Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
AKPIK: Arbeitskreis Physik, moderne Informationstechnologie und Künstliche Intelligenz
AKPIK 1: Reservoir Computing & Neural Networks
AKPIK 1.6: Vortrag
Dienstag, 19. März 2024, 10:45–11:00, MAR 0.002
Physics-Informed Deep Learning to Couple Reactive Diffusion and Swelling in Cellulose-based Porous Media — •Alexandra Serebrennikova1, Maximilian Fuchs1, Raimund Teubler2, and Karin Zojer1 — 1Institute of Solid State Physics, TU Graz, Graz, Austria — 2Institute of Analytical Chemistry, TU Graz, Graz, Austria
Simulating the reactive diffusion of fluids through porous media presents significant challenges due to the intricate geometries of real-world systems, particularly when the porous media itself undergoes changes, such as swelling of the solid matrix. Traditional numerical solvers often struggle to represent these complex details accurately and feasibly. Based on state-of-the-art extended physics-informed neural networks (PINNs), this contribution focuses on creating a mesh-free modeling framework for studying the reactive transport of volatile organic compounds (VOCs) through the complex microstructure of paper. PINNs serve us to implicitly incorporate the experimentally observed evolution of geometrical features of paper matrix into the formulation of the governing partial differential equations. This approach enables to study the spatio-temporal evolution of VOC concentrations in the porous environment of paper, while the geometry of the material dynamically adapts itself through swelling or shrinking as response to the current state of adsorption.
To our knowledge, this is the first contribution that applies PINNs to consider adaptive geometries during transport.
Keywords: pinn; paper; reactive trasnport; porous material; swelling