航空学报 > 2026, Vol. 47 Issue (16): 132983-132983   doi: 10.7527/S1000-6893.2025.32983

基于多任务学习的翼型结冰与气动特性预测

辛惠竹1,2, 李晨3, 桑为民1,2(), 侯靓1,2   

  1. 1.西北工业大学 航空学院,西安 710072
    2.飞行器基础布局全国重点实验室,西安 710072
    3.上海卫星工程研究所,上海 201109
  • 收稿日期:2025-10-28 修回日期:2025-11-17 接受日期:2025-12-15 出版日期:2025-12-29 发布日期:2025-12-29
  • 通讯作者: 桑为民 E-mail:sangweimin@nwpu.edu.cn
  • 基金资助:
    结冰与防除冰重点实验室开放课题(IADL20230302);国家自然科学基金(12572270);西北工业大学“1-0”重大工程科学问题项目(G2024KY0613)

Airfoil icing and aerodynamic characteristics prediction based on multi-task learning

Huizhu XIN1,2, Chen LI3, Weimin SANG1,2(), Liang HOU1,2   

  1. 1.School of Aeronautics,Northwestern Polytechnical University,Xi’an 710072,China
    2.National Key Laboratory of Aircraft Configuration Design,Xi’an 710072,China
    3.Shanghai Satellite Engineering Research Institute,Shanghai 201109,China
  • Received:2025-10-28 Revised:2025-11-17 Accepted:2025-12-15 Online:2025-12-29 Published:2025-12-29
  • Contact: Weimin SANG E-mail:sangweimin@nwpu.edu.cn
  • Supported by:
    Open Fund of Key Laboratory of Icing and Anti/De-icing(IADL20230302);National Natural Science Foundation of China(12572270);The “1-0” Major Engineering Science Problem Project of Northwestern Polytechnical University(G2024KY0613)

摘要:

针对翼型结冰预测中单任务学习精度有限且难以同时获得结冰形态和气动特性的问题,提出一种基于多任务学习的翼型结冰预测框架MTAF(Multi Task Airfoil Former)。该框架采用编码器-解码器架构,通过共享特征编码器提取翼型几何与飞行条件的深层表征,利用任务关联模块实现特征融合与任务交互,由专门的预测头分别完成结冰预测和气动系数预测。对比了Transformer和ConvNeXt两种骨干网络在该框架中的性能表现。基于625个NACA 2412翼型样本进行训练和测试,结果表明Transformer架构在结冰预测任务中达99.79%的像素准确率和0.990 6的决定系数,优于ConvNeXt的99.50%和0.977 2;在气动系数预测方面,Transformer的升力系数决定系数达0.975 4,阻力系数决定系数达0.987 4,与ConvNeXt的0.973 3和0.990 1相当,两种架构各有优势。在此基础上,进一步利用100组未参与训练的外部样本进行了泛化性验证,从像素和几何两个层面进行的对比分析表明MTAF框架在不同结冰类型和多种飞行工况下均保持较高精度和良好泛化能力,为翼型结冰工程应用提供了高效可靠的预测方法。

关键词: 飞机结冰, 图像预测, 气动特性, Transformer, 卷积神经网络

Abstract:

To address the problem of limited accuracy in single-task learning for airfoil icing prediction and the difficulty in simultane-ously obtaining icing morphology and aerodynamic characteristics, a multi-task learning based airfoil icing prediction framework called MTAF (Multi-Task Airfoil Former) is proposed. The framework employs an encoder-decoder architecture that extracts deep representations of airfoil geometry and flight conditions through a shared feature encoder, realizes feature fusion and task interaction via a task correlation module, and accomplishes icing prediction and aerodynamic coefficient prediction respectively through dedicated prediction heads. The performance of Transformer and ConvNeXt backbone net-works within this framework is comparatively studied. Based on training and testing with 625 NACA 2412 airfoil samples, results demonstrate that the Transformer architecture achieves pixel accuracy of 99.79% and coefficient of determination for 0.990 6 in icing prediction tasks, significantly outperforming 99.50% and 0.977 2 of ConvNeXt. For aerodynamic coefficient prediction, Transformer achieves coefficient of determination values of 0.975 4 for lift coefficient and 0.987 4 for drag coefficient, which are comparable to 0.973 3 and 0.990 1 of ConvNeXt, respectively, with each architecture showing distinct advantages. The MTAF framework maintains stable prediction performance across different icing types and various flight conditions, providing a high-precision prediction tool for airfoil icing engineering applications.

Key words: aircraft icing, image prediction, aerodynamic characteristics, Transformer, convolutional neural network

中图分类号: