Von Mises stress smaller than zero value

JRP

How is it possible to get Von Mises stress value smaller than zero? The definition of Von Mises stress shows that it should be always larger than zero. Am I reading someting wrong?

Victor

The values at the nodes should always be nonnegative. A common way nonphysical results like these can appear is through the interpolation for finding values between nodes. It uses the elements' shape functions to interpolate the colors on the display. If they're quadratic, they allow the values between nodes to be less than the all the values at the nodes. This is OK because the error vanishes with mesh refinement.

That's a common explanation. However, it looks from your pictures that there are a lot of negative values all over adjacent elements, which doesn't seem to fit it. Could you post the liml file or more detailed screenshots with high texture resolution (Tools -> Options -> Contour plot -> Resolution) showing the node locations (turn off Show thickness)?

JRP

Here are some detailed screenshots with resolution set to highest value. I forgot to mention that I'm working with shell offset - half the thickness of the tube. I use this because it is a lot more work to create offset surfaces with CAD software. I use quad8 and tri6 elements, analysis type is static 3D. 

This representation of results seems to occur only on bottom surface of the shell element. I don't see it on any of the upper surfaces.

One question here.

It is hard to check bottom surface of tube structures in Mecway, because of tube has closed shape. Is there any way I can show bottom surface results on the upper surface? Or any other easier way? I didn't try but if I reverse element normals this should be opposite - bottom surface on the outside of the tube, am I right?

Victor

From these pictures, the negative values only seem to be patches in the middles of elements, so it looks like just the interpolation used by the display and you can ignore it. The node value look to be all non-negative.

The existence of patches could be because the elements are flat, rather than having their midside nodes fitting the curved surface, so it's a faceted tube instead of a smooth one. Or it could be an effect that the quad8 shells show in some cases where the stress alternates across each element.

About your question of viewing the inside stress. Yes, you could invert the elements. You could also use the cutting plane slider to hide the elements covering them. I realize that this way of showing all the stresses together on one plot is difficult sometimes so I hope to have a way to separate them in a future version.

JRP

Yes, the node values are positive, I checked. I can check it with better resolution, thanks for info.

In which version was the cutting plane introduced? I can't find it in version 4. As of separating results - that would be great. I'm always thinking how these things were solved in Ansys, because I did a lot of FEM in it. You could show top, mid and bot results, which one you choose, separately.

Victor

The cutting plane has been in all versions and is the slider at the top-right of the toolbars. Slide it half way to hide the front half of the mesh.

JRP

Ok. Thanks. Didn't notice it before. It will make some things easier now.