Fundamental question about column design 3

[jay156]

Hey, this may seem like a dumb question.

I'm designing new mezzanines to go in this high bay at a factory and am checking the columns on my RAM Advanse structural software. They're failing. I told the client this and he said check the calcs from the engineer who worked here before me, who added a mezzanine to a different area, but with the same size and height columns.

So I looked at the old calcs and he designed his columns one level at a time. By that I mean he broke a 61' tall column into columns from 0' to 11', from 11' to 25', etc and considered the tops and bottom fixed. In my model, I have the entire 61' column modeled with mezzanines framing into it.

The question is, should this be making the kind of difference it is? Or is his modeling of the column not conservative? When I run an 11' column in the program with the mezzanine's reaction applied to it, then it says it passes. Could it be a problem with my program or how I'm using it?

What is the right way to design columns: the entire height with all loads applied at their respective levels, or broken up into multiple short columns the height of each level?

 

[nongamer12]

it depends on where the columns are braced against buckling. if you have multiple levels of mezzanine floor framing, you're probably able to say the columns are braced at those levels.

it's not just a matter of how it's put into the analysis software, it's a matter of conecptual understanding of the design and getting the analysis software to produce results in line with your concept. you can model the columns as being 61' tall and continuous, as long as the unbraced length is adjusted to framing levels (bracing levels in general). the other option is to model the columns from level to level, and make sure all proper fixity is modeled. make sense?

 

[csd72]

Not a dumb question.

Firstly, if your predecessor just designed it as a single length with total fixity top and bottom then I would say that was oversimplifying it and could be unconservative.

The classic method to isolate a footing is to take every member a bay past the member in every direction and then choosing a pinned or fixed condition at the ends depending on the situation. This is still an approximation.

It sounds to me like the way you are doing it is a much more correct way and should be giving accurate results. I would just suggest that you check that each floor os restrained laterally if it is a braced structure as rollers would give you a 61' effective length.

One thing that you may not have thought of is the fact that concrete strength increases with age and the actual strength may be considerably above the 28 day strength called up for in the design. Some in situ testing may be possible to get an estimate of the actual strength.

Also remember to reduce your live loads.

 

[a2mfk]

And with wide flange shapes, use caution on the weak and strong axis bracing assumptions in your design, and ensure it is shown on the drawings with the orientation you want. Its now always obvious what will control either. 20ft unbraced strong axis may be ok while 10 ft unbraced weak may control, depending on the shape.

I usually model columns as pin-pin unless I truly am providing a fixed base. As far as breaking up the column design into stories/levels, as long as you chase down the loads correctly for each story, and input the proper bracing at each height, this is how its usually done.

Take a step back from your model for a second and do some quick hand calcs (you could pull the reaction off your model) and pull out the AISC column tables for a sanity check. You should be within a shape or so if your model matches your hand calcs...

 

[jay156]

These columns are steel, W12x65's.

I had not entered the correct unbraced length, so that's part of the reason they were failing.

Now I'm confused as to what to put in for the K value. The entire column is pinned at the bottom, since it's just bolted to the foundation, and at the top, only the roof steel frames into it, but all along its height, there are floors that frame in. I don't think it's correct to just say that the whole thing is simply pinned-pinned, because there are all those floors.

Maybe it does make better sense to break the column up into sections, even in my model, so that I can specify the right K value for each one. Does that sound right?

 

[steellion]

In terms of computer modeling, if the column is modeled in 11' segments, it will generally treat each segment as a "member" (depends on the software!). If it is a continuous column with nodes, it may or may not assume bracing at nodes, you need to figure out what assumptions it is making. I like making one member and adding nodes and then checking the unbraced length. Either way, you need to make the computer work for you. And always do a sanity check with a hand calc. If you're not sure on the assumptions, ask a mentor.

 

[a2mfk]

I agree with steellion, these are discussions to have with your mentor, and more will come up.. This is normal.

If you are using one size column for the entire height:
1. Get your total axial load.
2. Go up to your first brace point, is it braced both strong and weak directions? This establishes your unbraced length for the axis in the direction it is braced. Often with mezzanines this may only be in one direction.
3. The lower level is usually your worse case (in most buildings), and if doing it by hand is where your column size would start. Then you go up the column to check the other unbraced lengths, as you have stated, you have a variety of unbraced lengths, and each must be checked. Good news is that your axial load is decreasing at every level, so your column may work all the way up.
4. K- If sidesway is inhibited, than it is usually K=1.0 or at least that's most conservative. Some people may model the first story column as fixed base and pin at the first story/mezzanine. If you have sidesway, than it gets a lot more complicated (you have not indicated if this is part of a moment frame or if lateral stability is provided by other means)...

Keep in mind Euler's buckling formula , EI*pi^2/ (KL)^2 , so the effect of increasing K or L is non-linear in the strength, obviously, inverse square. This is why its so crucial you are making the correct KL assumptions and inputting it correctly.

Sadly, a good illustration is the WTC failures. You can see in the video where once the exterior HSS columns lost floor lateral support (failed floor joist connection I believe), they outwardly buckled in a zipper-type sequence, and did so explosively...

 

[JAE]

I would use K = 1.0 with each mezzanine providing bracing in each direction (assuming the mezzanine connects to the column in each direction).

The K = 1.0 usage, even with a continuous, multi-story column, makes sense because the column could buckle in a continuous "S" shape at each floor - i.e. lowest story the column sweeps fully left, next story up it sweeps fully right, then left at the next story, etc.

Each of these buckled "shapes" is in a full arc from node to node - thus, you have a K = 1.0 condition at each floor.

 

[a2mfk]

JAE- sinusoidally ?

I agree with you in typical buildings.. The complication with jay's structure is it sounds like he has braces in different axes and at different heights.

 

[TTUengr51]

To add to Steellion, in Ram Advanse, if you model a continuous column (physical member) with internal nodes, by default it uses the full length of the physical member as the unbraced length. If you want to use the heights between levels/nodes, you have to enter this manually under the steel design spreadsheet. This spreadsheets allows entry of Lxx, Lyy, Kxx, Kyy, and top/bottom flange bracing for lateral torsional buckling.

Nick Deal, PE, SE
Michael Brady Inc.

http://www.michaelbradyinc.com

 

[JAE]

a2mfk - yes maybe you'd have something different with different orthogonal braces - but even then, if you just look at the axis under consideration (and ignore the other axis) you still might get that sine curve (sinusoidally sounds like a sneeze on a table cloth) reaction.

 

[trainguy]

When I was in 1st year Civil Engineering, my Structural Design prof said something like "column design is something that has engineers wake up screaming in the night".

I always took this quite seriously, given the catastrophic nature of column failures. I strongly agree with the above posts relating to mentors and hand calcs.

Using an internet message board (even this one) to get advice on column design is just plain wrong, and really worries me.

tg

 

[hetgen]

Jay,

Check the post by JAE [the form Guru] on this thread

http://www.eng-tips.com/viewthread.cfm?qid=201140

 

[JAE]

hetgen - thanks for the "guru" status but I would suggest that there are waaaay more high-end gurus on Eng-Tips than me. That fact is what makes this site so great - that there are so many who know so much.

 

[paddingtongreen]

nongamer12, has it right. The question at each level is, "Is the floor a node or just a load?" If it is pinned to the column and not braced against joint translation it is a load, if it is fixed to the column or braced it is a node. "k" is going to depend much on whether there is bracing. If that bottom segment is not braced against translation but fixed against rotation, k would default to 2.2 times that 11' piece. Since it is only partially restrained against rotation, it probably should be higher.

There used to be some nomographs on this, copies used to be in the AISC manual.

Michael.
Timing has a lot to do with the outcome of a rain dance.