Scholarlink

<p>Optimization of blade shapes plays a key role in enhancing the aerodynamic efficiency of rotorcraft. However, blade-shape optimization remains challenging due to the strong spanwise variation in local flow conditions experienced by each blade section during rotation. Designers typically employ low-fidelity aerodynamic solvers in rotor blade design due to their computational efficiency. However, these solvers rely on pre-computed airfoil performance lookup tables, making it difficult to treat airfoil shapes as design variables during optimization, thus limiting design flexibility. While surrogate modeling offers a promising solution for real-time generation of airfoil lookup tables, several challenges remain due to the need to account for diverse flow conditions across a broad airfoil design space. To address these challenges, this study proposes an efficient deep-neural-network framework, named <i>Airfoil Brain</i>, which integrates data-driven airfoil parameterization, scalable and reliable surrogate modeling, and an adaptive sampling strategy. Consequently, this framework allows simultaneous consideration of both airfoil and planform parameters in rotor blade-shape optimization. The effectiveness of the proposed framework is demonstrated through a multi-objective aerodynamic optimization of a proprotor.</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

Data-driven surrogate modeling framework for real-time prediction of airfoil lookup tables: application to proprotor optimization

  • 期刊论文

  • 发表日期：

    2026年1月22日

查看全文
  • Yu-Eop Kang,
  • Dawoon Lee,
  • Yoonpyo Hong,
  • Sunwoong Yang,
  • Kwanjung Yee

摘要

Optimization of blade shapes plays a key role in enhancing the aerodynamic efficiency of rotorcraft. However, blade-shape optimization remains challenging due to the strong spanwise variation in local flow conditions experienced by each blade section during rotation. Designers typically employ low-fidelity aerodynamic solvers in rotor blade design due to their computational efficiency. However, these solvers rely on pre-computed airfoil performance lookup tables, making it difficult to treat airfoil shapes as design variables during optimization, thus limiting design flexibility. While surrogate modeling offers a promising solution for real-time generation of airfoil lookup tables, several challenges remain due to the need to account for diverse flow conditions across a broad airfoil design space. To address these challenges, this study proposes an efficient deep-neural-network framework, named Airfoil Brain, which integrates data-driven airfoil parameterization, scalable and reliable surrogate modeling, and an adaptive sampling strategy. Consequently, this framework allows simultaneous consideration of both airfoil and planform parameters in rotor blade-shape optimization. The effectiveness of the proposed framework is demonstrated through a multi-objective aerodynamic optimization of a proprotor.