About this Abstract |
Meeting |
2023 TMS Annual Meeting & Exhibition
|
Symposium
|
Additive Manufacturing of Metals: Applications of Solidification Fundamentals
|
Presentation Title |
Bayesian Optimization of an Exponentially Modified Gaussian Heat Source Model for Laser-Based Additive Manufacturing |
Author(s) |
John Coleman, Gerry Knapp, Matt Rolchigo, Benjamin Stump, Alex Plotkowski |
On-Site Speaker (Planned) |
John Coleman |
Abstract Scope |
Melt pool scale models of additive manufacturing (AM) processes can provide insight into process-structure-property relationships for AM parts. These models generally assume a Gaussian distribution of power density, which tend to overpredict peak temperatures in the melt pool, especially in the keyhole regime. This assumption can lead to erroneous predictions of the solidification conditions, and in turn, the grain structure development. An exponentially modified Gaussian power density model provides a more uniform distribution of heat in the melt pool and is shown to better approximate thermal conditions during keyhole formation. A surrogate-based Markov Chain Monte Carlo (MCMC) algorithm is used to calibrate the heat source shape, effective absorption, and power density distribution against experimental melt pool shapes over a wide range of processing conditions in both the conduction and keyhole processing modes. Temperature data from calibrated models is used in a cellular-automata grain structure prediction code and compared to experiments. |
Proceedings Inclusion? |
Planned: |
Keywords |
Additive Manufacturing, Computational Materials Science & Engineering, Solidification |