Experience of a predictive adaptive controller on pilot and industrial plants with transport phenomena

Abstract

The existing experience on using MUSMAR, a predictive adaptive controller, on industrial and large scale pilot plants with transport phenomena is discussed. The processes to control have been selected because their dynamics depends not only on time, but also on space, being therefore described by partial differential equations, and implying increase difficulties for the controller. Case studies on an industrial boiler, an arc-welding machine, a distributed collector solar field and a water distribution canal are used to illustrate the main difficulties and the corresponding solutions when using MUSMAR. These include plant model uncertainty and start-up adaptation transients, large and uncertain plant i/o transport delay, existence of un-modelled dynamics, closed-loop response shaping and constraints. The emphasis of the presentation is on the practical impact of the theoretical properties of the MUSMAR algorithm and on their illustration by means of actual experiments on the real processes mentioned above.