Predictive Analysis Of Sand Shear Properties Based On Particle Characteristics

The shear strength of soil is an important mechanical property of soil.In geotechnical engineering,the shear strength of sand is affected by many factors,mainly including density,particle shape,particle size distribution,and saturation.In this paper,the relationship between particle shape and particle size distribution and the shear strength of sand was studied to analyze the influencing factors of shear strength.Explore the feasibility of machine learning method in the prediction of sand shear strength index.A prediction model for the shear strength index of binary mixture sandy soil was established.The main research contents and conclusions of this paper are:（1）In order to study the effect of particle shape on the shear properties of sand,spherical glass beads,irregular glass sand and Fujian standard sand with the same mineral composition and particle size range were selected as the research objects.Glass sand and Fujian standard sand were mixed with glass beads to prepare 12 samples with different particle shapes,and the indoor triaxial consolidation drainage test was carried out.The results show that the dilatation behavior,the peak state deviatoric stress and the internal friction angle φd of the samples are closely related to the particle shape,and they all increase with the increase of the irregularity of the particle shape.（2）Four shape parameters of length-width ratio AR,convexity C,sphericity S and overall regularity OR of glass beads,Fujian standard sand and glass sand were measured by QICPIC.Based on the correlation analysis and simple linear regression analysis,the results show that the shear strength index decreases with the increase of the shape parameter,and shows an obvious linear trend.Among four shape parameters,the overall regularity OR showed the strongest correlation,and the linear regression fit was the highest,which could be used as a representative parameter of particle shape characteristics.（3）In order to study the effect of particle size distribution on the shear characteristics of sand,the coarse particles of 1-2 mm in glass beads,glass sand and Fujian standard sand were mixed with the fine particles of 0.1-0.25 mm,0.25-0.5 mm and 0.5-1 mm,respectively.The particles were mixed according to the proportion of 10%incremental steps to form three binary mixtures with different particle size ratios,and the shear strength index was determined by the triaxial drainage test.The test results show that the particle size distribution has a significant effect on the strength index of the binary mixture,and the shear strength index first decreases and then increases with the increase of fine particle content.In addition,the position of the lowest point of the shear strength index changes with the particle size ratio.For text samples with different particle size ratios,the lowest point of the shear strength index（φd）min and the corresponding fine particle content fcrl all increase with the decrease of the particle size ratio of the binary mixture.（4）Four machine learning algorithms,K-nearest neighbor algorithm,support vector regression,lasso regression and extreme gradient boosting,were used to predict the shear strength index of sand-soil binary particle mixture.Comparing the prediction performance,it is found that the prediction accuracy of the K-nearest neighbor algorithm is excellent,and it can be used to predict the shear strength index φd of binary sand-soil mixtures.Lasso regression models has poor predictive performance.In addition,the importance analysis of characteristic parameters based on the Xgboost model shows that the overall regularity OR has the highest correlation with the sand shear strength index.（5）The influence of fine particle content,particle size ratio and particle shape on shear strength was analyzed.Based on the variation φd～fc curve of shear strength index of binary mixture with different particle size distribution,on the basis of Rolling-ball theory,the theoretical calculation formula of fine particle content at the lowest point of strength index fcrl is established.The relationship between the minimum shear strength index（φd）min and particle shape and particle size distribution was analyzed,and the functional relationship between（φd）min and D/d and OR was obtained by using Matlab R2020b fitting.Combined with each parameter,a prediction model for the shear strength index of sand-soil binary mixture based on particle shape and particle size distribution is established.The results show that the prediction model has high accuracy.