noslip wall BC not respected

[mark smith]

Elmer 8.2 Rev d3bf396 Windows mpi build

Hi
I have some 3D geometry built in Gmsh with a fluid volume with 2 rectangular inlets and 2 rectangular outlets. I put all the other fluid walls as noslip wall BC = True, the inlets have a uniform normal velocity applied, the outlets have zero pressure.
When I look at the velocity distribution on the inlets, one shows a uniform velocity & the edges are not set to zero, where as the other inlet has 2 edges with a uniform velocity and 2 with zero velocity? Changing the order which the boundary conditions are applied seems to have no effect i.e. if I set BC 1 as the no slip then the inlet/outlets as BC's 2-5 or set BC's 1-4 as the inlet/outlets and BC 5 as the no slip, the order make no difference?
This leads to different flow rates at the 2 inlets which are the same shape and size with identical uniform velocities prescribed.

I seem to recall from earlier postings that the BC priority obeys some rules but I cannot find anything on the topic (perhaps it was on the old forum?) so is this a bug ( I don't remember seeing this in version 7) or is there a way of enforcing BC priority at a common edge between 2 faces?
I tried to impose an edge BC of zero velocity but elmergrid complains when converting the gmsh mesh to elmer format that 1D BC are not allowed.

To make thing clearer see the attached paraview image.

Thanks in advance
Regards
Mark

no slip at edge problem

velocity BC problem.PNG (49.16 KiB) Viewed 3030 times

[mark smith]

Hi All,
After scouring the forum I discovered that I needed to put the following line in the Simulation section of my sif file.

Code: Select all

Set Dirichlet BCs by BC Numbering = Logical True

I cannot understand why this isn't the default as otherwise (as I found) the BC's at shared edges e.g. in cfd an inlet and wall, can be either one or the other with no apparent logic to which gets imposed?

This feature is not mentioned in either of the latest models or solver manuals ;-(

Regards
Mark

[raback]

Hi Mark

The order is as it is because by setting the BCs in the order that the elements are located in the boundary element list is simpler than adding a second loop over the BCs. Now, one might order the entities such that they are in a growing order which would make these two the same. However, no such thing is done and the node/element ordering is inherited from the mesh generator.

-Peter