Fault isolation in discrete event systems by observational abstraction

We propose a method for fault isolation in discrete event systems such as object oriented control systems, where the observations are the logged error messages. The method is based on automatic abstraction that preserves only the behavior relevant to fault isolation. In this way we avoid the state space explosion, and a model checker can be used to reason about the temporal properties of the system. The result is a fault isolation table that maps possible error logs to isolated faults, and fault isolation thus reduces to table lookup. The fault isolation table can also be used as an analysis tool at the design level to find both faults that cannot be isolated as well as redundant error messages.