[Michael-STM]

Hi, i'm trying to understand how a CFD app for my tablet works...there is something that is bugging me.
The grid looks like the one that i attached here, how do they have implemented the no-slip boundary condition here? It looks to me that there is no "real" normal on the grid from any point on the solid except for the very front...anyone willing to make me understand? Infos with diagrams is appreciated!!!
Thanks!

[FMDenaro]

Originally Posted by Michael-STM

Hi, i'm trying to understand how a CFD app for my tablet works...there is something that is bugging me.
The grid looks like the one that i attached here, how do they have implemented the no-slip boundary condition here? It looks to me that there is no "real" normal on the grid from any point on the solid except for the very front...anyone willing to make me understand? Infos with diagrams is appreciated!!!
Thanks!

Are you asking for the treatement of the BC.s at the corners? Even if this issue can be seen a special cases and treated separately, you can just set zero the velocity vector.

[Simbelmynë]

There are a few options. Assuming this is a FDM code and you use the nodes to store values. In this case you have a singularity at the corners if you have Neumann boundary conditions at the wall (e.g. for the Pressure Poisson equation). The singularity can be avoided by using an average of dP/dx and dP/dy at the corner node. A better option is to incorporate the Neumann condition as a source term in the interior nodes, then no averaging is needed. This means that interior nodes on the west and east of a corner would recognize the corner as a vertical wall and interior nodes on the south and north would recognize it as a horizontal wall.

If the nodes in the picture represents the cell centered values (similar to most FVM codes) then the real boundary is 1/2 cell inside the domain. In this case you do not have any singularity and the Neumann boundary conditions are easily found in a similar manner as the second approach described above (in this case the velocity at the cell centered wall location is not set directly by the Dirichlet condition).

[Michael-STM]

Thanks guys! Now my toughts about it are a little bit less cloudy