CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   Airfoil simulation with k-epsilon realizable, calculation of transition possible? (http://www.cfd-online.com/Forums/fluent/96145-airfoil-simulation-k-epsilon-realizable-calculation-transition-possible.html)

level January 13, 2012 11:02

Airfoil simulation with k-epsilon realizable, calculation of transition possible?
 
1 Attachment(s)
Hello CFD community,
I need your help with CFD for my universityproject... We have to simulate a NACA 4418 airfoil with a gurney-flap at windspeeds of 25m/s. The airfoil has a cordlength of 1 meter and because of that we have a reynoldsnumber of 1.700.000.
We simulated a "naked" airfoil with an angle of attack from -4 up to 24 with no problems (k epsilon realizable and 30<y+<60). Now at the simulation of this airfoil with a gurneyflap, which has a lenght of 2% of the cordlength, our simulations suddenly won't converge anymore when the angle of attack is bigger than 14. :eek:
If I understand this right the airflow starts to stall and i have a transition from a laminar to a turbulent flow. Normally the k-epsilon modell has no problems with turbulent and lamniar flows but when the flow starts to change from laminar to turbulent (transition) there are problems, is this correct? :confused:

Two questions:
- Is the k epsilon realizable turbulencemodell the right choice for this problem and is it possible to get a solution with this transitionproblem?
- We are running out of time, what do we have to do to get a converged solution?
i) Use another turbulencemodell, kw sst for example?
ii) Refine the Mesh and pay more attention to y+ values for every new angle of attack? Until now we used one mesh for all analysys which had good y+ values for 2 but y+ of course varies for new angle of attacks.
iii) Change the simulation from steady to transient?
iv) ...

I attached an image for visualization with an arrow which indicates the flowdirction. (unconverged result at 24 angle of attack)
I would appreciate your help very much! Thank you in advance! :)

LuckyTran January 14, 2012 00:24

Quote:

Originally Posted by level (Post 339195)
Hello CFD community,
I need your help with CFD for my universityproject... We have to simulate a NACA 4418 airfoil with a gurney-flap at windspeeds of 25m/s. The airfoil has a cordlength of 1 meter and because of that we have a reynoldsnumber of 1.700.000.
We simulated a "naked" airfoil with an angle of attack from -4 up to 24 with no problems (k epsilon realizable and 30<y+<60). Now at the simulation of this airfoil with a gurneyflap, which has a lenght of 2% of the cordlength, our simulations suddenly won't converge anymore when the angle of attack is bigger than 14. :eek:
If I understand this right the airflow starts to stall and i have a transition from a laminar to a turbulent flow. Normally the k-epsilon modell has no problems with turbulent and lamniar flows but when the flow starts to change from laminar to turbulent (transition) there are problems, is this correct? :confused:

Two questions:
- Is the k epsilon realizable turbulencemodell the right choice for this problem and is it possible to get a solution with this transitionproblem?
- We are running out of time, what do we have to do to get a converged solution?
i) Use another turbulencemodell, kw sst for example?
ii) Refine the Mesh and pay more attention to y+ values for every new angle of attack? Until now we used one mesh for all analysys which had good y+ values for 2 but y+ of course varies for new angle of attacks.
iii) Change the simulation from steady to transient?
iv) ...

I attached an image for visualization with an arrow which indicates the flowdirction. (unconverged result at 24 angle of attack)
I would appreciate your help very much! Thank you in advance! :)

The laminar to turbulent transition would occur even for no angle of attack.

Most turbulence models are not applicable to laminar flows or flows with mixed laminar and turbulent regions. The realizable k-epsilon model is not applicable to laminar flows. If you are certain that there is a laminar to turbulent transition, you should use one of the formulations that can account for this effect, Fluent has transitional models. If the flow is not laminar anywhere then the realizable k epsilon model can be used. Most turbulence models have trouble dealing with stall, even if they converge the results are likely unreliable. Away from walls, the kw sst model is the same as the standard k epsilon model (and not the realizable k epsilon model). It may provide less accurate results compared to the realizable k-epsilon model. It depends on the flow scenario, but you will have to try it out and see.

As long as your y+ satisfies your wall treatment then it is okay to keep using the same mesh, if it does not, then you will need to regenerate the mesh. What wall function approach are you using?

Is the picture zoomed in around the airfoil? Is there more to the computational domain?

Using a transient simulation is not recommended for beginners if you are having trouble with a steady state solution.

level January 14, 2012 11:29

2 Attachment(s)
I'm using the standard wall function with the k epsilon realizable model and my computational domain is a c grid with 12.5 cord length infront and 20 cord length after the airfoil .
I'm not realy sure if my problem is because of stall and if i have transition at all but this is my only explanation why i get results up to 14 and after that my simulation has a complete different convergence. At 14 the flow is parallel to the airfoil and at 16 i get a backflow and a small eddy at the top end of the airfoil which moves to the front when i enlarge teh angle of attack. Is this seperation also a transition...?
When i am simulating the 3% gurneyflap this behaviour already starts at an angle of attack of 14. This would make sense because the bigger gurneyflap produces an earlier seperation...
So my only solution would be to use a transitional model for example the k-kkl-omega or the SST-model?
I attached my residualplots at 14 and 16...

LuckyTran January 14, 2012 14:46

Quote:

Originally Posted by level (Post 339306)
I'm using the standard wall function with the k epsilon realizable model and my computational domain is a c grid with 12.5 cord length infront and 20 cord length after the airfoil .
I'm not realy sure if my problem is because of stall and if i have transition at all but this is my only explanation why i get results up to 14 and after that my simulation has a complete different convergence. At 14 the flow is parallel to the airfoil and at 16 i get a backflow and a small eddy at the top end of the airfoil which moves to the front when i enlarge teh angle of attack. Is this seperation also a transition...?
When i am simulating the 3% gurneyflap this behaviour already starts at an angle of attack of 14. This would make sense because the bigger gurneyflap produces an earlier seperation...
So my only solution would be to use a transitional model for example the k-kkl-omega or the SST-model?
I attached my residualplots at 14 and 16...

Separation and transition are two completely different physical phenomenon. Separation refers to the lift-off of the boundary layer from the surface. Transition refers to the development of the boundary layer from laminar to turbulent regime.

Is your boundary layer laminar at the leading edge or always turbulent? That should be known beforehand. With the current setup, it is likely you are turbulent everywhere and there is no transition. There is no need to use a transitional model if there is not any transition.

Turbulence models deal poorly with separated flows. You can see the residuals are worst for your turbulence model (k - epsilon). Your residuals for the continuity and momentum equations are not that bad.

Check the y+ values of your mesh and make sure they are still >30 everywhere since you are using standard wall functions and make sure that your use of standard wall functions is still appropriate. Is there any particular reason why you are using standard wall functions? With a sufficiently fine mesh, the enhanced wall function approach will likely produce more accurate results.

level January 15, 2012 10:33

Ok, I will check my mesh and the y+ values again.
I think that I have a full turbulent flow, because the reynoldsnumber is 1.700.000...
I'm using the standard wall function to save time during the simulation due to a coarser mesh.

LuckyTran January 15, 2012 12:24

Quote:

Originally Posted by level (Post 339393)
Ok, I will check my mesh and the y+ values again.
I think that I have a full turbulent flow, because the reynoldsnumber is 1.700.000...
I'm using the standard wall function to save time during the simulation due to a coarser mesh.

That is not what determines whether the flow is turbulent or not. Your Reynolds number is 0 at the leading edge and increases as the boundary layer grows (the boundary layer thickness is 0 at the leading edge because there is no boundary layer). The local Reynolds number (based on the local boundary layer thickness if you want to be strict) is what determines where transition will occur. If the boundary layer is initially laminar, then a transition to turbulence may occur. If the boundary layer is initially turbulent, then it is already turbulent. It is already in your definition of the problem and its setup.

level January 15, 2012 13:32

Thank you for your further explanations!
I just noticed that my main problem are the y+ values which dropped slightly under 30 for bigger angles of attack... Ok, it seems that I have to build a lot more different meshes for the different angles. :(


All times are GMT -4. The time now is 11:55.