IMC based automatic tuning method of PID controllers using neural networks

Abstract

In this paper, a new approach for neural PID tuning is presented based on the use of a neural network and the internal model control (IMC) principle. The neural network is used to adjust on line the PID controller after an off line training step. The developed approach is based on the use of a neural supervisor having as inputs the control signal, its correspondent output and the filter time constant and the PID parameters as outputs. The design of the supervisor is based on the procedures of linearization and IMC principle. We briefly outline the content of the tuning course and finish the paper with illustrative example where good performances have been obtained.