Development of a diagnostic expert system using fault tree analysis information

The majority of Knowledge Based Expert Systems (KBES's) applied to the field of system diagnostics employ compiled heuristic knowledge obtained from human experts and operating manuals. They often lack the ability of solve atypical and novel problems. To correct these and other shortcomings, more recent Diagnostic Expert Systems capture and use 'deep knowledge', which is that knowledge based on the design and structure of a system.; In this thesis a prototype diagnostic KBES for a nuclear auxiliary boiler feed system is described. A commercial Expert System shell accesses a developed heuristic knowledge base in order to resolve the majority of common faults. Supplementing the KBES are a series of C routines which have been designed to access the deep knowledge contained in a Fault Tree data base. The specific Fault tree was previously developed during the licensing process. These functions are called by the KBES shell and perform diagnostics, give detailed explanations of failure mechanisms, and identify system failure risks through examination of cut sets. Use of hashing routines and a lexical analysis mechanism have improved speed of execution and clarity of output. (Abstract shortened by UMI.)...