A New Approach for Construction of Diagnostic Bayesian Network

Bayesian network is one of the most effective theoretical models applied in the fault diagnostic decision. In order to realize the computer-aided construction of the Bayesian network for diagnostics, this paper proposes a simple, yet comprehensive failure knowledge representation model combined with the structure decomposition of the complex system. Using the polychromatic sets theory, the rock-bottom frame of this model is described in unified mathematical language by means of defining the failure modes as the inherent colors of the physical entities. Then, the Bayesian network is generated automatically with the iterative search process operated on the reasoning matrices of the polychromatic sets. A case study is presented to illustrate the validity of the proposed approach. The research shows that the polychromatic sets approach to the diagnostic Bayesian network (DBN) construction has formed mathematical foundation and can be readily expressed and operated by computer.