热轧带钢精轧过程的混合温度模型  

Hybrid temperature model for hot strip in finishing stands

作　　者：傅新[1] 陈水宣[1] 邹俊[1] 杨华勇[1] 

机构地区：[1]浙江大学流体传动及控制国家重点实验室,浙江杭州310027

出　　处：《浙江大学学报（工学版）》2008年第2期219-223,共5页Journal of Zhejiang University：Engineering Science

基　　金：国家"973"重点基础研究发展规划资助项目(2006CB705400);国家自然科学基金资助项目(50575200)

摘　　要：结合现场数据采集系统的实测数据,应用差分法替换精轧温度设定系统中计算误差较大的轧制传导温度模型,根据带钢各层温度的分布特点提出变长和变厚度网格划分的方法;对原精轧温度设定系统中空冷温降模型的辐射系数进行离线的钢种和厚度等级修正,并建立新的指数水冷温降模型,在线调整对流换热系数.由此构建的混合温度模型在保证计算速度的同时大幅度提高了预测精度.为满足现场生产精度和稳定性的需要,采用"接力式"初始化训练方法,建立了神经网络自学习系统.结果表明,改造后精轧温度设定系统的预测精度比原系统高,并且在变钢种、变规格轧制时误差波动小.Based on the production data from the field data acquisition system, a differential model was proposed to replace the former thermal model of rolling conduction in the finishing temperature setting system, characterized by a method of dividing elements in varied length and thickness according to the characteristic of thermal distribution in hot steel. Furthermore, an off-line adaption technique based on steel grade and thickness class was adopted to correct the radiation factor of air cooling model, and the original water cooling model was replaced by a new exponential model with on-line adjustment of the convective heat transfer coefficient. Consequently, the enhanced hybrid temperature model presented higher prediction accuracy as well as acceptable response speed. In addition, by the way of ＂Relay＂ initialization learning, a neural network adaption system was put forward to match the requirement of hot strip production. Both simulation and experimental results confirmed that the setup accuracy of finishing temperature was effectively improved and the error deviation was greatly decreased when steel grade and dimension were changed.

关 键 词：热轧带钢 精轧过程 混合温度模型 神经网络 自适应 

分 类 号：TH137[机械工程—机械制造及自动化]