Application of Entropy-Based Markov Chains Data Fusion Technique in Fault Diagnosis

This paper proposes an entropy-based Markov (EMC) chain fusion technique to solve the problem that the sample set is incompletion in fault diagnostic field. Firstly, the concept about probability Petri net is defined. It can calculate the fault occurred probability from incidence matrix based on the complemental information. Secondly, probability Petri net diagnostic model is designed from diagnostic rules that obtained by Skowron default rule generation method after the sample set is reduced by rough set theory. And in order to simplify the framework of the diagnostic model, Petri net model is designed as distributed form. Finally, depending on the diagnosis of distributed diagnostic model, EMC technique will be used to obtain consensus output if the places that represent fault in the model have several tokens. The diagnostic result is the consensus output that with the maximum of posterior probability after normalized treatment. The design is described by an example about rotating machinery fault diagnosis, and is proved availability by test sample set.