[Gmsh] Triangles when using 'Transfinite Line' and Blossom algorithm

[Eiser, Sebastian]

Dear all,

I'm currently testing the Blossom algorithm. I think it does a very good job and is much faster than other quad meshers I know. I'm generally very satisfied with the resulting unstructured mesh. But I have some rectangular surfaces which have a very high width/height ratio (e.g. a very thin structure). I found it impossible to set the characteristic lengths such that I get exactly two quadrilateral elements in the "thin" direction. For these surfaces, it is very handy to use transfinite mesh. I also use 'Progression 0.99' to achieve some sort of scale transition in the extended direction.

Whenever a surface meshed by the Blossom algorithm intersects with a transfinite surface, the Blossom algorithm would then not recombine all triangles to quads. It works only for heavily tuned 'characteristic length' values, as can bee seen in the attached uber-simplified example. In the example, I'm coupling a square with transfinite mesh to a square which is meshed with the Blossom algorithm. Slightly modify the 'cltest' parameter, and there will be triangles in the second square.

What can I do to either (A) use Blossom to mesh thin structures or (B) be able to successfully mix structured and unstructured quadrilateral meshes? 

thanks a lot, 
-Sebastian

PS: I made a very humble donation to gmsh earlier this week and found it impossible to choose EURO as currency. I'm not sure how much of my donation arrives if it is firstly converted to USD and then back to EURO. Surely paypal cashes for every conversion...
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ref2d-bloss-transf.geo
Type: application/vnd.dynageo
Size: 1421 bytes
Desc: ref2d-bloss-transf.geo
URL: <http://www.geuz.org/pipermail/gmsh/attachments/20120726/c3457104/attachment.geo>