Automatic Fuzzy Tuning of Proportional-integral Controllers Based on Time-domain Specifications

Abstract. In this paper, an online tuning method based on time-domain performance specification is proposed to determine the parameters of standard PI controllers. The method uses a process model to predict the future output and to detect the specification violation. The maximum performance violation along with general tuning guidelines formulated as fuzzy sets and rules are used to find the new PI parameter values. The procedure is repeated each sampling time to provide continuous automatic tuning of the PI controller so that it can preserve good performance over a wide range of operating conditions. The algorithm allows the usage of any form of models and uses four simple fuzzy sets with eleven simple fuzzy rules to maintain simplicity and minimum computational effort. Numerical testing of the algorithm on a CSTR and on an evaporator example shows that better performance than what can be achieved by standard PID controllers for both set point change and load disturbance. © 2005 (A.H. 1425) King Saud University