the fact that you don’t see a convergence in your data is related to the way how you do the convergence study by only looking at the total number of DOFs but at the same time having non-uniform mesh refinements.
I did a new convergence study of your problem using three distinct meshing approaches:
- first order without refinements
- second order without refinements
- second order with local refinements on the fillets (where the high smax appears):
Here the results:
From the results you can see that the mid point stress is converged pretty early whereas the max stress in the model does only converge for a very high nomber of DOFs. This is due to a highly localized stress concentration (see screen shot) where the max stress happens.
Hope this clears your doubts!
Ah, and here the link to the project as reference: https://www.simscale.com/projects/rszoeke/convergence_study/