About this Abstract |
Meeting |
MS&T22: Materials Science & Technology
|
Symposium
|
Additive Manufacturing of Metals: Microstructure, Properties and Alloy Development
|
Presentation Title |
Measurement and Classification of SLM Feedstock Powders by X-Ray Microscopy and Machine Learning |
Author(s) |
Daniel R. Sinclair, Eshan Ganju, Hamidreza Torbati-Sarraf, Nikhilesh Chawla |
On-Site Speaker (Planned) |
Daniel R. Sinclair |
Abstract Scope |
Selective laser melting (SLM) has received attention as a transformative metal 3D printing method in commercial, industrial, and defense applications. However, SLM parts are susceptible to reduced reliability due to defects retained from irregular powder feedstock. As sources of feedstock are diversified to supply expanded manufacturing, granulometry methods are needed which distinguish sources of irregularity. In recycled or composite powder blends, for example, standard stereography conflates fused and elongated particles. In this work, we studied an irregularly shaped, recycled AA7050 feedstock powder with titanium additives (AA7050-RAM2). Lab-scale x-ray microscopy was coupled with automated watershed segmentation to quantify the shapes and sizes of particles. Novel 3D shape factors were developed to describe nonstandard geometries based on concavity and convexity. Additionally, automated methods of classification were compared, using a combination of algorithmic clustering and the Random Forests machine learning algorithm, to maximize efficiency and capture a wide range of powder morphologies. |