Research On Correlation Data Mining Of Soil Physical And Mechanical Parameters

Soil is a natural material formed through complex actions such as weathering,transportation and sedimentation in a long geological history.It has the characteristics of strong spatial variability.Due to the lack of in-depth understanding of the spatial variability of soil,the current standard design assumes that the soil is homogeneous material,which is seriously inconsistent with the engineering practice.There are two main difficulties in solving this problem: one is the limited soil sampling in the project site,resulting in less soil parameter data,which is difficult to conduct in-depth analysis;the lack of reliable data mining methods for soil parameters makes it difficult to deeply understand the laws behind variability data.Therefore,how to deeply mine and analyze the soil parameter data based on the limited soil physical and mechanical parameter data is of great significance to understand the correlation and spatial variability of soil parameters.This study starts with the analysis of the correlation of soil physical and mechanical parameters.Firstly,this study collected and sorted out more than 10000 soil physical and mechanical parameter data from survey reports and published literature at home and abroad,and independently developed the corresponding database program to realize the classified storage,management and sharing of soil physical and mechanical parameter data,which provides a data basis for the subsequent data mining research on the correlation of soil physical and mechanical parameters.In addition,this study expands the geological database and adds the geological disasters and meteorological database such as landslide and rainfall in Shenzhen,which provides a platform for connecting soil physical and mechanical parameters with landslide disaster prediction.Secondly,based on a large number of existing data,the correlation of soil physical and mechanical parameters is deeply analyzed.Based on the copula function theory,the multivariate distribution functions of soil cohesion,internal friction angle,compression modulus,void ratio and water content are established,and the probability estimation of soil parameters that are difficult to measure is realized.On this basis,the influence of multivariate distribution function of different soil parameters on foundation failure probability is analyzed.The conclusion shows that it is necessary to determine the reasonable category of multivariate distribution function before calculating foundation failure probability.Finally,the autocorrelation of soil physical and mechanical parameters is further analyzed,the three-dimensional Kriging method is developed,and a deep mining method for predicting soil physical and mechanical parameters in unknown strata based on known soil physical and mechanical parameter data in three-dimensional space is proposed.Considering the widespread anisotropy in soil,an anisotropic ellipsoid nested model is proposed.The model can simulate the variation range and semi variogram of soil physical and mechanical parameters in three-dimensional space.The rationality of the ellipsoid nested model is verified by the established data,which provides a core algorithm for the establishment of the law of discreteness and variability of soil parameters.