Research On Stayed-cable Bridge Damage Identification Based On The Rotation Mode Analysis Of Wavelet-Neural Network Analysis

As the number of bridges is increasing In China, bridge health monitoring is becoming increasingly important, and identifying the location and extent of the possible damage to the bridge structure in an effective and rapid way becomes particularly important.Wavelet analysis is known as signal microscope of mathematical analysis, it has a good time and frequency localization characteristics. The neural network has good self-organization, self-learning, adaptive capacity, in signal processing, and it has highly nonlinear mapping ability. Therefore, combine their advantages, effectively identifying the damage location and extent of cable-stayed bridge by wavelet analysis and neural networks, has important theoretical significance and application value.This paper established a cable-stayed bridge damage identification principle, by finite element analyzing of the cable-stayed bridge, by continuously wavelet transforming the extracted modal parameter, wavelet coefficients can be obtained. The structure damage located by wavelet coefficients modulus maxima. Using the first four natural frequencies of the damaged cable-stayed bridge as the input parameters of neural network to build a neural network model, then train and test the sample data, hence enabling the identification of the location and extent of the damage to cable-stayed bridge.On the basis of finite element analysis to extract the angle and curvature of the modal parameters, through the research of single-tower cable-stayed bridge damage and multiple injuries, make the modal parameters continuously wavelet transform, then based on the figure of the mold maxima to identify the damage location. Using finite element analysis extract the first four modal frequencies, then build Two different structural damage training sample at 20%, 30%, 40% and 18%, 28%, 38%, use the network calculation and analysis of the structure of the BP neural network can calculate that, the error of the output in the training sample data are within 5%, the error of the test is within 5% too.This paper studies the cable-stayed bridge's single damage and multiple damages identification, by establishing the finite element model of damaged cable-stayed bridge, analysis the finite element it dynamic characteristic to get the modal parameter, and then to carry on the continuous wavelet transform to effectively identify the damage location. Using finite element analysis extract the first four modal frequencies, then build Two different structural damage training sample at 20%, 30%, 40% and 18%, 28%, 38%, use the network calculation and analysis of the structure of the BP neural network can calculate that, the results can satisfy the requirements of engineering precision.In this paper, the established wavelet neural network algorithm is not only able to identify the damage position of the main girder of cable-stayed bridge, but also can effectively identify the damage degree of the main girder of cable-stayed bridge. The method of damage detection of cable-stayed bridge has certain guiding significance to the engineering application.