My comments on differences in aerodynamic wind tunnel testing results when using scale models came from answers to questions I asked in conversation years ago with the race engineer and aerodynamicist on one of the most dominant IMSA GTP cars in history; I have no direct independent knowledge of the subject.

Wind tunnel testing is limited to 60% size and a max speed of 50 meters per second. Real world is 100% and a max speed of F1 cars are in the 220 mph range which makes the math really simple bc that is about 100 meters per second.
Now before everyone starts jumping at ratios let me lay out some other engineering equations onto the table. Check this out http://en.wikipedia.org/wiki/Buckingham_Ï€_theorem
The pi theorem spits out all kinds of parameters that help out in understanding fluid flows. SOME examples are Reynolds number, Mach number, Froude number, Weber number, Euler number, Prandtl number, Eckert number, Strouhal number, Grashof number, specific-heat ratio, roughness ratio, temperature ratio, pressure coefficient, lift coefficient and drag coefficient. All of these unitless numbers tell you one thing or another about the systems in which you’re trying to analyze and if two systems have similar parameters then they will behave similarly too. But instead of analyzing identical systems in a highly controlled environment a 1:1 wind tunnel the FIA tried to reduce costs by instituting a 60% size and 50% speed limitation while that actually only made the analysis and manufacturing more difficult.

There are two main areas of similarity that is primarily focused on in modeling of systems in wind tunnels: geometric and kinematic. Geometric similarity means that it is a true scaled model but that means that not only the general dimensions of the model has to be to scale but the gaps in between the body panels, the bolt heads, the rivets, the surface roughness also have to be to scale. The more that it isn’t an exact scale replication the less the results will align. Kinematic similarity is the quote from Langhaar in my previous post “The motions of two systems are kinematically similar if homologous particles lie at homologous points at homologous times” which means that the reality-model correlation has to apply to the length-scale ratio and the time-scale ratio such that both systems have the same Froude number. Length-scale equivalence is another simple application of the geometric similarity but the time-scale equivalent is a difficult thing to grasp for a lot of people. The length ratio from the geometric similarity is correlative to the square of the velocity and time. So if the model is a 60% scale model then one model second is equivalent to .7745 seconds in reality and 50 meters per second applied to a 60% model is equivalent to 64.55 scale meters per second but that only gets us to 145 mph so other tricks have to be used such as increasing the kinematic viscosity by using a different fluid or by increasing the mass flow rate in some other creative ways but even that has been closed down… wind tunnels are limited by physics of the system and that the FIA has limited the testing systems in such a way that equivalent speeds above about 150 mph can’t be tested in a wind tunnel which (now that I’ve done the math) explains why the McLaren airfoils had problems that should have been resolved in the wind tunnel last year above those speeds last season... because the FIA won't let them test their designs at top speed.

Enter CFD… full scale modeling, multiple iterations can be tested, and it is a lot cheaper to convert a preexisting closet into a computer lab for half a dozen glorified draftsman and add the machines to the IT departments maintenance list than it is to build a new wind tunnel every time the FIA changes what is allowed. Computers are comparatively cheap. Years ago FEA design optimization systems were getting pretty advanced about being able to provide a general design with allowable variables, loads and material selections, hit the pretty little optimize button, the computer churns away for a few hours and spits out the part that is minimized in the ways you wanted and more often performed better than expected. It isn’t that far of a stretch for the same algorithms to be applied to the CFD world and it wouldn’t surprise me one bit if the designers were being taken further and further out of the loop in favor of computer generated solutions the more we knock on future’s door.

Mariner, wow I haven’t read that name in a long time.

It was as you say, a failure.

And who was the skipper of Mariner? Ted Turner, of CNN fame. A truly great sailor but a jerk. What did Ted say about the innovative design after it proved to be bad, “even a turd is tapered at the end”. When Robert Director changed the back end of the boat in the middle of the challenger series he didn’t use a computer he used a marker and drew a line around the back and told the workers, “cut it off there”. Some improvement, still lost.

While Mariner was a disaster, and built about a mile from where I live, Courageous was an epic ship which just kept winning. Built about ten miles from where I live, in the opposite direction.

Oohh I could bore you for hours.

Let's put it this way. If you are able to test the car in a perfect condition (no wind, etc..) on the track, scaled wind tunnel model and CFD, you'll get 3 different answers.

CFD is a numerical solution meaning that there are no exact answers due to round off errors of numbers. These numbers are crunched through 100 million cells and 1000+ iterations.

The wind tunnel will have different factors all together such as calibrations of the measuring instruments and boundary layer of the test section etc.

In the real world, it's pretty unlikely to be able to test in a no wind condition. Plus the undulations in the surface of the road.

What you try to achieve is to get the solutions within a certain percentage of each other.

That being said, track testing is the best. Like all the teams say, by banning track testing to reduce costs, they are forced to investing in simulations at the factory. That money saved gets spent elsewhere. On the track testing is where you'll get the best result.

Track testing is the best but the reason for effecutally banning it is to increase the efforts to increase the accuracy of the CFD programs and computer systems.

impatientinventor - 03 December 2009 12:13 AM

Track testing is the best but the reason for effecutally banning it is to increase the efforts to increase the accuracy of the CFD programs and computer systems.

And how exactly does that decrease costs when these programs cost not only for the programs themselves, but also for the "experts" who know how to get the most from them and relate them to "real life". This addition of elements in analysis can supercede the actual cost of running the car in cases of "crossing the realms" of CFD-Simulation-data acquisition programs. And at some point the need to actually run the car to "prove" out what the programs say.
IMHO, it would be cheaper to have test days the day before official practice. Though having said that, the money teams would still use all of the above...

Definitely a necessary evil....

the teams may be paying millions of dollars for the development of the software and training of what will become the industry experts but without that expense being borne by F1 then how would they ever find out that the back tires of a windstar will lift off the ground at 120mph without spending tens of millions on tooling and development to scrap a car WAY too late in the development process. Now they can design the minivan put it in the same software and figure out how much they can detune the engine and get it to still barely go up a 1% grade hill at 45mph.

In the end the guys who paid $300 for calculator in the 70s helped get me my graphing calculator for under a hundred buck today and the F1 team paying for the development of CFD software today will be taking my job away from me in 5 years.

impatientinventor - 11 December 2009 01:20 AM

the teams may be paying millions of dollars for the development of the software and training of what will become the industry experts but without that expense being borne by F1 then how would they ever find out that the back tires of a windstar will lift off the ground at 120mph without spending tens of millions on tooling and development to scrap a car WAY too late in the development process. Now they can design the minivan put it in the same software and figure out how much they can detune the engine and get it to still barely go up a 1% grade hill at 45mph.

In the end the guys who paid $300 for calculator in the 70s helped get me my graphing calculator for under a hundred buck today and the F1 team paying for the development of CFD software today will be taking my job away from me in 5 years.

I thought the same thing with data acquisition some twenty years ago. I too thought the software would progress and be able to analyze on it's own and take my job away too. Still hasn't happen, real world experience along with the "tool" still reins over.
I can't see CFD taking positions away or replacing human intervention but rather improving the knowledge of the engineers that know how to use it and use it well. That makes the knowledge "factor" much stronger and will take the engineering to even greater heights.
Someone still has to tell the program what to do and add imagination....

All the computer power and sophisticated programs in the world are great ways to test a lot of things, mathematically, without having to build them or test them, either as models or in real life. However, they do not take the place of knowing what you're doing. The hundred dollar graphic calculator that will do all manner of calculus, for example, will do the functions and draw the picture, but you still need to know calculus in order to program it. You also need to know what to expect when it gives you the answer. If you just blindly plug in numbers, it will give you calculated garbage out from the un-calculated garbage you put in.

By the same token, you also need to be able to make sense of the results, recognize when you've reached the optimum point in your design and not go beyond it. The 12-Meter yacht "Mariner" was one example of getting conned by test results. The Lotus 80 and other "ground effects" cars of the early 80s that "porpoised" at speed are a second example. They looked great in theory, but failed badly in practice. I'm sure you can think of others.

In the end, a good computational fluid dynamics program is like a Stradivarius violin -- in the right hands, it can make beautiful music. But you still need to know where to put your fingers and how to draw the bow. And that is why we will always have "glorified draftsmen" and engineers.

yea but this engineer keeps getting displaced by the "glorified draftsmen" because I am just infront of the wave of technology that has been crashing on society for the past decade....

The valedictorian's address at my graduation summed it all up for me... He said "I may have the highest grades for the past four or five years but 95% of what we will use over our careers as engineers hasn't been invented yet so keep learning."