Reply To: Mesh convergence more mesh get lower stress

[Kaushal Vadnere]

 

Hi Patrick,

The stress value should rise with mesh refinement in case stress singularities are present. In the graph you shared, it is showing that the maximum stress values remain fairly consistent with increasing mesh count indicating absence of singularities.
In case where the mesh is coarse, bad elements can get highly distorted with large deformation and result in higher stress values. Mesh refinement in this region can lead to better element quality and hence, better stress distribution (This may result in lowering of stress values as well).

To answer your other question regarding the wrong mesh genration, by looking at the mesh, it seems there are no generation issues.

Before coming to any conclusion, I would want to know few more things: 

What is the use of the upper roller (bigger one)? because you have mentioned that you are applying joint-rotation to only the bottom roller. is there any contact defined between the bottom and upper roller? 

In the 2nd screenshot (Equivalent Von-Mises Stress) is display set to auto scale (>1x) or True Scale (1x) ? If it is at true scale, the element aspect ratios seems unreasonably high. By that I mean, the hex elements look more like a plate rather than cube which is an ideal shape of hex elements.

Also, you have used 7 elements through the thickness, is there any specific reason for that? if not, this will result in high element aspect ratios as mentioned above and un-neccessarily increase mesh count and solution time. I would suggest to go with 3-4 layers and try to keep aspect ratio close to 1 (This means keeping the length, width and depth of the element same). You can define element size = length of smallest edge/3 or use edge sizing in mesh to define no. of divisions for each edge.

What contact settings are you using?

Which support are you using and where? My assumption is that you have applied fixed support to the botton of the plate