A rheometry based calibration of a first-order DEM model to generate virtual avatars of metal Additive Manufacturing (AM) powders

Abstract

The technology of metal powder-bed Additive Manufacturing (AM) has been evolving rapidly over the last few years. Creating parts through additive manufacturing within acceptable tolerances for porosity, strength and roughness is contingent upon many process variables. One important factor is the ability to spread uniform layers of loose powder for a given thickness over a given coverage area. This is related to the angle of repose of the powder and its rheological behavior i.e., the flow or shear motion of the powder. As in the case of spreading which is essentially a shear flow under a load, in this work, a benchtop powder rheometer with the capabilities to study bulk flow performance of AM powders is utilized to characterize a metal AM powder. When used in tandem with powder dynamics modeling, the rheometer can provide the powder rheological parameters to quantify ‘spreadability’ i.e., the ease with which a powder will spread under a specified set of conditions. Powder dynamics modeling using the discrete element method (DEM) can simulate powder spreading and capture powder layer quality descriptions such as segregation, porosity and surface roughness but needs to account for true particle shapes and sizes. The requirement to simulate realistic particle shapes and sizes results in computationally expensive DEM simulations with millions of particles. This study is therefore directed towards modeling bulk powder performance and determining the minimum volume of powder to be simulated to understand its rheological properties. This work aims to calibrate a virtual media with monodispersed spherical particles against angle of repose and rheological properties of a real media made up of almost spherical particles and comparable sizes. This calibrated media, which is the “virtual avatar” of the real powder, is also tested for rheological properties at varying load conditions and can further be made available to study the ‘spreadability’ of AM powders. The model compares well with published results for a benchmark granular media (2 mm glass beads) and has moderate accuracy for in-house experiments on a well-known AM powder (100 − 250μm Ti-6Al-4V). CPU-based serial computing is used for modeling around 8000 2 mm glass beads and subsequently, significantly faster GPU-based parallel computing is used to model 1.3 million 250μm Ti-6Al-4V powder.