| Satellites, as one kind of the most important spacecraft, require a high reliability to continue their missions. As the space environment is harsh, satellites are unavoidably susceptible to possible faults, especially for the attitude control systems(ACS). The ACS is the essential subsystem of satellites and faults of ACS usually lead to serious consequences, such as system failure and even a catastrophe. Fault diagnosis provides an effective method to improve the safety and reliability of systems, thus many space technology personnels and scholars focus their attentions on fault diagnosis of satellite attitude control systems. In this paper, a satellite attitude control system is considered as the research object, and the corresponding fault diagnosis problems are studied. The main research contents and results are as follows:Aiming at the problem of fault isolation for actuators and sensors of the satellite attitude control system, fault isolation strategy based on multiple analytical model is proposed. According to the redundancy relationship of component structure between the dynamics and the kinematics, the proposed scheme is developed in two phases, fault preliminary detection and particular fault isolation. With the design of the two phases, the faults in satellite control system including actuators, gyros and star sensor can be isolated successfully. In the first phase, two fault detection observer are designed to achieve the determination of faulty part; in the second phase, two banks of fault isolation observers activated by the previous detection results are designed to isolate the particular fault.Since it is the nature of systems that model uncertain, disturbance, and noise are unavoidable in practice, traditional analytical model-based fault diagnosis can not achieve the detection of the fault with tiny magnitude. In this situation, this paper studied a combined model-based and neural network method to detect the tiny actuator fault in satellite attitude control system. The proposed approach is an extension of the observer-based method with artificial neural networks modeling technique to enhance the performance of the diagnosis system. The basic idea of our study is to use model-based method to decouple the possible disturbances and isolate faults, and then, appeal to neural network to further reduce the influence of remaining model uncertainties. It consists of a nominal model designed based on observer and a compensation model based on neural network. With the decision logic considered, the small faults of actuators can be detected successfully.Traditional analytical model-based fault diagnosis method does not have the ability for fault identification, meanwhile, expert system which is widely adopted encounters the difficulty of acquiring diagnosis rules. In this situation, fault identification method based on residual-analysis is proposed in the paper. Since the residual generated by the model-based method is independent with the initial state and input of the system, the residual in ideal situation represents the fault information. Residual-analysis method consists of residual generation, residual processing and diagnosis rules acquiring. Through further analysis and processing of the residual, the potential features can be obtained to represent fault, and then rough set theory is adopted to extract diagnosis rules. Finally, three types of actuator faults in satellite are taken as example, and the corresponding diagnosis rules are acquired by the proposed method.Fault estimation proposes the essential fault information for active fault-tolerant technique, which has been widely investigated in recent years. Traditional fault estimation method is designed more conservative since the fault is usually assumed as constant or slowly varying, thus it can only achieve an unbiased estimation for constant fault. Therefore, this paper studied a new nonlinear augmented observer. With a assumption that the fault is finite times derivatives, the proposed method can achieve a more accurate estimation of actuator fault in satellite attitude control system, and the estimation ability is illustrated by comparing it to the traditional method. |
