基于深度学习和MOEA/D的高升力翼型气动优化设计  

Research on Aerodynamic Optimization Design of High Lift Airfoil Based on Deep Learning and MOEA/D

作　　者：沈永强 王菡[2] 向纪鑫 李志强[2] SHEN Yongqiang;WANG Han;XIANG Jixin;LI Zhiqiang(College of Mechanical and Vehicle Engineering,Taiyuan University of Technology,Taiyuan 030024,China;College of Aeronautics and Astronautic,Taiyuan University of Technology,Taiyuan 030024,China)

机构地区：[1]太原理工大学机械与运载工程学院,太原030024 [2]太原理工大学航空航天学院,太原030024

出　　处：《太原理工大学学报》2024年第4期660-669,共10页Journal of Taiyuan University of Technology

基　　金：国家自然科学基金青年项目资助项目(12102290);山西省关键核心技术和共性技术研发攻关专项(2020XXX017)。

摘　　要：【目的】针对翼型优化气动参数间性能冲突问题,提出一种基于深度学习和MOEA/D相结合的混合优化模型,将卷积神经网络、遗传算法内插到MOEA/D框架之中,用于权衡各目标函数之间的相关性和复杂度。【方法】首先,尝试将深度学习方法作为常规流体力学分析方法的补充,建立高可信度的翼型气动特性CNN响应预测模型,用于快速评估翼型的气动参数;然后将响应模型、遗传算子内插到MOEA/D框架之中,构建基于MOEA/D的多目标混合优化模型,并以NACA某5位数高升力翼型巡航工况下的升阻比和力矩系数为优化目标,对其进行测试。【结果】结果表明:利用该混合模型可以高效地在设计空间内获得连续的Pareto前沿解,最后对获得的Pareto前沿面翼型进行气动性能和流场结构分析,研究不同翼型构型在前沿面的分布规律,进一步指导设计者在翼型选型中挖掘潜在的基础翼型。【Purposes】Aiming at the performance conflict between optimization parameters in prior optimization method,a hybrid optimization model based on MOEA/D is proposed,which integrates CNN and genetic algorithm into MOEA/D framework to balance the correlation and complexity between various objective functions.【Methods】First,the deep learning method is used as a supplement to the conventional fluid mechanics analysis method to establish a highly reliable CNN response prediction model for airfoil aerodynamic characteristics,which can be used to quickly evaluate the aerodynamic parameters of airfoil.Then,the response model and genetic operator are interpolated into the MOEA/D framework to construct a multi-objective hybrid optimization model based on MOEA/D.And the lift drag ratio and moment coefficient of a NACA high lift airfoil under cruise condition are taken as the optimization objectives for testing.Finally,through the analysis of aerodynamic performance and flow field structure of the airfoil on the Pareto front,the distribution law of different airfoil configurations on the front is studied,which further guides the designer to explore the potential basic airfoil in the airfoil selection.

关 键 词：气动优化 混合方法 MOEA/D CNN CFD 

分 类 号：V211.3[航空宇航科学与技术—航空宇航推进理论与工程]