Relationship Between Index Properties and CBR of Soil and Prediction of CBR

Abstract

The California bearing ratio (CBR) of soil is the ratio of test load to standard load. The California bearing ratio depends on several factors, i.e., liquid limit (LL), plasticity index (PI), plastic limit (PL), maximum dry density (MDD), and optimum moisture content (OMC). A relationship is developed between LL, PI, PL, MDD, and OMC with soaked CBR. The relationships are mapped using simple linear regression analysis. The correlation coefficients are also calculated for CBR. The experimental procedure to determine the CBR of soil is cumbersome and time-consuming. The California bearing ratio of soil is predicted using an artificial neural network (ANN), decision tree (DT), Gaussian process regression (GPR), support vector regression (SVR), and random forest (RF) AI approaches in the present research work. The performance of ANN, DT, GPR, SVR, and RF is 0.9736, 0.9052, 0.9468, 0.7502, and 0.9292, respectively. The artificial neural network model is identified as the best architecture model based on the model's performance, and soaked CBR of soil is predicted. The actual vs predicted CBR curve is plotted to determine the correlation coefficient, and the correlation coefficient of predicted CBR is 0.9731 determined. It is concluded that the LMNN_CBRs model has the potential to predict the soaked CBR of soil.KeywordsArtificial neural networksDecision treeGaussian process regressionSupport vector regressionRandom forestIndex properties