how can it affect accuracy when simplified by ignore and drop all mid nodes in definition?
Afraid I cannot follow what you try to say here :s
The model uses now C3D20 elements which contains mid nodes as well.
Your screenshots are made in a different software based on the surface? External software likely recognises or plots information not automatically recognised by Prepomax or Calculix?!
i understand curved shell element mentioned is quadratic element with midside nodes following geometry since it’s not flat as linear element.
yes right, many external software generated shell element local axes automatically by default but later it still allowed to modify.
Hi Frank,
Hope we see you more offen in the forum. 
You could ask in the Calculix forum too. https://calculix.discourse.group/
Rsmith user who wrotte orientations for the elements code is there and active as part of the community.
Regarding this particular problem I think your approach needs a warning.
Assuming that the symmetry condition is applied correctly (picture seems not), one should ask first if the model meets the conditions to apply a symmetry BC.
There must be symmetry in the geometry (ok), symmetry in the loads (ok) and symmetry in the material properties (ortothropic!!!).
Imagine that your fibers are the elements of the lattice in the upper picture. If you reduce the lattice by half and apply a symmetry BC condition, you are completely changing the lattice stiffness and enforcing a symmetric displacement field at the boundary which a priory shloud not with an ortothropic material.
1 Like
Good points @Anys! Room for thought. I believe, it will be fine for 0 and 90 degree orientations but the issue you describe will likely affect off-axis orientations indeed (e.g. +/-45). OF the top of my head, it likely needs a different type of symmetry in this case… rotational?! I’ll think about it and let you know.
That’s why I said about using some known benchmark. It’s hard to know what to expect and differentiate a mistake in the orientation form an unexpected behavior.
For example, a cylinder with one of the axes (+45) clearly stiffer than the other can make the cylinder twist under internal pressure. One could think the model is wrongly constrained and remove that twist when in fact it is its natural behavior.
Be sure to check all reaction forces on the BC are zero.
Just a quick edit on the original post. The orientation keyword for 0, 90,45 and -45 would need to look like:
*ORIENTATION, NAME=OR_X
1.0, 0.0, 0.0, 0.0, 1.0, 0.0
*ORIENTATION, NAME=OR_Y
0.0, 1.0, 0.0, 1.0, 0.0, 0.0
*ORIENTATION, NAME=OR_45
1.0, 1.0, 0.0, -1.0, 1.0, 0.0
*ORIENTATION, NAME=OR_NEG45
1.0, -1.0, 0.0, 1.0, 1.0, 0.0