Abstract: Aiming at the problem that the failure data or life data of the whole line of the flexible production line of aviation structure parts have the characteristics of small sample, and the performance state of its subsystem is difficult to define accurately, an effective fuzzy Bayesian network evaluation method based on multi-dimensional information fusion of components was proposed.In this method, the subsystem experiment data, on-site operation and maintenance data and similar system maintenance data were deeply fused by analytic hierarchy process. The fuzzy set theory was used to judge the various operation states of components, and the Bayesian network model of the whole production line with multiple states was constructed, which broke through the integration of subjective and objective information of Bayesian network based on hierarchy criteria. The application in a flexible production system of aviation structural parts shows that the forward analysis can make full use of multi-source fusion information to evaluate the reliability of high confidence operation, and the reverse analysis can locate the weak parts. It improves the efficiency of system operation reliability analysis and provides reference for condition based maintenance.