基于神经网络的连续热镀锌线拉矫机工艺参数设定模型  

Neural Network Setting Model for Tension Leveller on the Continuous Galvanizing Line

作　　者：丁军[1] 张清东[1] 常铁柱[1] 姜正连[2] 吴彬[2] 

机构地区：[1]北京科技大学机械工程学院,北京100083 [2]宝钢股份冷轧厂,上海200941

出　　处：《冶金设备》2007年第1期33-36,共4页Metallurgical Equipment

摘　　要：针对冷轧宽带钢连续热镀锌生产线上的拉矫机,为了建立可针对不同钢种、规格、来料板形的拉矫机插入深度、延伸率和工艺张力等重要工艺参数的设定模型,结合对拉矫变形过程的力学分析和现场实测数据统计与分析的结果,从板形良好的钢卷中择取有代表性的生产数据作为训练样本,将良好的操作经验转化为数学模型,利用BP网络模型的自适应和自学习优势设定带钢拉矫过程中的工艺参数。此模型可以将新的样本随时添加到网络的矩阵中,对新品种具有较好的扩展能力。经现场试用表明,它能准确地预设定拉矫机工艺参数,基本满足工厂生产要求。On the continuous galvanizing line of cold rolled strip, parameters setting of the tension leveller include inter-mesh, elongation and processing tension. To establish the tension leveller parameters setting model, it should consider the strip steehype variety, species and strip flatness. Combined the mechanism analysis of levelling process with practical statistical producing data and results, taking the representative coils data which have the well flatness as training sample of the BP neural network, and transforming the well operation experience of the workers into mathematics model, this superiority of auto-adaptive and self-study is used for the processing parameter setting of strip Tension leveller. New samples can be appended to the BP matrix in this model at any moment, so it has a expendding capability to new species. The autoptic results demonstrate that it can accurately preset the leveller processing parameter, and satisfy production requirement in factory.

关 键 词：神经网络 拉矫机 延伸率 数学模型 带钢 

分 类 号：TG333.23[金属学及工艺—金属压力加工]