| About this Abstract |
| Meeting |
MS&T22: Materials Science & Technology
|
| Symposium
|
Materials Informatics for Images and Multi-dimensional Datasets
|
| Presentation Title |
B-3: 3D Computer Vision and Machine Learning for Porosity Analysis in Additive Manufacturing |
| Author(s) |
Daniel Diaz, Yuheng Nie, Anthony Rollett, Elizabeth Holm |
| On-Site Speaker (Planned) |
Daniel Diaz |
| Abstract Scope |
Additive manufacturing (AM) is a promising novel technology that is revolutionizing the way we manufacture products, but many properties are limited by porosity produced during processing. In order to better understand the relationship between pore morphologies and properties, it is necessary to accurately identify the characteristic classes of pores observed. To this end we leverage the tools of 3D computer vision and convolutional neural networks to examine the pore morphologies present in datasets collected using X-ray computed tomography (CT). A transfer learning approach is utilized where 3D versions of EfficientNet are initialized with weights that have been trained on ImageNet and converted to a 3D format. Segmented CT image stacks are fed into this pretrained network, and the results are used to divide the pores into clusters that aid in identifying the various morphologies. This has the potential to become a valuable tool for automating the characterization of AM products. |