冷轧平整机的工作辊直接和带钢接触，其表面粗糙度衰减情况对带钢成品的板形和表面质量有重大影响。因此，分析轧辊磨损机制，对轧辊表面粗糙度的衰减进行精确预测十分必要。首先采用灰色关联度分析对影响平整机工作辊表面粗糙度磨损的因素进行分析，确定了工作辊表面粗糙度评估指标体系。进而应用优化在线稀疏最小二乘支持向量回归模型对冷轧平整机的上工作辊表面粗糙度进行在线预测。通过预测误差准则实现系统的前向递推，采用FLOO(fast leave one out)的修剪算法实现其后向删减，并且采用最速下降法实现了2个超参数的在线优化。经过仿真研究表明，系统预测的绝对误差平均值为0.0149，与其他方法相比具有明显的优越性，并且系统具有在线自适应的能力，能够随着时间而进化。
The working rolls in cold rolling temper mill are in direct contact with the strips. Therefore,the surface rough-ness of working rolls will affect the flatness and surface quality of strip. So it’s necessary to predict the attenuation of working roll’s surface roughness accurately by analyzing roll wear mechanisms. Firstly,the factors which influence the working roll’s surface roughness are analyzed by the means of grey relational analysis. A system for evaluating the work-ing roll’s surface roughness is determined. Then optimized OS-LSSVR model is used for on-line prediction of the surface roughness. The new key nodes are added recursively by using prediction error criterion,and the redundant key nodes are deleted following FLOO. Moreover,the gradient descent method is adopted to optimize the two hyper-parameters online. The results of simulation show that the average absolute error of the model is 0.014 9,much smaller than other models. In addition,the model has on-line adaptive ability,and is able to evolve over time.
参考文献
| [1] | 李晓燕,张杰,陈先霖,曹建国,赵新明,贾生晖,艾斌,关贵民.冷轧平整机板形问题的特点及对策[J].钢铁,2003(12):26-29,33. |
| [2] | 徐江华,李山青.基于模糊推理的板形自动控制算法及其应用[J].钢铁,2013(09):45-49. |
| [3] | 丁顺风,杨利坡,于丙强.基于模糊评判规则的冷轧带钢智能板形评价模型[J].钢铁,2014(09):61-63,70. |
| [4] | 王淑华,马光宗,吕长宝.热轧钢卷平整生产线挫伤缺陷的原因分析[J].中国冶金,2014(05):45-49. |
| [5] | 白振华,周庆田,窦爱民,徐俊,王骏飞.冷连轧高速轧制过程中摩擦因数机理模型的研究[J].钢铁,2007(05):47-50. |
| [6] | 牛培峰,李鹏飞,李国强,马云飞.基于万有引力优化的支持向量机模型在板形识别中的应用[J].钢铁,2012(12):45-49. |
| [7] | 杜鹃,韩传基,蔡开科.基于最小二乘支持向量机的结晶器液面预测控制[J].钢铁研究学报,2007(08):24-27. |
| [8] | 张勇,赵冰.灰度关联分析结合支持向量机用于近红外光谱研究[J].光谱学与光谱分析,2013(02):363-366. |
| [9] | Hossam Mohammad Khalil;Mohammad El-Bardini .Implementation of speed controller for rotary hydraulic motor based on LS-SVM[J].Expert Systems with Application,2011(11):14249-14256. |
| [10] | 徐喆,毛志忠.钢包精炼炉元素收得率的影响因素分析及预报[J].钢铁,2012(03):34-37. |
| [11] | Mao-xiang CHU,An-na WANG,Rong-fen GONG,Mo SHA.Multi-class Classification Methods of Enhanced LS-TWSVM for Strip Steel Surface Defects[J].钢铁研究学报（英文版）,2014(02):174-180. |
| [12] | 叶美盈,汪晓东,张浩然.基于在线最小二乘支持向量机回归的混沌时间序列预测[J].物理学报,2005(06):2568-2573. |
| [13] | 王定成,姜斌.在线稀疏最小二乘支持向量机回归的研究[J].控制与决策,2007(02):132-137. |
| [14] | 张浩然,汪晓东.回归最小二乘支持向量机的增量和在线式学习算法[J].计算机学报,2006(03):400-406. |
| [15] | 邹雷雷,吕庆,刘东辉,陈树军,李凤臣,白瑞国.基于加权灰色关联度法分析印尼海砂最佳配比的研究[J].中国冶金,2014(09):6-9. |
| [16] | 李恩颖,王琥,李光耀.基于支持向量机回归的材料参数反求方法[J].机械工程学报,2012(06):90-95. |
| [17] | S. Deng;Tsung-Han Yeh .Applying least squares support vector machines to the airframe wing-box structural design cost estimation[J].Expert Systems with Application,2010(12):8417-8423. |
| [18] | 刘毅,陈坤,王海清,李平.选择性递推LSSVR及其在过程建模中的应用[J].高校化学工程学报,2008(06):1043-1048. |
| [19] | 刘毅 .间歇过程的核学习自适应建模与控制研究工业应用[D].杭州:浙江大学,2009. |
| [20] | Cawley GC;Talbot NL .Fast exact leave-one-out cross-validation of sparse least-squares support vector machines.[J].Neural Networks: The Official Journal of the International Neural Network Society,2004(10):1467-1475. |
| [21] | Suykens J A K;Brabanter J D;Lukas L et al.Weighted least squares support vector machines:robustness and sparse approxi-mation[J].NEUROCOMPUTING,2002,48(1/2/3/4):85. |
| [22] | 陶少辉,陈德钊,胡望明.LSSVM过程建模中超参数选取的梯度优化算法[J].化工学报,2007(06):1514-1517. |
- 下载量()
- 访问量()
- 您的评分：
-
10%
-
20%
-
30%
-
40%
-
50%