Ahmed body calculator: drag coefficient, frontal area and Fluent reference values
Set the velocity, the slant angle and whether you model the whole body or half of it. You get the reference area to type into Fluent, the Reynolds number, the blockage of your enclosure and the drag force the wind-tunnel data predict. Free, no sign-up, runs in your browser.
The second tab checks the drag coefficient in your report: it re-bases it on the right area, velocity and density, and puts it next to the measured values and published RANS results at your slant. It opens on the Ahmed body challenge of Skill-Lync's CFD courses (25 m/s, Fluent's default air). For the first cell height, use the y+ calculator.
Ahmed et al. 1984 (as quoted) · Meile et al. 2011 · ERCOFTAC AC1-05 · Ansys Fluent guide · 805 student reports read · Updated 22 Nov 2026
Fluent's default air is 1.225 kg/m³ and 1.7894 × 10⁻⁵ Pa·s. The measured Cd values are on the body's frontal area (0.112 m², stilts left out); the drag force uses that area. The enclosure is the CFD box's cross-section; leave it blank to skip the blockage.
Set the case above to what you ran (velocity, slant, full or half model), then type the numbers from your report and the reference values from Fluent's Reference Values panel.
Circles: wind-tunnel values open sources quote at this slant (shaded: their range). Squares: published RANS results. Dashed: the 0.33 many reports compare with.
- Fluent, Reference Values: Area 0.11203 m², Length 1.044 m (only the moment coefficient uses it), Velocity 25 m/s, Density 1.225 kg/m³. Fluent does not set the area or length when you compute from a boundary: type them.
- At 25° the flow is three-dimensional: two strong C-pillar vortices roll up off the slant's side edges and drag rises with the angle; at 25° the flow separates near the top of the slant and reattaches on it.
The Ahmed body in numbers
The Ahmed body is a generic car shape: a rounded front, a box and a rear slant whose angle can be changed. S. R. Ahmed, G. Ramm and G. Faltin published their measurements on it in 1984 (SAE 840300), and it has been the standard check for car aerodynamics CFD since, because the slant angle switches the wake between very different flows.
| Dimension | Value | Note |
|---|---|---|
| Length L | 1044 mm | the length the Reynolds number uses |
| Width W | 389 mm | |
| Height H | 288 mm | body only, ground gap not included |
| Front radius | 100 mm | R/H = 0.347 (Krajnović and Davidson) |
| Slant length | 222 mm | the same at every slant angle; 201 mm along the ground at 25° |
| Ground clearance | 50 mm | four stilts, 30 mm diameter |
| Frontal area A | 0.112 m² | W × H; 0.115 m² with the front pair of stilts |
The rounded front doesn't change the outline seen from the front, so the frontal area is just width × height:
Meile et al. (2011), the AhmedML dataset and the Liu and Moser study all use 0.112 m². It leaves out the stilts: Meile's team measured the stilts' drag on their own and subtracted it.
What drag coefficient should the Ahmed body have?
It depends on the slant. Ahmed's paper isn't open, so the values below are the ones open sources quote, each named. They don't agree to the third decimal at 25°: each source read Ahmed's figures its own way.
| Slant | Cd | Source | Speed |
|---|---|---|---|
| 25° | 0.285 | Ahmed et al. 1984, as quoted by COMSOL | 60 m/s |
| 25° | 0.2875 | Ahmed et al. 1984, read off their Fig. 4 by SimScale | 60 m/s |
| 25° | 0.298 | Ahmed et al. 1984, as quoted by Krajnović (2014) | 60 m/s |
| 25° | 0.299 (Cl 0.345) | Meile et al. 2011, Graz, 40 m/s (as tabulated by AhmedML and SimFlow) | 40 m/s |
| 35° | 0.26 | Ahmed et al. 1984, as quoted by Liu and Moser (ETH Zurich) | 60 m/s |
Where the parts are quoted, pressure drag dominates. Ahmed's abstract puts it at almost 85% of the total:
| Part of Cd | 25° (Ahmed via COMSOL) | 35° (Ahmed via Liu and Moser) |
|---|---|---|
| Front (pressure) | 0.02 | 0.02 |
| Slant (pressure) | 0.14 | 0.09 |
| Base (pressure) | 0.07 | 0.095 |
| Friction | 0.055 | 0.055 (the rest) |
| Total | 0.285 | 0.26 |
At 25° most of the pressure drag sits on the slant: two strong vortices roll off its side edges and pull its pressure down. At 35° the flow has left the slant and the base takes over. Between them, at about 30°, is the critical angle. In Meile's words, "both the drag and lift coefficients change abruptly at the angle φ = 30°", and just below it the vortices are strongest and the drag peaks.
Two cautions. Krajnović and Davidson note that Ahmed measured the pressure-drag split at only three slants (5°, 12.5° and 30°) and interpolated the rest. And no open source quotes a measured Cd at other angles, so this calculator compares only at 25° and 35°. If your CAD has a 20° or 30° slant (the reports mention 20°, 25°, 30° and 35°), measure the angle before you compare.
Where does "Cd = 0.33" come from?
Skill-Lync read all 805 distinct public Ahmed body reports (839 pages) on 22 Nov 2026. 660 have text to read and 474 of those run the challenge at 25 m/s. 62 call 0.33 the experimental, literature or reference drag coefficient for the Ahmed body. Of the 114 reports at 25 m/s that quote a Cd of their own, the middle half land between 0.3 and 0.35, centred on 0.329.
No paper, database or validation case Skill-Lync could find gives 0.33, at 25 m/s or any other speed. The 25° body measures 0.285 to 0.299 and the 35° body 0.26. A grid study that keeps refining until it reaches 0.33 is tuning to a number with no source. Run the study for its own sake: refine until Cd stops moving, then compare that with the measured range and explain the gap.
Frontal area: 0.112 m², not 0.1315
Fluent works out the drag coefficient from the force and three reference values:
The area is the value reports get wrong most often. 55 of the reports give a number for the reference, frontal or projected area. 16 of them use 0.131482 m², or 0.065741 m² on a half model, several straight from Fluent's projected-area report. That is exactly 389 × 338 mm: the width times the height down to the ground, as if the 50 mm gap under the body were solid. The coefficient comes out 0.852 of the right one, 15% low, which is enough to turn a 0.30 into a 0.26. 9 use 0.112 to 0.116 m², which is right. 6 type a length (the 0.288 m height, or the width) where an area goes.
Two more slips the check tab catches:
- Area left at 1 m². Fluent uses 1 m² unless you change it, and computing the reference values from the inlet sets the velocity and density, not the area or the length. A Cd worked out on 1 m² is 8.93 times too small.
- The full area on a half model. A symmetry plane halves the force, so the area must be halved as well, 0.05602 m². With the full area the coefficient comes out half the body's.
To re-base a coefficient worked out with the wrong reference values:
Reynolds number: 25 m/s is not 60 m/s
| Case | U | ReL | ReH | Drag* |
|---|---|---|---|---|
| Skill-Lync challenge | 25 m/s | 1.79 M | 0.49 M | 12.2 N |
| Erlangen, Graz | 40 m/s | 2.86 M | 0.79 M | 31.3 N |
| Ahmed 1984 | 60 m/s | 4.29 M | 1.18 M | 70.4 N |
Re in millions, in Fluent's default air. *Drag on the full body at Cd 0.285 and 0.112 m². The experiments' own figures, in their own air, are 4.29 million (Ahmed, on the length) and 2.78 million on the length or 768,000 on the height (Lienhart and Becker).
The challenge runs at under half of Ahmed's Reynolds number. Meile et al. measured the 25° body at 10, 20, 30 and 40 m/s and fitted one curve of Cd against Reynolds number through their points and the earlier ones, over a tenfold range: the coefficient is not a property of the shape alone. Treat a 25 m/s result against 60 m/s data as a comparison with a few per cent of slack, not a match to three decimals. The drag force, though, goes with U²: the same body feels 5.76 times the drag at 60 m/s as at 25 m/s.
The enclosure matters too. The 40 m/s tunnels had 1.87 × 1.4 m and 2 × 1.46 m sections, so the body blocked about 4.3% and 3.8% of them. A box much tighter than that speeds the flow past the body and raises the drag. Type your enclosure's width and height and the calculator gives the blockage.
What RANS models get right and wrong on the Ahmed body
| Slant | Model | Cd | Run |
|---|---|---|---|
| 25° | k-ε | 0.297 | Hachimy et al. 2022, Fluent, 40 m/s, half model |
| 25° | k-ω SST | 0.2839 | Hachimy et al. 2022, Fluent, 40 m/s, half model |
| 25° | k-ε, wall functions | 0.3 | COMSOL verification model, 40 m/s, stilts included |
| 25° | k-ω SST | 0.313 | SimFlow (OpenFOAM simpleFoam), 40 m/s |
| 25° | k-ω SST | 0.2915 | SimScale, 60 m/s, A = 0.115 m² |
| 35° | k-ε, wall functions | 0.242 | Liu and Moser, transient RANS |
| 35° | k-ε-v2 | 0.264 | Liu and Moser, transient RANS |
| 35° | SSG Reynolds stress | 0.21 | Liu and Moser, transient RANS |
| 35° | RSM | 0.282 | Liu and Moser, transient RANS |
| 35° | SST | 0.241 | Liu and Moser, transient RANS |
The totals look close: every 25° run in the table is within 10% of the measured range. The flow behind them is often wrong. At 25° the real boundary layer separates near the top of the slant and reattaches on it. Guilmineau et al. (2018) report the "failure of RANS models to predict the flow for the slant angle 25°". The ERCOFTAC evaluation notes that k-ε type models keep the flow attached and SST type models let it stall from the start of the slant. COMSOL's own k-ε validation lands at 0.30 against 0.285, while its front pressure drag is three times the measured one (0.06 against 0.020) and its slant drag is too low (0.12 against 0.14).
So a total Cd within a few per cent doesn't validate the model. Plot the pressure on the slant and the base, look at where the flow separates, and compare the parts. At 35°, where the flow leaves the slant edge cleanly, RANS has an easier time: Liu and Moser's runs span 0.21 to 0.282 around the measured 0.26.
What this calculator does not cover
- Slant angles other than 25° and 35°. Ahmed's full curve is in the paywalled paper; no open source quotes it as numbers. The set-up tab still gives the area, Reynolds number and reference values at any angle.
- Lift. Meile et al.'s Cl of 0.345 at 25° is the only open measured lift value; the check tab compares drag only.
- Moving ground and wheels. All the data here are from a fixed floor with the body on stilts.
- The first cell height and inflation layers. Those are on the y+ calculator, which opens on the same challenge.
Setting reference values, sizing the domain and knowing what a validation case can and can't prove are the everyday work of a CFD engineer. The Ahmed body is the external-aerodynamics project in Skill-Lync's CFD courses.
Frequently asked questions
What is the drag coefficient of the Ahmed body?
It depends on the rear slant angle. For the 25° slant, open sources quote 0.285 to 0.299: Ahmed, Ramm and Faltin's 1984 value at 60 m/s as quoted by COMSOL (0.285), SimScale (0.2875) and Krajnović (0.298), and Meile et al.'s 2011 Graz measurement at 40 m/s (0.299). For 35°, Ahmed's value as quoted by Liu and Moser is 0.26. All of them use the body's frontal area, 0.112 m².
What reference area should I use for the Ahmed body in Fluent?
The frontal area, width × height: 0.389 × 0.288 = 0.112 m² for the full body, half of it (0.05602 m²) for a half model on a symmetry plane. Fluent's reference area is 1 m² unless you change it, and computing the reference values from the inlet does not set it. 0.1315 m², which 16 of the 55 Skill-Lync reports that state an area use (or its half), is 389 × 338 mm: it counts the 50 mm gap under the body as solid and makes Cd 15% too low.
Is the Ahmed body drag coefficient 0.33?
No open source Skill-Lync could find gives 0.33. 62 of the 660 public Skill-Lync Ahmed body reports with text call 0.33 the experimental value at 25 m/s, and the reports that quote their own Cd at 25 m/s cluster around it (middle half 0.3 to 0.35). The measured values for the 25° body are 0.285 to 0.299; for 35°, 0.26.
What velocity and Reynolds number did Ahmed use?
Ahmed et al. tested at 60 m/s, a Reynolds number of 4.29 million on the 1.044 m length. Lienhart and Becker (Erlangen) and Meile et al. (Graz) used 40 m/s, Re 2.78 million on the length or 768,000 on the height. The Skill-Lync challenge runs 25 m/s: in Fluent's default air that is Re 17,86,772 on the length, under half of Ahmed's.
Should the stilts be included in the Ahmed body drag?
Meile et al. measured the stilts' drag separately and subtracted it, so their Cd is on the body alone with A = 0.112 m². If your force report includes the stilts, include their frontal area too: 0.115 m² with the front pair (the rear pair hides behind them in the front view), 0.118 m² counting all four, as COMSOL's half-model 0.059 m² does. Better, leave them out of the force report.
Why does my k-ε result match the Ahmed body drag but k-ω SST does not?
On the 25° slant the real flow separates and reattaches. k-ε type models tend to keep it attached and SST type models tend to let it stall from the start of the slant, so neither gets the slant pressure right, and a total Cd close to the experiment can come from errors that cancel. Compare the slant and base parts, not only the total.
Related on Skill-Lync
- y+ calculator: first cell height and inflation layers for the Ahmed body
- Drag force calculator
- Flow over a cylinder calculator
- Mesh quality calculator
- Flow over an Ahmed body using open-source software
- Student project: Week 3, external flow over an Ahmed body
- CFD courses compared
- All free engineering tools
Sources
- S. R. Ahmed, G. Ramm, G. Faltin, "Some salient features of the time-averaged ground vehicle wake", SAE 840300 (1984): 60 m/s, ReL 4.29 × 10⁶, "almost 85% of body drag is pressure drag" (abstract). Paywalled; values as quoted by the sources below.
- ERCOFTAC Knowledge Base, Application Challenge AC1-05 "Ahmed body": description (geometry, stilts, 222 mm slant at every angle), test data (60 and 40 m/s, turbulence levels, ¾-open sections) and evaluation (F. Menter's 2024 note on k-ε and SST behaviour at 25°).
- COMSOL Multiphysics 5.6 Application Library, "Airflow over an Ahmed body": Table 1, Ahmed's measured parts at 25° (0.020, 0.140, 0.070, 0.055; total 0.285) against a k-ε model (0.30); half-model area 0.059 m² with the stilts.
- W. Meile, G. Brenn, A. Reppenhagen, B. Lechner, A. Fuchs, "Experiments and numerical simulations on the aerodynamics of the Ahmed body", CFD Letters 3(1), 2011, 32-39: Graz tunnel 2 × 1.46 m, A = 0.112 m², blockage 3.8%, 40 m/s, stilt tare loads subtracted, critical angle 30°; Cd 0.299, Cl 0.345 at 25° as tabulated by N. Ashton et al., AhmedML, arXiv:2407.20801 (2024), and by SimFlow's validation page.
- S. Krajnović, L. Davidson, "Flow around a simplified car, part 1", J. Fluids Eng. 127 (2005) 907 (R/H 0.347, Lienhart and Becker did not measure drag, Ahmed's pressure split measured at 5°, 12.5° and 30°); S. Krajnović, J. Fluids Eng. 136 (2014) 121103 (Ahmed's 25° Cd as 0.298).
- H. Lienhart, S. Becker, "Flow and turbulence structure in the wake of a simplified car model", SAE 2003-01-0656: 40 m/s, ReH 7.68 × 10⁵, 25° and 35°; test section 1.87 × 1.40 m (6.493H × 4.861H in Krajnović and Davidson).
- Y. Liu, A. Moser (ETH Zürich), "Numerical modeling of airflow over the Ahmed body": Table 1, Ahmed at 35° (0.020, 0.095, 0.090; total 0.260) and five transient RANS models.
- F. Hachimy, A. A. Omar, O. Elsayed, CFD Letters 14(5) (2022) 24-32: Fluent k-ε 0.297, k-ω SST 0.2839 at 25°. E. Guilmineau et al., Computers & Fluids 176 (2018) 302-319: RANS fails at 25°. SimScale and SimFlow Ahmed body validation cases (0.2875 read from Ahmed's Fig. 4; SST 0.2915 and 0.313).
- Ansys Fluent Theory Guide 2024 R2 §25.2.1 (force coefficient = force / ½ρv²A from the reference values); Fluent 12.0 User's Guide §30.11.1 (reference length and area are not set by computing from a boundary), §33.10, §8.3.2 and Tutorial Guide (default air 1.225 kg/m³, 1.7894 × 10⁻⁵ kg/m·s). Default reference area 1 m²: Ansys Learning Forum, answer by an Ansys engineer, Nov 2023.
- Skill-Lync, all 805 distinct public Ahmed body reports (839 pages), read 22 Nov 2026: velocities, the 0.33 claim, the Cd each quotes, the reference areas typed.
- Skill-Lync, scripts/verify/ahmed.py: counts the areas on a 0.1 mm raster of the front view, checks the quoted parts add up, re-computes every case and each slip; the page checks itself against it at build time.