About this Abstract |
| Meeting |
2027 TMS Annual Meeting & Exhibition
|
| Symposium
|
AI-Enabled Materials Processing: Integrating Accelerated Experimental Workflows and Processing-Aware Machine Learning
|
| Presentation Title |
Online Wire-arc Additive Manufacturing Process Optimization Using Active Learning |
| Author(s) |
Satyajit Mojumder, Pallock Halder, Tiana Tonge, Ballweber Wyatt |
| On-Site Speaker (Planned) |
Satyajit Mojumder |
| Abstract Scope |
Wire-Arc Additive Manufacturing (WAAM) enables large-scale metal component printing, but producing qualified deposits requires material- and location-specific process parameters. Reliable process-window development therefore requires substantial offline trial-and-error experimentation. We present an online process optimization framework that collects and analyzes process monitoring data in real-time to optimize the process parameters in the WAAM system. A microphone is used for acoustic process monitoring, and the data is used to determine the deposited bead quality through a trained convolutional neural network. These scores provide an in-process qualification metric without post-analysis and are used to train a Gaussian Process surrogate, while Expected Improvement selects the next experiment in an active learning environment. The process parameter search is conducted in a non-dimensional space for cross-material generalization. The active learning-guided online process optimization framework reduces costly trial-and-error parameter tuning and lowers labeling effort by introducing a new quality score. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Additive Manufacturing, Machine Learning, ICME |