Study On Seepage Failure Criteria Of Soil-Structure Interface Under Pressure-Seepage Coupling

At present,the contact surface of soil and structure generally exists in various basic projects.The problem of seepage failure of the contact surface of soil and structure under the action of seepage has always been the focus of the current research.Once the contact surface is damaged,it will not only bring huge losses to the country The economic loss will also affect the life and property safety of the people.Therefore,it is very important for the actual project to clarify the damage process and influencing factors of the soil-structure contact surface under the seepage action,but nowadays the seepage damage of the soil-structure contact surface The research on the seepage failure especially when considering the stress state has yet to be perfected,so this paper conducts an in-depth study of the infiltration failure process and phenomena under various influencing factors through a combination of laboratory tests and numerical simulations,and combines multiple The regression model and the neural network model analyze and verify the critical hydraulic slope of the seepage failure of the interface between the soil and the structure,so as to make a scientific and reasonable prediction of the actual project.Firstly,through summary analysis,this paper develops the indoor test device for seepage failure of the interface between soil and structure under the action of confining pressure,which mainly includes visualization test room,hydraulic pressure supply device,confining pressure supply device and monitoring device.At the same time,the laboratory test of seepage failure of the soil-structure contact surface under different confining pressures was carried out through the self-configured soil sample materials.The critical hydraulic slope of seepage failure at the interface between soil and structure is analyzed,and the seepage failure process at the interface between soil and structure under confining pressure is analyzed.Secondly,a three-dimensional numerical calculation model was established through the finite difference program FLAC3D,and the working conditions of the indoor laboratory were verified and analyzed,thereby proving the feasibility of the indoor experiment and the validity and accuracy of the numerical simulation.Furthermore,through numerical simulation,the effects of different confining pressures,the roughness of the contact surface and the content of material clay particles on the seepage failure of the interface between the soil and the structure were analyzed.At high conditions,the smaller the deformation of the interface between the soil and the structure,the greater the critical hydraulic slope of the seepage failure at the interface between the soil and the structure.Finally,based on Pytorch,a multiple regression analysis model and feedforward neural network model are established.On the one hand,based on the multiple regression model,the regression equation of the relationship between the critical hydraulic slope and the confining pressure,the roughness of the contact surface and the material clay content is set.By defining the loss function and using the back propagation method,the parameters are optimized to determine The specific parameters of the regression equation and the expression of the regression equation;on the other hand,based on the feedforward neural network model,using the MSE mean square error as the target function,the adaptive moment estimation(Adam)algorithm is used to train and verify the network to determine the critical hydraulic slope and The relationship between various factors.The two methods are used to verify and predict the critical hydraulic slope of the seepage failure at the interface between soil and structure.This study analyzes the seepage failure process of the interface between soil and structure by combining the laboratory test and numerical simulation,and obtains the critical hydraulic slope of the seepage failure of the interface under different conditions.Through the multiple regression equation and artificial neural network model,the relationship between the critical hydraulic slope and each influencing factor is fitted and trained to obtain the specific expression formula of the multiple regression equation and the best prediction state of the neural network model.The two prediction models can scientifically predict the critical hydraulic slope of the seepage failure of the soil-structure contact surface in the actual project under the known confining pressure,roughness of the contact surface,and material clay content,thus It provides theoretical support and experience reference for the rational design and safe operation of the interface between soil and structure in actual engineering,and reduces unnecessary waste in actual engineering.