Survey of real-time and on-line diagnostic expert systems

This paper reviews the published work on on-line and real-time diagnostic systems embedded in other processes which have advanced past the prototype stage and are in routine use. The basic differences between real-time, on-line, and off-line systems are discussed with the range of current applications being indicated. First generation expert systems used rules, event trees, and/or fault dictionaries to implement their expertise. Second generation expert systems, on the other hand, start with a model of the system and descriptions of how components of the model operate. The benefits and problems of both techniques are discussed. The use of different reasoning methods (e.g. 'shallow' vs. 'deep') is compared. The range of interaction between diagnostic systems and the monitored system, as well as between the diagnostic system and the user, are examined. Difficulties and other issues that arise in these systems are discussed. Finally, a proposal for the design of future diagnostic expert systems is presented.