| Attitude control system is one of the final subsystems in the satellite system,and as the complexity of its structure and the harshness of the working environment increase,the probability of failure during operation also increases.Fault diagnosis and health management technology can improve the stability and safety of the satellite system,reduce the occurrence of failures in the system,thereby reducing the occurrence of catastrophic accidents and ensuring that the satellite can complete the task safely and stably.This paper takes satellite attitude control system as the research object,and studies the fault diagnosis and health management of satellite attitude control system based on Bayesian network and machine learning.The main contents of the paper include:Firstly,this paper introduces the concepts of fault diagnosis and health management by investigating a large number of literature,and systematically summarizes the methods of satellite fault diagnosis and health management,as well as its research status at home and abroad.The advantages of Bayesian network and machine learning in fault diagnosis and health management are analyzed in detail,because they have better judgment on uncertain data and do not need to establish accurate system models.Therefore,this paper applies it to the fault diagnosis and health management of satellite attitude control system.Secondly,the basic theories of Bayesian networks and machine learning are summarized and introduced.The generation and development of Bayesian network and machine learning are expounded,and the construction process and related algorithms are analyzed,among which Bayesian network includes structure learning,parameter learning,inference and classification algorithms,machine learning includes various commonly used loss functions,Softmax,linear layer,convolutional layer,Bayesian neural network,and other algorithms to lay a solid theoretical foundation for the application of the following in satellite attitude control system.Then,this paper proposes the Bayesian Le Net neural network fault diagnosis method and applies it to the fault diagnosis of satellite attitude control system,and builds its model on the basis of relevant theories,so as to introduce uncertainty into the neural network.Through the analysis and debugging of the data of the semi-physical simulation platform,the four fault data of the reaction flywheel in the subsystem are classified and compared with the traditional neural network model.It is found that the Bayesian Le Net neural network model improves the accuracy,reduces overfitting,successfully achieves the purpose of fault diagnosis of the attitude control system,and lays a foundation for health management tasks.Finally,under the premise of fault diagnosis of satellite attitude control system,this paper proposes a Bayesian network-based health management method and applies it to the task of health management of the system.This method has two modules of fault prediction and health assessment,and the above fault diagnosis results are analyzed and calculated,and finally the health score of the system is quantitatively analyzed and expressed for different health levels,and experiments show that the method successfully achieves the purpose of health management of satellite attitude control system. |
