Concurrent multi-scale modeling of granular materials: Benchmarking volume-coupled DEM-FEM models across elastic and elasto-plastic regimes

We develop and benchmark a concurrent DEM–FEM scheme that couples particle and continuum physics via an overlapping region with penalty-enforced kinematic compatibility, implemented in Kratos Multiphysics. Analysis spans three levels: (i) a mono-disperse elastic column with an exact reference, (ii) an elastic polydisperse assembly, and (iii) an elasto-plastic polydisperse assembly whose continuum follows a Drucker–Prager law calibrated to DEM data through Bayesian inference. We quantify how penalty magnitude, spatial weighting, and mesh-to-particle size ratio govern stress/strain transfer, lateral stress ratio K , and the emergence of local overlap artefacts, and show that smooth weights and sufficient Gauss-point sampling suppress spikes without degrading accuracy. Across friction levels, the hybrid model closely reproduces DEM axial hysteresis and lateral–vertical stress paths, and macroscopic discrepancies align with contact-scale slip statistics, highlighting the limitations of an elastic–plastic (Drucker Prager) FEM model in capturing dilatancy evolution and path dependence. The results provide practical guidance on parameter selection and hybrid-zone design, demonstrate scalability to large granular systems, and indicate the specific flow regimes in which a more advanced continuum model (e.g. with non-associated plasticity, strain hardening/softening, or fabric evolution) would be needed to fully match DEM.