Abstract: In order to solve the problem caused by extra torque of aircraft rudder electric loading system, a hybrid control strategy is proposed where the improved credit assignment cerebellar model articulation controller is applied for feedforward control and the incremental proportion integral derivative (PID) is applied for feedback control. The feedforward controller of threedimensional input neural network structure is designed, which is combined with variable stiffness metalrubber buffer spring, the feedback of torque velocity and gradient loading method, in view of weight adjusting algorithm of variable balanced learning constant based on Sigmoid function. The feedback controller of incremental PID can not only solve the shortcoming of integral term overflow, but also provide training learning samples for the neural network. Finally, the convergence characteristic of the improved algorithm and the stability of the closedloop system are shown through the theoretical analysis. The simulation results show that the method can effectively improve the loading accuracy and the capability of online realtime control, and to some extent suppress the extra torque.

Key words: electric loading system, extra torque, neural network, Sigmoid function, metalrubber buffer spring