Comparative Soft Computing Approaches for Evaluating the Impact of Shearing Parameters on the Bearing Capacity of Shallow Foundations

The structural stability of foundations for most civil engineering infrastructures depends on the resistance properties of foundation soil to neutralize the possibilities of failure by shear. These resistance capacities are based on shear strength parameters. In contrast, the complexity of the analysis and its time-consuming nature pose challenges when analyzing foundation soil. Thus, developing models for computing the bearing capacity of soil foundations from existing laboratory data and field results will help minimize the predicaments and shortcomings associated with analyzing soil foundations. Therefore, this paper correlates observed data of shear strength parameters with predicted values obtained from models developed using Artificial Neural Network (ANN) and Multiple Linear Regression (MLR). The performance indicators obtained from ANN for cohesion (R2, MSE, RMSE, R) are R2 = 0.9749, MSE = 17.0160, RMSE = 4.1250, R= 0.9874, and predicted values of angle of internal friction also show good correlations from the history of the indicators R2 = 0.9499, MSE = 26.5235, RMSE = 5.1500, R= 0.9746. Likewise, MLR shows a promising correlation between dependent and independent parameters; the R2 of cohesion and angle of internal friction were obtained to be 0.8789 and 0.91415. Other indicators are MSE = (82.3801 and 44.8042), RMSE = (9.0763 and 6.6935), R = (0.9374 and 0.9561). The models developed between these parameters present suitable and reliable outputs for determining the bearing capacity of foundation soil using shear strength parameters. In addition, higher precision was observed from the result given by the ANN.