Need to determine capacity of steel assembly through testing - how many tests? 3

[tdstructural]

I’m working for a mechanical fabrication facility and we need to perform tests to determine the capacity of our lift lugs on our Air Handling Units (AHU). These units are commercial and some are extremely large and the lift forces can be 4k – 5k in some cases.

I’m not concerned about the lift lug itself – it’s really beefy and calcs out with a really small Demand to Capacity Ratios.

The lift lugs are welded to the exterior vertical face of HSS framing members in the bottom of the AHU. My concern is with the connection to the perimeter base member and the capacity of the perimeter base member. We have performed some pull tests on the assemblies and at thinner wall HSS members the vertical exterior face of the tube plastically deforms. With our thicker tubes we sometimes break the weld between the lift lug and the vertical exterior face of the tube. Those are the modes of failure I expect to see in our new tests.

I should tell you that I’ve tried to perform calculations for the assemblies but my results are extremely low and I know from testing that the true capacity is much greater than what the calculations yield. The welded connection includes some proprietary information so I can’t explain to you why the hand calcs don’t work and I also can’t provide a sketch – please just take my word for it – hand calcs are not adequate.

I’m trying to determine how many tests for each tube size we should perform. I can’t find any literature that gives advice so my first question – does anyone know of a guide I can use for steel assembly testing?

If there isn’t a guide how about your input for a debate I’ve been having with my co-worker. He is using the code requirement of 20 tests for concrete as a guide and thinks that since steel is more homogenous and also we have control over the fabrication that 10 tests for each assembly is sufficient.

I think utilizing the Aluminum Design Manual Appendix 1 method should be sufficient. They require a minimum of 4 tests to get a Factor of Safety (FS) of 2.5. You can perform more tests if you want to reduce that FS but because these are lift lugs we plan on using a minimum FS of 3.0 so we don’t need to increase the number of tests just to reduce the FS. I feel that steel and aluminum are similar and this method could work for steel.

We just want to make sure we have enough tests to support our design but of course don’t want to waste company money on too much testing.

 

[AnimusVox]

"please just take my word for it – hand calcs are not adequate."

Famous last words - I would either revise the connection at the HSS so you're not worrying about rupturing/yielding the face of the HSS member. This is a common problem in structural engineering with HSS columns & shear tabs inducing a lot of stress in a pretty thin wall of the HSS.

Also, if the connection with the lugs, welds & HSS are all made of steel, why are you referencing the Aluminum design manual? I would use the AISC 360 and look at failure modes for the HSS & weld (if you're dealing with steel)

I don't have any experience with testing requirements or procedures, but you definitely want everything to work out on a hand calc.

I can't provide any supplementary information on design guides or such but it sounds like you're wanting a multi-faceted and thorough answer to a pretty vague problem, and without a sketch or larger background knowledge I can't help anymore than the input I've mentioned.

 

[Teguci]

I don't think its appropriate to grab onto a material testing procedure for a system that is outside the rules of that material. For steel, we have factors for welding, fatigue, bending, compression, shear, etc... If you can't declare what failure you are designing against, then the design manual is not appropriate. For a testing program, consider that there are many variables in a fabrication process that will affect the success rate for an assembly. Some of those variables may greatly affect the final product.

Your best bet is to get a calculation procedure to properly model your assembly so that you can use the steel guide rules and factors. Outside of that, you are breaking new ground.

 

[tdstructural]

We as structural engineers already rely on tested capacity values when hand calcs don't suffice:
Concrete expansion and epoxy anchors, screws, Simpson holdowns, etc. The last airplane you flew on might very well have had 3M double sided tape helping to keep the wing skins in place. These are all tested products - not calculated values. I'm trying to recreate that process but have no experience in doing so. I was hoping to hear from someone on this forum that does have experience.

The only reason I referenced the aluminum method is because I can't find a steel method. In my opinion aluminum with it's homogenous properties and it's controlled fabrication environment is more similar to steel than concrete. I may very well be way off base - that's why I came to this sight. I need assistance.

I will contact AISC and possibly even research the ICC-ES testing procedures.

Thank you Teguci.

 

[SlideRuleEra]

I’m trying to determine how many tests for each tube size we should perform. I can’t find any literature that gives advice...

Use statistics, which is of course a field of study by itself. See Process or Product Monitoring and Control for general information. A long used public domain reference is MIL-STD-105.

If "hand calcs don't work", it is because the hand calcs used do not accurately mathematically model what is actually occurring.

http://www.SlideRuleEra.net

http://www.VacuumTubeEra.net

 

[mike20793]

The ICC reports have to come from accredited testing agencies; you can find a list of accredited agencies on the ICC website. Also, there is a big push by Simpson to make their products more in line with code calculated values rather than pure testing. Check out their new anchoring catalog and notice some of the testing tables have IBC approved crossed out.

 

[dhengr]

Tdstructural:
In part, the higher FoS for some lifting components is because of their misuse, abuse and reuse over their lifetime. Your lifting lugs will only be used a few times during the entire life of the AHU.
Who says you have to test these lifting lugs? I’ve designed all kinds of structures and details without testing 18 of them before we actually built one or many for real use. Many of them have been standing and functioning for 40-55 years. If 2,3, or 4 test all show essentially the same results, I’d write my final report. However, you might ask, how many tests does it take to make sure that no welder will shoot a bull on one of the lifting lugs some day? Also, the 20, 10 and 4 numbers of tests that you mentioned were more than likely pulled out of thin air, but with some of the logic you suggest making some sense, and factoring in.
The fact that you can’t show us any of your national secrets, we can’t see the details, we know nothing about the AUH configuration or it base frame or the member sizes...., that makes it pretty tough to offer any suggestions.
Is what you are doing now, the first of a new design? Is the plane of the lifting lug perpendicular to the web/face plane of the light HSS to which it is welded? If so, I can imagine some local yielding. Have a ½" pl. (or larger thickness?), 10" long, formed to fit around the outside of the HSS, with a lower horiz. leg ½" smaller than the HSS width and a vert. leg ½" shorter than the HSS ht. Weld this doubler angle to the HSS, and then weld the lifting lug to the doubler angle. You might also just rotate the lifting lug so its plane is parallel to, and right along side the HSS web face plane. Then weld them together all around. I’ll bet you are getting high stresses at weld terminations, or at the HSS corner radii, and/or some buckling of the thin HSS web face shell.

 

[Ron]

Never fewer than 3!

One test tells you a result, but you can't rely on it.
Two tests can create confusion if the results are disparate.

With three tests, you chances of corroboration are much greater!

 

[Agent666]

The way testing works in my country (New Zealand) is that number of tests isn't as important as establishing the coefficient of variation, this in turn depending on the number of tests should establish a factor by which you multiply your test loads by to demonstrate compliance.

The lower the number of tests, the higher this additional load factor is. The higher the coefficient of variation of the entire population of the test items is, the higher the additional load factor is.

So in general terms you test first item, recording and noting at what load your acceptance criteria are exceeded. You divide this load by the additional load factor (which is based on a coefficient of variation), and if this is greater than your required working load, you pass. If it doesn't pass, you do another test, with a subsequent test your additional load factor is lower. Again you then compare by dividing both results with the new lower load factor and compare to the working load. If both results pass, then you stop as you have acceptance, if not you rinse and repeat until you have all tests passing and with each test you can lower the load factor, because in effect you are establishing with more certainty the variation across the population of your test samples.

 

[dicksewerrat]

What happens if the machinery drops to the ground? Are there employees around? I would test until the company safety head is satisfied that no one will be killed by falling machinery. After that look at the OSHA section on rigging and lifting apparatus.

Richard A. Cornelius, P.E.

http://WWW.amlinereast.com

 

[BAretired]

It should be possible to check the capacity of the HSS wall with reasonable accuracy using Yield Line Theory.

BA

 

[Tomfh]

I'm definitely in the testing camp. Why second guess reality when you can ask it directly?

 

[Teguci]

I'm definitely in the testing camp. Why second guess reality when you can ask it directly?

I think we're all in the testing camp. This makes me think of those $500k+ testing programs that are required for concrete anchors, assembly fire ratings, UL listings, etc... All those different failure mechanisms that need to be looked at.

Take concrete anchors as an example - They needed to test for - edge distances (1 edge, 2 edges), shallow embedment, different concrete strengths, shear/tension variations, drill bit size, installation tolerances, long term creep for epoxy systems (they knew about this even before Boston), pullout in cracked concrete, dynamic loading, ... Way more than 3 tests that check for a capacity in a laboratory setting.

Who wouldn't want to piggy back on previously performed testing programs? Moving forward on this one, I would suggest modeling the lug design as best a computer can to formulate a capacity using existing steel design procedures. Then provide testing to corroborate the calculated design. If the testing doesn't fit the calculation, find the errors in the calculation and correct.

For a lug design, lifting angles and dynamic loading would be a couple of variations that should be reviewed.

 

[j19]

Take a look at Chapter F (Tests for Special Cases) in AISI S100, or NASPEC. It gives guidance on how many tests to run based on the variation in your results and also how to calculate a safety factor. However, since you will be testing lugs then I don't think I would depend on the AISI calculated safety factor. Maybe defer to the old AISC manual that used a safety factor of 5 for turnbuckles and clevises that are used in rigging. I do agree with Ron in that you should run at least 3 tests.