Fault detection, identification and reconstruction for gyroscope in satellite based on independent component analysis

Although satellites are designed with high reliability, faults do occur when satellites are in orbit. To avoid the important services being affected, redundancy is used in satellites. There are many sensors in satellites. In order to reduce the cost, space, weight and power consumption, redundant sensors should be added to satellite as few as possible. Analytical redundancy is an efficient way to optimize the application of redundant. The gyroscope is the attitude determination sensor of the satellite. The minimum redundant structure of the gyroscope system is as follows: three gyroscopes installed in three-axis orthogonally and one gyroscope installed with slantwise for redundancy(3o + 1S). To achieve fault detection, identification and reconstruction, hypothesis of statistical independence between the three-axis angular rates and hypothesis of statistical independence between the angular rates and fault are proposed. The scenario that only one sensor is faulting and there are only additive fault and full fault is supposed. Under these assumptions, firstly a threshold method is used for fault detection. After a fault is detected, independent component analysis (ICA) based algorithm for fault identification is employed. To overcome the ambiguities of ICA, correlation coefficients and prior information of the mixed matrix are used. Finally, the reconstruction matrix is obtained. By using this matrix fault signal is extracted so that the yaw, roll and pitch axes (three-axis) angular rates of the satellite can be recovered. Numerical simulations show this method can fulfill fault detection, identification and reconstruction of the gyroscope system. (C) 2010 Elsevier Ltd. All rights reserved.