| About this Abstract |
| Meeting |
MS&T22: Materials Science & Technology
|
| Symposium
|
2022 Undergraduate Student Poster Contest
|
| Presentation Title |
Computer Vision and Machine-learning Approaches to 2-D High Energy X-ray Diffraction Image Analysis for Phase Detection |
| Author(s) |
Zhuldyz Ualikhankyzy, Weiqi Yue, Pawan K. Tripathi, Nathaniel K. Tomczak, Gabriel Ponon, Laura S. Bruckman, Matthew A. Willard, Vipin Chaudhary, Roger H. French |
| On-Site Speaker (Planned) |
Zhuldyz Ualikhankyzy |
| Abstract Scope |
High energy X-ray diffraction (XRD) at synchrotron beamlines provides information about materials’ crystalline structure and properties such as crystalline phases, texture, and solute dislocation density, to name a few. Most of the XRD analysis software developed during the last decades required human assistance resulting in an inherited bias in handling large volumes of data. They also relied on transforming the 2-D grayscale images into 1-D intensity plots. Although functional, this method involves many steps reflected in its low analysis throughput, which may lead to terabytes of data never analyzed. We explored a novel approach to build a comprehensive pipeline based on Image Processing, Computer Vision, and Machine Learning (ML) algorithms utilizing distributed and high-performance computing and have developed a Convolutional Neural Network (CNN) to perform on-the-fly phase identification. A statistical analysis performed to ensure the quality of image processing provides insights into potential ways to improve the approach developed. |