Hi to you all!
This is my first posting on the forum and maybe the problems I encountered are somehow "stupid".
I started using Code_Saturne almost half a year ago and I can say I can manage to make it work in some problems. Now I want to simulate a flow with thermal phenomena included and I am stuck a little bit on y+. From what I have seen in the manual for high-Reynolds (I try to use the k-epsilon model) number I must have a y+ between 30 and 100, but for my case I obtain much higher values. I even calculated the needed y(distance to the wall) for a y+ ~ 90 and I obtained something like 0.5mm which I think is to small to have a mesh for an object of some meters long. In the manual it is said that the scalable-wall function should solve the problem if you do not have thermal phenomena involved, not my case.
So can you please guide me:
1. scalable-wall function is a solution?
2. how can I have a mesh refined let say at the wall to obtain the needed y+ and bigger in the rest of the volume? (I am not sure if in this case I do not obtain the "mesh to refined at the wall" message.
3. any other suggestion?
Thank you in advance for the help and please excuse my stupid questions.
Y+ problem
Forum rules
Please read the forum usage recommendations before posting.
Please read the forum usage recommendations before posting.
-
Yvan Fournier
Re: Y+ problem
Hello,
I'll let turbulence specialists answer for 1), but I'm not sure scalable wall functions are enough (recent work on thermal wall functions wont' be available in a full version before the end of the year, but some patches might be; still, I'm not sure here).
2) really depends on the meshing tool you are using. If you are using SALOME, the viscous layer option in the recently released 6.3 version should be a big help. If you are using major CFD oriented commercial tools, adding a layer of thinner cells near a boundary is also always an option, though how to do it will be tool-specific. If you are using a meshing tool which is more structural mechanics oriented, this may be much more difficult.
If you run your calculation Did you visualize the postprocessing output of y+ on the domain boundary ? With a complex flow, you may have local high or low values of y+, with an acceptable value in most places, so if you have not done so already, you may want to run the calculation without thermal effects first to estimate and visualize the "average" y+.
Best regards,
Yvan
I'll let turbulence specialists answer for 1), but I'm not sure scalable wall functions are enough (recent work on thermal wall functions wont' be available in a full version before the end of the year, but some patches might be; still, I'm not sure here).
2) really depends on the meshing tool you are using. If you are using SALOME, the viscous layer option in the recently released 6.3 version should be a big help. If you are using major CFD oriented commercial tools, adding a layer of thinner cells near a boundary is also always an option, though how to do it will be tool-specific. If you are using a meshing tool which is more structural mechanics oriented, this may be much more difficult.
If you run your calculation Did you visualize the postprocessing output of y+ on the domain boundary ? With a complex flow, you may have local high or low values of y+, with an acceptable value in most places, so if you have not done so already, you may want to run the calculation without thermal effects first to estimate and visualize the "average" y+.
Best regards,
Yvan
-
Eduard Pauna
Re: Y+ problem
Hello, Yvan!
Thanks for your quick answer.
I was using Salome for geometry and meshing (sorry for not mentioning it) version 5.1.5 and now I am investigating the possibilities of 6.3.0. I tried the viscosity layers option, manage to obtain them but I encountered a problem when dealing with the mesh. It seems that in the latest version of Salome the export med format is 3 and Code_Saturne seems not to recognize it.
I am also investigating the post processing of y+ (only on the boundaries from what I discovered) and I will let you know of the progress.
If you or someone else has any suggestion please tell me.
All the best,
Eduard
Thanks for your quick answer.
I was using Salome for geometry and meshing (sorry for not mentioning it) version 5.1.5 and now I am investigating the possibilities of 6.3.0. I tried the viscosity layers option, manage to obtain them but I encountered a problem when dealing with the mesh. It seems that in the latest version of Salome the export med format is 3 and Code_Saturne seems not to recognize it.
I am also investigating the post processing of y+ (only on the boundaries from what I discovered) and I will let you know of the progress.
If you or someone else has any suggestion please tell me.
All the best,
Eduard
-
Yvan Fournier
Re: Y+ problem
Hello Eduard,
Sorry, I forgot that Code_Saturne 2.0.1 does not recognize MED 3. Patch release 2.0.2 should be released within a few days, and can be compiled with either MED 3 or MED 2.6 (as well as CGNS 3.1 in addition to CGNS 2.5), so this will solve the issue (if it is delayed too much, I will post at least the preprocessor on this forum).
Best regards
Yvan
Sorry, I forgot that Code_Saturne 2.0.1 does not recognize MED 3. Patch release 2.0.2 should be released within a few days, and can be compiled with either MED 3 or MED 2.6 (as well as CGNS 3.1 in addition to CGNS 2.5), so this will solve the issue (if it is delayed too much, I will post at least the preprocessor on this forum).
Best regards
Yvan
-
César Vecchio
Re: Y+ problem
Eduard, be careful with the way you calculate Y+. Most of ad hoc calculators that you'll find on the web (there was one made by NASA) are base on a fully turbulent flat plate at zero AOA. I used that formula myself in a little helper code I put on this forums under the Examples section, you can check it out if you wish, but it hapened to me that at moderate or low Reynolds the estimated cell distance was smaller than actually needed.
Perhaps if you can run something like an Eppler or Drela codes (yeah, the ones used for airfoils) you can have a better estimate of the needed cell distance as long as there's not dettached flow. JavaFoil for instance already offers this info.
One suggestion: Salome, Gmsh and other meshers work easily with UNV, a format fully compatible with Saturne. It produces larger mesh files as data is stored as ASCII and not binary, but that's not usually trouble.
Perhaps if you can run something like an Eppler or Drela codes (yeah, the ones used for airfoils) you can have a better estimate of the needed cell distance as long as there's not dettached flow. JavaFoil for instance already offers this info.
One suggestion: Salome, Gmsh and other meshers work easily with UNV, a format fully compatible with Saturne. It produces larger mesh files as data is stored as ASCII and not binary, but that's not usually trouble.
-
Eduard Pauna
Re: Y+ problem
Thank you all for the answers. They helped a lot and for the moment I am a step forward.
If someone is in the same trouble as I was, I note here how I solve the problem.
First I used the viscosity layers (much thinner than the rest of the mesh) option in Salome 6, then I exported the mesh in UNV format and recompiled the cs_preprocessor (see the solution https://code-saturne.info/products/code-saturne/issues/64 ) in order to be able to read the UNV file and from that moment all went smooth. For an estimated y+ (using some calculator, Cesar I know they are not perfect) Code_saturne gave me a similar value and the thermal phenomena seems to be more accurate now (no more scalable wall functions, just the two scales one).
But still I have one more question - how many viscosity layers is the right way to choose? In the test I used 3 layers on the thermal wall and also on the adiabatic walls but I am not sure if it should not be more.
Best wishes,
Eduard
If someone is in the same trouble as I was, I note here how I solve the problem.
First I used the viscosity layers (much thinner than the rest of the mesh) option in Salome 6, then I exported the mesh in UNV format and recompiled the cs_preprocessor (see the solution https://code-saturne.info/products/code-saturne/issues/64 ) in order to be able to read the UNV file and from that moment all went smooth. For an estimated y+ (using some calculator, Cesar I know they are not perfect) Code_saturne gave me a similar value and the thermal phenomena seems to be more accurate now (no more scalable wall functions, just the two scales one).
But still I have one more question - how many viscosity layers is the right way to choose? In the test I used 3 layers on the thermal wall and also on the adiabatic walls but I am not sure if it should not be more.
Best wishes,
Eduard
-
Yvan Fournier
Re: Y+ problem
Hello,
2.0.2 has just been installed on EDF machines today, and should be posted tomorrow.
Best regards,
Yvan Fournier
2.0.2 has just been installed on EDF machines today, and should be posted tomorrow.
Best regards,
Yvan Fournier