Application Of ANFIS(Adaptive Neuro-Fuzzy Inference System) In Bridge Condition State Evaluation

Bridge management system(BMS) can manage numerous amount bridge's components in road net effectively, maintain safety and transportation function of bridge, and assist decision maker with the most appropriate maintenance project. at present, BMS has developed dramatically in developed country, but the relative research work is in process of underway stage in out country, so conducting relative theory research and practice of BMS has significant practical and economic importance in our country.The whole BMS has three basic function: status evaluation, degradation forecasting and optimizing maintenance plan, of which status evaluation is the foundation of the other two, and is the most important components in BMS. The major research content in this paper is the theories and methods of bridge's condition state evaluation, and make a profitable attempt for creating the genuine Bridge Management system aftertime.In this paper, existing methods of bridge condition state evaluation were compareded synthetically; a deep study on multi-factors that can influence the performance of bridges in reinforced concrete short/medium span bridges was given. An evaluation index system for durability and load-carrying capacity of existing reinforced concrete short/medium span bridges was established. The rating standard of evaluation index was defined, and a durability evaluation system for main girders of existing bridges based on ANFIS was constructed.A method to simulate the data of bridge inspection and expert attitude investigation based on statistics method was brought up, and the feasibility of this method by some instances was validated. In this system, the fuzzy inference process used by experts to evaluation bridges was implicit in the connections and weights of the neuro-fuzzy networks, which successfully avoid the neural networks becoming a "black box" . Furthermore, this system has the ability of learning that overcame lack of learning ability, which is a disadvantage of traditional fuzzy inferencesystem.We testified multi-performance of the system using many instances. The results demonstrated the agreement between evaluation results from the system after training and bridge experts, the system had well learning ability and practicality, could imitated the nonlinear fuzzy inference process of bridge experts fast and accurately, had good application foreground.