Research On Damage Identification Method Based On Bridge Health Monitoring Big Data

Bridges are an important part of traffic routes and key nodes in road engineering.Accurate assessment of bridge operation status is the key to avoiding bridge safety accidents.In the course of bridge service,the bridge health monitoring system has accumulated a large amount of monitoring data.It is very important for bridge health monitoring to adopt effective methods to analyze and mine the value of monitoring data.However,traditional monitoring systems face the problems of massive monitoring data storage pressure and difficulty in efficiently analyzing monitoring data.In the context of big data,the use of deep learning for bridge structure damage warning has a good application prospect.The research content of this paper is as follows:(1)Designed a bridge structure damage warning scheme based on big data.Analyzed the traditional bridge structure health detection system and measurement point layout plan,built a bridge structure damage early warning system based on big data,divided the system into data acquisition layer,big data platform layer and application layer,and described the bridge based on big data Structural damage early warning process.(2)Constructed a bridge structure damage warning model based on CNN-LSTM.Considering that the strain is more sensitive to the local damage of the bridge structure,the actual measured strain of the bridge is preprocessed;in order to avoid the influence of the temperature of different measuring points on the structure response,the vehicle-induced strain is extracted;the bridge structure response is constructed based on CNN-LSTM The inter-correlation model realizes the prediction of the structural response data of the key measuring points of the bridge;the early warning threshold is set according to the model prediction residual to carry out the early warning of the bridge structure damage;the comparative experiment proves that the model in this paper has better predictive performance.(3)Realize the parallelization of the CNN-LSTM bridge structure damage warning model.Deployed the Spark big data cluster,realized the parallel processing of abnormal data processing based on the isolation forest algorithm,the parallel extraction of vehicle-induced strain,and the parallelization of the CNN-LSTM neural network model;the comparison experiment proved that the parallelization mode Abnormal data processing and vehicle-induced strain extraction use shorter time,and the training efficiency of neural network models is higher.