In order to accurately predict the potential safety hazards of the Yellow River diversion tunnel, a fuzzy Bayesian safety evaluation model based on fuzzy set theory and subjective Bayesian method is proposed. Combined with the actual inspection data, this model uses Bayesian Network(BN) to process fuzzy information and uncertain information, fuzzies polymorphic fault events from the inspection data, and then finds out logical relationship among different fault events through conditional probability table of Bayesian network. Finally, the overall evaluation of the headrace tunnel safety model is obtained, and the feasibility of this method is verified by example analysis of the headrace tunnel safety evaluation model.