Model Updating For Suspension Bridge's Finite Element Model Based On Sensitivity Analysis

With the continuous improvement of finite element analysis software, the results of finite element analysis are always used for controlling and predicting the large-span bridges' construction. Engineers always use the theoretical value of parameters to create the finite element model, but actually, the values of many parameters are different from their theoretical value, which makes the initial finite element model can not reflect the real structural behavior, then leads deviation from the original theoretical goal. Therefore, it needs finite element model updating to ensure the results of analysis reflect the construction process accurately and achieve the goal of construction control. This thesis takes Balinghe Bridge as example, mainly studies the following area.First, the different thesis of construction control calculation methods are briefly introduced, those methods application scope is investigated, tangent displacement method when establishing model is emphatically introduced, and Balinghe Bridge forward analysis model is established on this base. Then, by comparing the results of forward analysis model with backward analysis model, it shows that the method proposed in this thesis can be applied in the forward construction analysis of this bridge.Secondly, on the basis of forward finite element model, the sensitivity of bridge's design parameters is analyzed, then reaches the conclusion that main cable stiffness and its density, weight per unit length of girder are the main design parameters. This conclusion has some reference value for other long-span suspension bridges'parameter sensitivity analysis.Last, the neural network toolbox of MATLAB is used for establishing the neural network of model updating, and design samples by using uniform design method, then train the network by the results of samples, and then the measured main cable total displacement is used for updating the forward finite element model. The results of updated model are anastomosis with measured results.The forward analysis method of large-span steel truss stiffening girder suspension bridge is studied, and this model is updated by using neural network method, so the model can truly reflect the structural behavior. Therefore, the method proposed in this thesis has some reference value for the long-span suspension bridge construction control.