| About this Abstract |
| Meeting |
2025 TMS Annual Meeting & Exhibition
|
| Symposium
|
Additive Manufacturing Modeling, Simulation and Machine Learning
|
| Presentation Title |
A Machine Learning-Based Approach for Process Optimization in Laser Based 3-D Printing of High-Performance Al-Alloys |
| Author(s) |
Jun Zheng, Ruobin Qi, John O'Connell, Jack White, Bhaskar Majumdar |
| On-Site Speaker (Planned) |
Ruobin Qi |
| Abstract Scope |
Although there have been major successes in laser-based 3D printing of metals, the process still suffers from defects such as voids or cracks generated during printing. High-performance Al-alloys are particularly prone to defects influenced by process conditions, namely laser speed, power level, hatch spacing, and layer depth. In this work, we have been categorizing defect types and sizes as a function of process conditions for an Al10SiMg alloy, and analyzing the connection between these conditions and defects using a machine learning-based approach. Specifically, for a given process condition, including laser speed and power level, we apply the mixture density network (MDN), which utilizes a neural network (NN) to predict the distribution of defect characteristics modeled as a Gaussian mixture model (GMM). The goal is to determine the optimum process conditions that minimize defects while enhancing strength and damage tolerance. Our approach and results will be presented. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Additive Manufacturing, Machine Learning, Aluminum |