Some Aspects of an Architecture for Real-Time Expert Systems in Industrial Plant Automation and Supervision

Abstract

The highest level of control and supervision of industrial plants, for instance fossile or nuclear power generating systems or chemical production units, still has to be done by human operators. To support these operators, there is an increasing demand for real time and on-line expert systems.Most of the expert systems in industrial use today are of the interactive type. They work on static situations and query the user to gather information. But when shifting to a time-varying environment, where the expert system is connected to the plant and incoming data change while the expert system infers, new architectural concepts of these systems are needed.By watching and analysing the behaviour of human operators in different plant stati, we developed an architecture for realtime expert systems centered around the situation concept. Not incoming data, but abstractions of these data (in it’s own a knowledge based operation), called situations, are the basis for human decisions in plant operation. This concept leads to a multi-parallel expert system architecture to be described in the paper.