Research On Structural Identification Of Small And Medium Bridges Based On Big Data Of Vehicles And Bayesian Theory

In recent decades,with the development of China’s economy,people’s demand for transportation has been increasing day by day.The increasing traffic load has posed a challenge to the safety performance of the existing bridge structures.Meanwhile,most of the existing small and medium-sized bridge structures do not have the conditions to install expensive bridge health monitoring facilities.It is necessary to establish a reliable,efficient and economic way to quickly detect and evaluate such bridge structures.As an important index reflecting the performance of bridge structures,bridge influence lines are widely used in bridge performance evaluation,bridge damage identification,bridge weigh-in-motion(BWIM)and other aspects.For small and medium-sized bridge structures that lack the support of BWIM system,the commonly used bridge influence line identification methods often have the disadvantages of closed traffic,high sensor maintenance cost,and short timeliness of identification results.Therefore,aiming at the current development status of small and medium-sized bridge structure health monitoring and the development trend of digital and intelligent bridge,this paper proposes a contactless small and medium-sized bridge structure identification method based on vehicle big data and Bayesian theory.This paper has carried out the relevant research in the following aspects:(1)In theory,this paper combines the unit influence line superposition method with the Bayesian parameter identification method,and establishes the bridge influence line inversion model based on the Bayesian theory.Under the condition that the vehicle parameter distribution in the vehicle database and a large amount of bridge dynamic response data sum are known,the posterior distribution of the ordinate value of the bridge influence line is obtained by using the Markov Chain Monte Carlo(MCMC)method.The maximum a posterior estimation influence line(MAPIL)of the bridge is finally obtained by connecting the maximum a posterior estimation value of the posterior distribution of the influence line of the bridge;On the other hand,according to the multivariate linear relationship among the bridge influence line matrix,axle load vector and bridge dynamic response vector,a BWIM model based on Bayesian theory is established.On the basis of knowing the posterior distribution of bridge influence line and bridge dynamic response data,the posterior distribution of vehicle axle load is inverted,and the maximum a posterior estimation(MAP)of axle load posterior distribution is used as the identification result of axle load;Finally,based on the property that the posterior distribution of bridge influence line and the posterior distribution of vehicle axle load can be inversed each other,a compound inversion mechanism of influence line and axle load is proposed to improve the accuracy of the identification results of bridge influence line and axle load.(2)In the aspect of numerical simulation,according to the vehicle classification method and the large data of vehicle parameters in the existing vehicle database,this paper establishes the mapping relationship between vehicle type and the distribution of gross weight and the distribution coefficient of axle load;In order to obtain a large number of bridge dynamic response data under multi-working conditions by numerical simulation,the finite element model of beam-slab bridge is established in ANASYS by using grillage method,and the half-vehicle numerical model is established in MATLAB by using the integrated form of concentrated mass-spring damping-suspended massspring damping;The feasibility and accuracy of the proposed bridge influence line identification method are verified by a large number of random working condition data generated by the vehicle-bridge coupling numerical model.By changing the vehicle speed,vehicle type,the number of operating conditions and other variables in the vehicle-bridge coupling numerical model,the influence of these variables on the posterior distribution of the bridge influence line and the MAPIL recognition result is explored.The feasibility and effectiveness of Bayesian BWIM method and composite inversion mechanism of influence line and axle load are verified by numerical example of vehicle-bridge coupling.(3)In the aspect of test,the simplified vehicle-bridge coupling test is carried out by using the single-span simply-supported T-beam bridge and the three-axle truck scale model test platform.The dynamic response data of the bridge model are obtained through laser displacement meters and strain gauges.By changing the location and size of the load on the model truck,the distribution information of the key parameters of the model truck is counted.Using the proposed method of bridge influence line identification based on Bayesian theory,the posterior distribution of the displacement and strain influence lines of the bridge model and the corresponding MAPIL are obtained,and the identification results are within an acceptable error range;The proposed BWIM method based on Bayesian theory is adopted to identify the axle load and gross weight of the vehicle model under different vehicle speed and load conditions,and the feasibility of the proposed method is verified.