Diagnosis with Incomplete Models: Diagnosing Hidden Interaction Faults

This paper extends model-based diagnosis (MBD) (de Kleer and Williams 1987; Reiter 1987) to systems with hidden interaction faults. An interaction fault is present if an interaction among a set of components leads to an observable failure, even though each individual component individually meets the specifications. A naive approach to address interaction faults is to simply account for all possible interaction faults in the system model. However, the naive approach presumes that all possible faults, both component and interaction faults, are known and addressed in the model. This assumption is violated by most real world systems, such as shorts in circuits (Davis 1984) or unmodeled connections (de Kleer 2007). That leads to incomplete system models, hence possibly hidden interaction faults. The problem of hidden interactions has been known for a long time (Davis 1984), but until now no general solution has been proposed. Instead of pushing for complete models (Preist and Welham 1990) or relying on additional structural information (Davis 1984; Bottcher 1995; de Kleer 2007) we approach the challenge differently. We allow system models to be incomplete and introduce a general, domain independent extension to model-based diagnosis to account for resulting hidden interaction faults. This extends model-based diagnosis to systems with incomplete models, in particular to models with incomplete structural information. In the paper, we demonstrate the proposed diagnosis framework on a logic circuit with a hidden interaction fault.