The invention provides a large-scale bridge network evaluationmethod based on a Bayesian network. ORDER-II and Dijkstra algorithms and vulnerability thoughts are adopted. The technical status rating of bridges in the network and the information of distance between cities are used to evaluate overall reliability of a large bridge network and recognize bridges at key road sections. The NP-hard problem for solving the reliability of large-scale bridge networks is transformed into the most probable failure combination and network connectivity state of the bridge under the required accuracy. By evaluating the bridge network in a national highway containing 1772 bridges, a proof is given that the method and vulnerability indictors can evaluate the overall reliability of the large-scale bridge network. Relative importance of all road sections is effectively recognized. The large-scale bridge network can be evaluated directly and efficiently.