About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Algorithms Development in Materials Science and Engineering
|
| Presentation Title |
The Variational Deep Materials Network: Efficient Extrapolation with Uncertainty of Homogenized Material Responses |
| Author(s) |
Andreas E. Robertson, Sam Inman, Ashley Lenau, Ricardo Lebensohn, Dongil Shin, Brad Boyce, Remi Dingreville |
| On-Site Speaker (Planned) |
Andreas E. Robertson |
| Abstract Scope |
Useful machine learned surrogate models must be designed to account for two common characteristics in materials science: uncertainty and limited data. The Deep Material Network (DMN) is a physics-informed machine learning framework designed to address the second. It can stably extrapolate to predict non-linear homogenized material responses even though it is trained on only cheap elastic data. In this talk, we present our extension: the Variational DMN. The VDMN accounts for uncertainty in its prediction – arising from the aleatoric uncertainty that is often present in a material system. Importantly, this uncertainty prediction also extrapolates, allowing the VDMN to quantify uncertainty in both linear and nonlinear material responses without the need for nonlinear data. We present the algorithmic advances necessary for these changes and then present a series of examples exploring the strengths and limitations of the VDMN as a tool for accelerated uncertainty quantification in materials science. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Machine Learning, Additive Manufacturing, Composites |