Integrating expert knowledge into industrial control structures

This paper is concerned with the improvement of existing industrial control structures by the integration of human experience and expertise. Through two case studies concerning mechanical and petrochemical industries, it is illustrated that conventional control techniques and knowledge-based strategies can complement each other. The objective is to derive maximum profit from available materials and equipment by combining them efficiently. From a concrete point of view, linguistic fuzzy systems (LFS) are used to achieve a collaboration between numeric and expert processing. Three distinct architectures are proposed to combine conventional regulators and knowledge-based procedures in a unified control structure. All of them are based on a context-oriented decomposition of the control problem under consideration. The choice of a particular architecture is then related to existing industrial installations and is finally dependent on company strategy. (C) 2003 Elsevier B.V. All rights reserved.