A data-driven virtual surface method for efficient DEM simulation of particle collisions with microstructured walls

Surface topography strongly influences particle-wall collision dynamics, affecting energy dissipation, adhesion, rebound velocity and angle, as well as rotational motion. To optimize granular processes, microstructuring techniques enable the fabrication of tailored surface topographies that can deliberately modify particle impact behavior. Achieving such targeted control requires a systematic investigation and quantitative characterization of the influence of surface microstructural features on collision dynamics. The Discrete Element Method (DEM) provides a powerful framework for studying these interactions. However, explicitly resolving microstructured surfaces in DEM simulations drastically increases computational costs due to the large number of resulting contact interactions. This work introduces a novel computationally efficient, reduced-order, data-driven DEM approach that accurately reproduces the rebound behavior of particles colliding with homogeneously microstructured surfaces. The proposed virtual surface (VS) method replaces explicit geometric modeling with a post-collision, data-driven correction based on precomputed, microstructure-specific parameters. The approach is validated through particle free-fall tests on two different microstructured surfaces manufactured via Direct Laser Writing. Compared to simulations with explicitly resolved geometries, the VS method reduces computational time by up to 85%, while incurring only a 31% increase relative to a flat wall. • A novel data-driven virtual surface method was developed to efficiently simulate particle collisions with homogeneously microstructured walls in DEM. • The method accurately reproduces microstructure-dependent rebound behavior, including coefficient of restitution, rebound angle, and particle rotation. • Validation against experiments and geometry-resolved DEM simulations showed very good agreement for microstructured surfaces. • The virtual surface method eliminates the need for explicit geometric resolution of microstructures while preserving collision physics. • Computational time was reduced by up to 85% compared to STL-resolved simulations, enabling efficient large-scale DEM simulations.