Learned Coarse Models for Efficient Turbulence Simulation

Examples of the Dil-ResNet trained on different domains, capturing a range of qualitatively diverse turbulent dynamics.

(Please see the menu for additional comparisons).

comparison_to_ground_truth.mp4

2D Incompressible Decaying Turbulence (Incomp-2D)

X- and Y- Components of Velocity and Vorticity for Ground Truth and Dil-ResNet model rollouts. (top row) Ground Truth, (middle row) learned model (Dil-ResNet) prediction, (bottom row) spectrum for each component.

example_prediction.mp4

3D Compressible Decaying Turbulence (CompDecay-3D)

Density, X-, Y- and Z- Components of Velocity, and Pressure for Ground Truth and Dil-ResNet model rollouts. (top row) Ground Truth, (middle row) learned model (Dil-ResNet) prediction, (bottom row) spectrum for each component.

example_prediction_L075.mp4

3D Mixing Layer(CoolMixLayer-3D)

Density, X-, Y- and Z- Components of Velocity, and Pressure for Ground Truth and Dil-ResNet model rollouts trained on box size L=0.75. (top row) Ground Truth, (middle row) learned model (Dil-ResNet) prediction, (bottom row) spectrum for each component.