Neural-Network solver of ideal MHD equilibria
Year: 2025
Type: article
Abstract: We present a novel approach to compute three-dimensional Magnetohydrodynamic equilibria byparametrizing Fourier modes with artificial neuralnetworks and compare it to equilibria computedby conventional solvers. The full nonlinear globalforce residual across the volume in real space is thenminimized with first order optimizers. Already, weobserve competitive computational cost to arriveat the same minimum residuals computed by exist-ing codes. With increased computational cost, lowerminima of the residual are achieved by the neuralnetworks, establishing a new lower bound for theforce residual. We use minimally complex neuralnetworks, and we expect significant improvementsfor solving not only single equilibria with neuralnetworks, but also for computing neural networkmodels valid over continuous distributions of equi-libria.
Source: Nuclear Fusion
Author: Thun, Timo
Institution: Max Planck Institute for Plasma Physics
Language: English
Subfield: Nuclear and High Energy Physics
Field: Physics and Astronomy
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,595
OpenAlex ID: W7083288081
FWCI:
Cites:
Cited by:
Related to: