DrivAerRim

DrivAerRim: A Computational Fluid Dynamics Dataset for Automotive Rim Aerodynamics

DrivAerRim is a computational fluid dynamics (CFD) dataset for studying how automotive rim geometry affects wheel flow and complete-vehicle aerodynamics. It contains steady Reynolds-averaged Navier--Stokes (RANS) results for all 904 rim geometries from the DeepWheel collection, installed on a fixed, full-scale DrivAer estateback.

Only the rim geometry is intentionally varied between cases. The vehicle body, closed cooling configuration, smooth underbody, deformed Rain tyres, wheel positions, open-road domain, boundary conditions, meshing strategy, solver settings, and export definitions are held constant. The same rim design is installed at all four wheel positions in each case.

The dataset is intended for controlled aerodynamic comparisons, wheel-flow analysis, and the training and evaluation of geometric deep-learning surrogate models.

Contact

Chao Xia
Department of Mechanical Engineering, Chalmers University of Technology
Email: chao.xia@chalmers.se

Key features

• 904 rim configurations named Rim0001 through Rim0904.
• A fixed, full-scale DrivAer estateback with a smooth underbody and closed cooling configuration.
• Deformed Rain tyres with longitudinal grooves and ground-contact patches.
• Rim rotation represented with moving reference frame (MRF) regions; tyre rotation represented with rotating-wall boundary conditions.
• Complete-vehicle and wheel-component force coefficients.
• Streamwise accumulated drag traces.
• Surface pressure and wall-shear-stress fields.
• Velocity, pressure, and total turbulent kinetic energy on fixed slices and in local wheel volumes.
• Preview visualisations and complete-vehicle drag-convergence histories.

CFD setup

Item	Setting
CFD software	Simcenter STAR-CCM+ 2506
Governing equations	3D steady incompressible RANS
Turbulence model	Realizable k-ε two-layer model
Wall treatment	Two-layer all-y+; target y+ < 1 on vehicle and wheel surfaces
Spatial discretisation	Second order for flow equations and turbulence convection
Flow solver	Coupled implicit solver with Roe flux-difference splitting
Inlet speed	140 km/h
Inlet turbulence intensity	0.1%
Ground	Fully moving, no-slip road at the inlet speed
Rim rotation	Cylindrical MRF region around each rim
Tyre rotation	Rotating-wall tangential velocity
Air density	1.204 kg/m³ at 20 °C
Dynamic viscosity	1.813 × 10⁻⁵ Pa·s
Reference area	2.16 m²
Iterations	3,000 per case
Reported means	Final 500 solver iterations

The open-road domain measures 15.2 vehicle lengths in the streamwise direction, 8.7 vehicle lengths in width, and 6.6 vehicle lengths in height. The vehicle is positioned 4.3 vehicle lengths downstream of the inlet and 9.9 vehicle lengths upstream of the outlet.

The baseline meshing strategy uses a predominantly hexahedral trimmed mesh with ten prism layers. External vehicle target surface sizes are 1--4 mm, with 1--2 mm local refinement around tyres, rim openings, wheelhouses, contact patches, and MRF volumes.

Repository structure

The current release is organised as follows:

DrivAerRim/
├── NumData/
│   ├── Force_Coefficients_Combined.csv
│   └── Cd_Accumulated_Combined.csv
├── Rim0001/
│   ├── Pictures/
│   ├── Slices/
│   │   ├── X/
│   │   ├── Y/
│   │   └── Z/
│   ├── Surfaces/
│   │   ├── Wheel_FL/
│   │   ├── Wheel_RL/
│   │   ├── WheelHouse_FL/
│   │   ├── WheelHouse_RL/
│   │   ├── WheelSupport_FL/
│   │   └── WheelSupport_RL/
│   └── Volumes/
│       ├── Wheel_FL_region/
│       ├── Wheel_FL_volume/
│       ├── Wheel_RL_region/
│       ├── Wheel_RL_volume/
│       └── Wake_volume/  # currently Rim0001--Rim0005
├── Rim0002/
│   └── ...
└── Rim0904/
    └── ...

RimXXXX is a zero-padded case identifier, where XXXX runs from 0001 to 0904. The location suffixes FL and RL denote front left and rear left. The front-right and rear-right installations are mirrored counterparts of the left-side installations; the released spatial records focus on the front-left and rear-left regions.

Numerical summaries

NumData/Force_Coefficients_Combined.csv

This file contains 904 rows, one per rim configuration. The Rim field stores the integer case index (1--904); it maps to case directory RimXXXX by zero-padding to four digits. All aerodynamic coefficients are dimensionless.

Field	Description
Rim	Rim case index
Cd	Complete-vehicle drag coefficient, C_D
Cl	Complete-vehicle lift coefficient, C_L
Clf	Front-axle lift coefficient, C_LF
Clr	Rear-axle lift coefficient, C_LR
Cs	Complete-vehicle side-force coefficient, C_S
Cd_FL, Cd_RL	Front-left and rear-left wheel-assembly drag coefficients
Cd_vent_FL, Cd_vent_RL	Front-left and rear-left ventilation-drag coefficients
Cd_wh_FL, Cd_wh_RL	Front-left and rear-left wheelhouse drag coefficients
Cd_ws_FL, Cd_ws_RL	Front-left and rear-left wheel-support drag coefficients
Cl_FL, Cl_RL	Front-left and rear-left wheel-assembly lift coefficients
Cl_wh_FL, Cl_wh_RL	Front-left and rear-left wheelhouse lift coefficients
Cl_ws_FL, Cl_ws_RL	Front-left and rear-left wheel-support lift coefficients

Across the 904 released rows, the complete-vehicle drag coefficient ranges from 0.2334 to 0.2549, with a mean of approximately 0.2427.

NumData/Cd_Accumulated_Combined.csv

This file contains a streamwise accumulated-drag trace for every case:

• X is the streamwise coordinate in metres.
• Rim0001--Rim0904 are the accumulated drag-coefficient values for the corresponding cases.
• The file contains 143 streamwise locations from approximately x = -0.838 m to x = 3.800 m.

The traces help identify the vehicle regions in which different rim configurations change the total drag.

Per-case spatial data

Each RimXXXX directory contains the following groups.

Surfaces

Six VTK PolyData (.vtp) files are provided per case for the front-left and rear-left wheel assemblies, wheelhouses, and wheel supports. Each file contains:

Array	Units	Description
MeanPressure	Pa	Static pressure averaged over the final 500 iterations
MeanWSS	Pa	Wall-shear-stress vector averaged over the final 500 iterations

Example path:

Rim0001/Surfaces/Wheel_FL/Wheel_FL_0001.vtp

Slices

Each case contains 128 fixed VTK PolyData (.vtp) planes:

• 80 planes normal to x, from -1.4 m to 6.5 m.
• 30 planes normal to y, from -1.5 m to 1.4 m.
• 18 planes normal to z, from -0.3 m to 1.4 m.
• Nominal spacing is 0.1 m.

The nested directory and file names record plane positions in millimetres. Each slice contains:

Array	Units	Description
MeanVelocity	m/s	Velocity vector averaged over the final 500 iterations
MeanPressure	Pa	Static pressure averaged over the final 500 iterations
MeanTotalTKE	m²/s²	Total turbulent kinetic energy averaged over the final 500 iterations

Example path:

Rim0001/Slices/X/X_2800/X_2800_0001.vtp

Volumes

Four standard VTK Unstructured Grid (.vtu) files are provided per case:

• Wheel_FL_region and Wheel_RL_region: the front-left and rear-left MRF regions around the rims and spokes.
• Wheel_FL_volume and Wheel_RL_volume: larger extraction boxes around the front-left and rear-left wheels.

Each volume contains MeanVelocity, MeanPressure, and MeanTotalTKE with the same definitions and units as the slice data.

An initial wake-volume release is also available for Rim0001--Rim0005 under Volumes/Wake_volume/. Each Wake_volume_XXXX.vtu file contains the primary MeanVelocity, MeanPressure, and MeanTotalTKE arrays, together with scalar MeanVelocityi, MeanVelocityj, and MeanVelocityk components. Wake volumes for the remaining cases are not yet included in the current release.

The larger extraction boxes cover:

Volume	x range (m)	y range (m)	z range (m)
Front left	-0.8 to 1.0	-1.25 to -0.25	-0.3 to 0.5
Rear left	2.0 to 3.8	-1.25 to -0.25	-0.3 to 0.5

Pictures

Each case contains 32 WebP preview images showing selected surface fields, slice fields, Q-criterion visualisations, total-pressure isosurfaces, and surface streamlines. It also contains Cd_steady_history.png, showing the complete-vehicle C_D history and the final 500-iteration averaging and convergence window.

Dataset splits

No official training, validation, or test split is provided at this time.

When constructing a split, keep every file associated with one RimXXXX case in the same partition to prevent information leakage between geometry, coefficient, surface, slice, volume, and image records.

Quality control and technical validation

Quality control was applied to geometry preparation, meshing, solver convergence, component-force partitions, and field export. A case passed the drag-convergence check when the range of complete-vehicle C_D over the final 500 iterations was below 0.002 (two drag counts). All 904 released cases completed and satisfied this criterion.

A separate validation configuration reproduced the detailed-underbody DrivAer model, restraint posts, and five-belt ground system used in the Volvo Cars Aerodynamic Wind Tunnel. Grid sensitivity was evaluated with 45.32, 97.88, and 180.43 million-cell meshes; the 97.88 million-cell baseline was selected as a balance between grid sensitivity and computational cost.

The realizable k-ε model was selected after comparison with wind-tunnel drag differences and surface-pressure measurements. The validation supports controlled comparisons of rim effects under this setup; it does not establish uniform accuracy for every local flow feature or every possible rim geometry.

Intended uses

DrivAerRim can support:

• analysis of rim-induced changes in complete-vehicle and wheel-component aerodynamics;
• geometric deep-learning models operating on point clouds or meshes;
• graph neural networks, neural operators, and transformer-based CFD surrogates;
• prediction of integrated force coefficients from rim or wheel geometry;
• prediction or reconstruction of local surface and flow fields;
• benchmarking case-wise generalisation across diverse rim shapes and topologies.

Limitations and responsible use

• The dataset represents a single full-scale vehicle configuration, one inlet speed, one ground condition, and one turbulence-model setup.
• The same rim is fitted at all four wheel positions within a case, so mixed front/rear rim combinations are not represented.
• Steady MRF retains one fixed spoke orientation and does not resolve transient spoke motion over a wheel revolution.
• The released arrays are averages over the final 500 steady-solver iterations, not physical time averages of an unsteady simulation.
• Native STAR-CCM+ solution files are not included, so the simulations cannot be restarted from the released data.
• Point and cell indices do not necessarily correspond across cases. Document any resampling, remeshing, interpolation, coordinate transformation, or normalisation.
• Models trained on this dataset should not be assumed to generalise to different vehicles, tyres, ride heights, Reynolds numbers, yaw angles, or numerical methods without additional validation.

Download

The full repository is approximately 2.72 TB. Targeted downloads are strongly recommended instead of cloning or downloading every case.

Install the Hugging Face client:

pip install -U huggingface_hub

Download only the numerical summaries:

hf download BeyondXia1212/DrivAerRim \
  --repo-type dataset \
  --include "NumData/*" \
  --local-dir DrivAerRim

Download one case:

hf download BeyondXia1212/DrivAerRim \
  --repo-type dataset \
  --include "Rim0001/*" \
  --local-dir DrivAerRim

Equivalent Python example:

from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="BeyondXia1212/DrivAerRim",
    repo_type="dataset",
    allow_patterns=["NumData/*", "Rim0001/*"],
    local_dir="DrivAerRim",
)

The CSV files can be read with standard tabular-data tools. The .vtp and .vtu files can be inspected with ParaView or other software supporting the Visualization Toolkit (VTK).

Citation

The accompanying manuscript is in preparation. Until a formal publication identifier is available, please cite the dataset as:

@dataset{xia2026drivaerrim,
  author    = {Xia, Chao and Lewerth, Carl and Zeng, Xi and Vdovin, Alexey and Sebben, Simone and Jia, Qing},
  title     = {DrivAerRim: A Computational Fluid Dynamics Dataset for Automotive Rim Aerodynamics},
  year      = {2026},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/datasets/BeyondXia1212/DrivAerRim}
}

Please update the citation to the peer-reviewed paper when its DOI becomes available.

Credits

The dataset was created by researchers at Chalmers University of Technology and Tongji University. The authors acknowledge Ford of Europe for access to the DrivAer model and Volvo Cars for tyre, rim, and wind-tunnel validation data. Numerical simulations were enabled by resources from the National Academic Infrastructure for Supercomputing in Sweden (NAISS), partially funded by the Swedish Research Council through Grant Agreement No. 2022-06725. The work was funded by the Transport Area of Advance Seed Project 2026.

License

DrivAerRim is released under the Creative Commons Attribution-NonCommercial 4.0 International license.