Abstract Details
Abstracts
Author: Tyler Mark
Requested Type: Poster
Submitted: 2026-03-19 09:55:42
Co-authors:
Contact Info:
University of Florida
1257 SW 9th Road APT 307
Gainesville, Florida 32601
United States
Abstract Text:
Descriptions of runaway electrons (REs) range from simple but efficient fluid models to computationally intensive kinetic descriptions. We introduce a novel approach that enables a multifidelity treatment of RE evolution within a single framework. This approach combines an adjoint formulation of the relativistic Fokker-Planck equation for RE evolution and a physics-informed neural network (PINN). The resulting framework yields rapid online inference, at the cost of expensive offline training, allowing for orders of magnitude reduction in prediction time compared with traditional kinetic solvers while retaining a high fidelity description of RE dynamics.
The adjoint deep learning framework can rapidly predict the RE density moment, with full kinetic fidelity, followed by higher order moments such as the RE current, average energy, and pressure anisotropy. Further, the framework enables direct prediction of the energy distribution through a careful treatment of the adjoint solution. While specific quantities necessitate tailored neural network architectures, the use of an adjoint solution allows predictions of RE moments for arbitrary initial conditions, while the deep learning treatment enables a single model to make predictions across a broad range of physical parameters.
Ongoing work includes reconstructing the RE momentum space distribution using a Green's function, enabling an
autoregressive rollout of the RE distribution function. This capability enables an efficient means of capturing the full RE kinetics under arbitrary time-varying plasma parameters where such a surrogate can be directly coupled with disruption codes at marginal cost. The resulting framework enables a trade-off between speed and physics fidelity, allowing the user to adapt the physics fidelity to the targeted application.
Characterization: 5.0
Comments: