Skip to main content
SearchLoginLogin or Signup

Emulating Magnetohydrostatic Models with Physics-Informed Neural Nets

Presentation #131.02 in the session Improving Understanding of the Sun-Earth System Through Advanced Statistical and Machine Learning Techniques.

Published onOct 20, 2022
Emulating Magnetohydrostatic Models with Physics-Informed Neural Nets

Real-time determinations of the true state of the coronal magnetic field requires fast forward modeling of complicated differential equations. Standard direct integration techniques are often too costly, especially when a highly nonlinear model, like magnetohydrostatics (MHS), is employed. However, physics-informed neural nets (PINNs) can be hard to train, and their outputs are less immediately interpretable. Preliminary results are presented towards using PINNS to emulate an MHS model to obtain fast, approximate solutions which could then be refined.

Comments
0
comment

No comments here