Different results with meshes generated by Gmsh and Genbox

[Jundi He]

Hi Neks,

I am trying to simulate a simple flow in a rectangle duct, the inlet velocity is 10m/s. At the beginning, I generated the mesh using Gmsh & gmsh2nek, and the simulation diverged quickly after starting. Then I used Genbox to generate the mesh and imposed the same boundary conditions. However, the simulation ran successfully without diverging, and the result seems reasonable. Does anyone know what is causing this difference? The geometry sizes and mesh resolutions built by Gmsh and Genbox are exactly the same.

The simple duct flow is described in this figure:

I have attached the two files for the mesh generation (ch.geo for Gmsh and ch.box for Genbox), and I've attached the case setup files, including the .usr, .par & SIZE file. It should be noted that, the .usr files for the Genbox case and the Gmsh case are different: For the Gmsh case, the boundary condition type for each face are specified in subroutine "usrdat2", for the Genbox case, the boundary conditions are already defined in the .box file. I've named them ch_Genbox.usr and ch_Gmsh.usr separately.

Many thanks for your kind help! 

Jundi

[hy...@anl.gov]

Hi, Jundi,

Why you initialize  uy = 0 ? can you initialize to also 10 ?

[Jundi He]

Hi Hyuan,

Many thanks for the useful suggestion, I initialized the flow to 10m/s and the flow develops faster rather than starting from 0m/s. 

This is an update on my previous problem, it seems that I made a mistake in the mesh, in the Genbox mesh, I used 160 nodes in y (streamwise) direction, and in the Gmsh mesh, I only used 80. This might be the reason for the divergence in the Gmsh case, after changing the Gmsh mesh to 160, the simulation can run successfully. And the result obtained using meshes from Gmsh and Genbox are exactly the same, the below figure shows the streamwise velocity distributions in the Gmsh case (upper) and the Genbox case (lower):

Regards!
Jundi