About this Abstract |
Meeting |
2022 TMS Annual Meeting & Exhibition
|
Symposium
|
Additive Manufacturing and Innovative Powder Processing of Functional and Magnetic Materials
|
Presentation Title |
Deep Learning with Generative Adversarial Network for Ti-6Al-4V Surface Roughness Improvement in Direct Energy Deposition Process |
Author(s) |
Im Doo Jung, Tae Kyeong Kim, Hyo Kyung Sung, Jung Gi Kim, Hyung Sub Kim |
On-Site Speaker (Planned) |
Im Doo Jung |
Abstract Scope |
The Direct Energy Deposition (DED) process is one of the additive manufacturing (AM) that sprays metal powder directly into a high-power laser and is useful for manufacturing large metal parts. However, there is a bottleneck that surface roughness is relatively low compared to other AM methods. In this work, we controlled three process variables that affect surface illumination during the DED process with titanium powder material and conducted convolutional neural network-based machine learning based on 2D scanned images of the sculptural surface manufactured accordingly. We predicted the process conditions of images with accuracy of more than 80% of MAPE based on test sets. In addition, we visually observed surface roughness with images generated by DC-GAN. It is expected that artificial intelligence will be actively used to improve surface roughness during the DED process for titanium alloy powder materials. |
Proceedings Inclusion? |
Planned: |
Keywords |
Additive Manufacturing, Machine Learning, Surface Modification and Coatings |