[Techies]

Hi guys,
I'm working with icem. I have a trouble that is when i specify spacing lower (5e-7 for y+<1), the quality elements become so bad. Actually, I try to modify tolerance and use project that I have found in the forum to bsplines but all don't work.
So anyone ever meet that can help me to solve this problem ? Thanks

[alainislas]

Could you post some images of your blocking (including the region of bad elements) ?

Some hints:

1) Double check that all edges, faces, vertices are properly associated to the corresponding geometry entities

2) You can go to Blocking tab -> Pre-Mesh Quality Histograms. And display the metric you are interested at. Then click on the histogram bars that are giving bad qualities, right click and create a subset. You can display again the edges and the subset only. This will give you impression of what's causing the trouble.

2) Probably there are some internal edges that are not properly splitted, or even misaligned. So you'll need to manually move them and reset the split.

[Puschkin]

I'm also facing the same problem while meshing a CD-nozzle with only hexa mesh. As soon as I set the spacing for the first layer thickness of my inflation layer lower than 1e-7 [m] for example 1e-8 [m], the first elements near the wall become very distorted (see attached pictures).


index.jpg


index 2.jpg

In my case I import the geometry data points into ICEM and generate a spline and lines which form my geometry. Then I do the blocking part as explained in the tutorials and define size of the pre-mesh. After smoothing is complete and health check of the mesh has passed all of my requirements, I extrude (rotational) the mesh one element deep (<5°) to perform a 2.5D simulation in CFX later. For smooth walls, the first layer thickness of 1e-7 [m] was totally enough for y+ values < 1 and my sim. converged easily. But as soon as I add roughness I have to refine the mesh even more since y+ values went up to 30-35. But I'm limited in sizing since the distortion of the prism elements starts.



Already tried to activate "Project to bslines",

changed topo tolerance and triangulation tolerance but since I dont have any surfaces it doesnt help.



(I also opened a thread earlier this day...but its not released yet in the forum.)

Hope to resolve this problem soon. Thanks for all the help so far!

[alainislas]

@Puschkin

From the image it looks like your near wall elements become a bit skewed. Does this happen only at this specific zone? Or all over the length of the wall?

I dont see your blocking, but my feeling is that these elements are skewed because there is a misaligned vertex, what causes these cells to stretch. Could you reference your tutorial? Also, some pictures containing your blocking would be helpful

[Puschkin]

Hello alainislas,


thanks for your reply! I updated my last post and added some information.
My blocking strategy is shown in the following picture:


index3.png


The vertices are associated with the points which are a part of the spline/geometry. So theoretically the associated vertex sits precisely on the spline. The edges are also associated with the curves. To mesh the boundary layer I splited the blocks "horizontally" with a defined ratio to the radius of the geometry to resolve the boundary layer as good as possible. The vertices which are seen "in the fluid" or the stream are defined via CAD so the block-edges are orthogonally aligned to the wall (as seen in the picture/converging part of the nozzle). After the pre-mesh is done I release the association of these inner points to allow more degrees of freedom for the smoother.


Regarding your question:


I just checked the constant part of the combustion chamber and the divergent part of the nozzle. The distorion of the elements is seen all along the wall, but much less in the constant part of the comb. chamber.

Especially in the divergent part of the nozzle (at this moment I dont know if this is a wrong portrayal since I have to zoom alot or just the fact that the steep curve is approximated with a fixed, relatively large element lenght) it seems that the mesh (even if the edge is associated with the curve correctly) is not alligned properly to the wall/curve.


Index 4.jpg


Could this be the origin of the distortion? Since the meshing algorithm is not able to capture/approximate the curve due the large element lenght in the axial direction compared to the radial one (radial growth)?

[alainislas]

@Puschkin

Could you try Blocking tab -> Edit Edge -> Split Edge -> Automatic Linear on your boundary (green) edges? See if that helps, your blocking looks good

[Puschkin]

Originally Posted by alainislas

@Puschkin

Could you try Blocking tab -> Edit Edge -> Split Edge -> Automatic Linear on your boundary (green) edges? See if that helps, your blocking looks good

@alainislas,


I tried your recommended fix/workaround but unfortunately the mesh is still distorted as it was before.



Kindest regards,
Puschkin

[alainislas]

Originally Posted by Puschkin

@alainislas,


I tried your recommended fix/workaround but unfortunately the mesh is still distorted as it was before.



Kindest regards,
Puschkin

Do you have all vertices properly snapped before trying to split the edges? If not, see if that helps

[mluckyw]

Originally Posted by Puschkin

Attachment 85993

From the association of points for the o-grid layer (I assume you do the o-grid for boundary layer) the vertex of which are essentially the start of the boundary layer are associated to a point since it says P in the blue vertex. have you tried to disassociate these vertex? I usually don't do association for the boundary layer vertices.

[Puschkin]

Sorry for my late reply guys. I'm very thankful for your help so far. I was busy with other work over the past few months but now I'm back and I'm still having trouble to generate my prism layer.

Originally Posted by alainislas

Do you have all vertices properly snapped before trying to split the edges? If not, see if that helps

The vertices are associated with the corresponding points and the edges are associated with the correspondig wall (if I remove the vertice-to-point association, I can move the vertice along the corresponding curve/wall). I did the edge splitting of the wall (green line) as you discribed.



1.PNG


The situation can be seen in the picture. The Edge (green line) is now following the geometry-spline. But the mesh is still distorted as I zoom in. Its not only a visual problem since the mesh-quality/determinant is showing me a bunch of bad elements.

Have you ever experienced such problems with a pre-mesh on a curvature while using a spacing of the first node of 1e-007 or 1e-008?

I did a bunch of simulation runs with diffierent first node spacings and it clearly shows me that I have to go that small for the first layer to reach my desired value of y+<1.

*I also tried to de-associate all the vertices from the points of the o-grid. Unfortunately it didnt help.