基于去噪扩散模型的翼型快速反设计研究

程明; 何磊; 蔺佳哲; 黄铭基; 张显才; 赵暾

doi:10.13224/j.cnki.jasp.20240425

基于去噪扩散模型的翼型快速反设计研究

doi: 10.13224/j.cnki.jasp.20240425

中国空气动力研究与发展中心计算空气动力研究所，四川绵阳 621000

详细信息

作者简介:
程明（1995-），男，助理研究员，硕士，从事机器学习与气动建模研究。E-mail：chengming_20@163.com

通讯作者:
何磊（1988-），男，副研究员、硕士生导师，博士，从事机器学习、气动数据建模与智能应用研究。E-mail：helei_email@163.com

中图分类号: V211.3
计量
- 文章访问数: 235
- HTML浏览量: 292
- PDF量: 29
- 被引次数: 0
出版历程
- 收稿日期: 2024-06-27
- 网络出版日期: 2026-01-24

Research on fast inverse design of airfoil based on denoising diffusion model

Computational Aerodynamics Institute，China Aerodynamics Research and Development Center，Mianyang Sichuan 621000，China

摘要

摘要:
为了实现翼型的快速反设计，利用去噪扩散模型强大的数据特征学习与样本生成能力，开展了基于去噪扩散模型的翼型快速反设计方法研究，直接以翼型坐标作为翼型的外形描述方式，实现气动性能条件控制下的翼型快速反设计。首先基于UIUC翼型库构建基础翼型集；以基础翼型集作为输入，训练无条件扩散模型；再基于无条件扩散模型生成大量新翼型，建立补充翼型集，增加数据的多样性；然后使用Xfoil翼型分析软件计算翼型气动性能；以翼型和气动性能作为输入，训练条件扩散模型，最终实现基于此模型的快速翼型反设计。实验结果表明：基于去噪扩散模型的翼型快速反设计方法的单个翼型设计仅需1.25 s，与标签翼型的气动性能偏差基本控制在了6%以内，同时生成翼型具有多样性，为创新设计更加高效的翼型提供了条件。
- 翼型反设计 /
- 翼型坐标 /
- 气动性能 /
- 去噪扩散模型 /
- 生成式模型
Abstract:
A fast inverse design method of airfoil based on denoising diffusion probability model is established, which directly uses coordinates as the description mode of airfoil, and utilizes the powerful data feature learning and sample generation ability of denoising diffusion probability model to realize fast inverse design of airfoil under aerodynamic performance control. The research of this paper is carried out by the following steps: firstly, the basic airfoil data set is constructed based on UIUC airfoil database; using the basic airfoil data set as input, the unconditional diffusion model is trained; then, based on the unconditional diffusion model, a large number of new airfoils are generated, and an extended data set is established to increase the diversity of data; then the aerodynamic performance of airfoil is calculated by Xfoil airfoil analysis software. Taking airfoil and aerodynamic performance as inputs, the conditional diffusion model is trained, and the rapid airfoil reverse design based on this model is finally realized. The experimental results show that the design of a single airfoil based on the denoising diffusion model takes only 1.25 s, and the aerodynamic performance deviation from the tag airfoil is controlled within 6%. At the same time, the generated airfoil shapes are diverse, which provides conditions for more efficient airfoil design.
- inverse design of airfoil /
- airfoil coordinates /
- aerodynamic performance /
- denoising diffusion probability model /
- generative models

HTML

图 1 去噪扩散模型原理

Figure 1. Principle of denoising diffusion model

下载: 全尺寸图片幻灯片

图 2 UDDPM的噪声估计网络结构

Figure 2. Noise estimation network structure of UDDPM

下载: 全尺寸图片幻灯片

图 3 CDDPM噪声估计网络结构

Figure 3. CDDPM noise estimation network structure

下载: 全尺寸图片幻灯片

图 4 本文翼型反设计方法流程

Figure 4. Flow of airfoil reverse design method in this paper

下载: 全尺寸图片幻灯片

图 5 部分翼型空间分布图

Figure 5. Spatial distribution map of partial airfoil

下载: 全尺寸图片幻灯片

图 6 翼型去噪生成过程

Figure 6. Generation process of airfoil denoising

下载: 全尺寸图片幻灯片

图 7 翼型集气动性能分布

Figure 7. Aerodynamic performance distribution of airfoil set

下载: 全尺寸图片幻灯片

图 8 两种方案训练损失曲线

Figure 8. Training loss curves of the two schemes

下载: 全尺寸图片幻灯片

图 9 3组气动指标对应的翼型设计结果

Figure 9. Airfoil design results corresponding to three groups of aerodynamic indexes

下载: 全尺寸图片幻灯片

图 10 设计翼型气动误差分析

Figure 10. Aerodynamic error analysis of design airfoil

下载: 全尺寸图片幻灯片

表 1 计算机配置

Table 1. Computer configuration

项目	内容
CPU	i9-10940x
GPU	RTX3090
操作系统	Windows 11
CUDA	11.3

下载: 导出CSV

表 2 模型误差统计

Table 2. Model error statistics

类型	数据集	L_RMSE	L_MRE/%
方案1	训练集	0.00267	5.64
方案1	测试集	0.00313	6.23
方案2	训练集	0.00194	4.86
方案2	测试集	0.00222	5.14

下载: 导出CSV

表 3 3组气动设计指标

Table 3. Three groups of aerodynamic design indexes

气动设计指标	气动系数	攻角/（°）
气动设计指标	气动系数	−2	−1	0	1	2
设计案例1	C_L	−0.0835	0.0075	0.0839	0.1538	0.2173
设计案例1	C_D	0.0367	0.0367	0.0385	0.0411	0.0448
设计案例2	C_L	−0.0867	0.0081	0.0942	0.1742	0.2485
设计案例2	C_D	0.0324	0.0322	0.0327	0.0338	0.0357
设计案例3	C_L	−0.0845	−0.0092	0.0639	0.1341	0.1984
设计案例3	C_D	0.0337	0.0336	0.0340	0.0350	0.0367

下载: 导出CSV

参考文献(21)

[1]	何磊, 钱炜祺, 刘滔, 等. 基于深度学习的翼型反设计方法[J]. 航空动力学报, 2020, 35(9): 1909-1917. HE Lei, QIAN Weiqi, LIU Tao, et al. Inverse design method of airfoil based on deep learning[J]. Journal of Aerospace Power, 2020, 35(9): 1909-1917. (in Chinese doi: 10.13224/j.cnki.jasp.2020.09.013 HE Lei, QIAN Weiqi, LIU Tao, et al. Inverse design method of airfoil based on deep learning[J]. Journal of Aerospace Power, 2020, 35(9): 1909-1917. (in Chinese) doi: 10.13224/j.cnki.jasp.2020.09.013
[2]	徐华舫. 空气动力学基础[M]. 北京航空学院出版社, 1987. XU Huafang. Fundamentals of aerodynamics[M]. Beijing University of Aeronautics Press, 1987. (in Chinese XU Huafang. Fundamentals of aerodynamics[M]. Beijing University of Aeronautics Press, 1987. (in Chinese)
[3]	BARRETT T R, BRESSLOFF N W, KEANE A J. Airfoil shape design and optimization using multifidelity analysis and embedded inverse design[J]. AIAA Journal, 2006, 44(9): 2051-2060. doi: 10.2514/1.18766
[4]	MASTERS D A, TAYLOR N J, RENDALL T C S, et al. Geometric comparison of aerofoil shape parameterization methods[J]. AIAA Journal, 2017, 55(5): 1575-1589. doi: 10.2514/1.J054943
[5]	高正红, 王超. 飞行器气动外形设计方法研究与进展[J]. 空气动力学学报, 2017, 35(4): 516-528, 454. GAO Zhenghong, WANG Chao. Aerodynamic shape design methods for aircraft: status and trends[J]. Acta Aerodynamica Sinica, 2017, 35(4): 516-528, 454. (in Chinese doi: 10.7638/kqdlxxb-2017.0058 GAO Zhenghong, WANG Chao. Aerodynamic shape design methods for aircraft: status and trends[J]. Acta Aerodynamica Sinica, 2017, 35(4): 516-528, 454. (in Chinese) doi: 10.7638/kqdlxxb-2017.0058
[6]	白俊强, 邱亚松, 华俊. 改进型Gappy POD翼型反设计方法[J]. 航空学报, 2013, 34(4): 762-771. BAI Junqiang, QIU Yasong, HUA Jun. Improved airfoil inverse design method based on gappy POD[J]. Acta Aeronautica et Astronautica Sinica, 2013, 34(4): 762-771. (in Chinese doi: 10.7527/S1000-6893.2013.0135 BAI Junqiang, QIU Yasong, HUA Jun. Improved airfoil inverse design method based on gappy POD[J]. Acta Aeronautica et Astronautica Sinica, 2013, 34(4): 762-771. (in Chinese) doi: 10.7527/S1000-6893.2013.0135
[7]	LI Jiaozan, GAO Zhenghong. Inverse design method of airfoil embedded optimization of pressure distribution [C]//2010 Third International Conference on Information and Computing. Piscataway, US: IEEE, 2010: 101-104.
[8]	LI Xiujuan, LIAO Wenhe, LIU Hao. An inverse method for 3d aerodynamic design of wing shape [C]//2009 IEEE 10th International Conference on Computer-Aided Industrial Design & Conceptual Design. Piscataway, US: IEEE, 2010: 625-630.
[9]	WICKRAMASINGHE U K, CARRESE R, LI Xiaodong. Designing airfoils using a reference point based evolutionary many-objective particle swarm optimization algorithm[C]//IEEE Congress on Evolutionary Computation. Piscataway, US: IEEE, 2010: 1-8.
[10]	刘俊, 宋文萍, 韩忠华, 等. Kriging模型在翼型反设计中的应用研究[J]. 空气动力学学报, 2014, 32(4): 518-526. LIU Jun, SONG Wenping, HAN Zhonghua, et al. Kriging-based airfoil inverse design[J]. Acta Aerodynamica Sinica, 2014, 32(4): 518-526. (in Chinese doi: 10.7638/kqdlxxb-2012.0165 LIU Jun, SONG Wenping, HAN Zhonghua, et al. Kriging-based airfoil inverse design[J]. Acta Aerodynamica Sinica, 2014, 32(4): 518-526. (in Chinese) doi: 10.7638/kqdlxxb-2012.0165
[11]	AHN T, KIM H J, KIM C, et al. Transonic wing design using genetic optimization of wing planform and target pressures[C]//IEEE SMC’99 Conference Proceedings. 1999 IEEE International Conference on Systems, Man, and Cybernetics. Piscataway, US: IEEE, 2002: 469-474.
[12]	SUN Gang, SUN Yanjie, WANG Shuyue. Artificial neural network based inverse design: Airfoils and wings[J]. Aerospace Science and Technology, 2015, 42: 415-428. doi: 10.1016/j.ast.2015.01.030
[13]	王超杰, 何磊, 李川, 等. 基于注意力机制的翼型反设计方法[J]. 航空动力学报, 2025, 40(1): 20230106. WANG Chaojie, HE Lei, LI Chuan, et al. Airfoil reverse design method based on self-attention mechanism[J]. Journal of Aerospace Power, 2025, 40(1): 20230106. (in Chinese doi: 10.13224/j.cnki.jasp.20230106 WANG Chaojie, HE Lei, LI Chuan, et al. Airfoil reverse design method based on self-attention mechanism[J]. Journal of Aerospace Power, 2025, 40(1): 20230106. (in Chinese) doi: 10.13224/j.cnki.jasp.20230106
[14]	吴明雨, 陈志华, 邱志明, 等. 条件生成对抗网络的翼型反设计方法[J]. 宇航学报, 2023, 44(10): 1512-1521. WU Mingyu, CHEN Zhihua, QIU Zhiming, et al. An inverse design method of airfoil using conditional generative adversarial network[J]. Journal of Astronautics, 2023, 44(10): 1512-1521. (in Chinese doi: 10.7527/S1000-6893.2024.31182 WU Mingyu, CHEN Zhihua, QIU Zhiming, et al. An inverse design method of airfoil using conditional generative adversarial network[J]. Journal of Astronautics, 2023, 44(10): 1512-1521. (in Chinese) doi: 10.7527/S1000-6893.2024.31182
[15]	SAMAREH J A. Survey of shape parameterization techniques for high-fidelity multidisciplinary shape optimization[J]. AIAA journal, 2001, 39(5): 877-884. doi: 10.2514/2.1391
[16]	HO J, JAIN A, ABBEEL P. Denoising diffusion probabilistic models[J]. Advances in neural information processing systems, 2020, 33: 6840-6851.
[17]	SONG Jiaming, MENG Chenlin, ERMON S. Denoising diffusion implicit models [EB/OL]. (2022-11-05)[2026-01-20]. https://arxiv.org/abs/2010.02502.
[18]	NICHOL A, DHARIWAL P. Improved Denoising Diffusion Probabilistic Models [EB/OL]. (2021-02-18)[2026-01-20]. https://arxiv.org/abs/2102.09672.
[19]	DHARIWAL P, NICHOL A. Diffusion models beat GANs on image synthesis[EB/OL]. (2021-06-01)[2026-01-20] https://arxiv.org/abs/2105.05233.
[20]	刘泽润, 尹宇飞, 薛文灏, 等. 基于扩散模型的条件引导图像生成综述[J]. 浙江大学学报(理学版), 2023, 50(6): 651-667. LIU Zerun, YIN Yufei, XUE Wenhao, et al. A review of conditional image generation based on diffusion models[J]. Journal of Zhejiang University (Science Edition), 2023, 50(6): 651-667. (in Chinese doi: 10.3785/j.issn.1008-9497.2023.06.001 LIU Zerun, YIN Yufei, XUE Wenhao, et al. A review of conditional image generation based on diffusion models[J]. Journal of Zhejiang University (Science Edition), 2023, 50(6): 651-667. (in Chinese) doi: 10.3785/j.issn.1008-9497.2023.06.001
[21]	HO J, SALIMANS T. Classifier-free diffusion guidance[EB/OL]. (2022-07-26)[2026-01-20] https://arxiv.org/abs/2207.12598.