Reference: Hewett, R. Transferability of Model-Based Reasoning Using Prime Models in a Reactor Core Cooling Diagnostic System. Knowledge Systems Laboratory, February, 1991.
Abstract: ICE (Identifying Causes and Effects) is a prototype system that exploits models of structures and functions of physical systems for use in multiple reasoning tasks including diagnosis, explanation, and prediction. ICE uses a model-based approach to reason about physical systems from models of physical structures and behaviors. It represents knowledge about physical systems at two levels of abstraction: domain and prime models. Domain models represent actual physical components and processes that make up particular systems whereas prime models represent abstractions of both normal and abnormal structures and behaviors of the systems. Each domain model can be viewed as an instance of a corresponding prime model. Thus, any properties of a domain model can be inherited model, any properties or causal behaviors of the corresponding prime model may entail instantiations. For example, a human respiratory system and an irrigation system are represented as domain models and both can be viewed as an instance of a prime flow model. In a prime flow model, we might represent abstract fault such as "a leakage of a flow structure can cause a decreased flow of a flow process that occurs in that structure." During reasoning about a respiratory system which can be viewed as a flow system, one can instantiate that "a leakage of a throat can cause decreased flow of inspired or expired air that flows through the throat."