Civil and Geoengineering Letters Editorial Office info@jcivilgeo.com mEDRA 20260525 DOI Registration 01

06 10.22034/CGEL.3.1.e100060 https://doi.org/10.22034/CGEL.3.1.e100060 Civil and Geoengineering Letters mEDRA 01 01 3115-8129 Civil and Geoengineering Letters Editorial Office IR 01 3115-8129 JB 3 1 01 202605 1 1 A01 Alireza Amiri Amiri, Alireza Alireza Amiri 2 A01 Arash Niromand Niromand, Arash Arash Niromand 3 A01 Ba-Phu Nguyen Nguyen, Ba-Phu Ba-Phu Nguyen 4 A01 Nhat-Phi Doan Doan, Nhat-Phi Nhat-Phi Doan 5 A01 Duy Triet Doan Doan, Duy Triet Duy Triet Doan 01 eng 202605 03 00 Accurate estimation of the bearing capacity of clean sand is a critical task in foundation design, especially in regions where granular soils dominate. Conventional empirical approaches may not fully capture the inherent variability and complex interactions among soil parameters. This study explores the use of machine learning (ML) techniques to predict the ultimate bearing capacity of clean sand based on Standard Penetration Test (SPT) data. A curated dataset of 150 geotechnical records specifically involving clean, well-graded sand (SP and SW according to the Unified Soil Classification System) was compiled. The data were collected from the NCEER database and validated case studies from peer-reviewed journals focusing on sandy soil profiles. Each data entry includes SPT N-values, relative density, dry unit weight, moisture content, and test depth. Four ML models, such as Linear Regression, Decision Tree, Random Forest, and Artificial Neural Network (ANN) were developed and tested using Python’s Scikit-learn and TensorFlow libraries. Model performance was assessed using R², RMSE, and MSE metrics. The Random Forest model yielded the most accurate predictions, with an R² value of 0.93 and a low RMSE, indicating a strong correlation between input features and bearing capacity in clean sand. This study demonstrates that ML-based prediction models can significantly enhance geotechnical design reliability in sandy soils and reduce the reliance on simplified empirical charts and conservative assumptions. 20 Clean sand; SPT; Machine learning; Bearing capacity; Random forest 01 00 20260310 02 00 20260428 2026 Alireza Amiri Arash Niromand Ba-Phu Nguyen Nhat-Phi Doan Duy Triet Doan