Repetitive Learning Observer Based Actuator Fault Detection, Isolation, and Estimation with Application to a Satellite Attitude Control System

An actuator fault isolation and estimation (FIE) scheme using a bank of repetitive learning observers (RLOs) for a class of discrete-time nonlinear systems is investigated in this paper. The parameters of these observers are repetitively updated using a proportional-derivative type learning algorithm at each sampling time. Based on the proposed RLOs, a group of diagnostic residuals are generated correspondingly. An actuator fault is located when only one residual goes to zero while the others do not. The parameter of the observer that locates the fault specifies the fault. Theoretically, sufficient conditions for the proposed fault detection, isolation and estimation scheme are derived. Practically, the proposed FIE scheme is applied to a satellite attitude control system, and the simulation results demonstrate its effectiveness.