FAULT DETECTION AND DIAGNOSIS METHOD BASED ON SLIDING MODE-NEURAL NETWORK OBSERVER

A new fault detection and diagnosis method based on sliding mode observer is presented for a class of non-linear system. Sliding mode term ensures that the observer has robustness under non-trouble conditions, and fault detection can be realized by making use of sliding boundary size. When the fault has been detected, the estimation part in the observer for the fault may be enabled. A radial basis function neural network is used to approximate the fault, so fault diagnosis can be finished online. Simulation results show the feasibility of the proposed approach.