DEM SIMULATION OF CEMENTED SAND UNDER DRAINED TRIAXIAL CONDITION

Abstract

This study presents a discrete element method (DEM)-based numerical investigation into the mechanical behavior of cemented sand under triaxial compression. The primary objective is to optimize the simulation parameters by minimizing the deviation between numerical and experimental stress-strain responses. Bayesian optimization via the Optuna framework was employed to systematically explore the parameter space, utilizing an objective function that combines root mean square error (RMSE) and Euclidean distance between experimental and simulated data. Sensitivity analysis was performed on 15 parameters, identifying 10 as the most influential in replicating cemented sand behavior. Through 200 optimization iterations, the best-fit parameters were determined, capturing essential mechanical properties such as stiffness, cohesion, and friction. Numerical simulations provided insights into displacement evolution and contact force distribution, demonstrating strong agreement with experimental observations. At the end, a parallel coordinates plot was employed to visualize the multidimensional parameter space, highlighting correlations and trade-offs among the variables. The post-optimization analysis validated the consistency of the sensitivity study, reinforcing the reliability of the calibrated parameters. The findings contribute to the broader understanding of DEM calibration for cemented geomaterials, particularly in applications requiring high-fidelity simulation of mechanical responses. This research provides a systematic approach for improving DEM model accuracy, reducing reliance on trial-and-error parameter selection, and paving the way for further studies in cemented sand behavior modeling.