A5051013095- Model Reduction and Neural Networks
- Meteorological Phenomena and Simulations
- Adversarial Robustness in Machine Learning
- Magnetic confinement fusion research
- Fusion materials and technologies
- Nuclear Engineering Thermal-Hydraulics
- Fluid Dynamics and Turbulent Flows
- Computational Physics and Python Applications
- Seismology and Earthquake Studies
- Generative Adversarial Networks and Image Synthesis
Max Planck Institute for Plasma Physics
2023
Abstract Simulating plasma turbulence presents significant computational challenges due to the complex interplay of multi-scale dynamics. In this work, we investigate use convolutional neural networks improve efficiency simulations, focusing on Hasegawa-Wakatani model. The are trained learn closure terms in large eddy providing a computationally cheaper alternative high-resolution numerical solvers for capturing effects high-frequency components. This study is first successfully apply...
Greif, R., (2023). HW2D: A reference implementation of the Hasegawa-Wakatani model for plasma turbulence in fusion reactors. Journal Open Source Software, 8(92), 5959, https://doi.org/10.21105/joss.05959
We investigate uncertainty estimation and multimodality via the non-deterministic predictions of Bayesian neural networks (BNNs) in fluid simulations. To this end, we deploy BNNs three challenging experimental test-cases increasing complexity: show that BNNs, when used as surrogate models for steady-state flow predictions, provide accurate physical together with sensible estimates uncertainty. Further, experiment perturbed temporal sequences from Navier-Stokes simulations evaluate...
Turbulence in fluids, gases, and plasmas remains an open problem of both practical fundamental importance. Its irreducible complexity usually cannot be tackled computationally a brute-force style. Here, we combine Large Eddy Simulation (LES) techniques with Machine Learning (ML) to retain only the largest dynamics explicitly, while small-scale are described by ML-based sub-grid-scale model. Applying this novel approach self-driven plasma turbulence allows us remove large parts inertial...