Currently when particle drag force is added to the gas phase it goes into FVX_D, FVY_D, etc and is then added to FVX etc.
At mesh interfaces the FVX values are matched by averaging... but the drag force at the mesh interface should be the sum, not average, of contributions for either mesh. The current approach leads to low values. This can be seen with a slice of FVX_D in the vegetation_drag_1.fds verification case (I split the mesh in the streamwise direction).
I have a fix where FVX/Y/Z_D values have an MPI exchange and summation at the mesh boundaries before being added to FVX/Y/Z ...and then flux matching can proceed as usual.
This is the result of the proposed fix
which now matches the case with no split mesh
Currently when particle drag force is added to the gas phase it goes into
FVX_D,FVY_D, etc and is then added toFVXetc.At mesh interfaces the
FVXvalues are matched by averaging... but the drag force at the mesh interface should be the sum, not average, of contributions for either mesh. The current approach leads to low values. This can be seen with a slice ofFVX_Din thevegetation_drag_1.fdsverification case (I split the mesh in the streamwise direction).I have a fix where FVX/Y/Z_D values have an MPI exchange and summation at the mesh boundaries before being added to FVX/Y/Z ...and then flux matching can proceed as usual.
This is the result of the proposed fix
which now matches the case with no split mesh