Page History

Step 7: Refine Mesh

It is very important to assess the dependence of your results on the mesh used by repeating the same calculation on different meshes and comparing the results. We will re-do the previous calculation on a 30 x 50 mesh as well as a 30 x 150 mesh and then compare the results with the 30x100 mesh used previously.

Modify Mesh in GAMBIT to a 30x50 mesh

The 30x100 mesh is saved as plate.dbs in your working directory. Bring up the command prompt window as in step 1. To copy plate.dbs to plate50.dbs, at the command prompt, type
copy plate.dbs plate50.dbs
We will work with plate50.dbs in order to retain plate.dbs as is. Launch GAMBIT with plate50.dbs as the input file by typing:
gambit plate50.dbs

...

We can see that the courser mesh produces slightly different results, although they are still reasonable. Some numerical error is introduced when the less-refined 30x50 mesh is used. As one would expect, resolving the boundary layer to the laminar sublayer, which we did with the orignial mesh, produces more accurate results than resolving only to the turbulent region. Resolving to the laminar sublayer is not always a reasonable thing to do, especially at high Reynolds numbers. The results from using the 30 x 50 grid show that a reasonable solution can still be obtained without resolving down to the laminar sublayer.

Modify Mesh in GAMBIT to a 30x150 mesh

Create a mesh that is finer than the original mesh to see if our original solution contained inaccuracies due to the mesh. Mesh the inflow and outflow edges at a Successive Ratio of 1.065 and an Interval Count of 150.

...