mesh convergence for a free mesh (ABAQUS)

[ChuckNorris1]

hello guys,
How do you go about doing a mesh convergence test for a mesh created through free meshing?

I picked a specific point in the model and tried to do a von mises stress convergence at that point. But the problem with this is that I have only approximate global mesh size on the x axis (free mesh). Also, I don't think this is reliable.

Has anybody got some ideas?

also here are my values from a dodgy convergence test I just carried out:

Global Size(Von mises)
100 (66.98)
80 (81.21)
50 (61.74)
25 (103.05)
20 (106.53)
15 (103.12)
10 (118.57)
5 (133.88)
2.5 (135.12)

http://postimg.org/image/657aanl9p/full/

pic: 'free meshing' is the only option for me

 

[corus]

Not sure what you're trying to achieve here as you've more or less proved you've got convergence to with 1 or 2%. You could get rid of the free mesh by partitioning around the holes using 2 perpendicular planes that divide the circle region into 90 degree segements, and repeating for the 2nd hole. You should get structured regions then. You'll still only have an approximate element size though, just as you have with the free mesh.

 

[ChuckNorris1]

Thank you for the reply,

I guess, I am looking unnecessarily for a nice smooth curve like the one below (but with stress):

http://postimg.org/image/ot2wjp8lv/

so I should work with percentages ? rather than trying to get a smooth graph.

because the stress went up and down on these global sizes :
100 (66.98)
80 (81.21)
50 (61.74)
25 (103.05)
20 (106.53)
15 (103.12)

 

[ChuckNorris1]

Nevermind, I should have clicked the number of nodes .. and plotted 'no. of nodes vs stress at a point'

For anyone interested :
Tools - Query - Mesh - Select face

although I am pretty sure you all know this ... but my name is Chuck Norris .. so do it !