推力矢量极性错误下的飞行控制自主重构技术  

Autonomous reconfiguration of flight control under thrust vector polarity errors

作　　者：潘豪[1] 胡瑞光 宋征宇 邵梦晗 Hao PAN;Ruiguang HU;Zhengyu SONG;Menghan SHAO(Beijing Aerospace Automatic Control Institute,Beijing 100854,China;China Academy of Launch Vehicle Technology,Beijing 100076,China;College of Control Science and Engineering,Zhejiang University,Hangzhou 310027,China)

机构地区：[1]北京航天自动控制研究所,北京100854 [2]中国运载火箭技术研究院,北京100076 [3]浙江大学控制科学与工程学院,杭州310027

出　　处：《中国科学：信息科学》2022年第5期870-889,共20页Scientia Sinica(Informationis)

基　　金：国防基础科研(批准号:JCKY2018203B022)资助项目。

摘　　要：飞行故障在线自主辨识与控制重构是实现航天智能飞行的一个重要标志.本文针对运载火箭飞行中的推力矢量极性错误,研究了基于人工神经网络的故障识别与自主重构方法.在明确故障模式基础上,结合扩张状态观测器(extended state observer,ESO)观测结果,提出了考虑模型偏差的仿真训练样本设计方法,并分别采用基于反向传播(back propagation,BP)神经网络和长短时记忆(long short term memory,LSTM)网络方法进行了故障辨识设计,仿真结果表明,两种方法均具有较高识别准确度,均可实现推力矢量极性错误下的平稳自主重构,相比而言,基于滑窗时间序列的LSTM网络方法更具优势,具有较高的识别准确度.Online failure autonomous identification and control reconfiguration are important features of aerospace intelligent flights.Aiming at handling thrust vector polarity errors in the flights of launch vehicles,this paper studies the fault identification and autonomous reconstruction method based on artificial neural networks(ANNs).On the basis of clarifying the fault modes and combining the observation results of an extended state observer(ESO),a design method for learning/verifying samples through simulations and considering model uncertainties and disturbances is proposed.Then,a fault identification design is carried out based on a back propagation(BP)network and a long short term memory(LSTM)network,respectively.The simulation results show that both methods can meet the requirements for failure mode identification,which provide the basis for a stable and autonomous reconstruction under thrust vector polarity errors.By contrast,the LSTM network method based on the sliding window has more advantages and higher recognition accuracy.

关 键 词：极性错误 自主重构 扩张状态观测器 反向传播 长短时记忆 

分 类 号：V249.1[航空宇航科学与技术—飞行器设计]