Damage Diagnosis Of Bridges Monitored Within One Cluster Under Time-varying Environment

The bridges located at local road network or urban ring road are defined as the bridge cluster.Different to single large-scale bridge,the bridges belonging to one cluster show the following three characteristics: i)the available data includes both the measured data obtained from the structural health monitoring(SHM)system of bridges and the measured data acquired by the load test of bridges without SHM system set up;ii)the data processing has to be implemented for every bridge which structural form is different to others among one cluster;iii)data correlation exists among the bridges which structural forms are similar with each other.Therefore,it is challenged to diagnose the structural condition of all the bridges belonging to one cluster by fully utilizing the special characteristics of the measured data obtained from these bridges.To address the above issue,this study focuses on the damage diagnosis of bridges under the time-varying environment from three factors,i.e.,the bridges set up with SHM system whose structural form are similar with others among one cluster,the bridges set up with SHM system which structural forms are different to others among one cluster,the bridges without SHM systems.The main contents of this study are described as following.Aiming at the issue how to establish the baseline finite element model(FEM)available for damage diagnosis of bridges under time-varying environment,the concept of probability baseline FEM of bridge was proposed.First,the distribution regulation of measured data of natural vibration frequencies of bridge under time varied environment was investigated,and the cluster analysis based on the Gaussian mixture method(GMM)was presented for classification of the massive data of natural vibration frequencies of bridge.Second,for each cluster,on the basis of stochastic FEM updating,the method of building the probability baseline FEM of bridge was proposed by importing the Kriging model.Third,with the theory of novelty detection,the index for damage diagnosis was proposed,and this index was implemented to diagnose the damage of the bridges set up with SHM system which structural forms are different to others among one cluster.The above work covered the shortage of traditional FEM technique unable to describe the regulation of massive data obtained from SHM system.To generate the data baseline model suitable for damage diagnosis of bridges under operational environment,the steady state data baseline model of bridge was proposed.First,with the singular value decomposition and reconstruction of matrix,the effect of measured noise was mitigated by using the method based on the reconstruction of structural strain of bridges.Second,the method of extracting the trend of measured data of structural strain was proposed by analyzing the influence regulation of varied traffic load on the measured data of structural strain of bridges,and the effectiveness of the presented method was demonstrated.And then the method of establishing the steady state data baseline model based on K-mean cluster analysis was proposed.Third,on this basis,we extended the method of extracting the trend of measured data of structural strain for diagnosing the damage of bridges.This method is suitable for the bridges set up with SHM system which structural forms are different to others among one cluster.The above work alleviated the effect of variation of operational environment on the data baseline model.Facing the issue how to diagnose the damage of bridges using the characteristics of all the bridges acted by the similar environmental load,the method without the data obtained by the reference state of bridge was proposed to diagnose the damage of the bridges monitored within one cluster.First,a cluster analysis algorithm considering the similarity among the strain monitoring data at different measured points is proposed to classify the strain monitoring data for all the bridges monitored within one cluster.Different classes of strain monitoring data are established by identifying similar probability distribution patterns.Second,for each class,a damage detection index is proposed based on the residual between the cumulative distribution Functions(CDF)of strain monitoring data.All the bridges monitored within one cluster are acted upon by similar environmental temperatures during the same monitoring period;hence,the effects of the environmental temperature on the proposed index are mitigated indirectly during the same monitoring period.Thus,the proposed index is implemented to effectively detect the damage in all the bridges belonging to each class.The above work avoided the dependence of the data obtained by the reference state of bridge.For diagnosing the damage of bridges without SHM system among one cluster,the strategy of fast damage diagnosis was proposed.First,on the basis of the load test of bridge using the quasi-static influence lines,the method of fast damage diagnosis based on the quasi-static displacement influence line was proposed,which kernel idea attributed to the theory of null space of matrix.Second,the key issues were discussed such as the inconsistent of loading vehicle before and after damage,robust and resistance to noise of the proposed method,and the effectiveness of the proposed method was verified.Third,considering the Brillouin optical time domain analysis technique,the method was extended to deal with the high density of measured points of strain of bridges,and then the method based on quasi-static strain influence lines was presented.The above methods were combined with the methods proposed previously in order to diagnose the damage of all the bridges belonging to one cluster.