MgFNO: Multi-grid Architecture Fourier Neural Operator for Parametric Partial Differential Equations

Zi-Hao Guo,Hou-Biao Li

2024-09-22

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: <a class="link-external link-https" href="https://github.com/guozihao-hub/MgFNO/tree/master" rel="external noopener nofollow">this https URL</a>.

Numerical Analysis

What problem does this paper attempt to address?

### Problems Addressed by the Paper The paper aims to address the issue of repeatedly solving large-scale high-resolution partial differential equations (PDEs) in science and engineering. Specifically, the paper proposes a Multigrid Fourier Neural Operator (MgFNO) to improve the training process of the traditional Fourier Neural Operator (FNO). #### Main Objectives: 1. **Increase FNO Training Speed**: Accelerate the FNO training process by introducing a triple V-cycle multigrid architecture. 2. **Address Slow Convergence of High-Frequency Components**: Use the multigrid method to convert high-frequency components on fine grids to low-frequency components on coarse grids, thereby speeding up the learning of high-frequency components. 3. **Achieve Resolution Invariance**: Ensure that FNO computations are simplified across different resolutions while maintaining good generalization capabilities. #### Method Overview: - **Multigrid Architecture**: Employ a triple V-cycle multigrid architecture, transferring the model from coarse grids to fine grids after multiple training sessions on the coarse grids to accelerate the training process. - **Frequency Principle**: Combine the frequency principle of deep neural networks to learn solutions from low frequency to high frequency. - **Compact Support Activation Functions**: Use activation functions with compact support to ensure that the multiscale neural network converges quickly and consistently across different scales. #### Experimental Validation: - Experiments were conducted on the Burgers' equation, Darcy flow, and Navier-Stokes equations, showing that MgFNO outperforms traditional Fourier Neural Operators. ### Summary By introducing a multigrid architecture and compact support activation functions, the paper significantly improves the training efficiency and performance of the Fourier Neural Operator. It addresses the slow convergence of high-frequency components in traditional FNOs and demonstrates superior performance on several typical partial differential equations.

MgFNO: Multi-grid Architecture Fourier Neural Operator for Parametric Partial Differential Equations

Gabor-Filtered Fourier Neural Operator for Solving Partial Differential Equations

Multi-Grid Tensorized Fourier Neural Operator for High-Resolution PDEs

Model-Parallel Fourier Neural Operators as Learned Surrogates for Large-Scale Parametric PDEs

Fourier Neural Operator with Learned Deformations for PDEs on General Geometries

MgNO: Efficient Parameterization of Linear Operators via Multigrid

Incremental Spatial and Spectral Learning of Neural Operators for Solving Large-Scale PDEs

Non-equispaced Fourier Neural Solvers for PDEs

Fourier Neural Operator for Solving Subsurface Oil/Water Two-Phase Flow Partial Differential Equation

A MgNO Method for Multiphase Flow in Porous Media

Transfer Learning Fourier Neural Operator for Solving Parametric Frequency-Domain Wave Equations

Enhancing subsurface multiphase flow simulation with Fourier neural operator

Enhancing Solutions for Complex PDEs: Introducing Complementary Convolution and Equivariant Attention in Fourier Neural Operators

Toward a Better Understanding of Fourier Neural Operators from a Spectral Perspective

Fourier-MIONet: Fourier-enhanced multiple-input neural operators for multiphase modeling of geological carbon sequestration

Diffeomorphism Neural Operator for various domains and parameters of partial differential equations

Deep Equilibrium Based Neural Operators for Steady-State PDEs

Pseudo-Differential Neural Operator: Generalized Fourier Neural Operator for Learning Solution Operators of Partial Differential Equations

Resolution invariant deep operator network for PDEs with complex geometries

A Born Fourier Neural Operator for Solving Poisson’s Equation with Limited Data and Arbitrary Domain Deformation

Quantum Fourier Networks for Solving Parametric PDEs