Geomaterials, such as soils and rocks, are discontinuous and multiphase in nature since they are a mix of particulate elements and different constituents. The discrete nature of geomaterials yields a complex media to deal with regarding predictive capabilities that can only be achieved through expensive computational simulations. The computational modelling of geotechnical problems may consider continuum mechanics or discrete element frameworks to simulate the behaviour of all the features involved. Undoubtedly, continuum-based simulation techniques are predominant in geotechnical engineering practice and have significantly influenced our comprehension of soil response. However, continuum simulations do not consider special features of granular matters at the grain scale that finally determine the characteristic behaviour of particulate materials. On the other hand, the Discrete Element Method (DEM) is in fact based on the inter-particle physical processes at the micro-scale and, in the last decades, has been widely used to simulate a large range of geomaterials, allowing quantitative and qualitative analyses.However, it lacks the scalability and efficiency of models based on continua.This investigation is focused on the modelling of the macroscopic mechanical response of granular assemblies, as a natural consequence of their micromechanical properties, but employing the micro-/mesoscopic method described above. Simulations of true triaxial tests (TTT), an important test in geotechnical engineering, using 3D virtual soil samples are executed with an advanced DEM code that can model assemblies of grains that are made of spherical or non-spherical particles. A DEM-based calibration methodology is thus proposed giving insights on the reliability of the discrete method to replicate soils' mechanical behaviour.ii