Assessment of interaction in process control systems
by
Abdelmadjid Khelassi
Boumerdes University
Coauthors: J.A. Wilson, R. Bendib

Interaction between control loops has long been recognised as an area for concern within the field of process control. The first formalised method for assessing the level of interaction present in a system of loops was Bristol’s relative gain array. Being based solely on steady state data, construction of the RGA is pleasingly simple and application of the technique has attracted a significant researchers. Where strong interaction effects arise as a problem in practice their elimination can be approached at a progression of levels each carrying a progressively increased cost implication.

The decision as to which of these levels of approach is warranted in overcoming any specific case of interaction involves gaining insight into the mechanics of the system itself. In the future, development in the use of detailed dynamic modelling and simulation may lead to the possibilities of evaluating the degree and form of interaction present and of developing compensating strategies.

In this paper we aim to draw attention to the fact that in considering interaction between feedback control loops an important distinction should be drawn between cases of true ( two-way) interaction and cases of what can be termed “one-way interaction “ or disturbance propagation “ from one loop to another. A frequency domain approach using magnitude/log frequency plots is presented which allows assessment of the direction and intensity of disturbance propagation present in a system. Some examples of interaction presented previously in the literature are analyzed on this basis and found to exhibit strong disturbance propagation features. In addition some anomalous results from use of the relative gain array are placed in context and related to the effects of using real rather than perfect controllers.

Date received: January 22, 2004

Copyright © 2004 by the author(s). The author(s) of this document and the organizers of the conference have granted their consent to include this abstract in Atlas Conferences Inc. Document # calu-62.