Learning Forum

[joao.faustino]

Hello everyone, I am looking for recommendations and best practices to improve simulation accuracy in Ansys Fluent without drastically increasing solve time or running into out-of-memory (RAM) issues. Specifically, I would like to know:

• Which meshing strategies (e.g., Poly-Hexcore, inflation layer setup) yield the best accuracy-to-cell-count ratio?

• How to effectively manage hardware resources (RAM/CPU cores) to prevent system crashes during meshing or solving?

• Recommended convergence criteria or solver settings to balance speed and stability.

[Rob]

Yes. Don't focus too much on overall cell count, you're looking to focus cell count where it's needed. So, a 1M cell mesh with uniform size is less useful than an 800k cell mesh where cells are clustered into regions with high gradients.  Inflation helps near the wall, but things like bodies of influence, local sizing and decompostion are better in the free stream. 

For meshing I'd use 1-2 cores generally. Whilst parallel will speed up some operations Fluent Meshing is not as good (yet) as the solver for optimal use of cores. 

Default convergence is usually sufficient but you also need to consider monitors. Residuals are only part of the story. For solver settings, default are usually fine, but I find in complex flows that the PBCS Courant Number/time factor are a little too high. Unfortunately there isn't a magic number or settings that'll make everything work: make sure the mesh is good before going near the solver will help more than messing with UR factors etc.