In foundations, we design for overturning by balancing the acting moment and the resisting moment to a minimum factor of safety. However, there are also cases in superstructure where we rely on bearing to transfer force and need to prevent net tension, using structural weight as ballast. I'm working on an existing building with slab on metal deck over bar joists for the roof, where the original designer balanced out uplift forces by providing just enough dead load at ~0.6D to counteract the ASD wind. Overturning from wind acting on the side of some new RTU's creates a small net uplift; my task was to check whether it would unzip the whole bay, assuming a bearing-only connection to the joists. As it turns out, the roof could maintain net downwards force on the joists at ASD wind (0.6D+0.6W), but it failed at LRFD 0.9D+1.0W. I was surprised to realize that the ASD uplift combination is less conservative on dead load resisting uplift (0.6:0.6 = 1:1 vs. 1:0.9). Strange.
When the net uplift is caused by dead load and resisted by dead load, it gets even stranger. It's not uncommon to have long cantilevers (three or four times the backspan) in contemporary construction, resulting in permanent uplift on the back column, under dead load alone. The back column may be stabilized by the dead load within its trib area: dead load against dead load. It got me wondering about design for uplift when dead load is both the bad guy and the good guy.
Say we model the above frame in analysis software with auto combinations--dangerous, I know. The software would create a demand envelope considering many combinations, using 1.2D for downward effects and 0.9D for uplift. If the wind and snow loads are very small by comparison (let's say it's a decorative canopy inside a shopping mall that experiences virtually no environmental hazards), there is a scenario where the dead load could be almost perfectly balanced; say 10 kips uplift due to dead load on the cantilever against 10.5 kips downwards due to the structural weight within the trib area of the column. In real life, the detrimental dead load could easily have been underestimated by a small margin, lifting the column under dead load alone. However, the software would show no net uplift at target reliability.
I'm curious as to whether anybody's come across this, and what method you would use to meet the standard of care. Is it appropriate to design to 0.9D_resisting + 1.2D_acting (equivalent FS = ~1.3)? Perhaps 0.9D_resisting + 1.4D_acting (FS = ~1.5)? Better to use a foundations approach with acting vs resisting forces at FS=2? I imagine this FS is intended to consider variability in soil weights, which are much higher than variability of dead load. Is it possible to perform such a check in a single software model?
Excited to hear some thoughts on this. Feel free to comment on just part of this very dense post.
Christian Parker P.E., M.ASCE
Structural Project Engineer