基于人工智能中间包内羽流及卷渣的数字孪生分析  

Digital Twin Analysis of Plume and Slag in Tundish Based on Artificial Intelligence

作　　者：李晓婷 金焱[1] 吴健舟 刘子钰 林鹏 黄正超 凌宏志 LI Xiaoting;JIN Yan;WU Jianzhou;LIU Ziyu;LIN Peng;HUANG Zhengchao;LING Hongzhi(Key Laboratory of Iron and Steel Metallurgy and Resource Utilization,Ministry of Education,Wuhan University of Science and Technology,Wuhan 430081)

机构地区：[1]武汉科技大学钢铁冶金及资源利用省部共建教育部重点实验室,武汉430081

出　　处：《特种铸造及有色合金》2024年第7期902-908,共7页Special Casting & Nonferrous Alloys

基　　金：国家自然科学基金资助项目(52174324)。

摘　　要：中间包长水口吹氩缺乏监控钢水面以下气泡羽流的手段,这成为利用长水口吹氩去除钢水中夹杂物的障碍。提出了一种神经网络模型的新型算法,由钢水水面的波动形态推断钢水内部气泡羽流的形态,由数学模型构建数据集,经2000个数据集学习后,推断精度达到0.998。建立了水模型验证该算法,并利用该算法将水模型和数学模型相结合,优化了长水口吹氩工艺,使夹杂物去除率提高同时防止了卷渣现象的发生。Tundish long nozzle blowing argon lacks the means to monitor the bubble plume below the molten steel sur⁃face,which becomes an obstacle to the use of long spout blowing argon to remove inclusions in the molten steel.A new algorithm of neural network model was proposed to infer the morphology of bubble plume inside molten steel from the fluctuation pattern of the molten steel surface,and the dataset was constructed by the mathematical model.The results indicate that inference accuracy reaches 0.998 after learning from 2000 datasets.A water model was established to verify the algorithm,and the water model and mathematical model were combined to optimize the argon blowing process at the long spout,improving the removal rate of inclusions and prevent the occurrence of slag rolling.

关 键 词：中间包 流场 神经网络 深度学习 机器学习 

分 类 号：TG249[金属学及工艺—铸造] TP311[自动化与计算机技术—计算机软件与理论]