Performance assessment of a foundation resting on reinforced collapsible Sabkha soil by deep soil mixing columns using machine learning analyses

Abstract

The current research aims to enable the safe construction of structures on collapsible sandy Sabkha soil, which is found both inland and along coastal regions. To reinforce this type of soil, and using the Tabuk region in Saudi Arabia as a case, the present study uses deep soil mixing columns (DSMCs) containing hydrated lime as a binder. Full-scale three-dimensional numerical models are created based on experimental tests to analyze the performance of footings on both untreated and treated collapsible soil with DSMCs. Various parameters such as column configuration, area replacement ratio and lime content are considered. The results demonstrate that DSMCs significantly increase the footing ultimate bearing capacity on Sabkha soil under immersion conditions, giving it a bearing capacity of up to three times that of the tested non-treated soil. Furthermore, with increases in the DSMCs’ area replacement ratio and lime contents, there is a reduction in the footing settlement and DSMC bulging as well as a rise in the stress concentration and stress transfer. Under the applied loads, the DSMCs are shown to yield before the surrounding soil. Multiple machine-learning (ML) models, including LR, NLR, SVM, GPR, RF, DT, and XGBoost, are also developed in the research. In addition to the parameters used in FEM analyses, collapse index, maximum dry unit-weight, and the California bearing ratio are utilized in ML analyses producing ultimate bearing capacity as an output. These models are found to exhibit strong performance in predicting the properties of the treated Sabkha soil, with R2-scores exceeding 0.90. The high accuracy of the selected models highlights their effective predictive abilities in assessing reinforced soil properties. Finally, an equation is formulated based on ML analyses to calculate the footing ultimate bearing capacity on the reinforced Sabkha soil with DSMCs from the used features. The use of hydrated lime as a sustainable binder in DSMCs offers environmental benefits and technical advantages by enhancing foundation performance while promoting sustainable construction practices.

Keywords

• Sabkha soil
• Deep soil mixing columns
• Lime
• Bearing capacity
• Machine-learning models