Adaptive fault detection and diagnosis for a class of nonlinear uncertain systems with on-line learning

The problem of fault detection and diagnosis (FDD) for a class of nonlinear systems with unknown uncertainty is studied in this paper. An adaptive FDD observer is proposed based on dead-zone operator, on-line learning and adaptive compensation techniques. The fault detection decision is made by evaluating the residual signals. After a fault is detected, a neural network estimator is constructed to approximate the real fault signal on-line. To improve the performance of the fault diagnosis, the adaptive term is applied to compensate the unknown disturbance, modeling uncertainties and optimal approaching error. Finally, the simulation results show the effectiveness of the proposed methodology.