A COMPARISON OF SEVERAL ON-LINE FAULT DIAGNOSIS SYSTEMS FOR A CSTR

摘要

This paper presents a comparison study of several knowledge based fault diagnosis systems for a CSTR (continuous stirred tank reactor) system. These diagnosis approaches include a rule based approach with rules developed from knowledge of system structures and component functions, a qualitative simulation based approach utilising a qualitative model of the process, a fuzzy neural network based approach, and a genetic algorithm based approach. Knowledge requirement and development effort of these approaches, as well as their performance, are compared.