| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Additive Manufacturing Modeling, Simulation and Machine Learning
|
| Presentation Title |
Machine Learning-Driven Predictions of Material Printability in Laser Powder Bed Fusion |
| Author(s) |
Sofia Sheikh, Brent Vela, Raymundo Arroyave |
| On-Site Speaker (Planned) |
Sofia Sheikh |
| Abstract Scope |
Additive manufacturing (AM), specifically Laser Powder Bed Fusion (L-PBF), faces challenges in optimizing process parameters and material properties for printability. This study evaluates various machine learning (ML) models to predict printability using a defined metric. CatBoost Regressor and Random Forest Regressor showed high predictive accuracy with minimal errors in metrics such as MAE, MSE, RMSE, and R2.
Feature importance analysis using SHAP highlighted key material properties influencing printability, like kinetic viscosity and electronegativity. While both models performed well, Random Forest demonstrated superior computational efficiency. This research emphasizes the importance of computational efficiency, interpretability, and robustness when selecting ML models for L-PBF material printability prediction. Integrating ICME methodologies and ML models can significantly enhance process parameter optimization and material properties in L-PBF, driving innovation in AM. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Additive Manufacturing, ICME, Machine Learning |