[jnanabrao]

Dear all,
I'm trying to start on this higher order opensource solver called "Flexi" that requires mesh to be prepared in hdf5 format by the HOPR (Higher Order Preprocessor) software. I executed a basic tutorial on importing a cylinder from a gmsh file and outputting a hdf5 mesh with HOPR which worked without issues. I tried the same steps with a much simpler gmsh lid driven cavity 2D mesh when I got an error that confused me.


On trying to read in a basic gmsh file, I get the error "Keyword $Nodes not found in file." This makes no sense to me as my .msh file clearly has "$Nodes" in it. Can someone help me out with this?


The mesh and parameter.ini files are attached. I'm using HOPR V2.1 and gmsh 4.4 for Linux.

[jnanabrao]

Apparently, the extensions are not accepted directly so I had to attach the zip files.

[Eifoehn4]

Dear jnanabrao,

i think you are using the wrong Gmsh export option. HOPR currently supports "Version 2 ASCII". In Gmsh choose (skip all other options):

File Export *.msh Version 2 ASCII ...

The exported ASCII file should look like:

$MeshFormat
...
$EndMeshFormat
$PhysicalNames
...
$EndPhysicalNames
$Nodes
...
$EndNodes
$Elements
...
$EndElements

Moreover, you have to define physical entities for volumes and surfaces.

Note that:

Currently FLEXI is based on pure hexahedral meshes with Mortar interfaces.

Regards

[jnanabrao]

Hi Eifoehn4,
I did realize that the error was caused by the .msh version. I am new to gmsh and did not have any idea about the option of the exporting to ASCII 2 version. I instead installed the gmsh 2.16 version and am now facing a different problem. One of my defined boundaries is not being recognized by HOPR while 3 others are. This mesh is the lid-driven cavity in 3D in a sense. The files are attached and I hope you can take a look.



Can you elaborate what you mean by HOPR works only with hex. Because I believe it would be pretty difficult to generate purely hex mesh in gmsh to begin with.I am going through the documentation available for HOPR and it says it is designed to generate unstructured 3D meshes, if I am not wrong. In my understanding, that indicates it should be able to handle tet elements as it would be difficult to have purely hex meshes for all sorts of geometries (especially those with highly curved edges). Is that a misunderstanding?



JBR

[Eifoehn4]

This was a bit misleading.

Sure you can convert Tet-meshes with HOPR. However, currently FLEXI is not able to use them.

Note that:

The spelling unstructured meshes has nothing to do with the supported element types. Moreover, Gmsh is able to generate pure unstructured Hex-meshes.

Regards

[Eifoehn4]

Two Remarks:

• Your "parameterfile.ini" seems to be corrupted. Do not use any nasty text editor!
• You are missing two normalvectors for the periodic BCs

Use this one:

!=============================================================================== !
! OUTPUT
!=============================================================================== !
projectname        = LID_DRIVEN_CAVITY
Debugvisu          = T
outputFormat       = 0
DebugVisuLevel     = 1
checkElemJacobians = T
!=============================================================================== !
! MESH
!=============================================================================== !
Mode               = 5
Ngeo               = 1
nZones             = 1
filename           = lid_driven_cavity_2.msh
!=============================================================================== !
! CURVED
!=============================================================================== !
useCurveds         = F
!=============================================================================== !
! BOUNDARY CONDITIONS
!=============================================================================== !
BoundaryName       = periodic_back
BoundaryType       = (/1,0,0,1/)
BoundaryName       = periodic_front
BoundaryType       = (/1,0,0,-1/)
BoundaryName       = walls
BoundaryType       = (/4,1,0,0/)
BoundaryName       = lid
BoundaryType       = (/4,1,0,0/)
vv                 = (/0,,0,,1./)
vv                 = (/0,,0,,-1./)

Regards

[jnanabrao]

Thanks a ton! That parameter.ini file worked. I haven't yet understood the boundary conditions in HOPR as I have only begun the documentation file. I was misled by the error message in to looking at the mesh file.



I have generated a pure hex mesh in gmsh as I do intend to run the mesh in flexi. However, I am not yet well versed with gmsh's refinement techniques to obtain the necessary quality, which I am working on right now. I have recently switched from openFOAM where I was exclusively using blockMesh and snappyHexMesh. In that block structured environment refining a mesh is quite different to what it is in gmsh. Thanks again for all your help.


P.S. I checked your Youtube channel out and it was quite impressive.

[jnanabrao]

Your "parameterfile.ini" seems to be corrupted. Do not use any nasty text editor!

I'm using gedit but I think it is creating problems. Could you suggest how I write the parameter files?

[Eifoehn4]

I generally use VI/VIM.

[jnanabrao]

Hi Eifoehn4,
Sorry to bother you again. But while your parameter file worked, I haven't had any success with writing my own. So I created a uniform hex mesh in gmesh and tried to read that in with a different set of BC names and definitions. But I don't understand the readin procedure used by HOPR. In cases 2 and 3 after reading in all the boundaries, the program says a side with undefined boundary condition is detected but doesn't specify which side or where it is. In case 4 it ignores two boundaries and then says later that they were not defined. The error file it asks me to check is written but is always blank. What am I missing?


PS: The meshes and boundary conditions in the cases with suffix 2,3,4 are practically the same.

[jnanabrao]

Forgot to click upload.

[jnanabrao]

The mesh files are wrong here. Please ignore the last two posts.