Research On Support Vector Machine-Based Faults Detection Theory And Simulation Platform For Satellite’s Actuators

Attitude control system (ACS) is one of the most complex satellite subsystems. To keep highprecision, high reliability and a long life, fault diagnosis technique is necessary. This thesis makesresearch on Support Vector Machine (SVM)-based faults detection theory and application for satelliteattitude control systems during collision avoidance. The key contributions are summarized as follows:The attitude control system model of satellite is established, and the elements of actuator areproposed.Various fault classification methods are given, and the models of satellite actuator faults areanalyzed. Moreover, the characters of fault diagnosis system in satellite attitude control systemsduring collision avoidance are surveyed, which illustrate the difficulties of the fault diagnosis systems.A fault detection and isolation approach based on Self-organizing fuzzy neural network (SOFNN)is proposed. SOFNN has good dynamic performance in estimating the actuator faults for theconsidered dynamics with external noises and systems parameters uncertainties.A fault detection and isolation approach based on SVM is proposed. The most predominance ofSVM is proper for limited samples decision. SVM is the more favorable for the practical engineeringproblem as fault diagnosis.The semi-physical simulation systems based on single axis air-bearing platform is studied todemonstrate the effectiveness and feasibility of the proposed fault detection and isolation methods inthis paper.