二十辊轧机轧制力动态预设定模型研究
|
摘要
针对某厂二十辊森吉米尔轧机生产不锈钢时存在非稳态过程轧制力设定偏差较大,带钢厚度超差段较长的问题,建立了摩擦因数数据库,采用迭代反算方式计算摩擦因数并存于数据库,在轧制力计算时提供摩擦因数;然后在机组原有ALSTOM轧制模型及Bland-ford公式的基础上,考虑了弹性区和温度变化的影响,建立了轧制力动态预设定模型,可在AGC投入前进行动态计算,时间间隔为1.5s;设定结果与实测数据比较表明,相对误差基本在±3%以内,厚度偏差2%以上的长度控制在5m以内,实现了森吉米尔轧机生产不锈钢时轧制力的精确设定。
For the production of stainless steel in 20-h Sendzimir mill,there are problems that the setting deviation of the rolling force is large in the unsteady stage and the length of thickness fluctuation of strip is longer.The friction coefficient database was established,and the coefficient of friction was calculated by the iterative inverse method and stored in the database.It was provided in the calculation of rolling force.Based on the Bland-ford formula and considering the influence of elastic zone and temperature change,the dynamic pre-setting model of rolling force was established by absorbing the original ALSTOM rolling model.Dynamic calculation could be made before the AGC was put into operation and the interval was 1 s.The comparison between the setting result and the measured data suggests that the relative error is less than ±3%,the length whose thickness deviation is more than 2% is controlled within 5 meters,and the precise setting of rolling force during the production of stainless steel in Sendzimir mill is achieved.
For the production of stainless steel in 20-h Sendzimir mill,there are problems that the setting deviation of the rolling force is large in the unsteady stage and the length of thickness fluctuation of strip is longer.The friction coefficient database was established,and the coefficient of friction was calculated by the iterative inverse method and stored in the database.It was provided in the calculation of rolling force.Based on the Bland-ford formula and considering the influence of elastic zone and temperature change,the dynamic pre-setting model of rolling force was established by absorbing the original ALSTOM rolling model.Dynamic calculation could be made before the AGC was put into operation and the interval was 1 s.The comparison between the setting result and the measured data suggests that the relative error is less than ±3%,the length whose thickness deviation is more than 2% is controlled within 5 meters,and the precise setting of rolling force during the production of stainless steel in Sendzimir mill is achieved.
引文
[1]周富强,曹建国,张杰,等.基于神经网络的冷连轧机轧制力预报模型[J].中南大学学报:自然科学版,2006,37(6):1155.
[2]薛涛,杜凤山,孙静娜,等.基于FEM-ANN的冷轧带钢轧制力预报[J].中南大学学报:自然科学版,2013,44(11):4456.
[3]刘欣玉,潘露,帅美荣.基于Matlab的BP神经网络轧制力预报模型及应用[J].重庆科技学院学报:自然科学版,2016,18(6):96.
[4]张俊明,刘军,康永林,等.应用RBF神经网络预测冷连轧机轧制力[J].钢铁,2007,42(8):46.
[5]梁勋国,贾涛,矫志杰,等.基于贝叶斯方法的神经网络应用于冷轧轧制力预报[J].钢铁研究学报,2008,20(10):59.
[6]刘承宝.基于最小二乘支持向量机的轧制力预测[D].北京:冶金自动化研究设计院,2016.
[7]杨景明,顾佳琪,闫晓莹.基于改进遗传算法优化BP网络的轧制力预测研究[J].矿冶工程,2015,35(1):111.
[8]张清东,徐兴刚,于孟,等.基于遗传神经网络的不锈钢带冷轧轧制力模型[J].钢铁,2008,43(12):46.
[9]郝心耀.基于机器学习算法的轧机轧制力预测[J].现代电子技术,2016,39(20):114.
[10]高雷,郭立伟,陈丹.冷连轧轧制力模型中变形抗力和摩擦系数的分析[J].轧钢,2013,30(4):12.
[11]白金兰,王军生,王国栋,等.提高冷轧过程控制轧制力模型的设定精度[J].钢铁研究学报,2006,18(3):21.
[12]高雷,郭立伟,李书昌,等.冷轧轧制过程摩擦系数模型优化方法研究[J].冶金自动化,2014,38(5):49.
[13]王伟,连家创.采用混合摩擦模型预报冷轧薄板轧制力[J].钢铁研究学报,2000,12(1):10.
[2]薛涛,杜凤山,孙静娜,等.基于FEM-ANN的冷轧带钢轧制力预报[J].中南大学学报:自然科学版,2013,44(11):4456.
[3]刘欣玉,潘露,帅美荣.基于Matlab的BP神经网络轧制力预报模型及应用[J].重庆科技学院学报:自然科学版,2016,18(6):96.
[4]张俊明,刘军,康永林,等.应用RBF神经网络预测冷连轧机轧制力[J].钢铁,2007,42(8):46.
[5]梁勋国,贾涛,矫志杰,等.基于贝叶斯方法的神经网络应用于冷轧轧制力预报[J].钢铁研究学报,2008,20(10):59.
[6]刘承宝.基于最小二乘支持向量机的轧制力预测[D].北京:冶金自动化研究设计院,2016.
[7]杨景明,顾佳琪,闫晓莹.基于改进遗传算法优化BP网络的轧制力预测研究[J].矿冶工程,2015,35(1):111.
[8]张清东,徐兴刚,于孟,等.基于遗传神经网络的不锈钢带冷轧轧制力模型[J].钢铁,2008,43(12):46.
[9]郝心耀.基于机器学习算法的轧机轧制力预测[J].现代电子技术,2016,39(20):114.
[10]高雷,郭立伟,陈丹.冷连轧轧制力模型中变形抗力和摩擦系数的分析[J].轧钢,2013,30(4):12.
[11]白金兰,王军生,王国栋,等.提高冷轧过程控制轧制力模型的设定精度[J].钢铁研究学报,2006,18(3):21.
[12]高雷,郭立伟,李书昌,等.冷轧轧制过程摩擦系数模型优化方法研究[J].冶金自动化,2014,38(5):49.
[13]王伟,连家创.采用混合摩擦模型预报冷轧薄板轧制力[J].钢铁研究学报,2000,12(1):10.