About this Abstract |
| Meeting |
MS&T25: Materials Science & Technology
|
| Symposium
|
Additive Manufacturing Modeling, Simulation, and Machine Learning: Microstructure, Mechanics, and Process
|
| Presentation Title |
Shallow Neural Network Informed Dwell Time Selection for Thermal History Control in Laser Hot Wire Thin-Walled Parts |
| Author(s) |
Logan-Samuel Maurer, Jack Beuth |
| On-Site Speaker (Planned) |
Logan-Samuel Maurer |
| Abstract Scope |
Laser Hot Wire Additive Manufacturing (LHWAM) offers precise thermal control, making it well suited for fabricating thin-walled parts. For thin-walled parts managing inter-layer dwell time is critical to mitigating thermal accumulation and creep. Commonly inter-layer dwell times are either constant or part of a thermal feedback control system. Predicting dwell time quickly before printing reduces machine sensor complexity while maintaining quality and production capacity. To achieve efficient inter-layer dwell time prediction, a hybrid approach of combining a finite element model (FEM) and shallow neural network is proposed. To validate the thermal FEM single bead Ti-6Al-4V walls with varied inter-layer dwell times were monitored using in-situ pyrometers. A shallow neural network is trained on the validated FEM to predict inter-layer dwell times. The trained network accurately predicts dwell times based on average top-layer temperature. This hybrid model approach reduces computational time and enables tailored thermal histories. |