Increase in packing density during multi-layer powder spreading: An experimental and numerical study

TL;DR

This work addresses the unexplained densification of powder packing observed during multi-layer spreading in metal additive manufacturing. It combines a custom experimental setup with large-scale DEM simulations in the Lethe framework to reproduce and analyze the phenomenon. The key finding is that densification scales with the build-plate length and arises when two vertical-wall–induced static zones merge in the powder bed, constraining particle motion during subsequent layers. DEM calibration with multiple parameter sets shows that PS1/PS2 can capture the experimental densification for longer domains, linking surface properties to the onset and magnitude of densification, with implications for recoater design and layer uniformity.

Abstract

A custom apparatus designed to isolate and replicate the spreading process of metal powder in additive manufacturing demonstrates a sudden and unexplained increase in packing density beyond layers 5 to 10. We replicate the experiments that lead to densification with the discrete element method (DEM) using \lethe{}, an open-source software framework. We show that large-scale multi-layer DEM simulations are able to reproduce the densification observed experimentally. Using the Lagrangian simulation results, we highlight significant particle displacement in the powder bed at lower layer number, accompanied by static zones generated by the vertical wall surrounding the powder bed. The amplitude of the densification and the layer number at which it starts to occur is correlated to the distance between those two vertical walls which delimit the powder spreading area. This study addresses the gap between mono-layer powder spreading studies on hard-flat surfaces and the actual metal powder-based additive manufacturing processes by providing a better understanding of how the powder bed behaves during multi-layer spreading.

Figures (11)

• Figure 1: Schematic of the increase in packing density observed over many experiments. Packing density as a function of layer number showing an initial increase due to the first layer effect followed by plateau region, densification and decrease during the 20 first layers.
• Figure 2: Functioning of the apparatus that isolate and replicate the spreading process. Side view schematic of the experimental apparatus showing each step of the spreading process. (a) Initial configuration, (b) Extrusion, (c) Spreading, and (d) Measuring steps.
• Figure 3: Overview of the apparatus(a) Isometric schematic view of the experimental apparatus showing important dimension variables of the equipment. (b) Photo of the experiment equipment.
• Figure 4: Powder particle size distribution and numerical modeling(a) Particle size distribution of the Ti-6Al-4V powder from AP&C used in this study with shaded area representing the excluded size range for the simulations. (b) SEM images, produced with a NANOS tabletop SEM from Semplor at 15kV, of the Ti-6Al-4V powder used in the experimental work. (c) Schematic representation of particle contact mechanics showing the interactions between particles i and j, including the normal force ($\bm{F}_{\mathrm{n},ij}$), tangential force ($\bm{F}_{\mathrm{t},ij}$), cohesive force ($\bm{F}_{\mathrm{c},ij}$) and torque due to the tangential force ($\bm{M}_{\mathrm{t},ij}$) and the rolling resistance torque ($\bm{M}_{\mathrm{r},ij}$). (d) Normalized force interaction plot showing total normal contact force as a function of the normalized overlap (excluding the damping term) with different contact conditions illustrated, i.e., no contact (left), zero overlap (center) and contact (right).
• Figure 5: Experimental and numerical results comparison. Cumulative relative density ($\mathrm{CRD}_{}$) and layer effective relative density ($\mathrm{LRD}_{}$) for every property set (PS1, PS2 and PS3) and domain length (L) combinaisons. Results for the domain length a) L$_1$, b) L$_2$, c) L$_3$ and d) L$_4$. When the domain length is increased, the $\mathrm{CRD}_{}$ and $\mathrm{LRD}_{}$ values of PS1 and PS2 approach the experimental measurements, while PS3 remains largely unchanged.