Elevation Prediction Of PC Cable-stayed Bridge Girder With Bayesian Network

PC cable-stayed bridge is widely used in many bridge types because of its large span capacity,mature construction technology,beautiful design line,relatively low cost and other advantages.PC cable-stayed bridge is a multi-time statically indeterminate structure,with complex theoretical calculation and many uncertain factors in the construction process,which may lead to the deviation of the main beam alignment from the design alignment in the construction process and affect the structural mechanical performance and driving comfort after the bridge is completed.If the alignment deviation of the main beam is too large,it will even affect the closure of the main bridgeIn this paper,the PC cable-stayed bridge of Kaiping Donghuan bridge is taken as the engineering background.Based on the bayesian network theory which combines probability theory and graph theory,the bayesian elevation prediction network is constructed to predict and analyze the linear control precision of the main girder during the construction stage.Considering the uncertainty of the factors affecting the deformation of the main beam,the bayesian network combines the prior experience of deformation control with the measured deformation data at the construction stage,constantly updates the parameters of the bayesian network,and obtains a more accurate predicted value of the deformation of the main beam to reduce the construction deviation.As a new deformation prediction agent model,bayesian linear control network studies the theoretical basis,application process and prediction effect of bayesian network.The main work of this paper is as follows:(1)Expatiate on the theory foundation of bayesian network and briefly introduce the functions of bayesian network construction,parameter table solving,updating and inferencing,etc.;(2)The finite element simulation analysis model of cable-stayed bridge construction process was established to analyze the error source of main girder alignment deviation in the construction process,and four main parameters including concrete bulk density,side span cable force,main span cable force and main girder temperature gradient were determined through parameter sensitivity analysis.(3)Using the D-optimal experimental design method based on the response surface,the functional relationship between the relative deformation deviation of the main beam and four important influencing parameters is fitted,which is used to solve the initial probability table of the bayesian network.(4)For the cable-stayed bridge of Donghuan bridge,the relative deformation deviation prediction model of bayesian network is established.By training and updating the bayesian network model with the measured data,the difference between the predicted deformation value and the measured deformation value is small,and the prediction accuracy is more accurate as the training sample increases.The deviation between the predicted value and the measured value of the middle and long cantilever beam segment is less than 20 mm.(5)According to the application process of bayesian network and the real situation of the Donghuan bridge,this paper,based on MATLAB platform,writes the elevation prediction program code of cable-stayed bridge construction stage,and realizes the updating prediction function of bayesian network.The important point of this paper is to make full use of the prior knowledge of the relationship between the deformation of the main beam and the four important influencing parameters and the information of the measured samples in the construction process,constantly update the bayesian network parameter probability table,and make the prediction more consistent with the actual situation of the specific structure.