Purpose Previous studies show the influence of sample preparation on the failure mechanisms of granular materials. This paper aims to investigate the effect of sample preparation formed by uniform spheres to be tested under triaxial compression. Design/methodology/approach A series of specimens is experimentally reconstituted at different sample preparation conditions using the air pluviation method. Perfect non-cohesive chrome steel spheres were selected in two sizes, 3 and 6 mm in diameter, aiming to minimise the effect of inter-particle force due to particle shape and surface roughness. A numerical discrete element model (DEM) based on particle-flow code in 3D is developed to replicate the experimental air pluviation sample preparation process, by following two modelling approaches: air pluviation and radius expansion. Findings The purpose is to validate the DEM sample preparation modelling approaches to reproduce the triaxial experimental stress-strain response. Results show that the DEM model based on the air pluviation sample preparation method provides a lower percentage error with respect to the experimental sample void ratio, when compared to the radius expansion method. Originality/value These finding highlights that the radius expansion method likely introduces a systematic under- and over-estimation of the modelling of strength and compressibility, respectively, in modelling the compression of granular materials. Notice that this study set the basis to produce data to produce alternative laboratory experiments. The ability to rapidly produce such data is particularly beneficial for training machine learning models This study adds to our experimental and numerical modelling research to leverage DEM-generated datasets to produce physics-informed neural networks.
Duong et al. (Tue,) studied this question.