Back Analysis Of M Value Of Soil In Foundation Pit Based On Neural Network Model

Soft soil deep foundation pit supporting structure design is always a very complex problem in geotechnical engineering.At present,the domestic method of foundation pit supporting design is the most commonly used method of m elastic foundation beam theory,and most of the specification also recommended.But the parameter selection method of m is still a big problem in the field of geotechnical engineering,design specification only gives the approximate range of m value.Because of the different geological conditions in different areas,the difference in m value is also great therefore,for a specific project,it is necessary to inverse analysis of the region foundation soil's m value on the basis of measured data for foundation pit excavation area or pit formation conditions similar supporting design and provide a reference basis,but also enrich the soil's m value experience.This article unifies the Shanghai dance center deep foundation pit project,the deformation law and characteristics of support structure are analyzed based on field measured data of horizontal displacement of support structure;based on the principle of orthogonal test and the large scale finite element simulation software MIDAS GTS,the orthogonal test table for sensitivity analysis of each soil layer parameter in excavation process of foundation pit in the third area is established and the sensitivity of the displacement of support structure to the parameters of each soil layer are also analyzed.Supporting structure displacement of each soil parameter,the Error Back-Propagation network model is established based on the principle of the artificial neural network and compiles the corresponding Matlab program,using orthogonal test data tables complete the network model training and inspection;The displacement measured data by using the network model has been established on the third area excavation stage of the soil's m values were based on the inversion,and the inversion results are used to predict the displacement of retaining structure under a condition.This paper mainly draws the following conclusions:(1)According to the displacement monitoring data of the third area foundation pit supporting structure in the excavation stage of the deep foundation pit of Shanghai DanceCenter,it is found that the deformation of the supporting structure under the two concrete supporting conditions shows "saddle",and the cumulative maximum horizontal displacement mostly appears at-4 meter.(2)Using MIDAS GTS,the displacement of the support structure under different soil parameters is obtained,and the orthogonal test table is established.Taking the displacement of supporting structure as the evaluation index,the effect of the soil parameters on the displacement of the supporting structure is determined by the range method.The order of the resistance coefficient m,the friction angle and the cohesion of the soil are determined.And the soil resistance coefficient m,which is the most sensitive,is used as the inversion parameter.(3)Based on the characteristics of this paper,a three-layer BP neural network soil parameter inversion model of four input and four output units is established,and a network error test program is compiled according to Matlab,and the hidden layer neurons The number is 10.The network model is trained and tested by the data of the orthogonal test,and the network model with reliable simulation capability is obtained.(4)The displacement data of the third area supporting structure measured in the excavation stage of the foundation of the Shanghai Dance Center are studied by using the established BP neural network inversion model.The soil's m values ??obtained from the inversion are more accurate than the norm values.(5)Using the soil parameters obtained from the inversion to calculate the displacement of the foundation pit in the next working condition,the calculated results are in good agreement with the field monitoring data.Which provides a strong guarantee for the design of the foundation pit support in the post-construction area and enriches the experience of the soil parameters in the area.