Abstract: Aiming at a class of control canonical form MIMO(multi-input multi-output) nonlinear systems which add nonlinear error and fault form,an adaptive control scheme combining backstepping technology for neural network is proposed. The method is used to estimate the model error and fault form on-line.In this paper,the robust term is utilized to compensate approximation model error.When the fault term is detected,the virtual control value is used to compensate the fault form by adjusting every step.The fault term effects on system is eliminated.By theoretical proof,every residual error signal is uniformly bounded,and finally converges to an arbitrarily small neighborhood around zero.Simulation results show the feasibility of the presented approach.