SCA: Eurographics/SIGGRAPH Symposium on Computer Animation

Permanent URI for this community

https://diglib.eg.org/handle/10.2312/423

Browse

Browsing SCA: Eurographics/SIGGRAPH Symposium on Computer Animation by Author "Azevedo, Vinicius"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Divergence-Free and Boundary-Respecting Velocity Interpolation Using Stream Functions
    (ACM, 2019) Chang, Jumyung; Azevedo, Vinicius C.; Batty, Christopher; Batty, Christopher and Huang, Jin
    In grid-based fluid simulation, discrete incompressibility of each cell is enforced by the pressure projection. However, pointwise velocities constructed by interpolating the discrete velocity samples from the staggered grid are not truly divergence-free, resulting in unphysical local volume changes that manifests as particle spreading and clustering.We present a new velocity interpolation method that produces analytically divergence-free velocity fields in 2D using a stream function. The resulting fields are guaranteed to be divergence-free by a simple calculus identity: the curl of any vector field yields a divergence-free vector field. Furthermore, our method works on cut cell grids to produce fields that strictly obey solid boundary conditions. Therefore, no artificial gaps are created between fluid particles and solids, and fluid particles do not trespass into solid regions.
  • Thumbnail Image
    Item
    Latent Space Subdivision: Stable and Controllable Time Predictions for Fluid Flow
    (The Eurographics Association and John Wiley & Sons Ltd., 2020) Wiewel, Steffen; Kim, Byungsoo; Azevedo, Vinicius; Solenthaler, Barbara; Thuerey, Nils; Bender, Jan and Popa, Tiberiu
    We propose an end-to-end trained neural network architecture to robustly predict the complex dynamics of fluid flows with high temporal stability. We focus on single-phase smoke simulations in 2D and 3D based on the incompressible Navier-Stokes (NS) equations, which are relevant for a wide range of practical problems. To achieve stable predictions for long-term flow sequences with linear execution times, a convolutional neural network (CNN) is trained for spatial compression in combination with a temporal prediction network that consists of stacked Long Short-Term Memory (LSTM) layers. Our core contribution is a novel latent space subdivision (LSS) to separate the respective input quantities into individual parts of the encoded latent space domain. As a result, this allows to distinctively alter the encoded quantities without interfering with the remaining latent space values and hence maximizes external control. By selectively overwriting parts of the predicted latent space points, our proposed method is capable to robustly predict long-term sequences of complex physics problems, like the flow of fluids. In addition, we highlight the benefits of a recurrent training on the latent space creation, which is performed by the spatial compression network. Furthermore, we thoroughly evaluate and discuss several different components of our method.