Study On Reasonable Construction State And Parameter Indentification Of Long-span Asymmetric Composite/Hybrid Grider Cable-stayed Bridge

The geometry and internal force state of long-span asymmetric composite/hybrid grider cable-stayed bridges are closely related to the construction process.Taking a Yangtze river highway bridge as the engineering background,starting from structural and mechanical characteristics of bridges,aiming at problems encountered in construction control work,reasonable construction state,random multi-parameter sensitivity analysis and parameter identification in the construction process of long-span asymmetric composite/hybrid grider cable-stayed bridges are studied.Firstly,reasonable construction state is studied.According to requirements of construction control and reasonable construction state,six optimized construction methods which can accelerate construction speed are proposed on the basis of the traditional construction method of composite girder.Finite element models which established by finite element software NLABS is used to calculate,analyze and compare the cable force,geometry,and internal force in completion state,stress and anchor cable tension amount in construction process.According to analysis results,two optimal construction methods is protocoled.Finite element models are established to confirm that the cable force,geometry and internal force in completion state of two optimal construction methods are close to the calculation results of traditional construction method,and the anchor cable tension amount of two optimal construction methods can meet the requirements of overall tension.At the same time,two optimal construction methods are shorter than traditional construction method in average construction time of each standard grider,and the whole construction process of them is safe.It can be seen that both of two optimal construction methods can simultaneously meet the requirements of construction efficiency of main girder,structural stress safety and feasibility,uniformity and safety of cable tension.Then,one of optimal construction methods is selected to carry out random multiparameter sensitivity analysis in the construction process of long-span asymmetric composite/hybrid grider cable-stayed bridges.By uniform experiment design,according to structural characteristics of construction process,the construction stages of double cantilever on the north bank,single cantilever on the north bank and single cantilever on the south bank are selected as study objects.The geometry,cable force and tower deviation variation in the construction process are selected as target response.The stiffness and weight of bridge decks,steel beams and stay cables,and the cable force of stay cable in construction process are selected as control parameter.According to reasonable range of control parameter variation,test level is divided.After normalizing data,uniform test is designed according to the uniform design table.The response of each control parameter combination is calculated by software NLABS.After normalizing calculation results,stepwise regression analysis with software SPSS is used to conduct the parameter significance test,and the regression model of control parameters related to each target response is established.According to regression results,the sensitive parameters,relatively sensitive parameters and insensitive parameters in the construction process of long-span asymmetric composite/hybrid grider cable-stayed bridges are determined.Finally,aiming at the sensitive parameters in the construction process of asymmetric composite/hybrid grider cable-stayed bridges,parameter identification in the construction process of asymmetric composite/hybrid grider cable-stayed bridges is studied.According to the characteristics and basic principles of BP(Back Propagtion)neural network and RBF(Radial Basis Function)neural network,two kinds of neural network are respectively established.The application of training data and test data proves that the training effect of the two kinds of neural network is relatively ideal,but the training accuracy of RBF neural network is slightly better than that of BP neural network.Using the trained RBF neural network,parameter identification of the test data in the actual construction process of the engineering background bridge is carried out.Compared with the data before parameter update,error is obviously improved,and updated finite element model is closer to actual response of the structure.