Research On Sampling Representativeness Of Rock Mass Structural Plane Based On K-Medoids Clustering Algorithm

The research shows that there are different structural planes with different directions,sizes and shapes,The instability of rock mass is often the failure of rock mass along the internal structural plane,Therefore,the research on the mechanical properties of rock mass structural planes has gradually become a hot topic of domestic and foreign scholars.And,the size effect exists in the rock mass structural plane itself.It is necessary to obtain different scale and small size structural plane models on the large size structural plane for experimental analysis.But in the actual experimental process,there are often a large number of small-size structural interview samples.If we can select a small number of representative samples from the structural interview samples for experiment,it will greatly reduce the cost.Therefore,it is necessary to study representative sampling methods for rock mass structure interview.In view of this,this paper based on the MATLAB program platform,starting from the representative sampling method of rock mass structure interview,has carried out a series of research.The main achievements include:?1?By analyzing the advantages and disadvantages of the current mainstream statistical methods of rock mass structural planes,the superposition statistical method with better statistical effect is selected as the acquisition method of original sample data for representative sampling of rock mass structural interviews.And based on the MATLAB program platform,the statistical program of the stacking statistics method is compiled,which improves the work efficiency.At the same time,when analyzing the sensitivity of parameters based on the traditional JRC parameter formula,we find that there are some problems in calculating the JRC value of contour curve with the traditional JRC parameter formula.In this paper,the data step size of the data matrix of wheel library curve topography information is uncontrollable when digitalizing the contour curve high-definition scanning image.And the traditional JRC parameter formula can not adapt to the change of sampling interval,which leads to the traditional JRC parameter formula can not accurately calculate the JRC value of contour curve samples.This will hinder the follow-up research,so a JRC calculation model suitable for various sampling spacing values needs to be established.?2?Establishment of JRC Adaptive Model with Multiple Sampling Spacing Values.The traditional JRC parameter formula value only adapts to the specific sampling interval value,it can not accurately calculate the JRC value of the contour curve samples obtained in this paper.In view of this,by analyzing the relationship between Z₂ and ?x of 10 standard roughness profiles proposed by Barton,the average value of Z₂ difference???adaptive function between ?x is fitted.The JRC-Z₂ parameter formula is modified by the adaptive function,and on this basis,the migration and rotation correction are further carried out,so that the JRC adaptive model suitable for various sampling intervals is established.It is proved that the model can still guarantee the accuracy of calculation when x varies arbitrarily in a certain range.Finally,the model calculates the JRC values of each sample and the whole contour curve.By comparing the relationship between the errors of each sample and the parameters D and d,the sensitivity and rationality of the parameters of the push-over sampling method are analyzed.?3?A representative sampling method for rock mass structure interview based on K-Medoids clustering algorithm is proposed.Several contour curves evenly covered in the structural plane are obtained by using simple longitudinal profiler,and the high-definition scanned contour curves are obtained by using engineering scanner.After digitizing the image,the contour curves are counted by the method of stacking statistics,and the original sample contour curve data of the representative sample of the structural interview in this paper are obtained.The Z₂ and?Rp-1?values of each sample contour curve were calculated and analyzed statistically.Considering the numerical relationship between Z₂ and?Rp-1?,the dimensionless values of Z₂^? and?Rp-1?^?are used as the eigenvalues of sample contour curves,and K-center point clustering is applied to the sample set.And the cluster center is the representative contour curve of the sample set.It has been proved that the error between the mean eigenvalue of representative contour curve and the mean of JRC and the mean of JRC of representative contour curve obtained by representative sampling in rock mass structure interview based on K-center clustering algorithm is less than 5% with the increase of K value.When K is greater than a certain value,the representative strength of representative samples obtained by this method and the rationality of this method can meet the requirements.