航空学报 > 2025, Vol. 46 Issue (10): 631172-631172   doi: 10.7527/S1000-6893.2025.31172

飞行器设计生成式模型专栏

基于无监督学习的飞行器表面网格平滑方法

王志超1,2, 陈新海1,2(), 邓亮3, 刘杨3, 庞宇飞3, 刘杰1,2   

  1. 1.国防科技大学 高端装备数字化软件湖南省重点实验室,长沙 410073
    2.国防科技大学 并行与分布处理重点实验室,长沙 410073
    3.中国空气动力研究与发展中心,绵阳 621000
  • 收稿日期:2024-09-09 修回日期:2024-12-10 接受日期:2025-01-09 出版日期:2025-02-11 发布日期:2025-02-06
  • 通讯作者: 陈新海 E-mail:chenxinhai16@nudt.edu.cn
  • 基金资助:
    国家自然科学基金(12402349);湖南省自然科学基金(2024JJ6468);国防科技大学青年基金(ZK2023-11);国家重点研发计划(2021YFB0300101)

A surface mesh smoothing method for aircraft based on unsupervised learning

Zhichao WANG1,2, Xinhai CHEN1,2(), Liang DENG3, Yang LIU3, Yufei PANG3, Jie LIU1,2   

  1. 1.Laboratory of Digitizing Software for Frontier Equipment,National University of Defense Technology,Changsha 410073,China
    2.Science and Technology on Parallel and Distributed Processing Laboratory,National University of Defense Technology,Changsha 410073,China
    3.China Aerodynamics Research and Development Center,Mianyang 621000,China
  • Received:2024-09-09 Revised:2024-12-10 Accepted:2025-01-09 Online:2025-02-11 Published:2025-02-06
  • Contact: Xinhai CHEN E-mail:chenxinhai16@nudt.edu.cn
  • Supported by:
    National Natural Science Foundation of China(12402349);Natural Science Foundation of Hunan Province(2024JJ6468);Youth Foundation of the National University of Defense Technology(ZK2023-11);National Key Research and Development Program of China(2021YFB0300101)

摘要:

在面向飞行器设计的数值模拟中,网格平滑方法是提升前处理流程中网格质量、减少模拟误差的重要手段。传统的优化式平滑方法受限于复杂的迭代求解过程,存在内存开销大、计算效率低等问题。为解决该问题,已有的智能化平滑方法采用神经网络拟合平滑过程,能够实现平滑效率、质量的平衡。然而,已有的方法在应用到三维表面网格时通常采用投影操作或有监督学习来保证网格点的贴体性,引入额外的计算或数据生成开销。基于无监督学习技术、局部曲面拟合,搭建了面向飞行器表面网格的智能化平滑代理模型GMSNet3D。模型设计了面向表面网格平滑的无监督损失函数,实现了无须高质量监督数据下的智能训练;模型还创新性地引入局部曲面坐标变换来保证平滑后网格点的贴体性。实验结果证明,GMSNet3D采用的局部曲面坐标变换方法相比已有方法的投影操作实现了13.82倍的加速比;同时,在保证网格平滑质量的同时,GMSNet3D相比传统的优化式平滑方法实现了29.81倍的优化效率提升。

关键词: 飞行器设计, 网格平滑, 局部曲面拟合, 优化式平滑方法, 无监督学习

Abstract:

In numerical simulations for aircraft design, mesh smoothing methods are crucial for enhancing mesh quality in the preprocessing stage and reducing simulation errors. Traditional optimization-based smoothing methods are limited by complex iterative solving processes, leading to high memory consumption and low computational efficiency. To address these issues, existing intelligent smoothing methods use neural networks to learn the smoothing process, achieving a balance between smoothing efficiency and quality. However, when applied to three-dimensional surface meshes, these methods often rely on projection operations or supervised learning to ensure mesh node conformity, which introduces additional computation or data generation overhead. This study develops an intelligent smoothing surrogate model, GMSNet3D, specifically designed for aircraft surface meshes, based on unsupervised learning techniques and local surface fitting. The model uses an unsupervised loss function tailored for surface mesh smoothing, enabling intelligent training without the need for high-quality supervised data. Furthermore, the model innovatively introduces local surface coordinate transformation to ensure the conformity of smoothed mesh nodes. Experimental results demonstrate that the local surface coordinate transformation method used in the GMSNet3D model achieves a speedup of 13.82 times compared to projection operations in existing methods. Additionally, while maintaining mesh smoothing quality, GMSNet3D achieves a 29.81-fold improvement in optimization efficiency compared to traditional optimization-based smoothing methods.

Key words: aircraft design, mesh smoothing, local surface fitting, optimization-based smoothing method, unsupervised learning

中图分类号: