CFD Online Discussion Forums

CFD Online Discussion Forums (
-   ANSYS Meshing & Geometry (
-   -   [ANSYS Meshing] Increasing number of blocks for parallel computing purposes (

Ivan August 3, 2012 18:27

Increasing number of blocks for parallel computing purposes
Dear friends,
I'm having some troubles in ICEM-CFD with the blocking and I hope you can help me. I'm running a simple geometry of a jet spreading into a coflow (I have the pipe into the domain). Up to now, I have divided the geometry in a small number of blocks. Since, however, i need to start a LES on combustion on that geometry, I need to increase the number of blocks as much as I can in order to split them between a large number of processors and run the case in parallel.

I have started from the geometry I have used up to know (where I have associations between edges and curves - no associations type point or surface) and tried to use the simple split along the axis direction. Nevertheless, after I do this, the grid cells are modified in an improper way, as in the blocks near the border of the block I have splitted, instead of going straight along the axis direction, they contract (approaching each other) in one block and enlarge in the next, like a "sousage".

I have reset all the parameters along the edges equal to each other, but I still have the same problem. I would like to know:

1) If there is an easy way to automatically split the block in sub-blocks without modifying any parameter (I do this just for parallel computation purposes, I do not need to use the new blocks into the geometry);

2) If not, which is a way to do this? Do I have to split each surface of the domain, create many other curves and associate each new edge with these curves?

Thank you!

P.s. I can send images representing the mesh if needed.

Ivan August 6, 2012 10:12

Solution of the posted problem
I have solved the problem! I'll write down the solution for who may have this issue in future. Practically I have a circular nozzle into the domain, made by two concentric circles in order to have the lips of the pipe into the domain as well. I used two O-grids, one for each of those (that is, for the internal diameter and for the external diameter). At the outlet, this O-grid is not associated with any circular curve, so it is a square (One border of the same block is at the nozzle exit, and one at the outlet). Now, if the block between the cells near the pipe wall and the outlet is enough long, there is no problem. When, however, I split this block in many blocks (for parallel computation purposes) what happens is that I have a circular block really close to a squared one, since by defeault the O-grid automatically generated by the split are not associated with any curve.

What happens now, is that the two edges of the first cell closer to the pipe wall connect both to the same edge of the next (squared) O-grid, that is to say, a hexa cell degenerate into a tetra cell. This of course affects significantly the solution.

To solve the problem I have created a couple of concentric cirles for each new block-border created and associated them with the respective O-grid by hand (painful, since the blocks are many, but at least the problem is solved!).


All times are GMT -4. The time now is 02:55.