电加热防冰系统是防止飞机结冰，保障飞行安全的关键技术之一，其防冰效率直接受加热元件的功率影响，开展电加热防冰系统优化设计的相关研究，对降低防冰系统能耗，提高系统的安全性有着重要意义。针对传统功率分布优化设计中，需多次计算稳态结冰、水膜和温度分布数据，导致耗时长，优化效率低等问题，本文提出了基于改进渐进分层提取网络的电加热防冰功率分布优化设计方法，利用多任务的联合学习能力，提高了深度学习模型的预测精度。该方法采用本征正交分解对数值计算得到的机翼表面水膜、结冰和温度分布数据进行降维，以功率分布参数作为输入，降维后的模态拟合系数作为输出，以改进的渐进式分层提取（Progressive Layered Extraction, PLE）多任务网络为基础，构建多任务学习模型。该模型对门控网络进行改进，显式添加独有任务专家信息，并在共享专家输出端引入注意力机制，增强模型对特定信息的提取能力，实验结果表明，改进的多任务学习模型在水膜、结冰和温度分布的预测任务中均表现出较高的准确度，在测试集上的均方根误差分别为0.0458、0.0848和0.7149。最后，结合遗传算法对电加热防冰系统功率分布进行单目标优化设计，以最小化总功率为目标，设定约束条件求解出较优的功率分布，与基准结果相比能耗下降约34.3%。数值计算验证表明，优化后的功率分布能在降低能耗的同时满足防冰要求。
Electric anti-icing systems are one of the key technologies for preventing aircraft icing and ensuring flight safety, with their anti-icing efficiency being directly influenced by the power distribution of heating elements. Research on the optimal design of such systems is of great significance for reducing energy consumption and enhancing system safety. To address the limitations of traditional power distribution optimization—namely, the requirement of repeated numerical simulations of steady-state ice accretion, water film, and temperature distributions, which result in high computational cost and low optimization efficiency—this study proposes an optimized power distribution design method for electric anti-icing systems based on an improved Progressive Layered Extraction (PLE) network. A multi-task learning framework is constructed by employing Proper Orthogonal Decomposition (POD) to reduce the dimensionality of numerical simulation data, including water film, ice accretion, and temperature distributions on the wing surface. The power distribution parameters are used as inputs, and the POD modal coefficients as outputs. Based on the PLE architecture, the network is enhanced by explicitly integrating task-specific expert information into the gating mechanism and introducing an attention mechanism at the shared expert output, thereby improving the model’s capability to extract task-relevant features. Experimental results demonstrate that the improved multi-task learning model achieves high accuracy across all three prediction tasks, with root mean square errors (RMSE) of 0.0458 for water film, 0.0848 for ice accretion, and 0.7149 for temperature on the test set. Finally, a single-objective optimization of the power distribution is performed using a genetic algorithm, aiming to minimize total power consumption under a set of physical constraints. Compared to the reference scheme, the optimized power distribution achieves approximately 34.3% reduction in energy consumption. Numerical simulations further verify that the optimized scheme meets anti-icing requirements while significantly reducing power demand.
[1] ZHANG Y, ZHANG Y, LUO G, 等. Research Progress of Aircraft Icing Hazard and Ice Wind Tunnel Test Technology[C/OL]//2021 12th International Conference on Mechanical and Aerospace Engineering (ICMAE). 2021: 412-416[2025-03-05]. https://ieeexplore.ieee.org/abstract/document/9522552. DOI:10.1109/ICMAE52228.2021.9522552.
[2] LYNCH F T, KHODADOUST A. Effects of ice accretions on aircraft aerodynamics[J/OL]. Progress in Aerospace Sciences, 2001, 37(8): 669-767. DOI:10.1016/S0376-0421(01)00018-5.
[3] APPIAH-KUBI P. U.S Inflight Icing Accidents and Incidents, 2006 to 2010.[J].
[4] SILVA G, SILVARES O, ZERBINI E. Airfoil Anti-ice System Modeling and Simulation[C/OL]//41st Aerospace Sciences Meeting and Exhibit. Reno, Nevada: American Institute of Aeronautics and Astronautics, 2003[2024-06-24]. https://arc.aiaa.org/doi/10.2514/6.2003-734. DOI:10.2514/6.2003-734.
[5] 苏媛, 徐忠达, 吴祯龙. 飞机积冰后若干飞行力学问题综述[J/OL]. 航空动力学报, 2014, 29(8): 1878-1893. DOI:10.13224/j.cnki.jasp.2014.08.017.
[6] WU YUNAN, ZHANG DALIN, CHEN WEIJIAN, 等. Optimization for electric heating power’s distribution on electro-thermal anti-icing surface[J/OL]. CSAA/IET International Conference on Aircraft Utility Systems (AUS 2018): 217. DOI:10.1049/cp.2018.0291.
[7] ARIZMENDI GUTIéRREZ B, DELLA NOCE A, GALLIA M, 等. Optimization of a Thermal Ice Protection System by Means of a Genetic Algorithm[C/OL]//FILIPI? B, MINISCI E, VASILE M. Bioinspired Optimization Methods and Their Applications. Cham: Springer International Publishing, 2020: 189-200. DOI:10.1007/978-3-030-63710-1_15.
[8] 刘宗辉, 卜雪琴, 林贵平, 等. 基于PHengLEI的非稳态电热除冰过程仿真[J]. 空气动力学学报, 2023, 41(2): 53-63.
[9] GUO X, YANG Q, ZHENG H, 等. Optimization of power distribution for electrothermal anti-icing systems by differential evolution algorithm[J/OL]. Applied Thermal Engineering, 2023, 221: 119875. DOI:10.1016/j.applthermaleng.2022.119875.
[10] POURBAGIAN M, HABASHI W G. Surrogate-Based Optimization of Electrothermal Wing Anti-Icing Systems[J/OL]. Journal of Aircraft, 2013, 50(5): 1555-1563. DOI:10.2514/1.C032072.
[11] POURBAGIAN M, HABASHI W. Power and Design Optimization of Electro-Thermal Anti-Icing Systems via FENSAP-ICE[M/OL]//4th AIAA Atmospheric and Space Environments Conference. American Institute of Aeronautics and Astronautics, [2025][2025-02-05]. https://arc.aiaa.org/doi/abs/10.2514/6.2012-2677. DOI:10.2514/6.2012-2677.
[12] NIU J, SANG W, QIU A, 等. AN OPTIMIZATION OF ANTI-ICING CHAMBER BASED ON POD AND KRIGING[J]. 2020: 1132.
[13] 杨倩, 郑皓冉, 程显达, 等. 基于引气控制的热气防冰优化设计方法[J]. 航空学报, 2023, 44(S2): 214-223.
[14] LIU J, KE P. Aero-Engine Inlet Vane Structure Optimization for Anti-Icing with Hot Air Film Using Neural Network and Genetic Algorithm: 2019-01-2021[R/OL]. Warrendale, PA: SAE International, 2019[2025-02-05]. https://www.sae.org/publications/technical-papers/content/2019-01-2021/. DOI:10.4271/2019-01-2021.
[15] 屈经国, 彭博, 易贤, 等. 基于深度神经网络的任意翼型结冰预测方法[J]. 空气动力学学报, 2023, 41(7): 48-55.
[16] 何磊, 钱炜祺, 易贤, 等. 基于转置卷积神经网络的翼型结冰冰形图像化预测方法[J]. 国防科技大学学报, 2021, 43(3): 98-106.
[17] WANG X (王旭), KOU J (寇家庆), ZHANG W (张伟伟). Unsteady aerodynamic prediction for iced airfoil based on multi-task learning[J/OL]. Physics of Fluids, 2022, 34(8): 087117. DOI:10.1063/5.0101991.
[18] 陈宁立, 易贤, 王强, 等. NNW-ICE软件的三维结冰模型及精度验证[J]. 航空学报, 2024, 45(12): 59-71.
[19] CHEN N, YI X, WANG Q, 等. Numerical Study on Wind-Driven Thin Water Film Runback on an Airfoil[J/OL]. AIAA JOURNAL, 2023, 61(6): 2517-2525. DOI:10.2514/1.J062618.
[20] CHEN N, YI X, WANG Q, 等. An analysis of heat transfer inside the ice layer and solid wall during ice accretion[J/OL]. International Communications in Heat and Mass Transfer, 2022, 137: 106276. DOI:10.1016/j.icheatmasstransfer.2022.106276.
[21] ROWLEY C W. MODEL REDUCTION FOR FLUIDS, USING BALANCED PROPER ORTHOGONAL DECOMPOSITION[J/OL]. International Journal of Bifurcation and Chaos, 2005, 15(03): 997-1013. DOI:10.1142/S0218127405012429.
[22] BERKOOZ G, HOLMES P, LUMLEY J. The Proper Orthogonal Decomposition in the Analysis of Turbulent Flows[J/OL]. Annual Review of Fluid Mechanics, 2003, 25: 539-575. DOI:10.1146/annurev.fl.25.010193.002543.
[23] 张钰, 刘建伟, 左信. 多任务学习[J]. 计算机学报, 2020, 43(7): 1340-1378.
[24] ZHANG Y, YANG Q. A Survey on Multi-Task Learning[J]. IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING, 2022, 34(12).
[25] VASWANI A, SHAZEER N, PARMAR N, 等. Attention Is All You Need[A/OL]. arXiv, 2023[2024-06-24]. http://arxiv.org/abs/1706.03762. DOI:10.48550/arXiv.1706.03762.
[26] MISRA I, SHRIVASTAVA A, GUPTA A, 等. Cross-Stitch Networks for Multi-Task Learning[C/OL]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2016: 3994-4003[2025-02-05]. https://openaccess.thecvf.com/content_cvpr_2016/html/Misra_Cross-Stitch_Networks_for_CVPR_2016_paper.html.
[27] TANG H, LIU J, ZHAO M, 等. Progressive Layered Extraction (PLE): A Novel Multi-Task Learning (MTL) Model for Personalized Recommendations[C/OL]//Proceedings of the 14th ACM Conference on Recommender Systems. New York, NY, USA: Association for Computing Machinery, 2020: 269-278[2025-03-07]. https://doi.org/10.1145/3383313.3412236. DOI:10.1145/3383313.3412236.
[28] 赵志鹏. 基于改进深度多任务学习的滚动轴承故障预测方法研究[D/OL]. 电子科技大学, 2023[2025-02-05]. https://link.cnki.net/doi/10.27005/d.cnki.gdzku.2023.001133. DOI:10.27005/d.cnki.gdzku.2023.001133.
[29] 张大海, 孙锴, 倪平浩. 基于耦合关系挖掘及渐进式分层提取多任务学习网络的风-光-荷短期预测[J/OL]. 电网技术, 2023, 47(9): 3537-3547. DOI:10.13335/j.1000-3673.pst.2022.1492.
CJA