陈志强,李翔,翟朝阳,等.基于改进PointNet++算法可变后掠翼飞行器表面压力分布快速预测[J].河南理工大学学报(自然科学版)，doi:10.16186/j.cnki.1673-9787.2025090006.

CHEN Z Q,LI X,ZHAI C Y,et al．Rapid prediction of surface pressure distribution on variable-sweep wing aircraft using improved PointNet++ algorithm［J］.Journal of Henan Polytechnic University( Natural Science) ，doi:10.16186/j.cnki.1673-9787.2025090006.

基于改进PointNet++算法可变后掠翼飞行器表面压力分布快速预测(网络首发)

陈志强，李翔，翟朝阳，王晓璐，张昕喆

（郑州航空工业管理学院 航空宇航学院,河南 郑州 450000）

摘要: 目的 新一代可变后掠翼飞行器凭借其独特的动态调整后掠角功能，能够灵活适应多样化的飞行任务需求，通过在不同飞行条件下改变机翼后掠角以优化气动性能。然而，后掠角的动态变化引发了气动外形与流场之间复杂的耦合特性，这极大地增加了设计周期的复杂性和不确定性。气动设计作为飞行器设计的核心环节，其关键在于快速获取全飞行包线内的压力分布数据，以支持设计方案的快速迭代与优化。因此，本研究旨在构建一种高效、准确的方法，实现可变后掠翼飞行器表面压力分布的快速预测，以缩短设计周期，提高设计效率。方法 本研究基于Fluent动网格技术，自建了涵盖多种后掠角状态的气动数据库，该数据库包含了丰富的表面压力分布数据。进一步，结合先进的PointNet++模型，构建了从动态外形到表面压力场分布的智能映射模型。该模型通过学习气动数据库中的数据特征，实现了对可变后掠翼飞行器表面压力分布的秒级预测。同时，为提升预测精度，本研究对CFD计算的表面压力数据进行了随机抽样与分组处理，有效减少了数据噪声，提高了模型的泛化能力。结果 经过大量测试验证，本研究构建的智能映射模型在预测可变后掠翼飞行器表面压力分布方面表现出色。与CFD计算的集中应力数据相比，升力系数的预测误差被控制在8%以内，显示出良好的准确性。结论 本研究提出的基于Fluent动网格技术与PointNet++模型的智能映射方法，为可变后掠翼飞行器的气动设计提供了一种高效、准确的解决方案，具备一定的应用价值。

关键词: 表面压力预测；数据驱动；神经网络；快速预测；变体飞行器

doi:10.16186/j.cnki.1673-9787.2025090006

基金项目: 国家自然科学基金资助项目（12302027，52206059）；郑州航空工业管理学院研究生教育创新计划基金资助项目（2025CX97）

收稿日期：2025-09-05

修回日期：2026-02-10

网络首发日期：2026-04-16

Rapid prediction of surface pressure distribution on variable-sweep wing aircraft using improved PointNet++ algorithm

Chen Zhiqiang, Li Xiang, Zhai Chaoyang, Wang Xiaolu, Zhang Xinzhe

( School of Aeronautics and Astronautics, Zhengzhou University of Aeronautics, Zhengzhou 450000, Henan, China)

Abstract: Objective The new - generation variable - sweep wing aircraft, featuring the unique capability of dynamically adjusting the sweep angle, can flexibly adapt to a wide range of flight mission requirements. Its aerodynamic performance was optimized by altering the wing sweep angle under different flight conditions. However, the dynamic changes in the sweep angle induced complex coupling characteristics between the aerodynamic shape and the flow field, which significantly increased the complexity and uncertainty of the design cycle. As the core aspect of aircraft design, aerodynamic design relied on the rapid acquisition of pressure distribution data across the entire flight envelope to support the rapid iteration and optimization of design schemes. Therefore, this study aimed to develop an efficient and accurate method for the fast prediction of surface pressure distribution on variable - sweep wing aircraft, so as to shorten the design cycle and enhance design efficiency. Methods In this study, a self - built aerodynamic database covering multiple sweep angle states was constructed based on Fluent's dynamic mesh technology. This database contained a wealth of surface pressure distribution data. Furthermore, by integrating the advanced PointNet++ model, an intelligent mapping model from dynamic shapes to surface pressure field distributions was established. This model was trained to learn the data characteristics within the aerodynamic database, enabling second - level predictions of surface pressure distribution on variable - sweep wing aircraft. Meanwhile, to improve prediction accuracy, random sampling and grouping techniques were applied to the surface pressure data computed by CFD, which effectively reduced data noise and enhanced the model's generalization ability. Results Extensive testing and validation were conducted, and it was demonstrated that the intelligent mapping model developed in this study performed excellently in predicting the surface pressure distribution of variable - sweep wing aircraft. Compared with the concentrated stress data computed by CFD, the prediction error of the lift coefficient was controlled within 8%, indicating good accuracy.ConclusionThe intelligent mapping method proposed in this study, which was based on Fluent's dynamic mesh technology and the PointNet++ model, provided an efficient and accurate solution for the aerodynamic design of variable - sweep wing aircraft and possessed certain application value.

Key words: surface pressure prediction; data-driven; neural network; rapid prediction; morphing aircraft