Our stress update is done using increments of deformation. The increments are calculated in the local coordinate system and used to calculate stress increments. This is very fast, but for large shear, the choice of coordinate system changes the result. It's not obvious in this problem because the motion is prescribed. However, if force loading was used instead, the deformation would not be at all the same. We really cannot solve this problem well using an updated Lagrangian type of approach. There is no good fix for this.
The transformations that we do in the element and output are a good match so there is nothing to fix there either. I'm going to close this as a "wontfix".
lpb
Bug 12322
Ticket#2016052610000153
______________________________________________________
Stress and strain output corresonds to integration point locations.
Thus Lobatto integration (see INTGRD in *control_shell) is required to get
output corresponding to the outer surfaces.
In the case of stress or effective plastic strain, MAXINT in *database_extent_binary
controls how many of the shell through-thickness integration points are output.
If MAXINT is negative, all 4 of the in-plane integration point stresses are output
(no intpt averaging is done) at each of the |MAXINT| through thickness locations (applies
only to fully integrated shell formulations; affects only d3plot, d3part, d3thdt).
For corresponding control in elout data, see INTOUT in *database_extent_binary.
***NEW ~ Sep 2012 ***
In *control_output:
EOCS Elout Coordinate System: controls the coordinate system to be used when writing out shell data to the elout file:
EQ 0: default
EQ 1: local element coordinate system
EQ 2: global coordinate system
***********************
Tensorial (true) strains at output at the uppermost and lowermost integration points
only and only when STRFLG=1 in *database_extent_binary.
See the tables under *section_shell for integration point locations.
See also:
shellstrain
shell_resultants