Study Of Reliability For Loess Slope Based On Random Finite Element

Analysis of loess slope stability has long been a hot topic in geotechnical engineeringstudies, but because of variability of loess strength parameters, the deterministic approach ofslope stability assessment is difficult to obtain reliable results. The study of verticalcorrelation length by the spatial variability and correlation characteristics of loess strengthparameters is the important part of this paper. Using Monte-Carlo simulation of stochasticfinite element method, the high loess slope stability factor and failure probability has beensystematically studied in Ganquan area, and the impact of correlation length on theprobability of failure and stability factor has been studied. Research results are as follows:(1) The engineering geological conditions in Ganquan area were analyzed, and the dataof soil test from Huangling-Yan’an expressway investigation reports were collected,counting the physical and mechanical parameters of loess in study area according tochronologic age.Meanwhile the probability distribution of strength parameters wereanalyzed,and the results showed that the area of values are acceptedlognormal distribution. The spatial correlation length were abtained by recurrence spaceaverage method, space mean method and space trend method, respectively,based on CPT data.The results of recurrence space average method and space mean method showed that thecalculations based on lateral friction data were the calculations based on tip resistance dataabout2times. Spatial correlation length based on tip resistance data obtained by these threemethods is0.76-2.85m, based on lateral friction data is0.64-6.32m.(2) The study of influence of geotechnical parameters related to spatial correlationlengths from the reliability analytical results that when the slope failure possibility is less than38.0%, with increasing spatial correlation lengths failure possibility increases, it means thatthe more uniform the loess slope will increase the slope stability and the heterogeneity willimprove slope stability. Conversely, when the slope failure possibility is greater than38.0%,with increasing spatial correlation lengths failure probability decreases, it means that themore uniform the loess slope will increase the slope stability and the heterogeneity willimprove slope instability. In simplify project,taking slope to be an uniform slope or bedivided into2-3large layers,the results will overestimate the stability of the slope, getting adventurous results. And for a more stable slope, slope stability is underestimated, getconservative results.