## About the author

Chao-Sheng Chen is the General Manager of <ExternalLink href="https://www.ksvalve.com/">Kong Jih Valve</ExternalLink> Ind. Co., Ltd.,
an established Taiwanese valve manufacturer. Together with Chin-Chi
Chang of the engineering department, they have benchmarked AirShaper
against both their internal calculations and physical test results.

## Ball valves

A ball valve is a popular fluid-control device -- it is durable and easy
to repair. But a valve also introduces resistance (in the form of a
pressure drop), which increases the required energy to pump a fluid
through the valve. The efficiency of a valve is indicated by the [Flow
Coefficient](https://en.wikipedia.org/wiki/Flow_coefficient) (Cv):

<ImageWithCaption
 src="/assets/images/ksvalve_case_1.jpg"
 alt=""
 width="457"
 height="199"
 positionCenter={true}
/>

This Cv can thus be used to calculate the flow rate (given a certain
pressure drop) or the pressure drop (given a certain flow rate). As this
impacts the required energy, pumps, ... for fluid systems, customers are
increasingly interested in obtaining accurate Cv values as input for
their engineering teams.

## Physical testing

The most direct approach to obtain Cv values is to perform a physical
test, i.e. integrating the valve into a test system where the flow
parameters (pressure upstream & downstream, flow rate) can be monitored.
Based on the output, the Cv value can then be calculated. But this
testing process is lengthy and costly. In addition, each design change
requires a new test.

## Approximate formulae

We have in-house formulae to predict the Cv values, but they are
one-dimensional calculations. They don't take the geometrical details
and complex flow aspects, i.e. boundary layers, flow separation, etc.,
into account. They are relevant to obtain the right order of magnitude,
but not precise enough to replace a test. For example, for the valves
tested in this study, we observed deviations up to 68% between the
calculated values and the physical test results. Using these, we'd have
to add a large uncertainty margin, ending up with conservative Cv values
for our customers.

## CFD simulations

To obtain more precise Cv values without having to test each design, we
turned to AirShaper for their simulation expertise. To gain confidence,
we provided AirShaper with the test data of 2 different valves at
various flow rates. We wanted to know the influence of different
parameters on the Cv values, so that we could select the right setup for
us to analyze other valves in the future.

#### Valves types

We analyzed two valve types for this benchmark.

- Sample A -- straight flow passage: this valve features very little
 obstruction on the inside. As such, the flow resistance is mainly
 determined by what happens in the boundary layer.

<ImageWithCaption
 src="/assets/images/ksvalve_case_2.jpg"
 alt=""
 width="521"
 height="201"
 positionCenter={true}
/>

- Sample B -- non-straight flow passage: this valve features a
 contraction as well as a more pronounced obstruction of the flow. As
 such, the flow resistance will be higher and more dependent on flow
 separation around these geometries.

<ImageWithCaption
 src="/assets/images/ksvalve_case_3.jpg"
 alt=""
 width="514"
 height="207"
 positionCenter={true}
/>

### CFD -- computational fluid dynamics

We added upstream & downstream pipes to allow the velocity profile to
develop before & after the valve. The pressure measurements were
performed at the start & end of these additional pipes. We added a flow
rate constraint at the inlet, and a 0-pressure constraint at the outlet.

<ImageWithCaption
 src="/assets/images/ksvalve_case_4.jpg"
 alt=""
 width="497"
 height="280"
 positionCenter={true}
/>

The results included the flow rate & pressure drop across the entire
system. This allowed the AirShaper software to calculate the pressure
difference, which we needed to calculate the Cv values.

<ImageWithCaption
 src="/assets/images/ksvalve_case_5.jpg"
 alt=""
 width="487"
 height="279"
 positionCenter={true}
/>

## Parameter sensitivity

### Layers

The addition of prism layers to the mesh provides a more realistic
approach to capturing the boundary layer.

- Sample A: The addition of layers leads to a big improvement,
 reducing the variations between the test data and the simulation to
 just 0.9% at the highest flow rate.

<ImageWithCaption
 src="/assets/images/ksvalve_case_6.jpg"
 alt=""
 width="483"
 height="272"
 positionCenter={true}
/>

- Sample B: the addition of layers again had a positive effect,
 although less pronounced. This is expected, as the impact of the
 boundary layer is relatively low due to the geometrical
 obstructions.

<ImageWithCaption
 src="/assets/images/ksvalve_case_7.jpg"
 alt=""
 width="483"
 height="272"
 positionCenter={true}
/>

### Surface roughness

Just like the prism layers, the surface roughness also impacts the
boundary layer and the friction on the wall. Although the internal
surface of the valve is quite smooth, we added roughness to see how
strong the effect would be (the analyses were done for 1 flow rate
only).

- Sample A: here too, sample A is quite affected by the roughness due
 to the relative importance of the boundary layer. Increasing the
 surface roughness by too much, lead to a larger variation again.

<ImageWithCaption
 src="/assets/images/ksvalve_case_8.jpg"
 alt=""
 width="483"
 height="272"
 positionCenter={true}
/>

- Sample B: as expected, the impact of surface roughness is less
 pronounced:

<ImageWithCaption
 src="/assets/images/ksvalve_case_9.jpg"
 alt=""
 width="483"
 height="272"
 positionCenter={true}
/>

### Turbulence intensity

The turbulence intensity at the inlet was increased, but no noticeable
change in Cv was observed.

<ImageWithCaption
 src="/assets/images/ksvalve_case_10.jpg"
 alt=""
 width="483"
 height="272"
 positionCenter={true}
/>

## Conclusion

By studying the sensitivity of different parameters, we concluded that
the Basic AirShaper simulations with layers, without roughness, were the
best compromise to obtain reliable values for sample A and sample B. For
sample A, this resulted in less than 1% deviation. For sample B, the
deviation was around 10%.

In general, the simulated values are close to the actual test values.
Although the deviation is slightly higher for sample B, this method is a
far more accurate method than our one dimensional calculation tool.
AirShaper has provided us with an efficient, accurate and economical way
to obtain Cv values through simulation.

Kong Jih Valve performed valve pressure drop tests and compared the results to AirShaper internal flow simulations.

Valve pressure drop test

## About the author

Tomáš Hájek holds a PhD in Aerospace, Aeronautical and Astronautical Engineering from the Brno University of Technology. He previously conducted wind tunnel tests, designed & tested ballistic recovery systems for aircraft, and is a fan of 3D printing and UAVs.

At Mejzlik Propellers, he heads the analysis and testing team, and that includes selecting & benchmarking new simulation & testing methods.

## Wind tunnel test data

To validate AirShaper, we used wind tunnel test data for one of our existing propellers, the <ExternalLink href="https://shop.mejzlik.eu/propeller-21x13-ccw-2b-e-l/">21x13 CCW 2B E L</ExternalLink>. It has a diameter of 21” and a pitch of 13”, so it’s a fairly large propeller. Yet, it weighs under 40g.

<ImageWithCaption
 src="/assets/images/mejzlik_case_1.jpg"
 alt=""
 width="560"
 height="420"
 positionCenter={true}
/>

The wind tunnel experiment was conducted in facilities at the University of Defence, Brno. The propeller was mounted onto a test stand in a puller configuration. A detailed description of the wind tunnel, test stand, and method is shown in an article of Rozehnal [1]. We then measured the power, torque & thrust while we increased the incoming wind speed from 0 to 35m/s (which translates into an advance ratio of up to 0.8). The propeller performed at a steady speed of 4900 RPM.

## In-house simulation tools

Together with the Czech government & partners, we have developed our simulation tool to predict propeller performance. Our tool is based on methods that do not require heavy computing. It combines Lifting line theory, Blade Element Momentum (BEM) Theory, and Vortex theory. The tool is extremely fast. But comes with the general limitations of such methods, which struggle to accurately capture blade tip losses, cope with non-axial flows, etc.

But because we design all of our propellers in-house in 3D software, the step towards CFD simulations on AirShaper was an easy one for us.

<ImageWithCaption
 src="/assets/images/mejzlik_case_2.jpg"
 alt=""
 width="800"
 height="350"
 positionCenter={true}
/>

## AirShaper simulations

We used AirShaper to run the same sweep of CFD (computational fluid dynamics) simulations:
 - Constant RPM of 4900
 - Incoming wind speed varying between 0m/s and 60 m/s (advance ratio up to 1.38).
 - 2 different accuracy levels: Basic (1 million cells) and Regular (10 million cells)
 - Without prism layers and with prism layers

The entire sequence (domain sizing, automated & adaptive meshing, ...) is automated, so the procedure was identical for all simulations. That allowed for a perfect one-to-one comparison of the results.

<VideoEmbed
 poster="/assets/images/mejzlik_validation_thumbnail.jpg"
 videos={[
 {
 src: 'https://airshaper.storage.googleapis.com/website/video/mejzlik.av1.mp4',
 type: 'video/mp4; codecs=av01.0.05M.08,opus',
 },
 {
 src: 'https://airshaper.storage.googleapis.com/website/video/mejzlik.hevc.mp4',
 type: 'video/mp4; codecs=hvc1',
 },
 {
 src: 'https://airshaper.storage.googleapis.com/website/video/mejzlik.h264.mp4',
 type: 'video/mp4; codecs=avc1.4D401E,mp4a.40.2',
 },
 ]}
/>

## Results

The most relevant data points have been summarized under the "Data" section at the end of this article. Below, the trends for the thrust & power coefficients as well as the efficiency are analyzed.

### Thrust coefficient

All results are quite close to each other for the thrust coefficient. In this context, close means that for example, at an advance ratio of 0.4 all predicted values are within 12% of the measurement. The BEM method features the largest deviation. The Basic simulation with layers underpredicts the thrust at low advance ratios. Overall, the Regular simulation with layers features the best correlation with the wind tunnel data with a near-exact match up to an advance ratio of 0.6. All methods capture the correct trend.

The wind tunnel data indicates that the propeller starts windmilling (where the wind starts to drive the propeller instead of the other way around, generating a negative thrust) at an advance ratio J just below 0.8. The simulations predict this to happen at an advance ratio of around 0.9.

<ImageWithCaption
 src="/assets/images/mejzlik_case_4.png"
 alt=""
 width="701"
 height="400"
 positionCenter={true}
/>


### Power coefficient

All simulation methods predict a higher power coefficient than the measured wind tunnel values. In the context of the power coefficient, values are within 16% of the measurement. The BEM method (black curve) overpredicts the most, closely followed by the Basic simulation without layers (with layers, the Basic simulation gets a little closer). All methods capture the correct trend.

At higher accuracies (the Regular simulations) the results correspond best, especially when layers are applied to the simulation – this results in a near-exact match. Beyond an advance ratio of 0.6, the simulation results (both BEM and CFD) start to diverge from the wind tunnel data.

<ImageWithCaption
 src="/assets/images/mejzlik_case_5.png"
 alt=""
 width="701"
 height="400"
 positionCenter={true}
/>

### Efficiency

For advance ratios below 0.6, the BEM method corresponds well with the wind tunnel data. The Basic simulation with layers underpredicts the efficiency. The Regular simulation with layers shows a good correspondence up to an advance ratio of 0.7. All methods capture the correct trend.

<ImageWithCaption
 src="/assets/images/mejzlik_case_6.png"
 alt=""
 width="701"
 height="400"
 positionCenter={true}
/>


## Wind tunnel effects

For this test, the propeller was placed in an open test section, see the image below. This helps to reduce the blockage effect (see this video), as the pressures before and after the propeller disk can equalize with the atmosphere. According to Glauert [2], unless the propeller disk is bigger than 70% of the area of the cross-section, no correction is required.

<ImageWithCaption
 src="/assets/images/mejzlik_case_7.jpg"
 alt=""
 width="179"
 height="119"
 positionCenter={true}
/>

In the case of a propeller rotating at an RPM close to the “windmilling” point, however, it will start acting like a disc blocking the flow. In that case, the air needs to curve around the propeller and fit between the propeller and the opening of the wind tunnel. This leads to a pressure build-up ahead of the propeller, reducing the thrust it can generate.

The simulation data doesn’t suffer from this blockage effect, and thus at advance ratios close to the windmilling point, the results start to diverge. The solution to this could be to use a larger wind tunnel, place the propeller further downstream of the opening, etc.


## Conclusion

We’ve seen that for advance ratios where the blockage factor doesn’t play a significant role (below the advance ratio of 0.6), all methods show a good overall correlation. As expected, the CFD methods outperformed the BEM method in accuracy, at the cost of a higher computational effort.

Within the CFD methods, the Regular simulations (10 million cells) with layers even provide a near-exact match between simulation and wind tunnel test data across the entire advance ratio range of 0.0-0.6. It is expected that in a properly sized wind tunnel, the correlation would further improve even for advanced ratios beyond 0.6.

The major benefit Mejzlik company sees in the CFD method is that it can be applied to full UAVs, including multiple propellers and their airframe. The interaction between propellers can be studied, the effect of prop wash onto the airframe, etc. This allows for designers to select the right propeller with the right wake structure for example, to improve the interaction with other propellers and the airframe. To us at Mejzlik, this allows us to advise our customers in much more detail on their overall design strategy, far beyond creating a better propeller. So we can provide a holistic aerodynamic design approach rather than simply providing them with propeller data.

## Data

| | Ct | | | |
| ------------------------------- | ------ |------- |------- |------- |
| Advance ratio J [-] | 0.2 | 0.4 | 0.6 | 0.8 |
| Mejzlik | 0.0994 | 0.0799 | 0.0510 | 0.0163 |
| AirShaper - Basic - wo Layes | 0.0957 | 0.0762 | 0.0493 | 0.0186 |
| AirShaper - Basic - w Layers | 0.0842 | 0.0686 | 0.0452 | 0.0174 |
| AirShaper - Regular - wo Layers | 0.0989 | 0.0780 | 0.0501 | 0.0185 |
| AirShaper - Regular - w Layers | 0.0925 | 0.0714 | 0.0425 | 0.0099 |
| Wind Tunnel | 0.0918 | 0.0711 | 0.0411 | -0.0035 |
| | | | | |
| Mejzlik | 108% | 112% | 124% | -459% |
| AirShaper - Basic - wo Layes | 104% | 107% | 120% | -523% |
| AirShaper - Basic - w Layers | 92% | 96% | 110% | -490% |
| AirShaper - Regular - wo Layers | 108% | 110% | 122% | -521% |
| AirShaper - Regular - w Layers | 101% | 100% | 103% | -280% |
| Wind Tunnel | 100% | 100% | 100% | 100% |


| | Cp | | | |
| ------------------------------- | ------ |------- |------- |------- |
| Advance ratio J [-] | 0.2 | 0.4 | 0.6 | 0.8 |
| Mejzlik | 0.0484 | 0.0485 | 0.0388 | 0.0204 |
| AirShaper - Basic - wo Layes | 0.0494 | 0.0465 | 0.0379 | 0.0247 |
| AirShaper - Basic - w Layers | 0.0467 | 0.0455 | 0.0384 | 0.0250 |
| AirShaper - Regular - wo Layers | 0.0467 | 0.0448 | 0.0363 | 0.0220 |
| AirShaper - Regular - w Layers | 0.0447 | 0.0425 | 0.0336 | 0.0196 |
| Wind Tunnel | 0.0417 | 0.0417 | 0.0321 | 0.0081 |
| | | | | |
| Mejzlik | 116% | 116% | 121% | 250% |
| AirShaper - Basic - wo Layes | 119% | 112% | 118% | 303% |
| AirShaper - Basic - w Layers | 112% | 109% | 120% | 307% |
| AirShaper - Regular - wo Layers | 112% | 108% | 113% | 271% |
| AirShaper - Regular - w Layers | 107% | 102% | 105% | 240% |
| Wind Tunnel | 100% | 100% | 100% | 100% |


| | Efficiency | | | |
| ------------------------------- | ---------- |------- |------- |------- |
| Advance ratio J [-]| 0.2 | 0.4 | 0.6 | 0.8 |
| Mejzlik | 0.4191 | 0.6637 | 0.7956 | 0.6777 |
| AirShaper - Basic - wo Layes | 0.3968 | 0.6520 | 0.7448 | 0.5545 |
| AirShaper - Basic - w Layers | 0.3643 | 0.6080 | 0.6627 | 0.5213 |
| AirShaper - Regular - wo Layers | 0.4375 | 0.6857 | 0.7981 | 0.6194 |
| AirShaper - Regular - w Layers | 0.4254 | 0.6574 | 0.7510 | 0.4967 |
| Wind Tunnel | 0.4452 | 0.6810 | 0.7745 | 0.2132 |
| | | | | |
| Mejzlik | 94% | 97% | 103% | 318% |
| AirShaper - Basic - wo Layes | 89% | 96% | 96% | 260% |
| AirShaper - Basic - w Layers | 82% | 89% | 86% | 244% |
| AirShaper - Regular - wo Layers | 98% | 101% | 103% | 290% |
| AirShaper - Regular - w Layers | 96% | 97% | 97% | 233% |
| Wind Tunnel | 100% | 100% | 100% | 100% |

## References

[1] ROZEHNAL, Dalibor, 2006. Experimental system for the small aircraft propeller testing. In: Engineering Mechanics 2006: Engineering Mechanics 2006. Book of extended abstacts. [online]. Svratka, Brno, Czech Republic: Institute of Theoretical and Applied Mechanics, Academy of Sciences of the Czech Republic [vid. 2024-05-20]. ISBN 80-86246-27-2. Dostupné z: https://www.engmech.cz/im/proceedings/show_p/2006/320

[2] GLAUERT, H., 1933. Wind Tunnel Interference on Wings, Bodies and Airscrews. In: [online]. [vid. 2023-03-15]. Dostupné z: https://www.semanticscholar.org/paper/Wind-Tunnel-Interference-on-Wings%2C-Bodies-and-Glauert/e1b4c6dc8ca7c627e7bae114821f45491fbb079c

Mejzlik benchmarked AirShaper simulations versus their in-house method and wind tunnel data.

Mejzlik Propeller Performance

## What is a standard model?

Aeronautical engineers have a long history of developing standardized models for wind tunnel calibrations and data comparisons between facilities. They are extremely useful in providing baseline datasets for correlation of results, data repeatability over time and verifying model installation or data acquisition systems. Reference models are also particularly relevant from the perspective of numerical analyses, where different codes can be directly compared with each other, or assumptions and solver settings can be experimented with to determine solution sensitivity to certain parameters. A standardized reference model typically fulfills two main criteria. Firstly, they are simplistic in shape with a precisely defined geometry, and secondly, they are representative of realistic configurations to ensure that the results are relevant. There are multiple examples of benchmark geometries which are available to aerodynamic engineers from various fields, including the Ahmed body for ground vehicles, the NASA CRM for commercial aircraft, the SDM for fourth generation fighters and the RFZ model, which is representative of modern launchers utilizing a re-usable first stage (similar to the Falcon 9).

<ImageWithCaption
 src="/assets/images/ssam_1.jpg"
 alt=""
 width="500"
 height="313"
 positionCenter={true}
/>

Because the idea of having standardized models is longstanding, it is often the case that most of these models and the accompanying experimental data are spread over multiple reports, spanning decades, typically appearing in the form of scanned copies in old repositories. This leaves the modern engineer, who is used to having digital copies of all required data, with the task of digitizing plots and regenerating the model geometry based on drawings using a CAD package. Not only is this a significant time investment, but it also introduces potential sources of error into either the model or the experimental datasets being copied, due to the nature of digitizing poor-quality images. These facts have been the main motivating factor in the generation of an open source aerodynamic model, where at the click of a button, CAD geometries and experimental data can be downloaded from a file sharing platform, bringing the idea of standard aerodynamic models into the 21st century.

## The SSAM-Gen5

The SSAM-Gen5 is a new test case for a fifth-generation fighter aircraft. The CAD, CFD grids and experimental data can be downloaded from the Zenodo page found in ref [5]. This model is based on the F-22 Raptor and was chosen over the F-35 as the majority of in service fifth-generation vehicles have a twin-engine configuration. The geometry was generated from drawings and images of the aircraft that are openly available in the public domain. As such, the geometry is not intended to reproduce the aerodynamic characteristics of the F-22 Raptor, but rather, to reflect the salient aerodynamic features typical of contemporary aircraft. Geometric details such as the chordwise camber and spanwise twist were omitted to allow the vehicle geometry to remain as simple as possible while still representing the complex outer mold line of a fifth-generation platform. The wing profiles are four percent thick biconvex sections with sharp leading edges and no incidence angle. The sweep angle of the horizontal stabilizer and the main delta wing is approximately 42° with the twin vertical stabilizers swept back approximately 25° and canted outwards at an angle of 27.5°. The thrust nozzles are retained, while the intakes are simplified and sealed. The full-scale length of the vehicle is 19 m.

<ImageWithCaption
 src="/assets/images/ssam_2.jpg"
 alt=""
 width="240"
 height="192"
 positionCenter={true}
/>

## Work conducted to date

Subsonic wind tunnel tests were performed at the University of Sydney, with force and moment balance measurements used to develop a database of experimental validation data for the platform at a freestream velocity of 20 m/s. Numerical simulations using a variety of turbulence models on a grid converged mesh were completed using the commercial CFD solver ANSYS FLUENT. Investigations show strong correlations between aerodynamic coefficients from the wind tunnel and CFD up to an angle of attack (AoA) of 15 degrees, indicating that the basic experimental and computational approaches are valid. Past this point, the flow field is dominated by unsteady effects due to separated flow and the majority of steady-state RANS models appear to underpredict the loads measured in the wind tunnel [6].

<ImageWithCaption
 src="/assets/images/ssam_3.jpg"
 alt=""
 width="400"
 height="400"
 positionCenter={true}
/>

## Airshaper Comparisons

To gain some additional perspective on the vehicle aerodynamics from a second solver, a series of “regular” simulations using the Airshaper platform were completed. The results for lift, drag, pitching moment and aerodynamic efficiency had an extremely good correlation with wind tunnel data and the existing CFD database.

<ImageWithCaption
 src="/assets/images/ssam_4.jpg"
 alt=""
 width="600"
 height="470"
 positionCenter={true}
/>

Analyzing the results, there are some clear aerodynamic characteristics which are typical of delta winged aircraft, such as the lack of stall behavior observed in the lift-coefficient plots. For aircraft with regular airfoils designed for subsonic flow (such as the NACA 4 or 5 series family), a lift coefficient peak is typically observed somewhere between 12 and 18 degrees AoA. As the AoA is increased past this point, severe flow separation occurs and causes the wing to stall. This manifests as a loss of lift and a negative lift curve slope. Contrary to this, the SSAM-Gen5 was tested up to 30 degrees AoA without showing any stalling behavior. This is due to the phenomenon of “vortex lift”, where the sharp, swept geometry of the leading edges promotes leading edge flow separation and vortex formation. Within the vortex cores, low pressure is observed and creates regions of suction along the wing. The reliance on vortices to generate lift, combined with the low aspect ratio wings results in a reduction in the lift-to-drag ratio, or aerodynamic efficiency, when compared with typical aircraft designed for subsonic operation. The final feature of note is the positive and non-linear gradient of the pitching moment coefficient, which indicates a statically unstable aircraft. An unstable aircraft tends to want to deviate from the trim point following a given control input, while a stable aircraft will want to return to the trim point. This fact can be easily explained using the ball analogy shown below, where any perturbation of the green ball will result in it resettling to its starting position, while the same action on the red ball will not.

<ImageWithCaption
 src="/assets/images/ssam_5.jpg"
 alt=""
 width="250"
 height="161"
 positionCenter={true}
/>

Commercial aircraft have strict design guidelines relating to static stability, however military aircraft often exploit unstable qualities to improve maneuverability. The positive pitching moment behavior suggests a nose up tendency of the vehicle and can be easily explained by looking at the surface pressure contours using the Airshaper online post-processing tool. The strength of the vortices increase with AoA, meaning that with every AoA increment, the centre of pressure (CoP) is shifting to a new location. For this vehicle, the regions of lowest pressure tend to appear towards the front of the vehicle (vortices formed at the leading-edge wing extensions/engine intakes), giving a CoP which is always located ahead of the center of gravity (CoG). The CoG has been estimated to be at about 35% of the mean aerodynamic chord.

<ImageWithCaption
 src="/assets/images/ssam_6.jpg"
 alt=""
 width="876"
 height="210"
 positionCenter={true}
/>

Between 15 and 20 AoA the formation of vortices along the nose are observed and can be seen in the image below for the 30 degree AoA case. Here it is clear that the systems of vortices result in an extremely complex flow field.

<ImageWithCaption
 src="/assets/images/ssam_7.jpg"
 alt=""
 width="350"
 height="277"
 positionCenter={true}
/>

Until now, the influence of vortices has been discussed only within the context of global aerodynamic coefficients, however it is important to understand that loads induced on the airframe by these swirling flows can cause issues. A classic case of this was earlier variants of the F-18 which saw burst vortices interacting with the fins causing buffeting (see image below from ref [7]). Because the frequency of the burst coincided with the first bending mode of the fin, significant structural damage was seen and some fleets were grounded. This problem was fixed through the addition of a fence on the leading-edge extension, creating a secondary vortex to interact with the primary vortex, strengthening its core and preventing it from bursting. This highlights the importance of understanding vortex structures, their interactions with other flow features, as well as the ability to accurately predict their paths.

<ImageWithCaption
 src="/assets/images/ssam_8.jpg"
 alt=""
 width="350"
 height="277"
 positionCenter={true}
/>

Moving back to the SSAM-Gen 5 results, there is a large spread in the CFD data when compared to the experimental data at high AoA. While the presented data has been corrected for blockage, there is some uncertainty about the accuracy of correction factors relating to non-traditional type aircraft, and this warrants further investigation. However, as strong agreement for lift and drag was observed across the AoA sweep, it is expected that the inherent sensitivity of pitching moment data, which relies heavily on both accurately predicting the total normal and axial forces as well as the locations where these forces are acting, plays a large role. It is well known that different RANS turbulence models are susceptible to large differences in modelling separated flows. Factors such as the size and shape of vortices as well as the prediction of vortex bursting is particularly critical. For the pitching moment coefficient, it was observed that there is a positively shifted curve in the experimental data at 0 degrees AoA. The gathered CFD results show unanimously that there is no pitching moment induced at this condition, meaning the drag of the fins above the CoG is offset by the drag of the sealed intakes below the CoG. It is suspected that there are some interference effects with the mounting strut in the wind tunnel that are causing this discrepency. Interestingly experiments for the SDM, a generic F-16 geometry, saw similar trends at the same condition where 0 degree AoA pitching moment predictions vary between -0.02 and 0.02 (see Figure 16 in ref [8]).

## Future work

Due to the open source nature of this project, the opportunities for fellow researchers, students or hobbyists to complete their own aerodynamic investigations and contribute to the knowledge base surrounding fifth generation fighter aircraft is now easily achievable. Recommendations for further investigations stemming from this work include the evaluation of blockage effects at high AoA and interference from the model mounting system. Other areas of interest include the of the influence of Reynolds number variation on the aerodynamic coefficients, the investigation of unsteady computational methods, as well as the updating of the wind tunnel model to allow surface pressure tapping, facilitating direct comparisons on the vortex location with RANS data. General recommendations for further avenues of investigation include the generation of lateral-directional, control and dynamic derivatives. Experimental data for validation purposes outside of the subsonic incompressible flow regime and at Reynolds numbers more closely representative of full-scale in-flight conditions is highly desirable but relies on collaborations with other research institutions.

#### References


[1] The RFZ Model - A standard model for the investigation of aerodynamic and aerothermal loads on a re-usable launch vehicle, https://zenodo.org/communities/rfz-model/?page=1&size=20

[2] NASA CRM homepage https://commonresearchmodel.larc.nasa.gov/?doing_wp_cron=1694606395.0442950725555419921875

[3] Huang, X. Wing and Fin Buffet on the Standard Dynamics Model; Technical Report Defense
Technical Information Center Compilation Part Notice ADPOI0722; Field Technology Inc.: Long Beach, CA, USA, 2000.

[4] Ahmed SR, Ramm G, Faltin G. Some salient features of the time-averaged ground vehicle wake. SAE Technical Papers ed. Warrendale: SAE International, 1984.

[5] SSAM-Gen5, https://zenodo.org/communities/ssam_gen5/?page=1&size=20

[6] Giannelis, N.F.; Bykerk, T.; Vio, G.A. A Generic Model for Benchmark Aerodynamic Analysis of Fifth-Generation High-Performance Aircraft. Aerospace 2023, 10, 746.
https://doi.org/10.3390/aerospace10090746

[7] F-18 Leading Edge Extension Fences, https://aerospaceweb.org/question/planes/q0176.shtml

[8] Erm, L., An Experimental Investigation Into the Feasibility of Measuring Static and Dynamic Aerodynamic Derivatives in the DSTO Water Tunnel, Defence Science and Technology Organisation, 2013, https://apps.dtic.mil/sti/pdfs/ADA588964.pdf

SSAM-Gen5 – an open source test case for a fifth-generation fighter - comparing AirShaper to wind tunnel data

Fighter Jet Test

## Introduction

At Wingtra, we develop autonomous surveying drones which are used for surveying, mining, agriculture, environmental monitoring and many more use cases. I am working in a 50/50 split between being lead-engineer and leading the Mechanics & Aerodynamics Team - being responsible for research and new developments.

## Aerodynamic performance

Our drone is a fixed-wing VTOL drone. That means it can take off and land vertically using the vertical lift of the propellers. But for forward flight, it will transition into horizontal mode, allowing the fixed wing to generate lift. In that scenario, the propellers no longer provide lift but mostly the thrust required to overcome aerodynamic drag. This makes it much more efficient compared to a conventional drone which constantly uses the propellers’ lift to keep it in the air.

<YouTubeEmbed
 videoId="Jq-yUvR0kpw"
 aspectWidth={4}
 aspectHeight={3}
 imgSize="hqdefault"
 size="md"
/>

At the beginning of a new project or iteration, we have a lot of variables to take into account: payload capacity (the weight a drone can carry, mostly camera equipment in our case), cost, support points for the vertical landing & take off, cruise speed, wingspan and so on.

That means we are looking at a lot of possible variations and it isn’t easy to determine a clearly preferred concept. In the definition phase of product development we weigh certain features or properties and then go through a score-card approach. Through the use of simulations, we see which one comes closest in terms of aerodynamic performance and flight behavior.

## Aerodynamic Design Tools

Several years ago, we had given the use of 3D CFD (computational fluid dynamics) tools a chance. When we later went to the wind tunnel for testing and verification, the correlation wasn’t very good. We kept an eye on the evolution in terms of CFD software, but until fairly recently, restricted ourselves to the use of fast but rudimentary tools like XFLR5 for the bulk of the initial conceptual work. Recently, we decided to re-evaluate CFD and work with AirShaper.

<ImageWithCaption
 src="/assets/images/wingtra-windtunnel.jpg"
 alt="Pictures of the previous wind tunnel test on the old Wingtra design."
 width="800"
 height="300"
 positionCenter={true}
/>

## Concept exploration

With strong goals in terms of both lifting capacity (requiring high absolute lift values) and lift-over-drag ratio (the aerodynamic efficiency: maximize lift and minimize aerodynamic drag) and many other requirements usually given by the product team the concept phase is featuring difficult trade-offs. A larger wingspan means more lift capacity and a higher efficiency, but also higher bending loads and difficulties in transportation. For this explorational concept phase the ease of a browser based simulation is great.

It is a very beautiful balancing act where we try to combine the function of different elements, like optimizing the wing tips for both aerodynamic performance and compactness for transportability.

It's only by running high numbers of simulations that you really start to see trends and find areas for improvement. In the past, this meant spending hours setting up a simulation, as it involved cleaning the CAD geometry (to make it watertight), creating the mesh (and manually improving it), processing the results into visuals and so on. With AirShaper, the process is automated, so we could simply launch dozens of design concepts at the same time in a matter of minutes and look at the results just hours later.

<ImageWithCaption
 src="/assets/images/wingtra-flowlines.jpg"
 alt="Example of the AirShaper simulation output (disclaimer: generic image)"
 width="602"
 height="295"
 positionCenter={true}
/>

## Flight testing

Design tools are highly useful, but in the end, the real drone flying through real air is still the ultimate test. So we literally have a dedicated team of multiple people flying the drones day & night to validate their performance, UX, reliability and many other factors out there in the field. We call the team the continuous testing (CT) team.

## Wind tunnel testing

One downside of flight testing though, is the limited control over the conditions and limited measurement data. In a wind tunnel, it’s much easier to position the drone at fixed angles of attack, vary the speed and other parameters in a very controlled way. At the same time, by mounting the drone on a measuring balance, we get highly precise insights into the lift & drag forces and the pitch, roll and yaw moments on the drone. This allows us to precisely map the aerodynamic properties of the drone, which are then used as input for our autonomous flight controller.

The major aspect of wind tunnel testing is that it allows you to verify your simulations. After all, the better the correlation the more you can rely on simulations for further design work.
To do so, we had to consider and calculate air densities, blockage factors and compensate for the influence of the mounting structure (the parts connecting the drone to the balance also generate drag & lift!).

Once we had done this, we were truly satisfied to see a very good correlation between the wind tunnel data and the CFD simulations:

<ImageWithCaption
 src="/assets/images/wingtra-drag-curve.jpg"
 width="369"
 height="278"
 positionCenter={true}
/>

<ImageWithCaption
 src="/assets/images/wingtra-lift-curve.jpg"
 width="369"
 height="278"
 positionCenter={true}
/>

Neither wind tunnel nor CFD are perfect: prototypes can (and will) have non-perfect surface finishing and there are balance measurements errors. Similarly simulations have uncertainty with modeling, discretization and iteration errors. But when both are performed correctly, they both feature a small error band and the correlation between them becomes apparent as visible in the graphs.

## Further work

Seeing this correlation, we have great confidence the drone will perform as projected. CFD has become an important part in our toolbox and we’ll use it with confidence for this and future projects. As the computing power & functionalities of CFD increase, we’ll surf the wave and benefit from it. It will not replace wind tunnel testing or flight testing all together but for sure its relative importance will grow and will help us to speed up developments. At our peak, we run over 100 simulations per person per month!

AirShaper customer Wingtra validated the AirShaper lift and drag characteristics through wind tunnel testing.

Wingtra fixed wind drone

## Why validate software?

Computer Aided Engineering (CAE) is a constantly growing force in the modern world. The ability to simulate and experiment virtually can save significant resources, time and money. Consequently, many industries have shifted their focus to utilising Computational Fluid Dynamics (CFD) software which is allowing engineers to conduct more detailed aerodynamic analysis and is even replacing wind tunnel testing in some cases.

However, as effective as these simulations are, the results must be validated to ensure they are accurate. Correlating results to real world data is a key part of this, but it is also important to benchmark software against existing techniques. This is what DARcorporation recently completed with AirShaper software.

---

## Using CFD in Aeronautical engineering

DARcorporation is an aeronautical engineering consultancy
based in Lawrence, Kansas. Specialising in aircraft design, they use proprietary
software in conjunction with tools from external suppliers.

<ImageWithCaption
 src="/assets/images/air-vy-400-commuter-aircraft.jpeg"
 alt="Air Vy 400 commuter aircraft concept in flight"
 caption={
 
 DARcorporation worked with Transcend on the Air Vy 400 commuter aircraft
 concept. CREDIT:{' '}
 <ExternalLink href="https://www.transcend.aero/">
 Transcend Air
 </ExternalLink>
 
 }
 width="700"
 height="394"
 positionCenter={true}
/>

Much of their work is in the conceptual and prototype stages of projects ranging from Unmanned Aerial Vehicles (UAVs) to multi-engine jet aircraft. Therefore, utilising accurate simulation tools is vital to the success of these projects as the majority of work takes place before any physical components or aircraft exist. Consequently, the engineers at DARcorporation use CFD to feed results into structural studies and control surface design.

---

## Benchmarking AirShaper

Alongside their own Advanced Aircraft Analysis (AAA) software, DARcorporation has extensive experience of a range of simulation tools including <ExternalLink href="https://www.darcorp.com/flightstream-aerodynamics-software/">FlightStream</ExternalLink>, <ExternalLink href="https://www.plm.automation.siemens.com/global/en/products/simcenter/STAR-CCM.html">STAR-CCM+</ExternalLink> and AirShaper. Flightstream is a surface vorticity-based simulation package that offers speed and convenience after meshing, but results in a less detailed simulation. STAR-CCM+ is a fully fledged CFD package which comes with the traditional overheads of meshing, long run times and post-processing requirements. Whereas, AirShaper achieves accurate CFD simulations quickly, whilst automating much of the set-up, so you don’t have to be an expert to benefit from using CFD.

<ImageWithCaption
 src="/assets/images/cessna-c210j-centurion-streamlines-side.jpeg"
 alt="CFD image of Cessna C210 Centurion showing coloured velocity streamlines and surface static pressure"
 caption={
 
 DARcorporation use a variety of simulation tools to carry out aircraft
 design studies. This{' '}
 <ExternalLink href="https://www.darcorp.com/projects/cessna-210j-centurion/">
 Cessna C210 Centurion CFD analysis
 </ExternalLink>
 shows velocity streamlines and surface static pressure. CREDIT: <ExternalLink href="https://www.darcorp.com/about-dar-corporation/">DARcorporation</ExternalLink>
 
 }
 width="700"
 height="258"
 positionCenter={true}
/>

DARcorporation wanted a CFD package that achieves high accuracy in short runtimes. This would allow them to generate aerodynamically efficient designs quicker, increasing the number of test cases they can accomplish. To identify the CFD software that best suited their needs, DARcorpoation conducted a validation study where they compared FlightStream, STAR-CMM+ and AirShaper against real-world wind tunnel data.

Real-world testing is expensive and slow, but benefits from using an actual aircraft or model. Whereas, simulations often have to be simplified which can lead to assumptions that ultimately affect accuracy. These errors cannot always be resolved, but a thorough knowledge of them means they can be accounted for in the design process.

It would be too impractical to conduct a bespoke wind tunnel test for this validation study, so another data source was needed instead. As discussed in our <Link href="/blog/best-aerodynamic-resources">Aerodynamic resources blog</Link>, NASA offers a wide range of open-source aerodynamic studies and data. One such study was carried out using a Cessna 210 Centurion light passenger aircraft in the Langley full scale wind tunnel. The Cessna has a relatively simple high-wing layout, common to many light aircraft, making it a suitable test case without being overly complicated.

- <ExternalLink href="https://ntrs.nasa.gov/api/citations/19730016341/downloads/19730016341.pdf">
 Read the original NASA wind tunnel report for the Cessna 210
 </ExternalLink>

<ImageWithCaption
 src="/assets/images/cessna-c210j-centurion.jpeg"
 alt="A white and red Cessna C210J Centurion flying over countryside"
 caption={
 
 The high-wing layout of the Cessna 210 Centurion made it a suitable test
 case for the validation study. CREDIT:{' '}
 <ExternalLink href="https://jackfleetwood.com/">
 Jack Fleetwood
 </ExternalLink>
 
 }
 width="700"
 height="326"
 positionCenter={true}
/>

---

## The study

Flightstream and Star-CCM+ both require the user to mesh a 3D model and carry out a meshing convergence study. This ensures that the mesh itself does not unduly influence the results. AirShaper on the other hand, automates this process when a model is uploaded, saving time and effort.

After running simulation sweeps at different airspeeds and angles of attack, the pressure or vortex (in the case of Flightstream) data is post-processed. This turns detailed node information into useful aerodynamic data. Three standard coefficients were chosen for analysis for the validation study: lift, pitching moment and drag.

---

## Lift coefficient results

As expected, the various simulation packages gave a spread of results. It’s also important to note that the wind tunnel results will differ slightly from reality due to the limitations of full-sized testing.

<ImageWithCaption
 src="/assets/images/airshaper-validation-figure-2-1-lift-coefficient.png"
 alt="Line graph showing the lift coefficient for different angles of attack for the three simulation packages and the reference wind tunnel data"
 caption={
 
 Lift coefficient vs angle of attack results for each CFD software and the
 wind tunnel reference data. CREDIT:{' '}
 <ExternalLink href="https://www.darcorp.com/about-dar-corporation/">
 DARcorporation
 </ExternalLink>
 
 }
 width="700"
 height="431"
 positionCenter={true}
/>

For the purposes of this study, the range of angles of attack up to around 10 degrees are of the most interest. Analysing the lift coefficients, we can see that all three softwares match the lift gradient well, but have differing offsets when compared to the wind tunnel data. FlightStream and AirShaper agree quite closely but Star-CCM+ has a further offset. The specific reason for this difference requires further study. But overall, the results for AirShaper are in the range of the two software packages that DARcorporation knows well.

At higher angles of attack, the software packages predict an earlier onset of stall compared to the measurement data. Software inaccuracies could be a cause (but both StarCCM and AirShaper predict this). Another cause could be geometry which was present on the real airplane (small seams or vortex generators which can delay stall) but not on the 3D model. As the NASA data is quite old, there is unfortunately no way of finding out.

---

## Pitching moment coefficient results

The pitching moment coefficient results follow a similar form to the lift coefficients. The slopes of all three simulations match quite closely to the wind tunnel data, with minimal offsets.

<ImageWithCaption
 src="/assets/images/airshaper-validation-figure-2-2-pitching-moment-coefficient.png"
 alt="Line graph showing the pitching moment coefficient for different angles of attack for the three simulation packages and the reference wind tunnel data"
 caption={
 
 Pitching moment coefficient vs angle of attack results for each CFD
 software and the wind tunnel reference data. CREDIT:{' '}
 <ExternalLink href="https://www.darcorp.com/about-dar-corporation/">
 DARcorporation
 </ExternalLink>
 
 }
 width="700"
 height="423"
 positionCenter={true}
/>

---

## Drag coefficient results

The drag results show a larger set of offsets between simulation and the wind tunnel, with all three CFD packages resulting in less drag than the wind tunnel data. This is not unexpected as the model in the wind tunnel has more imperfections and limitations compared to a perfect 3D CAD model. However, for this coefficient, AirShaper and Star-CCM+ are a close match.

<ImageWithCaption
 src="/assets/images/airshaper-validation-figure-2-3-drag-coefficient.png"
 alt="Line graph showing the drag coefficient for different angles of attack for the three simulation packages and the reference wind tunnel data"
 caption={
 
 Drag coefficient vs angle of attack results for each CFD software and the
 wind tunnel reference data. CREDIT:{' '}
 <ExternalLink href="https://www.darcorp.com/about-dar-corporation/">
 DARcorporation
 </ExternalLink>
 
 }
 width="700"
 height="424"
 positionCenter={true}
/>

---

## Reaching a conclusion

Experience tells us that it is not unusual to have four different answers for the same test. Quantifying and explaining these differences further are the next step in such an exercise. However, for DARcorporation, the benchmarking process proved that AirShaper produces results within the range they would expect from their normal tools, whilst being easier to use and set-up.

‘AirShaper takes a lot of effort out of the hands of an engineer,’ concludes Dennis van Dommelen, of DARcorporation. ‘It is simple and very straightforward to use. The uploaded aircraft does not require any modifications and after setting the right angles and airspeed, it is ready to run. Elimination of the mesh generation step is a major time saving and the accuracy of the standard run settings is comparable to other tools that are available.’

---

##### Interesting links:

 

<ExternalLink href="https://www.darcorp.com/statement-of-work-aircraft/">
 DARcorporation’s work flow for aircraft conceptual design
</ExternalLink>

 

<Link href="/blog/avl-racing-benchmarking-airshaper">
 AirShaper validation study with AVL Racing
</Link>

 

<Link href="/cases/optimization/cessna">
 Optimizing the aerodynamics of the Cessna 210 winglet
</Link>

 

<ExternalLink href="https://www.aviationconsumer.com/aircraftreviews/aircraftstepups/cessna-210-centurion-2/">
 A review of a Cessna 210 Centurion
</ExternalLink>

 

<ExternalLink href="https://www.darcorp.com/projects/transcend-air-corporation/">
 The design and analysis of the Transcend Air Vy 400 VTOL aircraft project
</ExternalLink>

---

As an independent 3rd party, DAR Corporation benchmarked AirShaper against StarCCM+, Flightstream and NASA wind tunnel data.

Cessna 210

## About the author

Michael Peinsitt is Skill Team Leader RACING at <ExternalLink href="https://www.avl.com/">AVL</ExternalLink>, a company which has been involved in racing for almost 25 years with their in-house developed vehicle dynamics simulation tool AVL VSMTM RACE. Inside VSM, all components of the car are modelled and characterized, such as the suspension, the tyres, the engine, and many more. One area is particularly important towards achieving a good correlation between simulation and reality: aerodynamics. Modern race cars produce quite a lot of downforce, which strongly influences performance. And with downforce significantly varying with ride height, it’s important to map it properly in function of the car’s setup.

## Mapping Aerodynamic Properties

When it comes to aerodynamics, the main challenge is to map downforce & drag in function of changes in input parameters. On top of the ride height, they also vary in function of the chassis slip & roll angle, the steering angle of the front wheels and so on. Grouping all of these correlations (or “sensitivities”) between aero coefficients and car setup results in what we call an <Link href="/videos/race-car-aerodynamics-3-aero-mapping/u-GyTRblNmc">Aero Map </Link> – a dataset of properties that can be used before & during the race weekend to setup & adjust the car in function of the track & the driver’s feedback.

## The Options

One option to obtain these sensitivities is **wind tunnel testing**, but this is typically expensive and requires quite some effort. Additionally, rules in certain race categories prevent teams to use wind tunnels.

The second option is to perform **aero measurements on the racetrack**. Apart from the efforts to equip the car with the required measurement equipment, track testing is always subject to varying external conditions. Changes in for example speed and orientation of the wind, or air temperature, make it difficult to compare results from different runs. Plus, in similar fashion as for the wind tunnel testing, in many race categories the rules ban, or severely restrict track testing.

A third option is to use CFD calculations. To execute a **“traditional” CFD workflow** is also time consuming, as the preparation and the meshing of the CAD model usually requires a certain amount of manual effort, plus the execution of the calculation itself needs considerable computing power to be conducted. Once done, post-processing is needed to convert raw simulation data into meaningful figures & visuals.

To capture the influence of the various input parameters on aero performance, a significant number of runs is needed – which means it takes time until the results are available. With preparation time always being a premium in racing, having to wait for results can seriously reduce the competitiveness of a team.

## Automating the simulation process

We turned to AirShaper to investigate the possibility of reducing or even eliminating many of the existing CFD challenges.

<VideoEmbed
 videos={[
 {
 src: 'https://airshaper.storage.googleapis.com/website/blog/videos/avl-setup.mp4',
 type: 'video/mp4',
 },
 ]}
/>

First of all, the need for “CAD-cleaning” – closing all gaps & holes in the 3D model of the car – can take a significant amount of time. AirShaper, however, runs directly on non-watertight 3D models, although in this particular example we had a watertight model available.

Second of all, we don’t need to worry about meshing, as this is entirely automated: the domain size, the initial mesh, adaptive mesh refinement, … are all included and, perhaps even more important, are repeatable. This means results between different models & setups are comparable.

Thirdly, we don’t need to manually post-process all the results. Frontal area, lift, downforce, aero balance, … are all post-processed automatically into a report. And using the online 3D result viewer, we can spot sources of drag, identify separation locations and so on. And if needed, we can always download the full results to do our own post-processing.

## Benchmarking AirShaper

To verify how realistic the output of these simulations would be, AVL conducted a simulation project with a race car that had already been analyzed using a conventional CFD approach (manual meshing, running, post-processing).

The model used for the simulation was a touring car provided by AVL RACING. The car was set up to drive in a straight line (chassis slip & roll angle as well as the steering angle were set to 0).

The exact same 3D model was provided to AirShaper and within 24 hours, the platform churned out CFD simulation results.

## The results

The first important comparison was a quantitative one, comparing the aerodynamic drag and lift coefficients. We were pleasantly surprised to see just a 5.2% difference in drag and an even smaller 3.6% difference in downforce.

<VideoEmbed
 videos={[
 {
 src: 'https://airshaper.storage.googleapis.com/website/blog/videos/avl-results-screen-capture.mp4',
 type: 'video/mp4',
 },
 ]}
/>

The second comparison was a qualitative one, comparing the surface patterns & flow structures in crucial areas on the car. A first example is the pressure map, which exhibits a near-identical pattern across both simulations.

<ImageWithCaption
 src="/assets/images/avl-airshaper-comparison-pressure-map.jpg"
 alt="Pressure map comparison between AVL and AirShaper"
 width="800"
 height="216"
 positionCenter={true}
/>

A second example is the local flow structure seen on a slice through the car, at 0.6m from the center. The way the flow enters the front wheel well, how it accelerates at the start of the diffuser, the velocity pattern around the rear wing, … all look highly similar.

<ImageWithCaption
 src="/assets/images/avl-airshaper-comparison-velocity-slice.jpg"
 alt="Velocity section comparison between AVL and AirShaper (0.6m from the centre)"
 width="800"
 height="207"
 positionCenter={true}
/>

## Conclusion

For us at AVL RACING, as an engineering & software provider in the racing industry, it is extremely important to first validate new software tools before we offer them to customers. Through this validation project, we have gained insight into the capabilities of AirShaper simulations to help speed up the aerodynamic simulations for our customers.
The good accuracy at a fraction of the cost & throughput time certainly makes AirShaper a viable option for future projects.

Interesting links:

<ExternalLink href="https://www.avl.com/">AVL</ExternalLink>
 

<ExternalLink href="https://airshaper.com/videos/race-car-aerodynamics/GApwSjKYUpQ">
 Race Car Aerodynamics
</ExternalLink>
 

<ExternalLink href="https://app.airshaper.com/simulations/new?accuracy=regular">
 Run Your Own Simulation
</ExternalLink>

AVL Racing compared automated AirShaper simulations to their conventional CFD software output.

AVL RACING

## About the Author 📢

Vincent Keromnes works at <ExternalLink href="https://www.a2mac1.com/">A2MAC1</ExternalLink>, a company specialized in benchmarking the performance of cars. Lots of cars. Vincent is in charge of what they call "Dynamic Benchmarking", covering Kinematic & Compliance testing, Energy Management Strategy, Battery Thermal management and so on. Recently, they added aerodynamics & aero acoustics to that list - the perfect timing for a blog article!

## Physical or Virtual Wind Tunnel? 🌬

We have several options to benchmark the aerodynamic performance of vehicles and we aim to get the best value out of each approach.

We analyze cars that are already available on the market. Running them through a physical wind tunnel test means you have the real shape (which can differ quite a lot from the digital mockup produced during development) in real wind conditions. But each wind tunnel has its own specific setup (attachment points, moving ground area, ...). This is ok when you're comparing the data with other vehicles tested in the same tunnel, but things get tricky when comparing with data from another tunnel or simulation data.

CFD (Computational Fluid Dynamics) simulations on the other hand can provide insights on digital models before the car is even built. Another advantage is the possibility to simulate real road conditions (without the limiations of a physical wind tunnel). But here too, comparability between different CFD methods is challenging, let alone the comparison to a wind tunnel test. The only solution for the latter is to include the full wind tunnel into the virtual CFD simulation.

In this blog, we'll compare the physical wind tunnel test with a CFD simulation applied on a high-quality exterior of the Tesla Model Y.

## Physical Wind Tunnel Setup ⚙

We took the Tesla Model Y to the <ExternalLink href="https://www.soufflerie2a.com/">S2A</ExternalLink> wind tunnel for a thorough airflow analysis. This tunnel has been used for many cars in the past and features a moving ground (although it is limited to the area between the wheels) and allows for the wheels to rotate.

<ImageWithCaption
 src="/assets/images/tesla-model-y-in-the-wind-tunnel.jpg"
 alt="The Tesla Model Y in the S2A Wind Tunnel"
 width="1438"
 height="959"
/>

Using smoke the flow pattern of the wind around the car was made visible. And using special techniques, the total pressure in a plane just 100mm behind the car was measured, giving good insight into the structure of the wake behind the car.

## The Virtual Wind Tunnel Setup 💻

To run a CFD simulation we needed a 3D model. In this case, a highly detailed scan of the exterior was made and then modified to increase realism: gaps between body panels were kept as geometrical discontinuity but were closed to avoid "false airflow" and the underhood flow was blocked by adding planes behind the grill. As for the aircurtains (small channels at the sides of the front bumper that provide a "shielding" layer of air across the front wheels), the inlets & outlets were part of the scan, but the duct connecting them to improve the flow was not (allowing for some air to escape "inside the car").

<ImageWithCaption
 src="/assets/images/tesla-model-y-scan-images.jpg"
 alt="3D Scan of the Tesla Model Y"
 width="1236"
 height="817"
/>

Once this was done, an AirShaper simulation was run. It also featured roating wheels and a moving floor (everywhere, not just between the wheels). The support arms used in for the wind tunnel test were not included.

<ImageWithCaption
 src="/assets/images/tesla-model-y-cfd.jpg"
 alt="CFD results for the Tesla Model Y"
 width="1298"
 height="594"
/>

## Results 🖇

We can't publish the full AirShaper report and the full S2A report (as these are documents available for purchase at either company). What I can share is the following:

CdA: the AirShaper simulation predicts a CdA which is 5-6% higher than the wind tunnel result. This can have many reaons, but in part this is because of the missing aircurtain ducts in the 3D model (increasing drag in the virtual wind tunnel simulation) and the narrow moving floor (reducing drag in the physical wind tunnel test).

Pressure distribution: below you can see the pressure distribution in a cut plane taken at 100mm behind the car. The global shape is very comparable, confirming the complementary & correlation between both techniques, despite their differences in setup & execution. The largest differences are close to the ground, where the physical wind tunnel (narrow moving ground) and virtual wind tunnel (full moving ground) feature a different approach.

<ImageWithCaption
 src="/assets/images/tesla-model-y-total-pressure.jpg"
 alt="Cut plot of the total pressure 100mm behind the Tesla Model Y"
 width="1928"
 height="649"
/>

## Cost 💲

Last but not least, I admit that the cost of doing a full vehicle scan and setting up a complete simulation manually yourself can cost as much as a physical wind tunnel test. By grouping all of these A2MAC1 scans & AirShaper simulations and offering the combined result to multiple users, the cost drops dramatically and understanding the aerodynamics of existing cars suddenly becomes much more accessible.

## Conclusion 💡

I will not enter into a debate discussing the difference between the two approaches - I've been trapped in this kind of discussion too many times during my career - it would never end 😏. The real value of vehicle performance analysis is not in the absolute values: benchmarking is about relative differences, with the goal of ranking the performance of different cars.

The most reliable approach is to use a robust protocol and a set of best practices applied in a systematic way, consistently using the same tools (whether they are physical or virtual).

<ExternalLink href="https://app.airshaper.com">
 Start your own simulation here
</ExternalLink>

Benchmarking company A2MAC1 took a Tesla Model Y to the wind tunnel and compared the results to AirShaper simulation data.

Validation casesBuilding trust by benchmarking AirShaper

Kong Jih ValveValve pressure drop test

Wind tunnel testMejzlik Propeller Performance

Wind tunnel testFighter Jet Test

Wind tunnel testWingtra fixed wind drone

NASA wind tunnel testCessna 210

CFD ComparisonAVL RACING

S2A wind tunnel testTesla Model Y

Trusted By