IAR-STSCKF-Based Fault Diagnosis and Reconstruction for Spacecraft Attitude Control Systems

This work investigates the actuator fault detection, isolation, and reconstruction (FDIR) for the spacecraft attitude control system (ACS) under unknown model uncertainty and colored measurement noise. First, a novel intelligent adaptive robust strong tracking square root cubature Kalman filter (IAR-STSCKF) is proposed to implement the parameter estimation. In this algorithm, two multifading factor matrices (MFMs) designed based on strong tracking theory make it achieve better robustness against non-Gaussian uncertainty and noise, as well as higher convergence speed and accuracy. In addition, the introduction of the support vector machine (SVM) discriminant module enables the two MFM to operate in coordination. Second, observers are designed based on IAR-STSCKF to estimate disturbance and fault. Also, a novel dual-observer FDIR scheme is proposed by introducing a gated recurrent unit neural network (GRUNN), which can further decouple fault from disturbance and achieve better robustness. Finally, the comparative simulation results are carried out on a spacecraft ACS to demonstrate the effectiveness and robustness of the proposed scheme.

Automated summary

This work investigates the actuator fault detection, isolation, and reconstruction for the spacecraft attitude control system under unknown model uncertainty and colored measurement noise. A novel intelligent adaptive robust strong tracking square root cubature Kalman filter is proposed for parameter estimation. Observers are designed based on the filter to estimate disturbance and fault, and a dual-observer FDIR scheme is proposed with the introduction of a gated recurrent unit neural network for better robustness.