Knowledge acquisition for fault diagnosis in gas chromatography

Abstract Knowledge acquisition is frequently considered to be a major bottleneck in the development of an expert system. An explanation-based method was applied to knowledge compilation in a gas chromatographic diagnostic system, GCdiagnosis. Initially, generally accepted gas chromatography fundamental variables were identified to provide the necessary understanding of the physico-chemical processes of a gas chromatography system with flame ionization detection. A theoretical model was built, which consisted of six components, namely carrier gas, injection, injector, column, fuel gas and detector. Interactions between each component were defined. Through the operation of this model, the knowledge used for diagnosing problem GC data can be generated by tracing changes in a specific component that is found to be faulty. Consistency verification of the compiled knowledge was carried out using a prototype. The knowledge is available for extension in future versions of GCdiagnosis.