Wind Tunnel Superceded with Total Switch to CFD

[hardingfv32]

CFD will be less expensive to use but only after it is fully developed and validated. There will be no cost savings getting to that point.

The only way this works is if a Spec CFD program is used by all teams and validation activities are prohibited.

Brian

[Just_a_fan]

Manor got the "when" wrong, but they did not get the "what" wrong. In the broad historical arc, I don't even think they were very far off with the "when".

They took a knife in to a gun fight...

One day CFD will be "better" than wind tunnels but how many days until that day comes is the big question.

Bear in mind that the teams use scale tunnels at scale speeds. Imagine if they used full speed tunnels at realistic speeds. CFD would have a long way to go to beat that because it will have to be DNS to do so. And car-scale DNS is a very, very long way down the road. A very long way.

[bill shoe]

CFD and Wind Tunnels are both simulations of the actual car's running aero performance at a grand prix weekend. CFD is a computational simulation, and Wind Tunnels are a simplified/model-based simulation.

Probably the cheapest and most accurate validation for either type of simulation is to look at data from the running car at grand prix weekends. If you can't relate either type of simulation to actual running car data then there's no point in the simulations. At that point you have the dreaded correlation crisis.

I think anti-CFD people (a.k.a. pro-Wind Tunnel people?) are setting up a bit of a straw-man argument. The assumption is that wind tunnels give the actual and correct value which can be compared to CFD to check the accuracy of CFD. If it was this simple there would never be the one or two teams per year that have a correlation crisis.

I don't buy the idea that CFD must simulate every micro bit of air dynamics to be a useful simulation tool. The simplifications in CFD, like the simplifications in a wind tunnel, accept theoretical imperfection in order to be practical and useful tool.

[djos]

Boeing (and previously other US customers) seem to regard activity there as money well spent
the power required for this type of work seems to be below 15 MW, ie only an order of magnitude greater than F1 tunnels

good thread though !

15MW of power draw @ say 10c p/kWh is about $1,094,976 per month to run or $13.14 million per year (all day everyday - worst case scenario) - if you dont have the expense (and turn around time required) of building 2/3rd's scale models I wonder if the costs roughly balance out?

PS, 15MW is a massive amount of power draw, the average commercial data center in Australia only draws about 6MW in total with a PUE of 1.4-1.6!!!

[Just_a_fan]

I think anti-CFD people (a.k.a. pro-Wind Tunnel people?) are setting up a bit of a straw-man argument. The assumption is that wind tunnels give the actual and correct value which can be compared to CFD to check the accuracy of CFD. If it was this simple there would never be the one or two teams per year that have a correlation crisis.

I don't buy the idea that CFD must simulate every micro bit of air dynamics to be a useful simulation tool. The simplifications in CFD, like the simplifications in a wind tunnel, accept theoretical imperfection in order to be practical and useful tool.

I'm pro-CFD (and wind tunnels) but I'm realistic that CFD has such high computation requirements that it is limited when it comes to large simulations.

A wind tunnel has limitations caused by the scaling of the model. It is a relatively easy task to build a full size tunnel (although F1's rules prohibit it) and run a full size car. It's far from trivial to do the equivalent with CFD.

I look forward to seeing a successful CFD-only F1 car. I just wonder how old I'll be when it happens.

[Tommy Cookers]

CFD and Wind Tunnels are both simulations of the actual car's running aero performance at a grand prix weekend. ....... If you can't relate either type of simulation to actual running car data then there's no point in the simulations. At that point you have the dreaded correlation crisis.
I think anti-CFD people (a.k.a. pro-Wind Tunnel people?) are setting up a bit of a straw-man argument. The assumption is that wind tunnels give the actual and correct value which can be compared to CFD to check the accuracy of CFD. If it was this simple there would never be the one or two teams per year that have a correlation crisis. ......

j a f suggests that F1 problems would be solved by using full scale and speed WTs ?
I guessed their (rather large) scale and speed is believed to be essentially as good as fullsize
(and without killer issues of fullsize fullspeed moving groundplanes)
though WT flow quality is always a likely suspect

some aviation WTs correlate much better than others
eg the early F-A 18 E/F carrier approach issues (though the makers stoutly claimed WT tests couldn't have predicted it)
a cynic might say that Europe has been more willing to fund the making of WTs than the making of planes (and vice-versa the USA)

the need for Re and other sweeps means that aviation WT work generally uses far less than the nominal maximum power

btw, in fun ...
the public domain now reveals that the 8'x8' supersonic Bedford UK tunnel (scrapped after 46 years use) was stated to be 60 MW
and the USA was trying to get for civil aviation 2 big new WTs in the Clinton era (iirc including cryogenic-cooled capability ?)

EDIT - Note to self
as mentioned by riff raff ? in post later, the NASA Ames 'Full Size' tunnels are more powerful
and seemingly used for testing actual aircraft and rotorcraft at full size
rotorcraft at model size cannot reconcile aerodyamic and aeroelastic factors

[godlameroso]

CFD will be less expensive to use but only after it is fully developed and validated. There will be no cost savings getting to that point.

The only way this works is if a Spec CFD program is used by all teams and validation activities are prohibited.

Brian

You got something against validation engineers?

[riff_raff]

some aviation WTs correlate much better than others
eg the early F-A 18 E/F carrier approach issues (though the makers stoutly claimed WT tests couldn't have predicted it)
a cynic might say that Europe has been more willing to fund the making of WTs than the making of planes (and vice-versa the USA)

the need for Re and other sweeps means that aviation WT work generally uses far less than the nominal maximum power

btw, in fun ...
the public domain now reveals that the 8'x8' supersonic Bedford UK tunnel (scrapped after 46 years use) was stated to be 60 MW
and the USA was trying to get for civil aviation 2 big new WTs in the Clinton era (iirc including cryogenic-cooled capability ?)

The 40' x 80' tunnel at NASA Ames uses over 104MW at max performance. Still available for use by any private company that is willing to pay the cost of operating the facility.

[djos]

some aviation WTs correlate much better than others
eg the early F-A 18 E/F carrier approach issues (though the makers stoutly claimed WT tests couldn't have predicted it)
a cynic might say that Europe has been more willing to fund the making of WTs than the making of planes (and vice-versa the USA)

the need for Re and other sweeps means that aviation WT work generally uses far less than the nominal maximum power

btw, in fun ...
the public domain now reveals that the 8'x8' supersonic Bedford UK tunnel (scrapped after 46 years use) was stated to be 60 MW
and the USA was trying to get for civil aviation 2 big new WTs in the Clinton era (iirc including cryogenic-cooled capability ?)

The 40' x 80' tunnel at NASA Ames uses over 104MW at max performance. Still available for use by any private company that is willing to pay the cost of operating the facility.

I'm guessing that's because it operates at 560 kph and can fit full size planes in it.

[Just_a_fan]

j a f suggests that F1 problems would be solved by using full scale and speed WTs ?

Not at all. I was merely pointing out that creating a bigger wind tunnel is easy compared to creating a massively more computationally expensive CFD implementation.

[Cold Fussion]

j a f suggests that F1 problems would be solved by using full scale and speed WTs ?

Not at all. I was merely pointing out that creating a bigger wind tunnel is easy compared to creating a massively more computationally expensive CFD implementation.

Is it really though? A bigger wing tunnel is just a matter of higher capital and running costs which is basically the same as building a larger computing cluster. Obviously though a wind tunnel will plateauin performance far quicker than a supercomputer.

[Just_a_fan]

A full size wind tunnel directly simulates every molecule of the air and every square millimetre of the car. To do that in CFD is currently impossible. That's the point.

Sure, the tunnel still has potential for correlation issues with the track, but it's closer than any CFD will get to "reality" in the near/medium future.

[Jolle]

i guess the connection with the Sim is a major advantage with CFD, if the use becomes less limited, teams can not only work with CFD during designing but also during race weekends finetuning the setup in the sim with all the variables (rain/temp/high/low pressure/running at the front, in the pack/wind direction, etc etc)

It will be a lot more dynamic.

[Vyssion]

i guess the connection with the Sim is a major advantage with CFD, if the use becomes less limited, teams can not only work with CFD during designing but also during race weekends finetuning the setup in the sim with all the variables (rain/temp/high/low pressure/running at the front, in the pack/wind direction, etc etc)

It will be a lot more dynamic.

There are a few simple approximations I have seen done with CFD currently to address some of those - though I cannot necessarily speak for their validity!!
For example, you can create a mesh of points which are seeded randomly over an inlet surface and then define a random velocity around some gaussian curve about your mean velocity to help with simulating minor variations. You can also do semi-dynamic simulations where you can stipulate an unsteady RANS solver to solve over some period of time where at a given timestep, the temperature jumps to a new value thus changing air density (which can help you look into humidity/pressure etc.). Rain is a little trickier to do, however what I have seen done before is create a moving ground which has a semi-fractal based distribution of peaks and troughs which are 5mm to -5mm in height respectively and then changing a coefficient of friction to some averaged value based on your distribution of puddles. Running in front or behind of another car is simple in that you just simulate two cars, one behind the other.

But yeah definitely not currently a substitute for real world testing!! One day maybe.........

[TwanV]

Myself being a CFD-engineer, my general impression is that people simply expect too much out of it. The general perception is that you simply import a geometry, press run and get everything you need in wondeful isocontours. At least that is what e.g. Ansys is trying to sell. The truth is much less glamourous, and the amount of labour and skill required to construct even a mesh that suits the needs of the problem you're trying to evaluate can be quite serious.
I guess the constructors are very sophisticated in automatic mesh generation and in combining all the various scaling problems into their setup, still I would be surprised if full scale transient simulations over all variations in the free stream velocity, turbulence patterns, etc would be practical, time effective and economical compared to windtunnel experiments for a significant time to come, if not forever, as turnaround time is of primary concern. Also I can imagine accuracy can be disputed for the important subtleties of the flow around a racing car. Even the most sophisticated turbulence models such as LES still are models, and have a run with reality at their very core.
As a design tool CFD is wonderful of course, either to (automatically) optimise design or check the qualities of a new concept upto the point that you're fairly certain that something is going to improve the car. But the fact that flowviz is in use during practice tells me that there is a significant accepted margin for error.