A robust fault detection and isolation method via sliding mode observer

A robust fault detection and isolation (FDI) approach for a class of nonlinear systems with uncertainty was presented. The FDI scheme was based on sliding mode observer, which was robust against system uncertainty. Fault detection can be realized by use of sliding boundary size. When the fault had been detected, the estimate part in the observer for the fault can be enabled. A radial basis function (RBF) neural network was used to approximate the fault, so making the fault isolation a simple task. The theoretic analysis guaranteed the convergence of the observer. Simulation results show the feasibility of the proposed approach.