[Joh]

Dear all,
How can you assess if the boundaries are sufficiently big enough to deal with your flow? So for unstructured mesh, how do you know how many prism layers to use. I use 3, split them into 6 and then rearrage them with a ratio of 1.4 and it seems to work. But how confident can I be that this will give the right results?
Regards,
Joh

[siw]

In answer to your question, you will need to conduct a simulation to see if the results are physically sensible/what you expect/match other sources of data. Only then can you know if the modelling of the boundary layer etc with your chosen generation of prisms was okay. If not you'll then have to make some mesh changes and repeat. Personally, I keep the expansion ratio to 1.2 or lower. Also, if using CFX (so I know it's in the manuals) but probably FLEUNT aswell it recommends 10 to 15 layers for RANS simulations and the y+ depending on the turbulence model and near wall treatment of choice.

[Joh]

Thanks, why do you keep the expansion ratio to 1.2 or lower?
What about for structured meshes, how do you know how far the O-ring has to be placed for laminar flow? Is it the same idea? Run the mesh and see if it gives a plausible solution?

[siw]

1.2 is the default value in ANSYS Meshing and it is the recommended value I've found in various user manuals, reports etc. However, sometimes I use lower values, but not hgher values. Give it a try and see how/if your results are dependent upon the value.

Again for the O-grid question, you'll only know once you've tried some different lengths for your particular simulation and its effect. I find with mesh generation, you just have to make what you judge to be a reasonable mesh for your simulation (to as best a quality as you can get it) and then run the simulation and repeat from there as you think best.