10–12 Nov 2026
Arcotel Wimberger
Europe/Vienna timezone

GPU-Accelerated Discrete Element Simulation of Powder Bed Formation in Binder Jetting

12 Nov 2026, 11:20
20m
Room 2

Room 2

Oral Presentation Digitalisation, Artificial Intelligence & Simulation Simulation

Speaker

Dr Jules Topart (Université of Picardie Jules Verne)

Description

Simulation has become an essential tool for understanding and optimizing
additive manufacturing (AM) processes. In Binder Jetting, part quality is
dictated to a large extent by the powder bed itself: local packing density,
layer homogeneity, and surface roughness directly govern binder penetration,
green-part integrity, and final density after sintering. Yet the recoating
step remains difficult to predict, and conventional simulation techniques are
often too slow and computationally demanding to be deployed in parameter
studies, embedded design pipelines, or digital twins. In this study, we
propose a high-performance, GPU-accelerated Discrete Element Method (DEM)
framework tailored for powder bed formation in Binder Jetting. The framework
resolves individual particle dynamics during spreading, capturing
inter-particle contact, friction, cohesion, and the interaction between the
powder and the recoater as it advances across the build area. By modeling the
powder at the grain scale, we naturally reproduce phenomena that mesh-based or
continuum approaches struggle to represent, including size segregation, void
formation, and the influence of particle size distribution and shape on
packing. The simulation is implemented entirely on the GPU, exploiting
massively parallel contact detection and integration to enable large particle
counts at interactive performance on high-end desktop hardware. This allows
fast prototyping and rapid feedback on recoating parameters such as blade
speed, layer thickness, and powder properties. The capabilities of the
approach are demonstrated through benchmark scenarios relevant to Binder
Jetting, including spreading dynamics, layer-by-layer bed build-up, and
packing-density analysis. The findings indicate that high-fidelity,
near-real-time powder bed simulation is attainable, provided that key
numerical parameters such as contact stiffness, time-step size, and
neighbor-search cutoff are carefully tuned to balance physical accuracy
against computational throughput. This trade-off, rather than raw hardware
performance alone, emerges as the main lever for reaching interactive speeds,
with the calibrated configuration validated against high-fidelity CPU-based
DEM reference computations and experimental packing measurements to ensure the
accelerated model remains physically faithful. Overall, this work unveils new
prospects for process optimization, recoater design, and simulation-based
control within digital manufacturing workflows.

Speaker Country France
Would you like to publish your paper in the special issue of BHM "Berg- und Hüttenmännische Monatshefte" Yes

Author

Dr Jules Topart (Université of Picardie Jules Verne)

Co-authors

Mr Hugues TORTAJADA (Université of Picardie Jules Verne) Dr Valery BOURNY (Université of Picardie Jules Verne) Prof. Stéphane PANIER (Université of Picardie Jules Verne)

Presentation materials

There are no materials yet.