Abstract Details
Abstracts
Author: Stefan Schnake
Requested Type: Poster
Submitted: 2023-04-24 08:06:16
Co-authors: E. Endeve, S. Hahn, C. Hauck, C. Kendrick, P. Snyder, M. Stoyanov
Contact Info:
ORNL
PO Box 2008
Oak Ridge, TN 37831
USA
Abstract Text:
Modeling magnetic confinement fusion devices requires accurate solutions to high-dimensional PDEs. Traditional Eulerian/grid-based methods are accurate but suffer from the "curse of dimensionality" which makes them intractable for 4D, 5D, and 6D simulations. To bypass the curse of dimensionality, we are actively developing ASGarD -- an adaptive sparse-grid discontinuous Galerkin solver for high-dimensional PDEs that was started by the Plasma Theory and Modeling group at Oak Ridge National Laboratory. The base framework is a high-order discontinuous Galerkin finite-element method that is attractive for its robustness, accuracy, and flexibility in the modeling of hyperbolic and elliptic equations. By using a multi-wavelet basis, adaptivity is handled though a sparse-grid selection process utilizing the solutions' coefficient data. Recent capabilities added to ASGarD include improved performance in evaluating operators on the GPU, the addition of matrix-free implicit and IMEX time-stepping methods, and the inclusion of non-linear operators such as Vlasov-Poisson and Lenard-Bernstein.
Comments: