Under relaxation parameter required for convergence of simple problem in Fluent
 

Hi All,
Recently, I was analyzing a 3D flat plate with very sharp leading edge (30 degree) using Ansys Fluent (Re=200, Laminar model used, water as fluid, 1 million nodes).
The run completed successfully without any issues with residuals coming down to 1e-6 easily.
Then I tried a case with even sharper edge (15 degree) and the flow started to diverge.
I adjusted the under relaxation factor (URF) for momentum (from 0.7 to 0.6) and the run completed, but took almost twice the time compared to the first case and the residuals never came down below 1e-5 (oscillating at 1e-5).
Can you please recommend any reasons as to why this might have happened, as in why the URF was required to solve such a simple problem? I have read about people using URF for complex processes, but did not expect the need to use them for something this simple.
Also, does the time requirement shift seems to be reasonable?
Any suggestion will be valuable.
Thanks.

For simple problems, the only reason you need to adjust the URF is because of a bad initial guess or poor mesh quality (skewed cells, etc). Usually it is a poor mesh.

Since you found that it diverged with default URF but was able to ~converge with a lower URF, it further supports that you have numerical instabilities (oscillations/wiggles) generated by bad cells. This could also explain why it takes more iterations to achieve the same level of residuals.

However, residuals are not a good indicator of convergence. The residuals are not dropping because they can't, the numerical oscillations prevent the solution from becoming better. If you monitor the velocity, pressure, etc. you might find that it actually converges faster (but not better).

Mesh quality is super super super important for high quality results even for simple problems. The fact that the problem is simple does not mean that it will automatically pass the sniff test.