[NightRippers]

Hello

I am new with Openfoam and fluid dynamic.

I am trying to simulate flow around an airfoil with RAS Kepsilon turbulence model. For that i decided to use SIMPEFOAM (From what i read, its good). I designed a mesh (Mesh is ok with checkMesh) and tried to configure it.

I would like to do a angle of atack of approximatly 9°, so i decided to put a horizontal velocity of 19 and vertical one of 3.

I put the pressure to 83333 (I read that it's Pressure/Density)

Now i am trying to simulate it but, i obtain some strange results without knowing why, if you can help me that could be great please

I put some documents in attachements

Have a good day
Florian

[cryabroad]

Please be more specific about your problem, what strange results did you observe?

From the files you attached, there are a couple of things that can be improved.
1. You specified a fixed pressure and velocity at the inlet at the same time, which is not normal. Try fixed velocity at the inlet (with pressure being zeroGradient), and fixed pressure at the outlet (with velocity being zeroGradient).
2. You are solving a transient simulation, as you specified Euler for ddtSchemes. I don't think you need this, a steady RANS would be good enough. Plus, and which is surprising is, simpleFoam by design is a steady state solver, I don't think it should be used for trasient calculations.
3. I think your boundary conditions may not be that great. You may need to use wall functions for turbulent variables (k, epsilon and nut), but typically these settings are likely to affect the accuracy only.

[NightRippers]

Originally Posted by cryabroad

Please be more specific about your problem, what strange results did you observe?

From the files you attached, there are a couple of things that can be improved.
1. You specified a fixed pressure and velocity at the inlet at the same time, which is not normal. Try fixed velocity at the inlet (with pressure being zeroGradient), and fixed pressure at the outlet (with velocity being zeroGradient).
2. You are solving a transient simulation, as you specified Euler for ddtSchemes. I don't think you need this, a steady RANS would be good enough. Plus, and which is surprising is, simpleFoam by design is a steady state solver, I don't think it should be used for trasient calculations.
3. I think your boundary conditions may not be that great. You may need to use wall functions for turbulent variables (k, epsilon and nut), but typically these settings are likely to affect the accuracy only.

Hello,

Thanks for your reply,

In the picture i send, no matter the time it's stay the same along time, without any turbulence or vortex, i will try your advice

[NightRippers]

Originally Posted by cryabroad

Please be more specific about your problem, what strange results did you observe?


2. You are solving a transient simulation, as you specified Euler for ddtSchemes. I don't think you need this, a steady RANS would be good enough. Plus, and which is surprising is, simpleFoam by design is a steady state solver, I don't think it should be used for trasient calculations.

Do you think that pisoFoam is better in my case ?

[NightRippers]

Hello again,

I tried with pisoFoam to solve this problem. My goal is to study turbulence, to see where in the wing the flow start creating vortex to implement active flow control.

So now, i tried with pisoFoam, but i dont have any turbulence over the wing and i dont found why. If you have some advice for me. I don't understand how to calculate k,epsilon and nut values, i put some "random" ones. I see and i have red the subject about htat but i see a lot of formulas with a lot of datas that i don't have (Or i don't know where to find it).

Thanks for your help

Have a good day
Florian

[mzzmrt]

What are the Re and the yPlus values?

[cryabroad]

I don't think you can observe turbulence by using RANS, if by turbulence you mean vortices in the velocity contour. Maybe by using a very small time step you will be able to see some, but RANS results are results that are averaged in time. To "see" turbulence you may need to go to LES.

You can check the k and epsilon distribution in your domain, those should indicate turbulence. And nut, which is the turbulent viscosity. Note that in RANS the ratio between nut and nu should be in the order of ~10 or higher.

As for boundary conditions at the inlet, please refer to this . For wall functions, just search in the forum, you should get a lot of information. There is a detailed documentation that talks about wall functions from Chalmers, just google "wall function chalmers".

[NightRippers]

Originally Posted by mzzmrt

What are the Re and the yPlus values?

Hello,

I didn't calculated y+, for the Re i calculated 1 200 000 for the air with a density of 1.2 and a speed of 20m/s

[NightRippers]

Originally Posted by cryabroad

I don't think you can observe turbulence by using RANS, if by turbulence you mean vortices in the velocity contour. Maybe by using a very small time step you will be able to see some, but RANS results are results that are averaged in time. To "see" turbulence you may need to go to LES.

You can check the k and epsilon distribution in your domain, those should indicate turbulence. And nut, which is the turbulent viscosity. Note that in RANS the ratio between nut and nu should be in the order of ~10 or higher.

As for boundary conditions at the inlet, please refer to this . For wall functions, just search in the forum, you should get a lot of information. There is a detailed documentation that talks about wall functions from Chalmers, just google "wall function chalmers".

Hello, Thanks for your reply,

To calculate k and epsilon i need I and l but i dont know how to determine them, in my case, I its a random value between 1 and 5 ? and i dont understand what is l (the turbulence length scale) and how to determine it.

I will try to do an LES simulation when epsilon and k and nu will be exactly know.

Thanks for your help ,
Florian

[mzzmrt]

For these conditions, it's better to use SpalartAllmaras model instead of kEpsilon and simpleFoam solver for various reasons so you can familiarize yourself to the case setup.

You have to resolve the full boundary layer so yPlus<1 must be achieved with enough grid points in the boundary layer.

Then you will need real 3D mesh (some grid resolution in the spanwise direction) because the real turbulence has a fully 3D nature and to work for the active flow control, 3D solution is required.

On the other hand I don't think RANS/URANS solutions will do the job for such a case, as cryaboard said above.

But LES will require very high computational power in this Re, if you have it no problem. Otherwise hybrid RANS/LES model may (or may not) help. If you start with SA model, you can adapt your case to DES easily.

[NightRippers]

Originally Posted by mzzmrt

For these conditions, it's better to use SpalartAllmaras model instead of kEpsilon and simpleFoam solver for various reasons so you can familiarize yourself to the case setup.

You have to resolve the full boundary layer so yPlus<1 must be achieved with enough grid points in the boundary layer.

Then you will need real 3D mesh (some grid resolution in the spanwise direction) because the real turbulence has a fully 3D nature and to work for the active flow control, 3D solution is required.

On the other hand I don't think RANS/URANS solutions will do the job for such a case, as cryaboard said above.

But LES will require very high computational power in this Re, if you have it no problem. Otherwise hybrid RANS/LES model may (or may not) help. If you start with SA model, you can adapt your case to DES easily.

Hello again

Thanks for your help, as it's an introduction to Active Flow control, my teacher ask us to use K epsilon solver in 2D (To do it easier i suppose ?)
I have a powerfull computer (Gaming one ahah but maybe it's not enough, but i can let it calculate a long time)

I don't use SIMPLEFOAM now but PISOFOAM, is it better ?


Thanks you very much for your help

Sorry if my questions can be idiots ones, i am discovering GMSH, openfoam and active flow control ahah

Have a good day
Florian

[mzzmrt]

Well, for LES or DES you will need pisoFoam solver but beginnig with simpleFoam may be easier to start with.


Your computer may be good for gaming but can not do a LES or DES study with this Re though. You will need say 1.000 times or more computational power.

[NightRippers]

Originally Posted by mzzmrt

Well, for LES or DES you will need pisoFoam solver but beginnig with simpleFoam may be easier to start with.


Your computer may be good for gaming but can not do a LES or DES study with this Re though. You will need say 1.000 times or more computational power.

In teacher exemple, he is using pisofoam and i read that SIMPLEFOAM is a steady state solver that is not really good for this exemple it's why i am using pisofoam, do you think i can use it ?

To decrease the Re, i can maybe reduce flow speed ? But that can affect results maybe ?

My teacher said taht the y+ values in reality should by less or egal to 1 but in our case (considering our computers) we can take one bigger, i will try to calculate it

[cryabroad]

Typical values for I (turbulent intensity) is 5%. The exact number shouldn't matter that much though. And typical values for l (turbulent length scale) is 7% of the flow characteristic length. Again, the exact number shouldn't matter that much, 10%, 15%, just pick one. What is your nut/nu ratio?

I think simpleFoam is a good solver to start with, transient simulations requires much more effort than steady-state ones. RANS results should provide enough information for you at this stage. Quick point, to get y+ I think you can just type yPlus in the command window. Depending on the version you are using it may or may not work, but even if it doesn't work the error message should tell you what is the right way of doing it.

By the way it's totally all right to ask "idiot" questions, as long as the questions are meaningful and specific. I mean, we all have went through this stage, and I myself still ask a lot of questions.

[NightRippers]

Originally Posted by cryabroad

Typical values for I (turbulent intensity) is 5%. The exact number shouldn't matter that much though. And typical values for l (turbulent length scale) is 7% of the flow characteristic length. Again, the exact number shouldn't matter that much, 10%, 15%, just pick one. What is your nut/nu ratio?

I think simpleFoam is a good solver to start with, transient simulations requires much more effort than steady-state ones. RANS results should provide enough information for you at this stage. Quick point, to get y+ I think you can just type yPlus in the command window. Depending on the version you are using it may or may not work, but even if it doesn't work the error message should tell you what is the right way of doing it.

By the way it's totally all right to ask "idiot" questions, as long as the questions are meaningful and specific. I mean, we all have went through this stage, and I myself still ask a lot of questions.

i choose to put 10 as you said to me before
I use openfoam v6, i will check that thanks

[NightRippers]

Hello,

I have an other question. why it's not correct to set a speed condition and a pressure condition at the inlet please ?

Have a good day
Florian