Hi,

I am having trouble implementing inlet, outlet and wall boundary conditions in my CVFEM (control volume based finite element) code which uses unstructured tetrahedral meshes. If anybody knows about some helpfull references about this, please let me know!

Thanks, Matthias

hi,Matthias

I have implemented the same recently on unstructured grids. Some tricks are necesscary to implement the B.C's . 1) all boundary conditions are given on nodes explicitly, not edges, although edge will carry a tag identifying it as a boundary edge. u,v,T, P at inlet are specified directly on nodes 2) I use ghostcells (also know as RIND cells).

For inflow the cell center value of ghost cell (for u,v ) are specified as average values of boundary nodes for edge shared by ghostcell and boundary cell 3) while reconstructing for getting nodal values u should skip for boundary node so that value remains

same throughout iterations. 4) for solid stationary wall specify u,v=0 and put

u,v=0 on cell center values of ghostcells. 5) to put Neumann du/dy=0 specify the same value on ghost cell center as that on the neighbouring cell center. I don't know the references but this is what i implemented after considerable R&D and seems to work well.

hope this helops Abhijit Tilak.

Hi Abhijit,

thanks for your usefull answer! One problem I encountered was how to create these ghostcells. I have curved boundaries and simply mirroring the tetrahedron cells didn't work out. How did you do this?

Thanks again, Matthias

hi matthais,

Actually our's is a 2-D code.The whole idea of ghostcells is to pickup correct boundary values and facilitate the specification of BC's. The ghostcells donot enter into the solver at all and no calculations are performed on ghostcells. but their presence is needed. U will also need to generate connectivity information for ghostcells. Actually in our code the coordinates of ghostcells are not needed. Though u may generate them by mirroring boundary cells. Its actually like iamgining that ghostcell center value is present at boundary edge midpoint. then u calculate cofficients by imagining a Cntrl Vol. formed by boundary cell center boundary edge nodes and the midpoint of that edge, Something similar to half-a-control volume used in structured grids. say ur interoir cells are from 0 to MaxNoOfcells then ghostcells index starts from MaxNoOFCells+1 and the solver runs from 0 to Max.. so ghostcells don't enter solver at all. Shape of boundary should'nt matter, right? if u still don't get it elaborate ur problem further. hope this helps.

/**************************/

Abhijit Tilak

Aerospace Engineering Dept.

IIT Bombay, Mumbai 400076

India.

/**************************/

Thanks a lot for the detailed explanation. I will give it a try now.

Regards, Matthias