[mahdypars]

hello to every body
i used flow 3D to simulate a wave break simulation, and the results was very good. after that for mesh independence study, i used the exact simulation with more mesh (2x cells in each direction). unfortunately the results was very (very) bad and unbelievable!
i was expecting that more mesh, would lead to more accurate result (and at least the same result), but it didn't.
please help me, why more mesh lead to worse answer ?!?
thank you in advance.

[Davahue]

Dear mahdypars,
I noticed something similar a while ago

http://dx.doi.org/10.1016/j.envsoft.2016.02.018

If you check Figure 3c (model apparent order, according to Celik et al. 2008), FLOW-3D converges faster than OpenFOAM. However, if you refine excessively, you can get still some flow relocation. While I'm not sure why this happens, the conclusion from this Figure 3c would be: keep it simple, works better. If you do not have access to this paper, I think one coauthor shared it in Research Gate, maybe you can check.

But I fully understand the point on conducting a mesh sensitivity analysis. So, I worked a bit more on that for a different case and got something interesting..

http://dx.doi.org/10.15142/T340628160853

This is open access. If you check Figure 2, two meshing strategies are compared and results are quite promising. These numerical oscillations disappear leading to a more stable convergence even for turbulence quantities, which are usually more sensitivity.

Sorry if I over focus in my own work, but it is easier to talk about my own experience than others. Sure you can find very interesting works on that from others!

Best regards,
Daniel