I m analysing the turbulent flow through rectangular duct with heat transfer through one principal wall. There r several models for wall bounded flows.I m using mesh size as per manual guide lines which is giving comparable results with experimental one but still not getting through grid independence test values r abruptly changing by increasing the density of mesh, I welcome your valuable suggestions.

In my case, results varied very slightly when cell number changed from 26820 to around 50000. Is the solution with 26820 cells already grid-independent solution? I am also getting confused. Thanks!

Yes, if deviation is very small,but some more sets of different grid size should be run.

Have you cosidered your wall function. If you are using standard wall fucntion and you increase the density of the nodes near the wall you may have to start using the two layer model. Check your Y+ values. If you don't know what I am talking about check the manual for more information on wall functions.

I m using enhanced wall treatment with 2layer model. I have taken y+ equals to 1 through grid adaption panel and followed other guidelines too. The results obtained with 10000 and 300000 cells. But got results four fold in later case and similar trend in intermediate steps too.