| About this Abstract |
| Meeting |
TMS Specialty Congress 2025
|
| Symposium
|
3rd World Congress on Artificial Intelligence in Materials & Manufacturing (AIM 2025)
|
| Presentation Title |
Development of a Microstructure Image Generation Technique and Machine Learning Model for Property Prediction of As-Cast Materials |
| Author(s) |
Chanwoo Park, Wonjoo Lee, Howon Lee, Seong-hoon Kang, Jonghun Yoon |
| On-Site Speaker (Planned) |
Chanwoo Park |
| Abstract Scope |
This study presents a novel approach for predicting the properties of as-cast materials through the development of a microstructure image generation technique and a machine learning (ML) model. The microstructure generation is achieved using a hybrid Lattice Boltzmann Method–Cellular Automaton (LBM-CA) model, enabling realistic simulations of solidification processes and the resulting microstructural features. By generating diverse and high-resolution microstructure datasets, we constructed a comprehensive training set for ML-based property prediction. The proposed ML model employs CNNs to extract critical features from microstructure images, allowing for the accurate prediction of mechanical properties such as hardness, tensile strength, and ductility. To ensure model generalization, data augmentation techniques and transfer learning strategies were applied. The model's predictive performance was validated using experimental data from cast specimens with varying compositions and process conditions. Our approach offers significant potential for accelerating the design of cast metallic materials, reducing experimental costs, and optimizing process parameters. |
| Proceedings Inclusion? |
Undecided |