针对氢能电动无人机(Hydrogen powered-Unmanned Aerial Vehicles，H-UAVs)的总体设计阶段的优化问题，探讨了在异构多源域情况下如何有效利用迁移学习技术建立代理模型并进行优化。为了解决氢能电动无人机演化过程中异构样本所导致的建模样本获取成本高昂的问题，提出了一种基于数据生成的多源域融合(DG-MSF)代理模型建立框架，通过测地线流式核方法将异构的源域和目标域映射到高维空间中确定多源域的关系；使用基于边缘分布的数据生成方法，实现了对源域信息的有效融合；构建了一个基于多层感知器的神经网络作为代理模型，并通过预训练与微调的方法完成迁移，实现了对氢能电动无人机性能的高效预测，最后对氢能电动无人机进行了优化设计。分析结果表明，方法能够有效利用多源域数据，提高模型训练的效率和预测的准确性，实现了氢能电动无人机总体优化性能的提升，为氢能电动无人机的发展提供了有力的技术支持。
This paper addresses the optimization problem of the overall design stage of hydrogen-powered unmanned aerial vehicles (H-UAVs) in the context of heterogeneous multi-source domains. It explores how to effectively utilize trans-fer learning technology to establish a surrogate model and optimize H-UAVs in the presence of heterogeneous samples. To solve the problem of high cost of building a surrogate model due to heterogeneous samples during the evolution of hydrogen-powered unmanned aerial vehicles, this paper proposes a framework for establishing a mul-ti-source domain fusion (DG-MSF) surrogate model based on data generation. The geodesic flow kernel method is used to map the heterogeneous source and target domains to a high-dimensional space to determine the relation-ship between multi-source domains. The joint distribution-based data generation method is used to effectively inte-grate source domain information. A multi-layer perceptron neural network is built as a surrogate model, and it is trained and fine-tuned through pre-training and fine-tuning methods to achieve efficient prediction of hydrogen-powered unmanned aerial vehicle performance. Finally, the optimization design of the hydrogen electric UAV is carried out. The analysis results show that the proposed method can effectively utilize multi-source domain data, improve the efficiency of model training and prediction accuracy, and achieve the improvement of overall optimized performance of hydrogen-powered unmanned aerial vehicles, providing powerful technical support for the devel-opment of hydrogen-powered unmanned aerial vehicles.
[1]刘莉, 曹潇, 张晓辉, 等.轻小型太阳能氢能无人机发展综述[J].航空学报, 2020, 41(3):6-33
[2]Tan C, Sun F, Kong T, et al.A survey on deep transfer learning[C]//Artificial Neural Networks and Machine Learning–ICANN 2018: 27th International Conference on Artificial Neural Networks, Rhodes, Greece, October 4-7, 2018, Proceedings, Part III 27. Springer Interna-tional Publishing, 2018: 270-279.
[3]史耀耀, 张碧瑶, 史恺宁, 等.基于实例迁移学习的跨工况刀具剩余寿命预测-“航空发动机高性能制造”专栏[J].航空学报, 2023, 44(X):X29038-1-12
[4]廖奕校.基于多源域知识迁移的旋转机械智能故障诊断方法研究[D]. 华南理工大学, 2023.
[5]Gupta L, Edelen A, Neveu N, et al.Improving surrogate model accuracy for the LCLS-II injector frontend using convolutional neural networks and transfer learning[J].Machine Learning: Science and Technology, IOP Pub-lishing, 2021, 2(4):045025-1-18
[6]Tian K, Li Z, Zhang J, et al.Transfer learning based variable-fidelity surrogate model for shell buckling prediction[J].Composite Structures, 2021, 273(1):114285-1-16
[7]Peng L, Wu H, Gao M, et al.TLT: Recurrent fine-tuning transfer learning for water quality long-term prediction , , : .[J].Water Research, 2022, 225(1):119171-1-12
[8]Seo W, Park S, Kim N, et al.A personalized blood glu-cose level prediction model with a fine-tuning strategy: A proof-of-concept study[J].Computer methods and programs in biomedicine, 2021, 211(1):106424-1-8
[9]Nasir M U, Ghazal T M, Khan M A, et al.Breast Cancer Prediction Empowered with Fine-Tuning[J].Computa-tional Intelligence and Neuroscience, 2022, 2022(1):5918686-1-9
[10]Yang D, Liu L, Bai W, et al.Conceptual design and configurations selection of S/H-UAVs based on Q-Rung Dual Hesitant double layer FQFD[J].Chinese Journal of Aeronautics, 2024, 1(1):1-13
[11]Liu L, Bai W, Yang D.Flight Endurance Increasing Technology of New Energy UAV Based on a Strut-Braced Wing[J].International Journal of Aer-ospace Engineering, 2022, 2022(1):1-13
[12]蔚光辉.绿色能源小型电动无人机总体设计[D]. 北京理工大学, 2021.
[13]Ge W, Yu Y.Borrowing Treasures from the Wealthy: Deep Transfer Learning through Selective Joint Fine-Tuning[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu, HI: IEEE, 2017: 10–19.
[14]Gong B, Shi Y, Sha F, et al.Geodesic flow kernel for unsupervised domain adaptation[C]//2012 IEEE Con-ference on Computer Vision and Pattern Recognition. 2012: 2066–2073.
[15]Long M, Wang J, Ding G, et al.Transfer feature learn-ing with joint distribution adaptation[C]//Proceedings of the IEEE international conference on computer vi-sion. 2013: 2200-2207.
[16]Borgwardt K M, Gretton A, Rasch M J, et al.Integrat-ing structured biological data by Kernel Maximum Mean Discrepancy[J].Bioinformatics, 2006, 22(14):e49-e57
[17]Gretton A, Sejdinovic D, Strathmann H, et al.Optimal kernel choice for large-scale two-sample tests[C]//Advances in Neural Information Processing Systems. Curran Associates, Inc., 2012, 25.
[18]Rumelhart D E, Hinton G E, Williams R J.Learning representations by back-propagating errors[J].Nature, Nature Publishing Group, 1986, 323(6088):533-536
[19]Hinton G E.Deep belief networks[J].Scholarpedia, 2009, 4(5):5947-5949
[20]Scabini L F S, Bruno O M.Structure and performance of fully connected neural networks: Emerging complex network properties[J].Physica A: Statistical Mechanics and its Applications, 2023, 615(1):128585-1-17
[21]Jain H, Deb K.An Evolutionary Many-Objective Op-timization Algorithm Using Reference-Point Based Nondominated Sorting Approach,Part II: Handling Constraints and Extending to an Adaptive Approach[J].IEEE Transactions on Evolutionary Computation, 2014, 18(4):602-622
[22]Li K, Deb K, Zhang Q, et al.An Evolutionary Many-Objective Optimization Algorithm Based on Domi-nance and Decomposition[J].IEEE Transactions on Evolutionary Computation, 2015, 19(5):694-716
[23]Tian Y, Cheng R, Zhang X, et al.Diversity Assessment of Multi-Objective Evolutionary Algorithms: Perfor-mance Metric and Benchmark Problems [Research Frontier][J].IEEE Computational Intelligence Maga-zine, 2019, 14(3):61-74
CJA