Column Splice Criteria

[RichRook]

I have a simple pin/pin column that needs to be installed in two pieces. What would be the criteria for the splice. Obviously I don't want another pin cnx at the splice, but I'd rather not develop the full moment capacity either.

I realize that I'll need to put some nominal load laterally just for things leaning on the column. Also, I'll Fix the base as much as possible since that will make installing the upper sections easier. But, I'm uncertain if there's a rational method for the spice without other considerations.

Would using 2% of the axial force apply? (in my particular case I'd use more)

It's more of a post than a column . . about 14' high, maybe 6x6 tubes with less than 10 kips. But it is repetitive so I'd like to be economical.

Thanks,
Rich

 

[KootK]

I'll assume that this is a splice near mid-height. Ick. A similar condition occurs when truss compression chords are spliced. And that's why they are generally spliced near panel points.

I'm not sure what criteria ought to apply to the joint but joint flexural stiffness will be every bit as important as strength. In a non-welded scenario, I'd lean towards thick end plates with a copious amount of pretensioned bolts.

I bet some criteria for this exists out in the industrial world somewhere...

The greatest trick that bond stress ever pulled was convincing the world it didn't exist.

 

[mike20793]

I've used Lindapter bolts (Hollo-Bolt) and plates to splice an HSS column in an existing wood building where welding wasn't possible. The contractor didn't seem to have a problem installing it correctly. The column was loaded pretty lightly, around 15 kips if I remember correctly.

 

[KootK]

Most multi-tiered steel columns have splices around 4' above every other floor. While I've yet to locate it, there's got to be something in AISC other than just the recommended details.

The greatest trick that bond stress ever pulled was convincing the world it didn't exist.

 

[RichRook]

It's not so much the detail itself I'm looking for - I can design that. Rather, what is the criteria for stability? How much of the member's flexural capacity do I need to develop for the column to be considered stable.

I have some minimum loads I want to apply laterally . . maybe 400#. So I can easily design the splice for a moment generated by the 400# lateral load. But is that enough for stability? It would be more than 2% of the axial load, but I'm not sure if that's the right criteria.

I've spliced multi story columns often enough, but I think this is the only time I've spliced a single story column.

Thanks,
Rich

 

[KootK]

Stability was precisely my point above Rich. The most important part of the connection design will be the connection's stiffness because that will significantly affect the column's buckling load. I've got the Shanley theory buckling model in mind as I type.

I get where you're coming from with the 2% business but I don't think that it's applicable here. The 2% stuff is about nodal bracing forcing a second mode buckling shape on a continuous column. The problem here is about reconstituting flexural stiffness across the joint. Whatever the solution is, connection stiffness has to factor in somehow.

The greatest trick that bond stress ever pulled was convincing the world it didn't exist.

 

[TLHS]

There's got to be a better way, but you could likely do it with the direct analysis method. You'd have to come up with the stiffness of your connection, though, and use that to put a spring in. That would be the hard part...

Then check what moment shows up at the connection point and make sure your system's stable.

 

[wannabeSE]

If I remember correctly, one model for calculating the strength and stiffness for stability bracing of columns is based on an analysis assuming the bracing acts as a spring support at a hinge. It may be possible to adapt this to your connection design.

 

[KootK]

If one could know the stiffness of the spring representing the connection, it would be a fairly simple matter to come up with a revised K such that standard column buckling equations could be used. Unfortunately, knowing the connection stiffness with any accuracy is hopeless. Consequently, the only rational strategy is to make the connection very stiff.

This has me seriously rethinking tiered column buildings where the splices are made 4' above every other floor. I've always assumed that the splices provided flexural continuity across the splice joint. Is that not the case? Should column splices be treated as flexural pins? I've worked on buildings with 12' floor heights where a typical 4' splice is pretty darn near mid span.

The greatest trick that bond stress ever pulled was convincing the world it didn't exist.

 

[structSU10]

The best I could find, going into my copy of Structural Stability Theory and Implementation, for an initially crooked column, the internal moment is M = (1/(1-P/Pe))*P*do*sin(pi*x/L) where Pe is the euler buckling load. You can assume the standard h/500 for do, the initial deflection of the column, and go from there? I would imagine that is the splice is midspan, in order to achieve the normal pin-pin behavior, you would need to ensure the splice is at least the same stiffness as the column. You may be able to have less stiffness away from the midspan, but I would imagine If the stiffness gets too low, it would try to buckle at that point rather than midspan.

 

[structSU10]

A paper I just found on the subject, that at least gives some criteria

 

[paddingtongreen]

I'm a little confused.

Obviously I don't want another pin cnx at the splice, but I'd rather not develop the full moment capacity either

What are you doing to transfer vertical load? If these are heavy columns they should be machined, and then an AISC minimum tension applied via welds or splice plates. If they are lighter weight columns, then the load can be transferred via weld or splice plates. These connections, to the best of my memory, have always been considered sufficient for stability.

Michael.
"Science adjusts its views based on what's observed. Faith is the denial of observation so that belief can be preserved." ~ Tim Minchin

 

[paddingtongreen]

I intended to add that the flexibility of the joint is not very important in stability calculations. The flexibility of the joint is over such a short part of the column than the decreased radius of curvature has little effect on the stability of the column. Again, I have no idea how you intend to make the splice but you could model the column, with the design load and the short flexible section and see what happens. If you weld the joint, there is no increase in flexibility and I believe there is a required minimum design tension that should more than handle the lateral moment. If you are carrying the load with splice plates, you should have more than enough capacity.

Michael.
"Science adjusts its views based on what's observed. Faith is the denial of observation so that belief can be preserved." ~ Tim Minchin

 

[JStephen]

Can you just design the whole column as a lighter column, then design for that moment capacity?
Suppose it's a 6"x6"x3/8" tube, you check your numbers and conclude a 6"x6"x0.035" tube would actually work, so you design that splice for the moment capacity of the 6"x6"x0.35" tube. You're just neglecting the excess material for that part of the check.
This assumes that load is independent of stiffness (should be, for pinned-pinned). And assumes that actual load is a lot less than capacity, or you wouldn't need to bother.

 

[KootK]

A paper I just found on the subject, that at least gives some criteria

Nice find StructSU10. It even covers exactly the case that I mentioned above (tiered columns). Unfortunately, the criteria proposed still requires one to know the connection stiffness. Based on the testing, however, the right bolted detail seems apparent: thick end plates and tight bolts.

It's also interesting to consider that the flexural stiffness of the splice essentially matches the un-spliced section until the prestress effect of the axial load is overcome. If only we could count on that from a reliability perspective. I'm sure that it helps many column splices to work in real life.

I intended to add that the flexibility of the joint is not very important in stability calculations. The flexibility of the joint is over such a short part of the column than the decreased radius of curvature has little effect on the stability of the column.

I disagree with this statement strongly. Deflection is not impacted substantially by spot reductions in flexural stiffness. However, stability is a different animal. Euler column buckling, at its heart, is a mathematical statement that, at all sections along the column length, restoring moments develop faster than P-Little Delta moments develop. As an extreme case, consider a column splice with no flexural stiffness at all.

And that's just elastic column buckling theory. If you consider inelastic buckling theory as proposed by Shanley and incorporated into modern code provisions, the importance of splice connection stiffness becomes even more apparent.

The greatest trick that bond stress ever pulled was convincing the world it didn't exist.

 

[paddingtongreen]

I was thinking of elastic buckling and I'm sorry you disagree. I see that "deflection", as I used it is misleading. I have checked many damaged columns using Newmark's iterative method. This can be used to find the highest load just before the column buckles and hence the safety factor against buckling under the design load. It can, of course, be done more easily reiterating with a structural program.

The diagram you show is so overstated as to be nonsense, we are discussing reduced stiffness, not zero stiffness.

Michael.
"Science adjusts its views based on what's observed. Faith is the denial of observation so that belief can be preserved." ~ Tim Minchin

 

[KootK]

There's no need to be sorry Paddington. Disagreement is productive so long as it's prosecuted respectfully and in the spirit of furthering understanding.

If you read through the paper that StructSU10 posted, you'll see that they also claim that localized stiffness reduction due to column splices has a significant impact on stability.

I contend that the sketch that I posted is quite sensible and germane to our discussion. Columns buckle when their flexural stiffnesses, including little P-delta effects, approach zero. If we're not going to discuss the attainment of the zero stiffness condition, then we're not discussing buckling.

The greatest trick that bond stress ever pulled was convincing the world it didn't exist.

 

[KootK]

Quote (Paddington)
I intended to add that the flexibility of the joint is not very important in stability calculations. The flexibility of the joint is over such a short part of the column than the decreased radius of curvature has little effect on the stability of the column.

I've come around, partially, to your way of thinking on this one Paddington. I'll try again. more carefully:

flexibility of the joint is not very important in stability calculations

The paper above says otherwise. We're kind of stuck with that.

However, stability is a different animal.

Nope, it's not. I was wrong about this. P-delta moments grow as the column deflects laterally and the entire length of the column contributes to the flexibility that leads to that lateral deflection. So then, why does having such short lived increase in flexibility at the splice seem to matter so much? More on that in a moment.

The flexibility of the joint is over such a short part of the column than the decreased radius of curvature has little effect on the stability of the column.

I think that this would be a true statement were the splice an element that could be said to have a radius of curvature. A welded splice, as you suggested above, would be an example of this.

However, when the splice is modeled as a spring element as was done in the paper above, the radius of curvature is effectively zero. Any rotation within the spring results in rigid body rotation of the segments of the column above and below the spring. And that leads lateral deflection to accrue fast. I think that this is why splice stiffness has such a significant impact on buckling load. Of course, whether a column splice ought to be modeled this way is another question altogether.

I like to debate structural engineering theory -- a lot. If I challenge you on something, know that I'm doing so because I respect your opinion enough that I want to either change it or adopt it.

The greatest trick that bond stress ever pulled was convincing the world it didn't exist.

 

[paddingtongreen]

I woke up in the middle of the night before last, wide awake and wondering why I didn't answer the OP directly, I had the answer. Okay, I had my answer.

The model is of a strut with an axial load and which has been given a lateral push in the middle. The critical load is one that can just support and remain in equilibrium. On the application of a small increase will cause the strut to fail. Big deal, we already can get that load easily. But we need to know the lateral deflection at that point to know the design moment, and shear if the splice is anywhere but in the center. OP suggest he will have some fixity at the base so I am not sure where the "center" is although full fixity would be covered in the method.

By applying the conjugate beam method we can calculate both P_Euler and the offset and thus, the design moment.

I am a generation or more removed from many of you and I am often puzzled by your nomenclature until I figure out that you are simply using a different name for something I know, E.G. when I started out in the UK, we simply had "long columns" and "short columns" and the implications were well understood. Other times, I stay puzzled.

Michael.
"Science adjusts its views based on what's observed. Faith is the denial of observation so that belief can be preserved." ~ Tim Minchin