Back Analysis Of Soil Strata Parameters In Foundation Pit And Deformation Prediction Based On Neural Network

Foundation pit excavation is a typical soil unloading process, it will exert aninfluence on the design of foundation pit using conventional triaxial test method todetermine soil deformation and strength parameters, which ignores the specialconditions of deep foundation pit engineering. Apart from the retaining structures andsoil characteristics, many other factors contribute to the deformation of foundationpit, such as excavation construction. So it will be of great importance in using backanalysis method to establish relatively reasonable soil strength parameter. Based onthe monitoring data of a huge deep foundation pit located in Yujiabao,Tianjin, thesoil parameters are analysied using ANSYS optimized back analysis method and asobe analysied by neural network intelligent analysis method with MATLAB.According to the monitoring data of this huge deep foundation pit, whichexcavation scale reaches to100,000square meters, the real working condition of thisexcavation is exposed. In order to determine the relatively reasonable soil parametersfor finite element calculation, two steps are used in analyzing the deformation andstrength parameters sperately with back analysis method.By matlab-bp neural network, a loop program is designed to judge the best unitnumber in hidden layer and16group learning samples is also designed by usingorthogonal test. A inverse calculation method neural network is obtained.