横向喷流与超声速来流干扰的机器学习预测研究  

Prediction for Lateral Jet Interaction with Supersonic Freestream Using Machine Learning

作　　者：韩天依星 胡姝瑶 蒋崇文[1] 高振勋[1] 李椿萱[1] 刘杰平[2] 蔡巧言[2] Han Tianyixing;Hu Shuyao;Jiang Chongwen;Gao Zhenxun;Li Chunxuan;Liu Jieping;Cai Qiaoyan(Beihang University,Beijing 100191,China;China Academy of Launch Vehicle Technology,Beijing 100076,China)

机构地区：[1]北京航空航天大学,北京100191 [2]中国运载火箭技术研究院,北京100076

出　　处：《气动研究与试验》2024年第5期58-68,共11页Aerodynamic Research & Experiment

摘　　要：横向喷流与超声速来流干扰是高超声速飞行器中反作用控制系统工作时发生的重要流动现象。本文基于机器学习方法对横向喷流干扰流动中的壁面干扰压力分布进行了建模,首先依据干扰压力系数分布特征进行数据采样;然后采用特征正交分解方法将采样数据降维;最后通过神经网络方法对干扰压力系数分布与工况参数的相关关系进行学习。本文采用的机器学习方法适用于喷流与超声速来流干扰流动,能够对壁面干扰压力分布进行较为准确的预测。The lateral jet interaction with supersonic freestream is an important phenomenon involved during the operation of the reactive control system for hypersonic vehicles.To predict the distribution of the interferential pressure on the wall for the lateral jet interaction,this paper utilized machine learning approaches as follows.First,data sampling according to the pressure distribution;Second,dimension reduction realized by proper orthogonal decomposition;Third,correlation between operating conditions and the pressure distribution using neural network.The results of the prediction by machine learning are almost the same with the numerical results,which indicates that the approaches adopted are applicable to the lateral jet interaction with supersonic freestream.

关 键 词：横向喷流干扰 机器学习 特征正交分解 神经网络 

分 类 号：V219[航空宇航科学与技术—航空宇航推进理论与工程]