Scholarlink

<p>Generative design is transforming aerodynamic geometry optimization by enabling the exploration of design spaces beyond conventional direct or inverse paradigms. However, existing reviews seldom synthesize how generative priors integrate with physics-based fidelity and multidisciplinary constraints. This paper provides a comprehensive survey of recent advances in direct, inverse, and generative aerodynamic design. We categorize representative generative models—including conditional diffusion, operator learning, and multi-fidelity optimization—and highlight their coupling with physics-informed surrogates and feasibility control strategies. A unified four-stage pipeline is proposed to summarize the end-to-end process of generation, evaluation, optimization, and validation across aerodynamic applications. The survey further discusses challenges in scalability, data efficiency, and physical consistency, offering actionable recommendations for future research. By consolidating over one hundred recent studies, this review bridges the gap between generative machine learning and aerodynamic design practice, aiming to guide practitioners toward robust, physics-consistent generative frameworks.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Comprehensive Review of the State of the Art in Generative Models for Aerodynamic Configuration Design of Aircraft

  • 期刊论文

  • 发表日期：

    2026年3月6日

查看全文
  • Huan Zhao,
  • Chunyun Li,
  • Xiaomeng Zhang,
  • Youyong Luo,
  • Bowen Shu,
  • Zhiyuan Gong,
  • Yongbo Li

摘要

Generative design is transforming aerodynamic geometry optimization by enabling the exploration of design spaces beyond conventional direct or inverse paradigms. However, existing reviews seldom synthesize how generative priors integrate with physics-based fidelity and multidisciplinary constraints. This paper provides a comprehensive survey of recent advances in direct, inverse, and generative aerodynamic design. We categorize representative generative models—including conditional diffusion, operator learning, and multi-fidelity optimization—and highlight their coupling with physics-informed surrogates and feasibility control strategies. A unified four-stage pipeline is proposed to summarize the end-to-end process of generation, evaluation, optimization, and validation across aerodynamic applications. The survey further discusses challenges in scalability, data efficiency, and physical consistency, offering actionable recommendations for future research. By consolidating over one hundred recent studies, this review bridges the gap between generative machine learning and aerodynamic design practice, aiming to guide practitioners toward robust, physics-consistent generative frameworks.