A5082835285- Magnetic confinement fusion research
- Fusion materials and technologies
- Particle accelerators and beam dynamics
- Superconducting Materials and Applications
- Lattice Boltzmann Simulation Studies
- Fluid Dynamics and Turbulent Flows
- earthquake and tectonic studies
- Laser-Plasma Interactions and Diagnostics
- 3D Shape Modeling and Analysis
- Computer Graphics and Visualization Techniques
- Model Reduction and Neural Networks
- Neural Networks and Applications
Beihang University
2023
Institute of Plasma Physics
2021-2022
Chinese Academy of Sciences
2021-2022
Hefei Institutes of Physical Science
2021
The linearized Bhatnagar–Gross–Krook equation is widely used to describe low-speed rarefied gas flows and can be solved numerically using deterministic methods such as the discrete velocity method (DVM). With help of physics-informed neural networks (PINNs), which have experienced rapid development in recent years, we develop a new surrogate model that combines PINNs DVM (PINN–DVM) simulate flows. In proposed PINN–DVM model, directly encoded into residual an artificial network. A loss...
Abstract Quasi-periodic pedestal burst instabilities (PBIs), featuring alternative turbulence suppression and bursts, have been clearly identified by various edge diagnostics during I-mode to H-mode transition in the EAST Tokamak. The radial distribution of phase perturbation caused PBI shows that is localized pedestal. Prior each PBI, a significant increase density gradient close top can be distinguished, then generated, accompanied relaxation profile, induces an outward particle flux....
Abstract Simultaneous control of the damaging erosion induced by transient and steady-state heat/particle fluxes on divertor target material is one critical issues for next-step magnetic fusion devices. H-mode operation without large edge-localized modes has been achieved in EAST with an ITER-like tungsten divertor, while being compatible partial pronounced detachment via either ramping-up bulk density or injection low/high- Z impurities. The pedestal characteristics during transition from...
We propose a non-intrusive method to build surrogate models that approximate the solution of parameterized partial differential equations (PDEs), capable taking into account dependence on shape computational domain. Our approach is based combination two neural networks (NNs). The first NN, conditioned latent code, provides an implicit representation geometry variability through signed distance functions. This automated encoding technique generates compact, low-dimensional representations...
Abstract By scanning toroidal rotation with a combination of co- and counter-current direction neutral beam injection (NBI) on the Experimental Advanced Superconducting Tokamak, it is found size edge localized mode (ELM) decreases increasing in direction. The synergistic effect plasma collisionality ELM behavior also studied by statistical analysis. Three-field module BOUT++ framework employed to study impacts rotation/ E × B flow shear behaviors. simulation results show that both Co-...