| About this Abstract |
| Meeting |
MS&T24: Materials Science & Technology
|
| Symposium
|
Additive Manufacturing: Artificial Intelligence and Data Driven Approaches
|
| Presentation Title |
Chemical Composition Based Machine Learning and Multi-Physics Model to Predict Defect Formation in Additive Manufacturing |
| Author(s) |
Ankit Roy, Stephanie Barbara Lawson, Mohan Sai Kiran Kumar Yadav Nartu, Nahal Ghanadi, Somayeh Pasebani, Isabella Van Rooyen |
| On-Site Speaker (Planned) |
Ankit Roy |
| Abstract Scope |
This study presents a novel approach to enhancing additive manufacturing processes for reactor materials through the integration of machine learning and multi-physics simulations. A chemical composition-based machine learning model was developed to predict the printability of alloys in laser powder bed fusion, focusing on defect avoidance. The model achieved high accuracy in predicting balling defect formation (92.3%) and porosity percentage (R2 score of 0.97), with neural networks and random forest regressors demonstrating notable performance. Key insights revealed carbon’s significant roles in defect occurrence, particularly in steel alloys. Multi-physics simulations in Flow-3D, focusing on carbon content variations in SS 316 L and H alloys, provided deeper understanding of printing quality variations. This research offers a swift, chemistry-driven tool for designing experiments and optimizing alloy compositions for additive manufacturing. |