| About this Abstract |
| Meeting |
MS&T24: Materials Science & Technology
|
| Symposium
|
Integrated Computational Materials Engineering for Physics-Based Machine Learning Models
|
| Presentation Title |
Deep Generative Model for Reproducing Microstructure of Low-Carbon Steel During Continuous Cooling |
| Author(s) |
Junya Inoue, Satoshi Noguchi |
| On-Site Speaker (Planned) |
Junya Inoue |
| Abstract Scope |
This paper investigates the application of a deep generative model for efficiently reproducing the complex microstructure of low-carbon steel formed during continuous cooling. The generative model is composed of two deep neural networks, VQVAE and PixelCNN. VQVAE reveals the spatial arrangement of features representing constituent microstructures, while PixelCNN reveals the linkage between the spatial arrangement and the evolution of phase fraction. This approach not only reduces computational cost but also achieves high accuracy in microstructure reproduction. The generative model further leverages the evolution of phase fraction estimated by a separate phase transformation model, such as JMAK. This approach offers a computationally efficient method for predicting microstructure, which is crucial for understanding and optimizing steel properties. The results highlight the importance of integrating domain-specific knowledge with data-driven frameworks for materials design. |