Abstract: In order to prevent the surplus torque of servo load simulator which had influence in the performance of the torque load, an intelligent controller was put forward based on the improved self-learning functional expansion wavelet neural network. The Bang-Bang control is used when the error is big, and if the error is small, function expansion based on wavelet neural network and fuzzy compensation control is used; At the same time, the improved differential evolution algorithm is for estimating the parameters of the controller. Considering the computational complexity and performance of the control system, the number of hidden neurons of the learning algorithm was designed. The results of simulation show that the dynamic and static performance wholly achieves double-ten indicator, this control strategy with the feasibility and rationality can improve the tracking performance and control precision of the torque load system.

Key words: load simulator, surplus torque, wavelet neural network, functional expansion, differential evolution, self-learning