Robust fault detection and diagnosis based on neural network nonlinear observer

A robust fault detection and diagnosis strategy based on observer for nonlinear systems with unknown uncertainty is presented. A neural network is constructed to approximate the fault on-line. The nonlinear observer can not only detect fault, but also realize the fault diagnosis. It is proved that the scheme has good robustness against modeling error and uncertainty. At last, simulations of a three-tank system illustrate the effectiveness of the proposed methodology.