Random Field Simulation Based On Kriging Model And Reliability Analysis Of Soil Slope

Slope stability analysis and reliability calculation have always been the focus of civil engineering research.The landslide caused by slope instability has become one of the most frequent geological disasters in the world.Considering the natural conditions of mountainous and hilly areas and the human factors of the rapid development of infrastructure projects in China,there are a large number of natural and artificial slopes.At the same time,China is located at the junction of plates,and the slope instability induced by earthquake often brings disasters to engineering construction and ecological environment,which seriously threatens the safety of life and property.Therefore,in-depth study on the stability and reliability of slope under earthquake conditions is of great significance for disaster prevention and mitigation.Firstly,a simple random variable model is used to describe the uncertainty of geotechnical parameters and analyze the reliability of slope under static and dynamic conditions.On the basis of equivalent extreme events,the time history curve of safety factor of slope under dynamic condition is equivalent to equivalent safety factor,which realizes the unification of limit state equation of slope under dynamic and static states,and extends the static reliability calculation method to dynamic field.The finite element method and two kinds of limit equilibrium method are used in the static safety factor calculation,and the pseudo-static method and time history analysis method are used in the dynamic safety factor calculation.The calculation results show that the limit equilibrium method can get a reasonable safety factor with less calculation cost,and it can be used for stability analysis in the case of limited resources;the quasi-static method will overestimate the failure probability of the slope,so as to make a conservative judgment.Then,in view of the shortcomings of current random field simulation methods,this paper proposes a random field simulation method based on uniform point set and Kriging model,which can be directly used for the realization of Gaussian and non-Gaussian random fields.On the basis of expounding the selection rules of representative points and sample sampling method,Kriging approximation model is introduced,and the implementation steps of the method are given.The applicability of the proposed method in Gaussian and non-Gaussian random fields,isotropic and anisotropic random fields is verified by two numerical examples.The results show that when the tangential radius is0.25 times of the fluctuation range,both accuracy and calculation efficiency can be achieved;for different types of correlation function,the simulation accuracy has slight differences,but the overall trend is good.Finally,the random field simulation method is combined with the reliability calculation method based on artificial neural network,and the detailed calculation steps and implementation flow chart are given.The results show that neglecting the spatial variability of the parameters will overestimate the failure probability of the slope,which is more obvious in the static reliability calculation.