PINNeAPPle-Labs/cylinder-wake-lbm

Confined cylinder wake, lattice-Boltzmann, Re 20-100 (9 runs, 720 vorticity snapshots, 240 x 100)

2-D flow past a cylinder (diameter D = 20 lattice cells) 3 D downstream of a uniform inlet in a channel 5 D high and 12 D long (blockage 0.2), no-slip walls, pressure outlet; D2Q9 BGK lattice-Boltzmann, inlet velocity 0.1 (Mach 0.17), no turbulence model. Each run covers 200 convective units (t U / D); the last 40 are stored (saturated wake), a vorticity snapshot (normalised by U/D, 0 inside the cylinder) every 0.5 units, labelled by Reynolds number, wake regime (steady or shedding), Strouhal number and wake amplitude. For regime classification, interpolation in Re and wake forecasting. The channel is far shorter than the entrance length, so the cylinder sees a nearly uniform inflow with thin wall boundary layers. Grid: x = i, y = j lattice cells, cylinder centre and solid cells in meta.json.

cylinder_wake_lbm v1.0.0 (revision v1.0.0) · licence cc-by-4.0 · part of the PINNeAPPle zoo

Configs and splits

Config	Split	Samples	Parquet rows	Parquet files	Content
default	train	480	480	1	6 runs, Re 20, 30, 50, 60, 80, 90 (80 snapshots each)
default	test	240	240	1	3 held-out Reynolds numbers, Re 40, 70, 100 (80 snapshots each)

Split rule: by Reynolds number: every snapshot of a run is in the same split

Columns

Column	Unit	Type	Description
Re	-	float64	Reynolds number U D / nu
regime	-	string	wake regime of the run: steady or shedding
strouhal	-	float64	Strouhal number f D / U of the run (0 for a steady wake)
t	D/U	float32	convective time of the snapshot
vorticity	U/D	float32[240, 100]	vorticity on the 240 x 100 lattice (x along the channel), 0 in the cylinder
wake_amplitude	-	float64	rms transverse velocity / U, 2 D behind the cylinder

Load

With the library (the npz files, SHA-256 checked; arrays with their full shape):

pip install "pinneapple @ git+https://github.com/PINNeAPPle-Labs/PINNeAPPle"

import pinneapple as pp

d = pp.datasets.load("cylinder_wake_lbm", split="test")
print(d.n_rows, d.columns)
print(d["vorticity"].shape)

With datasets (the Parquet files of this repo):

from datasets import load_dataset

ds = load_dataset("PINNeAPPle-Labs/cylinder-wake-lbm", split="test", revision="v1.0.0")
print(ds)
ds = ds.with_format("numpy")
print(ds[0]["vorticity"].shape)

Provenance

python scripts/zoo/cylinder_data.py generate --out DIR; pinneapple_simulation.numerical_solvers.cylinder_wake (the lab experiment cylinder_lbm runs the same function)

Verification

Solver checks (python scripts/zoo/cylinder_data.py verify; pinneapple_simulation.numerical_solvers.cylinder_wake.verify): developed low-Re channel flow against plane Poiseuille, max profile error 7.6 % / 4.3 % on the 2x coarser / dataset grid (first order, full-way bounce-back walls), centreline over mean velocity 1.43 / 1.46 vs 1.5. Resolution study at Re 100 (D = 20, 30, 40 lattice cells, same physical time and sampling): Strouhal 0.2444 / 0.2458 / 0.2460, so the dataset grid (D = 20) is within 0.7 % of D = 40; vorticity rms 2 D behind the cylinder within 0.3 %; transverse-velocity rms within 6.7 % (0.9 % at D = 30). At Re 120, D = 20 gives a Strouhal number 4.4 % below D = 30 (relaxation time 0.55, near the stability limit), so the dataset stops at Re 100. Every shedding run is saturated: the probe rms changes by at most 5 % across the stored window; below the onset (Re <= 50, confinement delays it from the unconfined ~47) disturbances decay. Strouhal number from a Hann-windowed, peak-refined spectrum (raw FFT bins are 0.025 apart). All fields finite.

Files

The npz files are the library's format and the source of truth: pp.datasets.load downloads them from this repo at the tagged revision and refuses any file whose SHA-256 differs from the manifest of the package. The Parquet files under data/ are derived from them by pinneapple_hub.zoo.parquet.write_dataset and hold the same numbers (a test round-trips npz -> Parquet -> arrays bit for bit); they are what the dataset viewer and load_dataset read.

File	Size	SHA-256
train-00.npz	28.3 MB	7932f317a12d52e77abb0cebbcf32f750562650999279b32f83dbdbf6008127f
train-01.npz	14.2 MB	86f534293e63c57ec2fa9e0c0213a386b3b245b64e69a5dfb82d66938e38b672
test.npz	21.3 MB	6c50430490bfd1493067cf016b11fb1073034f2ab5b6ca97afb09a40ac46e130

Total 63.8 MB of npz.

Citation

Cylinder wake dataset generated with PINNeAPPle (pinneapple_simulation.numerical_solvers.cylinder_wake, D2Q9 lattice Boltzmann); PINNeAPPle, Zenodo DOI 10.5281/zenodo.23343062.

Cite

If you use this dataset, please cite PINNeAPPle (software, Zenodo DOI 10.5281/zenodo.23343062; see CITATION.cff):

@software{pinneapple,
  author = {Barros, Yan and Schiassi, Enrico},
  title  = {PINNeAPPle},
  doi    = {10.5281/zenodo.23343062},
  url    = {https://github.com/PINNeAPPle-Labs/PINNeAPPle}
}