Loads on Adjacent Double Fillet Welds 3

[Ussuri]

This is one that occurred in the office and it got me thinking. It is a question about twin fillet welds placed close together. I have looked around online but couldn’t find anything specific.

Please see the attached sketch.

The background here is seafastening equipment to a vessel deck however, I have simplified the sketch so as to remove surplus detail. The sketch is intended to present the question so the dimensions are approximate.

Say you have a stiffened flat plate which is continuous and you apply a vertical load (P) to the plate at some point on one of the spans. Globally speaking you could assess the reactions at each of the stiffener points through a continuous beam analysis. This gives you a single reaction load (Fr) at the support point. The weld between the steel plate and the stiffener can then checked to ensure that it can transmit the load.

If the load is applied directly above the stiffener then all the load is passed directly through.

Now consider that the applied load is moved a short distance toward the centre of the span (say 40mm in the sketch). The reaction at the nearest stiffener is now marginally less than the applied load, P. There is also the moment continuous over the support.

Now looking at it more closely the load path is actually through two adjacent welds close together (Frw1 and Frw2 on the attached sketch). If you did the same continuous beam analysis replacing the single reaction point (Fr) as two points close together (Frw1 and Frw2) then you get a very high load on Frw1. Which would be the shorter load path. Now this area is stiffer so I’m not convinced that approach is valid.

Now the question is how would you assess the load distribution between the two welds, Frw1 and Frw2?

1) Check the weld under the singular reaction load distributed 50/50 over the welds on either side of the stiffener. Moment continuous over the support.
2) Assume all the load goes into the nearest weld? Ignore the other weld and effect of eccentricities.
3) Consider the two weld points as individual ‘supports’ and work out load on each accordingly?
4) Assess the rotation at the support point and work out how much bending (out of plane) is developed on the stiffener associated with this. Ensure the weld can carry that moment in addition to the Fr reaction?

 

[SAIL3]

I would assume all the P load goes into the Frw1 weld plus load from eccentric moment on welds of Pxdistance to center of welds......Section modulus of 2 parallel weld is Ld..d is distance between centerline of welds and L is the length of the welds...add this load from moment on Frw1 weld to the already P load...then make Frw2 weld the same size as Frw1 weld...

 

[KootK]

Gotta love structural engineering. Look at anything hard enough and it becomes impossible to resolve.

#1) Sadly, this is probably what I would do in real life.
#2) This will be unconservative due to the prying effect that you've rightly identified.
#3) This will be too conservative. Crazy prying action.
#4) I believe that this solution has the most technical merit.

Fillet welds generally do not enjoy resisting flexure: Link To bad, so sad for them.

To complicate matters further, I've known engineers who wouldn't even like this detail if the plates were right behind one another. The concern seems to be lamellar tearing across the horizontal plate. I create similar conditions all the time in bracing connections so I certainly hope that it's not a problem.

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 to either change it or adopt it.

 

[Ussuri]

Sail3, that would be an option 2, with moment/eccentricity allowed for assuming a cantilever.

KootK, The through thickness issue is one raised occasionally, but unfortunately it is not something that can typically be avoided due to the nature which vessels are constructed. Vessel decks are continuous steel plate, with all manner of things welded to them. I understand that because most of the time the deck plates are thin then lamellar tearing is not so much of concern with thin plates.

 

[dhengr]

Ussuri:
That’s a great sketch, except for one thing, the proportions in the blown-up view seem a little out of wack. I know you said they maybe aren’t the exact dimensions. I’m not giving you a hard time for the sketch showing the concept and the question at hand, but suggesting that small errors in proportion can lead to erroneous first impressions of how things look and act or might work, so be careful. It’s a little akin to saying, ‘I have a cantilever problem,’ and then showing a sketch of a 3' long, 6' deep canti. (corbel), with a point load out at the tip, but my actual problem is a W8 beam which is 10' long, with the point load. They are both cantilevers, but two drastically different problems and the sketch is very misleading given my actual problem. From an experienced engineer, at first glance, you will get a completely erroneous early/starting discussion with the corbel sketch. The 10 and the 6 on the loaded/top pl. are the same size, and the 40 btwn. the pl. centers seems to small, and these proportions are very important in the way this whole thing works. I’ve assumed the 10mm pl. thickness is accurate as my basis of scale. So the welds are smaller than shown, and I think I see much more deck pl. bending than your sketch would suggest. These types of details are really difficult to analyze to the nat’s ass, when you really look at them closely, as KootK suggests. I’d likely take a quick look at deck pl. bending btwn. the two weld groups. For example, the deck pl. essentially fixed-fixed at the two weld groups, and what happens with a little vert. displacement of the loaded pl.?

I don’t have an exact numbers answer for you, but rather a design philosophy or general approach. At first glance, I’d say the welds are loaded thus: .3P & .7P and .7Fr & .3Fr, from left to right, maybe .25 & .75. And, Fr = P, almost no load will go to another stiffener 6.5' away. I would make the loaded pl. as long and as stiff as practically possible, to reduce the load per unit length. Maybe it should be a 3x3 angle with the toe pointing to the right, so it can be welded to the right of the deck stiffener also. I would hope for good penetration at the roots of the welds and inspect for any undercutting where the welds meet the loaded pl., the deck pl. and the deck stiffener. I would allow that the two .7P welds would yield under max. load and as soon as that strain started happening more load would be transferred to the outer two welds, thus limiting the yielding. I’d probably take a look at full load on the two nearest welds at ultimate stress. Then I’d go get a coffee or better yet a beer, and let this whole problem simmer a bit. Then, I know I can’t spend more than 2 weeks on this one detail. I know FEA won’t help me much on this type of problem, it raises more questions than it answers, because there will be great big red areas at every corner or surface direction change on the detail, and we don’t know what to do with von Mise stresses anyway. Finally, I know that any mis-location of the load pl. w.r.t. the deck stiffener pl. or any imperfections in the details (weld defects, etc.) will affect the way things act much more than changing my assumptions a little bit, but that’s what FoS are there for. Another beer and I’m done.

I don’t look at a detail like this, at first, and think what are the moments, shears and reactions so much as, how does this detail deform under load, and how do those deformations cause the forces to be distributed. What yielding and redistribution can I tolerate and still protect the entire detail? That doesn’t give me exact stresses either, but it helps me understand how to approach the problem. Through plate tension can be a problem if there is a slag inclusion or lamination in the plate. If it’s important enough, tensile stresses high enough, a quick UT exam will reveal any laminations, slag inclusions less so. These laminations are usually not a problem becuase of the way plate is usually used, they are parallel to the stress flow. This should probably be less of a problem today, what with continuous casting and all. The plates do have slightly different properties through the plate as opposed to properties in the direction of rolling or across the grain. But, in normal structures these variations are usually not significant. You should just know that the best stress orientation is in the direction of rolling. This detail and these plate thicknesses and weld sizes should not usually cause a lamellar tearing problem. Generally, the plates must be thicker, the welds much larger, and the joint more highly restrained within itself, for lamellar tearing to be a problem. This detail is not restrained enough in that way.

 

[Ussuri]

Thanks dhengr. The original discussion came about when trying to assess what might be a realistic load distribution. Many engineers had many different takes on the matter. When a finger in the air (or complete guess) is needed I find Pareto 80/20 rule works well enough.

Your point about not actually knowing what the state of stress is true. We would have no idea what the stress in this area would be before we applied the load. We generally do not consider the global effects due the ship girder flexing when at sea. There are half a hundred other things going on as well such as residual stress from the fabrication of the ship itself.

So an envelope type approach including a number of assumptions would negate knowing what the actual distribution is, suffice to say its somewhere between two boundaries.