摘要
通过提取高炉风口视频数据的帧图像,并结合先进的图像识别算法监测高炉风口区域的工作状态,实时分析相应的高炉调参策略,有利于降低高炉风口的休风率与时长,同时弥补现阶段依靠人工经验判断风口状态存在的响应滞后、结果不准确的工艺缺陷,保证高炉的长期稳定顺行。以国内某钢铁厂在2023年6月1日—6月31日的风口视频数据为基础,梳理了高炉炼铁过程中4种常见的风口异常状况,并结合炼铁原理分析了主要原因及应对措施,包括挂渣、涌渣、断煤和漏水。然后总结了图像识别技术在高炉风口识别与监测中的应用路线,包括图像预处理、风口的识别与预警和专家经验的植入3个阶段,同时对目前应用较广的图像识别算法进行了介绍,包括卷积神经网络、Transformer机制和图神经网络,并对后2种算法进行了肯定。最后,基于图卷积神经网络开发了关于高炉风口的监测与分析系统1.0版本,并对其功能进行了简要介绍。秉持低延时与高精度的发展原则,通过对风口异常和图像识别算法的梳理探究未来高炉风口图像识别的应用路线,为钢铁企业选择合理的风口监测技术,提升风口识别与监测领域的智能化水平提供一种理论参考。
By extracting the frame image of BF tuyeres’video data,and combining with advanced image recognition algorithm to monitor the working state of tuyeres’area,and analyzing the corresponding blast furnace parameter adjustment strategy in real time,which is conducive to reducing the air rest rate and duration,and at the same time making up for the process defects of the response lag and inaccurate results in judging tuyere’s state by relying on manual experience at this stage,so as to ensure the long-term stability and smooth movement of blast furnace.Based on the tuyeres’video data of a domestic steel plant between June 1~June 31,2023,this article sort out four common tuyere abnormalities in ironmaking process,and analyzed main causes and countermeasures based on the principle of ironmaking,including hanging slag,slag inflow,coal cutoff and water leakage.Then,the application route of image recognition technology in blast furnace tuyere recognition and monitoring is summarized,including image preprocessing,tuyere identification and early warning,and implantation of expert experience,and the widely used image recognition algorithms are introduced,including convolutional neural network,Transformer mechanism and graph neural network,and the latter two algorithms are affirmed and respected.Finally,based on the graph convolutional neural network,the monitoring and analysis system 1.0 of the blast furnace tuyere is developed,and its functions are briefly introduced.Adhering to the development principle of low latency and high precision,this article aims to explore image recognition’s application route of blast furnace tuyere in the future by combing the tuyere anomaly and image recognition algorithm,so as to provide a theoretical reference for China's steel enterprises to select reasonable tuyere monitoring technology and improve the intelligent level of tuyere identification and monitoring.
作者
段一凡
刘然
刘小杰
李欣
袁雪涛
吕庆
DUAN Yifan;LIU Ran;LIU Xiaojie;LI Xin;YUAN Xuetao;LÜQing(College of Metallurgy and Energy,North China University of Technology,Tangshan,063210,Hebei,China;Tangshan Branch,HBIS Group Co.,Ltd.,Tangshan 063020,Hebei,China.)
出处
《钢铁》
CAS
CSCD
北大核心
2024年第5期56-70,共15页
Iron and Steel
基金
国家自然科学基金青年基金资助项目(52004096)。
关键词
高炉风口监测
风口异常
图像识别
智能化建设
高炉炼铁
blast furnace tuyere monitoring
tuyere abnormalities
image recognition
intelligent construction
blast furnace ironmaking
