[PeterShi]

Originally Posted by Flowkersma

Hi Peter,

Interesting results. I wonder from which experiment are these results? Is the boundary layer tripped?

My divergence schemes are following:

div(phi,U)          bounded Gauss linearUpwind grad(U);
div(phi,k)          bounded Gauss upwind;
div(phi,omega)      bounded Gauss upwind;
div(phi,gammaInt)   bounded Gauss upwind;
div(phi,ReThetat)   bounded Gauss upwind;
div((nuEff*dev2(T(grad(U))))) Gauss linear;

From my experience there is no big difference in using first and second order schemes for the turbulence quantities. However, first order converges much faster and therefore I am using it. To find out the cLmax I guess there is no better way to finding it out iteratively.

Best,
Mikko

Hi Mikko,

To see more details of this paper, please download the paper below:
http://arc.aiaa.org/doi/pdfplus/10.2514/6.1997-1932.

Usually, I will utilize following divergence first.
divSchemes
{
default none;
div(phi,U) bounded Gauss linearUpwindV grad(U);
div(phi,k) bounded Gauss linearUpwind limited;
div(phi,omega) bounded Gauss limitedLinear 1;
div(phi,gammaInt) bounded Gauss linear;
div(phi,ReThetat) bounded Gauss linear;
div((nuEff*dev2(T(grad(U))))) Gauss linear;
}

If my calculation is divergent, I would like to turn to:
div(phi,U) bounded Gauss linearUpwind grad(U);
div(phi,k) bounded Gauss upwind;
div(phi,omega) bounded Gauss upwind;
div(phi,gammaInt) bounded Gauss upwind;
div(phi,ReThetat) bounded Gauss upwind;
div((nuEff*dev2(T(grad(U))))) Gauss linear;
After thousand iterations, maybe I will reuse my first div scheme again.

Near separation, neither of schemes can make my case convergent, I have to employ:
div(phi,U) bounded Gauss upwind;
div(phi,k) bounded Gauss upwind;
div(phi,omega) bounded Gauss upwind;
div(phi,gammaInt) bounded Gauss upwind;
div(phi,ReThetat) bounded Gauss upwind;
div((nuEff*dev2(T(grad(U))))) Gauss linear;

In general, when I use above scheme, my case will be convergent, and CL decreases dramatically. Even when you use this div scheme on medium AOAs, which are far from stall, the CL ALSO decreases dramatically (You can have a try). Moreover, the above scheme is the only option that can continue my program near stall.

As you can see, my results look like well in predicting separation, however, I am not sure whether the separation does occur at that AOA, or the aforementioned scheme causes the stall of CL.

What do you think?

BTW, do you have any ideas of the incapability of the transitional model in predicting CL&CM well in my cases? (CD is good)

Best,
Peter

[PeterShi]

Originally Posted by Flowkersma

Hi Peter,

Interesting results. I wonder from which experiment are these results? Is the boundary layer tripped?

My divergence schemes are following:

div(phi,U)          bounded Gauss linearUpwind grad(U);
div(phi,k)          bounded Gauss upwind;
div(phi,omega)      bounded Gauss upwind;
div(phi,gammaInt)   bounded Gauss upwind;
div(phi,ReThetat)   bounded Gauss upwind;
div((nuEff*dev2(T(grad(U))))) Gauss linear;

From my experience there is no big difference in using first and second order schemes for the turbulence quantities. However, first order converges much faster and therefore I am using it. To find out the cLmax I guess there is no better way to finding it out iteratively.

Best,
Mikko

Hi Mikko,

More questions for you:
1. Number of lambda iterations exceeds maxLambdaIter(10).
Now, have you solved this issue yet?

2.Have you conducted calculations of the compressible solver?
Previously, I thought it will perform better, but it is even worse. I do not know why. And I used the compressible solver to simulate the flow around simple case NACA0012, and the result is good.

Unsteady simulations are underway.

Best,
Peter

[Flowkersma]

Thanks for the paper. They trip the boundary layer of the main element at x/c=0.125. Have you compared this with your simulation? I think your simulations agree well with the experiments and also capture the critical point well. However, I am not sure if transitional model should improve your results?

Usually, I will utilize following divergence first.
divSchemes
{
default none;
div(phi,U) bounded Gauss linearUpwindV grad(U);
div(phi,k) bounded Gauss linearUpwind limited;
div(phi,omega) bounded Gauss limitedLinear 1;
div(phi,gammaInt) bounded Gauss linear;
div(phi,ReThetat) bounded Gauss linear;
div((nuEff*dev2(T(grad(U))))) Gauss linear;
}

If my calculation is divergent, I would like to turn to:
div(phi,U) bounded Gauss linearUpwind grad(U);
div(phi,k) bounded Gauss upwind;
div(phi,omega) bounded Gauss upwind;
div(phi,gammaInt) bounded Gauss upwind;
div(phi,ReThetat) bounded Gauss upwind;
div((nuEff*dev2(T(grad(U))))) Gauss linear;

Have you compared the results of these two scheme settings? Are there significant difference in results? I would not use the first order upwind scheme for velocity.

Unfortunately, I have not find the reason for the maxLambdaIter(10). Also, I have no experience in using OpenFOAM with compressible flows.

[PeterShi]

Originally Posted by Flowkersma

Thanks for the paper. They trip the boundary layer of the main element at x/c=0.125. Have you compared this with your simulation? I think your simulations agree well with the experiments and also capture the critical point well. However, I am not sure if transitional model should improve your results?




Have you compared the results of these two scheme settings? Are there significant difference in results? I would not use the first order upwind scheme for velocity.

Unfortunately, I have not find the reason for the maxLambdaIter(10). Also, I have no experience in using OpenFOAM with compressible flows.

Hi Mikko,

All right. That is my point: Truly KOmegaSSTLM does a good job, but it did not improve the results of KOmegaSST "significantly", which was stated in many papers.

How to get the transition point and visualize it? I mean, I only set the value for Retheta and gamma, but I do not know where the transition occur.

For the above quoted divSchemes, there is only minor difference of results. However, setting upwind scheme is more easily to be convergent. But there is a big gap between the results using first order or second order schemes for velocity. But as I said, that is the only way that can converges my program at large AOAs. From the vector profile, I observed that the flow does separate in the cove region, which confirms me that my computation is reasonable.

Using compressible and unsteady solvers might improve our results.

Best,
Peter

[Flowkersma]

My point is that you have two different transition modeling approaches in your arsenal. One assumes that the boundary layer is always fully turbulent (kOmegaSST) and the other assumes free transition (kOmegaSSTLM). However, in the experiment the boundary layer is tripped at a certain point which is not at the beginning of the airfoil. So both assumptions are wrong. Now the question is, which one simulates better the experimental setup? Your simulation results indicate that always fully turbulent boundary layer is a better assumption in this case. However, in my opinion the transition model cannot be expected to improve the results in this case unless you really know that it should really simulate your case better. If you had experimental results from an airfoil with free transition then I would expect kOmegaSSTLM to give better results.

That being said, the paper by Langry and Menter state

The present transition model is built on a transport equation for intermittency, which can be used to trigger transition locally.

and

In both the experiment and in the CFD prediction, the boundary layer was tripped near the plate leading edge. In the CFD computation, this was accomplished by injecting a small amount of turbulent air into the boundary layer with a turbulence intensity of 3%. The same effect could have been accomplished with a small step or gap in the CFD geometry.

So it should be possible to trigger the transition at the right point which should improve the results.

Look at the the intermittency field to estimate where the transition occurs.

[PeterShi]

Originally Posted by Flowkersma

My point is that you have two different transition modeling approaches in your arsenal. One assumes that the boundary layer is always fully turbulent (kOmegaSST) and the other assumes free transition (kOmegaSSTLM). However, in the experiment the boundary layer is tripped at a certain point which is not at the beginning of the airfoil. So both assumptions are wrong. Now the question is, which one simulates better the experimental setup? Your simulation results indicate that always fully turbulent boundary layer is a better assumption in this case. However, in my opinion the transition model cannot be expected to improve the results in this case unless you really know that it should really simulate your case better. If you had experimental results from an airfoil with free transition then I would expect kOmegaSSTLM to give better results.

That being said, the paper by Langry and Menter state

and

So it should be possible to trigger the transition at the right point which should improve the results.

Look at the the intermittency field to estimate where the transition occurs.

Hello Mikko,

Got it. Thanks, you are so inspiring.

Best,
Peter

[Chrisi1984]

Hello Peter,

I also tried to compile the kOmegaSSTLMmodel in OpenFOAM 4.1. But I did not succeed up to now. I am using the code that is linked in the thread and which is also the current code in OpenFOAM 5.

Can you please describe or even provide the code with that you compiled the model in OpenFOAM 4.1?

Thanks in advance!

Kind regards
Chrisi