In order to improve the response speed and accuracy of a certain follow-up load simulator,a study on neural network Sliding Mode Control (SMC) is conducted.Based on the model of the servo system,Non-singular Terminal Sliding Mode Controller (NTSMC) responds to its nonlinear dynamic changes.The output of nonlinear quantitative Cerebellar Model Articulation Controller (CMAC) is used to compensate for SMC output,and the gradient descent method is used to update its weight.Nonlinear quantitative CMAC has the advantages of strong generalization ability and fast convergence,and NTSMC has the advantages of strong robustness.The combination of the two can effectively reduce the influence of nonlinear factors in the load simulator.Simulation experiments show that the proposed method can ensure the systems stability,accelerate dynamic response and improve the control accuracy.