A sliding mode observer based FDI scheme for a nonlinear satellite systems

This paper proposes a method to identify and isolate actuator faults in a nonlinear benchmark satellite system Mars Express (Mex), subject to model uncertainties and sensor noise. A nonlinear sliding mode observer is designed to estimate the faults induced in the thrusters. The nonlinear Mex system gives rise to a special case of a sliding mode observer with no reduced order dynamics. The equivalent injection signal is filtered to generate a residual signal which is used to identify the faults present by means of threshold selection. Fault isolation is based on directional residual set evaluation using the equivalent injection signal. Finally performance of the designed FDI is verified using Monte Carlo simulations.