| About this Abstract |
| Meeting |
2024 TMS Annual Meeting & Exhibition
|
| Symposium
|
Additive Manufacturing Modeling, Simulation and Machine Learning
|
| Presentation Title |
Fast and Scalable Method to Generate Reduced Order Models of Metal-based Additive Manufacturing Simulations Using a Hypercomplex-based Automatic Differentiation Finite Element Method |
| Author(s) |
Mauricio Aristizabal Cano, Juan-Sebastian Rincon-Tabares, Matthew Balcer, Arturo Montoya, David Restrepo, Harry Millwater |
| On-Site Speaker (Planned) |
Mauricio Aristizabal Cano |
| Abstract Scope |
Simulation of metal Laser Powder Bed Fusion Additive Manufacturing (LPBF-AM) process is challenging due to the number of non-linear physics involved. Thus, reduced order models (ROMs) are necessity for uncertainty quantification, optimization, etc. However, traditional ROM generation requires a large number of full model evaluations. In this talk, we show a novel method to reduce the cost of ROM generation by efficiently computing high-order derivatives of the LPBF-AM model using a HYPercomplex-based Automatic Differentiation Finite Element Method (HYPAD-FEM). The methodology consists of three steps: First, the finite element model is evaluated at a single value of input parameters. Second, the exact tangent matrix is computed. Finally, the residual of the problem is evaluated using hypercomplex inputs to compute the derivatives of the solution with respect to the parameters of interest. Applications to transient thermal and transient thermomechanical 3D simulations of the LPBF-AM process are presented. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Additive Manufacturing, Modeling and Simulation, Computational Materials Science & Engineering |