CFD vs Wind Tunnel Results
Explore Our Tested Simulations and Compare Them Against Real Experimental Data
In this study, all simulations were performed using Bramble’s CFD workflows built around OpenFOAM (OpenFoam v1806), ensuring a robust, repeatable, and industrially scalable simulation framework.
In computational fluid dynamics, accuracy is everything. Engineering decisions ranging from aerodynamic performance to thermal management and aeroacoustics are increasingly driven by simulation results. However, modelling assumptions such as compressibility effects, thermal effects, and turbulence modelling can introduce uncertainty or loss of accuracy.
This is precisely why CFD validation against high-quality experimental data is essential for real engineering confidence
Correlation is not a formality. It is the foundation of credibility.
Without systematic validation against wind tunnel measurements or well-documented benchmark datasets, CFD predictions remain theoretical. Correlation builds confidence in:
- Aerodynamic coefficients (Cd, Cl, Cm)
- Surface pressure distributions
- Velocity fields and wake structures
- Separation behaviour
- Flow topology and vortex dynamics
In industrial practice, especially in automotive aerodynamics, tight correlation between CFD and wind tunnel data directly impacts design cycles, cost reduction, and performance optimisation. Reliable CFD validation enables teams to reduce physical testing dependency while maintaining engineering certainty.
To explore this topic in depth, we are launching a technical blog series focused on CFD-to-wind-tunnel correlation using well-established aerodynamic benchmark models. Each post will examine a specific model, discuss the numerical setup, and compare high-fidelity CFD results with experimental data, (wind tunnel data).
We begin with one of the most important modern automotive benchmarks: the DrivAer model.
The DrivAer Model: A Modern Automotive Aerodynamic Benchmark
The DrivAer model was developed as a collaboration between the Technical University of Munich (TUM) and Audi AG to provide an open, realistic, and highly detailed automotive benchmark geometry for aerodynamic research.
Unlike simplified academic bodies such as the Ahmed body, the DrivAer model represents a more production-like passenger vehicle. It includes:
- Realistic vehicle proportions
- Detailed underbody geometry
- Multiple rear-end variants (fastback, notchback, estate)
- Detailed surface features
The model was designed specifically to close the gap between simplified bluff-body research models and proprietary production vehicle geometries.
DrivAer provides:
- High-quality CAD geometry
- Detailed wind tunnel measurement datasets
- Surface pressure measurements
- Aerodynamic force data
- PIV (Particle Image Velocimetry) wake measurements
This makes it an ideal benchmark for evaluating:
- Turbulence model performance (RANS, DES, LES)
- Boundary layer resolution strategies
- Wake prediction capability
Because of its realism and available experimental data, DrivAer has quickly become a reference case in the automotive CFD community.
bramble CFD Results: Correlation with Wind Tunnel Data
In this first study of our series, we present bramble CFD results for the DrivAer model from the AutoCFD04 Conference and compare them against available wind tunnel measurements.
The objectives of this analysis are:
- Evaluate drag coefficient prediction accuracy
- Compare surface pressure distributions
- Assess wake structure prediction
[https://autocfd4.s3.eu-west-1.amazonaws.com/test-cases/case2/AutoCFD4_Case2_Intro_240409.pdf]
[https://autocfd4.s3.eu-west-1.amazonaws.com/test-cases/case2/AutoCFD4_Case2_Intro_240409.pdf]
Numerical Setup Overview
The simulations were conducted using:
- ANSA committee provided grid
- bramble with an OpenFOAM solver
- DDES-SA
- RunTime = 4s and results averaged over the last 2s
- Time-step = 0.00025s
- Forces convergence validated with meanCalc
- In–house all Y+ treatment with lookup tables
Results and Discussion
Aerodynamic Coefficients
The predicted drag coefficient (Cd) for both geometries were close to the experimental values, with an over-prediction of 43 counts.
The effect of geometry change, or delta prediction showed strong correlation with the experimental data, with the CFD results having an 8 deltaCd counts against 12 deltaCd counts for the experimental results:
Surface Pressure Distribution
Surface Cp comparisons demonstrated excellent correlation across key areas of the vehicle:
- Base pressure along upper centrebody
- Base pressure along under centrebody
- Rear body side pressure
Case 2a Results:
[Cp upperbody centreline]
[Cp upperbody at centreline]
[Cp upperbody at sidewall]
The rear-end pressure recovery region shows slight variation, over predicting the wind tunnel experimental values. The results with the largest spread between participants and all CFD codes were in this region on the notch, highlighting the difficult and sensibility of CFD codes in the prediction of complex flow separation with mild pressure gradients.
[Cp upperbody centreline]
For the delta comparison between Case 2a and Case 2b, the flow field variations within the wheelhouse region exhibit improved agreement between the CFD predictions and the wind tunnel measurements. In particular, the computed ΔCp values reproduce the experimentally observed trends consistently across the probe locations, both in magnitude and directional behaviour.
The relative changes between the two configurations are captured more accurately than the corresponding absolute values, indicating that the CFD model reliably resolves the incremental aerodynamic effects associated with the geometric modification.
Velocity plots
A series of PIV measurements of velocity along horizontal and vertical lines across the model were available for results comparisons:
Velocity lines along the underfloor for both horizontal and lateral measurements had good agreement with experimental results, demonstrating high capability in the code of predicting the complex flowfield under the vehicle:
[L1 plot Velocity]
[U1 and U2 plot Velocity]
[U3 and U4 plot Velocity]
[U5 plot Velocity]
In addition, wake velocity profiles and recirculation zones show good qualitative agreement with the experimental PIV data.
[V1 and V2 plot Velocity]
[V3 plot Velocity]
The largest difference was observed further downstream, where the CFD results had less velocity deficit in relation to the experimental data, which could be attributed as the wake diffusing more rapidly in CFD, but still keeping a similar topology.
[V5 plot Velocity]
Why This Matters
Correlation studies such as this serve multiple purposes:
- Build trust in CFD-driven design decisions
- Identify modelling limitations
- Guide turbulence model improvements
- Establish best practices for industrial workflows
For engineering teams relying on CFD to reduce wind tunnel testing cycles, validated simulation frameworks are essential.
What’s Next in This Series
This article marks the beginning of our benchmark CFD validation series. Upcoming posts will cover:
- Other automotive benchmark models
- Imperial Front Wing
- Turbulence model comparison studies
- Other canonical benchmark models
Each case will present transparent modelling details and direct comparison to experimental data.
Our goal is not only to demonstrate performance, but to provide insight into why certain modelling strategies correlate better than others.
Because in CFD, accuracy is not assumed.
It is demonstrated.