About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Artificial Intelligence Applications in Integrated Computational Materials Engineering (AI-ICME)
|
| Presentation Title |
Generative AI–Driven Process–Structure–Property Optimization for Materials Discovery |
| Author(s) |
Avanish Mishra, Brenden Hamilton, Mashroor Nitol Shafat, Arindam Debnath, Nithin Mathew, Tim Germann, Saryu J Fensin |
| On-Site Speaker (Planned) |
Avanish Mishra |
| Abstract Scope |
A comprehensive understanding of the process–structure–property relationship is crucial for materials discovery. We established these links by generating a large, realistic dataset of simulated microstructures under varied processing conditions and producing Inverse Pole Figure (IPF) images via virtual texture analysis that replicate information obtained from Electron Backscatter Diffraction (EBSD) experiments. This process generated three-channel images that were used to train a Generative AI model. Based on a variational autoencoder (VAE), this model learns compact microstructural representations and enables smooth interpolation between processing conditions. The trained model successfully characterized experimental IPF micrographs and correlated latent features with processing parameters. Furthermore, to forecast material properties, we trained a Vision Transformer that processes IPF images and VAE latent vectors, along with system embeddings via large language models (LLMs), to produce 2D property maps and predict average atomic properties. This GenAI-driven workflow offers a realistic, end-to-end approach for accelerated materials discovery. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Computational Materials Science & Engineering, ICME |