Design Forces from Response Spectrum Analysis

[fracture_point]

When the modes are combined with various methods (CQC, SRSS etc.) the sign of the displacements and forces are lost within most software, so often when analysing forces due to a response spectrum load case the laws of static are often not obeyed.

I have a shear wall being supported on a transfer structure of two columns. The overturning moment from the lateral displacement should cause compression in one column and tension in another, but the forces under the RSA show both columns going into tension (due to the aforementioned sign loss). I am using ETABS, so there is no apparent option to use the 'sign of the most dominant mode'.

How do you typically determine appropriate design forces from a RSA?

 

[Agent666]

Run an equivalent static analysis beside it, and use some judgement to determine the sign of a given action for a given loading direction. I often find you need to manually review and interpret the static cases to make sure you are interpreting the sign correctly for the response spectrum cases.

 

[JoshPlumSE]

So, the challenge here isn't that either of the column forces are incorrect, it's just that their RELATIVE signs are incorrect (or not useful because they're so far removed from obeying statics).

In cases like this there are two things to worry about:
1) Run the response spectra load combinations twice. Once with a positve and negative load factor. That way, you're checking the columns for both tension and compression. Right?
2) Identify cases where the relative signs matter.... maybe if you've got a combined footing that connects the two columns. Then use engineering judgement to determine relative signs. Maybe using Agent 66's idea of using an ELF method to determine relative signs of the two column components when designing that combined footing.

 

[fracture_point]

Thanks for the responses. I actually just realized when you place the RS load case into a load combination in ETABS, it automatically gives you an envelope ranging from max to min, which is essentially the same as creating the negative load case that JoshPlum suggest.

Cheers