Study on Soil–Reinforcement Interaction and Field Performance of Reinforced Soil Walls for Geotechnical Applications Using Computer Applications and Artificial Intelligence

Abstract

In geotechnical engineering reinforced soil walls have proven to be a viable means of increasing slopes' stability and retaining structures. This study examines the soil-reinforcement interactions and field performance of soil-reinforced walls thru the use of numerical analysis, computational tools, and artificial intelligence (AI). The methodology proposed assesses horizontal displacement, settlement, and tensile force, strain of reinforcement, pullout resistance and factor of safety for various load conditions. Three AI models are created to predict the behavior of the wall, namely “Artificial Neural Network (ANN)”, “Random Forest (RF)”, and “Extreme Gradient Boosting (XGBoost)”. The ANN model results in a high coefficient of determination (R2 = 0.98) with a value of MAE and RMSE of 1.21 mm and 1.68 mm respectively, whereas the training and validation harms were equal to 0.0021 and 0.0032 respectively. XGBoost model had the highest prediction accuracy of 98.6%, whereas ANN and RF models had prediction accuracy of 97.9% and 95.8%, respectively, while all other models had prediction accuracy less than 97.3%. The results show the ability of AI based computer applications to predict soil–reinforcement interaction with reliability and their ability to improve the design, stability analysis and field performance assessment of reinforced soil barriers.