Applied Research Of Neural Network In Construction Monitoring Of Qinhe Bridge In Chongzhou City

The prestressed concrete continuous beam is a kind of bridge structure widely used in the current bridge construction.However,this bridge type has the characteristics of high difficulty in the line control,large hidden danger of cantilever construction,and high requirements for construction accuracy in construction.Therefore,monitoring and control of bridge construction is particularly important.Usually,the initial finite element model of the bridge structure is based on the design drawings.There is a difference between the finite element model and the actual state of the bridge.This will inevitably lead to inconsistency between the model calculation results and the actual measurement results.As a result,it will have an uncertain impact on construction control.Therefore,it is of great engineering application research value to carry out the study of the neural network in the correction of the finite element model of the prestressed concrete continuous beam bridge.In this paper,with the "Qinhe Bridge" as the engineering background,the whole process of the main bridge hanging cantilever construction method is monitored,and the sensitivity analysis is combined with the neural network algorithm to modify the finite element model of the bridge.The main work and results of this paper are summarized as follows:1.The concept,purpose,calculation and analysis methods of bridge monitoring are summarized,and a systematic study of the construction monitoring theoretical system for long-span prestressed concrete continuous beam bridges is conducted;2.Midas Civil 2015 is used to establish the bridge construction finite element mode.Based on the model,the sensitivity analysis of the cumulative displacement of the completed bridge to the design parameters is carried out.The results show that the sensitivity of the cumulative displacement of the completed bridge to the parameters is in descending order of the prestressed tendon tension control stress,the structural weight coefficient,the steel elastic modulus,and the prestressed pipe wall friction coefficient;3.In this paper,the four main design parameters identified by the sensitivity analysis are the test objects.Uniform Design 300 Crack is used to uniformly design the test objects and then the test samples are extracted.The BP network model and the generalized regression network model are established on the MATLAB platform.The test samples are used for training neural network model.The training results show that the generalized regression network model has faster error convergence speed and better generalization performance.Finally,a generalized regression neural network was used to modify the finite element model.The modified finite element model achieved good results in construction monitoring and control of Qinhe Bridge.