几何可调燃烧室虽然能够满足超燃发动机对动力性能的需求，但其具有多变量、强耦合的特性，不仅难以针对性地开发高性能控制算法，而且成为阻碍超燃发动机领域发展的难点、痛点问题。为缓解控制器设计与机械耦合之间的矛盾，本文首先针对传统的几何可调超燃冲压发动机，重新设计了多点燃料注入机械架构，使得几何可调燃烧室获得了更多的可操作自由度，确保了理论上提升超燃发动机性能的能力。同时，在重构的多点燃料注入模式基础上，本文提出了一种改进的LSTM/DDPG控制方案，用以解决在超声速燃烧情况下的精准压力分布式控制难题，进而充分发挥出超燃发动机在宽速域、长时间飞行下的潜在性能。仿真结果充分验证了提出方案的有效性，即本文提出的针对多点燃料注入架构的LSTM/DDPG改进控制方案，能够有效地改善/优化几何可调燃烧室在不同工况下的燃烧性能。最后，通过硬件在环 (Hardware-In-the-Loop，HIL) 仿真验证了本文提出的控制方法的实用性，在不同扩张比下压力分布均可有效跟随指令控制，实现了宽速域几何可调燃烧室多点压力均方差的高精度控制目标。
Generally, variable geometry combustor can ensure the suitable performance for scramjets. But its multivariate and strong coupling characteristics make it difficult to develop high-performance control algorithms, and also become an intractable problem hindering the development of the scramjet field. In order to solve the contradiction between control design and mechanical coupling, this paper firstly proposes the multipoint fuel injection structure for the conventional variable geometry combustor. The newly proposed structure can provide more number of operable inputs, thereby ensuring a theoretical ability for improving the system performance. On the basis of the multipoint fuel injection structure, a new LSTM/DDPG-based control scheme is proposed to solve the problem of distributed pressure control in the case of supersonic combustion. This can effectively explore the potential performance of the scramjet in wide range velocities and long flight times. The effectiveness of the proposed scheme is then verified with the simulation results, i.e., the LSTM/DDPG-based control scheme can significantly optimize the combustion performance of variable geometry combustor under different operating conditions. Finally, the practicality of the proposed scheme is validated on the hardware-in-the-loop simulation platform, and the pressure distribution can be effectively controlled under different expansion ratios. It is obvious to realize the high-accuracy control of multipoint pressure distribution in a wide range velocity for a variable geometry combustor.
[1]刘小勇, 王明福, 刘建文, 任鑫, 张轩. 超燃冲压发动机研究回顾与展望[J]. 航空学报, 2024, 45(5): 529878-529878.
Xiaoyong LIU, Mingfu WANG, Jianwen LIU, Xin REN, Xuan ZHANG. Review and prospect of research on scramjet[J]. Acta Aeronautica et Astronautica Sinica, 2024, 45(5): 529878-529878. (in Chinese)
[2]Ji C, Wu S, Yi Y, et al. Effects of hydrogen injection strategies on the flow field and combustion characteristics in a hydrogen-fueled rotary engine with the swirl chamber[J]. Fuel, 2024, 364: 130951.
[3]王璐，钱战森，高亮杰, Wang L, Qian Z S, Gao L J. 基于边界层燃烧方法的宽速域飞行器内流道减阻研究. 推进技术, 2022, 43(6): 137-146.
Drag Reduction by Boundary Layer Combustion on Internal Flowpath in a Wide-Range Mach Numbers Vehicle[J]. 2022, 43(6): 137-146. (in Chinese)
[4]Hu R, Wang W. Numerical Study of Pipeline Distribution Effect on the Performance of Pasty Propellant Rocket Motor[J]. Space: Science & Technology, 2023, 3: 0083.
[5]BOUCHEZ M, LEVINE V, FALEMPIN F, et al,. Airbreathing space launcher interest of a fully variable geometry propulsion system and corresponding french-russian partnership[C]. AIAA Paper ,2000:3340.
[6]Zhang J, Lin L, Yao Z, Bao W. Research on combustion characteristics of the strut-equipped variable geometry scramjet combustor at Mach 2 condition[J]. Acta Astronautica et Astronautica Sinica, 2022, 201: 12-21.
[7]杨庆春. 宽速域超声速燃烧室的热力与几何调节方法研究[D].哈尔滨: 哈尔滨工业大学，2015.
YANG Qingchun.The research on thermal and geometric adjustment methods of wide-speed supersonic combustion chamber[D].Haerbin:Harbin Institute of Technology,2015.(in Chinese)
[8]杜宪, 马艳华, 王欣悦, 徐羚, DU X, MA Yh, et al. 基于神经网络时延预测的航空发动机内模控制器设计.
Design of Aero-Engine Internal Model Contr ol System Based on Neural Network Time-D elay Prediction. 推进技术. 2023;44(11):214-2 6.
[9]Volodymyr M., Koray K., David S., et al. Human-level control through deep reinforcement learning[J]. Nature, 2015, 518(7540): 529-533.
[10]肖冰, 张海朝. 航天器姿态稳定强化学习鲁棒最优控制方法[J]. 航空学报, 2024, 45(1): 628890-628890.
Bing XIAO, Haichao ZHANG. Reinforcement learning robust optimal control for spacecraft attitude stabilization[J]. Acta Aeronautica et Astronautica Sinica, 2024, 45(1): 628890-628890.
[11]盛汉霖，顾至诚，陈芊，刘祁，尹炳雄，王喆，张天宏, SHENG Hl, GU Zc, CHEN Q, LIU Q, YIN Bx, et al. 嵌入式分布式架构下航空发动机模型预测控制方法研究
Research and Implementation of Model Predi ctive Control for Aircraft Engines with Emb edded Distributed Architecture. 推进技术. 20 23;44(11):198-213.
[12]Liu Q, Zhang H, Zhang X, Yuan D. Improved DDPG Based Two-Timescale Multi-Dimensional Resource Allocation for Multi-Access Edge Computing Networks[J]. IEEE Transactions on Vehicular Technology, 2024, 73(6): 9153-9158.
[13]Zhu M, Tian K, Wen Y-Q, et al. Improved PER-DDPG based nonparametric modeling of ship dynamics with uncertainty[J]. Ocean Engineering, 2023, 286: 115513.
[14]金哲豪,刘安东,俞立.基于GPR和深度强化学习的分层人机协作控制[J].自动化学报，2022,48(09):2352-2360.
JIN Zhehao,LIU Andong,YU Li,Hierarchical human-robot cooperative control based on GPR and deep reinforcement learning[J].Acta?Automatica?Sinica,2022,48(09):2352-2360. (in Chinese)
[15]WU H, SONG S, YOU K, et al. Depth control of model-free AUVs via reinforcement learning[J]. IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2018, 49(12): 2499-2510.
[16]丰硕.宽速域冲压发动机几何可调燃烧室性能影响规律研究[D].哈尔滨工业大学,2017.
Feng Shuo.Study on the Effect of Variable Geometry Combustor on the Performance of Ramjet with Wide Velocity Range [D].Harbi n Institute of Technology, 2017.
[17]Ji M, Xu G, Duan Z, et al. Cooperative pursuit with multiple pursuers based on Deep Minimax Q-learning[J]. Aerospace Science and Technology, 2024, 146: 108919.
[18]Mousa A. Extended-deep Q-network: A functional reinforcement learning-based energy management strategy for plug-in hybrid electric vehicles[J]. Engineering Science and Technology, an International Journal, 2023, 43: 101434.
[19]Jung-Wook P, Harley R G, Venayagamoorthy G K. Adaptive-critic-based optimal neurocontrol for synchronous generators in a power system using MLP/RBF neural networks[J]. IEEE Transactions on Industry Applications, 2003, 39(5): 1529-1540.
[20]Jin H, Qi H, Zhao J, et al. Software Systems Implementation and Domain-Specific Architectures towards Graph Analytics[J]. Intelligent Computing, 2022.
[21]Hou Y, Liu L, Wei Q, et al. A novel DDPG method with prioritized experience replay[C]. 2017 IEEE International Conference on Systems, Man, and Cybernetics (SMC). 2017: 316-321.
[22]Shi D, Gan W-S, Shen X, et al. What is behind the meta-learning initialization of adaptive filter? — A naive method for accelerating convergence of adaptive multichannel active noise control[J]. Neural Networks, 2024, 172: 106145.
CJA