Robust Fault Diagnosis for Satellite Attitude Systems Using Neural State Space Models

Abstract

In this paper, a robust fault detection and diagnosis scheme using neural state space models has been developed for a class of nonlinear systems. The neural state space models are adopted to estimate the modeling uncertainties in the states and outputs of the system. Subsequently, a residual is generated to identify the characteristics of the fault. Moreover, the robustness, sensitivity and stability properties of the proposed fault detection and diagnosis scheme are rigorously derived. Finally, the neural state space model based fault detection and diagnosis scheme is applied to a satellite attitude control system and the simulation results demonstrated its good performance.