On-line Model Updating And Validation Of Cable-stayed Bridges Based On Health Monitoring System

Finite element model updating is one of the hot research directions in the field of bridge health monitoring.An accurate benchmark finite element model can provide a good foundation for fine analysis of dynamic characteristics of bridge structures or for modelbased safety assessment and damage identification.With the continuous accumulation of massive health monitoring data,the demand for data processing efficiency is increasing.Timely acquisition of bridge status information means that the update cycle of finite element model needs to be accelerated.Based on the health monitoring data of Guanhe Bridge in Shenhai freeway,this paper attempts to establish an online frequency identification method for long-span cable-stayed bridges,which can be further applied to online model updating and validation of large-span cable-stayed bridges.The main research work of this paper includes:(1)Based on the stochastic subspace identification of covariance,the automatic identification method of modal parameters based on offline data is studied on eliminating false modes.(2)A recursive on-line identification method of instantaneous frequency based on least square method is explored.According to the relative noise reduction ability of NEx T method,a multi-degree-of-freedom on-line frequency identification method based on free attenuation signal is studied.Finally,it is applied to the on-line frequency identification of Guanhe Bridge.(3)The uncertainty quantification and transmission of temperature,vehicle load,wind and modal frequency of main environmental factors of cable-stayed bridges are studied.Based on the analysis of the mechanism of environmental factors affecting modal frequencies and single-factor statistics,the multivariate linear model based on PCA and ANN model based on NLPCA are used to carry out multivariate regression analysis,and finally the modal parameters are revised.(4)A two-stage method suitable for online model correction and confirmation of cablestayed Bridges was studied.In the first stage,the interval response surface method based on sensitivity analysis was used to establish the uncertain reference finite element model by combining the offline data within a time scale.In the second stage,the online model of Guanhe bridge is revised and confirmed by using the deterministic model correction method based on the online frequency identification results.The main conclusions are as follows:(1)Based on the modal parameter identification method of environmental vibration,an automatic modal parameter identification method based on stochastic subspace method is further established.The results of modal parameter identification of Guanhe Bridge show that the frequency variation caused by the variability of the operating environment of Guanhe Bridge in one day can be accurately identified by this method.(2)An improved four-parameter method for on-line frequency identification of single-degree-of-freedom system and multi-degree-of-freedom system is proposed.Compared with HHT method,the improved four-parameter method is more stable in instantaneous frequency identification and can be used to predict the time-varying characteristics of modal parameters of time-varying structures.The mean value of modal parameters of time-varying structures can be identified by least square fitting of phase angle.The results of identification agree well with those of HHT and PP methods.This method can be applied to the online frequency identification of Guanhe Bridge.(3)The influence of ambient temperature on the modal frequencies of the cable-stayed bridges is basically the same,and they have good negative linear correlation.The influence of seasonal temperature difference at 40? on the modal frequencies of Guanhe Bridge is about 1.3%.On the time scale of one week,the influence of vehicle load and ambient temperature on the dynamic characteristics of Guanhe Bridge is basically the same,and the effect of one day on the modal characteristics of Guanhe Bridge is generally less than 1%.The PCA-based multivariate linear regression model and NLPCA-based ANN regression model can better transfer the environmental factors and modal frequencies of Guanhe Bridge,reduce the cognitive uncertainty of measured frequencies,and have a better correction effect on the quantification of modal frequencies in a short time scale.(4)Based on the optimal Latin hypercube experimental design method and multi-island genetic algorithm,an interval response surface model updating and validation method based on sensitivity analysis is established.The validated model reflects the uncertainty of model parameters in the past period,and the average relative error of target frequency does not exceed 0.5%.It can be further used as a reference finite element model for on-line model updating and validation.(5)On the basis of the fitted interval response surface,according to the on-line modal frequency identification results of Guanhe Bridge,on-line model updating and validation are carried out based on the reference finite element model.The results show that the maximum error of online model updating is less than 1%.The uncertainty of the parameters to be amended is greatly reduced.The reliability of the updating results is improved,and the shorttime scale model updating and validation are realized.