Neural-network-based self-tuning PI controller for precise motion control of PMAC motors

Abstract

In general, proportional plus integral (PI) controllers used in computer numerically controlled machines possess fixed gain. They may perform well under some operating conditions, but not all. To increase the robustness of fixed-gain PI controllers, we propose a new neural-network-based self-tuning PI control system. In this new approach, a well-trained neural network supplies the PI controller with suitable gain according to each operating condition pair (torque, angular velocity, and position error) detected. To demonstrate the advantages of our proposed neural-network-based self-tuning PI control technique, both computer simulations and experiments were executed in this research. During the computer simulation, the direct experiment method was adopted to better model the problem of hysteresis in the AC servo motor. In real experiments, a PC-based controller was used to carry out the control tasks. Results of both computer simulations and experiments show that the newly developed dynamic PI approach outperforms the fixed PI scheme in rise time, precise positioning, and robustness.