Hi,

I am using CFX-10 for airfoil computations. Because I want to resolve the boundary layer at the airfoil in detail, I use ICEM CFD for the generation of a hexa mesh. I am starting with 2D computations, but will later go on to 3D.

If I want to resolve the boundary layer, I need very thin elements near the surface (in the order of 1E-5m). When I create such a mesh, I often get problems with mesh quality on the elements near the airfoil surface. I have problems with inverted elements (negative determinants) at the airfoil, meanwhile the element quality is good in the remainder of the mesh.

I have tried several solution strategies for resolving the problem:

1. To run the check/fix command, and inverting the inverted blocks. I have had some success with that, but most of the time it does not really help me.

2. I have done pre-mesh hexa smoothing, but I am not sure which of the smoothing types would be most efficient (quality, orthogonality or multiblock smoothing?). There is a lot of possibilities for different smoothing algorithms and so on. So far no one of the smoothing approaches I have used have been very successfull. Often I have to apply the same smoothing procedure many times to get rid of the inverted elements near the airfoil surface. But this often result in a mesh that is smoothed too much in the remainder of the mesh, and often it is not possible at all to remove all of the bad elements.

3. I have done a little smothing under the "edit mesh" capabilities. But I am not sure if this is better than the pre-mesh smoothing.

Is there any other approach beside smoothing techniques to remove the bad elements near the airfoil surface?

This might seem a long story from me, but I think the topic is quite complex, but very necessesary in order to get a good quality mesh. So I would like to hear any others experience with the topic.

Best regards CFDworker

You probably have an issue where the geometry is not resolved enough for the detail in the fine mesh spacing. Go to the Model/Global Mesh Parameters and look for the Triangle Tolerance entry. Reduce this by 2 or 3 orders of magnitude, save the geometry file, and read it back in - then see if the mesh behaves better. (This value controls the resolution of the geometry when reading it in - a smaller value makes the geometry closer to the true geometry.)

Hi Myron, Thank you for the response. I have tried your suggestion today, and it might help a little, but I am not totally convinced yet. The problem is a little strange. To me it appears like the mesh has excellent quality with respect to determinants before the smoothing (no values below 0.6). After smoothing, the quality is often worsened, often with negative determinants at a few cells close to the surface of the airfoil. So, in my opinion the smoother works opposite as intended in this case...

Since the quality of the mesh is good before smoothing, you might ask why I want to do smoothing... The reason is that I think the unsmoothed mesh follows the edges of the blocks too much, making the cells on the block interfaces change "direction" in a too abrupt manner, ie. the structure of the blocking dominates the mesh too much. I don't know if that defines my problem to you more precisely?

The pre-mesh smoothing definitely has limited applicability. It would be easier to adjust node spacing/distributions. You could also try edge splits to manually shape the edges a bit more to your liking.