Transitional Boundary Layer

Description

These snapshots are derived from a Direct Numerical Simulation (DNS) of a spatially developing, incompressible transitional boundary layer over a flat plate, which is publicly available through the Johns Hopkins Turbulence Database (JHTDB). The simulation captures the full transition process of a zero-pressure-gradient boundary layer — from a laminar Blasius inflow, through the growth and breakdown of disturbances, to a fully developed turbulent state — driven by free-stream turbulence. The flow field stored in the database is defined on a grid of $3320 \times 224 \times 2048$ points in the streamwise (x), wall-normal (y), and spanwise (z) directions respectively.

For this machine-learning-oriented release, we subsampled the original JHTDB transitional boundary layer fields and re-saved them in a new, ML-friendly format. Each snapshot retains the complete three-dimensional flow field with all three velocity components (u, v, w) and pressure. Two versions of the dataset are provided:

• Sequential version — 20 consecutive snapshots sampled at the stored time interval, preserving the temporal evolution of the transition process. This version is intended for tasks that require temporal coherence, such as spatio-temporal forecasting, super-resolution in time, or dynamics/transition-onset learning.
• Random version — 10 snapshots drawn at random (non-consecutive) times from the statistically stationary regime. This version is intended for tasks that benefit from diverse, decorrelated samples, such as single-frame reconstruction, statistical modeling, or generative learning.

Quick Info

• Contributors: Charles Meneveau
• N_ɸ = 4 + 3
• N_x = 3320, N_y = 224, N_z = 2048
• DOI
• .bib
• Download.sh
  

Links to different cases

ID	Description	Size (GB)	Links
0	Sequential Snapshots	561	Kaggle1-8, info.json1-8, Kaggle9-16, info.json9-16, Kaggle17-20, info.json17-20,
1	Collection of snapshots at different time	281	Kaggle1-8, info.json1-8, Kaggle9-10, info.json9-10,