In this thesis, we developed an expert system for computer fault diagnosis based on the heuristic classification method proposed by William J. Clancey. In order to apply the heuristic classification, diagnostic process of micro-computer systems is analyze to be identified as a sequence of actions; data abstraction, heuristic match, and solution refinement. The heuristic match, which associates abstracted data with an abstracted solution, is identified as the key step and a rule is a suitable format to represent the knowledge involving heuristic match. Various other diagnostic knowledge is encoded in terms of symptom data, hypotheses, frames, abstraction rules, and refinement rules. A prolog-based expert system shell running on the IBM-PC is used to build the expert system. The shell provides various knowledge representation schemes such as frames and rules. User-friendly interface was emphasized by implementing windows, menus, and various explanation facilities.
The use of the heuristic classification as a diagnostic model supports modularity of knowledge base and, therefore, it facilitates easy modification and expansions of knowledge base. By the successful development of the expert system, the heuristic classification is confirmed as a well-suited model for diagnostic problems.