Process control using the G2 Real-Time Expert System

The authors consider the application of expert system technology in process plants. Particular attention is given to the problems of knowledge representation, including the need to represent dynamic qualitative knowledge, dynamic analytic knowledge, and the deep structure of the process. It is noted that the application of inference in real-time requires paradigms that go beyond rote pattern matching, using metaknowledge to focus the inferencing resources of the expert system. The application of truth maintenance requires a temporal model of the time dependence of the truth of data and inferred results. A structure that takes these considerations into account is presented. This application environment, called the G2 Real-Time Expert System, has been placed at over 100 sites. The first large online process plant installations are discussed. The results demonstrate that engineers, using G2, can build and manage large real-time expert systems. The effective use of G2 in real-time operation, including access to online data and the running of significant knowledge bases in real time, has also been demonstrated.<>