Knowledge engineering for model-based diagnosis: An experience-based development approach

Falut diagnosis has proved to be one of the most rewarding application areas for the introduction of knowledge-based systems. Initially all diagnostic systems were based upon shallow knowledge, and many proved to be highly effective within a narrow task-specific domain. The current thrust of research, however, is also focusing on the use of model-based reasoning, which provides deeper knowledge of the structure and function of the device under diagnosis. Whilst much has been written with regard to development methods for shallow knowledge-based systems, relatively little has been published specifically relating to model-based approaches. This paper describes the approach adopted for the development of a model-based application for the diagnosis of hydraulic systems, paying particular attention to the lessons learned.