Controller tuning via rule based identification

Abstract

Based on the recently developed rule based identification algorithm of Wang and Jones (1991), this paper presents a new algorithm for controller tuning in unit feedback control systems. In the paper, the rule based identification algorithm is used directly to identify the model between the reference step input and the integral of the tracking error. Two stages of tuning are proposed. In the first stage, the correct structure of the controller is chosen by monitoring the integral of the tracking error. This enables the realisation of perfect steady-state tracking. In the second stage, the estimated model is used to tune the controller gains in order to provide an improved dynamic performance for the closed-loop response. The results of applying the algorithm to the stimulation of a real process are given.