An adaptive internal model control system of a piezo-ceramic actuator with two RBF neural networks

This paper presents a neural network based positioning control system of a piezo-ceramic actuator which exhibits hysteretic behavior. Proposed control system utilizes two neural networks with radial basis function (RBF) as their activation functions: one is used for modeling hysteretic behavior of the actuator and the other is assigned the role of a feedback controller for hysteresis compensation and tracking. The particle swarm optimization algorithm has been applied to the training of RBF-NN for modeling PZT dynamics to achieve high precision, whereas back propagation has been used for online controller parameters update. An internal model control (IMC) structure is employed which combines aforementioned two neural networks for positioning control of the actuator. Results of the positioning control simulation of PZT will be shown to indicate the validity of the proposed two RBF-NN internal model control system.