|About this Abstract
||2018 TMS Annual Meeting & Exhibition
||Additive Manufacturing: Building the Pathway towards Process and Material Qualification
||A-25: Applied Machine Vision and Machine Learning to the Characterization and Qualification of Additive Manufacturing Powder Feedstock
||Anna C. Smith, Brian L. DeCost, Elizabeth A. Holm
|On-Site Speaker (Planned)
||Anna C. Smith
Metal additive manufacturing is an emerging technology providing flexibility and rapid innovation for prototyping and low volume manufacturing. Understanding the correlation between metal powder physical characteristics and the effect on process parameters and outcomes is dependent on robust characterization and qualification methods. Standard methods of characterization such as particle size distribution exclude important physical characteristics, specifically powder morphology, surface texture, and defects.
To address these challenges, we have developed an automated system to characterize metal powder feedstock materials by employing computer vision and machine learning methods. In this work, we analyze various gas-atomized powders and characterize them using convolutional neural networks and clustering algorithms to group powder particles based on morphological similarities. This approach may enable future applications in powder qualification, quantifying the effects of recycling, selecting build parameters based on powder characteristics, and defining objective material standards based on microstructure.
||Planned: Supplemental Proceedings volume