[mazhar16823]

Hi everyone,



I used a mesh size which provided me the converged solution while resulting in the correct trend for monitoring quantities such as skin friction coefficient. But when I reduced the mesh by 10% of its previous cell size, the trend for the same quantity becomes quite random. I never expected this, since refinement has to go in favour of considerable accuracy. Any suggestions, why does this happen?


Thanks for your time.



Regards,

[CFDfan]

Originally Posted by mazhar16823

Hi everyone,



I used a mesh size which provided me the converged solution while resulting in the correct trend for monitoring quantities such as skin friction coefficient. But when I reduced the mesh by 10% of its previous cell size, the trend for the same quantity becomes quite random. I never expected this, since refinement has to go in favour of considerable accuracy. Any suggestions, why does this happen?


Thanks for your time.



Regards,

what software did you use?

[mazhar16823]

[QUOTE=CFDfan;779975]what software did you use?[/QUOTE


I used STARCCM+;

[bluebase]

Hi MA,

mesh refinement does not always lead to a more converged solution, such as in text book cases.
Especially in turbulent cases, the resolvable flow features are determined by local mesh density. It can happen, that you have a domain, where a finer mesh will resolve more of flow features which might alter the simulated flow. The consequence could be a seemingly stable flow shifts to unstable/transient flow detaching.

Having only two simulations with such a small density difference mesh seldom lead to sufficient confidence that you achieved a "good" solution.
In general make at least 3 different simulations to assess whether these simulation can be assumed to be converged. Go further, if you haven't found a trustworthy solution. This is especially required in academic problems.


Best regard,
Sebastian

[CFDfan]

[QUOTE=mazhar16823;779978]

Originally Posted by CFDfan

what software did you use?[/QUOTE


I used STARCCM+;

I read somewhere about that and it was related to reaching a threshold of the mesh cell aspect ratio that screws up the solver, thus it might not be an issue with the physics but with mathematics. If you make the mesh even finer and the solution converges then you are in business again

[mazhar16823]

[QUOTE=CFDfan;780128]

Originally Posted by mazhar16823

I read somewhere about that and it was related to reaching a threshold of the mesh cell aspect ratio that screws up the solver, thus it might not be an issue with the physics but with mathematics. If you make the mesh even finer and the solution converges then you are in business again

That could be a good reason but I wouldn't go for "further refinement" because it will bloat the cell count so I am making the farfield a little finer to achieve the fine mesh since the geometry has Y+ within 5.