Column Effective Length

[mkrei]

I have an existing column that is just above a column splice location. The lower column extends 4' above the finish floor elevation and is a W14x132. The upper column that is spliced on top of the W14x132 is a W14x53. The floor to floor height is 14'-0". Can the effective length for buckling etc for the upper W14x53 be taken as 10' (14' floor to floor - 4'-0" splice length) since the W14x53 is actually only 10' long from the top of the splice to the top of the next floor. And along those lines I have 24" beams framing into the minor axis of the W14x53 and 16" beams framing into the strong axis of the W14x53. The slab is 6 1/2" thick. As I see it you could reduce the unbraced length for the column even further by the depth of the slab and the clip angle connection of the beams to the columns. Any thoughts would be appreciated as this is an existing condition for W14x53 that is getting extra load and to coverplate the column would be very difficult. Thanks for your thoughts

 

[JAE]

As long as the column below is designed for cantilever action (k=2) then the upper column can be considered with k=1.0 with the 10' length. You can use the center of the intersecting beam connection as the brace point.

 

[271828]

JAE, I personally wouldn't do it that way, but perhaps it's ok. The 4' piece of column probably has K>>2, for one thing. What's Ig for use in the alignment charts if one tried to calculate Kx or Ky?

I think the best model, assuming the column splice transfers moment, is a single column with K=1 (assuming this is just a gravity column) and different A and I from 0' to 4' and from 4' to 14'.

mkrei, do you have access to SAP or another program that'll do buckling analysis? It should take literally 5 min. to get Pcr from the eigenvalue buckling analysis, divide that by A to get Fe for use in Section E3 of the 13th Ed. Spec. Conservatively use A from the W14x53.

 

[mkrei]

I have Risa 3D. Will that perform the correct analysis

 

[271828]

Unfortunately, probably not. It didn't the last time I checked anyway.

You *might* be able to do the following. (Do some experimenting to see if RISA gives reasonable results for known problems.)

Create your column in RISA and subdivide it a bunch of times along its length. Give it a tiny initial lateral displacement by modifying the coordinates. Make it a half sine wave. Then apply a vertical load that you're confident is below Pcr and see what lateral displacement you get a midspan using a second order analysis (this is the question -- I don't know what kind of second order analysis RISA does and some types won't work for this). Increase the load and repeat. At just a hair under Pcr, you should have a small deflection, say 0.01". At just a little closer, it'll increase by a lot, say to 10". In some ways, this is actually the best way to use a program to find Pcr. It gives more control than an eigenvalue analysis which finds ALL buckling modes whether you want them or not. Some will be unreasonable. Remember that this gives the elastic load which you use to get Fe and adjust for inelastic buckling using the Spec. equations.

Good luck.

 

[JAE]

RISA's second order analysis does match the AISC test problems in the manual when using sub-divided axial members.

I was assuming a pinned splice. Yes, if it is a full moment splice then what 271828 states is correct.

I don't usually use RISA's automatic K generator as it is based on rudementary end conditions that may or may not match what is really happening.

 

[mkrei]

Thanks for the advice 271828. I ran the column in risa with .5' long members with the W14x132 up 4' and the W14x53 up the remaining 10' and developed a load that when we exceeded it slightly was on the order of 600 kips. I then reduced that appropriatly to an allowable of approximately 367 kips. Then just I ran the column as only 12.5 feet tall 4' of W14x132 and the remainde W14x53 to account for the 6" slab and the 24" beam framing into the side of the column. The critical buckling load in this case appeared to be about 825 kips which seemed reasonable with a shorter unbraced length. The allowable load based on this was about 494 kips. All these results seemed to be reasonable and somewhat agreed with the 10' unbraced length and 8.5' unbraced length doing a normal column design. If you have any further thoughts please let me know and thanks for your help. This solved a serious ooops for me in the field. Thanks

 

[mkrei]

The columns plice was not a moment connection but it is 3/4" flange plates with (6) A325 bolts above and below the splice on each flange so I think it can transfer the small amount of moment required to act as a unit. Thanks so much for your guys input. It is nice to know there are other practical engineers out there to help out with a problem.

Thanks

 

[271828]

mkrei, before you trust the computer output, put a simply supported column in there iwth the same Ix and Iy over the entire length. Then solve for Pcr. It should almost exactly match Pcr=pi^2*E*I/L^2. Otherwise, the procedure isn't working correctly.