Predicting Failure Load

[lancer360]

I have the rare opportunity to test a couple of old a-frames to failure. These are old a-frames that have been inspected and determined to be unfit for refurbishment due to a combination of weld cracks and corrosion. I have been asked to predict the theoretical failure load of these two a-frames when they were new. I primarily use Visual Analysis to check the overall design using AISC code checks combined with some hand calcs and every once in a while Ansys for potential problem areas. How much "extra" is left between a unity of "1" and theoretical failure? Does it vary depending on the particular AISC equation used, or are they fairly consistent? The cross member is loaded primarily in bending and the legs are a combination of bending and compression.

After the units fail we are going to send some of the material off for testing to get an accurate yield strength of the material that failed and use that info to adjust the model and see how close our models are.

 

[frv]

In theory at least, the failure load of the structure should be the resistance of the weakest part of your truss under your test loading conditions, NOT including any resistance factor.

Note that this approach ignores both helpful parameters (over strength, strain-hardening, etc.) as well as factors that may reduce your strength (for example, the cracking in the welds and reduced cross-section you mentioned).

Note: to come up with a failure load, you must also come up with failure criteria (i.e. what constitutes a failure? excessive deflections, the first member yielding, or the whole thing going "boom")

 

[lancer360]

From my general knowledge of the structure, I am pretty sure the failure is either going to be yielding of the legs right above the connection for the hydraulic cylinders or it is going to be a catastrophic "boom" of the weld holding the hydraulic cylinder padeye to the leg and the entire structure will come crashing down. I'm not going to begin to try and predict the failure load of the padeye weld because MPI inspection has shown cracks in the weld.

My hope is to predict the failure load based on the yielding of the leg and hope the padeye weld will hold on one of the two frames so I can get a correlation between our models and the existing frame. This is a secondary goal though so I am not allowed to strengthen the connection. The primary goal is to find out what the failure load is "as is".

I'm thinking that I should combine the bending and shear stress to get the Von Mises stress and compare that to the theoretical yield stress of the material and after the test compare that the actual yield stress once we complete the material testing. Thoughts?

 

[csd72]

It depends if you are talking abot LRFD or ASD.

The theoretical failure load is when the stress equals Fy or Fu (depending on how it fails) whereas, for example, the allowable stress for bending is 0.6Fy for design.

 

[lancer360]

I use ASD. I guess I am trying to figure out what the factors are that are built into the various AISC ASD equations compared to the yield strength.

 

[frv]

csd72..

I disagree. The failure load should be independent of the method of analysis. The true resistance of any member is some nominal value. This value is, in theory at least, the expected failure load. ASD and LRFD simply are means of determining what load can safely be applied to the member.

 

[youngstructural]

I wouldn't apply a code... They are by nature empiracle and safety-oriented. What you want is fundamentals, such as raw bending strength and tensile strength in welds.

You don't need to apply LRFD or ASD. Neither should be applied... What you DO need is to include the Overstrength which frv mentioned (need to get this from the rolling mill who produced the steel), or in a pinch you could use whatever the standard overstrength factor currently is in the USA. Here in New Zealand it's 1.25, so you can get a lot more strength.

I would suggest:

Mn = 1.25ZxFy (where Zx= Plastic section capacity)
Welds = Code Weld Strength/Resistance Factors Applied

Good luck, Take LOTS of phtos, and please please please let us know how it goes!

Cheers,

YS

B.Eng (Carleton)
Working in New Zealand, thinking of my snow covered home...

 

[lancer360]

Thanks YS. I think I will look at just the stress values. Question is on how to combine the bending stress (both axis) and the shear stress. I figure going back to the basics and using these to calculate the Von Mises stress would be the easiest and then comparing that to the yield strength would give me a reasonable estimate without spending days, if not weeks look at a non-linear anaylsis, plastic deformation, etc.

We haven't made a final decision yet on whether we are going to use water bags or rig up a hydraulic cylinder. The hydraulic cylinder will be more contolled, but may not give us the catastrophic failure we are after. Water bags will be quite dramatic, but I have a feeling we will damage the bags if we have a sudden catastrophic failure and the frame comes crashing down on top of them. There would also be a good chance of the water bag splitting open like a giant water balloon when it hits the ground. Water bags are not that cheap. A hydraulic cylinder could be put at the other end of the wire away from the a-frame so it wouldn't be damaged.

We are going to video tape the entire affair, but I doubt my client will make the results public and publish the video. I have done some cool stuff in the past, but this will take the cake. I grew up listening to my dad telling stories of destroying things in the lab (on purpose) while testing down hole oilfield tools. Now I will have a good story of my own!

 

[youngstructural]

Sounds like a good plan, too bad about the photos though!

Cheers,

YS

B.Eng (Carleton)
Working in New Zealand, thinking of my snow covered home...

 

[civilperson]

Yield in the extreme fiber is rarely the limiting load. That is why the plastic section modulus was referenced in previous posts. Compression buckling would be my guess as to the failure mechanism.

 

[SlideRuleEra]

Concur with YS, this is a "Strength of Materials" situation, not a Code issue. Anyway, failures almost always occur at the connections, not the members. Since the structures are "unfit for refurbishment due to a combination of weld cracks and corrosion", connection failure is what I would bet on.

http://www.SlideRuleEra.net

http://www.VacuumTubeEra.net

 

[wannabeSE]

AISC Seismic Provisions for Structural Steel Buildings provides factors for determining the expected yield strength for different steels. I am not up to speed on these provisions, but it may provide some useful information for your analysis.

http://www.aisc.org/Content/Content...c_provisions/SeismicProv_second_print_web.pdf

 

[pba]

This really is somthing I have no experience of - so I may be totally off beam.

Can you fit stress gauges to the frame before testing? As has been said, failure is likely to occur at a connection and these are known to be damaged. You might get some good correlation to theoretical behaviour prior to failure from the gauges.

If possible, you should also aim to be measuring deflections at lots of different points.

Finally, make sure you don't overload the foundation before you get the A frame to fail...

 

[asixth]

I would also like to hear of results, particularly when you talk about test failure loads versus calculated failure loads, and I am hoping the calculated failure loads are below the test results.

I had a friend of mine who did some work designing freestanding transmission towers. Government regulations require these towers to be load tested to collapse. This person got the failure load close enough to the mark but most importantly, predicted the failure mode. Most impressive of all, he did it all using the nonlinear analysis program he wrote in FORTRAN. Not bad at all.

I was amazed watching the video of the testing to see steel buckle out like a piece of spaghetti.

 

[csd72]

lancer360,

You asked "...How much "extra" is left between a unity of "1" and theoretical failure? .."

Do you still need to know what the comaparison is with allowable loads, or just to predict the failure load?

What are you hoping to conclude from the failure loads?

 

[lancer360]

These "a-frames" as we call them are just a simple u-boom made from pipe with square corners that are reinforced with a smaller piece of pipe forming a triangle in the corner and then capped with plate. About 1/3 to 1/2 way up the pipe there is a padeye welded onto the leg with a hydraulic cylinder connected used to move the boom in and out. All of this is mounted on a steel base. In the middle of the cross pipe is a padeye with a hanging sheave. According to my model the cross beam and the legs are pretty close when comparing unity numbers. The resultant load is pretty close to inline with the legs, so they are mostly in compression with a little downwards bending and some outwards bending due to the deflection of the horizontal cross member. Assuming the cracked welds hold, I expect the legs to fail from combined bending/compression.

I got a response from Charlie Carter with AISC and he said to use the 2005 AISC calcs and omit the omega and phi to get the predicted failure load. Seems to be very straight forward except I don't have the latest AISC manual. I guess I need to break down and buy it. I'm still using the 9th Edition ASD.

 

[lancer360]

There are two things we need from the predicted failure load.

1. How much load are we going to need to fail the a-frame so we can size the water bags or hydraulic cylinder.

2. Try and correlate how much factor of safety on failure do we actually have with our SWL using our current design methods.

If it fails early at the connection we won't get the data for #2, but the ultimate goal is to find out what the failure load is of this existing a-frame and compare that to the load that was being used on the a-frame when it was taken out of service.

 

[csd72]

You can download the 2005 specification from the aisc website for free.