code_saturne User's Forum

Hello all,

I am just looking for some advice on the best way to go about meshing the geometry I have attached - a central nacelle (hollow), a disk representing turbine blades and a long rectangular channel.

I have read through the mesh generation advice on this site but am wondering whether anyone has experience with similar geometry. My issue is with trying to use a fully structured hex mesh only. More specificially how to transfer from an oGrid on my disk (and possibly the nacelle too) to a bricked out channel.

What I have tried so far:

1.Boxing in the turbine and nacelle (which are hex meshed themselves), using unstructured tet in that region with quad faces so that I can simple slot this into the channel.

2. Extending a cylindrical oGrid throug the length of the channel to allow hex mesh everywhere.

3. Using block decomposition to cut up the domain up into 6 sides composite parts so that the whole thing kind of looks like a 3d British flag!

Anyone have any ideas/advice on this? I am going to be using similar geometries for quite a while so I want to try and construct the mose robust set up I can now to save time later!

thanks in advance,

Susan Tully

Hi

The problem is not so hard to make structured although it depends on the meshing software. ICEM CFD could deal with this quite quickly. What mesher are you using?

The stages I would go for would be:

1 - O-mesh along length of channel fitting around disk radius.

2 - Cuts at front and back of disk and at back of nacelle. (Removing the block containing the disk).

3 - O-mesh along length of channel inside the first o-mesh.

4 - Cut in second o-mesh just upstream of nacelle nose. In the block containing the nose use an o-mesh as you would about a half sphere.

5 - Cut in second o-mesh at back of nacelle. (Removing block containing the nacelle).

6 - O-mesh in third o-mesh spanning from the back of nacelle to the outlet(x minimum).

You could also cut the outer channel (before the first o-mesh) to fit a smaller rectangle about the disk which would give you a nicer mesh at the channel corners.

Hope that was some help - it is probably easier to draw than explain but if you need any more ideas feel free to ask.

James

Hi James, thanks for your reply. What you are describing is exactly what I tried with option 2 in my post though I didnt go into much detail when I wrote it (I tried with the additional square and without too ). My issue with this method is that it seems to be very inefficient regarding mesh size. I need a fine resolution over my disk region and extending this along the length of my entire domain means that I have a lot of wasted computational effort far up and down stream. This is a scaled down model, I will be eventually be working on a 30km long channel containing a ~30m diameter turbine so want to be as frugal as possible with mesh elemtents far upa nd down stream. Also, for me this leaves me with thin vertical elements at the left and right sides of my channel when flat elements would be better to capture the vertical shear profile of a tidal channel

I throught about boxing in the turbine and extending out the oGridded disk and nacelle to quads on the face of this box. But I am stil trying to get my head around doing this Also I havent tried yet with concentrating my radial mesh lines out to either side to reduce the tall thin elements I mentioned above which I am also going to try.

I am using Salome to build the mesh. I am fairly new to it but am liking it so far!

thanks again for your reply. Good to know someone else would take the approach I did

Hello,

I'll just add that all you options are valid: a hybrid mesh will require less time to build, but with a fully hexahedral mesh, you'll usually have as good or better results with slightly fewer cells. If you run many calculations with that mesh, the investment in meshing time might be worth it. There are cases where highly "warped" purely hexahedral meshes provide worse convergence than "good" hybrid meshes, so I would not recommend "hexahedral at all costs".

Note also the SALOME 6.6 and 7.2 (released last week) have a new "HEXABLOCK" module as a more specialized alternative to SMESH for generation of blocked hexahedral meshes. I have not tested it yet, and it might not be fully mature yet, but you are welcome to test it (i.e. we are interested in your feedback ).

Regards,

Yvan