I'd say yes, it does. I create my structured walls by piling boxes on each other in Salomé, then creating a partition of the various boxes, then creating respective groups in this partition to be able to attribute different (intermediate) surfaces with BC in Elmer later on. The whole process is described rather well in http://www.elmerfem.org/wiki/index.php/ ... e_to_Elmer .It would be better if the mesher cleaned up the duplicate surfaces and created a continuous mesh. Do you know if Salome can handle this problem?
The UNV mesh that is exported for the complete calculational domain, consisting of all participating bodies, carries the body indexes over to Elmer, and the junctions between them consist of only one surface. The UNV net can be directly loaded into Elmer, but beware of potential "locale" issues, namely the necessity to exchange "." against ",". For this, see some other thread here.
There is the necessity to decompose individual surfaces again in ElmerGUI, though. Here an angle value of 70° always worked for my basically rectangular geometries.
All in all, I feel Salomé to be more practical, and presently we are trying to go the whole way round from parametric geometry definition to scalar values extraction to couple the simulation to an evolutionary black-box optimizer module on the longer run.
Regards,
Peter