[13]欧阳福莲,王俊,周杭霞.基于改进迁移学习和多尺度CNN-BiLSTM-Attention的短期电力负荷预测方法[J].电力系统保护与控制,2023,51(2):132-140.

Ouyang F L, Wang J, Zhou H X. Short-term power load forecasting method based on improved hierarchical transfer learning and multi-scale CNN-BiLSTM-Attention[J]. Power System Protection and Control，2023,51(2):132-140.

[14]那峙雄,孙涛,来广志，等.多尺度特征融合的光伏电站故障诊断[J].计算机工程与应用,2022,58(10):300-308.

Na Z X, Sun T, Lai G Z, et al. Fault diagnosis for photovoltaic power station by multi-scale features fusion[J]. Computer Engineering and Applications,2022,58(10):300-308.

[15]Li M T，Lu Y，Cao S X, et al.A hyperspectral image classification method based on the nonlocal attention mechanism of a multiscale convolutional neural network[J].Sensors,2023,23(6):3190.

[16]梁涛,陈春宇,谭建鑫，等.基于多方面特征提取和迁移学习的风速预测[J].太阳能学报,2023,44(4):132-139.

Liang T, Chen C Y, Tan J X, et al. Wind speed prediction based on multiple feature extraction and transfer learning[J]. Acta Energiae Solaris Sinica,2023,44(4):132-139.

[17]Md Rashedul Islam，Momotaz Begum，Md Nasim Akhtar. Recursive approach for multiple step-ahead software fault prediction through long short-term memory (LSTM)[J].Journal of Discrete Mathematical Sciences and Cryptography,2022,25(7):2129-2138.

[18]Jrges C, Berkenbrink C, Stumpe B.Prediction and reconstruction of ocean wave heights based on bathymetric data using LSTM neural networks[J].Ocean Engineering,2021,232: 109046.

[19]贾睿,杨国华,郑豪丰，等.基于自适应权重的CNN-LSTMGRU组合风电功率预测方法[J].中国电力,2022,55(5):47-56，110.

Jia R, Yang G H, Zheng H F, et al. Combined wind power prediction method based on CNN-LSTM&GRU with adaptive weights[J]. Electric Power,2022,55(5):47-56，110.

服务与反馈：

【文章下载】【加入收藏】