Issues regarding the design and acceptance of intelligent support systems for reactor operators

The central premise is that conventional expert systems which encode experimental knowledge in production rules are not a suitable vehicle for the creation of practical operator support systems. The principal difficulty is the need for real-time operation. Intelligent support systems will have to be heavily instrumented and the associated knowledge bases will require a hierarchical organization to emulate human approaches to analysis. Arguments to support the findings are made by first examining the needs of licensed reactor operators and then reviewing the capabilities of conventional expert systems. Differences between existing expert systems and proposed intelligent support tools are identified by comparing certain design features including the source and content of the knowledge base, the mode of interaction with the user, knowledge organization, the inference engine, and the man-machine interface. Issues related to operator acceptance of intelligent support tools are reviewed. Possible applications are described and the relative merits of the machine- and human-centered approaches to the implementation of intelligent support systems are enumerated