[Shrap]

Hi,

I'm doing a CFD model of HAWT with a structured mesh. I have a 120° sector (periodic boundary condition) and a MRF domain close the blade.

My problem is that I can't make the uppper face of the blocks to follow the curved boundary. I've tried "Split Edge" with "Spline" option. Doing so changes the edge but the block stays the same.

Do you have any idea or suggestion to extend the blocks to the curved face of the domain? I have the same issue for the MRF domain.


blocks1.jpg

blocks2.jpg


Have a nice weekend!

[bluebase]

Hi Pascal,


in Associations tab, you can find a feature which allows you to map selected faces to selected surfaces.
With this feature, you should be able to get the desired mesh.


Best,
Sebastian

[Shrap]

Hi Sebastian,

Thank you for your answer!
I tried what you suggested without any success. The association worked but the block did not change geometry/appearance.

AssociateFaceToSurface.jpg

I tried a different way of assigning the vertices from the 3D bounding box. Instead of assigning the two lower vertices of the box to the point on the sharp edge, I assigned only one of them. Then I have one "extra vertex" that I could put on the arc. Ideally, I would like to have two points on the arc, maybe I can achieve that by splitting the blocks and moving vertices?

It looks like this now:
DifferentWayOfAssigningVertices.png

Now, I would like to create the block for the MRF domain. Do you have any suggestion on doing that? I've created a curve that outline the boundary. Could I use those curves maybe?

CreatingMRF.jpg

Have a nice weekend!

[shereez234]

can you post a picture of your pre-mesh and a scan plane viewing through

[Shrap]

Hi shereez234,

Thank you for your answer!

When checking the premesh it follows the curved part as it should. I thought that the block would fit perfectly after associating. I guess it's just a confusion from my part about the preview of the block I'll check the premesh next time.

Premesh.jpg

Premesh2.jpg

[bluebase]

Originally Posted by Shrap

I tried what you suggested without any success. The association worked but the block did not change geometry/appearance.

Attachment 74659

Hi Pascal,

you might have a look at the menu again, it seems you have used project to part. Two radio buttons down, there is also the feature to project to specific surfaces. In same edge cases this direct association is necessary when the automatic association misinterpretates locations of a face.

In ICEM, the shape of the block is always drawn minimalistically. If you want to understand the actual shape of the mesh, you'll need to use the computed premesh (and scan planes).
More often though, you'll just draw the edges and activate the feature to show the projected edges (in the context menu of edges).

Originally Posted by Shrap

I tried a different way of assigning the vertices from the 3D bounding box. Instead of assigning the two lower vertices of the box to the point on the sharp edge, I assigned only one of them. Then I have one "extra vertex" that I could put on the arc. Ideally, I would like to have two points on the arc, maybe I can achieve that by splitting the blocks and moving vertices?

You might need to check your solver whether it can deal with wedge shaped elements.
In case your solver is capable to use it, such a blocking structure might be simpler because you'll need fewer splits to resolve the domain of interest.

However, your second approach should work as well. You'll likely need to add some o-grid layers to deal with trapazoidal shape of elements at the "free" corner of the block.
I suggest, you go through the few tutorials for hex mesh generation. This will likely help you to understand what to do.

Best regards,
Sebastian

[Shrap]

Hi Sebastian,

Thank you for the tips!

- I'll tick the "surface" and not part next time for the association.
- The projected edge is a good feature indeed, thanks!

You might need to check your solver whether it can deal with wedge shaped elements.
In case your solver is capable to use it, such a blocking structure might be simpler because you'll need fewer splits to resolve the domain of interest.

Yes, on top of that the orthogonality of the elements was not good so I abandoned this idea and went for the splits.


I've done some ICEM tutorials beforehand, the closest tutorial of what I'm doing now being hex-mesh of a wing. Here's the result of this tutorial:
FinalMesh.png
FinalMesh2.jpg


I think I'm done with the trailing edge for the HAWT, I will do the leading edge now. I'm a bit unsure about how to do the far domain. I'll try this way: create a bigger block around the MRF domain then create a part named "Fluid_far" and select the block that I just created. Is that how you would do it?

Here's how it looks for now:
Ogrid.jpg

Is anything blatantly wrong? I should maybe link the edges on the trailing edge to the curved sector.

Have a nice day!

[bluebase]

Originally Posted by Shrap

Is anything blatantly wrong? I should maybe link the edges on the trailing edge to the curved sector.

You may have a look at the guide "Simons Tips and Tricks", on slide 99 and 100.
You'll find that guide in the sticky thread Before you start a thread, see if you can find answers here

The cross-section of your blocking along the wing should approximately look like that blocking by Simon. Though, it depends on the wing shape whether you need a c-grid downstream or wether there is a full o-grid around your wing.
Nevertheless, your blocking in front of the tip seems to give quite distorted elements.
Use Scanplanes and the quality reports to inspect the mesh. Note that by default the determinant report is showning only the range from 0 to 1. However, when there are very bad elements, they are likely below zero. So, start by drawing the range -1 to 1 in 2 columns and only draw the negative elements. That way you'll likely identify the regions which need adjustment.

The farfield is of secondary priority because it can easily accept the core (the blocking around the blade).
I'd suggest you to start playing with a 2D section of your wing until you get the blocking right, then it will be a straightforward step to create that blocking in 3d. For example you could later extrude the 2D blocking into 3D. Then you'd only have to create the block at the axis.



Another, comment, do not merge vertices unless you really need to. It seems you already have created a degenerated block (fewer than 8 corners) at the layer of the wing root.

[Shrap]

Thanks again for your answer!

I'm checking page 99 and 100 but I don't really see the problem with the mesh on the left below:

Comparison.jpg

To my untrained eye they look very similar. I guess the mesh on the right is better if you have to do heat transfer and you need the mesh inside. But for exterior aerodynamics, if you delete the mesh on the inside, both meshes look very similar? I'm missing something maybe?

On the picture of my last reply I had not associated vertices to points on the curve of the leading edge. After associating and moving the vertices to have orthogonality between curve and edge it looks like this now:

Blocking1.jpg

Ok, I didn't know if was a problem with blocks with less than 8 corners. I have blocks with 6 corners at the root and tip unfortunately, as you can see here:

Blocking2.png

Is that game-over? Will the blocks with 6 corners be impossible to mesh?

Interesting this 2D-blocking with extrusion strategy. The twist is fine if I do some splits and associate again, right?
I think I'll try that if I'm forced to redo everything

[Shrap]

As you predicted, I have negative determinant elements at the root:

NegativeDeterminant.jpg

I think I need to associate the edge at leading edge to the curved surface of the leading edge too. I only associated the edge to the curve but then the leading edge looks rather blunt as you can see above.

I'm trying to do the 2D extrude block method you mentioned but I'm running into some problems. When I choose "2D Planar" the block appears in the XY-plane and I can't find a way to move it to be at either the tip or root of the blade. Do you have any suggestion on positioning the 2D block?

[Shrap]

I managed to associate the leading edge It seems that I missed quite a few edge sizings at the root and tip. I suspect that's why I got so many bad elements.

If you want to take a look at the mesh, here's a link to download the workbench archive (2019 R3 version):
https://drive.google.com/file/d/1_Hz...ew?usp=sharing