Checking if a structure is sway or nonsway (for deciding buckling lengths)

[zeeblake]

I'm looking for some systematic and reliable procedure for checking if the assumption of applying buckling length coefficients as "1" to all bars in a structure is on the safe side or not. That assumption can be considered to be on the safe side if the structure is nonsway.

Searching the forum provided me with some hits, and it seems the most "right to the point" topic is this:

https://forums.autodesk.com/t5/robot-structural-analysis-forum/buckling-analysis/m-p/6879200

However, I still have some important questions after reading it:

1- Why is it advised to use only the most critical vertical load case for the buckling analysis? I always thought that horizontal loads (i.e: wind) have a leading role when assessing if a structure is sway or nonsway. My first guess would be that the sway/nonsway check should be performed in all ULS combinations, and that the structure should be assumed to be nonsway only and only if it's nonsway in all ULS combinations.

2- Do you need to split bars for getting an accurate result for the sway/nonsway check? Reading other forum topics about buckling analysis I found that Artur greatly advises to split bars (in 6 parts for example). Splitting bars would be non-convenient for my purposes, because it would basically mean having two models: one for design, and another for the nonsway check. I'd prefer to have one model only. If the accuracy of the sway/nonsway check depends on splitting bars, what happens if I don't split them: will the check be on the safe side for the nonsway result? Or maybe it's not possible to say if the inaccuracy goes to the safe or unsafe side?

3- In the link above, it's said that you don't need to change the combinations back from "buckling analysis" to "linear analysis", because the buckling results contain the linear results as well. However, is there some chance that a result from buckling analysis is more critical than the one from linear analysis, and thus it will be used for design? I prefer that members design uses results from linear analysis only. Thus, should I have duplicated ULS combinations, so that I perform member design on the ULS combinations that have linear analysis only?

4- Last, but not least: Is buckling analysis the best way for deciding the sway/nonsway behaviour of the structure? Or is there other more direct way of checking it in Robot? Also, would a P-Delta analysis be better suited for the check, or is the buckling analysis enough for this purpose?

Thanks a lot!!