Coverplating Existing Columns 2

[structuralecstasy]

I am coverplating existing columns (power plant application), and I have an issue with unbraced lengths. First, the background information: We are coverplating columns with plates welded across the flange tips, essentially forming a tube-shape. About 95% of the existing axial load in the column is dead load that cannot be unloaded. We also have new axial loads that are approximately 20% of the existing axial load.

The question of the hour is: Will adding coverplates to a column that is already loaded increase the ry value (weak axis radius of gyration), and thus the allowable axial stress? One side of the argument is that the column is already loaded in its deflected shape and coverplates would do nothing to help with the existing load. The other side is that the coverplates will still help restrain the column from buckling under the existing load, and the column's allowable axial stress can be increased. In either case, the coverplates will not be used to take existing load, only part of the additional load, so for the purposes of this post the coverplates are there only to increase the ry of the column.

Thanks in advance! -Nathan

 

[damstructural]

The existing column should not be "significantly" deflected by the existing dead load. If the column was properly designed, there should be significant capacity remaining in the column for unseen live load.

I agree that adding plates without unloading the column will not help to carry the existing dead load because the existing column strain is what it is at the time of welding the plates. However, you shouldn't need the extra plates for the existing dead load anyway.

In my opinion, building up the sectional properties by adding plates will help to carry additional load. As new load is applied, the incremental stress will distribute across the entire built-up section (but I would expect the existing section to always have a higher stress than the plates).

I would expect that the concern should be for buckling rather than pure axial stress. I don't know the details of this problem, but I would probably be comfortable ignoring the "prestress" effect due to exisiting dead load. My approach would probably be to design for the total load (exisisting and new loads) using the new built-up section properties. I'd be conservative with load assumptions and allow a comfortable margin of safety.

 

[MotorCity]

Structuralecstasy,

A word of caution...be careful when welding a loaded column. The extreme heat generated by the welding will create an area of local weakness and/or residual stress. This is a concern when the column is at or near its design capacity and most of the load is dead load. If most of the load was live load, there would not be much concern because the welding could be performed when live load was not present. Good luck.

 

[Taro]

This was discussed a few months ago:

thread507-92489

 

[structuralecstasy]

I wish we were just adding new loads, because then I could check the "core" column section for existing DL plus a proportional amount of the new DL and existing LL against the design yield stress, 0.6*Fy (ASD). Then I could check the reinforced column section for the the total load at the buckling allowable stress, based on the new section properties.

Unfortunately, I neglected to mention that part of the structural modifications include removing 2 consective beams that brace the weak axis of the existing column, so I am essentially tripling the unbraced length of the column's weak axis. We are adding the coverplates mainly to reinforce this weakened section of the column.

Our proposed process is to install the coverplates first, then remove the existing bracing beams, and finally patch the discontinuties in the coverplates at the existing beam locations. When the bracing beams are removed, will the new built-up column section be fully effective (i.e., can I use the new section properties for the existing load)?

Thanks, Taro, for the reference to the previous discussion. The information there was very helpful with some good resources I have yet to read but look promising. Unfortunately it appears there is no method that everyone has decided is the right method.

Sorry about the missing information from the first post, I was just trying not to bring up too many issues in the first post. "Ask one question at a time if you want them all answered."

 

[damstructural]

Whoa Structuralecstacy, tripling the effective length and adding 20% more load changes everything. My comments were based on retrofitting a well-designed existing column for 20% more load. You're going to lose axial capacity and possibly induce local stresses before you get to a final built-up column section. You've got some homework to do.

Taro,

I have to respectfully question your comment from the previous thread. You said,

"For an Euler column, the buckling load is:

Pcr = pi^2 * EI / L^2

...when the existing column reaches yield, it's stiffness, E, goes to zero (assuming elastic-perfectly plastic behavior). Therefore, it has zero buckling resistance."

Notice that the buckling equation doesn't include a yield stress term because buckling is a geometric phenomena, not a function of yield stress.

Also, you are neglecting the built-up section properties when you state that,..."the existing column stiffness goes to zero." The entire built-up section must buckle together (assuming continuous construction). The existing column can't buckle without taking the plates with it.

Thanks for the reference to that testing you describe by Yura. Correct me if I read it wrong, but it appears to be for columns that are preloaded to near-capacity before adding plates. If that's the case, I can see where adding plates to a column that is ready to buckle just isn't going to save it. The stress redistribution would be very high between the inflection points. I'd like to learn more about that research.

 

[damstructural]

Structuralecstacy, by "homework" I mean that this is obviously a theoretical problem that will require research and judgement. My two cents was based on what I think is sound theory, but I don't think it goes far enough to really help you.

Once you decide how to handle the preload condition for your situation, it should be a fun problem calculating capacities under a variety of construction stages as the bracing is removed and the plates are spliced together (assuming that approach works). It would be interesting to know what you decide to do and how this turns out.

 

[jheidt2543]

Thinking about structuralecstacy's description of the sequence of operations raises the following question. With the cover plates in place above and below, and removing the bracing beams, but before the "patching" plates are installed, seems to be a VERY critical point in the work. Isn't the location of the bracing beams at the point of inflection of the column? I would think some type of temporary bracing would be required, probably just below that point, to support the column while the "patching" plates are being installed.

I would be interested in more description of how structuralecstacy is planning to handle this.

 

[Taro]

damstructural,

Buckling absolutely IS a function of yield stress. Examine the Euler buckling equation more carefuly. The term "E" is Young's modulus of elasticity. It is defined as stress divided by strain. After the steel yields, it continues to experience increased strain at a constant stress. The slope of the stress-strain curve (which equals E) goes to zero. Now substitute a value of zero for E in the Euler equation. It is obvious that the yielded steel provides no resistance to buckling.

 

[damstructural]

Taro,

I respectfully disagree that buckling is a function of yield stress. E, Young's modulus, is not a function yield stress, its the ratio of stress/strain. Since they both have the same E value, 36ksi steel will theroetically buckle at the same point as 50ksi steel, all things being equal. The critical buckling stress is almost always well below yield.

The modulus of the material does not actually "go to zero," that's just one of several mathematical models to describe instability. You could also say that the moment of inertia, or better yet, EI (stiffness), goes to zero. You could also say that the effective length goes to infinity. In fact, changing the effective length was the approach that led to K-factors to account for the reduced capacity based on geometric conditions (fixity).

Respectfully,

damstructural

 

[Taro]

damstructural,

I am talking about inelastic buckling here. For inelastic buckling, the yield stress is very important. "E" may be a poor choice of words because it usually is used for elastic stiffness, but it is useful because you can see how it is used in the Euler formula. What we are really interested in is the material stiffness. After the steel yields, it's stiffness (differential stress divided by differential strain) goes to zero. Zero stiffness means zero buckling resistance. Go back and review the development of column strength curves including residual stress effects and you will see that yielding plays a huge factor in buckling resistance.

 

[damstructural]

Taro,

If we're talking about post-buckling failure mechanisms (inelastic buckling), I think we're beyond the scope of this thread and into ultimate strength, ductility, strain hardening, and the yield plateau, etc. Column coverplates are not going to help much at that point.

For the purposes of design, which is what I thought this was about, yield stress is based on percent strain, within the linear portion of the stress/strain (E) curve (elastic range). For purposes of buckling theory and column design, E is a constant and yield stress is not in the equation.

Respectfully,
damstructural

 

[Taro]

damstructural,

Inelastic buckling is a fundamental concept of standard column design. Anytime a column has a Kl/r value less than C-sub-c (for ASD design) or has a lambda-sub-c value less than 1.5 (for LRFD design), then inelastic effects come into play. That's why there are separate curve-fitting column strength equations depending on which side of the elastic buckling limit you are on. I suggest you research column buckling further independently because it is apparent that I cannot get through to you on this.

 

[damstructural]

Taro,

You don't need to "get through to me". I don't think either one of us is educating the other here, we're just restating different things and straying farther and farther from the coverplate topic.

I'm fully aware of the inelastic/elastic ranges of the lambda-c curve. I even programmed the curves and formula logic into my HP48 so I can calculate the column tables faster than looking it up, and in the field without a manual.

I think we're both correct, but we're writing about different things. I hoped you could agree that for pure theoretical Euler buckling, yield stress doesn't matter. That was what I was writing about, because that's how I read your initial comment.

I agree with you that yield stress does matter for short columns (no buckling)and for intermediate-length columns (buckling-yielding). The lambda-c formula is an empirical relationship for this "hybrid" range, not a theroetical one. When you get to true Euler buckling at larger KL/r values, the capacities merge regardless of the yield stress.

As for the topic at hand, we don't have enough information about the column in question. It may be so short and fat that he doesn't even need cover plates.

Respectfully,
damstructural

 

[3dboy]

Guys,
you have gotten way off topic, the question is has two parts, the first being whether or not adding plates across flange tips will increase the ry and, yes, it will. You are increasing both I and A and Iy by a greater amount than A and there for ry increases. What makes the question tricky is the second part which is whether or not the stresses due to the existing deadload are going to make this "not work". I think the way to handle this is to check the allowable buckling stress with the reinforcing plates added, subtract out the "pre stress" due to deadload and if the remaining allowable stress is greater than that induced by live loads, call it good. The construction sequence is also critical. You absolutely have to make sure the thing is braced after the intermediate beams come out, and before the holes are patched up with the plates. I know this isn't the most technical way, but it is a safe conservative way and besides when you get in deep to what the "power's that be" say about residual stresses and such, most of it is a guess anyway.

 

[damstructural]

Your suggestion is similar to what I recommended before getting sidetracked.

I hope the theoretical banter helped in some strange way. I hate to get caught up in semantics, but sometimes good comes out it.

Interesting project.

Cheers