Research On The Fault Diagnosis Of The Satellite Attitude Control System

As the structure and performance of the space technology become more and more complex, itis also harsh to the fault diagnosis technology. First of all, we establish the attitudecontrol system of satellite (ACSS) model, which is studied from the theoretical aspect. Then the fault of the actuator is diagnosed by combining the advantages of the adaptive fault diagnosis observer and the Bayesian network. The main tasks are included as follows:First, the concept and the common methods of fault diagnosisare introduced. Besides, the methods and development direction ofspacecraft fault diagnosis are systematically analyzed.Secondly, several kinds of the satellite coordinate system and the attitude descriptive methods are described. The satellite attitude kinematics equations and dynamics equations are given according to the physical description of the satellite. The instruments and actuators of theACSS are described combined with the fault. Then the simulation model of the ACSS is constructed for the study of fault diagnosisstrategies.Combining the advantages of artificial neural network, the adaptive fault diagnosis observerof the actuator is designed based on the Radial Basis Function (RBF). The RBF neural network adaptive fault diagnosis observer is proved stability by using Lyapunov stability theorem. By comparing with fault detection observer, the effectiveness and superiority of the adaptive observer areverified,fault diagnosis is completed from system level to component level.Finally, an fault diagnosis method based on Bayesian network is investigated for the ACSS because of the uncertainty of the momentum wheel.The probability of Bayesian network is introduced with the method to obtain fault by failure mode and effects analysis (FMEA). Combined with Bayesian network model, FMEA and "Cause-Mode-Effect", a method is put forward. Bayesian network fault diagnosismodel of the momentum wheel isestablished, what’s more, the convenience of the method is illustrated by simulation.Fault diagnosis is completed from component level to element level.