ETABS 18 underestimates/overestimates steel columns/beams compressive strength (AISC 360-10)?

[sadato9224]

Hi, everyone.

Title says it all. I checked the axial compressive strength of a compact W-shape column/beam in ETABS 18.1.1 and compared it to manual calculations based on AISC 360-10 Section E3 and E4. It seems that ETABS calculates the compressive strength differently than AISC 360-10 even though the design manual clearly has the same formulae as AISC 360-10.

I've checked the flexural, shear, and tensile capacities reported in ETABS and they are equal to those I've calculated manually. Only the compressive strength results given in ETABS are different.

I've followed the calculations exactly and changed the input parameters accordingly (K-factor, unbraced length, etc.) and the results are always different no matter what dimensions I use. Is anyone facing this issue as well or am I missing something?

 

[JoshPlumSE]

Post your calculations. And, the result printout for the column. Chances are there is a difference in assumptions somewhere.

 

[sadato9224]

I have a more complete spreadsheet calculation attached here if anyone is interested in the other values I've calculated.

 

[sadato9224]

Thanks for the suggestion, JoshPlumSe. Probably should've started with that. I kinda agree that there might be different assumptions used, but I've not been able to pinpoint the source of the differenc.

So, I only checked for Flexural Buckling (Section E3) since the torsional unbraced length is equal to the lateral unbraced length and therefore flexural-torsional buckling (Section E4) is not the controlling limit state.

Based on my manual calculations, the axial compressive reduced strength is 10869 kN while ETABS reports the compressive strength to be 9937 kN, quite a difference. This report is for a beam, so all the K-factors are taken as one.

I've attached the manual calculations and ETABS report below.

Edit: Had to remove the license information from the report for reasons.

 

[Agent666]

Is etabs using 20000mm for L, and your hand calc is noting 18000mm for the major axis buckling length?

So are you sure your calculations are considering the same effective lengths for compression buckling?

If the L factor is the ratio of full length to buckling length, then 0.936 x 20000 does not equal 18000?

 

[sadato9224]

@Agent666

Yes, the beam span is 20000mm in both ETABS and my hand calculations, and since there were columns at each end, the unbraced length in the major axis is taken as 18800 mm which is equal to 0.939 x 20000 in ETABS. Maybe I should've used a more accurate value of 18780mm for the major axis unbraced length, but the difference between 18800 and 18780 is quite small to have any significance in my opinion.

So yeah, I've used the same unbraced lengths in ETABS and my manual calculations for both the major and minor axes.

 

[Agent666]

That's what I get for reading it on my phone, could have sworn you had 18000mm for the length, but it was 18800mm!

If you do the hand check at 20000mm do you get closer to the Etabs value, maybe it is neglecting the reduction in length for some reason?

 

[sadato9224]

@Agent66

Yup, I tried that as well, but the reduced compressive strength in my hand calculations is still about 10759 kN, a little bit closer but still far from the value of 9937 kN reported by ETABS.

I'm beginning to wonder whether there are other limit states I should consider. I'm using an I-shaped welded beam which is doubly symmetrical, so I believe only section E3 and E4 applies.

 

[sadato9224]

So, I tried tinkering with the K-factor values from 1 to 10, and the higher the K-factors, the closer the values are from my hand calculations and ETABS, but still not exactly equal. I'm convinced that ETABS uses a different formula, either accidentally or deliberately.

 

[Agent666]

Have you created a test file that is basically one single compression member and compared? Take out all other variables?

One thing Etabs might be doing is a buckling check, which is coming up with a different effective length that allows for the stiffness of interconnecting members making the K factor effectively higher.

Similarly it may be using the direct analysis method (notional loads), your output suggests this with the 0.8 factor noted on stiffness? Therefore could be working back from the combined actions check to the maximum axial load that can be carried by the system.