CBR of stabilized and reinforced residual soils using experimental, numerical, and machine-learning approaches

This research employed coir fiber and activated carbon as eco-friendly additives and waste materials while using lime as a traditional binder to improve the properties of two residual soils. The Finite Element Model (FEM) was done to predict the California Bearing Ratio (CBR) value according to the shear strength parameters and elastic modulus, which were determined by the direct shear and unconfined compressive strength tests, respectively. Furthermore, the CBR value was predicted using the machine-learning technique's random forest (RF) model. The results indicated that the combination of lime, activated carbon, and coir fiber improved CBR value in both soils by 92.79% and 91,41% for heavy compaction. In comparison, the CBR value in lime-treated specimens reached the CBR values of 82.36% and 60.61%, while in activated carbon with coir fibre specimens, the CBR enhanced to 23.24% and 26.78%, respectively. The performance of the FEM model and RF model were evaluated using error indices. The plot's coefficient of determination R2 between laboratory and FEM model results was more than 0.95. Moreover, the RF model demonstrated acceptable predictive ability and a high level of accuracy, as evidenced by its R2 values of 0.925 for the training data and 0.937 for the testing data. The analyses indicate that the combination FEM and RF method effectively predicts the CBR value.

Automated summary

This research used coir fiber and activated carbon as additives and waste materials, and lime as a binder to improve the properties of two residual soils. The Finite Element Model (FEM) and random forest (RF) model were employed to predict the California Bearing Ratio (CBR) value, and the results showed that the combination of lime, activated carbon, and coir fiber significantly improved the CBR value in both soils.