Research On Reliability Analysis Methods Of The Existing Bridges Based On Monitoring Data

Bridge health monitoring systems have accumulated a huge amount of monitoring data in the long-term service periods.How to assess and predict the reliability of the existing bridges with these data,has become one of the main scientific problems in the field of structural health monitoring.In this thesis,the single monitoring point is considered as the bridge component,and the system composed of the multiple monitoring points is considered as the bridge system.Based on the monitoring data,reliability analysis methods of the existing bridges are thoroughly studied.The main contents are as follows:(1)The dynamic reliability prediction method of bridge components is proposed through the fusion of dynamic monitoring data and the particle filter algorithm.Firstly,the moving average method and the least square method are used to decouple the coupled load effects,and the dynamic linear models of decoupled high/low frequency load effects are respectively established;secondly,the particle filter algorithm is adopted to predict the decoupled high/low frequency load effects with high accuracy,and the hybrid particle prediction of coupled load effects is realized by adding the predicted results of high-frequency and low-frequency load effects and the constant load effects caused by the dead weight of the bridge.finally,with the first order second moment method,the time-variant reliability prediction of bridge key components can be made,and an engineering example is provided to illustrate the feasibility of the proposed method.(2)The dynamic reliability prediction method of a bridge system is proposed through the fusion of multivariate Bayesian dynamic linear models and Gaussian copula functions.Firstly,with the monitoring data,the multivariate Bayesian dynamic linear models of load effects for the few monitoring points are built,further,with Bayesian updating and prediction theory,the dynamic load effects of the few monitoring points and the time-variant correlation coefficients among the few monitoring points are predicted;secondly,the Gaussian copula function is adopted to build the time-variant nonlinear correlation model among the failure modes of the few monitoring points,further,the time-variant reliability prediction of bridge systems considering the dynamic nonlinear correlation among the failure modes of the few monitoring points can be made;finally,an engineering example is provided to illustrate the feasibility and application of the proposed method,and the results show that the predicted failure probability of the bridge system considering the nonlinear correlation of failure modes is lower than that without considering the correlation of failure modes,the predicted reliability of the bridge system without considering the nonlinear correlation among the failure modes is more conservative,which indicates that for time-variant reliability prediction of the bridge system,it is necessary to consider the nonlinear correlation of the failure modes.(3)The dynamic reliability prediction method for bridge systems is proposed through the fusion of Bayesian dynamic linear models and the Vine-Copula theory.Firstly,the C-Vine Copula models and D-Vine Copula models are introduced to describe the nonlinear correlation among the failure modes of the multiple monitoring points,further,the Bayesian dynamic vine copula model of load effects at the multiple monitoring points is built through the fusion of the Vine-Copula technology and Bayesian dynamic linear model recursive processes;then,with the first order second moment method,the time-variant reliability prediction of bridge systems can be made;finally,an engineering example is provided to illustrate the feasibility of the proposed method,and the results show that the predicted results of the proposed method are more reasonable than that without considering the dynamic nonlinear correlation among the failure modes.(4)The time-invariant reliability analysis method of the bridge system is proposed based on optimal R-Vine Copula models.In order to solve the ergodicity problem of CVine Copula models and D-Vine Copula models,the optimal R-Vine Copula models are established to describe the nonlinear correlation among the failure modes of the multiple monitoring points,further,the time-invariant reliability of the bridge system is thoroughly studied,and the feasibility of the proposed method is illustrated by an engineering example.The above research results will provide theoretical foundation and application methods for the safety assessment and preventive maintenance decision-making of the existing bridges.