Sensitivity Analysis on Critical Combinations of Input Parameters in DEM Granular Flow Analysis

Abstract This study aims to capture the critical combinations of key parameter in granular flow simulations using Discrete Element Method (DEM) considering the particle size distribution. For that purpose, XGBoost feature importance is employed to quantify the importance of four DEM input parameters, such as friction angle between elements, bottom friction, coefficient of restitution, and spring coefficient in consideration of the particle size distribution. The dominant parameters are identified and then comprehensively explored through the Gaussian process regression (GPR) response surfaces obtained from the results of a series of DEM simulations. By clarifying these mechanisms, this study attempts to identify critical parameter sets for landslide risk assessment throughout the entire parameter space.