Skip to content

Commit 39eea46

Browse files
jenzenhojenzenho
committed
Renamed the dataset URLs to be more descriptive.
1 parent 769ded7 commit 39eea46

18 files changed

+1792
-0
lines changed

‎_datasets/blastnet_momentum.md‎

Lines changed: 34 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,34 @@
1+
---
2+
layout: datapage
3+
title: BLASTNet Momentum128 3D SR Dataset
4+
excerpt:
5+
header:
6+
teaser: /assets/img/ico_chung2022.png
7+
description: Collection of BLASTNet Simulations
8+
categories:
9+
- numerical
10+
- nonreacting
11+
- turbulent
12+
- hit
13+
- pipe
14+
- channel
15+
- jet
16+
- benchmark
17+
---
18+
19+
![image](./assets/img/diversity.png)
20+
21+
## Description
22+
The BLASTNet Momentum128 3D SR Dataset is a benchmark dataset for developing and evaluating 3D super-resolution (SR) methods on turbulent flow data. It is a curated subset of the larger BLASTNet datasets, specifically designed to facilitate high-fidelity reconstruction of velocity fields from low-fidelity fields. This dataset includes 2,000 volumetric samples of 128$$^3$$ grid points, each containing the three components of the velocity field (u, v, w) and the density. The sub-volumes were extracted from high-resolution direct numerical simulations (DNS) which span a range of flow regimes and statistical variations.
23+
24+
Each sample is accompanied by pre-computed low-resolution inputs at multiple downsampling ratios (e.g., 8×, 16×, and 32×), enabling the evaluation of SR models under different reconstruction challenges. Data is provided in binary .dat format using single-precision floating point (little-endian) ordering. The dataset is split into training, validation, and test sets, with metadata stored in accompanying CSV. These include physical statistics summary (e.g., skewness, kurtosis, variance).
25+
26+
27+
## Quick Info
28+
* <a href="https://www.kaggle.com/datasets/waitongchung/blastnet-momentum-3d-sr-dataset">Kaggle Link</a><BR>
29+
* Contributors: Wai Tong Chung, Bassem Akoush, Matthias Ihme
30+
* N<sub>x</sub> = 128, N<sub>y</sub> = 128, N<sub>z</sub> = 128, N<sub>&#632;</sub> = 4
31+
* Size = 75.15 GB
32+
* <a href="https://proceedings.neurips.cc/paper_files/paper/2023/file/f458af2455b1e12608c2a16c308d663d-Paper-Datasets_and_Benchmarks.pdf">DOI</a><BR>
33+
* <a href="./assets/bib/chung2022.bib">.bib</a><BR>
34+
<!-- * <a href="./assets/json/chung2022_info.json">info.json</a> -->

‎_datasets/ch4air_forced_hit_flame.md‎

Lines changed: 33 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,33 @@
1+
---
2+
layout: datapage
3+
title: Reacting Forced HIT
4+
excerpt: (1 case)
5+
description: CH4-Air Flame Interaction With Forced HIT DNS
6+
header:
7+
teaser: /assets/img/ico_poludnenko.png
8+
categories:
9+
- reacting
10+
- hit
11+
- turbulent
12+
- numerical
13+
---
14+
15+
![image](./assets/img/poludnenko.png)
16+
17+
## Description
18+
19+
TThe DNS study involves a statistically steady, isotropic, and homogeneous turbulent flow in an unconfined space. The flame is initialized by a planar surface separating half of the domain containing methane/air mixture at 700 K and 3.04 × 107 erg cm−3 pressure, and another half with hot products, and is immersed in a high-intensity turbulent flow field with Kolmogorov type spectrum. The idea is to investigate the process of flame interaction with steady homogeneous isotropic turbulence. However, the flow needs to be constantly stirred at the largest scale to ensure a steady energy cascade to smaller scales so that the turbulence-flame interaction at the quasi-steady state can be studied. A spectral turbulence- driving method is used in the study, the details of which are available in Poludnenko and Oran. This driving method produces statistically steady forced-HIT flows with arbitrarily complex energy spectra. In particular, it is possible to achieve Kolmogorov type turbulence with inertial range of energy cascade extending up to energy injection scale. The other advantage of this method is that it does not introduce any artificial large-scale anisotropy, compression, or rarefaction. Prior to ignition, all domain boundaries are periodic. At ignition, boundary conditions along the left and right z-boundaries (as shown in Figure 9) are switched to zero-order extrapolation to prevent any non-physical pressure build-up in the domain and the formation of artificial large-scale rarefaction waves at the boundaries.
20+
21+
The computational domain aspect ratio is 1 × 1 × 16, with a grid size of 257 × 257 × 4097, including 16 grid points per unit laminar flame thermal thickness. The cell size is 2.62 × 10−4 cm. The turbulent velocity at energy injection scale (L = 0.067 cm) length scale is 213.92 cms−1 with turbulent root-mean-squared (RMS) velocity of 245.83 cms−1, resulting in an eddy turnover time of 3.14×10−4 s. The same velocity quantities corresponding to the integral length scale (l = 0.0196 cm) are 141.93 cms−1 and 132.2 cms−1. The ignition delay time of the mixture is three times the eddy turn-over time, and the total simulation runtime is 16 times the eddy turn-over time. The Damköhler and Karlovitz numbers are 0.66 and 9.97, respectively.
22+
23+
The DNS calculation is performed using the code Athena-RFX, which implements higher-order fully conservative Godunov-type methods for integration of fluid equations. The numerics in this work are third-order accurate in space and second-order accurate in time. More details are available in the original paper. The foundational fuel chemistry model (FFCM-1) with 22 species and 107 reactions is used as the chemical mechanism.
24+
25+
26+
## Quick Info
27+
* <a href="https://www.kaggle.com/datasets/waitongchung/forced-hit-ch4-air-ffcm">Kaggle Link</a>
28+
* Contributors: Alexei Y. Poludnenko
29+
* N<sub>x</sub> = 257, N<sub>y</sub> = 257, N<sub>z</sub> = 4097, N<sub>&#632;</sub> = 6 + 21
30+
* Size = 30 GB
31+
* <a href="https://doi.org/10.1016/j.combustflame.2009.11.018">DOI</a><BR>
32+
* <a href="./assets/bib/poludnenko2010.bib">.bib</a><BR>
33+
* <a href="./assets/json/poludnenko_info.json">info.json</a>

‎_datasets/compressible_inert_ch4o2_hit.md‎

Lines changed: 28 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,28 @@
1+
---
2+
layout: datapage
3+
title: Non-reacting HIT
4+
excerpt: (1 case)
5+
header:
6+
teaser: /assets/img/ico_chung2022.png
7+
description: Compressible Inert CH4-O2 Homogeneous Isotropic Turbulence DNS
8+
categories:
9+
- nonreacting
10+
- numerical
11+
- turbulent
12+
- hit
13+
---
14+
15+
![image](./assets/img/chung2022.png)
16+
17+
## Description
18+
19+
The HIT DNS simulation is performed on a 3D cubic domain of length L, where a spherical gaseous-oxygen core of radius r = 0.25L at 300 K is initialized in gaseous methane environment of 300 K at 1 atm pressure, providing an idealized representation of an inert gaseous fuel-air mixture in a rocket engine. Periodic boundary conditions are used at all boundaries. A synthetic turbulence generator by Saad et al. based on von Kármán-Pao energy spectrum with zero mean velocity is used to generate the initial velocity profile. Ideal gas law is used as the equation of state (EoS) to relate pressure, temperature and density. The simulation is performed in an unstructured compressible finite-volume solver [64]. The solver uses a fourth-order accurate central spatial finite difference scheme. For the time integration, a stable third-order Runge-Kutta scheme is employed. Mixture-averaged transport properties are used in the DNS.
20+
21+
## Quick Info
22+
* <a href="https://www.kaggle.com/datasets/waitongchung/inert-ch4o2-hit-dns">Kaggle Link</a><BR>
23+
* Contributors: Wai Tong Chung, Matthias Ihme
24+
* N<sub>x</sub> = 129, N<sub>y</sub> = 129, N<sub>z</sub> = 129, N<sub>&#632;</sub> = 6 + 2
25+
* Size = 6 GB
26+
* <a href="https://doi.org/10.1016/j.combustflame.2021.111758">DOI</a><BR>
27+
* <a href="./assets/bib/chung2022.bib">.bib</a><BR>
28+
* <a href="./assets/json/chung2022_info.json">info.json</a>

0 commit comments

Comments
 (0)