Abstract Details
Abstracts
Author: Brendan C Lyons
Requested Type: Consider for Invited
Submitted: 2025-02-21 19:03:27
Co-authors: O. Meneghini, L. Stagner, A. Ghiozzi, T. Neiser, T. Slendebroek, M.G. Yoo, H. Anand
Contact Info:
General Atomics
PO Box 85608
San Diego, CA 92121
United States
Abstract Text:
The FUsion Synthesis Engine, FUSE (https://fuse.help), is an open-source framework for integrated whole-facility modeling of fusion devices. Originally developed for pilot-plant design, FUSE has been extended to model the controlled time evolution of both plasma and plant. We provide an overview of FUSE capabilities, with a particular focus on the latest work towards pulse-design simulations of DIII-D. QED (a 1D, time-dependent, current-diffusion code) and VacuumFields (a grid-free electromagnetic model for poloidal-field coils) have been coupled using a novel algorithm to simulate the evolution of the plasma current profile and the currents in coils and conducting structures. This allows for self-consistent modeling of inductive current drive by ohmic coils. A new free-boundary solver, FRESCO, solves both the forward and inverse equilibrium problems. The forward problem can be solved by allowing eddy currents in conducting structures to stabilize vertical instabilities. Iterating this eddy-current-stabilized equilibrium with the QED-VacuumFields current evolution allows FUSE to simulate VDEs. The vertical motion and the plasma shape can then be controlled using internal algorithms or coupling to TokSys. TJLF, Julia translation of TGLF allows for rapid database generation to train accurate neural networks to accelerate FUSE’s time-implicit, flux-matching transport solver. In addition, neural networks have been trained to provide accelerated solutions to the forward equilibrium problem. We will present progress on combining these capabilities to create a modern, Grad-Hogan-like solver in FUSE, which promises a novel capability coupling first-principles-based, time-dependent equilibrium evolution with high-fidelity transport models. Together with a new ability to read DIII-D data, FUSE is prepared to provide digital-twin tools for both experimental analysis and predict-first pulse design.
This work was supported by General Atomics corporate funding.
Characterization: 5.0
Comments: