Energy Error Norm as aid in convergence check?

[321GO]

Hello guys,
-----------------------------------------------------------
on a high stress location(small fillet in transition) i have an Energy Error Norm of 75.

This means 75% stress deviation between 2 adjecent elements, correct?
Element vs nodal value also show a (large)deviation.
-----------------------------------------------------------

Question:
When to call it converged, where to draw the line?

Second question:
can one base the decision on wether or not to re-run with finer mesh, on these Energy Error Norm results?
(in the case of long model solve times)


Thank you, greatly appreciated.

 

[EngAddict]

To my understanding, energy norm averages the stress across the nodes of the element to give element stress, then compares it to the nodal stress. You can do something similar by plotting stress using nodal and then repeat with element and compare the difference.

I would not say you could use it for convergence. Generally you would plot stress vs mesh size, refine your mesh and repeat. Continue to refine your mesh until your stress remains about constant. Make sure not to measure at a singularity otherwise your stress will increase infinitely.

Energy norm may only indicate that your mesh has bad structure, i.e. it does not fit the structure well, it may be too stretched, etc. But your mesh size would be ok. I would normally refine my mesh in areas of high stress and coarsen it in others, then run a simple convergence to check the results.