基于迭代学习的轧辊偏心补偿控制  

Roll eccentricity compensation control based on iterative learning

作　　者：乔继柱 史鸿剑 孙杰[1] 彭文[1] 李霄剑 张殿华[1] QIAO Ji-zhu;SHI Hong-jian;SUN Jie;PENG Wen;LI Xiao-jian;ZHANG Dian-hua(State Key Laboratory of Rolling and Automation,Northeastern University,Shenyang 110819,Liaoning,China;Equipment Department,Ningbo Baoxin Stainless Steel Co.,Ltd.,Ningbo 315800,Zhejiang,China;College of Information Science and Engineering,Northeastern University,Shenyang 110819,Liaoning,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]宁波宝新不锈钢有限公司设备部,浙江宁波315800 [3]东北大学信息科学与工程学院,辽宁沈阳110819

出　　处：《中国冶金》2023年第5期102-108,共7页China Metallurgy

基　　金：国家重点研发计划资助项目(2022YFB3304800);国家自然科学基金资助项目(U21A20117,U21A20475,52074085);辽宁省兴辽英才计划资助项目(XLYC1907065);中央高校基本科研业务费资助项目(N2004010)。

摘　　要：针对轧机普遍存在的轧辊偏心问题,结合偏心信号的特点及补偿控制方法,将一种在工业机器人重复性控制领域广泛使用的迭代学习控制方法应用在轧辊偏心补偿控制中。基于对周期误差数据不断学习,轧机偏心控制系统的压下控制调整更加精准。对构造的轧辊偏心信号进行测试,其相比于快速傅里叶变换补偿方法,迭代学习10~50次后的补偿精度提高20.1%~56.4%;经过迭代学习律参数的优化,迭代学习后补偿精度达到50%以上的同时,收敛速度大幅增加。考虑到在轧机加减速状态时偏心信号频率发生改变,通过迭代学习控制对变频信号进行补偿,经过10次的快速学习补偿,误差评价函数值达到最初的13.6%。利用冷连轧一道次的生产线实际数据进行测试,结果表明迭代学习补偿控制能够在学习过后有效提升偏心补偿精度。研究基于迭代学习控制方法,为轧辊偏心的补偿控制提供一定的参考。Aiming at roll eccentricity in rolling mill,an iterative learning control method which was widely used in industrial robot repeatability control was applied to roll eccentricity compensation control by combining the characteristics of eccentricity signal and compensation control method.On the basis of learning the cycle error data,the adjustment of rolling mill eccentricity control system was more accurate.The test results of constructed roller eccentricity signal show that compared with the fast Fourier transform compensation method,the compensation accuracy is improved by 20.1%-56.4%after 10-50 iteration learning.After iterative learning law parameters are optimized,the compensation accuracy reaches more than 50%after iterative learning,and the convergence speed is greatly improved.Considering that the frequency of eccentricity signal changed when the rolling mill was accelerating and decelerating,the frequency conversion signal was compensated by iterative learning control.After 10 times of fast learning compensation,the value of error evaluation function reached 13.6%of initial.The actual production line data of one pass in continuous cold rolling are used for testing,and the results show that the iterative learning compensation control can effectively improve the accuracy of eccentricity compensation after learning.Study results based on the iterative learning control method provides some reference for the compensation control of roll eccentricity.

关 键 词：轧辊偏心 厚度控制 迭代学习 参数优化 变频补偿 

分 类 号：TG333.17[金属学及工艺—金属压力加工]