Rheological Characterization Of Water Atomized Tool Steel Powders Developed For Laser Powder Bed Fusion By Supervised And Unsupervised Machine Learning

Authors:

Denis Mutel (Université Laval, Canada),
Simon Gélinas (Université Laval, Canada)

Abstract:

Metal powders developed for powder-bed additive manufacturing processes need to achieve specific flow characteristics to be considered suitable. However, the relationship between powder flow and the morphological characteristics of individual particles can be difficult to establish. In this context, artificial intelligence appears to be the perfect tool to clarify the imprecision surrounding this type of interaction. The work summarized in this manuscript first uses a neural network architecture (Mask R-CNN) allowing the segmentation of individual water-atomized particles in micrographs acquired in scanning electron microscopy. The micrographs of individual particles or their shape descriptors (roundness, elongation, etc.) are then processed using different machine learning (ML) strategies, supervised or not, to relate the information collected on individual particles with the rheological properties of powder specimens. The approach developed aims to acquire new knowledge regarding specific particle characteristics that are required to optimize powder flowability for laser powder bed fusion.

DOI:

https://doi.org/10.59499/WP225368779