About this Abstract |
| Meeting |
MS&T25: Materials Science & Technology
|
| Symposium
|
Additive Manufacturing Modeling, Simulation, and Machine Learning: Microstructure, Mechanics, and Process
|
| Presentation Title |
Accelerating Cellular Automata Grain Structure Predictions via Surrogate Thermal Modeling in Laser Powder Bed Fusion |
| Author(s) |
Michael Paleos, Berkay Bostan, Albert C. To |
| On-Site Speaker (Planned) |
Michael Paleos |
| Abstract Scope |
Cellular automata (CA) models have been successful in approximating the true physics of melt pool solidification in laser powder bed fusion (LPBF). However, they face a significant bottleneck in the computational expense of thermal process simulations and typically rely on replicated thermal data from single tracks or layers, failing to capture layer-to-layer variations. We propose a deep learning surrogate modeling framework for the scan-resolved thermal process simulation which attempts to bypass those bottlenecks. Following the time-efficient capture of thermal signature data through the surrogate model, the as-built microstructure can be predicted by leveraging the open-source ExaCA software in a decoupled manner. Ultimately, by circumventing the computational cost of thermal process simulations, this framework facilitates large-scale CA simulations and, thus, LPBF process optimization. |