Mesh problem - ElmerGrid vs gmsh

General discussion about Elmer
Post Reply
sebastien ROUQUETTE
Posts: 97
Joined: 05 Nov 2011, 16:50
Antispam: Yes
Location: Nimes, France

Mesh problem - ElmerGrid vs gmsh

Post by sebastien ROUQUETTE »

Good evening Dear Elmer fella,

I hope that someone can fix or explain me my problem.

Here it is: i tested the free surface case called "freesurf_sloshing_2d" in the elmer test database. A very interesting one!

I slightly modified it. I added the heat equation solver to study the effect of the surface tension coefficient on the fluid flow motion (i also applied a heat flux on the free surface).

It is working pretty well when i use the geometry '"square.grd" provided in the test folder. When i generate a similar geometry with gmsh (so a square of same size as the one defined in square.grd)... the simulation led to a different result: an awful computation of the free surface. The size of my square are identical for elmergrid and gmsh. When i apply "ElmerGrid 14 2 square.msh -autoclean", the generated mesh as 3600 elements as for square.grd and 3482 nodes instead 3742 for square.grd!

It seems that so informations are missing in gmsh file for a good computation of the free surface but i don't understand what? i also don't really understand the syntax used for the generation of the mesh with ElmerGrid and more exactly the section devoted to the boundary definition in square.grd (these values -1, -2, -3 and -4 for the column "out")...

Anyway, i attach the sif file, a udf file that i use for the heat flux definition, the square.grd and square.geo + square.msh and the free surface at 3 differents iterations (or step time): 100 iters, 200 iters and 300 iters....

Thank you for your reading and hope you will get me the answer!
Sebastien
Attachments
freesurf_surfacetension_2d.tar.gz
(275.45 KiB) Downloaded 338 times
square.grd
(636 Bytes) Downloaded 313 times
caseNSFSHT.sif
(6.63 KiB) Downloaded 342 times
Rouquette Sébastien
University of Montpellier / Mechanical and Civil Engineering Lab. / Welding Group
sebastien ROUQUETTE
Posts: 97
Joined: 05 Nov 2011, 16:50
Antispam: Yes
Location: Nimes, France

Re: Mesh problem - ElmerGrid vs gmsh

Post by sebastien ROUQUETTE »

Well Well, that is obviously extremely weird!!!!

i understood the difference of nodes between the 2 meshes (elmergrid / gmsh).
When we specify "element division 1 = 60" in square.grd (elmergrid mesher), it will divide the boundary in 60 segments so we have 61 nodes ...
I specified 60 in square.geo with the keyword "Transfinite Line {1, 3} = 60 Using Progression 1;" (there is 2 lines: 1 and 3 here). the boundary will have 60 nodes and thus 59 segments!!!! that is why i got 3600 elements and 3721 (61 * 61) nodes with elmergrid and 3600 elements & 3481 nodes with gmsh (59*59=3481).

The weird part is ... if i set "element division 1 = 59" in square.grd ... the computation goes wrong as for the meshgrid generated with gmsh.... It seems that the computation gives some interesting results only if the boundary (where is defined the free surface condition) is divided with an "even" number of elements!!!!

I don't understand why? i will perform further tests to confirm it!

See ya all!
Rouquette Sébastien
University of Montpellier / Mechanical and Civil Engineering Lab. / Welding Group
Post Reply