Study On Evolutionary Law Of Dynamic Properties And In-Service Condition Assessment Of The Long-span Cable-stayed Bridge Using Long-Term Monitoring Data

The structural health monitoring system(SHMS)can provide important data sources for evaluating in-service conditions of long-span bridges and making reasonable maintenance decisions.Currently,due to the anomalies of monitoring data and absence of effective automated analysis methods,the monitored data is difficult to be utilized to generate real-time actionable structural information,hence creating gaps between SHMS and in-service condition assessment.Therefore,aiming to provide technical supports for real-time structural health monitoring,this paper takes Sutong Bridge,the first kilometer-level cablestayed bridge around the world,as the engineering background.Based on the long-term monitoring data recorded by its SHMS,studies are carried out on the evolutionary law of dynamic properties and inservice condition of a long-span cable-stayed bridge with its emphasis on automated bridge monitoring and massive data analysis.The main contents of this research are summarized as follows.(1)Anomaly detection method for bridge monitoring data based on GANs and AE.Gramian Angular Field is used to transform the monitored time series into grayscale images.Subsequently,two unsupervised deep artificial neural networks,i.e.,generative adversarial Nets(GANs)and Autoencoder(AE)are trained,and the accuracies of the neural networks are verified according to the prediction errors of the testing dataset.According to the training and testing datasets,the optimal indexes for anomaly detection are determined.On that basis,the anomaly detection method for bridge monitoring data is established with the combination of the cumulative sums.Finally,the long-term monitoring data recorded by the SHMS of Sutong Bridge is utilized to verify the effectiveness of this method.Results show that compared with traditional supervised-learning methods,this method can learn the features of training examples without labelling,thus simplifying the training process and effectively improving the efficiency of data anomaly detection.(2)Study on automated modal parameter identification method for the long-span cable-stayed bridge based on principal component and cluster analysis.Initially,the noise components included in the modal validation criterion are separated based on principal component analysis,and the false modes in the stabilization diagram are pre-eliminated using of k-means clustering method.Subsequently,the hierarchical clustering method is used to separate the modes of different orders,based on which the relationship between the number of truncated clusters and the number of effective modes is studied.The obtained relationship is utilized to determine the optimal number of clusters for hierarchical clustering.On that basis,an automated modal identification method for the long-span cable-stayed bridge is established.The effectiveness of this method is verified by scaled-model tests and SHMS data of Sutong Bridge.Results show that this method can automatically and accurately identify the modal parameters of the long-span bridge,as well as effectively reducing the dispersion of identified damping ratios.(3)Study on environmental effects on tracked modal parameters of a long-span cable-stayed bridge.Based on the gaussian mixture model,a novel method is established for automatically determining and updating the baseline of modal parameters of the long-span bridge.It is then validated using numerical simulation and SHMS data of Sutong Bridge.Subsequently,based on the automated modal identification method,the automated modal tracking method for the long-span bridge is established.The established method is then validated using the long-term monitoring data recorded by the SHMS of Sutong Bridge.Finally,variations of modal parameters with monitored environmental factors(e.g.,temperature and wind speed)is studied during long-term operation of the bridge.Results show that the proposed method can be used to automatically track the modal parameters of the long-span cable-stayed bridge during long-term operation.The obtained relationship between modal frequencies and environmental factors can lay the research foundation for calibrating the finite element model and evaluating the operational condition of the bridge.(4)Study on the variation of dynamic characteristics of a long-span cable-stayed bridges during typhoons.Initially,the relationship between the static/dynamic bridge responses(e.g.,acceleration and displacement)and environmental factors(e.g.,wind speed,temperature)is analyzed according to the SHMS data of Sutong Bridge during three typhoons,i.e.,Haikui,Damrey,Bolaven.Afterwards,the modal parameters of the bridge are identified and tracked.On that basis,the relationship between modal parameters of the bridge and wind speed,temperature,and vibration amplitude(including displacement and acceleration)is studied.Results show that the low-order frequencies of the bridge are obviously influenced by the wind speed,while the high-order frequencies are controlled by the temperature.The damping ratios of the first three vertical bending modes are correlated with the wind speed,but weakly correlated with the displacement and acceleration amplitudes.The obtained conclusions can provide references for wind-resistance performance evaluation and vibration control of similar bridges.(5)Fatigue reliability assessment of a long-span cable-stayed bridge based on long-term monitoring strain data.The following analysis is carried out based on the long-term strain data recorded by the SHMS of Sutong Bridge.Initially,dynamic and static components of the monitored strain data of the main girder are separated;the main controlling factors on dynamic strain responses are analyzed.Subsequently,the rain-flow counting method is used to calculate the stress amplitude and the number of cycles of each monitoring point.The lognormal function is utilized to fit the distribution of the measured stress amplitudes;consequently,the equivalent stress amplitude and number of cycles are estimated.On that basis,the fatigue reliability of three typical welding details at the key U rib of the main girder of Sutong Bridge is evaluated according to the S-N method.Results show that the local deformation of the top flange of the main girder caused by the wheel loads is the main reason for the fatigue damage of welding details,and the designed cutout of the diaphragm could improve the fatigue reliability of the welding details between the U ribs and the diaphragm.Conclusions can provide references for fatigue-resistance design,manufacture,and maintenance of the steel box girder.(6)Safety assessment of the riding vehicle on the long-span cable-stayed bridge during typhoons based on monitoring data.Initially,deduce the dynamic equation of a single vehicle running on the bridge under wind loads and apply the monitored bridge responses and wind loads on the vehicle model,hence realizing the efficient evaluation of dynamic responses of the vehicle on the bridge deck.The method is verified using the numerical simulation of a simply supported girder bridge,and the influence of different polynomial interpolation methods on estimating vehicle responses is analyzed.Finally,the safety of the riding vehicle on the deck of Sutong Bridge during typhoon Haikui is evaluated based on the recorded SHMS data.Results show that this method can accurately estimate dynamic responses of vehicles on the bridge deck and provide an effective alternative for evaluating the safety of driving vehicles.Wind loads,lateral and torsional vibration of the bridge deck are the main factors that control the vehicle safety.The limitation of driving speed is an effective means to ensure the vehicle safety.