Stochastic Back Analysis Of Geotechnical Parameters And Reliability Analysis For Slope Using Bayesian Method

With the expansion of Chinses water conservancy and hydropower,Sichuan-Tibet railway,highways,deep foundation pits and the scale of other basic construction,the important slope projects are emerging.Due to the high concealment,high spatiotemporal variability,and high instability of slope engineering,slope reliability analysis and disaster risk control had becomed the focus of geotechnical engineering research.Accurately determine the geotechnical parameters and analyse the reliability of slope can provide important theoretical basis and technical support for slope treatment,which has a very important theoretical and practical significance for Chinses development.For this reason,this paper aims at the conventional back analysis methods of parameters not consider the uncertainty of soil parameters,the calculation accuracy and efficiency of the existing probability analysis methods of slope stability is low,and how to comprehensively use limited site data and select the optimal calculation model of geotechnical engineering to carry out the research on slope stability analysis,the major research results as follows;(1)Briefly introducing the method and theory of Bayesian,summarizing prior distribution of parameters and likelihood function model,which are commonly used in geotechnical engineering,systematically compare the basic principles of three practical stochastic back analysis methods and the calculation of posterior failure probability,and discusses the merits and drawbacks and applicability of each method by two slope cases,which can lay the foundation for stochastic back analysis of parameters,especially spatial variation.(2)Systematically analyzing the influence of the selection of likelihood function models and the prior distribution of parameters on the evaluation of slope reliability calculations,and applied to the reliability analysis of slope with limited test data.The research results can provide reference for rationally choose the prior probability distribution of soil parameters and likelihood function of Bayesian analysis process.(3)Analyzing the uncertainty of parameters and the error of the geotechnical conversion model,base particle swarm neural network to construct the proxy model of slope output response.On this basis,based on the Bayesian method with the monitoring displacement data to stochastic back analysis the soil parameters and evaluate slope reliability,solved the problems of the back analysis of engineering parameters and deformation prediction of the vertical slope of Changchun West Passenger Station.(4)Developing an optimized selection method of soil-water characteristic curve model based on adaptive Bayesian updating.Based on the model evidence value obtained by Bayesian updating,directly discriminate the optimal soil-water characteristic curve model,and infer the probability distribution of posterior parameters of the soil-water characteristic curve model,which can provides an effective way to optimize the selection of calculation models of geotechnical engineering with limited data.