license: cc-by-4.0
tags:
- 2D Steady-State Laminar Flow
- Flow Field Prediction
language:
- en
- zh
DeepCFD
Dataset Overview
The DeepCFD dataset is a publicly available two-dimensional steady-state laminar-flow computational fluid dynamics dataset created by a research team affiliated with the German Research Center for Artificial Intelligence (DFKI). The dataset was generated with OpenFOAM's simpleFoam solver and contains approximately 1,000 samples of flow around randomly generated obstacles in a channel. It is suitable for CFD solver acceleration, evaluation of flow-field generalization across different obstacle configurations, prediction-error analysis, and flow-field visualization.
Supported Tasks
This standardized dataset repository organizes DeepCFD's input geometry and signed distance fields in dataX.pkl and the corresponding ground-truth CFD solutions in dataY.pkl. Each sample contains steady-state flow variables including computational mesh coordinates, horizontal velocity (U_x), vertical velocity (U_y), and pressure (p). It also provides a signed distance function (SDF) and flow-region class labels to represent obstacle geometry, the computational domain, and boundary conditions. The data can be used as input for training, inference, evaluation, and visualization with the OneScience/DeepCFD model.
Data Format
| File | shape | dtype | Channel Description |
|---|---|---|---|
data/dataX.pkl |
[981, 3, 172, 79] |
float32 |
Channel 0: signed distance to the obstacle surface; Channel 1: fluid-region label; Channel 2: signed distance to the top/bottom surfaces |
data/dataY.pkl |
[981, 3, 172, 79] |
float32 |
Channel 0: horizontal velocity Ux; Channel 1: vertical velocity Uy; Channel 2: pressure pressure |
Each sample in this dataset is represented on a 172 × 79 regular grid. The input data consists of the signed distance to the obstacle surface, fluid-region labels, and signed distances to the top/bottom surfaces. The output data consists of the corresponding horizontal velocity Ux, vertical velocity Uy, and pressure pressure.
How to Use the Dataset
This dataset is compatible with the OneScience-Sugon/DeepCFD model. During DeepCFD runs, scripts in the model package can organize the data from this repository into the model's working directory.
Download the dataset:
hf download --dataset OneScience-Sugon/deepcfd --local-dir ./data
Official OneScience Information
| Platform | OneScience Main Repository | Skills Repository |
|---|---|---|
| Gitee | https://gitee.com/onescience-ai/onescience | https://gitee.com/onescience-ai/oneskills |
| GitHub | https://github.com/onescience-ai/OneScience | https://github.com/onescience-ai/oneskills |
Citation and License
- Original DeepCFD paper: DeepCFD: Efficient Steady-State Laminar Flow Approximation with Deep Convolutional Neural Networks
- This dataset was converted from the original DeepCFD dataset and is licensed under the same CC-BY-4.0 license. The original copyright notice and license text must be retained when using, modifying, or redistributing it.