Magnetohydrodynamics with Physics Informed Neural Operators

The modeling of multi-scale and multi-physics complex systems typically involves the use scientific software that can optimally leverage extreme scale computing. Despite major developments in recent years, these simulations continue to be computationally intensive time consuming. Here we explore AI accelerate at a fraction computational cost classical methods, present first application physics informed neural operators model 2D incompressible magnetohydrodynamics simulations. Our models incorporate tensor Fourier as their backbone, which implemented with TensorLY package. results indicate accurately capture describe laminar flows Reynolds numbers $Re\leq250$. We also applicability our surrogates for turbulent flows, discuss variety methodologies may incorporated future work create provide efficient high fidelity description broad range numbers. developed this project is released manuscript.