基于深度学习和影响函数法的机翼非定常流场预测  

Prediction of Wing Unsteady Flow Field based on Deep Learning and Influence Function Method

作　　者：于煜斌 落龑寿 缪佶 曹隽喆[2] 李京杰 YU Yubin;LUO Yanshou;MIAO Ji;CAO Junzhe;LI Jingjie(Beijing Institute of Astronautical Systems Engineering,Beijing,100076;Dalian University of Technology,Dalian,116024)

机构地区：[1]北京宇航系统工程研究所,北京100076 [2]大连理工大学,大连116024

出　　处：《导弹与航天运载技术（中英文）》2023年第5期32-37,共6页Missiles and Space Vehicles

摘　　要：为兼顾非定常流场求解的精度与速度,基于深度学习和影响函数法,提出了一种机翼非定常流场特性预测方法,将影响函数嵌入深度学习网络的训练过程中,并引入早停机制和Dropout机制,实现了机翼流场参数的快速预测。结果表明,该方法具有较高的预测精度并且其计算效率较CFD方法有大幅提高。Based on the deep learning technology and influence function method,a prediction approach for the unsteady flow field characteristics of the wing is proposed.In this method,the influence function is adopted in the training process of the deep learning network,and lead into the early stop mechanism and the dropout mechanism to realize the fast prediction of the wing flow field parameters.The results show that this method has higher prediction accuracy and its computational efficiency is greatly improved compared with the CFD method.

关 键 词：卷积神经网络 深度学习 影响函数 非定常流场 机翼 

分 类 号：V221.3[航空宇航科学与技术—飞行器设计]