Elman dynamic neural network control for direct-drive feed system in advanced CNC machine tools

Abstract

Permanent Magnet Linear Synchronous Motors (PMLSM) are gaining increasing interest as direct-drive actuator for CNC machine tools. However, the servo control system of PMLSM is multi-variable, non-linear and strong-coupled and sensitizes to various force disturbances. For high speed and high precision to CNC machine tools, Elman dynamic Neural Network(NN) is introduced and the compound controller of Proportional(P) controller and Elman NN is proposed based on on-line training. Elman NN is trained by P controller at the beginning of control process. Then, speed loop of PMLSM servo system is controlled by Elman NN. The work process of proposed compound controller is exposed and dynamic characteristic of Elman NN is proved. The simulation results show that the proposed control scheme not only has strong robustness to uncertainties of the linear servo system, but also has good tracking precision and disturbance rejection abilities.