A computational framework that utilizes the discrete element method (DEM) was developed to conduct triaxial compression tests on lithium-based pebbles. In addition, the process for determining optimized simulation parameters for performing stress-controlled numerical hydrostatic and triaxial compression tests using LIGGGHTS has been outlined. This framework was employed to study the influence of the polydispersity of pebbles on the Drucker-Prager (D-P) parameter, which is the friction angle. The study revealed that the effect of polydispersity, measured by the polydispersity index ( \(\lambda \) ) was negligible, as the value of friction angle ( \(\beta \) ) remained constant at \(\approx 29^\circ \) for \(\lambda \le 0.5\) . However, in highly polydisperse samples with \(\lambda =0.8\) , \(\beta \) increased to \(\approx 31^\circ \) . Additionally the dilatancy angle ( \(\psi \) ) decreases as \(\lambda \) increases. The difference between \(\beta \) and \(\psi \) increases with \(\lambda \) , and thus, the associative flow rule is not suitable for highly polydisperse samples. To conduct a more detailed analysis of the mesoscopic mechanics, three parameters based on the mean number of contacts, which governed the microstructural similarity of the sample and the extent of particle participation, were examined. Additionally, two local parameters based on Voronoi tessellation were investigated. These parameters highlight how changing \(\lambda \) influences the local deformation and characterizes the local structural variation in the granular samples. In particular it was found that, the participation of particles to the total deformation was higher in samples with high polydispersity.
