MgFNO: Multi-grid Architecture Fourier Neural Operator for Parametric Partial Differential Equations (2407.08615v2)

Abstract: In science and engineering, there is often a need to repeatedly solve large-scale and high-resolution partial differential equations (PDEs). Neural operators are a new model that can map between function spaces, allowing trained models to emulate PDE solution operators. This paper introduces a novel Fourier neural operator with a multigrid architecture (MgFNO). The MgFNO combines the frequency principle of deep neural networks (DNNs) with the multigrid idea of solving linear systems. To speed up the FNO training process, a three-layer V-cycle multigrid architecture is used. This architecture involves training the model multiple times on a coarse grid and then transferring it to a fine grid to accelerate the training of the model. The solver based on DNN learns the solution from low to high frequency, while the multigrid method acquires the solution from high to low frequency. Note that the FNO is a resolution-invariant solution operator, so the corresponding calculations are greatly simplified. Finally, experiments are carried out on Burgers' equation, Darcy flow, and Navier-Stokes equation. The results show that the proposed MgFNO outperforms the traditional Fourier neural operator. Code Availability: https://github.com/guozihao-hub/MgFNO/tree/master.