Study On Prediction Of Deck Elevation Of Continuous Rigid Frame Bridge During Cantilever Construction Based On Dendrite Net

As the preferred bridge type for high-speed and economical application in mountainous areas,continuous rigid bridge has been widely recognized by the engineering community for its many advantages.With the progress and maturity of construction technology,the use of neural network and machine learning and other methods to effectively guide the main beam elevation and realize the setting of pre-arch to form a smooth line has attracted everyone’s attention.In this thesis,a right-width T structure of a 150 m five-span continuous rigid bridge is taken as the research object,and the main influencing factors of elevation are determined by using solid analysis software.An elevation prediction method based on dendritic network is proposed,which proves that the proposed method can provide a guiding basis for the construction elevation control of continuous rigid bridges.The main research contents and conclusions of this thesis are as follows:(1)The main influencing factors affecting elevation prediction were determined by numerical simulation using solid analysis software,and the bulk density,longitudinal tensile stress,concrete elastic modulus,temperature gradient,main pier stiffness,hanging basket load,pipeline deviation coefficient,pipeline friction coefficient,vertical prestress,and overall temperature were listed in order of importance.On this basis,the variation analysis of single-segment parameters is carried out according to the construction order,the elastic modulus of concrete with little influence on the elevation is removed,and the bulk density and longitudinal tensile stress of concrete with large influence are retained as the main design parameters for predicting the elevation.(2)Build and compile MATLAB program of Dendrite Net for parameter identification and train the network;Identify the two main design parameters that affect the elevation;Input the parameter identification results into the model to recalculate the elevation change,and the deviation between the identified elevation and the measured elevation is less than that before identification;Based on the measured elevation data of the project,the error of vertical formwork elevation,the error of hanging basket deformation and the error of roof pouring thickness are calculated and identified,which are used for the elevation prediction of Dendrite Net.(3)The Dendrite Net prediction program based on MATLAB is compiled.The Dendrite Net is trained by using the identification value of the poured segment parameters and the field measured data.The elevation of the main girder after tensioning is predicted by predicting the five main influencing parameters of the next segment.The predicted elevation is basically consistent with the measured elevation,indicating that the Dendrite Net can be used for the elevation prediction of continuous rigid frame bridges.(4)The Dendrite Net prediction program based on MATLAB is compiled.The Dendrite Net is trained by using the identification value of the poured segment parameters and the field measured data.The elevation of the main girder after tensioning is predicted by predicting the five main influencing parameters of the next segment.The predicted elevation is basically consistent with the measured elevation,indicating that the Dendrite Net can be used for the elevation prediction of continuous rigid frame bridges.