Neural Network Adaptive Control of Piezoelectric Actuator in Scanning Tunneling Microscope

Piezoelectric ceramics actuators have ultra high resolution, but the positioning precision is depressed by the inherent characteristic of hysteresis and nonlinearity. Accurate tracking is difficult to be achieved. An adaptive control scheme combining neural network estimator with traditional PID controller is proposed to improve the performance of piezo actuator in Scanning Tunneling Microscope. The mathematical model of the actuator is established based on the analysis of displacement mechanism. A neural network estimator is used to predict the displacement in the next step, and parameters in the traditional PID controller are modified through adjusting the weight values in the neural network with the function of self-learning and adaptability. A low frequency sine reference signal with the amplitude of 30μm is applied to the control system to validate the effectiveness of the proposed method. The results show that the mean tracking errors are reduced from 0.45μm under traditional PID control to 0.22μm of neural network control. The project requirement is satisfied.