基于CF-PSO-SVM的冷连轧非稳态工作辊弯辊模型优化  

Optimization of work roll bending model in unsteady process of tandem cold rolling based on CF-PSO-SVM

作　　者：宋君 任廷志 王奎越 王军生 SONG Jun;REN Ting-zhi;WANG Kui-yue;WANG Jun-sheng(Education Ministry Engineering Research Center of Rolling Equipment and Complete Technology,Yanshan University,Qinhuangdao 066004,Hebei,China;Ansteel Beijing Research Institute,Ansteel Group,Beijing 102211,China)

机构地区：[1]燕山大学轧制设备及成套技术教育部工程研究中心,河北秦皇岛066004 [2]鞍钢集团北京研究院有限公司,北京102211

出　　处：《钢铁》2021年第11期78-86,共9页Iron and Steel

基　　金：国家重点研发计划资助项目(2017YFB0304100)。

摘　　要：板形质量是冷轧带钢重要的技术质量指标,同时工作辊弯辊是改善冷轧带钢板形质量的最有效的控制手段之一。冷连轧机组在高速稳定的轧制过程中板形控制精度能够达到较高的水平,但在升降速非稳态的轧制过程中板形控制效果非常不理想,这也成为制约冷轧带钢产品质量的不利因素。为了提高冷连轧机在升降速非稳态轧制过程中带钢板形的控制精度,在深入研究了冷连轧弯辊力设定原理的基础上,利用智能算法和包括出入口带钢厚度、机架间口张力、轧制速度、中间辊窜辊、带钢宽度、轧辊倾斜以及轧制力等现场实际轧制大数据样本,提出了一种基于粒子群算法优化支持向量机的工作辊弯辊力预测模型。同时阐明了粒子群优化算法和支持向量机的基本原理,引入压缩因子的概念,提升了粒子群算法参数寻优的效率,选取冷连轧机组五机架为研究对象,利用拉依达准则对轧制数据样本进行处理,通过平均绝对误差、均方差误差和平均绝对误差百分比等评价指标对比预测模型的性能。结果表明,改进的预测模型具有良好的模型预测性能和泛化能力,同时根据实际生产数据样本,回归出基于轧制速度和辊间弹性系数的弯辊力缝补偿模型,并验证了模型的有效性,模型的投入降低了板形控制系统的负荷,改善了非稳态轧制过程中的板形控制精度,产品头尾部的质量合格率提高了5.1%。Flatness quality is an important technical quality specifications of cold rolled strip.At the same time,work roll bending is one of the most effective control methods to improve the flatness quality of cold rolled strip.The strip flatness control precision of tandem cold mill can reach a high level in the process of high speed and stable rolling,but the effect of shape control is not ideal in the process of unsteady rolling,this has also become a negative factor restricting the quality of cold rolled strip.In order to improve the control precision of flatness in the unsteady process of acceleration and deceleration,the principle of bending force setting in tandem cold rolling was studied,combined with the intelligent algorithms and the actual rolling data samples including the strip thickness at the entry and outlet,the tension between the stands,the rolling speed,the middle roll shifting,the strip width,the roll inclination and the rolling force,a bending force prediction strategy based on the support vector machine theory of which parameters were optimized by the particle swarm optimization algorithm was proposed.The principle of support vector machine(SVM)theory and particle swarm optimization(PSO)algorithm was studied.In order to improve the parameters optimization ability,the concepts of compression factor were introduced in the PSO algorithm.Moreover,the PauTa criterion and five-point three-time smoothing method were used to process the relevant rolling data,and then the performance of the prediction strategies was compared by the evaluation indexes,such as mean square error and mean absolute error.The results show that the proposed algorithm was verified with good predictive performance and excellent generalization ability,at the same time,according to the actual production data sample,the compensation model of bending force gap based on rolling speed and roll elasticity coefficient was regressed,and the validity of the model was verified,the model reduced the load of the flatness control system and improv

关 键 词：冷连轧 板形控制 弯辊力模型 粒子群算法 支持向量机 

分 类 号：TG333.17[金属学及工艺—金属压力加工]