The Feedback Analysis Of Slope Stability Based On BP Neural Network And The Finite Difference Method

With the development of computer and intelligence,research scholars are keen to choose numerical simulation as a means of analysis to arrange layouts and design schemes and select construction measures in slope engineering.But the authenticity of numerical simulation depends on accuracy of material parameters,which leading that the key to the stability analysis of slope engineering is to obtain the mechanical parameters of rock and soil mass.In this paper,based on the finite element difference numerical calculation method,combined with the orthogonal test design to analyze the sensitivity of mechanical parameters to slope stability and deformation and determine the main influence factors and secondary influence factors.Combined with actual monitoring point displacements,main mechanical parameters of slope rock mass were inversed by the BP neural network back analysis based on Matlab programming.Aim to study mechanical parameters affecting slope stability and deformation largely.Analyzing the prediction of the slope in the future and guiding the construction and design of the proposed slope by applying the numerical simulation analysis results to the actual slope engineering.(1)Analyzing sensitivity of 6 mechanical parameters to slope stability and deformation based on a classical homogeneous slope by using orthogonal experimental design to arrange the test for bulk density,tensile strength,cohesion,friction angle,elastic modulus and poisson ratio.Analyzing sensitivity based on a heterogeneous slope again,which confirming bulk density,cohesion,friction angle and elastic modulus were the main factors of slope stability and deformation.(2)In this paper,BP neural network was selected as back analysis method.BP can contact relation of arbitrary precision nonlinear function based on Matlab programming and with 3 layers network.(3)Take Jin-ping Hydropower Station,left abutment slope section 4-4 as the object,according to the scope of geotechnical Parameters obtained by survey and based on uniform design requirements,mechanical Parameters were designed to different levels to fix the scheme of numerical simulation.FLAC3 D was used to calculate the displacement of monitoring points under different mechanical parameters,and the monitoring points were taken as network inputs,and the parameters were taken as the outputs of the network.Then the trained sample of BP neural network was obtained.The function relation between the displacement and the inversion parameters was obtained by the trained neural network.With back analysis of the actual monitoring displacement,9 main mechanical parameters were obtained.Comparison between inversion calculation displacements and the actual monitoring displacements,the error is less than 5%,which proving the rationality of inversion calculations.(4)Analyzing and calculating numerical simulation results of the slope section 4-4 under conditions of excavation without support and with support.analysis of the variation law of slope stability and displacement in the process of excavation to forecast the stability and displacement in the process of slope construction and to provide reference guidance for other proposed slope engineering...