Spurious axial forces in shells

[Jatkinson]

Hi Dr. Mckenna and others,

I was wondering if anyone can think of a reason for spurious axial loads appearing in my shear wall results. I made a short image album summarizing these results and the problem here: http://imgur.com/a/UGWG1

Basically, I have modeled shear walls with ShellMITC4 elements and ElasticMembranePlateSections, and analyzed them using a response spectrum procedure that I wrote. I am trying to take 'section cuts' through the wall to get resultant forces [Fx,Fy,Fz,Mx,My,Mz]. I do this by reading the results from an element recorder with the -forces option specified and resolving the forces to the gross centroid of the section. The results seem to match ETABS for shears and moments, but there is this weird axial force that has me very puzzled (see last figure in album). Theoretically, it should be zero, as ETABS predicts. I've tried removing rigid diaphragms, removing columns, removing coupling beams, but even when I have a bare wall with just static lateral loads, the axial force result shows this pattern! Could this be the result of the geometric transformation used in the shells? Some other part of the shell formulation?

I was hoping that someone might recognize the source of this discrepancy, as I can't think of a reason.

[fmk]

what convergence test did you use?

[Jatkinson]

the norm unbalance test is being used.

Edit: also tried the norm disp increment, with no change to the results.