Abstract: Electromechanical actuators are widely used in industrial production and military fields. However, the model uncertainty of electromechanical actuators will reduce the accuracy of model-based nonlinear controllers. At the same time, due to the limitation of installation space and cost, it is often impossible to install speed sensors in system. To solve these problems, a sliding mode output feedback control strategy based on neural network is proposed. The constant disturbance and parameter uncertainty in the system are estimated by extended state observer(ESO). The universal approximation property of radical basis function neural network is used to estimate the time-varying disturbances in the system, and then the feedforward compensation technology is used to compensate the former. At the same time, the speed of the system observed by the ESO is used to design the control quantity, so as to realize the output feedback control. By using Lyapunov stability theorem, it is proved that the designed controller can achieve bounded stability of the system. A large number of simulation and experimental results prove that the controller designed can improve the control accuracy by an order of magnitude compared with the traditional PID and nonlinear controller.

Key words: electromechanical actuator, output feedback control, neural network, sliding mode control, model uncertaint