Fault Diagnosis For Spacecraft Attitude Control Systems Using Bayesian Network

The attitude control system is one of the most important subsystem of the spacecraft, due to the spacecraft works in the harsh environment, it is prone to faults. Fault diagnosis technology can make the spacecraft response system abnormalities very quickly, so as to reduce or even avoid catastrophic accidents. With the spacecraft attitude control system as the research object, the importance of system components is analyzed by using the Bayesian Network. On the basis of this, fault diagnosis problems of the spacecraft attitude control system are studied deeply. The main work of this paper as follows:First of all, the concept of fault and fault diagnosis are introduced after investigating a large number of literatures. The methods and the research status at home and abroad of spacecraft fault diagnosis technology are analyzed systematically, at the same time, the future development direction of the technology is also given. After analyzing applications and advantages of the Bayesian Network in fault diagnosis area, we find that the Bayesian Network model can deal with the uncertainty information very well and it is not dependent on the mathematical model of the system, so the model is introduced into the study of fault diagnosis problem in the spacecraft attitude control system in this paper.Secondly, the basic theory of the Bayesian Network is summarized. Its generation、 development and category are introduced simply, its learning algorithms and inference algorithms and tool softwares are analyzed systematically, then the construction process of the Bayesian Network and Bayesian Network classifiers are analyzed emphatically, which provide a solid theoretical foundation for the later studies.Then, in order to save fault diagnosis resources and increase fault diagnosis speed, the node selection problem of the attitude control system based on the sensitivity function in Bayesian Network is investigated. For the spacecraft attitude control system, the Bayesian Network model of the attitude control system is established by using expert experience and historical fault data, which avoids the influence of uncertainty factors. With combining the Bayesian Network model and the Bayesian sensitivity function, a node selection algorithm for the spacecraft attitude control system is put forward, it can effectively find out the key components of the system through calculating the node importance, which guides the later research of the fault diagnosis algorithm.Finally, for the typical fault of the key components in the spacecraft attitude control system, a fault diagnosis method for the spacecraft attitude control system is proposed by using the Bayesian Network classifiers, which combines the signal feature extraction technology and the Bayesian Network classifiers. The method uses the time domain and frequency domain signal feature extraction technology to extract the fault signal characteristic information, then it uses the attribute reduction algorithm to remove redundancy attributes of the feature information, and on the basis of these feature attributes, a modified naive bayesian classifier is designed to classify the data and realize fault diagnosis. Finally, the method is applied to research the single fault diagnosis and multiple faults diagnosis of the spacecraft attitude control system, the effectiveness of the proposed method is verified by fault diagnosis examples analysis.