Abstract: In order to accurately evaluate the safety performance of bridge, an evaluation index system was established for safety assessment of existing reinforced concrete bridges based on the structure properties. Rating standards were developed for the evaluation indexes. RBF neural network was used to replace the traditional BP neural network, improving the learning speed and scope of application. A safety evaluation system was established for existing bridges based on RBF neural network and adaptive fuzzy inference. The system was used to evaluate the safety of a certain steel concrete bridge. The results show that a large number of experts evaluation data can provide sufficient input data for the evaluation system, and that the system after studying can imitate the actual evaluation results of experts quickly and efficiently, and can accurately evaluate the actual safety condition of the bridge.

Key words: reinforced concrete bridges, safety assessment, adaptive fuzzy inference, radial basis function(RBF) neural network, expert evaluation