Research On Damage Identification Of Long Span Continuous Rigid Frame Bridge Based On Monitoring Data

Large-span continuous rigid-frame bridges are an important part of modern highspeed highway and railway transportation networks.The structural safety of such bridges is of great significance for ensuring transportation safety and national economic development.However,due to the long-term exposure of bridge structures to external weathering and vehicle loads,damage to bridges is inevitable.Rapid and accurate identification of the degree of damage is crucial for ensuring the safety of bridge structures.Therefore,research on damage identification technology for large-span continuous rigid-frame bridges based on monitoring data has important practical significance and scientific value for improving the safety of bridge structures.In fact,early warning and evaluation technology based on health monitoring systems has become a hot research topic in the academic and engineering fields,especially in large-span bridge structures.Processing and analyzing the long-term and large amount of monitoring data collected by bridge health monitoring systems and performing damage early warning and location are currently major challenges.To address these challenges,this paper conducts the following research:(1)Summarizing common damages and defects in large-span continuous rigid-frame bridges,and studying the impact of these damages and defects on monitoring parameters.The loss of prestressing force has a significant effect on the structural response of largespan continuous rigid-frame bridges,with the full-bridge prestressing loss having the greatest impact,followed by the top plate prestressing loss and the bottom plate prestressing loss.The stiffness loss of different beam segments has a relatively large impact on the structural response of the beam segment itself,but has a relatively small impact on the structural response of other beam segments,and increases with the increase of stiffness loss.(2)Based on the quasi-static signals in the bridge health monitoring system,a moving principal component analysis is performed to propose two sensitive damage indicators for damage early warning.This method is implemented using Python programming language and demonstrated using simulated quasi-static deflection data from a large-span continuous rigid-frame bridge under temperature effects.The early warning timeliness and robustness of the proposed method are analyzed and its feasibility is demonstrated.(3)For damage location,a large number of damage datasets are generated using ANSYS APDL,and four neural network models with different damage location accuracy are trained using deep learning methods.Finally,the feasibility of damage identification and location based on deep learning is verified.(4)A signal preprocessing method is proposed for continuous signals obtained from long-term online monitoring.Real data from a bridge health monitoring system are processed and the trend is analyzed.The proposed sensitive damage indicators based on moving principal component analysis are used for damage early warning,and the trained damage model based on deep learning is used for damage early warning and location.The feasibility of these two methods for practical monitoring projects is verified.