Speeding Up Model-based Diagnosis by a Heuristic Approach to Solving SAT

Model-based diagnosis of technical systems requires both a simulation machinery and a logic calculus. The former is responsible for the system’s behavior analysis, the latter controls the actual diagnosis process. Especially when pursuing qualitative simulation, it makes sense to realize the simulation machinery with a logic calculus as well. Say, a qualitatively described hypothesis can directly be mapped onto an instance of the well-known SAT problem. Likewise, an entire diagnosis process, i. e., a sequence of hypothesis refinements, represents a set of SAT problems. This paper reports on the operationalization of such a SAT-based diagnosis approach. A specific characteristic here is the idea to exploit an ordering of the logical formulas according to their likeliness of being satisfiable. This idea is new in the context of qualitative reasoning, and it leads to a considerable speed up of the diagnosis process. Its applicability has been evaluated in the domain of hydraulic circuit diagnosis.