Composite kinked column analysis challenge

[trainguy]

All,

We are looking at repairing a kinked aluminum tubular compression member (1 inch kink over 60 inches, so slope is 1 in 30) with a 15 foot long straight steel plate) on 1 face. This is an 80 foot long horizontal member on a railcar. For argument's sake, lets assume it's an approx 5" (horiz) x 10" (vertical) tube. We have resolved corrosion and temperature stress issues, and are looking to finalize the axial load capacity of this new composite member. Buckling is fully restrained in the strong direction. In the weak axis, lateral supports are at approx 60 inches. Our aim is to transfer a portion of the axial load carried by the aluminum into the steel doubler by deliberate end connections with tight, well installed rivets at the end regions, where there are no gaps. At the central, kinked region, we are installing rivets through shims that make up the gap, but using oversize holes which serve only to stabilize the steel plate against local buckling, in all directions. Yes, we're talking many, many rivets.

We have determined, by displacement compatibility, how much axial load is taken up by the aluminum portion, and how much by the steel.

Any idea on the easiest way to determine the adequacy of this member? We have NL FEA capacity if required. Design standard to we we are working states "No permanent deformation", so neither yielding, nor buckling are permitted. Should we "simply" do a buckling analysis with FEA? Should initial geometry include the kink? Should we include geom nonlinearity?

Any other, simpler ideas using hand calcs and code spec. beam-column equations?

tg

 

[canwesteng]

If the kink doesn't pop back when unloaded it seems like inelastic buckling has occurred, I can't visualize how your repair addresses this. At that point I would expect all load needs to be taken by your repair

 

[SAIL3]

from your description it does not look like you have a composite member...why the oversize holes?...sounds like you are using an "already kinked" member to stabilize a flat pl reinforcement...there should be some shear capacity in the connection between the flat pl and the member in order for it to act as a composite member in buckling...how much?......maybe assume a buckled shape..calculate the resulting moment and from that the resulting horiz shear between the flat pl and member..just throwing out ideas here...requires more thought...to be really confident I would try and use the full capacity of the pl in calculating this shear

 

[rb1957]

i think it's a "mistake" to attach the steel plate on 5' spans, 'cause between the attachments the deformed rail and the steel plate are working independently. if they were continuously attached you could say they're working as one, determine the centroid position along the length and analyze the deflected column.

instead of plate, i'd've used a channel.

instead of living with the deflected rail, i'd've repaired it ... cut out the bent section, fish in a good section. how are you accounting for the stresses in the section resulting from the bend ? (plastic deformation)

another day in paradise, or is paradise one day closer ?

 

[trainguy]

All very interesting questions.

Sail3 - agreed it's not really composite and it's difficult to get good composite action because we are essentially riveting thru 2 faying surfaces - the shims are not welded to either material. Your interpretation is correct, I am trying to use the kinked member to stabilize the flat plate. The reason I felt this may work is that the main member has stiff lateral supports at 5 feet centers. The apex of the kink is approx. 14 inches from one of these, which has been rebuilt slightly shorter to accommodate this new shape. The kink, not being at mid-span, is therefore mostly stabilized by a nearby lateral member.

rb - I can't use a channel because this will foul the maximum vehicle envelope (the railcar's clearance diagram). I can't cut away and repair this easily, because it's an old aluminum that has very poor weldability and the extrusion's shape is quite hard to match.

Latest update - we have noticed, by a 2D beam nonlinear FEA, that it may make more sense to merely add a lateral member (or truss) at the peak kink location than to rely on a steel reinforcement, and all the head-aches that result.

tg

 

[dhengr]

Trainguy:
I take it these are side sills on 89' flats, actually roll-formed shapes, maybe extruded or is it part of the side construction on a passenger car? What kind of railcar is it, and what member? Is the whole member deformed, or just one flange, the top flange, or what? Was the member impacted by something to cause the damage, involved in a derailed, or was this an over-stress buckling situation from fairly normal operation and loading? Why not alum. cover pls., why stl? I’d want to know a great deal more about the actual details and conditions before I said much more.

Mixing steel and aluminum would worry me because of the potential for galvanic corrosion btwn. the two different materials. We always put a bitumen/asphaltic material (painted on, with some thickness) btwn. the two different materials to try to control the electrolytic interactions (corrosion), but these were usually not a slip critical type joint application. And, I always wondered how that really worked out in the bolt holes and btwn. the bolts and pls. The joints were usually predominantly in compression or bearing with fairly low shearing loads at the faying surfaces. At the ends, you are wanting what amounts to a slip critical connection to transfer the compression load, but over time any galvanic corrosion could change that. I would weld (no... rivet, so as not to lose temper) shear blocks (alum. blocks to the alum. sill/tube) and then cut the stl. bar long, to fit btwn. the shear blocks, so you had to gag it down at the shimmed center rivets, press it in, to fit tightly lengthwise. This would sorta prestress the alum. side sill, putting a little tension in the alum., and putting the stl. bar into initial compression. Alum. or stl. rivets? They all (too many of them) have a negative net area affect on the primary member. Shear block rivets take the compression and rivets in the ends of the cover pl. take tension components. Rivets in the center brace the cover pl. and tie the two elements together, but not too many as this is also the max. stress location. Contact me at rwhaiatcomcastdotnet if you want to talk some more on this problem.

Before we had regular access to computers and appropriate software I did some of these beam/column problems, by hand, with Newmark’s Numerical Integration methods and had pretty darn good luck. We actually tested a couple of different car designs to 1.25 million pounds and my deflection predictions were right on the money, and there were no detail problems, no buckling problems or permanent set issues. But, there were some pretty significant deflections over the length of the cars; enough to cause most of the people witnessing the tests to back away after we got to about 750kips compression. As I look back, I may have been a damn fool or else cock sure of my design abilities, as I stood so close to the side of the car to read the deflection gages. The car body returned to its starting point when unloaded.

If you can get at it from both sides to drive rivets, can you do anything to straighten the member with force and some heat, as long as you stay below the tempering temps?

 

[rb1957]

supporting the kink sounds reasonable to me ...

another day in paradise, or is paradise one day closer ?

 

[rb1957]

nd you might need to consider the other ends of the kink too (where there'll also be a lateral kick load)

another day in paradise, or is paradise one day closer ?

 

[trainguy]

Latest updates:

Using NL FEA, starting out with the measured imperfections (read: kink), we have determined that adding 2 lateral braces almost entirely removes any bending stresses. I'm thrilled about not adding a steel doubler. Now, to finalize, do you guys have any suggestions on how to determine a factor of safety against buckling, using the results of geometric nonlinear runs done to date? Specifically, I'm trying to determine kl/r, etc. for use in classical beam-column equations. Should I just take the distances between my points of inflection?

As you may infer, I'm trying to avoid doing an FEA buckling run.

tg

 

[trainguy]

Anyone...?

 

[KootK]

I feel that effective buckling lengths are not interchangeable with distances between inflection points. Apple and oranges. Well, there's some correlation I suppose. Apples and pears.

I think that the member strength and stability checks can proceed as usual with the following exceptions:

1) No need to amplify design actions for big P-delta effects as you've already captured that in your NLA.

2) Probably no need to amplify design actions for little P-delta effects either. This assumes that you've broken up all of your members into at least four segments with a node being present around mid-span of each member.

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

 

[trainguy]

That makes sense KootK, I was also against using the amplifications.

I'll just use the distances between my support points with K=1.0. I have also managed to run FE buckling runs, both linear (Eigenvalue) and non-linear, which I was already partially running.

Youtube is amazing, esp. the Siemens videos on Femap functionality.

tg

 

[KootK]

Sounds good TG. Just ensure that your NL buckling voodoo captures material / cross section non-linearity in addition to geometric non-linearity.

My tweenage kids are so addicted to YouTube that we're contemplating intravenous feeding. I'm glad to hear that something useful's come of it. How's your twerking progressing?

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

 

[rb1957]

is "twerking" like "torking" ?

i guess what you're trying to do is to redirect the load along the kink, so you have kick loads at three places, the kink and the two other ends of the damage. is this rail a "stand alone" structure ? or is it attached to something that can help it react transverse (kick) loads ?

another day in paradise, or is paradise one day closer ?