1: \begin{abstract}

2:   We investigate scaling and efficiency of the deep neural network multigrid method (DNN-MG).

3:   DNN-MG is a novel neural network-based technique for the simulation of the Navier-Stokes equations that combines an adaptive geometric multigrid solver, i.e. a highly efficient

4:   classical solution scheme, with a recurrent neural network with memory.

5:   The neural network replaces in DNN-MG one or multiple finest multigrid layers and provides a correction for the classical solve in the next time step.

6:   This leads to little degradation in the solution quality while substantially reducing the overall computational costs.

7:   At the same time, the use of the multigrid solver at the coarse scales allows for a compact network that is easy to train, generalizes well, and allows for the incorporation of physical constraints.

8:   Previous work on DNN-MG focused on the overall scheme and how to enforce divergence freedom in the solution.

9:   In this work, we investigate how the network size affects training and solution quality and the overall runtime of the computations.

10:   Our results demonstrate that larger networks are able to capture the

11:   flow behavior better while requiring only little additional training time.

12:   At runtime, the use of the neural network correction can even reduce the computation time compared to a classical multigrid simulation through a faster convergence of the nonlinear solve that is required at every time step.

13: \end{abstract}

14: