Wind tunnels - dumb aero question

[Lycoming]

For aero purposes I understand that we should consider the air to be an incompressible fluid. Whilst I (demonstrably!) don't know much about aero, I have done a considerable amount of rheology in the past and I guess what I'm getting at is that the air on the track surface could be considered to be static relative to the road surface. The air passing under the car body is moving at high speed relative to the car itself. In a fluid model this would be analogous to a shear stress being applied to the fluid, where one part of the system is stationary and the other is moving, creating a shear stress across the gap.

I don't know if the air (our fluid in this analogy) is shear sensitive, or behaves as a Newtonian fluid and what the flow condition is from static at the road surface, to fast moving air at the vehicle surface (which it should be if it is generating d/f via the Bernoulli principle.)... or is it static at both the track surface and the car boundary, but moving in between?

Air can generally be considered to be a newtonian fluid. The shear stress you mention is a consequence of viscosity. You should read up on boundary layers... basically, it's static relative to the road surface, and at the surface, static relative to the undertray of the car. You then have a gradient of increasing velocity relative to the car until you are far enough away, at which point it's basically the same as the free stream.

To move on to a different topic, a bigger issue with wind tunnels is modelling corners. A car going 'round a corner is not the same as a car in a sideslip condition. To properly simulate cornering, you should essentially put the car on a centrifuge and spin it around at some radius comparable to real world cornering radius (though in most corners, the radius of the path you track is not constant). Of course, doing this will not get you useful results because at meaningful speeds, you will constantly be running into your own wake and it will just become a mess. You can do it in CFD of course, but it requires a rather large mesh.

But people seem to have gotten along just fine for many decades now despite this.

[Greg Locock]

I don't know if 'just fine' really describes the state of the art.

I went to a seminar on the results of a recent round robin series of tests on most of the full size wind tunnels in Europe, and a couple in the States. These include a wide variety of designs, almost all were 3/4 open with one true closed loop. Some had one moving belt, some were static floor, some had 5 moving belts. They also had different methods of BL control.

Then they tested a whole bunch of production cars and a van, and made some mods to some of them. The improvement in CL values due to fitting a dam at the nose was HUGELY dependent on the exact tunnel configuration. They also ran each tunnel at different speeds, of course the 101 equation says that shouldn't affect Cl or Cd. it does.

Sadly I can't give any numbers as this is all internal to the round robin group (Eurpoean OEMS and a few stragglers) and proprietary, but I think the standout was that the estimates of Cd had a standard deviation of 0.01 at 140 kph. So using your pet tunnel, no names, gives you an advantage of 0.02 in aero bragging rights. Different companies also process the data in different ways, so the same car measured in the same windtunnel by two different regular users of that tunnel will give different numbers.

The interesting bit for me was comparing the hold down systems used, on a moving belt system the tire and springs allow the car to pitch, which affects the result vs speed, whereas the static floor cars always maintained attitude 'perfectly', but incorrectly.

[Holm86]

I think what he is talking about is that in a windtunnel the air blown at the car has a laminar flow. When a car is driving through the air in the real world its not driving through laminar air. The air is moving in several different directions. Its turbulent, has different temperatures and pressures etc. In a windtunnel the flow hits the car at a perfect angle.

i was thinking the same, offcourse a windtunnel brings a close representation of reality, however, the fact remains that
there is an enforced stream of air powered to the car, whilst in fact in real life, there is no wind 'aimed' at the car, the car
breaks the wind, and indeed, wind can come from several directions.
a windtunnel does provide a good representation of how it 'cuts' through air, which is offcourse important [otherwise there would not be so much invested].

the fact that windtunnel testing is very tricky is that ferrari last year made various parts that in the windtunnel performed like they were hoping, but in real life didn't provide the same results.

Precisely. A windtunnel is still the best way to simulate how a car travels through the air. Its not perfect though thats why we see so many teams having correlations problems though out the years.

[AnthonyG]

The rules are clear that only 60% models can be used for aero testing

How is that rule enforced?

[autogyro]

It still baffles me as to why there is not sufficient data available now on aero to undertake all calculations in CFD.
This thread has been an eye opener to me.
I can see that the laminar air used in a wind tunnel reacts nowhere near the same as the static air (or low speed turbulent air) a car experiences in the real world.
To me aero is simply yet another area of modern science that if the BS is removed remains simply qualified guess work.
Quantum Physics is another example. no facts just theories.

[CBeck113]

It still baffles me as to why there is not sufficient data available now on aero to undertake all calculations in CFD.
This thread has been an eye opener to me.
I can see that the laminar air used in a wind tunnel reacts nowhere near the same as the static air (or low speed turbulent air) a car experiences in the real world.
To me aero is simply yet another area of modern science that if the BS is removed remains simply qualified guess work.
Quantum Physics is another example. no facts just theories.

The biggest issue with aerodynamics is the fact that you cannot measure it without influencing it. If you could measure pressure differences and flows in the air on and around the car with a device like an IR thermometer (which actually only works on a surface, so it has a similar issue), then we'd be much further than we are today. But, since that doesn't exist yet, we have to estimate the actual pressures and flow vectors to then calculate (= numerical simulation) what the air is doing. It is also being influenced by the laminary flow (easy to simulate) and turbulent flow (very difficult to simulate, see chaos theory). The turbulent flow is the problem, and it turns up almost everywhere around the vehicle. You can say how this turbulent flow volume will influence the lanimary flow around it, but it is almost impossible to say where the lowest pressure areas are in the turbulence, and therefore it can't be simulated accurately. No big issue for street cars, but for F1 or aircraft this can have a huge impact on the real-world performance.
We are much further than Quantum Physics, but we have a long way to go before we know enough for realistic simulations to replace wind tunnels.

[Greg Locock]

I'm not so sure I agree there. You could model the distrubance that the instruments make in CFD as well. We do this all the time - you don't just model the item under test, you model the test rig as well.

Then when it is all correlated properly, run the model without all the test rig mods.

[autogyro]

The FIA could give the teams six Months to put together an aero data base for mutual use and then ban wind tunnels.
Now that would save a few wasted pennies.

[Tommy Cookers]

just how 'laminar' is the tunnel flow ?
it has been accelerated by a highly-powered fan that has a small number of discrete blades
its velocity is less than consistent
flow quality is always a factor, so (it has been said) no two tunnels give the same result
Boeing and Airbus etc have spent millions comparing in detail in-flight flow to tunnel flow (eg stagnation)
F1 can use 60% models, not 6%

turbulence effects are detectable eg on 'stalling' AoA in real life light aircraft handling
in really still air, on landing there is 'float' beyond that related to airspeed/momentum variation

[Holm86]

just how 'laminar' is the tunnel flow ?
it has been accelerated by a highly-powered fan that has a small number of discrete blades
its velocity is less than consistent
flow quality is always a factor, so (it has been said) no two tunnels give the same result
Boeing and Airbus etc have spent millions comparing in detail in-flight flow to tunnel flow (eg stagnation)
F1 can use 60% models, not 6%

turbulence effects are detectable eg on 'stalling' AoA in real life light aircraft handling
in really still air, on landing there is 'float' beyond that related to airspeed/momentum variation

Right behind the fan there is a stator changing the flow direction to a more direct rearward flow. And there is a section before the air enters the test chamber which consists of a large number of pipes fittet in an array. This causes the flow to be laminar.

Because even thoug laminar flow is not 100% as in the real world its still 1000 times better than turbulent air comming directly from a fan.

[autogyro]

It is still nothing like the still air (with wind and turbulence on occasion) that the moving car meets on track.
Float when landing a light aircraft is caused by boundary layer effect.
I once flew a tiger moth that had been covered in really light fabric and trying to land on a large flat grass airfield proved almost impossible.
The aircraft just would not go down.

[CBeck113]

I'm not so sure I agree there. You could model the distrubance that the instruments make in CFD as well. We do this all the time - you don't just model the item under test, you model the test rig as well.

Then when it is all correlated properly, run the model without all the test rig mods.

Do you work for Ferrari?
Sorry, jokes aside, yes it is possible (and done, especially for road cars), but when we look at the flow over the entire F1 car (i.e. to determine the diffusor sealing), you would have to make dozends of measurements, relocating the sensors after every pass, always measuring forward of the rig in order to get the exact flow & form from each vortex. Before doing this you have already calculated the expected results, but still need the wind tunnel data for confirmation. Anywhere you have turbulent flow, and for an F1 car the open wheels are causing the most issues, you cannot accurately simulate, so you measure. You also cannot accurately simulate the influence your sensors in the turbulent flow, which is why I made my statement above.

[Holm86]

It is still nothing like the still air (with wind and turbulence on occasion) that the moving car meets on track.
Float when landing a light aircraft is caused by boundary layer effect.
I once flew a tiger moth that had been covered in really light fabric and trying to land on a large flat grass airfield proved almost impossible.
The aircraft just would not go down.

You should look at what I wrote on page 1. I agree with you

[Lycoming]

It is still nothing like the still air (with wind and turbulence on occasion) that the moving car meets on track.

Is it? I would think it uncommon to encounter perfectly still, completely nonmoving air on a racetrack plied by multiple cars, whereas the flow entering the test section of a wind tunnel is usually fairly laminar.

[autogyro]

It is still nothing like the still air (with wind and turbulence on occasion) that the moving car meets on track.

Is it? I would think it uncommon to encounter perfectly still, completely nonmoving air on a racetrack plied by multiple cars, whereas the flow entering the test section of a wind tunnel is usually fairly laminar.

The air in a wind tunnel is moving at high speed.
The air at the track is stationery or subject to a low wind speed.
Low speed turbulence can be introduced to the air before the car goes through it but it will be low speed.
Where is the similarity?