On a 20‑story RC flat plate with an L‑shaped plan and a big atrium, a rigid diaphragm in ETABS gives fast iterations and clean collector/chord design, but switching to semi‑rigid shells blows up solve time about 5x and bumps story drifts 6-8% with clear redistribution of diaphragm shear. Do you accept the rigid assumption and backstop with targeted hand checks, or eat the runtime/meshing pain to capture in‑plane flexibility and torsion for wind/seismic?
I keep rigid for global runs, then clone the model and do semi‑rigid only at the L‑shaped/atrium levels with a membrane slab and about 6–8 ft mesh; that drops the 5× hit to about 2× while still catching the diaphragm shear shuffle for collectors/chords. On a 20‑story flat plate we saw the same 6–8% drift bump, and the semi‑rigid check only nudged a couple perimeter collectors about 10%, so we detailed those and called it good.
On my last L-plate, ‘drift about 6–8% without the 5× slog.’ matched my experience — cracking the diaphragm membrane to about 0.4E and plate M11/M22 to about 0.3 on just the atrium levels let a 10‑ft mesh hit strip hand‑check deflections and kept collector forces within about 5%. Small caveat: model the chords/collectors as frames with realistic stiffness or the shell will over‑relax and inflate drift, @OP — have you tried modifier calibration?
@OP I stick with rigid for the 20‑story global model, then spin up a stripped semi‑rigid “check” with only the atrium stories active, single‑step P‑Delta, and about 30 Ritz vectors so ETABS still shows the shear shuffle without the long runtimes. If the drift bump there’s noticeable, I crank accidental torsion to about 7–10% just at those levels in the rigid run and size collectors/chords to the semi‑rigid shears.
Quick example: when I need the redistribution around an atrium without the runtime hit, I switch the slab at those levels to a membrane property (no plate bending) in ETABS so I keep only in‑plane DOFs; drifts land within single‑digit percent of a full shell but it solves close to the rigid model. I still use rigid for the global, and this “membrane check” has flagged collectors I would’ve missed — just don’t use it for punching or vibration decisions, @OP.
I’ve had good luck keeping the full model rigid but assigning semi‑rigid only to the atrium stories, with a 3–4 ft mesh just around the opening and coarser elsewhere; that caught the shear redistribution and kept the solve near rigid times. I also cut column rigid zones back to the face of slab at those levels — otherwise the slab reads stiffer in‑plane than it really is. If your accidental torsion ratio or chord demand at the re‑entrant corner jumps, @OP, I’d run one full semi‑rigid case as a backstop; CSI’s note helped: CSI Knowledge Base.
I keep the global model rigid and fake the atrium flexibility by splitting the floor into two rigid diaphragms tied with axial links sized off a slab strip’s AE/L; it captures most of the shear shuffle without the 5x runtime. Before trusting it, @OP, I compare the ‘drift bump’ to the +/-5% accidental torsion envelope and tune the link stiffness with one bare-bones shell run of that level.