Drum: Reasoning without Connicts and Justiications

In this paper we describe the most important concepts of the belief revision system DRUM and show its use in model-based diagnosis. DRUM works directly on (logical) models and uses neither connicts nor justi-cations. The system description plus observations are seen as integrity constraints, while the inferred facts as well as assumptions about behavioral modes are revised whenever they connict with an integrity constraint. To reason eeciently on large examples, DRUM uses a combination of static and dynamic inconsistency detection methods, which are triggered selectively by revising or updating the knowledge base by new information. Our experiments so far have shown that DRUM is easily scaleable to thousands of components without problems and thus represents an interesting alternative to existing systems based on connicts and justiications.