Relation Between Wind Tunnel and CFD
 

Hello everyone,

Can someone please tell me why still use wind tunnel when CFD is already exist?

Is it because of large memory needed of its large mesh?

Thanks in advance

Dido

I would say it is because CFD is unreliable and not properly validated for many applications.

CFD + Wind Tunnel = Complimentary Relationship
 

"To many observers the rivalry between wind tunnels and Computational Fluid Dynamics (CFD) is a zero-sum game - as CFD matures it simply replaces wind tunnels. However, this is far from the truth. Often you'll find wind tunnels and CFD used together in a symbiotic process where one technique fills in knowledge gaps left by the other." For more try:
http://www.symscape.com/blog/wind-tunnels-and-cfd

CFD is only unreliable when you don't benchmark (calibrate) - agreed. However, the same principle applies to wind tunnels when you don't benchmark too. So without benchmarking most modeling (physical or virtual) is unreliable. For more try:
http://www.symscape.com/blog/lessons...fluid-dynamics

Full Disclosure: I authored both articles I linked to above.

I understand and agree with what you are trying to say, but "symbiotic" is a bit twee.


That's not the only time CFD is unreliable.

Many CFD codes (not necessarily the under-lying mathematics or assumed physics) fail to pass what some would regard as essential scientific tests, e.g. reproducibility of results by different, unbiased experimenters.

if there are tow results before the goverment officer who make important decision,one is from wind tunnel,the other is from CFD,he will choose the former.

When both are well done, I'll trust wind tunnel over CFD. I'd use wind tunnel to calibrate CFD, I would almost never use CFD to calibrate a wind tunnel.

Why does CFD hava an issue? Turbulence and turbulence models. For example, only the highest fidelity CFD runs can capture transition from laminar to turbulent, and that is for academic problems. Another example, most people use RANS solvers. The turbulence models are just models. They create eddy viscosity in an effort to model turbulence. The eddy viscosity can be hundreds of thousands times larger than the laminar viscosity term in the NS equations. So basically all that is being done is to throw gobs of viscosity at a problem to model turbulence.

Of course CFD has it's place in the design cycle. It helps engineers explore a problem, come up with preliminary designs, and get the best value out of wind tunnel runs. Wind tunnel runs can be very expensive and time consuming. I believe in the need for CFD since I write CFD codes.

Wind tunnel tests have their issues too. For example modeling flight Reynolds numbers and transonic flow. But I don't necessarily trust CFD to model flight conditions either. For example, it is very hard to model gaps in geometries, such as those that occur around flaps, slats, and ailerons. The reason is both griding and physics. Many times those gaps are fared over in CFD.

Here is a link to a problem where a RANS solver can give you two answers, depending on how the problem is converged, and neither one of them is correct.

http://www.hegedusaero.com/examples/...eShoulder.html