Tube/WF beam double angle connection 2

[SteelPE]

I have a friend who is a steel detailer using Tekla. Recently he was able obtain a trial version of the newest RISA connection software for review because he is interested in purchasing the product as it now links directly with the Tekla software he is using.

He contacted me yesterday with some concerns regarding a connection he was looking at on one of his jobs. In this instance he had a W24x162 beam framing into the side of a HSS8x8x1/2" column. He has the beam attached to the column with 2-L4x3x1/2" angles that are 20.5" long. RISA fails this connection stating that it is in violation of the Hb/Bb requirements of table K2.2A of the AISC 14th edition. Further investigation into the problem reveals a little explanation by RISA on their design philosophy:

http://risanews.com/how-does-risaconnection-design-hss-columns-with-shear-connections/

My question is whether or not RISA is correct in their interpretation? I'm sure they worked extremely hard on this matter and did the best that they could with the information they had. I also agree that axial loads on double angles are not really addressed well in the AISC, however, now we have a connection that technically fails. What are the options given this interpretation....

-use a single tab (not a good as a double angle in my opinion)
-increase the angle and column size such that Hb/Bb ratio is greater than 0.5 (never going to fly with the EOR or the architect)
-use a clip that is in compliance with the ratios provided by AISC/RISA, in this instance the clip would need to be less than 13-3/8" long or essentially a 11-1/2" clip with 4 bolt clip (now the clip isn't strong enough).

 

[SteelPE]

hokie66

Simple, you cope out the bottom flange and the beam web will slide in-between the angles that are installed in the shop

 

[hokie66]

OK, guess I was thrown by the picture not showing the cope. I have never seen that type connection.

 

[Ken62465]

In regards to Risa not showing that bottom cope. When I first saw the missing cope exported to Risa from Tekla I figured it wasn't necessary from an engineering standpoint and I could see why Risa left it out but found myself saying "you watch somebody is going to question this"...and there you have it before it ever left my office. JoshPlum, personally I think this should be shown accurately to what is modeled in Tekla, or from a manual/standalone Risa Connection input, because as hokie66 got confused by this missing cope I'm sure it will continue. It just would look more professional being true to form.

 

[JoshPlumSE]

Nutte -

You say the following:

"I'm talking about providing, and thus endorsing, a connection type that lacks published literature, lacks consensus in the engineering community, and has problems with it. On top of this, you're expecting detailers, not engineers, to make an informed decision about this non-standard connection type. I think Risa overreached on this connection. It should have been left as shear-only."​

Huh? My belief is that these shear connections are widely used. I think what you're objecting to the the consideration of axial force in the connection. But, what would you have us do? Should RISAConnection ignore the axial force completely (as, I believe, most other programs do)? Because, that a much less conservative and less responsible route.

Now, you might be arguing that we should fail any shear connection like this which has a non-zero axial force. That would certainly be conservative. I just don't think you have really reviewed what we're doing and what the code basis for it is.

Personally, I think using the well established failure modes and procedures of Chapter K to evaluate the HSS local wall failures is reasonable. It is certainly an extension of chapter K because the "branch member" connecting to HSS is a Wide Flage with a clip angle, not another Tube. But, it is a logical extension. You've got a rectangular cross section welded to the HSS member. Local failure modes in the HSS wall should be essentially the same.... I think most engineers will agree with our logic, or think that we are being a little more conservative than they would be.

But, the most important thing for me to point out is that RISAConnection is a program written for engineers. I understand that the integration with Tekla (which was requested by engineers who want to work more closely with their CAD/BIM/detailers) means that detailers may be using the program to some extent. We cannot prevent that. Thankfully, detailers (however knowledgeable or experienced) cannot stamp drawings. So, some engineer will need to take responsibility for these connections. That's who we really expect to be looking in detail at the RISAConnection input and output.

 

[nutte]

I feel I’ve been pretty clear with my comments. I think Risa has made a mistake by allowing this connection type, clip angles bolted to the beam and welded to the support, to take axial load. AISC does not endorse this connection taking axial load. If someone asked me to provide a connection like this for both shear and axial load, I would not do it. I would offer a shear tab.

What should the program do with this connection if there is an axial load on it? You shouldn’t ignore the axial load and say it’s OK. You should say the connection is not OK, or at the very least fail to provide any connection capacity, axial or shear.

I haven’t commented on the application of HSS to HSS truss equations to this connection, because there is a fundamental problem with this connection (under axial load) in the first place. This fundamental problem has nothing to do with the support member being an HSS column. Bending on the outstanding legs of the clip angles will cause the vertical welds to rotate about their longitudinal axis, resulting in tension stresses at the root of the fillet weld. This should be avoided. That’s why there is no literature on these types of connections taking axial load.

I also think we’re kidding ourselves if we say this program is not going to be used primarily by detailers.

 

[WillisV]

Josh

I agree with nutte on this one. This type of connection should not be used for transferring axial loads due to rotation of the angle tips at the weld (again irregardless of rather it is attached to an HSS or a Wide Flange). You don't have to take my word for it though, see Bill Thornton's response to the same question here:

http://www.modernsteel.com/Uploads/Issues/September_2011/092011_Sept_si.pdf

 

[JoshPlumSE]

Sorry if I have been obtuse here. But, I was primarily concerned with the context of the original OP's questions / comments, specifically related to the applicability of chapter K. Plus, I'm not really the "weld expert" in our office, and didn't spec out this behavior. Therefore, I was slow to pick up on Nutte's concerns, which have some validity.

To be clear, RISAConnection does NOT now and has never allowed the type of weld configuration that Bill Thornton was talking about in that MSC exchange. This was immediately identified as an issue during initial development because that weld configuration cannot resist the applied load. The statics just don't work. At least not unless welds can somehow be said to resist torsion. From the RISAConnection help file we came up with a way to address the statics issue:

When axial tension is present in the beam, the weld configuration of a single vertical weld on each clip angle (at the support) is inadequate, as that weld would have to resist the tension via torsion in the weld throat. Because no adequate methodology for the torsional strength of weld exists, the program will automatically place a weld on the bottom of the clip angles as well. The capacity of the weld is then taken as double that of a single clip angle shear connection.​

The additional weld along the bottom of the clip angles fully addresses the concerns regarding statics. Between the return o the top of the connection and the full length weld at the bottom of the connection this (at least partially) addresses the issue Bill Thornton mentions in that Steel exchange.

One thing that RISAConnection must deal with that code writers and academics rarely appreciate is that the real world isn't separated into shear and tension connections. Because of that, a truly robust program MUST be capable of dealing with tension in a rational way. Otherwise, you start failing connections that have a tension of 15 lbs and a shear of 15 kips. We'd like to be able to rationally address that case rather than blindly failing it.

Also, keep in mind that section 12.1.4 of ASCE states that all connections (for seismic resistance) must be designed for a minimum axial force equal to 5% of the demand shear force. How do you meet that restriction if you say that these connections can't ever take axial force. I guess you could say that these clip angle connections should only be used for projects in seismic design category A. But, that just doesn't feel right. I think most engineers would agree.

All that being said, Nutte's comments (now that I'm understanding them better) have been good food for thought for me.

In response, I've been running a couple of tests on these types of connection. It's looking like the tension capacity of these connections (as calculated by RISAConnection) is coming out to be something approaching 20% of their shear capacity (loading purely in shear or purely in tension). That's higher than I would have thought. So, perhaps we should add in a restriction on how high it is allowed to get. Some percentage of the applied shear, presumably. But, what percentage to use? It's gotta be greater than 5% based on that ASCE requirement. If we made it as high as 20% it would be irrelevant because it would already be failing. I guess that puts us at 10%, not exactly scientific though.

The key is for lower axial forces, RISA is using a rational approach with a weld configuration that works in theory. I don't have test data to back it up, but we don't have any code requirements specifically forbidding this. Then at higher levels of axial force the program will be getting a failure invalidating the connection because it is not primarily a "shear" connection as we have assumed.

 

[WillisV]

Also, keep in mind that section 12.1.4 of ASCE states that all connections (for seismic resistance) must be designed for a minimum axial force equal to 5% of the demand shear force. How do you meet that restriction if you say that these connections can't ever take axial force. I guess you could say that these clip angle connections should only be used for projects in seismic design category A. But, that just doesn't feel right. I think most engineers would agree.

Support welded double angle connections should not be used to meet the minimum integrity requirements. See Gustafson 2009 (

http://www.aisc.org/store/p-1818-meeting-the-2009-ibc-structural-integrity-requirements.aspx)

which states:

The configurations where the angles are welded to the support, often used as ‘knifed
connections’, are not well tailored to resist axial beam forces. The welds at the tips of the
outstanding legs, being eccentric to the beam web, are subjected to a torsion stress for
which there is no documented reliable resistance. These connection configurations should
be avoided where the Structural Integrity axial tension force is stipulated, unless
additional reinforcement is provided. When adding reinforcement, due consideration to
the effect on the rotational flexibility to accommodate the vertical reaction should be
evaluated.

Connections which have been evaluated successfully for integrity (and axial load requirements) are single plates (

http://www.aisc.org/content.aspx?id=29220),

end plates, and all manner of bolted-bolted type connections (e.g. single angles

http://ascelibrary.org/doi/abs/10.1061/(ASCE)ST.1943-541X.0000678?journalCode=jsendh)

Note that satisfying the integrity provision still does not mean it is a good idea to have day-to-day tension in the connection as part of say the lateral load resisting path. Generally meeting the integrity requirement just means the connection has to be able to handle that tensile load in a collapse type scenario where severe deformations and strain hardening are acceptable.

Finally, even with the weld across the base of the angles, I still strongly disagree with presenting this connection as one able to resist axial loading as to my knowledge there is zero literature or testing as to the complex stresses that would be induced on the welds, the deformation behavior of the angles (they will still pry/put torsion on the top of the vertical welds and could unzip the connection), and the more basic fact that no welder is going to pick up on this deviation from standard connection and it will not get done in the field anyway.

Intuitively, can you put 100 lbs on it - sure, but at what point do you make that cutoff in a computer program - well, that's why I don't write programs!

 

[JoshPlumSE]

WillisV -

I believe Gustafson's quote is almost identical to the quote from our help file regarding no documented resistance for weld torsion. Which is why RISAConnection doesn't use that weld configuration. Let me say that agin. RISAConnection doesn't use that weld configuration. No matter how many times someone objects to that weld configuration, that doesn't make it directly applicable to this discussion. It's not irrelevant, just not directly applicable.

That traditional weld configuration is thrown out even if the tension load is as little as 1% of the shear load. That's because (as Gustafson and the RISAConnection help file point out) there is no theoretical way to calculate / document the resistance of the weld. However, if you change the weld configuration by adding in a weld along the bottom, there is a very welld documented way to calculate resistance of the weld (the ICR method for an L shaped weld configuration). Due consideration has been made for rotational flexibility per Gustafson's comment. RISAConnection refers to these checks as "rotational ductility" calculations (which I believe most other programs ignore).

Note also that ANY shear load will already induce bending / prying / deformation on the tension side of the weld due the the eccentricity of the applied shear. And, that's without ANY tension load applied. So, contrary to what you imply, the presence of tension in this weld is NOT exclusive to the connection having axial force. Therefore, the concern over the effect of angle bending / prying also should apply to the pure shear connection. Perhaps the weld return (which is shown graphically, but not considered in the RISAConnection calculations) helps to alleviate this concern at low stress levels. I'd be willing to accept that argument.

I'm not certain that this is a "good" connection. But, there is a theoretical capacity of the weld weld group that we are reporting based on the well proven ICR weld calculations. That's doesn't make it a good connection, but it ceases to be an "irresponsible" connection that Nutte was talking about earlier.

Now, this discussion is the first objection that I have yet heard to this non-standard weld configuration in the two years that we've had it as a possibility. That may mean that no one is using this connection for tension, or that they are just blinding accepting it, or that RISAConnection is clear enough in presenting what it's doing that they don't have questions. Contrary to Nutte's belief, the vast majority of technical support communication for this program is with engineers, not detailers. Maybe that will change as this link with Tekla becomes more widely used.

Over here at RISA, we always try to give our user base the behavior it wants.... regardless of what I think should be done. Provided, of course, it doesn't clearly violate the code. So, I ask you what would be your suggestion for how the program should treat a 50 kips shear plus 0.5 kip tension for this type of connection? Automatic fail? If that's what our engineering users want, then we'll give it to them. But, we've had these feature out there for about 2 years and I had not heard that request yet. At least not until this discussion. Mostly, I have gotten responses like, "FINALLY, a program that is considering axial force in the design. Everyone else just pretends that it doesn't exist."

Based on this discussion, I now think the program should throw up a flag if the tension becomes "significant" in comparison to the shear. And, we will do that for a future release. Currently, I am defining that as 10% of the applied shear. Maybe that should drop down to 5%. I don't know.... I'll let our users decide that.

However, what happens when the tension is somewhere between "virtually non-existent" (defined as less than 1% of the shear) and "significant"? The current method in the program is a well documented weld configuration (L shaped welds). Is that enough to justify using it even though we can't point to any published examples or full scale testing of the weld configuration using clip angles? There isn't any specific code provisions that prevent it. There aren't any articles (even the ones cited in this thread) which would prohibit this revised weld configuration. It's an engineering judgment call. If our users would prefer that this be removed, then we'll do it. Though I have not yet had that request from a user.

 

[nutte]

Josh, I don’t think there is much more to be added to this conversation, but I respectfully offer the following.

The Gustafson article clearly recommends not using angles welded on the outstanding leg for connections with tension loads. If I was the Engineer of Record and it was my job to approve and ultimately accept responsibility for this connection, I would not.

I understand you’ve added a full return on the bottom. As I mentioned early on, and WillisV mentioned more recently, you still have the problem at the top of the weld where the axial force is trying to twist the weld. The instantaneous center of rotation method provides a capacity for this weld loaded in shear. I do not know of an application of the IC method for the out-of-plane tension load.

Note also that ANY shear load will already induce bending / prying / deformation on the tension side of the weld due the the eccentricity of the applied shear.
Technically this is true. But practically speaking, the shear-only version of this connection has worked well for a long time. The shear capacities we calculate have been shown to be conservative. So I would not worry about this nominal, unquantified tension stress. When we get to actual, quantified tension load, that’s where I would draw the line.

I understand what you’re doing, and I can appreciate the efforts you and Risa have undergone in developing procedures for designing these connections. This is the same thing I do when developing my own calculations. Ultimately an Engineer of Record has to approve these connections. When engineering judgment leads me apart from the norm in the industry, that’s when the chain of command comes into play: The EOR’s engineering judgment trumps mine every time.

 

[JoshPlumSE]

Nutte -

Let me just point out that I am actually very thankful that we've had this conversation. Because it has raised some red flags here at RISA, even if it took me awhile to recognize them. But, that's still a good thing. It keeps us honest and forces us to think about issues in greater detail than most engineers would ever consider.

During this discussion, I located a section in the AISC manual that specifically references adding tension to these connections (page 10-141 of the 14th edition). Based on that section of the manual (and this discussion) we will be making some revisions to this connection when it is subjected to tension. The current plan for the next major release is to do the following:

1) Add an automatic fail whenever the tension loads exceed a certain percentage of the demand shear. To emphasize our intent that this must be primarily a shear connection.

2) Add a limit state for "angle leg bending". We have to research this a bit further, because I don't yet know of equations for this limit state. However, this is the one limit state mentioned in that section of the AISC manual (Simple Shear Connections Subject to Axial Forces) that RISAConnection is not currently considering.

Personally, I would not be surprised if we find that the angle leg bending requires thicker angles to resist the axial force.... which then violates the rotational ductility checks.

 

[JoshPlumSE]

Nutte -

Let me just point out that I am actually very thankful that we've had this conversation. Because it has raised some red flags here at RISA, even if it took me awhile to recognize them. But, that's still a good thing. It keeps us honest and forces us to think about issues in greater detail than most engineers would ever consider.

During this discussion, I located a section in the AISC manual that specifically references adding tension to these connections (page 10-141 of the 14th edition). Based on that section of the manual (and this discussion) we will be making some revisions to this connection when it is subjected to tension. The current plan for the next major release is to do the following:

1) Add an automatic fail whenever the tension loads exceed a certain percentage of the demand shear. To emphasize our intent that this must be primarily a shear connection.

2) Add a limit state for "angle leg bending". We have to research this a bit further, because I don't yet know of equations for this limit state. However, this is the one limit state mentioned in that section of the AISC manual (Simple Shear Connections Subject to Axial Forces) that RISAConnection is not currently considering.

Personally, I would not be surprised if we find that the angle leg bending requires thicker angles to resist the axial force.... which then violates the rotational ductility checks.

 

[RFreund]

This was a good thread, one quick side question - Why the weld across the bottom of the OS Legs as opposed to the top or both when subject to axial loads?
Thanks

EIT

http://www.HowToEngineer.com

 

[nutte]

Josh, you're right that most engineers don't go into this much detail on simple shear connections. Connection engineers and software writers do, though. As for bending on the OSL in the welded case, Salmon and Johnson discuss it.

RFreund, the weld is left off the top of the angle leg to provide for rotational ductility in the connection. As the beam tries to rotate, the angle legs can flex away from the support, providing some rotation so the pinned support assumptions remain valid. A weld at the top of the angle would lock the angle in, preventing this rotation.

 

[RFreund]

Thanks for the clarification. That makes sense, but at the same time if the top were welded, wouldn't this be the same as a single plate connection. Apparently not I guess, but it seems like they would be similar except you now have plates on both sides?

EIT

http://www.HowToEngineer.com

 

[a2mfk]

Excellent thread!

Following up with what Nutte's discussion points, the shear tables in AISC have been developed over the decades based on research, testing, history of usage, etc. The fundamental premise is that the connection be flexible enough to allow shear transfer and no appreciable moment transfer, and it is a factor of the size and length of the welds, connecting members, the bolts, etc. When you deviate from these specifications, you get into unqualified territory, and though you may be able to prove it through careful analysis, most of us would rather not take this risk.

However, I was digging through my library and found this article from 1995 that addresses this very subject. I do not know whether this practice is now officially discouraged by AISC in lieu of other types of connections that can better and more predictably transfer both axial tension and shear?

Could you not transfer a axial tension/compression force by simply adding an angle clip to the top flange, welded to the column and top flange of the beam? The angle and weld could be sized and detailed in a manner where the angle would flex like a hinge with beam rotation but still transfer axial forces. I think I remember doing this (or something similar) before when I needed to transfer axial forces when a beam was being used as a drag strut in a diaphragm due to wind forces. I for one would rather add a small piece of steel to a connection to deal with this force than try and make a shear connection work for multiple forces.

Anyway, once again, good conversation.