[panos_metal]

Hello

I simulated a case using 2D(symmetry) and 3D geometry. The 3D case is consisted of about 300.000 cells and the 2D of 10.000. I compared the results of the two cases and i found very very little diference ( about 3% in temperature).
Can i be sure now that i have grid independent solution? it would be more proper to compare the 3d model with another 3D which consisted of lets say 50000cells?

Thanks in advance

[Logan Page]

Start with a coarse mesh and and keep doubling/increasing the number of nodes until the relevant property (say heat transfer) changes by less than 1%.
|Q(i) - Q(i-1)|/|Q(i)| < 0.01 where i is the step in refining your mesh.

Then you know you the solution is insensitive to further grid doubling at (i-1).

[panos_metal]

Hello Logan and thank you for your reply.

Assuming that i have finished with the simulation and that i have validated my results , it wouldnt be proper to check the grid independence by comparing a 2d and a 3d simulation results?
I totally agree that your recommandation is right but it will be time consuming for me now to do this just to ensure that my solutionb is grid independent. So this is the only way?
Thank you in advance

[skullCFD]

This might be very late but may be useful for some newbies. For meshing in CFD, the norm is to never mesh more than what is required i.e., the smart way vs. the brute force way of using an over kill mesh and wasting computer memory and your time. When there is symmetry (2D/axisymmetric) of any type, it has to be used but not a full blown 3D simulation which gives you the same result. For 'panos_metal': you have to check grid independence by refining your 2D mesh further.