Research On Fault Detection Methods Of The Reaction Wheel Of Satellite

As an important actuator of satellite,reaction wheel has a long on-orbit lifetime.It works through the whole life cycle of the satellite,and affects the success of the entire mission.Therefore,it is necessary to detect the health status of reaction wheel.Two fault detection methods are studied in this paper,one is based on analytical model and the other is based on knowledge.In the aspect of analytical model-based methods,fault detection based on Dual-UKF is proposed.Double unscented Kalman filters are used to detect the faults by estimating the key parameters of reaction wheel.This method has higher precision and less computation than methods based on extended Kalman filter.Because this method has certain requirements for sampling frequency,so it is more suitable for fault detection in ground running test.In the other aspect,a fault detection method of combined residual is proposed.Based on the characteristics of telemetry data,the proposed method uses XGBoost regression model to predict the rotation speed and generate residual with available data.The sensitivity of viscous friction coefficient to sudden change of friction torque is also combined with the method.This method overcomes the shortcomings of neural network methods,has lower requirements for data,and has strong robustness for incomplete data,so it is suitable for on-orbit telemetry data monitoring system.Simulation results show that the proposed fault detection methods both can effectively detect the common faults of reaction wheel.