基于流形学习的卫星姿态控制系统故障检测技术研究
发布时间：2018-08-26 18:53
【摘要】：卫星在轨运行期间会向地面传送大量遥测数据,这些数据真实地反应了卫星的有效载荷与运行状态,通过挖掘高维遥测数据低维特征信息,可以有效提高卫星异常状态检测能力和可靠性水平。本文针对“天巡一号”卫星姿态控制系统遥测数据,开展了基于局部线性潜入流形学习法的高维数据特征提取与故障检测方法研究,并建立了卫星姿态控制系统故障检测快速半物理仿真系统,验证故障检测方法可行性与适用性。在总结分析卫星遥测数据的一般特征与分类基础上,深入研究卫星姿态控制系统遥测数据特性,设计基于主元分析的卫星姿态控制系统故障检测方案。从在轨卫星的星上工作模式与遥测参量影响因素两方面,介绍卫星姿态控制系统遥测数据基本特征;针对遥测数据高维特性,通过主元分析方法,开展高维数据降维与特征提取方法研究,同时利用统计量实现对低维特征信息的异常检测。针对一般线性特征提取方法无法挖掘非线性高维遥测数据特征信息问题,本文研究了基于局部线性嵌入流形学习法的数据特征提取与故障检测方法。非线性高维遥测数据的低维嵌入几何结构难以通过一般线性特征提取方法获得,采用局部线性嵌入流形法,设计高维遥测数据降维与特征提取方案;针对获得的低维特征信息,结合统计量SPE和2T设计故障检测方法;通过“天巡一号”遥测数据验证所设计的特征提取与故障检测方案的有效性。针对在线样本数据不断更新,传统批处理工作模式的局部线性嵌入难以更新完善数据库等问题,研究了基于增量式局部线性嵌入法的数据特征提取与故障检测方法。通过在线样本更新权值矩阵进而完善数据库,在此基础上设计了在线样本特征提取与故障检测方案,通过“天巡一号”遥测数据验证方法的有效性。针对地面获取遥测数据不完整、卫星在轨故障模拟困难等问题,设计卫星姿态控制系统故障检测快速仿真平台。采用PC104和AD7011-EVA单板机分别模拟星载控制计算机、模型计算机,采用激励源端信号注入方式实现故障模拟与故障注入,将多配置飞轮接入闭环回路,通过对飞轮系统的故障模拟与故障注入,模拟飞行、获取仿真数据基础上验证上述故障检测方法的可行性以及检测结论的一致性。
[Abstract]:A large number of telemetry data will be transmitted to the ground during the orbit operation. These data truly reflect the payload and operation state of the satellite. By mining the low-dimensional characteristic information of the high-dimensional telemetry data, It can effectively improve the ability and reliability of satellite abnormal state detection. In this paper, the feature extraction and fault detection method of high-dimensional data based on the local linear submersible manifold learning method is developed for the telemetering data of the satellite attitude control system of "Tianxuan-1" satellite. A fast semi-physical simulation system for fault detection of satellite attitude control system is established to verify the feasibility and applicability of the fault detection method. On the basis of summarizing and analyzing the general characteristics and classification of satellite telemetering data, the characteristics of satellite attitude control system telemetry data are deeply studied, and the scheme of satellite attitude control system fault detection based on principal component analysis is designed. This paper introduces the basic characteristics of satellite attitude control system telemetry data from the two aspects of onboard working mode and influencing factors of telemetry parameters, aiming at the high dimensional characteristics of telemetering data, the method of principal component analysis is used to analyze the characteristics of satellite attitude control system. The methods of dimensionality reduction and feature extraction of high-dimensional data are studied. At the same time, anomaly detection of low-dimensional feature information is realized by using statistics. Aiming at the problem that the general linear feature extraction method can not mine the feature information of the nonlinear high-dimensional telemetry data, this paper studies the data feature extraction and fault detection method based on the local linear embedded manifold learning method. The low dimensional embedded geometric structure of nonlinear high dimensional telemetry data is difficult to be obtained by general linear feature extraction method. The method of local linear embedding manifold is used to design the dimensionality reduction and feature extraction scheme of high dimensional telemetry data. Combined with statistic SPE and 2T, the fault detection method is designed, and the effectiveness of the designed feature extraction and fault detection scheme is verified by "Tianyun-1" telemetry data. Aiming at the problem that the online sample data is constantly updated and the local linear embedding of the traditional batch processing mode is difficult to update and perfect the database, the data feature extraction and fault detection method based on incremental local linear embedding method is studied. By updating the weight matrix of online samples to perfect the database, an online sample feature extraction and fault detection scheme is designed, and the validity of the method is verified by the remote sensing data of "Tianyun-1". Aiming at the problems such as incomplete acquisition of remote sensing data on the ground and difficulty in simulation of satellite fault in orbit, a fast simulation platform for fault detection of satellite attitude control system is designed. The PC104 and AD7011-EVA single board computer are used to simulate the spaceborne control computer and the model computer respectively. The fault simulation and fault injection are realized by the excitation source signal injection method. The multi-configuration flywheel is connected to the closed-loop. Through the fault simulation and fault injection of the flywheel system, the simulation flight is simulated, and the feasibility of the above fault detection method and the consistency of the detection conclusion are verified on the basis of obtaining the simulation data.
【学位授予单位】：南京航空航天大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：V448.22;V467
[Abstract]:A large number of telemetry data will be transmitted to the ground during the orbit operation. These data truly reflect the payload and operation state of the satellite. By mining the low-dimensional characteristic information of the high-dimensional telemetry data, It can effectively improve the ability and reliability of satellite abnormal state detection. In this paper, the feature extraction and fault detection method of high-dimensional data based on the local linear submersible manifold learning method is developed for the telemetering data of the satellite attitude control system of "Tianxuan-1" satellite. A fast semi-physical simulation system for fault detection of satellite attitude control system is established to verify the feasibility and applicability of the fault detection method. On the basis of summarizing and analyzing the general characteristics and classification of satellite telemetering data, the characteristics of satellite attitude control system telemetry data are deeply studied, and the scheme of satellite attitude control system fault detection based on principal component analysis is designed. This paper introduces the basic characteristics of satellite attitude control system telemetry data from the two aspects of onboard working mode and influencing factors of telemetry parameters, aiming at the high dimensional characteristics of telemetering data, the method of principal component analysis is used to analyze the characteristics of satellite attitude control system. The methods of dimensionality reduction and feature extraction of high-dimensional data are studied. At the same time, anomaly detection of low-dimensional feature information is realized by using statistics. Aiming at the problem that the general linear feature extraction method can not mine the feature information of the nonlinear high-dimensional telemetry data, this paper studies the data feature extraction and fault detection method based on the local linear embedded manifold learning method. The low dimensional embedded geometric structure of nonlinear high dimensional telemetry data is difficult to be obtained by general linear feature extraction method. The method of local linear embedding manifold is used to design the dimensionality reduction and feature extraction scheme of high dimensional telemetry data. Combined with statistic SPE and 2T, the fault detection method is designed, and the effectiveness of the designed feature extraction and fault detection scheme is verified by "Tianyun-1" telemetry data. Aiming at the problem that the online sample data is constantly updated and the local linear embedding of the traditional batch processing mode is difficult to update and perfect the database, the data feature extraction and fault detection method based on incremental local linear embedding method is studied. By updating the weight matrix of online samples to perfect the database, an online sample feature extraction and fault detection scheme is designed, and the validity of the method is verified by the remote sensing data of "Tianyun-1". Aiming at the problems such as incomplete acquisition of remote sensing data on the ground and difficulty in simulation of satellite fault in orbit, a fast simulation platform for fault detection of satellite attitude control system is designed. The PC104 and AD7011-EVA single board computer are used to simulate the spaceborne control computer and the model computer respectively. The fault simulation and fault injection are realized by the excitation source signal injection method. The multi-configuration flywheel is connected to the closed-loop. Through the fault simulation and fault injection of the flywheel system, the simulation flight is simulated, and the feasibility of the above fault detection method and the consistency of the detection conclusion are verified on the basis of obtaining the simulation data.
【学位授予单位】：南京航空航天大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：V448.22;V467
【参考文献】
相关期刊论文 前10条
1 李维铮;孟桥;;基于遥测数据动态特征的卫星异常检测方法[J];空间科学学报;2014年02期
2 王健;冯健;韩志艳;;基于流形学习的局部保持PCA算法在故障检测中的应用[J];控制与决策;2013年05期
3 ;Design and simulation of fault diagnosis based on NUIO/LMI for satellite attitude control systems[J];Journal of Systems Engineering and Electronics;2012年04期
4 单长胜;李于衡;王荔斌;;在轨卫星异常报警和故障诊断方法研究[J];飞行器测控学报;2011年03期
5 赵圣占;王t熌，
本文编号：2205838
本文链接：https://www.wllwen.com/kejilunwen/hangkongsky/2205838.html
