Title: Grad–Shafranov equilibria via data-free physics informed neural networks

Journal Article · · Physics of Plasmas
DOI: https://doi.org/10.1063/5.0188634 · OSTI ID:2567811

A large number of magnetohydrodynamic (MHD) equilibrium calculations are often required for uncertainty quantification, optimization, and real-time diagnostic information, making MHD equilibrium codes vital to the field of plasma physics. In this paper, we explore a method for solving the Grad–Shafranov equation by using physics-informed neural networks (PINNs). For PINNs, we optimize neural networks by directly minimizing the residual of the partial differential equation as a loss function. We show that PINNs can accurately and effectively solve the Grad–Shafranov equation with several different boundary conditions, making it more flexible than traditional solvers. This method is flexible as it does not require any mesh and basis choice, thereby streamlining the computational process. We also explore the parameter space by varying the size of the model, the learning rate, and boundary conditions to map various tradeoffs such as between reconstruction error and computational speed. Additionally, we introduce a parameterized PINN framework, expanding the input space to include variables such as pressure, aspect ratio, elongation, and triangularity in order to handle a broader range of plasma scenarios within a single network. Parameterized PINNs could be used in future work to solve inverse problems such as shape optimization.

Research Organization:
University of Maryland, College Park, MD (United States)
Sponsoring Organization:
USDOE; USDOE Office of Science (SC), Basic Energy Sciences (BES). Scientific User Facilities (SUF)
Grant/Contract Number:
AC02-05CH11231; FG02-93ER54197
OSTI ID:
2567811
Journal Information:
Physics of Plasmas, Journal Name: Physics of Plasmas Journal Issue: 3 Vol. 31; ISSN 1070-664X
Publisher:
American Institute of PhysicsCopyright Statement
Country of Publication:
United States
Language:
English

References (28)

Three ways to solve partial differential equations with neural networks — A review journal May 2021
On the limited memory BFGS method for large scale optimization journal August 1989
Hydromagnetic equilibria and force-free fields journal September 1958
Multilayer feedforward networks are universal approximators journal January 1989
Surrogate modeling for fluid flows based on physics-constrained deep learning without simulation data journal April 2020
A comprehensive study of non-adaptive and residual-based adaptive sampling for physics-informed neural networks journal January 2023
Deep neural operators as accurate surrogates for shape optimization journal March 2024
Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations journal February 2019
DAS-PINNs: A deep adaptive sampling method for solving high-dimensional partial differential equations journal March 2023
Physics-informed machine learning journal May 2021
“One size fits all” analytic solutions to the Grad–Shafranov equation journal March 2010
Noncircular, finite aspect ratio, local equilibrium model journal April 1998
DESC: A stellarator equilibrium solver journal October 2020
Real time equilibrium reconstruction for tokamak discharge control journal July 1998
Hybrid neural network for density limit disruption prediction and avoidance on J-TEXT tokamak journal March 2018
Deep neural network Grad–Shafranov solver constrained with measured magnetic signals journal December 2019
Proof of concept of a fast surrogate model of the VMEC code via neural networks in Wendelstein 7-X scenarios journal August 2021
Neural net modeling of equilibria in NSTX-U journal July 2022
Surrogate models for plasma displacement and current in 3D perturbed magnetohydrodynamic equilibria in tokamaks journal November 2022
Physics-regularized neural network of the ideal-MHD solution operator in Wendelstein 7-X configurations journal April 2023
On the potential of physics-informed neural networks to solve inverse problems in tokamaks journal November 2023
Neural Network Differential Equation and Plasma Equilibrium Solver journal November 1995
Physics-Informed Neural Networks for Heat Transfer Problems journal April 2021
DeepXDE: A Deep Learning Library for Solving Differential Equations journal January 2021
Understanding and Mitigating Gradient Flow Pathologies in Physics-Informed Neural Networks journal January 2021
Physics-Informed Neural Networks with Hard Constraints for Inverse Design journal January 2021
Algorithm 778: L-BFGS-B: Fortran subroutines for large-scale bound-constrained optimization journal December 1997
The Old and the New: Can Physics-Informed Deep-Learning Replace Traditional Linear Solvers? journal November 2021