Reduced order process modelling in self-tuning control

Abstract

One major objection to the model-based auto-tuning and self-tuning control is based on the observation by Bristol that a reduced order model being identified would not yield good control of the actual high order process. In this paper we have re-examined the work of Bristol and argued that this observations could not be generalized. We then consider reduced order modelling which incorporates a deadtime element. It is shown that the deadtime could help to provide a close match of the freqency response near the critical frequencies. Extensive simulation studies have also confirmed that reduced order modelling with deadtime included could be an adequate model for practical self-tuning control.