Bayesian Network Algorithm Based on Dynamic Dependency and Multi-source Information Nodes

Aiming at the problem of target decision error caused by the deviation and lack of the fault information sources in a complex equipment, we propose a Bayesian diagnosis network based on the dynamic fusion of multi-source information nodes and node dependency adjustment. First, we carry out the multi-attribute decision analysis of the multi-source static information through a real-time dynamic information source, and adjust the priori probability of the node according to the evidence fusion rule to make it close to the actual probability, so that the diagnosis error is reduced. Then we obtain the dependency similarity by comparing the results of different Bayesian network diagnosis, and dynamically adjust the dependence between nodes to reduce the effect of the missing information source. Experimental results on the Weifang Diesel Engine R6105AZLD show that using the proposed method, the accuracy of fault diagnosis is improved by 21%, and the iterative error representing robustness is reduced to 0.01.