航空学报 > 2025, Vol. 46 Issue (10): 631194-631194   doi: 10.7527/S1000-6893.2024.31194

飞行器设计生成式模型专栏

生成式模型赋能飞行器技术应用研究进展与展望

陈树生1,2, 贾苜梁1,2, 林家豪1,2, 金世轶1,2, 高正红1,2, 王岳青3,4(), 马志强5, 李铮6, 段辰龙7, 李佳伟8   

  1. 1.西北工业大学 航空学院,西安 710072
    2.飞行器基础布局全国重点实验室,西安 710072
    3.中国空气动力研究与发展中心 计算空气动力研究所,绵阳 621000
    4.空天飞行空气动力科学与技术全国重点实验室,绵阳 621000
    5.西北工业大学 航天学院,西安 710072
    6.中国运载火箭技术研究院 空间物理重点实验室,北京 100076
    7.中国航空研究院,北京 100029
    8.沈阳飞机设计研究所 扬州协同创新研究院有限公司,扬州 225000
  • 收稿日期:2024-09-12 修回日期:2024-10-09 接受日期:2024-10-25 出版日期:2024-11-19 发布日期:2024-11-18
  • 通讯作者: 王岳青 E-mail:yqwang2013@163.com
  • 基金资助:
    国家自然科学基金(92371109);民机科研项目;西北工业大学“1-0”重大工程科学问题项目(G2024KY0613)

Empowering aircraft technology applications with generative models: Research progress and prospects

Shusheng CHEN1,2, Muliang JIA1,2, Jiahao LIN1,2, Shiyi JIN1,2, Zhenghong GAO1,2, Yueqing WANG3,4(), Zhiqiang MA5, Zheng LI6, Chenlong DUAN7, Jiawei LI8   

  1. 1.School of Aeronautics,Northwestern Polytechnical University,Xi’an 710072,China
    2.National Key Laboratory of Aircraft Configuration Design,Xi’an 710072,China
    3.Computational Aerodynamics Institute,China Aerodynamics Research and Development Center,Mianyang 621000,China
    4.State Key Laboratory of Aerodynamics,Mianyang 621000,China
    5.School of Astronautics,Northwestern Polytechnical University,Xi’an 710072,China
    6.Science and Technology on Space Physics Laboratory,China Academy of Launch Vehicle Technology,Beijing 100076,China
    7.Chinese Aeronautical Establishment,Beijing 100029,China
    8.Yangzhou Collaborative Innovation Research Institute Co. Ltd,Shenyang Aircraft Design and Research Institute,Yangzhou 225000,China
  • Received:2024-09-12 Revised:2024-10-09 Accepted:2024-10-25 Online:2024-11-19 Published:2024-11-18
  • Contact: Yueqing WANG E-mail:yqwang2013@163.com
  • Supported by:
    National Natural Science Foundation of China(92371109);Civil Aircraft Research Project;“1-0”Major Engineering Science Problem Project of Northwestern Polytechnical University(G2024KY0613)

摘要:

在自然语言处理和计算机视觉领域取得颠覆性应用的生成式模型正成为数智化的新型技术基座,是未来驱动飞行器技术智能化发展的重要引擎。综述了生成式模型赋能飞行器技术应用进展。首先,总结了生成式模型架构的发展历程,详细介绍了变分自编码器、生成对抗网络、扩散模型、Transformer等基本原理架构和改进方向。其次,归纳了生成式模型在飞行器空气动力学、航迹预测和目标检测等领域的典型应用和变革情况;关注了参数化建模、气动预测模型、反设计等飞行器空气动力设计关键技术的发展趋势;探讨了实时航迹预测、完整航迹预测、协同航迹预测和预测误差补偿的智能实现方法;从现有目标检测方法改进角度分析了生成式模型在多尺度融合、超分辨率增强和数据增强中的作用。最后,从模型方法和应用场景拓展角度展望了生成式模型赋能飞行器技术未来的研究方向,针对构建可解释的通用大模型和推动垂直领域应用等方面提出了发展建议。

关键词: 生成式人工智能, 飞行器技术, 空气动力设计, 航迹预测, 目标检测

Abstract:

Generative models, which have achieved disruptive applications in the fields of natural language processing and computer vision, are becoming the cornerstone of digital intelligence technologies, serving as a crucial engine driving the future development of intelligent aircraft technology. This paper reviews the application progress of aircraft technologies empowered with generative models. Firstly, the development history of generative model architectures is summarized. Detailed introduction to the fundamental principles and improvement directions of variational autoencoders, generative adversarial networks, diffusion models, and Transformers is provided. Secondly, typical applications and transformative impacts of generative models in aircraft aerodynamics, trajectory prediction, and target detection are generalized, with a focus on the development trends in key technologies of aircraft aerodynamic design, including parameterized modeling, aerodynamic prediction model and inverse design.Intelligent implementation methods of real-time trajectory prediction, complete trajectory prediction, collaborative trajectory prediction and prediction error compensation are studied. From the perspective of improving existing target detection methods, the roles of generative models in multi-scale fusion, super-resolution enhancement and data enhancement are analyzed. Finally, we propose future research directions for aircraft technologies empowered with generative models from the perspectives of model method and application scenario expansion. Development suggestions are proposed for building interpretable general models and promoting vertical domain applications.

Key words: generative artificial intelligence, aircraft technology, aerodynamic design, trajectory prediction, object detection

中图分类号: