[chandra shekhar pant]

Dear Foamers,


I am trying to simulate a 3D aerofoil case, in which the mesh is generated by the gmsh. The overall geometry, boundary condition, fvschmes, fvsolution and plot of convergence are attached herewith.

I am unable to get the decreasing residual for Uz and Pressure, rest parameters seems reasonable. Is this because of the improper meshing in the z-direction? Since I am not sure if the meshing is done correctly in the z direction by the gmsh. Although the checkMesh looks ok (also attached herewith).



Any suggestion or comment is highly welcomed.



Thanks a lot!

[chandra shekhar pant]

The checkMesh is attached herewith.

[xpqiu]

Originally Posted by chandra shekhar pant

The checkMesh is attached herewith.

Totally tetrahedra mesh? Without prism layer mesh? May be you can try leastSquare gradient calculation by changing the default gradient scheme from "Gauss linear" to "leastSquare".

[chandra shekhar pant]

Thanks Qiu Xiaoping a lot for the help.



I am trying to implement this, to start with I would say that instead of "leastSquare" this is "leastSquares" :-) I will update you once the sims are done.

[chandra shekhar pant]

Hello Qiu Xiaoping,

I had tried to modify the mesh and using your inputs tried the case. The residual indeed is falling but not as quick as expected. Surprisingly, if I use the "checkMesh" utility it shows the mesh is OK, but while using " checkMesh -allTopology -allGeometry ' it shows that Failed 1 mesh checks, probably because of Cells with small determinant (< 0.001) found, number of cells: 210. The complete report and the residual plots are attached herewith. Any comment about this ?


Thanks a lot for your help.

[xpqiu]

When activate "-allTopology" and "-allGeometry", checkMesh will do more checks, and for incompressible steady-state simulation, "Cells with small determinant (< 0.001) found" is not a big issue. Did you run potentialFoam to initialize the fields? If not, try it and this may help.

[chandra shekhar pant]

Hi Qiu Xiaoping,



Thanks for your help. Will this issue be a problem for unsteady case like PIMPLE or INTERPHASECHANGEFOAM?



No, I haven't used potential Foam to initialize the fields, I will do that and update.


Thanks again for your kind help.

[chandra shekhar pant]

Hello Qiu Xiaoping,



Using your inputs I have initialized my SIMPLE case using the potentialFoam, I had added lines in the fvsolution and put the initial Phi in the 0 folder. To start with, it was giving errors, since something was missing in the fvSolution, but eventually I make it run OK.

Please find the fvSolution and the initial Phi herewith. But still I could not get a good convergence for the Uz and pressure, particularly the Uz, is there something bad in the mesh for Uz? Is there a way to figure it out or something I am missing in fvSchmes, fvSolution etc?

Thanks for your time and help.

[HPE]

get rid of tetrahedron mesh.

[chandra shekhar pant]

Thanks HPE for your kind help. But if you check the checkMesh report then there are no tetrahedron cells only prisms are there. So what you recommend which kind of cells should I have to get better convergence?
Also I am using gmsh how to control these, I have no idea, if I use Quad cells then it throws an error and my meshing doesn't complete. Any idea in gmsh will be very helpful to me.
Thanks a lot for your time and help.

[chandra shekhar pant]

Oh Sorry I think you saw the previous report. This is the updated report. Please find the checkMesh report and also of checkMesh -allTopology -allGeometry reports herewith.

[HPE]

Never used GMSH, I am afraid. There was a OpenFOAM mesh tutorial in OpenFOAM Wiki, exemplifying mesh generation with quad elements in GMSH.

I suggest you to use hex elements, for which you can either use snappyHex or even blockMesh of OpenFOAM since your geometry is very simple.

I meant also triangular prism elems by tetrahedron as both are similar and problematic for OpenFOAM simulations afaik. May be foam-extend can handle tetra meshes. But need to check.

[chandra shekhar pant]

Thanks a lot HPE.
Since, I am having the solid boundary (aerofoil) in which I have to define some patches inside that solid boundary. If I import the aerofoil as single stl file in SHM, the SHM will treat it as a single wall, but if I want some other patches (like slotting, blowing, suction etc.) to be included inside that wall/aerofoil, then I guess I won't be able to do that in SHM (please correct me ). Therefore I am using gmsh.
In most of the gmsh tutorials I have seen that they are meshing in 2D while I need the 3D meshing (by extruding in 3D). I haven't found any tutorial for 3D meshing to be used as input for CFD, however there are some which shows gmsh utility for solid meshing. Please suggest some reference if you find any.



I think you are suggesting that quad mesh or hexahedral (meshes formed by blockmesh and SHM) mesh are good for OpenFoam.


Thanks a lot for your time and help.

[HPE]

>> some other patches (like slotting, blowing, suction etc.) to be included inside that wall/aerofoil, then I guess I won't be able to do that in SHM (please correct me )

It is completely possible. I personally have meshed an entire wind turbine containing many STLs constituting the entire turbine seamlessly when combined. Might be painful for a novice, or enjoyful. I found it joyful (kidding).

>> https://openfoamwiki.net/index.php/2...ial_using_GMSH

Might this help?: https://openfoamwiki.net/index.php/2...ial_using_GMSH

>> I think you are suggesting that quad mesh or hexahedral (meshes formed by blockmesh and SHM) mesh are good for OpenFoam.

Yes, I believe so.

[chandra shekhar pant]

Hello HPE,


Many thanks for this information. I know that I am just at a starting point of learning the SHM but I am just excited to know how this works..


If I have an aerofoil (which is a complete stl file), and suppose at the stagnation point I want a jet, primarily I want to introduce a patch (or boundary condition) with some outward velocity at the aerofoil itself, how it could be done in the SHM? It will be a great thing to compare the convergence from the two meshes (one from gmsh and other from SHM).



Thanks for your help and time.

[HPE]

>> how it could be done in the SHM?

It is very broad question, I'm afraid. Don't think I can give an answer on that. Might others give some insight?

[chandra shekhar pant]

Hi HPE,



As per your suggestion, I tried to change my meshing from prism/tetrahedra to hexahedra, but still the convergence is an issue. Please find the residual and the latest checkMesh log herewith.

I would be happy to implement some other suggestion/comments from you.


Just wondering if this is the final fate of convergence, since the problem could be inherently unsteady. Does that makes sense?
Many thanks for your help and time.

[simrego]

Hi!


Don't trust only the residuals. Monitor some physical quantities too. For example drag, inlet pressure, velocity at a point, etc... Also 60 interation is nothing. Let the solver run.
Also you can try 1st order shcemes, and if you have a converged solution switch to second order.

[chandra shekhar pant]

Hello,



Thanks a lot for your reply. But is it ok to say that, if the flow is inherently non-steady then the residual won't converge? Could you please comment over this?


Many thanks for your time and help.

[simrego]

Most of the flows are inherently non-steady, but most of the time you can reach a steady-state solution which is kind of a "time-averaged" solution. But the reason for your "non converged case" is that you just give up after 60 iterations which is nothing... And in your case you should have a steady state solution...
BTW i think you have a really slow airplane.