Satellite Attitude Active FTC Based On Geometric Approach and RBF Neural Network

This paper proposes a novel satellite attitude Active Fault Tolerant Control System. The Active Fault Tolerant Control System is based on a Fault Detection and Diagnosis module providing fault estimates decoupled from the aerodynamic disturbance thanks to the application of the NonLinear Geometric Approach. The diagnosis module is a bank of adaptive filters based on a Radial Basis Function Neural Network. The controller reconfiguration exploits directly the online estimates of the fault signals. To the authors’ knowledge, the joint use of the NonLinear Geometric Approach and Radial Basis Function Neural Network is not present in literature. The use of a Radial Basis Function Neural Network allows to design generalized fault estimation filters which do not need any a priori information about the fault internal model. The use of the NonLinear Geometric Approach allows to obtain very accurate fault estimates, independently from the knowledge of the aerodynamic disturbance parameters. Extensive simulation results based on a detailed nonlinear satellite attitude model embedding disturbance description are given. The obtained results highlight that the proposed satellite Active Fault Tolerant Control System can deal with reaction wheel faults of different types (i.e. step faults, ramp faults, sinusoidal faults) thus obtaining good attitude control performances