航空学报 > 2025, Vol. 46 Issue (10): 631679-631679   doi: 10.7527/S1000-6893.2025.31679

飞行器设计生成式模型专栏

飞行器生成式模型气动设计研究进展与展望

林杰1,2,3, 唐志共3(), 钱炜祺3, 王岳青2,3, 张鹏2,3, 徐炜遐1, 刘杰4,5   

  1. 1.国防科技大学 计算机学院,长沙 410073
    2.空天飞行空气动力科学与技术全国重点实验室,绵阳 621000
    3.中国空气动力研究与发展中心,绵阳 621000
    4.国防科技大学 高端装备数字化软件重点实验室,长沙 410073
    5.国防科技大学 并行与分布处理重点实验室,长沙 410073
  • 收稿日期:2024-12-18 修回日期:2025-01-13 接受日期:2025-01-17 出版日期:2025-02-06 发布日期:2025-02-06
  • 通讯作者: 唐志共 E-mail:tangzhigong@126.com
  • 基金资助:
    国家重点研发计划(2023YFA1011704);国家自然科学基金(12472300)

Research progress and prospects of aircraft aerodynamic design based on generative models

Jie LIN1,2,3, Zhigong TANG3(), Weiqi QIAN3, Yueqing WANG2,3, Peng ZHANG2,3, Weixia XU1, Jie LIU4,5   

  1. 1.College of Computer Science and Technology,National University of Defense Technology,Changsha 410073,China
    2.State Key Laboratory of Aerodynamics,Mianyang 621000,China
    3.China Aerodynamics Research and Development Center,Mianyang 621000,China
    4.Laboratory of Digitizing Software for Frontier Equipment,National University of Defense Technology,Changsha 410073,China
    5.Science and Technology on Parallel and Distributed Processing Laboratory,National University of Defense Technology,Changsha 410073,China
  • Received:2024-12-18 Revised:2025-01-13 Accepted:2025-01-17 Online:2025-02-06 Published:2025-02-06
  • Contact: Zhigong TANG E-mail:tangzhigong@126.com
  • Supported by:
    National Key Research and Development Program of China(2023YFA1011704);National Natural Science Foundation of China(12472300)

摘要:

生成式模型技术作为深度学习领域发展最为迅速的方向之一,在计算机视觉等领域取得巨大成功,也为空气动力学等科学领域研究提供了新的模式和方法。聚焦生成式模型在飞行器气动设计领域研究进展,系统总结近年来相关研究成果。首先,建立了包含“表示—生成—评估”3个环节的飞行器生成式气动设计框架;其次,针对气动布局表征、生成式气动布局生成和设计质量评估等环节技术发展现状和涉及的关键技术问题分别进行梳理和深入讨论;然后,简要介绍了气动数据构造方法和典型气动数据集,为开展生成式气动设计提供数据支撑;最后,结合气动设计需求和大模型技术在气动设计领域研究趋势,对发展融合型生成式模型架构、构建气动设计大模型和领域智能体、建立生成式气动设计质量综合评估体系、生成式气动设计模型领域知识融入等未来重点发展方向进行展望。

关键词: 气动布局设计, 生成式模型, 气动布局表征, 气动性能评估, 深度学习

Abstract:

As one of the fastest-growing directions in deep learning, the generative model has achieved remarkable success in realms such as computer vision and has also introduced novel paradigms and methodologies for research endeavors within the scientific fields like aerodynamics. This paper focuses on the latest research advancements of generative models in the field of aircraft aerodynamic configuration design, and systematically summarizes the relevant research achievements in recent years. Firstly, a representation-generation-evaluation framework for generative aerodynamic configuration design of aircraft is established. Subsequently, the key technologies and current development progress involved in aerodynamic configuration design are examined and discussed from the perspectives of aerodynamic configuration representation, the development of generative aerodynamic configuration design models, and methods for evaluating design quality. Additionally, a brief overview of aerodynamic data construction methods and typical datasets is provided, serving as a data foundation for generative aerodynamic design. Lastly, the future key development directions in the field of generative aerodynamic configuration design are discussed, including exploration of hybrid generative model architectures, construction of large models and domain-specific agents for aerodynamic design, establishment of a comprehensive evaluation system for generative aerodynamic design quality, and integration of domain knowledge into generative aerodynamic design models.

Key words: aerodynamic configuration design, generative model, aerodynamic configuration representation, aerodynamics performance evaluation, deep learning

中图分类号: