SMuK 2023 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

P: Fachverband Plasmaphysik

P 19: Magnetic Confinement V/HEPP VIII

P 19.2: Vortrag

Donnerstag, 23. März 2023, 18:00–18:25, CHE/0089

Physics-informed machine learning to approximate the ideal-MHD solution operator in Wendelstein 7-X configurations — •Andrea Merlo, Daniel Böckenhoff, Jonathan Schilling, Samuel Aaron Lazerson, Thomas Sunn Pedersen, and the W7-X Team — Max-Planck-Institute for Plasma Physics, 17491 Greifswald, Germany

The stellarator is a promising concept to produce energy from nuclear fusion by magnetically confining a high-pressure plasma. Magnetohydrodynamics (MHD) describes how plasma pressure, current density and magnetic field interact. In a stellarator, the confining field is three-dimensional, and the computational cost of solving the 3D MHD equations currently limits stellarator research and design. In this work, we present data-driven approaches to provide fast 3D MHD equilibria: we describe an artificial neural network (NN) that quickly approximates the ideal-MHD solution operator in W7X configurations. The model fulfils equilibrium symmetries by construction and the MHD force residual regularizes the solution of the NN to satisfy the ideal-MHD equations. The model predicts the equilibrium solution with high accuracy, and it faithfully reconstructs global equilibrium properties (e.g., magnetic well depth). We also optimize W7X magnetic configurations, where desiderable configurations can be found in terms of fast particle confinement. Moreover, preliminary results from solving the ideal-MHD equations for a generic stellarator geometry with a physics-informed model without any ground-truth data will be presented.