Research On Girdge Bridge Damage Identification Method Based On Support Vector Machines

I In the process of bridge operation, many damages and local failures occurred due to the frequent load, overload and erosion of natural environment as well as human disturbance such as traffic accident. To assure the use safety and durability of the bridge, it is necessary to carry out damage identification to bridge structure. Firstly, in this paper, we make some research to fix the Shihe bridge sensor well, which will provide data for bridge damage identification. Secondly, we put forward to use the support vector method to make Shihe bridge damage identification, and we get good identification performance. Several aspects of the Research work are listed as following:Firstly,we fix measurement sensor in the bridge, so this can support base for identification. In the paper, we put forward optimal layout research of the static sensor basing on damage identification theory, and we write the program to realize the method. In the last, we test validity and superiority of the method by applying the method to Shihe bridge.Secondly, in this paper, we put forward to apply the support vector to the Shihe bridge structure damage identification. We make contrast between the later simulator data and former data, and compared to the former structure damage identification prediction precision data,the data using the support vector method increases by12.5%.Finally, in the selection of support vector core function and parameter relevant optimization, we make contrast in prediction precision that uses different core function, and we get the conclusion the prediction precision using the RBF core function to training group and test group is more accurate than other core functions, such as linear core function, polynomial core function, Sigmoid core function, at the same time, the model about the RBF core function is optimization. We optimize the support vector with crossing testing method, and we get the optimization parameter. We test the validity of support vector method in the bridge damage identification with the simulator and experiment result.