基于迁移学习的三维飞行器气动力智能预测-2026科协年会论文

田锋; 张一鸣; 胡宁; 覃建秀; 李国良; 冯亦葳; 朱德华; 陈思员; 艾邦成

doi:10.7527/S1000-6893.2026.33658

航空学报 >

0 1 - 0

DOI: https://doi.org/10.7527/S1000-6893.2026.33658

基于迁移学习的三维飞行器气动力智能预测-2026科协年会论文

田锋 ,
张一鸣 ,
胡宁 ,
覃建秀 ,
李国良 ,
冯亦葳 ,
朱德华 ,
陈思员 ,
艾邦成

展开

中国航天空气动力技术研究院

收稿日期: 2026-04-01

修回日期: 2026-05-16

网络出版日期: 2026-05-19

基金资助

XX支持项目

收起

Intelligent Aerodynamic Prediction for Three-Dimensional Aerospace Vehicles based on Transfer Learning

TIAN Feng ,
ZHANG Yi-Ming ,
HU Ning ,
QIN Jian-Xiu ,
LI Guo-Liang ,
FENG Yi-Wei ,
ZHU De-Hua ,
CHEN Si-Yuan ,
AI Bang-Cheng

Expand

Received date: 2026-04-01

Revised date: 2026-05-16

Online published: 2026-05-19

Supported by

XX Support Project

Fold

摘要

本文提出了一种基于迁移学习和误差补偿的点云双分支Transformer气动力预测模型，用于弹箭类三维旋成体飞行器气动力的智能端到端预测。针对三维飞行器高精度仿真数据获取困难、样本数量不足的问题，首先采用点云自注意力主干网络训练大规模低精度数据作为预训练模型，以把握气动力变化的基础特征与趋势；然后采用点云自注意力+交叉注意力分支网络学习小规模高低精度数据之间的残差作为微调模型，以实现低精度数据的误差修正与高低精度数据融合。本文基于工程计算和数值仿真分别构建了低精度和高精度气动力数据集，在该数据集上进行的实验结果表明，本文提出的方法相比于直接使用高精度数据训练的模型平均相对误差降低了74.8%，比直接微调基线模型平均相对误差降低了16.5%，验证了迁移学习和误差修正分支网络相结合的有效性和优势。该方法为飞行器设计阶段复杂飞行器外形气动性能的快速可靠预测提供了有效的解决方案，有望推广至其他工程领域的少样本学习问题。

关键词： 点云; 注意力机制; 迁移学习; 误差补偿; 三维飞行器; 气动性能; 智能预测

本文引用格式

田锋 , 张一鸣 , 胡宁 , 覃建秀 , 李国良 , 冯亦葳 , 朱德华 , 陈思员 , 艾邦成 . 基于迁移学习的三维飞行器气动力智能预测-2026科协年会论文[J]. 航空学报, 0 : 1 -0 . DOI: 10.7527/S1000-6893.2026.33658

Abstract

This paper proposes a point cloud dual-branch Transformer aerodynamic prediction model based on transfer learning and error compensation for intelligent end-to-end prediction of aerodynamic forces of three-dimensional rocket and missile type axisymmetric vehicles. To address the challenges of difficulty in acquiring high-fidelity simulation data and insufficient sample size for three-dimensional vehicles, a point cloud self-attention backbone network is first trained on large-scale low-fidelity data as pre-trained model to capture the fundamental characteristics and trends of aerodynamic variations. Subsequently, a point cloud branch network combined with self-attention and cross-attention is employed to learn the residuals between small-scale low-fidelity and high-fidelity data as a fine-tuning model, enabling error correction of low-fidelity data and fusion of multi-fidelity data. Low-fidelity and high-fidelity aerodynamic datasets are constructed using engineering calculation methods and computational fluid dynamics (CFD) simulations, respectively. Experimental results on these datasets demonstrate that the proposed method reduces the mean relative error by 74.8% as compared to models trained directly with high-fidelity data, and achieves a 16.5% reduction compared to baseline models with direct fine-tuning. These results thoroughly validate the effectiveness of combining transfer learning with error correction. The proposed method provides an effective approach for fast and reliable prediction of aerodynamic performance of complex aircraft configurations during the aircraft design phase, and is expected to be extended to few-shot learning problems in other engineering domains.

Key words： point cloud; attention mechanism; transfer learning; error correction; 3D aerospace vehicles; aerodynamic performance; intelligent prediction

参考文献

[1] 唐志共,朱林阳,向星皓,等. 智能空气动力学若干研究进展及展望[J]. 空气动力学学报, 2023, 41(7): 1-35.
TANG Z G, ZHU L Y, XIANG X H, et al. Some research progress and prospect of Intelligent erodynamics[J]. Acta Aerodynamica Sinica, 2023, 41(7): 1-35(in Chinese).
doi: 10.7638/kqdlxxb-2023.0128
[2] 何开峰, 钱炜祺, 汪清, 等. 数据融合技术在空气动力学研究中的应用[J]. 空气动力学学报, 2014, 32(6): 777-782.
HE Kaifeng, QIAN Weiqi, WANG Qing, et al. Application of datafusion technique in aerodynamics studies[J]. Acta AerodynamicaSinica, 2014, 32(6): 777-782.
doi: 10.7638/kqdlxxb-2014.0072
[3] 李瑛, 颜廷龙. 航空大数据的融合及挖掘技术综述[J]. 航空计算技术, 2020, 50(6): 124-128.
LI Ying, YAN Tinglong. A survey of fusion and mining technologyof aeronautical big data[J]. Aeronautical Computing Technique,2020, 50(6): 124-128.
doi: 1671-654X(2020)06-0124-05
[4] 邓晨, 陈功, 王文正, 等. 基于不确定度和气动模型的气动数据融合算法[J]. 空气动力学学报, 2022, 40(4): 117-123.
DENG Chen, CHEN Gong, WANG Wenzheng, et al. Aerodynamicdata fusion algorithms based on aerodynamic model and uncertainly[J]. Acta Aerodynamica Sinica, 2022, 40(4): 117-123.
doi: 10.7638/kqdlxxb-2020.0151
[5] HE L, QIAN W Q, ZHAO T, et al. Multi-Fidelity Aerodynamic Data Fusion with a Deep Neural Network Modeling Method[J]. Entropy, 2020, 22(9), 1022.
doi: 10.3390/e22091022
[6] 陈昊涵，王霄，闫盼盼，李怀璐，张伟伟．飞行器多舵面气动力智能融合建模方法[J/OL]．哈尔滨工业大学学报. https://link.cnki.net/urlid/23.1235.T.20251106.0939.002
CHEN H H, WANG X, YAN P P, et al. An intelligent fusion modelling method for aerodynamic forces of multi-control-surface aircraft[J/OL]. Journal of Harbin Institute of Technology.
https://link.cnki.net/urlid/23.1235.T.20251106.0939.002
doi:10.11918/202509098
[7] 杨依峰,沈霖,邱恩鹏等. 基于PINN的多源气动数据融合建模方法[J]. 导弹与航天运载技术,2025,413(5):73-81.
YANG Y F, SHEN L, QIU E P, et al. Multi-source Aerodynamic Data Fusion Modeling Based on PINN[J]. Missiles and Space Vehicles, 2025,413(5):73-81.
doi: 10.7654/j.issn.2097-1974.20250510
[8] KOU J, ZHANG W. Multi-fidelity modeling framework for nonlinear unsteady aerodynamics of airfoils[J]. Applied Mathematical Modelling, 2019(76): 832-855.
doi:10.1016/j.apm.2019.06.034
[9] 王旭, 宁晨伽, 王文正, 等. 面向飞行试验的多源气动数据智能融合方法[J]. 空气动力学学报, 2023, 41(2): 12-20.
WANG Xu, NING Chenjia, WANG Wenzheng, et al. Intelligentfusion method of multi-source aerodynamic data for flight tests[J].Acta Aerodynamica Sinica, 2023, 41(2): 12-20.
doi: 10.7638/kqdlxxb-2021.0428
[10] 黄红亮，闫昊，张鲸超，等. 多源多精度数据融合与气动特性智能外推［J］. 力学学报， 2024， 56（9）： 2775-2787.
HUANG H L， YAN H， ZHANG J C， et al. Multisource and multi-fidelity data fusion and intelligent ex?trapolation of aerodynamic characteristics［J］. Chinese Journal of Theoretical and Applied Mechanics， 2024， 56（9）： 2775-2787（ in Chinese）.
doi: 10.6052/0459-1879-24-110
[11] MENG X H, KARNIADAKS G E. A composite neural network that learns from multi-fidelity data: application to function approximation and inverse PDE problems[J]. Journal of Computational Physics, 2020, 401(15): 109020.
doi: 10.1016/j.jcp.2019.109020
[12] LI K, KOU J Q, ZHANG W W. Deep learning for multifidelity aerodynamic distribution modeling from experimental and simulation data[J]. AIAA Journal, 2022, 60(7): 4413-4427.
doi: 10.2514/1.J061330
[13] 颜廷龙, 李瑛, 王凤芹. 基于改进支持度的飞行数据记忆融合滤波算法[J]. 计算机与数字工程, 2023, 401(3): 635-638.
YAN Tinglong, LI Ying, WANG Fengqin. Flight data memory fusion filtering algorithm based on improved support[J]. Computer & Digital Engineering, 2023, 401(3): 635-638.
doi: 10.3969/j.issn.1672-9722.2023.03.018.
[14] Qi C R, Su H, Mo K, et al. PointNet: Deep learning on point sets for 3D classification and segmentation[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 652-660.
doi:10.1109/CVPR.2017.16
[15] Kashefi A, Rempe D, Guibas L J. A point-cloud deep learning framework for prediction of fluid flow fields on irregular geometries[J]. Physics of Fluids, 2021, 33(2).
doi: 10.1063/5.0033376
[16] Qi C R, Yi L, Su H, et al. PointNet++: Deep hierarchical feature learning on point sets in a metric space[J]. Advances in neural information processing systems, 2017, 30.
doi:10.48550/arXiv.1706.02413
[17] Shen Y, Huang W, Wang Z, et al. A deep learning framework for aerodynamic pressure prediction on general three-dimensional configurations[J]. Physics of Fluids, 2023, 35(10).
doi:10.1063/5.0172437
[18] Massegur D, Da Ronch A. Graph convolutional multi-mesh autoencoder for steady transonic aircraft aerodynamics[J]. Machine Learning: Science and Technology, 2024, 5(2): 025006.
doi:10.1088/2632-2153/ad36ad
[19] 冯怡忆, 刘珊珊, 董守鸿, 等. 基于迁移学习的点云机翼气动性能预测[J]. 空气动力学学报, 2025, 43(X): 1-9.
FENG Y Y, LIU S S, DONG S H, et al. Research on aerodynamic performance prediction of point cloud wings based on transfer learning[J]. Acta Aerodynamica Sinica, 2025, 43(X): 1-9(in Chinese). doi: 10.7638/kqdlxxb-2024.0221
[20] 张鹏, 徐丹, 孙国鹏, 等. 面向三维飞行器的气动性能智能预测模型及迁移学习框架研究[J]. 空气动力学学报, 2025, 43(X): 1-11.
ZHANG P, XU D, SUN G P, et al. An intelligent prediction model and transfer learning framework for aerodynamic performance of 3D aircraft[J]. Acta Aerodynamica Sinica, 2025, 43(X): 1-11(in Chinese). doi: 10.7638/kqdlxxb-2025.0049
[21] 王鹏飞，曾丽芳，邵雪明，等．基于预训练微调的机翼气动载荷多源数据融合建模方法［J］．航空学报， 2025， 46（19）：532297．
WANG P F， ZENG L F， SHAO X M， et al. Multi-source data fusion modeling method for aerodynamic load of aircraft wing based on pre-training and fine-tuning[J]. Acta Aeronautica et Astronautica Sinica， 2025, 46(19):532297( in Chinese).
doi：10. 7527/S1000-6893. 2025. 32297
[22] 陈苏, 丁毅, 孙浩, 等. 物理驱动深度学习波动数值模拟方法及应用[J]. 力学学报, 2023, 55(1): 272-282.
CHEN S, DING Y, SUN H, et al. Methods and applications of physical information deep learning in wave numerical simulation[J]. Chinese Journal of Theoretical and Applied Mechanics, 2023, 55(1): 272-282 (in Chinese).
doi:10.6052/0459-1879-22-401
[23] Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, ?. Kaiser, I. Polosukhin, “Attention is all you need,” Advances in neural information processing systems 30 (2017).
doi:10.48550/arXiv.1706.03762
[24] Dosovitskiy, A., Beyer, L., Kolesnikov, A., Weissenborn, D., Zhai, X., Unterthiner, T., Dehghani, M., Minderer, M., Heigold, G., Gelly, S., et al.: An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929 (2020).
doi: 10.48550/arXiv.2010.11929
[25] SOOY T J， SCHMIDT R Z. Aerodynamic predictions, comparisons, and validations using missile DATCOM(97) and aeroprediction 98(AP98)[J]. Journal of Spacecraft and Rockets, 2005, 42(2): 257-265.

Options

文章导航

模态框（Modal）标题

摘要

本文引用格式

Abstract

参考文献