Abstract

Powder bed additive manufacturing (AM) is comprised of two repetitive steps: spreading of powder and selective fusing or binding the spread layer. Powder-bed AM can be sub-categorized as fusion-based where electron beams or laser beams are used to fuse the spread powder layer and binder-based where a liquid binder is used to bind the spread layer at areas specified by the governing CAD model. The latter process, commonly referred to as binder jet additive manufacturing (BJAM), outperforms fusion-based methods with respect to cost, build time, and material suitability; however, the parts are prone to shear-induced deformation during the powder spreading stage. Unlike fusionbased AM, the strength of BJAM parts is not fully developed until sintering and infiltration during postprocessing. This results in BJAM parts being more susceptible to deformation or even breakage due to the shearing action of the spreader. This shear-induced deformation can affect the precision and thereby performance of 3D printed parts. The binding step in BJAM is a complex function of binder viscosity, density, droplet size, impact speed, and drying time. The spreading step is a complex function of spreader speed and spreader shape, topography of spread and bound layer, and the rheology of the AM powder. This study presents a first-order model to simulate BJAM using a weak concrete-like, non-local, multilayer bonded DEM model. The DEM model has been parallelized using the massive parallelism offered by GPUs. An industry-grade metal powder is used to print physical cuboids at varying spreader speeds. The model is qualitatively verified against experiments on a real 3D printer. The model can be used to provide layer-wise spreading process control to minimize spreader shear-induced deformations.