Actuator fault estimation using neuro-sliding mode observers

Reformulated principle for designing actuator fault estimation for continuous-time linear MIMO systems, based on neuro-sliding mode observer structure, is presented in this paper. Radial basis function neural network is used as a model-free fault approximator of the unknown additive fault. The method utilizes Lyapunov function and the steepest descent rule to guarantee the convergence of the estimation error asymptotically, where the design parameters can be obtained using LMI techniques. Finally, the proposed fault estimation scheme is applied to a nonlinear water tank system and simulation results illustrate its satisfactory performance.