Abstract Details
Abstracts
Author: Byoungchan Jang
Requested Type: Poster
Submitted: 2023-04-07 10:13:54
Co-authors: A. A. Kaptanoglu, M. Landreman
Contact Info:
University of Maryland
8223 Paint Branch Dr.
College Park, 20740
United States
Abstract Text:
A device-agnostic approach is proposed for addressing the Grad-Shafranov equation in plasma physics using Physics-Informed Neural Networks (PINNs). Magnetohydrodynamic (MHD) equilibrium codes are essential in plasma physics, but extensive computations can be resource-intensive. PINN surrogate solvers offer the potential for computationally efficient solutions in inverse problem-solving, uncertainty quantification, and stellarator optimization. The exploration of parameter space involves varying model dimensions, collocation point quantities, and boundary conditions. Tradeoffs between reconstruction accuracy and computational performance for different parameter configurations are assessed. The findings highlight the efficacy of the device-agnostic PINN method in solving the Grad-Shafranov equation.
Comments: