SHIFT-Pump-sample: High-Fidelity Computational Fluid Dynamics Dataset for Centrifugal Pump Design and Analysis

We're excited to introduce the SHIFT-Pump dataset — a high-fidelity fluid dynamics simulation dataset developed as part of the Luminary SHIFT Models initiative. This dataset enables the training and benchmarking of real-time physics AI models for pump design and performance analysis.

Website: shift.luminarycloud.com

Contact: shift@luminarycloud.com

Note: this is a sample dataset of the first ~100 samples in the full SHIFT-Pump. It provides the same information as provided in the full dataset. This should be sufficient to assess pipelines and test whether you want access to the full dataset. The full dataset is available at https://huggingface.co/datasets/luminary-shift/Pump.

Summary

Physics AI models can transform early stage pump design by giving users real-time feedback on the physics-based performance implications of design decisions. However, the lack of high-quality training data has been a barrier to their development. Luminary SHIFT Models provide access to both high-quality datasets and pretrained models for a variety of applications and industries.

SHIFT-Pump is a massive step forward in this direction: purpose-built for high-fidelity pump performance inference, without requiring CFD expertise or meshing. This dataset is based on thousands of parametrically varied pump geometries operating at and outside design flow conditions, allowing systematic exploration of the design space for centrifugal pump impellers and volutes.

This dataset supports training surface-based or volume-based pump performance surrogate models, real-time inference systems, and exploring geometry-performance correlations for pump design optimization.

Applications

• Rapid pump design prototyping and performance optimization
• Research in fluid dynamics inference, point cloud learning, or physics-aware generative models for turbomachinery
• Training and fine-tuning Physics AI models for pump design

Attribution

Please attribute Luminary Cloud for the SHIFT-Pump model and dataset.

An article is being prepared so users can cite this dataset - we will update this accordingly when available. Until then you can use this citation:

@misc{shift_pump_2026,
  author = "{Luminary Cloud}",
  title = {SHIFT-Pump: High-Fidelity Computational Fluid Dynamics Dataset for Centrifugal Pump Design and Analysis},
  year = {2026},
  url = {https://huggingface.co/datasets/luminary-shift/Pump/}
}

Contents

This repository contains the SHIFT-Pump dataset with over 5,600 unique pump geometry and operating condition configurations. We will continue to push newly computed samples to this repository periodically, on our path towards comprehensive pump design space coverage. The data generation and organization within the repository is described below.

Geometry Creation

The geometry creation process transforms design parameters into a complete pump geometry consisting of an impeller and volute casing. Geometric curves that define blade profiles and volute cross-sections are computed from the input parameters, generating curve data files for blade profiles, periodic boundaries, and meridional flow paths. Three-dimensional blade geometries are created by lofting profile curves to create blade surfaces, generating a periodic sector volume, and extracting individual blade solids. The volute creation process loads a parametric template, applies volute geometric parameters and generates the volute casing geometry. The volute geometry is then imported into a cloud CAD system where fillets are applied to specified edges before being exported. The final assembly stage imports the impeller sector as base geometry, imports and patterns blades around the rotation axis, subtracts blades to create flow passages, and imports the volute.

CFD Solver

The simulations are run on a fixed mesh with boundary layers applied to both impeller and volute surfaces.The simulation setup configures a rotating reference frame simulation with water as the working fluid. The impeller domain rotates at the specified rotational speed while the volute domain remains stationary, connected through a sliding interface. Boundary conditions include a total pressure inlet, walls with surface roughness on both impeller and volute surfaces, and an outlet with target mass flow rate based on the operating conditions. example pump flow field visualization

Files

At the root level of the repository you will find directories containing the simulation data organized using unique sample IDs. Each directory corresponds to a unique pump geometry configuration or operating condition. The directory structure looks like:

sample_00001
sample_00002
sample_00003
...

The full dataset currently contains over 5,600 unique geometry + operating condition configurations. In each geometry directory you will find the following files:

• merged_surfaces.stl: STL file with the pump's geometry (impeller and volute surfaces)
• merged_surfaces.vtp: surface field solution file with pressure and wall shear-stress from the CFD simulation
• merged_volumes.vtu: volume field solution file with pressure and velocity fields from the CFD simulation
• params.json: JSON file containing the parametric design values including flow_rate, head, teAngleDelta, leAngleDelta, flow_rate_op_condition, rotational_speed, and other geometric parameters
• metadata.json: JSON file containing simulation metadata including simulation_id, geometry_id, variant_name, and parametrization details
• forces.json: JSON file containing computed quantities of interest like torque and outlet total pressure

Downloading

You can use HuggingFace to gain access to the entire repository, but will require the associated storage available locally. Note you will need to have git lfs installed first, then run

git clone git@hf.co:datasets/luminary-shift/Pump
# to ensure future `git pull` commands won't checkout full files, you'll want to ensure the skip is active in this repo
cd <path/to/repo>
git lfs install --skip-smudge --local

You can then pull down files you want to interact with in multiple ways:

# pull a specific file
git lfs pull --include="path/to/your/file"
# pull a directory
git lfs pull --include="path/to/file1,path/to/dir/*"
# pull, but exclude certain paths
git lfs pull --exclude="**/*.mp4"

and remove those files and reset them to pointers when done using them:

rm path/to/your/file
git checkout -- path/to/your/file

Credits

Luminary Cloud

The SHIFT-Pump dataset was developed by Luminary Cloud as part of the SHIFT Models initiative, providing comprehensive pump design data for physics AI model training and pump design optimization.

Please attribute Luminary Cloud for the SHIFT-Pump model and dataset.

License

This dataset is distributed under the CC-BY-NC-4.0 license, which is also included in the dataset itself. By downloading the dataset you acknowledge the terms of this license.