The application of self-tuning control

Abstract

Thanks to cheap computing power, self-tuning control has matured rapidly from an academic topic to a position in which it can be seriously considered as an alternative to the ubiquitous PID regulator for many industrial process-control loops. The purpose of this paper is to discuss some of the key ideas in self-tuning - from the point of view of an industrial control engineer without excessive use of mathematics. The use of predictive control laws is explored first in the non-adaptive context in which the closed-loop response, the control input behaviour, and the response to load disturbances can be (to a certain extent) individually tailored by the user. This is contrasted with the usual tuning of PID coefficients based on the subjective evaluation of control performance. The role of adaptation is then discussed in terms of its possible applications, for example for commissioning fixed controllers or for keeping a critical loop tightly tuned. Some of the likely pitfalls in applying adaptive control are outlined, and alternatives to the approach based on discrete-time predictor theory are briefly mentioned.