| About this Abstract |
| Meeting |
Materials Science & Technology 2020
|
| Symposium
|
Additive Manufacturing: Qualification and Certification
|
| Presentation Title |
Connecting Metal Powder Morphological Characteristics with Flowability Properties Using Machine Learning |
| Author(s) |
Srujana Rao Yarasi, Andrew Kitahara, Ryan Cohn, Elizabeth A Holm, Anthony D Rollett |
| On-Site Speaker (Planned) |
Srujana Rao Yarasi |
| Abstract Scope |
Computer vision and machine learning techniques are used to quantify the morphological characteristics of several different metal powders used in Powder Bed Fusion AM and connect them with their flowability. A framework is constructed to understand their differences in flowability based on their powder morphology distributions (PMDs), which includes powder sizes. These PMDs are obtained from SEM images of the powder particles. This is accomplished by using pre-trained convolutional neural networks to generate a feature vector for each powder particle, sets of which are then clustered according to morphological similarity, leading to a powder morphology distribution (PMD) for each powder system. Several machine learning algorithms are tested to correlate the PMDs to their flowability properties. Powder size metrics are also explored as a way to predict flowability behavior in powder bed fusion machines. |