This work in progress is aimed at providing a framework for model-based diagnostic expert systems, considering knowledge and strategies that are useful for diagnosing faults. As the first step to reach the goal, we provide a framework for hypothesis generation, one of the three subtasks of diagnosis, classifying information needed for diagnosing faults into three classes: the domain model, additional domain knowledge, and diagnostic strategy. Three sorts of knowledge belonging to the domain model, namely, device model, process model, and topological relative position are used to model the device being diagnosed. Heuristics and naive physics, which are classified in the additional domain knowledge, are used as additional information to diagnose faults. As strategies for diagnosing faults, the framework applies three sorts of diagnostic strategies: the qualitative value propagation, direct path of causality, and structural fault localization. An algorithm for coordinating the sorts of knowledge and the strategies in diagnosing faults is also provided. Furthermore, an example of diagnosing a structural fault in the domain of refrigeration plants is given to illustrate how the framework works.
ASJC Scopus subject areas
- Computer Science Applications
- Artificial Intelligence