Uncovering turbulent plasma dynamics via deep learning from partial observations

Abstract

One of the most intensely studied aspects of magnetic confinement fusion is edge plasma turbulence which is critical to reactor performance and operation. Drift-reduced Braginskii two-fluid theory has for decades been widely applied to model boundary plasmas with varying success. Towards better understanding edge turbulence in both theory and experiment, we demonstrate that physics-informed neural networks constrained by partial differential equations can accurately learn turbulent fields consistent with the two-fluid theory from just partial observations of a synthetic plasma's electron density and temperature in contrast with conventional equilibrium models. These techniques present a novel paradigm for the advanced design of plasma diagnostics and validation of magnetized plasma turbulence theories in challenging thermonuclear environments.

11 pages, 8 figures

References in corpus (4)

• On the eigenvector bias of Fourier feature networks: From regression to solving multi-scale PDEs with physics-informed neural networks
• Global turbulence simulations of the tokamak edge region with GRILLIX
• Constructing a new predictive scaling formula for ITER's divertor heat-load width informed by a simulation-anchored machine learning
• Turbulent field fluctuations in gyrokinetic and fluid plasmas

Cited by in corpus (15)

• A unified scalable framework for causal sweeping strategies for Physics-Informed Neural Networks (PINNs) and their temporal decompositions
• 2022 Review of Data-Driven Plasma Science
• Integrated Finite Element Neural Network (I-FENN) for non-local continuum damage mechanics
• Unveiling the optimization process of Physics Informed Neural Networks: How accurate and competitive can PINNs be?
• Error convergence and engineering-guided hyperparameter search of PINNs: towards optimized I-FENN performance
• A machine-learning-based tool for last closed-flux surface reconstruction on tokamaks
• Data-driven model for divertor plasma detachment prediction
• Data-driven modeling of Landau damping by physics-informed neural networks
• EAST discharge prediction without integrating simulation results
• Deep modelling of plasma and neutral fluctuations from gas puff turbulence imaging
• Turbulent field fluctuations in gyrokinetic and fluid plasmas
• Deep electric field predictions by drift-reduced Braginskii theory with plasma-neutral interactions based upon experimental images of boundary turbulence
• Leveraging turbulence data from fusion experiments
• Physics-informed neural networks viewpoint for solving the Dyson-Schwinger equations of quantum electrodynamics
• Quantifying experimental edge plasma evolution via multidimensional adaptive Gaussian process regression