[sukye]

I modified cavity sample a little by narrowing its moving lid.

The mesh was created successfully. But blockMesh exited with floating point exception for a few runs.

the Courant Number has increased far > 1, and the U field has become saturated in the box.

Any suggestion?

I also changed the dimension to 3D by moving the faces in frontAndBack to fixedWalls. It seemed to me nothing evolved after the initial set-up.

[colinB]

Hey there,

the information you provide is very little. Therefore it would help if you could post at least:

- blockMeshDict
- log of the creation of blockMesh / error message (?)
- results of checkMesh

and since you mentioned your courant number explodes:

- log of the calculation process (error message)
- controlDict

Assuming that all solver/BC settings are set right this should bring us to
the solution.

Few comments on what you've got so far:

But blockMesh exited with floating point exception for a few runs.

what do you mean by that? blockMesh did not exit properly, but with a FPE
but you used it and then the calculation crashed?

I also changed the dimension to 3D by moving the faces in frontAndBack to
fixedWalls.

I'm not sure if changing the BC from empty to wall is sufficient to get a 3D
mesh, it might be that you also have to add some more cells in
the direction where you've got so far just 1 cell layer.

Finally you mentioned that Co and U explode:
since deltT (if you use the standart settings of the controlDict)
and deltX are fixed and U explodes this might be more an BC issue than a
meshing issue.

I hope I could contribute

regards

[sukye]

Thank you Colin for your reply.

I think theCourant number overflow problem is related to mesh size. The mesh at one corner of triangle is very small generated by blockMesh with default settings. How can I create triangle mesh? I searched in the forum to find someone mentioned it without solution.

I will add more cells in the box to try to work 3D. But it is unclear to me why one box is sufficient for 2D while not for 3D.

[colinB]

Hi again

I think theCourant number overflow problem is related to mesh size

I guess you mean cell size (shape?)

The mesh at one corner of triangle is very small generated by blockMesh with
default settings.

Actually there are no triangulars in the lid driven cavity so I don't get this
point.

How can I create triangle mesh?

If you mean wedge shaped blocks for the domain use the keyword
wedge in stead of hex in blockMesh

If you mean tetrahedra cells, there are no native OF meshers which could
do that, but Salome and enGrid can create them.

Again: a little bit more detailed description of your problem could help us
to help you more easy.
Including some pictures/sketches could clarify a matter pretty fast
as well as some output form some log files.
For more questions on that see the sticky post in the main OF Forum!

regards
Colin

[sukye]

Thank you Colin,

It turns out I use gmsh to generate mesh.

From your post, I will guess that we need wedge shaped blocks to make triangle mesh possible. Right?

I also mention in another post
http://www.cfd-online.com/Forums/ope...ientation.html
that my mesh by gmsh cant pass checkMesh.