Study On Spatial Variability Simulation Of Soil Properties And Practical Reliability Analysis Method Of Soil Slope

The variability of soil properties is the fundamental uncertainty in reliability study of soil slopes, and the spatial variability of soil properties can not be ignored. The research presented in this thesis focuses on the simulation of spatial variability of soil properties, its effects on slope reliability and practical reliability analysis methods of soil slopes considering spatial variability. The emphases are as follows:A new slope reliability analysis method - LHS-BP neural network is proposed. LHS method, improved constraint optimization method for FORM and Harr's PEM are applied to slope reliability analysis.Some common problems in slope reliability study are presented and discussed, such as systematic reliability, critical slip surface, reliability evaluation index and problems related to FOSM and FORM and etc.The most widely-used random variable model is only applicable to particular circumstances in slope reliability analysis. The techniques of random field theory and geostatistics can be used to quantify and model the spatial variability of soil properties.The constraint random field (CRF) is first introduced in slope reliability analysis and implemented utilizing random field theory and regionalized variable theory of geostatistics to ensure that the random field realizations match the parameters at the sample locations exactly.The spatial variability of soil properties is mainly dominated by its geological genesis, thus geological genesis of soil can be incorporated in reliability analysis method. The random field model is employed to simulate spatial variability of soil properties and slope reliability methods considering spatial variability is proposed such as FORM by optimization approach and Monte-Carlo simulation with LHS technique.Sensitivity analysis method for slope reliability under random field model is proposed. The sensitivity analysis results can play an important role in slope design and reinforcement.The influence of spatial variability of soil properties and simulation methods on slope reliability is first thoroughly studied such as scale of fluctuation, correlation function, equivalent isotropic correlation structure, discretization scale of random field and etc. Simplified variance reduction method and random field discretization methods in slope reliability analysis are compared, and results are quite valuable for practice.The application of CRF is introduced through Monte-Carlo random finite element method. The results show that the CRF can significantly reduce variance of the random field simulation and get lower probability of failure when high variability of soil properties exists.Vector sum factor of safety is a new definition for slope factor of safety. A simple method for solving vector sum factor of safety is proposed under equilibrium framework and then applied to slope reliability analysis.