Probabilistic Analysis Method Of Soil-water Characteristic Curve And Its Application

Soil Water Characteristic Curve(SWCC)is an important curve showing the relationship between the suction and the volumetric water content(or effective saturation)of unsaturated soils.It is used to study soil moisture retention and unsaturated soil mechanical properties.At present,SWCC is usually measured by laboratory testing or in-situ testing methods.Direct testing method is time consuming,so we can only obtain a limited number of data and it is difficulty in obtaining a complete SWCC.Therefore,curve fitting is usually performed on a limited number of SWCC test data using a parametric model.However,there are many SWCC parametric models in literature.Determination of SWCC parametic model with a limited number of test data is inevitably affected by model selection uncertainty and parameter uncertainty.How to reasonably select the SWCC model based on a limited number of experimental data and determine the model parameters is one of the major challenging.In addition,how to properly incorporate the uncertainty of SWCC model selection and parameters into to the safety and reliability analysis of geotechnical engineering structures(such as earth-rock dam)is also an urgent problem to be explored.To address the above problems,this paper compares and analyzes how to select the SWCC model which commonly used methods(such as Bayesian method,Akaike information criterion,Bayesian information criterion),and develops the software of WHUSWCCPCS for the cases with a limited number of test data.The software is used to analyze 90 sets of SWCC test data of loess,which provided a reference for determining the SWCC model of loess.Then,the influence of SWCC model selection uncertainty and parameter uncertainty on the reliability of earth-rock dam slope stability is studied.This study and major conclusions are summarized as follows:(1)This paper conductes a comparative study of SWCC model selection methods.From the perspective of model fitting and penalty,Bayesian method,AIC and BIC for model selection are compared.It is found that the Bayesian method is superior to the other two methods for model selection with a limited number of test data.The model fitting may dominate the information creterion of SWCC model selection.The penalty is relatively small,but its effects cannot be ignored.Based on the above methods,WHUSWCCPCS was developed,which provides an effective tool for SWCC model selection and parameter identification with a limited number of data.(2)A number of 90 sets of loess SWCC test data were collected.Each set of SWCC data was analyzed by WHUSWCCPCS.The SWCC model of each set of data and its corresponding parameters were determined based on the Bayesian method.The estimates of effective degree of saturation from the models identified by Bayesian method were faverbally compariable with test data.This indicates the rationality of the Bayesian method.According to the analysis results of 90 sets of data,the prior information of the SWCC model parameters is updated.It is found that the most probable model corresponding to68.9% of sets of data is the Fredlund–Xing model,which provides a reference for the model selection and parameter identification of the SWCC.(3)A robust reliability analysis method is proposed for earth-rock dam slope stability considering SWCC model selection uncertainty and parameter uncertainty.The method is illustrated by taking clay core wall dam as an example,and two type of uncertainties and their effects are discussed.The proposed method enriches the analysis method of stability and reliability of unsaturated earth-rock dams,and provides technical supports for stability safety and reliability control of earth-rock dams.