Fault identification through the combination of symbolic conflict recognition and Markov chain-aided belief revision

Fault identification is a search for possible behaviors that would explain the observed behavior of a physical system. During this search, different possible models are considered and information about the interaction between possible behaviors is derived. Much of this potentially useful information is generally ignored in conventional pure symbolic approaches to fault diagnosis, however. A novel approach is presented in this paper that exploits uncertain information on the behavioral description of system components to identify possible fault behaviors in physical systems. The work utilizes the standard conflict recognition technique developed in the framework of the general diagnostic engine (GDE) to support diagnostic inference through the production of both rewarding and penalizing evidence. In particular, Markov matrices are derived from the given evidence, thereby enabling the use of Markov chains to implement the diagnostic process. This work has resulted in a technique, which maximizes the use of derived information, for identifying candidates for multiple faults that is demonstrated to be very effective.