Hi CFD Community!

Reading papers about CFD validations one question has come to me. All papers I've read come with an experimental data gathering to be able to have the CFD data validated, of course, but many of them allways concludes that a CFD investigation/report/analysis should allways be complemented with experimental tests due to CFD is still nowadays not considered as a pure quantitative tool to know the flow characteristics.

The fact is that it is known cars manufacturers use CFD to make different models, run simulations and chose those who are the the best in terms of having good aerodynamics in the earlier stage of the shape of the car. However, wind tunnel tests and/or test-road runs are allways needed at the end, but CFD has helped to reduce, at least, the developing time. Unfortunatelly, I have not been able to find any paper that explains with some (quantitative) detail what is the economical saving that is achieved with the use of CFD (if any!) a part from a save in the developing time (by the way, any paper that I have (not many though) doen't explains in a qualitative way that developing time reduction)

I would be pleased if anyone could refer or send to me any paper that you might have, explaining with some detail what I exposed above. I believe that the time and economical saving that comes from the use of CFD must be considerable in comparison to a wind tunnel test (the example that I'm most interested), because it would have no sense to waste money in a licence of a CFD code that doesn't give conclusive quantitative results if wind tests were cheaper, knowing that any experimental data done is a decisive result.

Thanks in advance!

Hi Freeman,

I see that you have a keen interest in automotive CFD applications.

To be rather candid, the views I have seen over the past 5 years have not been overly complimentary to the accuracy levels predicted for drag & lift coefficients. The perceptions are that CFD can give results +- 20% from a known wind-tunnel solution & even then, with some careful manipulation.

Many folks have the opinion that CFD merely provides 'nice pictures'. To others, CFD provides a general 'feel' of how the airflow paths can change, given a vehicle geometry change.

A lot of comparative study has been performed by, for instance, Prof. Turek's team in Germany <www.featflow.de>, where they performed benchmark tests for many commercial & research codes - the standard test being flow over s cylinder, in 3d. They have also performed soem work with a local 'well-known' automotive manufacturer.

My personal speciality is automotive heat-transfer applications eg. flow through micro-finned heat-transfer devices. At this point, I have developed a rather accurate simulation environment - but, only after first establishing a tried-&-tested testing framework - where real wind-tunnel data is used to compare cfd predictions.

There are so many factors that will affect your results, that it is generally almost impossible beforehand, to say which factors cause errors. This takes a lot of time & patience. My personal 'pet hate' is convection-stabilisation like 'upwinding' & its family - these influence the pressure-drop predictions in a very dramatic way - for the worse.

For some reasonable background information, can I refer you to the SAE website?

diaw...

Thanks for your participation diaw. Yes! I've also heard few professors in my University saying that CFD is good only at making some "nice pictures", but not taking conclusive data over the quantitative flow fields. But I believe that if automotive, aerospace and even naval industry are spending time and economical resources in CFD is for something...

I was searching in SAE website few days ago for any paper comparing CFD costs with wind tunnel ones, but with no success; after reading your post I will go there again, but I wrote here to see if somebody knew any specific paper on this topic. I've written to also MIRA and Pinninfarina, but there's still no answer.

Thanks for your time

Regards, Freeman.

Hi again Freeman,

In many ways, we try to make up for our lack-of-knowledge across the full fluid-flow spectrum, by assuming that the N-S equations will somehow contain some 'hidden magic' that will provide us the insights we need.

At incompressible high speed flows, & low-speed (ships) things have been explored in a lot of detail & the required adjustments made - but, in the middle is a veritable no-mans-land which is going to take a new paradigm to fully understand & model.

To be brutally honest, the N-S equations, in their current form, will probably need some modification to cope with the no-mans-land... This is the area in which you are working. Have fun - it is a challenge which most of academia chooses to skirt at some discreet theoretical distance.

diaw...

MIRA have an excellent 'climatic wind-tunnel'. When I went there a few years ago, I was very impressed with their work on the CFD side, but, if my memory serves me correctly, they were using CFD for flow inside engine-bays & exhaust systems.

Perhaps they have expanded their cfd work since then? Tey would be an excellent source of information.

diaw...

This is almost more of a philosophical question than anything else. There is a silly saying, which contains more than a grain of truth: "Nobody will believe a calculation, except the person doing the calculation, but everybody will believe an experiment, except the experimentor." It's difficult to make a cost comparison between wind-tunnel testing and CFD, because the two processes really compliment each other, and should not be seen as being in competition with each other. To put it differently, if you are trying to choose between CFD and wind-tunnel testing to answer the same question, you are in serious trouble.

In practice it helps to think of the outcome of a CFD calculation (or a wind-tunnel test, for that matter) as being an engineering decision, rather than a number. Used together, intelligently, CFD and wind-tunnel testing can help you make better engineering decisions. The value of that is difficult to estimate, but avoiding engineering mistakes is extremely cost-effective. If, by avoiding making the engineering mistake in the first place, you can knock a problem-fixing wind-tunnel test series out of the development process, the CFD program has already more than covered its own cost. So, if you have to motivate a CFD budget, this is an aspect you should probably focus on.

It is good to look at what the top Formula 1 teams do. They operate in a competitive environment, so the success of what they are doing can to a certain extent be measured. The top teams have all invested massively in both wind-tunnel testing and CFD, so that should be telling you something.

In my own experience, CFD does not really lie to you. It is easy to ask the wrong questions, ask them the wrong way, or misinterpret the results, but the CFD result is seldom "wrong".

Very valuable comments, diaw and Charles: many thanks.

Charles, you've said some things I wanted to arrive at:

>> "It's difficult to make a cost comparison between wind-tunnel testing and CFD, because the two processes really compliment each other, and should not be seen as being in competition with each other"

This is what most of the papers I've read concludes: CFD is becoming a useful tool to reduce developing times by a reduction in the experimentation (specially in wind tunnels), but the fact that every CFD calculation should be validated with real experimentation sometimes makes me to be confused and to ask to myself if CFD is really useful/helpful: in plain words, if we NEED an experimentation to validate results, why to do a CFD simulation? The next paragraph shed me some light into this.

>> "If, by avoiding making the engineering mistake in the first place, you can knock a problem-fixing wind-tunnel test series out of the development process, the CFD program has already more than covered its own cost"

Any of you know any paper related to this or have it? This is what I was looking for! Successful cases that show a little how it has been able to knock experimental tests out of the developing process thanks to CFD. I believe that it should be any that has investigated the benefits both in economical and time terms between a CFD run and a wind tunnel one, for example... but since now, I've been only able to know that this benefits are currently happening, but I haven't seen how and their importance in a quantitative way.

Thanks a lot for your participation!

Hi again Freeman,

A thought... I wonder if approaching a few large automotive manufacturers would help? There are a few groups large-enough to have in-house climatic wind-tunnels. They are sure to have some sort of comparative information. I know that it is not common for them to hand out this type of info to just anyone, but, dependening on what you require the information for, coupled with some useful feedback to them after your study is complete, could make them useful allies.

Your contact to MIRA, coupled with a visit, would be very useful. I have fond memories of their kindness during my visit to their premises - a high-speed drive around their test-track. They were very, very well organised. Thoroughly nice folks.

Their climatic wind-tunnel was absolutely huge & very well instrumented.

Good luck with your search. The results will be very useful.

diaw...

Hi folks,

just a minute for fun about this subject...

http://pc.freeshell.org/cfd/cfd_windtunnel.html

Enjoy

Renato.

Freeman wrote:

" but the fact that every CFD calculation should be validated with real experimentation sometimes makes me to be confused and to ask to myself if CFD is really useful/helpful: in plain words, if we NEED an experimentation to validate results, why to do a CFD simulation? "

I would modify the above statement somewhat. Not EVERY CFD calculation needs to be validated with a real experiment. If you have determined the accuaracy of your CFD code for the parameters of interest for a particular case with experimental data, then performing simulations on "similar" cases (ie, major problem features do not change "too much") is expected to provide simulation results with similar accuracy. In the aircraft combustor industry that I am familiar with, they would typically call the "anchoring the CFD code" for a particular type of combustor design. Of course, a lot of experience is needed to determine how "similar" a particular simulation needs to be compared to the "anchored" simulation in order for similar simulation accuracy with respect to the design parameters of interest.

The benefit of the above "anchoring" is to significantly reduce the number of rig tests needed to develop the product. An example of significant reductions for aircraft engine combustors is given in the AIAA paper 2004-331 from Pratt & Whitney in which rig tests per engine program were reduced by 50% from 1985 to 2001 by increased use and improved accuracy of their in-house CFD tools for combustor design. The other major benefit of CFD can be helping to determine/solve problems with existing designs since a completed product typically does not permit detailed measurements (at for internal flows). Even if the CFD is only showing trends correctly, I certainly know of examples where CFD helped solve "problems in the field" without starting up a major and costly experimental program for a product already in production.