Research Of Electric Load Simulator Control Method Based On Neural Network

Load simulator is an important ground loop simulation equipment, which can simulate aerodynamic load on aircraft. Electric load simulator has the features of small volume and been widely used. With the development of aviation technology, there put forward a higher requirement of controlling performance and accuracy for loading system. The controlling strategy based on neural network was researched:(1) Regarding the output of rudder as external disturbance of load simulator,and then the mathematical model of load simulator was obtained according to the physical characteristics of components. In consideration of the parameter perturbation and uncertain situation of system model, analyzing that the un-modeled dynamics, parameter perturbation, the extra torque and external interference had an impact on system performance.(2) Studying the extra torque and its inhibition method of simulator. In order to effectively suppress the extra torque, a complex suppression strategy based on speed feed-forward and disturbance observer was presented. The simulation results show that the complex compensation strategy can effectively inhibit the extra torque.(3) For the instruction torque tracking accuracy problem of load simulator, a dynamic neural network generalized inverse control scheme was proposed. The dynamic networkconsisted of a three-layer feed-forward network and integrator. Firstly, training neural network offline to get inverse model; and then putting it before the simulator to be a pseudo-linear system; implementing closed-loop control with additional controller. The simulation results show the control strategy can improve torque tracking precision of load simulator effectively.(4) The neural network generalized inverse method had two deficiencies, that can't guarantee the real-time control and the non-ideal characteristic of pseudo-linear system.According to these, the online regulation and common model control(CMC) were introduced to improve system performance. Simulation results show that the composite method improves the torque tracking accuracy and robustness of load simulator.