Prediction Of Bridge Deformation By Particle Swarm Optimization Neural Network

Real-time bridge deformation monitoring and its change prediction are essential for the study of the factors affecting induced bridge displacements and displacement patterns.As an important part of the transportation network,bridges facilitate people’s daily life and work needs while maintaining the steady growth of the national economy.However,the structural deformation of bridges varies due to the effects of construction,operation,improper maintenance,and geological environment changes,which can seriously affect people’s daily life and cause national economic losses.Therefore,it is especially important to monitor and predict the deformation of bridges by understanding their stability status.In this thesis,LSTM and BP neural network are introduced to bridge deformation prediction by using particle swarm algorithm to optimize the model,and the bridge displacement prediction model by PSO optimized LSTM neural network and bridge settlement prediction model by PSO optimized BP neural network are proposed.The research related to bridge displacement prediction was carried out.The research contents are as follows:(1)Based on the geological profile survey and bridge stability analysis,a monitoring network was deployed and monitoring points were placed at landmark locations.The collected monitoring data are analyzed,and the s-t curve is presented to analyze the displacement influencing factors and curve characteristics.(2)For the problems of abnormal data and missing data in the collected monitoring data,the standard deviation method is used to eliminate the abnormal values and the Newton polynomial interpolation method is used to supplement the missing values to improve the quality of bridge displacement monitoring data.Finally,the improved data are analyzed by the angle-cutting method and the corresponding forecasting rules are formulated.(3)To address the problems of low accuracy and weak prediction effect capability of bridge deformation prediction by LSTM and BP neural network models,the parameters of the two prediction models are improved by PSO algorithm,and the bridge displacement prediction model using PSO-LSTM neural network is related to the bridge settlement prediction model using PSO-BP neural network.The data analysis results of the engineering examples show that the combined model has obvious advantages and has improved effects in terms of fitting ability,etc.The root mean square errors of the constructed models are within 5%.It shows the feasibility and effectiveness of the combined model in bridge displacement prediction in this thesis.