Login
- Dokumenttyp:
- Forschungsdaten
- Verantwortlich:
- Winchenbach, Rene
- Autorinnen / Autoren:
- Winchenbach, Rene; Thuerey, Nils
- Institutionszugehörigkeit:
- TUM
- Herausgeber:
- TUM
- Titel:
- Symmetric Basis Convolutions for Learning Lagrangian Fluid Mechanics
- Enddatum der Datenerzeugung:
- 01.02.2024
- Fachgebiet:
- DAT Datenverarbeitung, Informatik; PHY Physik
- Quellen der Daten:
- Simulationen / simulations
- Datentyp:
- Datenbanken / data bases
- Anderer Datentyp:
- Compressed datasets consisting of hdf5 files, one per simulation instance.
- Beschreibung:
- This archive contains spatio-temporal data from simulations of the Navier-Stokes equations split into four different test cases
Test Case I: One-dimensional pseudo-compressible SPH simulations in periodic boundary conditions with random initial conditions. Contains training and testing data with different initial conditions.
Test Case II: Two-dimensional weakly-compressible SPH simulations in closed boundary conditions with random initial velocitie fields. Contains training and testing data with different frequency spectra.
Test Case III: Two-dimensional incompresisble SPH simulations with no boundary conditions of randomly located and accelerated fluid blobs in free-space. Contains training and testing data with different initial locations and speeds.
Test Case IV: Three-dimensional test data of regularly sampled particles in periodic boundary conditions for verification of kernel and kernel gradient estimates. - Schlagworte:
- flow prediction; PDEs; numerical simulation; Graph Neural Networks; Continuous Convolutions; autoregressive models
- Technische Hinweise:
- Projekt: https://ge.in.tum.de/publications/
Publication: https://arxiv.org/abs/2309.01745
Source Code: https://github.com/tum-pbs/SFBC - Sprache:
- en
BibTeX