Research On Finite Element Model Updating Of A Through Tied Arch Bridge

The finite element model of the structure which has the advantages of high efficiency and low budget can be used in static and dynamic analysis,damage location and quantitative identification,bearing capacity evaluation,structural optimization design and reinforcement scheme selection,etc.But the initial finite element model of the structure can not completely reflect the actual conditions of the structure.The finite element model updating is to adjust the parameters of the model according to the static and dynamic response of the structure,so that the modified model response is closer to the actual structural response,and thus a more precise finite element model can be obtained.Few studies have been carried out to modify the finite element model of through tied arch bridges with single support plane.The research of the through tied arch bridge with single supporting plane(The net span is 247.0807 m and the height is 55.5 m),which has the largest span in the same type of bridges in China at present,is of typical significance to the correction of the finite element modeling of the bridge.The bridge was modeled by ANSYS finite element software.The modal parameters of the bridge are obtained by modal testing on site.Then,the modal frequencies of typical modes are selected and the finite element model of the bridge is modified by the radial basis function neural network method and the response surface method,respectively.The design parameters,neural network and response surface model,which are of universal significance to this type of bridge,are obtained.The main works are as follows:(1)The background and significance of this study are elaborated,and various methods of finite element model updating and their advantages and disadvantages,as well as the application status demestic and abroad are summarized and analyzed.Taking the through tied arch bridge with single supporting plane as the research object,the finite element model of the bridge is established by using ANSYS.The static analysis of the model is carried out under the conditions of self-weight and design load,and the internal forces such as deformation,bending,shear and axial force are obtained.The first ten modal frequencies,modes and dampers are obtained by structural dynamic analysis.(2)The step-by-step cumulative method was used to optimize the layout of bridge measuring points,17 measuring points were arranged on both sides of the bridge deck.Moving point method was used for field modal testing.The self-reciprocal power spectrum modal parameter identification method was used to identify the modal frequency,mode shape,damping and other parameters of the bridge.Comparing the measured modal parameters with the calculated modal parameters,the deviation between them is confirmed.(3)The finite element model of the bridge is modified by the radial basis function neural network method.Firstly,the material density and elastic modulus of the model are selected as design parameters by sensitivity method,and uniform design method is used to set up parameter samples.Then,the modal frequency is taken as the input sample of the neural network,and the design parameters are taken as the output sample.The training of the neural network is completed by MATLAB.Finally,the model of the bridge is modified by the trained neural network.(4)The second-order incomplete polynomial response surface model updating method is used to modify the finite element model of the bridge.Firstly,the functions of response surface models such as first order polynomial,second order complete polynomial and second order incomplete polynomial are compared.Then,the sample data determined by the neural network are fitted with the undetermined coefficients of the second-order incomplete polynomials by the least square method,and the accuracy is tested.The objective function based on frequency is set up,and the modified optimal design parameters are obtained by quadratic sequence programming.Comparing the model updating effects of the two methods of RBF neural network and response surface,it is found that both the methods can improve the accuracy of the finite element model.