A method of data fusion system for fault detection based on model-based diagnosis and evidence theory

This paper applied first Principles that is brought forward by Reiter etc, to obtain the conflict set and hitting set. Based this result, it introduce date fusion system with D-S evidence theory to solve the problem of detecting multi-fault through multi-symptom of these faults. The method is to seek for conflict set candidates through SD (system description) of the device off-line, and then discerns minimal conflict set and minimal hitting set on-line through the action data of the device based on the fault symptoms. Afterwards, it assigns prior probability to each possible fault component and process data fusion on the set of possible fault components using D-S evidence theory. Finally the probability order of the possible fault components is proposed.