Study On The Rodman Filed Model Of Xi’an Loess Parameter

In the process of reliability analysis of geotechnical engineering, we will find that thereliability index calculated are often too small, the failure probability is too large—oftenmore than10%, even if which has been proved to be security. In addition to the largedispersion of the rock and soil material properties, the main reason that caused this departurefrom the theoretical and actual condition is due to the use of random variable theory todescribe the probabilistic characteristics of soil, where exist a lot of problems: Soilparameters in the same layer are treated as random variables, which ignores the correlationbetween parameters of different test points, at the same time，the statistical index does notreflect the spatial average characteristics of sillmass. Therefore, the use of traditional randomvariables theory to calculate the probability distribution of geotechnical parameters is notdependable. At present，method of simulating parameters of soil spatial variability andcorrelation is Vanmarkce (1977) random field theory.This paper introduced the random field theory into the statistical calculation ingeotechnical parameters. Described the basic concept of random field theory and its digitalcharacteristic, its stationarity and ergodicity. Introduced the concept of variance reductionfunction Γ2(h) and the correlation distance δ. Two methods of correlation distance calculation,namely recursive subspace method and correlation function method and its improvement areprovided. And put forward a new method.Collected the geotechnical test data of Xi’an area, gain the vertical distribution structureof Xi’an loess on different geomorphic units, and a large number of cone penetration test(CPT)data. By collected data to established the random field model of Xi’an loess. Homogeneoustreatment on the raw data is discussed; the applicability of the model was tested by theMatlab language programming; correlation distance of each layer of xi’an loess wascalculated, used different methods and different test indexes as samples; According to thecalculation results, Analyze the reasonable value of spatial scale h that used to calculate thecorrelation distance of xi’an loess; discussed the differences between the different testparameters and different calculation methods for the relational distance, And analyze the reason why differences exist; Statistical distribution characteristics of the calculatedcorrelation distance of each layer of Xi’an loess was discussed; Probability distribution modelof Xi ’an loess correlation distance were established, model parameters were estimated.Goodness-of-fit test on the model is Conducted， using the Pearson chi-square method, testresults show that its comply with the Beta distribution.