[mazhar16823]

Hi everyone,



I have never been able to be a part of wind tunnel experiments, however I have conducted CFD simulations where usually I get to see that experimental data never shows the location of transition from laminar to turbulent boundary layer however usually it provides very low amount of data points compared to CFD. So, does it mean that flow everywhere in the wind tunnel is turbulent that is why the experimental data does not even give an indication of transition?

[t.teschner]

Wind tunnels these days are designed with a heavy emphasis on reducing turbulence as much as possible. If you measure the free stream turbulence intensity, i.e. u'/U_average, you would generally expect those to be less than 1%. Good wind tunnels will have much less than that. Have a look at this paper (it is open access): https://www.sciencedirect.com/scienc...752?via%3Dihub , in particular Figure 5, where they measure the turbulence intensity for a specific wind tunnel. It is, as one would expect, depending on the Reynolds number, but their values are generally less than 0.5% (measured at the centerline of the wind tunnel), for some cases even below 0.1%, which is not uncommon.

The reason why wind tunnel data experiments don't provide the transition point is that it is not a very useful quantity to measure (and also you would need to specify what you mean by that). For example, measuring the forces on an airfoil would be done by suspending the airfoil int he center of the wind tunnel. Typically it will be attached through turntables at the wall and your question may be why we don't measure the boundary layer at the wall here? well, there are always going to be wind tunnel effects (pressure can't expand indefinity like in a cfd simulation, for example, where we can place the boundary condition arbitrarily far away), so there are blockage effects here as well. The boundary layer is just another thing we have to deal with. But just like we have to do some tricks in CFD to get our results correct (for example using wall functions in RANS models and, if you think about it, is quite a crude approach and may not be necessarily physical), wind tunnel experimentalist do user similar tricks as well to prevent turbulent boundary layers and their influence ont he result. You could for example suck away the boundary layer, you could have a moving ground and I am sure there are more techniques available, but I guess you see the point. That is also why it is sometimes very difficult to get a good correltation between wind tunnel results and CFD simulations. If you are a motorsport fan, you may know, for example, that year after year F1 teams are having difficulties understanding the aerodynamics around their car and you often hear that their CFD results don't correlate to their wind tunnel experiments. Again, both approaches introduce approximations and errors and it is our job as CFD / wind tunnel practitioners to ensure their correctness.

So to conclude, yes, turbulent boundary layer can develop in wind tunnels which we may want to address but generally speaking wind tunnels themselves have a very low level of background turbulent intensity and it just becomes a matter of preparing the geometry and the set up in a way that any boundary layer that may develop does not change the actual measurement. The best that we can hope for is that some correlations studies are done (for example, by simulating the geometry within the wind tunnel and not using far-field boundary conditions) or at least that who ever is doing the wind tunnel experiments ensures is noting the influence of any quantities that may have an influence on the final result.

[mazhar16823]

Originally Posted by t.teschner

Wind tunnels these days are designed with a heavy emphasis on reducing turbulence as much as possible. If you measure the free stream turbulence intensity, i.e. u'/U_average, you would generally expect those to be less than 1%. Good wind tunnels will have much less than that. Have a look at this paper (it is open access): https://www.sciencedirect.com/scienc...752?via%3Dihub , in particular Figure 5, where they measure the turbulence intensity for a specific wind tunnel. It is, as one would expect, depending on the Reynolds number, but their values are generally less than 0.5% (measured at the centerline of the wind tunnel), for some cases even below 0.1%, which is not uncommon.

The reason why wind tunnel data experiments don't provide the transition point is that it is not a very useful quantity to measure (and also you would need to specify what you mean by that). For example, measuring the forces on an airfoil would be done by suspending the airfoil int he center of the wind tunnel. Typically it will be attached through turntables at the wall and your question may be why we don't measure the boundary layer at the wall here? well, there are always going to be wind tunnel effects (pressure can't expand indefinity like in a cfd simulation, for example, where we can place the boundary condition arbitrarily far away), so there are blockage effects here as well. The boundary layer is just another thing we have to deal with. But just like we have to do some tricks in CFD to get our results correct (for example using wall functions in RANS models and, if you think about it, is quite a crude approach and may not be necessarily physical), wind tunnel experimentalist do user similar tricks as well to prevent turbulent boundary layers and their influence ont he result. You could for example suck away the boundary layer, you could have a moving ground and I am sure there are more techniques available, but I guess you see the point. That is also why it is sometimes very difficult to get a good correltation between wind tunnel results and CFD simulations. If you are a motorsport fan, you may know, for example, that year after year F1 teams are having difficulties understanding the aerodynamics around their car and you often hear that their CFD results don't correlate to their wind tunnel experiments. Again, both approaches introduce approximations and errors and it is our job as CFD / wind tunnel practitioners to ensure their correctness.

So to conclude, yes, turbulent boundary layer can develop in wind tunnels which we may want to address but generally speaking wind tunnels themselves have a very low level of background turbulent intensity and it just becomes a matter of preparing the geometry and the set up in a way that any boundary layer that may develop does not change the actual measurement. The best that we can hope for is that some correlations studies are done (for example, by simulating the geometry within the wind tunnel and not using far-field boundary conditions) or at least that who ever is doing the wind tunnel experiments ensures is noting the influence of any quantities that may have an influence on the final result.

Hi,

Thanks for your detailed thoughts on this topic. Actually, I got to read the similar examples as you told like tunnel blockage, 3D wall effects of the wind tunnel etc. However, recently I did a CFD study on a wind turbine where I used the Reynolds 1/10 of full scale turbine Reynolds number because I had to make a comparison with the LES. During this, when I looked at the pressure distribution and skin friction - the CFD results showed the point of transition however in the experiment I couldn't see any transition phenomena. As soon, I moved to higher Reynolds CFD (same Reynolds as the experiment) I saw the different trend in the pressure distribution and skin friction but same as in the wind tunnel data without indicating any transition. Now, the BIG question is that if there are some wind tunnel effects which made the low Reynolds simulations change from the experimental results. Then why higher Reynolds simulation (using same Re as experiment) match exactly with experiment?
In short, first I conducted low Reynoldssimulations with 1% of free stream turbulence (however experiments were conducted at 7%). Then I performed higher Reynolds simulations with the free stream turbulence same as in the experiments i.e. 7%.

[t.teschner]

Well, if you are looking at LES, i.e scale resolved simulations, boundary conditions become hughly important. The freestream turbulence intensity is a very critical parameter to determine the correct transition point, so all of these things combined can give you quite different results and trends if you don't match Reynolds number and boundary conditions correctly. Don't forget that the Navier-Stokes Equations are non-linear, by resolving most of the turbulent field through LES you will feel these non-linearities even stronger (i.e. they are not necessarily damped through some numerical viscosity) and small differences can have a strong influence on the overall outcome of your simulation.

[mazhar16823]

Originally Posted by t.teschner

Well, if you are looking at LES, i.e scale resolved simulations, boundary conditions become hughly important. The freestream turbulence intensity is a very critical parameter to determine the correct transition point, so all of these things combined can give you quite different results and trends if you don't match Reynolds number and boundary conditions correctly. Don't forget that the Navier-Stokes Equations are non-linear, by resolving most of the turbulent field through LES you will feel these non-linearities even stronger (i.e. they are not necessarily damped through some numerical viscosity) and small differences can have a strong influence on the overall outcome of your simulation.

Thanks but I am a little concerned here " At higher Reynolds RANS simulations (same Reynolds as the experiment) I noticed the SAME trend in the pressure distribution and skin friction as in the wind tunnel data without indicating any transition.

Whereas low Reynolds RANS simulations indicate the location of BL Transition.

Now, the BIG question is that if there are some wind tunnel effects which made the low Reynolds simulations show BL transition. Then why higher Reynolds simulation (using same Re as experiment) matches exactly with experiment?

In short:


Low Reynolds RANS using free-stream turbulence 1% predicts transition location.
Higher Reynolds RANS using free stream turbulence 7% (same as in the wind tunnel) does not predict the transition however the CFD results well agree with the experiment.

[t.teschner]

well, maybe I am not understanding your problem then entirely, but in general, yes, if you have a low enough Reynolds number, then it makes sense to state the transition point (because there is a well-defined transition region) whereas the higher Reynolds number is so energetic that the flow transitions immediately to a fully turbulent flow (technically speaking there is of course still a transition region but it may be rather small in comparison to a meaningful reference length, like the cord length on your wind turbine). So the higher your reynolds number the earlier your flow will transition from laminar to a turbulent flow. Hope this makes sense?

[mazhar16823]

Originally Posted by t.teschner

well, maybe I am not understanding your problem then entirely, but in general, yes, if you have a low enough Reynolds number, then it makes sense to state the transition point (because there is a well-defined transition region) whereas the higher Reynolds number is so energetic that the flow transitions immediately to a fully turbulent flow (technically speaking there is of course still a transition region but it may be rather small in comparison to a meaningful reference length, like the cord length on your wind turbine). So the higher your reynolds number the earlier your flow will transition from laminar to a turbulent flow. Hope this makes sense?

Ah yes yes I got it. I was just mixed up with a lot of stuff Thanks a lot. It's been a nice discussion.

[FMDenaro]

I think that the main misunderstanding is about the concept of transition joined to the Re number. I suppose you should consider the local Re number and discover the location for which the Re is enough increased to drive to the onset of transition.
That has nothing to do with a certain value of the Re number based on the geometry of the wind tunnel.