Deformation of geogrid-reinforced soil wall: Usage of random forests regression analysis

An essential component of designing geosynthetic reinforced soil walls (GRSW) is deformation analysis. Nonetheless, research highlights how artificial intelligence techniques may be used to solve geotechnical engineering problems. This study's primary goal was to investigate the potential use of machine learning-based techniques for GRSW deformation (Dis) estimate. This paper presents and validates new methods that combine random forests (RF) with the ant lion optimization (AnOA), the chimp optimization algorithm (ChnO), and the gannet optimization algorithm (GAOA). The dataset for this purpose was created by combining 166 finite element studies that have been done in the literature. The findings presented that the RF(AnOA), RF(ChOA), and RF(GaOA) methods have a significant ability to accurately predict the 𝐷𝑖𝑠 of GRSW with R2 values larger than 0.976. The value of Theil inequality coefficient (TIC) was 0.0463 and 0.0282 in the learning and examining sections for 𝑅𝐹(𝐺𝑎𝑂𝐴), remarkably smaller than those of RF(ChOA) at 0.0523 and 0.063, and RF(AnOA) by 0.0564 and 0.0799, respectively. In conclusion, the results suggest that the suggested models may be used to evaluate the effectiveness of geosynthetic reinforced soil structures. This research provides a significant contribution by establishing a scalable and efficient framework for evaluating the deformation performance of GRSW structures, bridging the gap between computational geomechanics and machine learning. The proposed RF models can replace or complement traditional numerical methods for estimating GRSW deformation, saving time and computational resources. Using these models, engineers can predict GRSW deformations with high precision, enabling more accurate design and better safety assessments of geotechnical structures.