Back-analysis Of Soil Parameters In Metro Site And Deformation Prediction Based On BP Neural Network

With the city traffic more crowded,the subway construction represents the general trend.As we all know,subway tunnels are built in the compliticated ground mass.The great randomness,fuzziness,uncertainty and other features of soil are difficult to determine the mechanical parameters of ground mass.The displacement back analysis method provides an effective way to obtain the mechanical parameters of soil.This paper is based on the Beijing metro line 7 JiuLongShan Railway Station~Dajiaoting station interval,on the basis of field monitoring data,the use of numerical simulation software,establishing soil parameter inversion system by BP neural network toolbox of MATLAB.Taking deep foundation pit of Dahongmen Railway station of Beijing Metro Line 8 as the research object,the deformation characteristics of foundation pit during excavation are analyzed,based on a large number of field monitoring data,using the time-series forecasting of BP neural network to predict the short-term deformation.This paper's main contents and research results are:1.This paper summarizes and analyses the concept of neural network and neuron.Introducing and analyzing BP neural network and its application in MATLAB neural network toolbox emphatically.Back analysis of soil parameters and deformation prediction are carried out by BP neural network.2.Back-analysis of soil parameters takes the Beijing metro line seven JiuLongShan Railway Station~ Dajiaoting station interval for example,introduces engineering and geological conditions and hydrogeological conditions,necessity of construction monitoring and the specific monitoring measurement scheme.According to the actual situation of the site,deformation of each excavation step are obtained.After the tunnel excavation,the deformation tends to be stable,and the final deformation stability value is obtained.According to the actual situation of the project,determine the soil Parameters inversion.3.This paper introduces the characteristics of MIDAS/GTS,the characteristic and development of LAC3 D and analyzes its advantages and disadvantages.Modeling and meshing JiuLongShan Railway Station~Dajiaoting station interval by MIDAS/GTS,importing into FLAC3 D by the compiling small program,then simulating excavation and support process.4.According to the elastic modulus E and the cohesion c range,the orthogonal design method was used to design the different test combinations,and the FLAC3 D was used to simulate the experiment,establishing BP neural network learning and training samples,parameters of soil are calculated by displacement back analysis.Compared with the measured data,the optimal solution of the parameters is obtained.5.Deformation prediction is based on the deep foundation pit of Dahongmen metro station.Firstly,the deformation characteristics of the foundation pit supporting structure during excavation are analyzed,and then the deformation prediction is carried out by using the time-series model of BP neural network.