热轧板坯头部翘曲机器视觉检测与BP神经网络预测控制  

Prediction and control for head warpage of hot-rolled slab based on machine vision detection and BP neural network

作　　者：黄华贵[1] 刘迎港 黄海林 陈光 王晓东 HUANG Hua-gui;LIU Ying-gang;HUANG Hai-lin;CHEN Guang;WANG Xiao-dong(National Engineering Research Center for Equipment and Technology of Cold Strip Rolling,Yanshan University,Qinhuangdao 066004,Hebei,China;Technology Center,Shougang Jingtang Iron and Steel United Co.,Ltd.,Tangshan 063009,Hebei,China)

机构地区：[1]燕山大学国家冷轧板带装备及工艺工程技术中心,河北秦皇岛066004 [2]首钢京唐钢铁联合有限公司技术中心,河北唐山063009

出　　处：《中国冶金》2022年第9期85-89,96,共6页China Metallurgy

基　　金：中央引导地方科技发展资金资助项目(216Z1602G)。

摘　　要：为解决热轧粗轧阶段板坯头部翘曲自动控制问题,以某2250 mm热轧线粗轧机组为对象,介绍机器视觉检测原理及其在热轧生产线检测位置及检测结果的表征方式。通过分析影响板坯头部翘曲的因素,建立了9-10-6-1的4层BP神经网络预测模型,以现场检测数据为样本对网络进行了训练和预测验证。结果表明,神经网络模型预测精度符合现场的控制要求,这为热轧过程头尾翘曲自动控制和生产的无人值守奠定了基础。In order to solve the problem for automatic control of slab head warpage in the roughing stage of hot rolling,the principle of machine vision inspection and its characterization of inspection position and inspection results on the hot rolling line were introduced taking a 2250 mm hot rolling mill roughing unit as the object.A 9-10-6-1 four-layer BP neural network prediction model was established by analyzing factors affecting the head warpage of slab,and the network was trained and validated using the field inspection data as samples.The results show that the prediction accuracy of the neural network model meets the control requirements in the field,which lays the foundation for the automatic control of head and tail warpage in the hot rolling process and the unattended production.

关 键 词：热轧 头部翘曲 机器视觉 BP神经网络 自动轧制 

分 类 号：TG335.11[金属学及工艺—金属压力加工] TP183[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程]