Study And Field Monitoring On The Temperature-and Train-induced Responses Of The Long-span Steel Truss Arch Bridge

Bridges serve as critical components of railroad infrastructure.These bridges must operate in harsh environments,experiencing large loads with every train crossing.Any unexpected closures of railway bridges can significantly affect the traffic capacity of the corresponding line.Maintenance and repair activities,which can effectively prevent structural damage or even more serious accidents,are crucial to ensuring operational performance.Accordingly,understanding the bridge behavior under in-service train loads will help operators to focus limited resources on maintaining safe and reliable operation of railway bridges.Thus,the object of this work is to investigate the behavior of long-span high-speed railway bridges under in-service temperature and high-speed train loading using long-term monitoring data produced by the structural health monitoring(SHM)system,with the goal of understanding issues affecting the behavior of the high-speed railway bridge.This study addresses steel truss bridges,which are frequently used for longer-span high-speed railway bridges.The contents and associated conclusions are presented in more detail in the following paragraphs.(1)Investigation on the extremes of temperature in the long-span steel truss arch bridge using the environmental contour(EC)method.The temperature variations in a steel truss arch bridge,including the average temperature and temperature difference within structural members,are systematically investigated based on the long-term field monitoring data.Accordingly,the probability density distributions of the average temperature and temperature difference are calculated.Subsequently,the joint probability distribution for the effective temperature and temperature difference within members in the long-span steel truss arch bridge is contoured using the highest density contour(HDC)method.Additionally,the temperature contours with the return period of 10,20,and 50 years are presented,which revealed the extreme temperature variations in the bridge.The results show that the probability distribution of the average temperature can be described using the normal distribution;the weighted sum of two lognormal distributions can describe the distribution of temperature difference.The temperature contours with the temperature difference as marginal distribution and the average temperature as the conditional parameter is shown to be a good approximation for the extreme temperature difference,whereas it may overestimate the effective temperature in the tails.(2)Mapping for temperature-induced responses of the long-span steel truss arch bridge.According to the spatial temperature distributions in the long-span steel truss arch bridge,three standard temperature measures(i.e.,the average temperature of the bridge,temperature differences between members,and temperature gradients within members)are used to represent the spatial temperature distributions;the characteristics of these three temperature measures are examined based on field monitoring data.Then,to understand the underlying mechanisms of the temperature effects on such long-span bridges with spherical bearings,a general correlation model for temperature-induced responses and the three temperature measures is determined based on the elastic beam theory.Subsequently,the mapped relationship is validated using long-term field monitoring data.Results show that the temperature-displacement-strain mappings produce unique flat planes in 3D space;the plane parameters,including the orientation and boundary,are determined by the structural parameters,especially the bridge cross-section properties and bearing properties.(3)Predication of temperature-induced responses for long-span steel truss arch bridges.This study first proposes that the relationship between increments in temperature and the associated increments in responses can be used as a surrogate to assess the bridge performance.Simulation results show that the statistical distribution of the error between measured and predicted response increments can be employed for identifying abnormal structural behavior.Accordingly,this study explores four typical mappings with different inputs for both displacement increments and strain increments based on the gated recurrent unit neural network(GRU).In addition,the calculations produced by these mappings are validated against the field monitoring data.Results show that the statistical distribution of the error between measured and predicted response increments can be employed for identifying abnormal structural behavior.The mapping with all temperature sensors performs the best;principal component analysis(PCA)can effectively reduce the dimension of input without compromising accuracy.In addition,the recorded time of temperature data is validated to be a useful indicator of the spatial temperature distribution in bridges,which can be employed to improve the performance of the mappings when the bridge has only a few temperature sensors.(4)Investigation on the vibration behavior of a long-span high-speed railway steel truss arch bridge.The train-driving frequencies are extracted using the frequency domain decomposition(FDD)method.Accordingly,the relationship between train speed and train-driving frequencies is systematically mapped.Then,the train-induced bearing displacement and structural strain during normal operation are explored to demonstrate the effects of train loads and bearing properties.Subsequently,the impact factor(IF)in typical members is examined;a sensitivity analysis is performed for the IF of strain responses with respect to the train speed,train weight,and temperature to identify the fundamental issues affecting these responses.Additionally,a similar sensitivity analysis is conducted for the peak acceleration.The results indicate that the friction force in bearings provides residual deformations when two consecutive trains are in opposite directions.In addition,the IF and peak acceleration are primarily affected by train speed,particularly near train speeds that result in the resonance of the bridge response.(5)Investigation on the pattern for train-induced strains of a long-span steel truss railway bridge.The underlying mechanisms behind the train-induced responses and the utility of influence lines(IL)are illustrated using elastic beam theory.In particular,the effect of train-induced longitudinal force and bearing restraint is considered.Accordingly,two condition indicators base on the axial strain in a pair of structural members are given(i.e.,the ratio between integral areas of pseudo-static strains and the ratio between average pseudo-static strain amplitude).In addition,the statistical patterns of the indicators are studied using numerical simulation.Subsequently,the strain patterns are illustrated for field monitoring data from a long-span steel truss railway bridge;a sensitivity analysis is performed for the condition indicators with respect to the temperature to identify the fundamental issues affecting these indicators.Results show that the ratio between integral areas of pseudo-static strains and the ratio between average pseudo-static strain amplitude can be employed as condition indicator of the long-span steel truss railway bridge.Note that the probability distribution of the condition indicators can be described using the normal distribution;these distributions do not change with the variation of train loads(i.e.,train weight and train speed).(6)Predication of train-induced responses for long-span steel truss arch bridges.The correlation between the characteristics of train-induced responses(i.e.,the peak accelerations and the IF of structural strains)and train parameters(i.e.,train speed and train weight)is evaluated using the maximal information coefficient(MIC).Then,this study explores typical mappings for both peak acceleration and IF based on the GRU.On this basis,the correlations among characteristics of train-induced responses are employed to improve the accuracy of the modeled GRUs.The results indicate that the mappings using the train parameters and the average temperature of the bridge as input perform poorly.The modified mappings take the characteristics of train-induced responses from other measurement points as additional input.These mappings for the peak acceleration show good agreement with the field recorded data,while the poor accuracy for the IF.