Proceedings of the Canadian Artificial Intelligence Conference, May 21-24,1996., Toronto, Ontario, Canada
The paper addresses the problem of automated model acquisition through the re-use of fault knowledge. The Diagnostic Remodeler (DR) algorithm has been implemented for the automated generation of behavioural component models with an explicit representation of function by re-using fault-based knowledge. DR re-uses as its first application the fault knowledge of the Jet Engine Troubleshooting Assistant (JETA). DR extracts a model of the Main Fuel System using real-world engine fault knowledge and two types of background knowledge as input: device dependent and device independent background knowledge. To demonstrate DR's generality, it has also been applied to a coffee maker fault knowledge base to extract the component models of a full coffee device.