Application of FEM and Artificial Intelligence Techniques (LRM, RFM & ANN) in Predicting the Ultimate Bearing Capacity of Reinforced Soil Foundation

Abstract

In this research paper, the behavior of shallow footing with square and rectangular shapes over geosynthetic reinforced soil was studied. A novel geogrid called “3D tube-geogrid” was utilized for this work. The impact of various reinforcement parameters, including the depth of the final layer (z), length (l), inclination (α), filler material used inside the geogrid tube, relative soil density, and the tensile stiffness of the geogrid (EA), were analyzed by running numerical simulations using PLAXIS 3D V20 software. The simulated data were used to quantify the relationship between the ultimate bearing capacity of the soil and the reinforcement parameters. Several artificial intelligence (AI) techniques, such as linear regression analysis, a random forest model, and an artificial neural network (ANN), were employed on the generated dataset. To evaluate the preciseness of these techniques, various statistical indicators, such as the squared correlation coefficient (R2), mean absolute percentage error (MAPE), mean squared error (MSE), and root-mean-square error (RMSE), were calculated, and error percentages of 20.98%, 12.5%, and 6.4% were obtained for the linear regression, random forest, and ANN, respectively. The numerical study determined the optimal values of the reinforcement parameters length, z/B, inclination, and filling material to be 4B, 3, 0°, and aggregate, respectively.