A neural network multigrid solver for the Navier-Stokes equations

TL;DR

This paper introduces DNN-MG, a neural network-enhanced multigrid solver that accelerates Navier-Stokes simulations by combining classical multigrid methods with neural network corrections, achieving faster computation and good generalization.

Contribution

The paper presents a novel neural network multigrid solver that improves efficiency and generalization for Navier-Stokes equations by integrating neural corrections into classical multigrid methods.

Findings

DNN-MG reduces computation time by about 50% compared to full multigrid solutions.

The neural network corrects interpolated solutions effectively on fine levels.

DNN-MG generalizes across different mesh domains and flow configurations.

Abstract

We present the deep neural network multigrid solver (DNN-MG) that we develop for the instationary Navier-Stokes equations. DNN-MG improves computational efficiency using a judicious combination of a geometric multigrid solver and a recurrent neural network with memory. DNN-MG uses the multi-grid method to classically solve on coarse levels while the neural network corrects interpolated solutions on fine ones, thus avoiding the increasingly expensive computations that would have to be performed there. This results in a reduction in computation time through DNN-MG's highly compact neural network. The compactness results from its design for local patches and the available coarse multigrid solutions that provides a "guide" for the corrections. A compact neural network with a small number of parameters also reduces training time and data. Furthermore, the network's locality facilitates generalizability and allows one to use DNN-MG trained on one mesh domain also on different ones. We demonstrate the efficacy of DNN-MG for variations of the 2D laminar flow around an obstacle. For these, our method significantly improves the solutions as well as lift and drag functionals while requiring only about half the computation time of a full multigrid solution. We also show that DNN-MG trained for the configuration with one obstacle can be generalized to other time dependent problems that can be solved efficiently using a geometric multigrid method.