Fault Diagnosis Method of Satellite Attitude Control System Based on Stacked Autoencoder Network

摘要

In view of the fault diagnosis problem of satellite attitude control system, a fault diagnosis method based on stacked autoencoder (SAE) network is proposed. This method uses SAE network to learn the historical data of system state variables, mines the correlation between state variables and establishes the normal state reconstruction model of the system. Once the fault occurs, the relationship between the state variables changes, and the fault detection is carried out according to the reconstructed residuals of the state variables. Then, the fault diagnosis rules reflected by the component residuals are extracted by the decision tree, and the fault isolation location is carried out according to the rules. The simulation results show that the proposed method, which has good robustness, can accurately reflect the associated relationship between the state variables, and can diagnose the minor faults hidden in the disturbance.