About this Abstract |
| Meeting |
MS&T26: Materials Science & Technology
|
| Symposium
|
Additive Manufacturing Modeling, Simulation, and Machine Learning: Microstructure, Mechanics, and Process
|
| Presentation Title |
A Material-Agnostic Framework for Rapid Generation of Updatable Process Maps for Powder-Blown Laser Directed Energy Deposition (L-DED) |
| Author(s) |
James Hanagan, Raymundo Arróyave |
| On-Site Speaker (Planned) |
James Hanagan |
| Abstract Scope |
Parameter development for metal additive manufacturing (AM) processes is a time-consuming process requiring numerous experimental trials at various processing parameters and/or costly simulations to build a reliable process parameter map. In laser powder bed fusion (LPBF), this has led to the development of methods for rapid generation of defect processing maps capable of being updated with experimental data. Powder-blown L-DED, on the other hand, has lagged behind on this front. In this talk, we present a method for rapidly generating L-DED defect processing maps as a function of laser power, scan speed, and mass flow rate using simple analytical expressions for melt pool dimensions and dilution. With the help of uncertainty quantification from Gaussian processes trained on melt pool dimensions, these maps are capable of being used as a starting point for experimental active learning studies where the map can be updated with data from strategically selected single track studies. |