We are developing a versatile maintenance expert system (VMES) for troubleshooting circuits. Like several other research teams we are using structural and functional descriptions to avoid difficulties of empirical-rule-based diagnosis systems in knowledge acquisition, diagnosis capability, and system generalization. Our diagnosis system has successfully pinpointed the faulty part of a multiplier/adder board, a favorite example for researchers in this field. We have embedded VMES in the SNePS Semantic Network Processing System, using it as a form of expert system tool. A central part of VMES is the 'SENDING' graphical interface. While troubleshooting, it displays the given device and dynamically indicates the state of the reasoning process. All knowledge used by 'display' is directly retrieved from the semantic network. This operation of 'display' is comparable to natural-language generation from a knowledge base. An important aspect of our research is to find a good knowledge-representation scheme to support diagnosis and display.
All Science Journal Classification (ASJC) codes