Construction Control Of The High-speed Rail Beam Bridge And The Application Of Artificial Neural Networks

With the rapid development of China’s high-speed rail, lots of the high-speedrail bridge has been built.And the continuous beam bridge is mostly used in thehigh-speed rail bridge structure, the degree of statical indeterminacy is small, simplydesigned,also the construction technology is very mature.So it is widely used.Becausethe difference of parameters between the actual state with the design state,to make thebridge’s line and stress meet the design requirements,we must impel the constructioncontrol when the bridge is building.In this paper,the Sha Heyi bridge (60+100+60m) of Jin Qing passenger Line asa practical engineering background,to study the continuous beam bridge constructioncontrol, The following are the main conclusions of this paper:①Reading a large number of literatures, to summarize the historical developmentand current situation of bridge construction control,also to elaborate the importance andnecessity of bridge construction control by using engineering examples.②To introduce the theory of bridge construction control,derive the gray theoryGM(1,1) model error estimate; describe the concept and structure of artificial neuralnetworks in detail,and introduce the specific algorithm of BP neural network based ongradient descent.③To analyze the Sha Heyi bridge construction by using bridge structuralanalysis software MIDAS.Combined with the existing engineering experience and thesimulation results,we believe the Sha Heyi bridge’s simulation results aboutstress,deformation and etc are reliable and accurate,So they can be to used to control theactual construction.④To introduce the actual construction control process on the stress,the linear andthe stability,to calculate the bridge formwork elevation,and use GM(1,1) model toestimate the measured stress.Eventually the Sha Heyi bridge elevation error is less than10mm,the stress is controlled in reasonable limits,the project progressing is smooth,andwithout a safety incident,therefore we consider that the construction control of Sha Heyibridge is successful,it achieves the expected monitoring purposes.⑤Using the BP neural network to predict the actual formwork elevation of ShaHeyi ‘s11#,12#and13#beams,by analyzing the terms,we get that the best BP neuralnetwork model—the hidden layer contains only eight neurons,and use momentum and adaptive gradient descent method (traingdx).At last the actual formwork prediction of11#,12#and13#beams is satisfactory,the error only0.53%,0.50%and0.83%.