| About this Abstract |
| Meeting |
2024 AWS Professional Program
|
| Symposium
|
2024 AWS Professional Program
|
| Presentation Title |
Machine Learning-guided Models for Optimization of Wire Arc DED Manufacturing Parameters |
| Author(s) |
Stephen Price, Kiran Judd, Matthew Gleason, Kyle Tsaknopoulos, Danielle Cote, Rodica Neamtu |
| On-Site Speaker (Planned) |
Stephen Price |
| Abstract Scope |
Traditional manufacturing of large-scale parts requires a costly process of developing, certifying, storing, and maintaining physical casts and molds. Wire Arc Directed Energy Deposition (wire-arc DED) offers a potential improvement to this process. However, as a new technology, wire-arc DED could benefit from an optimized parameter selection process, enabling quicker and more efficient manufacturing. This work presents a data-driven machine learning approach to relate processing parameters to bead shape with high-performing models capable of predicting bead width and height. Additionally, this work highlights techniques to improve performance including a design of experiment specifically tailored to machine learning and novel feature engineering to include chemical and physical properties. Lastly, this work highlights the capabilities of transforming individual bead characteristics such as width or height and creating composite three-dimensional models for improved comparison. |
| Proceedings Inclusion? |
Undecided |