Environmental Effects On Bridge Damage Detection And Its Elimination

Damage identification is an important part of Structure Health Monitoring(SHM)systemic problems.At present,the global detection technology based on structural vibratio n characteristics has become the core of the SHM field.Process the vibration data of the structure monitored in real time,extract the characteristic parameters or dynamic fingerprints,and use the difference between the characteristic parameters of the non-destructive state and the lossy state to judge the structural state.However,the engineering structure is under complex environmental factors and operating conditions,including temperature,relative humidity,and varying loads.The influence of these factors often overwhelms the changes in structural characteristic parameters.If you ignore the existence of this effect,it will inevitably lead to misjudgment and missed judgment of the structural state,making the traditional damage identification method lose its effect in practical applications.The specific research contents of this article are as follows:(1)There are various types of environmental factors,and each environmental factor has different impacts on dynamic fingerprints.This article first analyzes the relationship between the modal frequency of an actual y monitored bridge and the temperature,wind speed,and traffic volume.Study the influence of each environmental factor on the modal frequency;then,a polynomial model is used to fit the relationship between the frequency and several environmental factors,and a polynomial model of the frequency and environmental factors in the non-damaged state is established.The modal frequency data was reduced to simulate six damage conditions,using polynomial model residuals for damage identification,and using principal component analysis and factor analysis to compare the effect of polynomial model in damage identification for separating environmental factors.(2)Damage identification based on separation of environmental factors based on kernel principal component analysis.Principal Component Analysis(PCA)has a better effect in processing the linearly correlated feature parameters,but the actual monitoring data often has a non-linear relationship between the variables,and it is difficult for the linear method to extract the non-linear information.In order to overcome this shortcoming of PCA,this paper combines PCA and Kernel Analysis method and proposes a method based on Kernel Principal Component Analysis(KPCA).According to the orthogonality of the secondary component in kernel space and environmental factors,it is proposed to use the secondary component in kernel space as a new feature parameter for damage identification.First,using the feature that the feature quantit y is more easily linearly separable in the high-dimensional space,assuming that there is a nonlinear mapping,the non-linear data in the low-dimensional space is mapped to the highdimensional space,and the PCA analysis is performed in the high-dimensional space;then the kernel regeneration Principle and kernel function,construct the nuclear space in the highdimensional space,and apply PCA in the nuclear space.The effectiveness of KPCA in handling variable nonlinearity is verified by the experimental data of Yonghe Bridge,the experimental data of wooden truss bridge.(3)Dynamic damage identification based on moving window-reduced kernel principa l component analysis.KPCA is a static modeling method.The static model needs to be trained with as many data sets as possible including all environmenta l factors within a period of time,which makes the kernel matrix dimension and calculation complexity in the modeling process very high,and the actual monitoring of early data is difficult to meet such requirements.In addition,the actual monitoring data often exhibits nonlinear and nonstationary dynamic changes.The static model based on early monitoring data is difficult to describe the normal fluctuation of the monitoring data caused by new environmental factors,which may lead to misjudgment.In view of the shortcomings of the static KPCA method,two methods of handling dynamic nonlinear nonstationary processes in the industrial field are introduced to separate environmental factors for structural health monitoring of civil engineering structures.Introduced Reduced Kernel Principal Component Analysis(RKPCA)to improve the defects of KPCA kernel matrix with high dimensions and high computational complexity;introduced a moving window method to carry out the old model based on the new data dynamica l ly monitored update,the improved model cannot describe the normal fluctuations caused by environmental factors in the new data,and it is easy to cause defects of damage misjudgme nt;in addition,the introduced sliding window method proposes a new model update criterion,which reduces the number of model updates and improves dynamic monitoring.The calculat io n efficiency is convenient for online monitoring.The effect of the improved KPCA in dynamic monitoring was verified by comparison between the wooden truss bridge experiment and the standard model data of the Z24 bridge.