Realtime Decision Support System for Plant Operators Using Concept Formation Method

Abstract

This paper presents an interface system for a plant diagnosis that has a capabilities of forming the hierarchically organized concepts on the anomalous behaviors from the incoming stream of measurement data. The system is implemented by usage of a machine learning method of concept formation. Especially, we put emphasis on the importance of realtime decision support environment, where a diagnosis system presents information anytime with a variety of abstraction levels catching up with the evolution of system's anomalous states and must avoid the duration of presenting no response to the operator.