The shear strength parameters, cohesion force(c) and angle of internal friction (ф) are two important properties required to determine the bearing capacity of soil in building/construction of geotechnical structures. In this paper, an attempt was made to predict the soil shear strength parameters. A comparison of machine learning tools like Classification and Regression tree (CART) and Feed Forward Neural Network (FFNN), considering the significant index properties of soil was done. The index properties considered were Plasticity Index (PI), Bulk Density (γ), Silt Percentage (STP), Sand Percentage (SP), Clay Percentage (CP) and water content (w). The performance of the models was validated by calculating the statistical parameters, d (Index of agreement), FB (Fractional Bias), NMSE (Normalized Mean square Error) and MB (Model Bias). A sensitivity analysis was carried out to determine the influence of the input parameters on ‘c’ and ‘ф’. Results, showed that STP and CP have a major influence in prediction of cohesion force, while, SP greatly influences‘ф’. The results also showed that CART is better in prediction of ‘ф’ while FFNN is good at prediction of ‘c’. However, a combined (cand ф) test by FFNN model did not produce any significant results.
Volume 11 | 09-Special Issue
Pages: 882-894
DOI: 10.5373/JARDCS/V11/20192647