Automated production of information models for use in model-based diagnosis

The approach to system maintenance uses a model of the system as a foundation for its knowledge base. A particular type of model, the information flow model, is very useful for diagnosing problems. The authors explore a learning approach to assist the modeling process. They describe a system in which a simulation model was used as a teacher to identify test attributes automatically for the system to be diagnosed. The automated modeling system begins with a simulation model and evaluates tests in a nominal situation to determine the limits and tolerances on these tests. The system then sequentially fails the components in the system and reruns the simulation to determine which tests will fail. The results of these simulations define an attribute map for the system that becomes the basis for an information flow model to be processed by the system testability and maintenance program (STAMP) and the portable interactive troubleshooter, (POINTER). These tools function together and use an information flow model to assess system testability and diagnose faults