Emulating Coronal Magnetic Fields with Physics-Informed Neural Networks
Abstract
Accurate models of coronal physics can be very expensive to run, and some applications require a high number of evaluations. We present a methodology to produce coronal magnetic fields with a physics-informed neural network. The model is trained on realizations of a plasma-parameterized mesh-free Magnetohydrostatic model; the produced training set will be made available online. This physics-informed emulation approach to nonlinear plasma solutions could be adapted into space weather forecasting implementations or as a subroutine to ease a variety of challenges in Magnetohydrodynamic simulations.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2022
- Bibcode:
- 2022AGUFMNG52A0173M