[and others..] How would you mesh this geometry?
 

Hi everybody!
I'm helping a member of this forum in meshing a geometry; it is a small cilinder with an impeller.
We would like to have a structured hexa mesh, and we would like to know how you would mesh it and or split the geometry.
As you can see from the attached pictures, high skewness cells are generated near the semicircle.
Inner cylinder (which has small semicircles) is the rotor (walls), outer cylinder is rotating fluid.
What do you think?

Thank you,
Daniele

4 blocks approach?

Yes. It is perfect idea.

I am also working on the similar case. Attaching blocking https://dl.dropboxusercontent.com/u/...tank_Jamal.zip

http://imageshack.us/a/img541/183/stridtank2008.jpg
http://imageshack.us/a/img687/9444/stridtank2001.jpg
http://imageshack.us/a/img33/6047/striiringtank001.jpg
http://imageshack.us/a/img541/6766/stridtank2002.jpg
http://imageshack.us/a/img843/3479/stridtank2004.jpg
http://imageshack.us/a/img444/1461/stridtank2007.jpg
http://imageshack.us/a/img201/6255/stirrtank2002.jpg
http://imageshack.us/a/img5/6041/stridtank2003.jpg
http://imageshack.us/a/img4/7914/stridtank2006.jpg

Thank you Far for your answer, basically I was using the same approach as yours, but in gambit :)
Looking at your picture, what do you think, in your experience, of sudden change in size of adjacent cells?Near the rotor walls I can see a change in size near 10 times: can this affect convergence?

Daniele

Yes you are correct, spacing jump is large and needs to be corrected. In fact this blocking is not final and still few operation are needed. Like making ogrid in remaining two drilled holes or blades. O-grid is also needed outside these drilled holes to make the mesh of high quality.

My intention to upload blocking at this stage is to invite forum fellows to try further and improve this version. Meanwhile as I get time I will improve it but it may take two to three days. At the moment mim quality is 0.2 in the holes where o-grid is not implemented otherwise mesh is of higher quality.

The problem with bottom - up approach is considerable effort in making vertices, edges, faces and volumes which is avoided in ICEM CFD. Moreover ICEM is flexible in sense that we can change location of any vertex, edge and face at any time to make the quality better without deleting everything and starting over.

Edit: O-grid at bottom is implemented and can be seen in attached pics.

http://imageshack.us/a/img442/2223/stridtank2009.jpg
http://imageshack.us/a/img441/7810/stridtank2010.jpg

Daniele: I would suggest that put a boundary layer in drilled holes and mesh with tetra to control the quality. Other portions (except drilled holes) may be meshed with full hexa.

Yes, of course!
Thanks

Or you may join us :D The ICEMers ;)

I would really enjoy the "icem world"; I have to install it and try; I remember when I opened the first time gambit and all was new to me; but after some time I was able to understand the most of the functions.
I need only time to look at icem and move on..gambit is good in my opinion, but it is a bit old and new software is coming on with new features.

Daniele

Yes. Gambit is good software. I have used it for seven years. It has very good features.

Regarding the sudden change of size, I would try to attach a size function (type meshed) with fine entities as source. Gambit should smooth the mesh in this area.
Far's option is also a good alternative.

Hi dear Daniele
a question: How do you create this volume? (The volume that has been mesh in right hand picture)

Best Regards

Hi,
in gambit, under edges, project the rotor curved lines on the external face (minimum distance), with the split face option enabled.
Then create all the remaining faces to create the volume.
If you want to follow the "4 blocks approach", then split the edge of the upper face and create a small square, create the edges and then split the face with edges.
Change if necessary vertex types (under face mesh in gambit) to have the faces mappable and mesh them with hex->map.
A mappable face has 4 'end vertex'.
Mesh one of the "vertical" edges of the volume and mesh it with the cooper scheme.

Daniele