Speaker
Yen Chen Chen
(National Center for High-Performance Computing, National Applied Research Laboratories, Taiwan)
Description
We apply the Physics-informed Neural Networks (PINNs) to the magnetohydrodynamic (MHD) simulations. We build a neural network to find a solution of the MHD equations. We use the coordinate of space-time as model inputs and the outputs of the model are the magnetic field, the plasma bulk velocity, the plasma mass density, and the plasma thermal pressure. The MHD equations are combined into the loss function for a physics constrain. We show our training results and compare the computation time of the PINNs method and MHD numerical method.
Author
Yen Chen Chen
(National Center for High-Performance Computing, National Applied Research Laboratories, Taiwan)
Co-authors
Dr
Chen W. S.
(National Center for High-Performance Computing, National Applied Research Laboratories, Taiwan)
Dr
Lee K. H.
(Institute of Earth Science, Academia Sinica, Taiwan)
Dr
Tsai T. C.
(National Center for High-Performance Computing, National Applied Research Laboratories, Taiwan)
