| About this Abstract |
| Meeting |
MS&T24: Materials Science & Technology
|
| Symposium
|
Materials Informatics for Images and Multi-dimensional Datasets
|
| Presentation Title |
Feature Extraction from SEM Images of Fatigue Fracture Surfaces |
| Author(s) |
Anthony Lino, Kristen Hernandez, Austin Ngo, Tu Pham, Roger H French, Pawan Tripathi, John Lewandowski, Laura S Bruckman |
| On-Site Speaker (Planned) |
Anthony Lino |
| Abstract Scope |
Imaging tools such as Scanning Electron Microscopes (SEM) are widely used for material characterization because the morphological images they produce can detect defects from the micrometer to the nanometer scale. This broad range is useful for identifying the optimal process window for Additively Manufactured (AM) parts, but analyzing morphological images requires image segmentation. Image segmentation in fatigue fracture surfaces is laborious because of the quantity and variety of defects on each image, limiting throughput. This work develops a data analysis pipeline for automatically processing SEM images, featuring a U-Net segmentation model designed to differentiate between keyhole and lack of fusion defects and segments the overload and fatigue regions of the fracture surface. These defects can then be associated with either the overload or fatigue regions, which can give insights into the distribution and importance of these defects. This pipeline can now be used in a variety of different applications. |