Optimizing nailing parameters for hybrid retaining systems using supervised learning regression models

Abstract

The work focuses on creating a hybrid retaining wall using geocell, geogrid, and soil-nailing techniques for a road embankment in Mangalore, India. Soil nailing reinforces the soil, geogrids give extra support, and geocell serves as a protective facia against external weathering impacts, decreasing the requirement for conventional shotcreting and lowering the carbon footprint of concrete. This promotes the United Nations’ Sustainable Development Goals (SDGs). The usage of concrete and steel in soil nailing can be minimized using supervised learning regression models (SLRMs), a branch of machine learning (ML). The soil properties in the site were collected by standard penetration tests (SPT). From the limit equilibrium method (LEM) study, 600 iterations are carried out to estimate the factor of safety (FoS), which serves as input training and testing data for the ML model. The surrogate model produces findings for the entire site to identify ideal nail parameters. The random forest (RF) model was found to be useful with a mean square error (MSE) value of 0.009. The finite element method analysis (FEM) yields a modest overestimation of roughly 4.5% while validating the results of the RF model in a typical slope. This study demonstrates the practical application of sustainable methodologies and machine learning to meet crucial development goals, explicitly improving slope stability and road development in the study area through environmentally conscious engineering practices. © The Author(s), under exclusive licence to Springer Nature Switzerland AG 2024.