Convergence

Hi,

I'm having troubles with convergence. I'm fairly new to FEA so its difficult to trust my results. I understand the concept of convergence but I'm not entirely sure what qualifies as a converged result.

I have a model that I have meshed with a second order Netgen 3D mesh. I have re-entrant corners but these are not showing any signs of singularities to start off with.

If I have my mesh set to moderate and I start to decrease my element size I start to converge towards an answer with reasonable results but when the mesh gets too fine I start to get higher stresses again, not in the corners but just off to the side of them. If I run the mesh at coarse I get similar results but if I run it at Fine or Very Fine I get the offset corner stresses.

I am unable to fillet the corners due to restrictions in the model. I have tried to model this as a plastic material but made no difference (probably did it wrong).

So what my question is, is if I start to converge towards a reasonable answer using a moderate mesh by decreasing elements size can I say that the result has converged or do I have to eliminate any chance of singularities before I can assume anything? And should the result converge to the same answer whether a coarse...very fine mesh is chosen or not or is element size alone good enough?

Any suggestions would be great!

James