About this Abstract |
| Meeting |
MS&T25: Materials Science & Technology
|
| Symposium
|
Additive Manufacturing Modeling, Simulation, and Machine Learning: Microstructure, Mechanics, and Process
|
| Presentation Title |
The Melt Pool Spatter Problem in Additive Manufacturing |
| Author(s) |
Jack L Beuth, Nicholas O'Brien, Christian Gobert, Satbir Singh, David Deisenroth, Jordan Weaver, Amir Barati Farimani |
| On-Site Speaker (Planned) |
Jack L Beuth |
| Abstract Scope |
In this talk, two applications of machine learning are applied to the spatter problem. The first application involves in-process monitoring of spatter generation by machine learning analyses of high-speed videos at the melt pool scale. This application is allowing determination of spatter counts and trajectories as spatter is emitted from melt pools as a function of process parameters. Two sets of experiments are analyzed, one early set from work at Carnegie Mellon and one more recent set carried out at NIST. The second application involves development of a machine learning surrogate model trained by results from detailed CFD simulations of argon flow and spatter transport in the build chamber. Where single CFD simulations require 1-5 hours to execute, results from the surrogate model can be obtained in a few seconds. This has allowed the development of a real-time build simulation tool of spatter transport in the build chamber. |