Multivariate Probability Distributions And Spatial Variability Of Clay Parameters In Shenzhen

Making full use of geotechnical engineering survey data to achieve refined characterization and modeling of stratigraphic and geotechnical parameters can provide parameter guarantees for geotechnical engineering reliability design,and is an important part of geotechnical engineering digitization,informatization,and intelligence.This paper draws on the idea of geotechnical big data,strives to reduce the uncertainty of geotechnical parameters through survey data in a large area,and establishes a refined model of multi-dimensional distribution of geotechnical physical and mechanical parameters.The main research contents are as follows:(1)Collected and sorted out 34 engineering survey reports in Shenzhen,established the global database SZ-CLAY/13/3542 of 13 geotechnical parameters of Shenzhen clay.the Shenzhen clay and the global clay are compared and analyzed.the statistical characteristics of each parameter and the marginal probability distribution function of a single parameter are analyzed.Shenzhen clay has the characteristics of strong compressibility,high natural water content and high saturation,and most parameters do not conform to normal distribution.(2)Considering the cross-correlation between multiple parameters,the global multivariate probability distribution model of seven clay parameters is established by using equal probability transformation and multiple sampling techniques.Based on Bayesian update,the global probability distribution model is used to update the parameter probability distribution of a project site,which significantly reduces the uncertainty of parameters.(3)Considering the spatial autocorrelation of clay parameters,the spatial variability of multiple clay parameters was analyzed,and the autocorrelation distance in the case of sparse data was calculated by multiple sampling technique and autocorrelation function method.The results show that in the same site,the autocorrelation distances of different parameters are relatively close.