Studies On Several Key Problems Of Finite Element Model Updating Based On The Sensitivity Analysis

Finite element model updating is one of important research contents in bridge health monitoring system.As we all know,finite element model may be different from the real structure because of simplied premise,uncertainty of boundary conditions and effect of none-structure members.The calculated results are not very close to the experiment measured data.It leads to more difficult if we apply finite element calculated results to assess bridge condition and predict the residual life of bridge.Therefore,both the researchers and civil engineers pay more attention on the bridge finite element model updating.This dessertation has studied several key problems on finite element model updating by several modal tests.The test members include two simply-supported beam,one simply-supported and cantilever beam,one space truss bridge,respectively. Then Wuhekou cable-stayed bridge is updated using optimization method.The detail research contents of paper are as follows:Firstly,the sensor optimization localization has been studied because it is important to acquire the effective test data and make correlation analysis between the analytical results and test ones.Thus,two methods are adopted to optimize the sensor localization.One is effective independance (EI)method,the other is an united method presented by author.This method takes full advantage of MAC method based on QR decomposition and EI method.The advantage of united method can ensure not only the linear independence but also maximum space angle between two measured vectors.Then the program in Matlab7.0 is developed for both the EI method and the united method,and it is applied to realize the sensor optimization localization of Wuhekou cable-stayed bridge.Some factors such as different order shape to be monitered,rank condition of matrix and matrix value are considered.Secondly,the studies on several key problems of finite element model updating based on the sensitivity analysis are carried out using modal test data and an example,and several factors including dynamic character selection,parameter selection,and difference of globe parameter and local parameter are considered. To solve the under-determination equation set,Bayesian method is adopted. The study results show if we use modal frequency,modal frequency+MAC, modal frequency+modal shape to update the finite element model,modal frequency+MAC is best,then modal frequency and modal frequency+modal shape.The study results also show the quantity of parameter selection also has an important influence on model updating.But we can't draw conclusion that the more parameter is selected to update model,the better model updating quality is acquired.The research also shows that it is better to update finite element model using local parameter than using globe prarameter.Thirdly,the model updating based on frequency response function is carried out,considering frequency band,noise level and quantity of frequency response function.The study results demonstrate that both narrow frequency band and broad frequency band all can realize the model updating.But using narrow frequency band to update model can describe the minor details of frequency response function.Using broad frequency band to update model can also get model very close to test frequency response function as a...