Abstract: To solve time-varying and nonlinear problem of the compound actuator ,the neural-network adaptive control principle was proposed. Establish the dynamic model of compound actuator using nonlinear auto-regressive moving average model, multi-layer perception neural network was used to identify the nonlinear model. The controller was based on the theory of generalized inverse and the network weights were deduced with respect to the error between the object and the identified model. The self-turning control was realized in combination with control-algorithm modify online. The simulation results show that this control scheme has good effect on the control of compound actuator under the model having time-varying and nonlinear character.