About this Abstract |
| Meeting |
MS&T25: Materials Science & Technology
|
| Symposium
|
Integrated Computational Materials Engineering for Physics-Based Machine Learning Models
|
| Presentation Title |
Machine Learning Model for Estimating the Number of Grains in Ti–6Al–4V XRD Patterns |
| Author(s) |
Gabriel Obsequio Ponon, Mohommad Redad Mehdi, Ozan Dernek, Hemant Sharma, Pawan Tripathi, Roger French |
| On-Site Speaker (Planned) |
Gabriel Obsequio Ponon |
| Abstract Scope |
Grain distribution information is essential for characterizing the mechanical properties of metallurgical samples. Though computer vision-based methods abound in literature, they often operate on SEM and TEM micrographs, limited to a localized sample surface region. This work explores the use of High-energy XRD patterns to identify the number of grains in metal samples of Ti–6Al–4V (Ti-64). We developed an automated approach for synthesizing Ti-64 microstructures with equiaxed β grains and α laths using DREAM.3D, which we used to simulate textured and non-textured XRD patterns with known “ground truth” grain counts. Machine learning models are trained on this synthetic dataset for identifying grain count in experimentally derived XRD patterns. This work demonstrates that ensemble simulations and ML models can be used for characterizing grain distributions in real XRD datasets. The methods shown also have the potential to enable in-situ characterization in synchrotron facilities. |