When We Model the Spandrel Beams in ETABS

When we model the spandrel beams in ETABS, we can model them as shell elements or as frame elements. How and why the spandrels internal forces outputs can vary using one than the other? Peter Placzek There is no difference and/or no comparison when modelled correctly. For deep beam L/d<3.5 finely meshed shell element is the only correct model. For shallow beams L/d>3.5 frame element is the most "economical" way of modelling. If shell element is used its stiffness will depend on how fine it is meshed. (The finer the mesh the softer it will be) But extreme fine meshing is not desirable - it slows the analysis too much. I have tested it and come up with F12=0.3 stiffness reduction of spandrels instead of extreme meshing. For L/d<1.0 no issue (spandrel is very stiff) For L/d=1 to 3.5 use shell with stiffness reduction F12=0.3 (actually came up with formula F12 =0.44-0.09L/d, but 0.3 does good enough job) For L/d>3.5 use frame element to avoid any issues. Same incidentally applies to choice between shell pier and frame column (for example the piers in the lift core at the doors).

Giorgio Albieri Great. Thank you Peter. At the moment I'm modelling the spandrel beams as shells and what I've found that their axial force output gets an odd diagram. Tensile at one end of the beam and compressive at the other end. Is this correct? Is this caused by the floor diaphragm? Peter Placzek must be the floor Giorgio Albieri Yes. It must be the floor. But, theoretically, shouldn't we expect a constant compressive force in the spandrel beam? How can we deal with this odd axial force output (tension at one end and compression at the other end)? Peter Placzek ETABS spandrel design takes the axial loads into account. They are secondary forces of little concern. Actual value will again depend on the meshing of the floor. You could factor the spandrel F11 stiffness down to minimise the axial load. Syed Rummaan Ahmed, M.Eng. EIT Giorgio Albieri if you remember i suggested you (at London Tall Design Course) to assign different spandrel label through out the length at small intervals. In this way you might achieve what you are looking for. Spandrel forces are integrated at begin and end of each spandrel assigned with the same label while they vary linearly in between end value. Hope it helps

Nikolaos K. Gkogkos The f11 must be modified to incorporate the cracked stiffness properties (moments come up as an integral of s11 normal stresses). No comment on f12 modification. On the other matter, if you have assigned a Spandrel name on an element and you are using shell elements to simulate the floor slab then the axial loads at the edges comes from the interaction between the slab and the spandrel (a product N•e where e is eccentricity between the mean plane of the slab and the centreline of the spandrel is defined due to the interaction). The same issue (and the same results) will come up if a linear element representing the spandrel will be "placed" eccentrically to the slab through the insertion point and the stiffness modification is requested (if not it will just be a picture and not a consideration of the actual stiffness of the beam based on the Steiner's term A•e). Furthermore, shells (thin) and frame elements have sufficiently different matrices. For in plane actions, shell are more accurate but if you are dealing with short spandrels (that is aspect ratio less than two that they are probably governed by shear) the current codes do not treat them properly with the exception of NZS3101 that requires increased shear stiffness.

Nikolaos K. Gkogkos In addition, spandrels that extend over both sides of a slab attribute their mass to the story that they are assigned (i.e above the level 1 slab the mass of the spandrel will be encountered to the weight of the level 2 slab) and therefore you need to be careful with the weight automatically assigned to the floor levels. The out-of-plane stiffness (in plane to diaphragm) provided by the shell simulation in comparison to that of the frame one is another issue as well.

Mahinda Ellegala CPEng MIPENZ Connectivity to the walls will be different based on the selection.