Abstract

In this thesis, the problem of fault detection and isolation in the attitude control subsystem of spacecraft formation flying is considered. For this purpose, first the attitude dynamics of a single spacecraft is analyzed and a nonlinear model is defined for our problem. This is followed up by generating the model of the spacecraft formation flight using the attitude model and controlling the formation based on virtual structure control scheme. In order to design the fault detection method, an extended Kalman filter is utilized which is a nonlinear stochastic state estimation method. Three fault detection architectures, namely, centralized, decentralized, and semi-decentralized are designed based on extended Kalman filters. Moreover, the `residual generation and threshold selection techniques are proposed for these architectures. The capabilities of the architectures for fault detection are studied through extensive numerical simulations. Using a confusion matrix evaluation system, it is shown that the centralized architecture can achieve the most reliable results relative to the semi-decentralized and decentralized architectures. Furthermore, the results confirm that the fault detection in formations with angular velocity measurements achieve higher level of accuracy, true faulty, and precision, along with lower level of false healthy misclassification as compared to the formations with only attitude measurements.
In order to isolate the faults, structured residuals are designed for the decentralized, semi-decentralized, and centralized architectures. By using the confusion matrix tables, the results from each isolation technique are presented for different fault scenarios. Finally, based on the comparisons made among the architectures, it is shown that the centralized architecture has the highest accuracy in isolating the faults in the formations. Furthermore, the results confirm that fault isolation in formations with angular velocity measurements achieve higher level of accuracy when compared to formations with only attitude measurements.