Research On Fault Diagnosis Of Satellite Attitude Control System Based On Data Driven

As one of the most important subsystems in the satellite operation process,the attitude control system determines whether the mission can be successfully completed.However,the challenging operating conditions of the satellite increase the possibility of failure of the attitude control system.Therefore,to improve the reliability of the satellite attitude control system,it is of great significance to research fault diagnosis technology.In recent years,with the rise of data-driven methods in various fields,it has also become one of the research hotspots in the field of fault diagnosis.In this paper,the satellite attitude control system is taken as the research object.Aiming at the characteristics of time series and multivariate strong correlation of the telemetry data of the satellite attitude control system,the simulation data is obtained using the dark matter satellite platform,and the data-driven algorithm based on the attention mechanism is used for fault diagnosis.Research.The research contents of this paper mainly include:(1)A fault diagnosis method based on time-variable Transformer is proposed for the characteristics of time-series and multi-variable strong correlation of telemetry data of satellite attitude control system.First,the method uses the time attention module to find the correlation between the data at a certain time and the data before the time,and effectively identify the time series dynamic characteristics of the data.At the same time,the variable attention module is introduced to capture the degree of correlation between different variables and realize multivariate decoupling.Then,dynamic data decorrelation is achieved by fusing the time-series dynamic characteristics of the data and the correlation between variables with the help of the interactive attention layer.In this way,the fault feature extraction effect of the satellite attitude control system is improved.(2)Aiming at the problem that the attention mechanism has a general ability to extract time-domain features,resulting in a poor fault diagnosis effect of the satellite attitude control system.Based on(1),this paper proposes a fault diagnosis method based on Attention-LSTM.First,the implicit correlation information between highdimensional variables is mined through the variable attention layer,and then the deep time series features are extracted by the LSTM network layer.The problem of too many features to be forgotten.In this way,the recognition accuracy of the slowly changing faults of the satellite attitude control system is improved.(3)Because of the problems of incomplete satellite telemetry data obtained on the ground and difficulty in obtaining satellite in-orbit faults.In this paper,with the help of the dark matter satellite platform,the experimental simulation data required for the above two fault diagnosis methods are obtained respectively,and the fault diagnosis simulation experiment is carried out to verify the feasibility of the two fault diagnosis methods.