Abstract: Based on the temporal analytical redundancy between the available star sensor and gyro measurements, two filters are utilized to isolate the star sensor faults in a threeaxis stabilized attitude control system with two star sensors. Considering that gyro measurement noise is multiplicative noise in the attitude kinematics equations, a residual generator based on the linear minimum mean square filter is designed to obtain residuals containing the fault message of the target star sensor. The target star sensor fault detection is realized by comparing the residual evaluation with the corresponding thresholds, which is computed according to the residual variances in fault free case. Star sensor faults are located by setting one sensor as the target each time and checking the detection logic afterwards. Numerical simulation results show that either abrupt faults or slow degradation of the measurement accuracy of star sensors can be isolated by the designed method.

Key words: fault diagnosis, linear minimum mean square filter, star sensor, multiplicative noise, bilinear system