h1

TY  - CHART
AU  - Nista, Ludovico
AU  - Schumann, Christoph David Karl
AU  - Vivenzo, Marco
AU  - Fröde, Fabian
AU  - Grenga, Temistocle
AU  - MacArt, Jonathan F.
AU  - Attili, Antonio
AU  - Pitsch, Heinz
TI  - Homogeneous isotropic turbulence database for training super-resolution data-driven turbulence closure models
M1  - RWTH-2024-03259
PY  - 2024
AB  - The aim of publishing this data is to facilitate the advancement of super-resolution (SR) methods in turbulence closure modeling, with a specific emphasis on data-driven approaches. The database includes various homogeneous isotropic turbulence (HIT) configurations, specifically forced and decaying HIT, at different Reynolds numbers obtained using direct numerical simulations (DNSs). These configurations are selected to represent a broad spectrum of turbulence behavior, enabling a detailed investigation into the effectiveness of SR reconstruction methods under diverse flow conditions. For each configuration, both low-resolution (LR) and high-resolution (HR) data are ingested into the database. The corresponding LR data are generated by explicitly filtering the DNS high-resolution (HR) fields with three different filter kernels: box, Gaussian, and spectral. This variety allows exploration of the architectures' potential across various filtering methods and flow conditions. Additionally, for each filter kernel, different filter widths are provided, offering extra flexibility for researchers aiming to adapt their SR models to the specifics of their turbulence simulations. This database serves as a valuable resource for users involved in developing SR methods tailored to turbulent closure modeling. Researchers in turbulence modeling and deep learning can utilize this database to train and test their architectures, contributing to the advancements of SR techniques. Furthermore, this database can be also used as a reference for comparing various SR methods proposed. Finally, as the routines for generating and post-processing the data are provided, users can flexibly compute quantities of interest for their applications by simply modifying the user-defined functions.
LB  - PUB:(DE-HGF)32
DO  - DOI:10.18154/RWTH-2024-03259
UR  - https://publications.rwth-aachen.de/record/981830
ER  -