Robust fault diagnosis based on on-line adaptive approximator for nonlinear systems

In this paper, a robust fault detection and diagnosis(FDD) scheme which is based on neural network on-line adaptive approximator for a class of nonlinear uncertain systems, is proposed. The proposed scheme can be able to realize FDD by using the neural network in the feedback path to capture only the nonlinear fault information of the estimated system. It is proved that the proposed scheme has good robustness against uncertainties including modeling error and unknown external inputs. Finally a simulation of three-phase current motor, which was utilized in flexible link robot and maybe occur often normal faults, demonstrates the effectiveness of the proposed methodology.