About this Abstract |
| Meeting |
MS&T24: Materials Science & Technology
|
| Symposium
|
Materials Processing and Fundamental Understanding Based on Machine Learning and Data Informatics
|
| Presentation Title |
Design of Microstructure in Zn-Al-Mg Alloys Using Integrated Finite Element Analysis and Deep Learning Techniques |
| Author(s) |
Khushahal Thool , Preetham Alluri, Ki-Seong Park , Wi-Geol Seo, Shi Hoon Choi |
| On-Site Speaker (Planned) |
Shi Hoon Choi |
| Abstract Scope |
The Zn-Al-Mg (ZAM) alloy wires are renowned for their superior mechanical strength, corrosion resistance, and electrical conductivity. This study introduces a novel integrated method that combines finite element simulation (FEM) and deep learning via convolutional neural networks (CNN) to shed light on the interplay between microstructural heterogeneity and ductility in extruded ZAM alloy wires. A specialized machine learning (ML) model—a UNet-based CNN architecture with a ResNet-34 backbone—has been meticulously trained on an extensive dataset of extruded ZAM microstructures. A dataframe is compiled and continuously updated with segmentation and morphological features derived from each image. Furthermore, a custom-developed meshing algorithm simultaneously transforms the ML model's segmentation results into a batch file compatible with ABAQUS for further FEM analysis. The maximum effective strain value—a marker of the material's ductility—is then extracted at the conclusion of each simulation and added to the dataframe. |