[jf_vt]

Hi
I was under the ( possibly wrong) impression that one of the uses of mapField was to generate an initial guess on a fine mesh from a converged solution on a coarser mesh that will allow for faster convergence, as illustrated by the cavity exemple in the documentation.

The reason of this post is that based on this assumption I tried to use mapField in a complicated and large 3d unstructured grid with acceptable orthogonality quality ( max orthogonality is always under 64 and average around 40) and all checkmesh test are passed for all grids without errors.

The solver is a variation on simpleFoam with basic k-epsilon model and wall function treatment.

Both cases converge slowly but steadily if I use potentialFoam to generate an initial guess.

I am trying to speed up convergence of larger case using smaller case solution and I fail miserably:

Indeed, if I use mapField -consistent with a converged solution on a coarser mesh as source, the utility works but then the case blow up very quickly even if I strongly lower the relaxation factors.

So my question:
Is there some known limit to the use of mapField to speed up convergence after a mesh refinement?
maybe my refinement is too agressive as number of mesh on the fine mesh is ~3X ( 9 10^6 for the coarse mesh and 26 10^6 for the finer one)?
is there some options to "smooth" the initial guess i am ignorant of?
Other?

I searched this forum and google before posting but found no discussion of such limit. Apololgies if I missed something.

Thanks
JF