摘要
热连轧过程中，为了提高轧制力预设定精度，提出一种新的修改轧制力模型参数的方法·利用ＢＰ 神经网络对以往的大量生产数据进行训练、预测·对ＢＰ 神经网络的预测结果利用最小二乘法，回归出轧制力模型中的温度相关系数ｍ １ 和变形速度相关系数ｍ ３·现场生产实验证明，应用修改后的轧制力模型系数，提高了轧制力预设定精度，从而使头部厚度精度有较大提高·对于象本溪钢铁公司热连轧厂这样的老企业，这种新方法更具有在线应用的可行性·
A new method of modifying parameters of the rolling load models was developed to improve the precision of predicted rolling load in a hot strip mill. Measured data of on line production were used to train and test the BP neural networks. The coefficient m 1 and m 3 , which relate to strip temperature and deforming rate in the rolling load models respectively,were regressed by means of the least square method according to the predictions of the BP neural networks. The field experiment indicates that, after putting the modified parameters into use, the precision of predicted rolling load is improved, so the strip head end accuracy is much better. The new method shows potential in the on line applications for some long service equipments such as Benxi Hot Strip Mill.
出处
《东北大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
1999年第5期522-524,共3页
Journal of Northeastern University(Natural Science)
基金
国家"九五"科技攻关项目
关键词
热连轧
轧制力
数学模型
神经网络
回归
hot strip mill, rolling load, mathematical models, neural networks, regression.
