modeling a box section using shell elements

huangyhg

modeling a box section using shell elements
folks,
recently i was experimenting modeling a box shape using shell elements in sap and etabs. i ran into the problem that the computed moment of inertia (moi) was different from what the program was computing.
i modeled a pin-roller condition of a given span and backtracked i using the deflection equation. the actual i was 20% higher than the backtracked i.
i modeled the box using center-lines and thickness of the elements.
has anyone else run into such a problem and if you did, what was your approach to modeling it right?
try a 4' wide x 4' deep box with 2'x2' void (= 12" thick box).
was this a concrete element or steel?  is it possible you used a different e than the program?
a few things come to my mind. one, how long was your span? if you had a 4'x4' box with a short span, shear deformations could be controlling rather than moment deformations, which would not be accounted for in your equation. i would use a span/depth ratio of at least 10. how fine was your mesh? if you are using 4 node elements they may not be acting fully composite over their length. especially with such a large cross section. i'd refine your mesh and see if the values converge. also, how is your load applied? if you apply a point load at a single node in that type of model, you'll get large local deformations as well, making values around that area invalid. one more thing. are you using the correct offsets for you elements? if you extrude the view does it create the box section your assuming? re  
e was correctly accounted for.
span was 60'
load was applied as a uniform force on the top slab.
yes, extruded shape is almost the same as section assumed (except for missing chunks at centerline intersections.
what about your mesh? i would suggest at least two rows of elements per side to place a node at mid height each side.
the mesh was very fine. it was a 6"x6" mesh. i would appreciate if someone could try modeling in their fea program to verify results.
span = 72'
loads = 100 psf x 4' wide (400 plf)
width of top slab = 4'
did you try simpler problems first?  if not, the you should.
start with a cantilever plate bending about its weak axis with a point load at the tip.
then do the same problem with the plate bending about its strong axis.
i suspect that the difference is either shear deformation or something slightly wrong with the boundary conditions in the shell model.  try to use a cantilever so you eliminate the chance of screwing up your boundary condition at hte roller.
for what it's worth, i've done this kind of thing and after messing with it a while, the two models should be closer than 20%.  maybe 2-3% different.
i just read your last post.  your loading method might be part of the problem.  you applied the load as 4 psf across the beam.  this will make the top shells bend between the vertical walls and i don't know what this will do to the beam's overall stiffness.  this is an example of introducing too many variables at once.
you could also try a cantilever with an applied moment at the tip.  that eliminates shear.
start simple and build up to the more complex problem.
slickdeals,
you could model it as the sum of two square areas (one is negative):
  i = (4*4^3 - 2*2*3)/12 = 20 ft^4
does that agree with your calculation?  you will not get the same result using centerlines.
  
best regards,
ba
@ba,
i know what the moi should be. i want to model it right using shell elements so as to capture torsion in the box due to uneven loading etc.
as a result, i started the model but in making quick checks, i found discrepancies and hence, wanted to know how people typically model them.
thanks
not familar with thin shell structure in computer model, but know it behave differently from plate element. have you checked the manul and make sure your elements meet the definition/limitation of thin shell element?