Study On Deformation Prediction Of Foundation Pit Based On Finite Element And Elman Neural Network

The development of the urbanization process is accompanied by the increasing scale of infrastructure engineering.As a common project in infrastructure construction,deep foundation pit has high risk because of the complicated factors of working environment.Most of the deep foundation pit of Metro is in the central area of the city,so the potential risk factors are more complicated,so it is necessary to strictly control the safety of deep foundation pit construction.Foundation pit monitoring can directly judge the stability of foundation pit,it is an important techmical means to prevent engineering safety accidents,foundation pit monitoring technology and equipment after continuous improvement and upgrading has been quite mature.However,since the monitoring is carried out at the time of construction,the subsequent changes in construction and unexpected conditions can not be accurately judged.Aiming at the limitation of the field monitoring function,this paper puts forward the use of historical monitoring data,combined with finite element and neural network methods to predict the subsequent deformation of construction,relying on the deep foundation pit project of Zhongyuan Road station of Zhengzhou Metro Transit line Line 5,mainly carries out the following work:This paper combs the theoretical and methodological research results of domestic and foreign scholars on the deformation and prediction of foundation pit deformation.This paper expounds the defommation mechanism characteristics of deep foundation pit,collates the support type of foundation pit,summarizes the failure form of support structure and the influencing factors of deep foundation pit deformation of Metro.This paper introduces the general situation of the deep foundation pit project of Zhongyuan Road station,clarifies the engineering monitoring grade,expounds the monitoring project and introduces the implementation method of some monitoring projects,and extracts the surface subsidence of some monitoring points and the deep horizontal displacement of the enclosure structure for analysis.Using Midas GTS finite element software,three-dimensional dynamic model of Zhongyuan Road station is established according to geological survey parameters,support design parameters and construction process,and the excavation of foundation pit is simulated in layered and segmented form.The three-dimensional model cloud map of displacement and surface subsidence of enclosures is analyzed,the simulation values of surface subsidence and deep horizontal displacement of enclosures are extracted and compared with the measured data to verify the usability of the model and provide data support for the fusion model.Because of the prediction limitation of the ideal environment parameters of the finite element model and the difference between the measured data and the finite element simulation value,the rolling prediction model of the Elman neural network is established,and the difference between the two in the later construction is predicted,and the modified predicted values are made and obtained by the fusion model calculation with the original element analog value.Comparing the predictive value,finite element simulation value and measured value of fusion model,it is found that the prediction result of fusion mode reduces the error of finite element analog value by 50%,corrects the deformation trend of smooth simulation value,and selects multiple sets of data to predict to verify the reliability of fusion model.It is found that the prediction result of surface subsidence by fusion model is better than that of horizontal displacement prediction of envelope structure.The research results have a certain reference function for the deformation prediction of foundation pit.