Abstract:

The multiple fault diagnosis problem is important, since the single fault assumption can lead to incorrect or failed diagnoses when multiple faults occur. It is challenging for continuous systems, because faults can mask or compensate each other’s effects, and the solution space grows exponentially with the number of possible faults. We present a qualitative approach to multiple fault isolation in dynamic systems based on analysis of fault transient behavior. Our approach uses the observed measurement deviations and their temporal  orderings to generate multiple fault hypotheses. The approach has polynomial space requirements and  prunes diagnoses, resulting in an efficient online fault isolation scheme.

Download: DaigleEtAl-AAAI-2007.pdf