Ứng dụng trí tuệ nhân tạo trong dự đoán sức chống cắt của đất sau biến dạng

Abstract

After being deformed due to disasters such as erosion, landslides, the soil will significantly change its shear strength. Therefore, it is necessary to forecast the reduction of the shear strength of these soils to predict the possibility of recurrence of unstable erosion with these deformed layers. In this paper, artificial intelligence (RF) will be applied to predict the remaining shear strength of soil after deformation. To perform the simulation, 131 experimental data were collected from literature. The data set consists of four input variables: LL liquid limit, PI plasticity index, Casagrande’s classification deviation ∆PI, CF clay content. The evaluation of the models was made and compared on training data set (70% data) and control data set (30% remaining data) by criteria of Pearson correlation coefficient (R) and RMSE error. The results of the study showed that the random forest model is feasible in determining the remaining shear strength of soil after soil deformation with a correlation coefficient for the training model is 0.97 and verified as 0.78. At the same time, the random forest model can show the importance of each soil property to the remaining shear strength of deformed soil, respectively in the order of Liquid Limit> Casagrande classification deviation ∆PI> Clay Fraction > Plasticity index.