Control Strategy Of Electric Loading System For Actuator Based On RBF Neural Network

As the ceaseless development of aeronautical technology, it is an inevitable trend to take automated equipment as a substitute for manual equipment to test the actuator. The function of electric loading system is to simulate torque loading of actuator output shaft of helicopter in flight under test laboratory condition, while also to verify whether the actuator achieve the design standards. During the test procedure, the loading motor needs to follow the actuator’s movement while at the same time putting torque on it, and the extraneous torque then become inevitable, which is a common problem in dynamic loading system. The existence of extraneous torque would badly affect the loading precision and dynamic response, therefore it is significantly important to find a proper control strategy to eliminate the extraneous torque. It is difficult to use the traditional control strategy based on accurate mathematic model to achieve the desired results because of the limitation of strategy.Model reference adaptive RBF control strategy is used to realize closed-loop control of electric loading system in this paper. The main works and research are as follows:(1)Firstly, the mathematic model for electric loading system is built to analyze the cause of extraneous torque on the level of transformation function. The influence of extraneous torque is illustrated through simulation, which shows that the extraneous torque would make the tests inaccurate.(2)The problem of changing mathematical model exists in the electric loading system. To solve this problem, an approved RBF neural network learning algorithm, with features of electric loading system fully integrated, is proposed on the basis of neural network performance simulation results. The concept of significance for the hidden layers is introduced, and a new kind of growing and pruning strategy of hidden layer that leads to faster studying speed is also presented. The model reference adaptive RBF control strategy, based on the improved RBF learning algorithm, is also presented in this paper, which constructs neural network identifier to identify the model of the object in real time. The identifier provides the training basis for the controller. RBF neural network is also taken as neural network controller at the same time, whose goal is to minimize the error between outputs of reference model and controlled object, therefore the extraneous torque is eliminated and the torque control instruction is precisely tracked.(3)The simulation and analyzing are carried out in the MATLAB environment, the validity of the control strategy is verified, and the tool of learning curve is also used to check the sanity of learning algorithm. The simulation result shows that the proposed control strategy can suppress the extraneous torque effectively under either different torque instruction or different interference.(4)The hardware structure and bundled software of the electric loading system are accomplished, platform is built up through using PXI Bus system, ART 9606 DAQ card and 2307 DAQ card, software is also developed by LabVIEW Virtual Instrument technology, and then the highly automated, user-friendly and fully functional test equipment is implemented.In this paper the strategy based on RBF neural network is applied to eliminate extraneous torque of electrical loading system, which relatively well solve testing problem of a certain model actuator. The research has certain reference value for future study.