航空学报 > 2010, Vol. 31 Issue (8): 1614-1621

基于扩展卡尔曼滤波的动量轮故障检测方法

李知周1,4, 张锐1,2,3, 朱振才2,3, 梁旭文1,2,3   

  1. 1. 中国科学院 上海微系统与信息技术研究所 2. 中国科学院 微小卫星联合重点实验室 3. 上海微小卫星工程中心 4. 中国科学院 研究生院
  • 收稿日期:2009-08-17 修回日期:2009-10-16 出版日期:2010-08-25 发布日期:2010-08-25
  • 通讯作者: 梁旭文

Extended Kalman Filter-based Fault Detection for Momentum Wheel

Li Zhizhou1,4, Zhang Rui1,2,3, Zhu Zhencai2,3 , Liang Xuwen1,2,3   

  1. 1. Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences 2. Microsatellites State Key Joint Laboratory, Chinese Academy of Sciences 3. Shanghai Engineering Center for Microsatellites 4. Graduate University,Chinese Academy of Sciences
  • Received:2009-08-17 Revised:2009-10-16 Online:2010-08-25 Published:2010-08-25
  • Contact: Liang Xuwen

摘要: 动量轮作为卫星姿态控制系统的关键执行部件,对其故障检测对维持卫星的正常运行具有重要意义。首先对动量轮系统的故障进行分析,在建立动量轮线性离散状态空间模型的基础上,把动量轮的故障检测作为时变参数系统的跟踪来处理,将动量轮的模型参数作为扩展的状态空间中的状态量,使得动量轮物理模型参数与状态空间中的状态量有对应关系,通过在扩展了的状态空间上采用扩展的卡尔曼滤波,完成时变参数的跟踪。然后,将离散空间的状态量变换回连续空间中,利用物理参数与状态量的对应关系,实现对动量轮物理参数的跟踪。此方法物理意义明确,为系统的物理参数提供了定量的估计值,为进一步诊断故障原因提供了良好的基础。数值仿真表明,此方法能够通过同时检测多个故障参量,实现故障的检测并满足卫星实时性要求。

关键词: 系统辨识, 故障检测, 扩展卡尔曼滤波, 动量轮, 时变系统跟踪, 离散系统连续化

Abstract: Momentum wheel is a key actuator of the attitude controlling system of a satellite. Thus, it is very meaningful to accomplish its fault detection to maintain the working order of the satellite. First,this article analyses the fault of the momentum wheel. Based on building the linear discrete-state-space model, parameter tracking is used as fault detection of the momentum wheel. The state space is extended to include the parameters of the momentum wheel, which makes the physical parameters of the momentum wheel establish a relationship with the variables in state space. By using an extended Kalman filter in the extended state space, time-varying parameter tracking is accomplished. Then the variables in the discrete state space are converted to the continuous state space. The relationship between physical parameters and state variables is used to realize the tracking of the physical parameters of the momentum wheel. This method has clear physical meanings and provides a quantitative estimate for the physical parameters, which serves as a good basis for further fault diagnosis. The numerical simulation shows that this method can detect the fault by detecting multi-fault variables and can satisfy the real time requirement of satellite.

Key words: identification, fault detection, extended Kalman filters, momentum wheel, tracking of time-varying systems, continuation of discrete system

中图分类号: