[heavy_user]

Hi Folks,

is it possible to create a mesh like a spiderweb (see picture attached) ?

best regards

heavy

[sega]

With blockMesh you can only create an O-Grid for the discretization of a circle.

If you have some other netgenerator you can try to mix tet- with hex-elements.

[heavy_user]

Originally Posted by sega

With blockMesh you can only create an O-Grid for the discretization of a circle.

If you have some other netgenerator you can try to mix tet- with hex-elements.

Hi sega,

thx for your reply, i will try to do so...

regards

[linnemann]

Hi

Using salome. 3 mins.

[heavy_user]

Nice!

But with blockmesh it takes a "little" more time, i guess...

[Poltak Stanggang]

Hi,

see attach the grid (3-D) for ducted. You can used very good blockMesh in OF to generate Mesh. First I used FORTRAN to organize blockMeshDict and transfer to OF. Try to learn how to organize blockMeshDict for OF.

[dratkins]

Originally Posted by heavy_user

But with blockmesh it takes a "little" more time, i guess...

Don't guess.

If you'd start reading manuals instead of complaining that they are not useful, you could do it in 3 minutes with blockmesh. Also in Salome you can't create a mesh in 3 minutes unless you read about it or you play around for a loooong time before you begin real work.

By the way, I don't know why you are so heavy. I recommend that you try a diet.

Robert

[linnemann]

I agree with you dratkins, but the amount of work you need to put into using/learning either Blockmesh or Salome would be somewhat the same. When you get to very complex geometries it is not a feasible approach to use blockmesh. At least not in the industry, maybe for scientific work blockmesh will suffice in a day to day basis.

[linnemann]

Or even better if you want pure hexes

[heavy_user]

HI All,

thx for your nice answers and images!!

but i am wondering....

Since i have no manual (which is great, by the way!) i am not shure what to do..

lets take the posted mesh of linnmann.
If you create the mesh with blockmesh you can chose the meshpoints on the x- y- and z-axis of each hex block.
Like this: hex (vertices) (100 100 50) simpleGrading (1 1 1)

But in the case the lines of the geometry are not parallel to the axes of the coordinate-system the lines of the geometry can be described by a combination of x and y (or x and z, or y and z; depending of how you have your coordinate system).
so i end up, not being able to mesh the radial direction independently.


How can i independently determin the radial mesh-points?
Can i do it in polarcoordinates or what is the trick?
(and if, how can i do so)

Or can I mesh each edge individualy (like in gambit, sorry for comparing) ?

I could manage to alligne all the adges of the volumes surrounding the inner tube (the geometry is a cylinder in a hex box) to the coordinate-axes by creating a Volume with 10 vertices.
If I look at UG page 127 this seems not possible.
Is there a way to do so with blockmesh?

(@dratkins: thats why i spend so much time looking for manuals and documentation...my small experience tells me that you might need them one day)

thx in advance!

bg

[jploz]

@heavy_user:
Have a look at section "5.3 Mesh generation with the blockMesh utility" of the OpenFOAM User's guide (v1.5). Features and usage of the blockMesh tool are described there (like in a manual).

HTH,
Jean-Peer.

[heavy_user]

Originally Posted by jploz

@heavy_user:
Have a look at section "5.3 Mesh generation with the blockMesh utility" of the OpenFOAM User's guide (v1.5). Features and usage of the blockMesh tool are described there (like in a manual).

HTH,
Jean-Peer.

Hi Jean-Peer,

thx for your answer!

Section 5.3 describes the use of blockMesh (partly).

In 5.3.3 it is described how to create blocks with LESS than 8 vertices, so I asume doing different is not possible (which is not explicitly stated there)...

In 5.3.1.3 in "nuber of cells" it sais: "...number of cells in each direction of the x_1 x_2 and x_3 directions.." which does not help (see description of problem above...) since i have edges in the x-y-direction that can not be graded independently (???) from the edges parallel to the x- or y- direction.

I did some research and found a paper "rotatingGGI" which shows an entry of an dynamic mesh dict:

coordinateSystem
{
type cylindrical;
origin ();
axis ();
direction ();
}
....
(but that is pretty much the only informatino I could gather regarding the use of cylindrcal coordinates)

But I dont know how to use cylindrical coordinates in the blockmeshdict, yet... I will try to find out (and post if i succeed) but thats the thing i have been complaining about...the UG and the PG seem not to describe the prog. completly...so i will go on guessing..

best regards!

[sega]

Originally Posted by heavy_user

But I dont know how to use cylindrical coordinates in the blockmeshdict, yet... I will try to find out (and post if i succeed) but thats the thing i have been complaining about...the UG and the PG seem not to describe the prog. completly...so i will go on guessing..

This is because OpenFOAM does not work with cylindrical coordinates.
OpenFOAM is restricted to cartesian!

Citation of the UG (you didn't find usefull):

2.1.1.1 Mesh generation

OpenFOAM always operates in a 3 dimensional Cartesian coordinate system and all
geometries are generated in 3 dimensions.

[heavy_user]

Originally Posted by sega

Citation of the UG (you didn't find usefull):

I read that also, but I was confused by the thing in the dynamicmeshdict, which obviously uses something else .
So how can OF not be able to do it, when somebody uses it in the dynamicmeshdict...

And I figuered there might be a trick to it, something like define a local-system for the whole geoemtry... but I thought I ask before shooting...

[sega]

Originally Posted by heavy_user

I read that also, but I was confused by the thing in the dynamicmeshdict, which obviously uses something else .
So how can OF not be able to do it, when somebody uses it in the dynamicmeshdict...

And I figuered there might be a trick to it, something like define a local-system for the whole geoemtry... but I thought I ask before shooting...

I'm not familiar with gliding interfaces, so I can't tell you anything about it.
Lets get back to your original problem: You want to mesh a cylinder!

Why don't you use the meshes & o-grids suggested by us?
They are popular and widely used.

[heavy_user]

Originally Posted by sega

I'm not familiar with gliding interfaces, so I can't tell you anything about it.
Lets get back to your original problem: You want to mesh a cylinder!

Why don't you use the meshes & o-grids suggested by us?
They are popular and widely used.

I meshed the cylinder with the o-grid.
Then I ran in the problems with not being able to mesh the edges independently since i have edges which are obviously seen as a combination of x and y... so i came back to findig a way of making a cylindrical mesh...and then the question of creating volumes with 10 vertices...

I created a 2D geometry and I am trying to get the case running right now.

The only remaing question is the indepentent-meshing problem i described before...
Did I make the problem clear or should i attach a draft??

regards

[sega]

Originally Posted by heavy_user

The only remaing question is the indepentent-meshing problem i described before...
Did I make the problem clear or should i attach a draft??

I didn't understand the problem. Maybe a draft would be useful.

[heavy_user]

Originally Posted by sega

I didn't understand the problem. Maybe a draft would be useful.

HI sega,

here the draft (sorry for the poor draft, but I hope it fits the cause...)

Lets say we name it from left to right, the first draft is A, the second B and the third C.

All drafts are to show a cylinder in a box.
The Z-direction is out of the screen and the axes poking your nose...

For creating the volumes around the cylinder there are 3 possebilities.

A: the lines between the cylinder and the box are not parallel to one of the axes and can therefore not be graded individual.

B: you could grade the edges individual, since they are parallel to the axes, but this kind of volume needs 10 vertices. (but how to determine the mesh of the circle ??)

C: (i dont have photoshop at work, so sorry for cheating :-P, i should have turned the box, not the axes... ) You can grade the edges connecting the box and the cylinder with the "hex....(Nx Ny Nz) " but how do you determine the mesh of the circle and the edge of the box? They are not parallel to any of the axes..(and i know that a circle can not be parallel to a line, but you know what i am trying to say...)

Perhaps it is really easy to solve and I am stuck somewhere...but i cant figure right now...

thx for your efforts!

regards

[sega]

This makes it more clearly to me, now.
But now you want to mesh the outside of the cylinder, not the inside?!

Originally Posted by heavy_user

A: the lines between the cylinder and the box are not parallel to one of the axes and can therefore not be graded individual.

As far as I know the edges of the hex don't have to be parallel to any axis.
You can simply set the appropriate grading factor to the corresponding direction.
Remember: each hex block (which forms a topological square) has its own coordinate system when dealing with blockmesh!
And this is the reason the grading should work in this case.
How the local coordinate system is determined depends on how you choose your vertices, but this can be re-read in the UG!

[Wolle]

Maybe this helps?

http://www.opencfd.co.uk/openfoam/do...#x6-380002.2.1