[usv001]

Dear Santos,

Thanks for the page on blockMesh. You have introduced me to the world of m4 programming. ;-)

I have a question for you: what is the algorithm used by blockMesh? I am trying to mesh a varying cross-section nozzle and the current code does not perform well when I use a spline along the length.

I found out that it is long standing problem in blockMesh. So, I want to try to correct it but I am lost at the code found in 'blockCreate.C'. It uses weights, importance factors, correction vectors, etc. I do not quite understand it. Do you know of any source (website/paper) that describes the principles behind blockMesh?

Once again, many thanks!

[wyldckat]

@usv001 - Quick answer: I'm not aware of any document that explains how do the internal systems work in blockMesh. But the code is open source and it should be just a matter of inspecting the source code.

If you feel that this is really a bug, then open a bug report on the bug tracker, but please provide more details for the problem itself. If you offer to help with fixing this issue, Henry might be able to provide you with some description of the current limitation that algorithm has got... at least beyond the ones that are already described on the bug tracker. You can find several if you Google:

site:openfoam.org/mantisbt blockMesh

But AFAIK and remember from a previous analysis I did on it: http://www.openfoam.org/mantisbt/view.php?id=1396 - the algorithm is pretty straight forward: after the points along an edge are calculated, using the standard position along a straight line, they are then transformed with a single transformation if applicable. One limitation with this is that it doesn't take into account multiple space transformations.

[usv001]

Dear Santos,

Once again, thanks for the reply. Here is what I have done up to now:

I tried to mesh a nozzle of varying cross-section. The current blockMesh algorithm does not scale/propagate the arcs properly along the length of the nozzle. You can see the pictures attached (mesh_before_correction_1 & 2). It is the same bug that was raised earlier.

I tried using the patch provided by Henry (Thanks to Henry!). It seems much better now (mesh_after_correction_1 & 2). But I want to make sure that it really does what it looks like it's doing - that is, it really scales the arc properly.

Earlier I have been trying to mesh just a quarter of the nozzle and using mirrorMesh to obtain the full nozzle. The fix by Henry does not seem to work for that case.

- - -

I have gone through the code in blockCreate.C. I understand that the process starts with identifying the a point on each of the 12 edges which corresponds to the (i,j,k) of the internal point we would like to find. But after that, I am not sure what exactly happens with importance factors, correction vectors, etc.

Thanks for all the help!

[wyldckat]

Quick answers:

Originally Posted by usv001

I tried using the patch provided by Henry (Thanks to Henry!).

You might want to double-check who posted what in the "Issue History" section of the bug report: http://www.openfoam.org/mantisbt/vie...d=1396#history

Originally Posted by usv001

I have gone through the code in blockCreate.C. I understand that the process starts with identifying the a point on each of the 12 edges which corresponds to the (i,j,k) of the internal point we would like to find. But after that, I am not sure what exactly happens with importance factors, correction vectors, etc.

Back then when I was trying to diagnose what was going on, I had to have the simplest geometry possible, so that I could manually calculate the desired mesh points and then compare the manual calculations with the ones made by the method "Foam::block::createPoints()". And in order to see the values calculated by that method, I used Info calls to output the values. For example:

Info << "k:" << k << " -> corz4: " << corz4 << endl;

Beyond this, as discussed in the bug report, it's possibly easier to use dynamic meshing, instead of relying on blockMesh. The other possibility is to do the calculations for all points with an outside application and then use an utility to replace the existing points that are created by blockMesh, with the ones created by the other application.
Such an example already exists in OpenFOAM... let me see if I can find the thread... ah the utility is datToFoam and the thread is this one: http://www.cfd-online.com/Forums/ope...rmat-help.html

[kalle]

This might be an alternative:
http://www.cfd-online.com/Forums/ope...blockmesh.html

Or look at extBlockMesh as discussed in the bug report. extBlockMesh may not handle grading, though.