Please wait a minute...
工程设计学报
Chinese Journal of Engineering Design  2024, Vol. 31 Issue (3): 292-300    DOI: 10.3785/j.issn.1006-754X.2024.03.143
Theory and Method of Mechanical Design     
Deep learning-based method for parametrized modeling of airfoil
Jianxiong SHEN1(),Yingyuan LIU1(),Leqin WANG2
1.College of Information, Mechanical and Electrical Engineering, Shanghai Normal University, Shanghai 201400, China
2.College of Energy Engineering, Zhejiang University, Hangzhou 310027, China
Download: HTML     PDF(5118KB)
Export: BibTeX | EndNote (RIS)      

Abstract  

In order to solve the problems of low efficiency and heavy computational workload during the optimization design process in the existing airfoil geometric parametrized modeling methods, a deep learning-based airfoil parametrized modeling method was put forward. In this method, the two-dimensional airfoil images converted from coordinate points of airfoil upper and lower surfaces in the airfoil database of the University of Illinois at Urbana-Champaign (UIUC) were taken as the input. Firstly, the convolution operations were used to extract geometric features of a large amount of airfoil images. Then, the extracted geometric features were classified and compressed by multi-layer perceptron, and the airfoil shape was compressed into several simplified fitting parameters. Finally, the airfoil image was restored and the coordinates of points on the upper and lower surfaces of airfoil were output by a decoder. On this basis, the influence of the number of fitting parameters on the geometric accuracy of airfoil was discussed, and a convolutional neural network (CNN) structure with six fitting parameters was determined. At the same time, the fitting accuracy of the proposed method was verified by the computational fluid dynamics numerical simulation. Finally, the visual airfoil geometry design software was developed to adjust and modify the fitting parameters, and the influence law of each fitting parameter on the airfoil shape was summarized. The results indicated that all the six fitting parameters had a global impact on the airfoil shape, and the new airfoil design space could be obtained by adjusting the six fitting parameters individually or jointly. This research results can provide technical support and theoretical guidance for airfoil optimization design.

Key words： airfoil parameterization      geometric feature      deep learning      convolutional neural network     
Received: 31 March 2023      Published: 27 June 2024
CLC:  TH 122  
Corresponding Authors: Yingyuan LIU     E-mail: 1000511952@smail.shnu.edu.cn;yyliu@shnu.edu.cn
Service
E-mail this article
Add to my bookshelf
Add to citation manager
E-mail Alert
RSS
Articles by authors
Jianxiong SHEN
Yingyuan LIU
Leqin WANG
Cite this article:

Jianxiong SHEN,Yingyuan LIU,Leqin WANG. Deep learning-based method for parametrized modeling of airfoil. Chinese Journal of Engineering Design, 2024, 31(3): 292-300.

URL:

https://www.zjujournals.com/gcsjxb/10.3785/j.issn.1006-754X.2024.03.143     OR     https://www.zjujournals.com/gcsjxb/Y2024/V31/I3/292


基于深度学习的翼型参数化建模方法

为解决现有翼型几何参数化描述方法优化设计效率低、计算工作量大的问题，提出了一种基于深度学习的翼型参数化建模方法。该方法以伊利诺伊大学厄巴纳-香槟分校（University of Illinois at Urbana-Champaign, UIUC）翼型数据库中翼型上下表面坐标点转化的翼型二维图像作为输入，首先使用卷积运算提取大量翼型图像的几何特征，然后通过多层感知机对提取的几何特征进行分类和压缩，将翼型形状压缩成若干个简化的拟合参数，最后通过解码器恢复翼型图像并输出翼型上下表面的点坐标。在此基础上，探讨了拟合参数数量对翼型几何精度的影响，确定了含6个拟合参数的卷积神经网络（convolutional neural network, CNN）结构，并基于计算流体力学数值仿真验证了所提出方法的拟合精度。最后，开发了可视化翼型几何设计软件，实现了拟合参数的调整与修正，并分析了各拟合参数对翼型形状的影响规律。结果表明，6个拟合参数均会对翼型形状产生全局影响，单独或联合调整6个拟合参数可获得新的翼型设计空间。研究结果可为翼型的优化设计提供技术支持与理论参考。


关键词： 翼型参数化,  几何特征,  深度学习,  卷积神经网络 
Fig.1 Generation schematic of airfoil data set
Fig.2 Structure and main parameters of airfoil parametrized model based on CNN
Fig.3 Influence of number of fitting parameters on fitting accuracy of NACA0021 airfoil
Fig.4 Influence of fitting parameter adjustment on airfoil shape under different number of parameters
Fig.5 Fitting effect of different airfoils by CNN-6 model
Fig.6 Fitting error of upper and lower surfaces of different airfoils
Fig.7 HQ109 airfoil water tunnel simulation model
Fig.8 Comparison of pressure cloud map between original airfoil and fitted airfoil
参数HQ109翼型RAE5214翼型
原始翼型拟合翼型相对误差/%原始翼型拟合翼型相对误差/%
升力系数0.980.991.021.081.052.78
阻力系数0.0170.01700.0180.0195.56
Table 1 Comparison of lift coefficient and drag coefficient between original airfoil and fitted airfoil
Fig.9 Comparison of surface pressure coefficient between original airfoil and fitted airfoil
Fig.10 Influence of single fitting parameter adjustment in CNN-6 model on RAE2822 airfoil shape
Fig.11 Visual airfoil geometry design software
[1]   高永，朱飞翔，李冰，等.改进CST方法在翼型优化设计中的应用［J］.海军航空大学学报，2017，32（5）：426-430. doi：10.7682/j.issn.1673-1522.2017.05.002
GAO Y， ZHU F X， LI B， et al. Application of improved CST parametric method in airfoil design［J］. Journal of Naval Aviation University， 2017， 32（5）： 426-430.
doi: 10.7682/j.issn.1673-1522.2017.05.002
[2]   SALUNKE N P， JUNED A R A， CHANNIWALA S A. Airfoil parameterization techniques： a review［J］. American Journal of Mechanical Engineering， 2014， 2（4）： 99-102.
[3]   VISWANATH A， FORRESTER A I J， KEANE A J. Constrained design optimization using generative topographic mapping［J］. AIAA Journal， 2014， 52（5）： 1010-1023.
[4]   QIU Y， BAI J， LIU N， et al. Global aerodynamic design optimization based on data dimensionality reduction［J］. Chinese Journal of Aeronautics， 2018， 31（4）： 643-659.
[5]   GREY Z J， CONSTANTINE P G. Active subspaces of airfoil shape parameterizations［J］. AIAA Journal， 2018， 56（5）： 2003-2017.
[6]   CHEN W， RAMAMURTHY A. Deep generative model for efficient 3D airfoil parameterization and generation［C］// AIAA Scitech 2021 Forum. Nashville， Jan. 11-15 & 19-21， 2021.
[7]   吴则良，叶建川，王江，等.基于深度自动编码器神经网络的飞行器翼型参数降维与优化设计［J］.兵工学报，2022，43（6）：1326-1336.
WU Z L， YE J C， WANG J， et al. Parameter dimensionality reduction and optimal design of aircraft airfoil based on deep autoencoder neural network［J］. Acta Armamentarii， 2022， 43（6）： 1326-1336.
[8]   王建军，高正红.HicksHenne翼型参数化方法分析及改进［J］.航空计算技术，2010，40（4）：46-49. doi：10.3969/j.issn.1671-654X.2010.04.013
WANG J J， GAO Z H. Analysis and improvement of HicksHenne airfoil parameterization method［J］. Aeronautical Computing Technique， 2010， 40（4）： 46-49.
doi: 10.3969/j.issn.1671-654X.2010.04.013
[9]   戴雅雪，应培，王波，等.带下表面射流的超临界翼型优化［J/OL］.工程力学：1-9（2023-02-27）［2024-03-16］. .
DAI Y X， YING P， WANG B， et al. Optimization of supercritical airfoil with jet on the lower surface of trailing edge［J/OL］. Engineering Mechanics： 1-9 （2023-02-27） ［2024-03-16］. .
[10]   ZHANG Z K， DE GASPARI A， RICCI S， et al. Gradient-based aerodynamic optimization of an airfoil with morphing leading and trailing edges［J］. Applied Sciences， 2021， 11（4）： 1929.
[11]   余彬，贺翔，邓浩.基于CST法的新能源无人机翼型优化设计［J］.航空科学技术，2019，30（10）：24-29.
YU B， HE X， DENG H. Optimization design of airfoil on new energy UAV based on CST method［J］. Aeronautical Science & Technology， 2019， 30（10）： 24-29.
[12]   WU X J， ZHANG W W， PENG X H， et al. Benchmark aerodynamic shape optimization with the POD-based CST airfoil parametric method［J］. Aerospace Science and Technology， 2019， 84： 632-640.
[13]   SEKAR V， JIANG Q， SHU C， et al. Fast flow field prediction over airfoils using deep learning approach［J］. Physics of Fluids， 2019， 31（5）： 057103.
[14]   BEN F H， BOUGUEZZI S， SOUANI C. Face recognition in unconstrained environment with CNN［J］. The Visual Computer， 2021， 37： 217-226.
[15]   NARANJO T J， MORA M， HERNÁNDEZ G R， et al. A review of convolutional neural network applied to fruit image processing［J］. Applied Sciences， 2020， 10（10）： 3443.
[16]   CHEN Y P， DAI X Y， LIU M C， et al. Dynamic ReLU［C］//16th European Conference on Computer Vision （ECCV 2020. Glasgow， Aug. 23-28， 2020.
[17]   LILLICRAP T P， SANTORO A， MARRIS L， et al. Backpropagation and the brain［J］. Nature Reviews Neuroscience， 2020， 21（6）： 335-346.
[18]   KINGMA D P， BA J L. Adam： a method for stochastic optimization［C］//3rd International Conference on Learning Representations. San Diego， CA， May 7-9， 2015.
[19]   WANG B， LIU Z， LI H， et al. On the numerical simulations of vortical cavitating flows around various hydrofoils［J］. Journal of Hydrodynamics， 2017， 29（6）： 926-938.
[1] SHANG Zhi-wu, ZHOU Shi-qi. Research on micropipetting technology based on image monitoring[J]. Chinese Journal of Engineering Design, 2021, 28(4): 495-503.
[2] WANG An-bang, SUN Wen-bin, DUAN Guo-lin. Research on intelligent method of manufacturing and processing equipment based on digital twin and deep learning technology[J]. Chinese Journal of Engineering Design, 2019, 26(6): 666-674.
[3] LU Wan-jie, FU Hua, ZHAO Hong-rui. Equipment recognition of mining patrol robot based on deep learning algorithm[J]. Chinese Journal of Engineering Design, 2019, 26(5): 527-533.