基于随机配置网络-蜣螂优化算法的硅钢热轧过程弯辊力和窜辊量优化策略  

Optimization strategy of bending force and rolling shift for hot rolling process of silicon steel based on stochastic configuration network-dung beetle optimization algorithm

作　　者：杜昊展 丁敬国[1] 孙建红 曹国屿 赵健 李旭[1] 张殿华[1] DU Haozhan;DING Jingguo;SUN Jianhong;CAO Guoyu;ZHAO Jian;LI Xu;ZHANG Dianhua(State Key Laboratory of Rolling and Automation,Northeastern University,Shenyang 110819,China;Hot Strip Mill,Bengang Steel Plates Co.,Ltd.,Benxi 117000,China;Cold Rolled Silicon Steel Mill,Anshan Iron and Steel Group Co.,Ltd.,Anshan 114031,China)

机构地区：[1]东北大学轧制技术及连轧自动化国家重点实验室,辽宁沈阳110819 [2]本钢板材股份有限公司热连轧厂,辽宁本溪117000 [3]鞍山钢铁集团有限公司冷轧硅钢厂,辽宁鞍山114031

出　　处：《冶金自动化》2024年第4期33-45,63,共14页Metallurgical Industry Automation

基　　金：国家自然科学基金区域创新发展联合基金项目(U21A20475)。

摘　　要：热轧硅钢板形对冷轧板形和边降具有显著的遗传效应,减小热轧硅钢产品横向同板差可以有效提高冷轧产品的质量。在硅钢热连轧过程中,换规格或换牌号会出现弯辊力和窜辊量预设定值不准确的问题,使板形控制的效果降低。针对该问题,本文提出了一种基于随机配置网络(stochastic configuration network,SCN)的硅钢板凸度预测模型。为了提升模型对数据的拟合能力,增加了模型的隐藏层层数(deep stochastic configuration network,DeepSCN),并在SCN的建模过程中引入了流形正则化项(regularization stochastic configuration,RSC)。以数据驱动模型的预测结果为导向,采用蜣螂优化(dung beetle optimizer,DBO)算法对弯辊力和窜辊量进行优化。根据优化结果可知,该方法可以使比例凸度的波动控制在较小范围内,并将精轧出口凸度偏差在±5μm以内的数据提高到92.2%。这不仅有效提高了硅钢的板形质量,也为硅钢板形控制提供了新的研究方向和技术手段。Hot rolled silicon steel plate shape has significant genetic effect on the cold rolled plate shape and edge drop,and reducing the hot rolled silicon steel products transverse with the same plate difference can effectively improve the quality of cold rolled products.However,in the process of hot rolling of silicon steel,changing specifications or steel grades can lead to inaccurate preset values of bending force and rolling shift,which leads to a poor plate shape control effect.To address this problem,this paper proposes a prediction model for silicon steel crown based on stochastic configuration network(SCN).To improve the model's capacity to fit the data,the number of hidden layers in the model was increased(DeepSCN),and incorporated the manifold regularization term in the SCN modeling process(RSC).Guided by the datadriven model's prediction results,the dung beetle optimization(DBO)algorithm was used to optimize the bending force and rolling shift.The results show that this method can control the fluctuation of the proportional crown within a small range,and the data of the finishing mill exit plate crown deviation within依5滋m can be improved to 92.2%.This not only effectively improves the quality of plate shape of silicon steel,but also provides a new research direction and technical method for the control of plate shape in silicon steel.

关 键 词：硅钢板形 随机配置网络(SCN) 流形正则化 弯窜优化 蜣螂优化算法(DBO) 

分 类 号：TG335.11[金属学及工艺—金属压力加工] TP18[自动化与计算机技术—控制理论与控制工程]