Research On Electric Loading System Control Strategy Based On Neural Network

Electric load simulator is the equipment for rudder loading test, which is realized by motor. And it is used to simulate the load that the aircraft rudder surface suffered during flight in laboratory conditions, so it is important ground simulation equipment for weapon systems. As the loading motor following the position changes of the rudder motor passively, so the influence of surplus torque becomes inevitably which is the mutual problem of passive servo systems. The surplus torque seriously constraints on the loading system performance, making the output torque tracks the load curve imprecisely. Therefore, how to restrain the surplus torque, research and develop high performance loading equipment for rudder motor becomes urgent.As the electric loading system is a typical nonlinear system which is strongly interfered by factors such as coupling, delay and mechanical friction, and the loaded objects are usually uncertain, so it is often imperfect in restraining the surplus torque by using traditional control strategy based on precise mathematical model of the controlled object. As for the complexity of the electric loading system control, neural networks controlling is introduced in the paper. By using the excellent nonlinear function approximation, the output torque will follow the order accurately.Firstly, the whole electric loading system project is designed, using a DC torque motor as the actuator, and the mathematical models of each part are established. On the basis, the relationship between surplus torque and rudder motor's movement, and the main factors affecting surplus torque are analyzed. Combining the operating characteristics of electric loading system, an improved structure of the single neuron PID control strategy is proposed. In order to have a better performance on restraining surplus torque, a complex control strategy that combines PID control with the diagonal recurrent neural network control is provided subsequently based on the compensation control idea. Through introducing the rudder position disturbance signal to the compound controller, the feed-forward control is applied using neuron network, which can get good performance without identification of the system model. At the same time, the size of network and the learning speed have been optimized, which leads to high generalization ability and low calculating burden. Simulation results show that the compound control strategy has good performance, no only in the elimination of the surplus torque, but also in the system response rapidness and robustness.The main control interface of the electric loading system is designed based on the VB language, and a motion controller card with digital signal processor inside is used for the main control of the system, then corresponding experimental research is carried out based on the above theory.