LRFD or ASD? The saga continues... 12

[vmirat]

I have a project being designed for us by another contractor. It involves a simple support bracket system for HVAC duct work. The structural engineer did calcs using AISC ASD method.

Whenever I see ASD used on design, I usually ask why, out of professional curiosity. Here was their response:

"Designing a large structure with large quantities of steel one should use LRFD to take advantage of the cumulative weight savings. That is not the case here, so ASD was used for simplicity and speed."

I'm wondering if we are schooling new engineers in both methods for this reason? I didn't take this any further, but I wonder how they decide the break point for LRFD vs. ASD.

 

[frv]

I still don't know why everyone keeps repeating the same argument. Savings have little to do with it.

LRFD will save you money only if your LL to DL ratio is less than 3 and serviceability is not the controlling factor. It has nothing to do with small or large. LRFD isn't any less conservative and you can just as easily modify it.

I really don't understand the argument about designing something to a "gnats ass". If, using ASD, you design whatever member for a unity check of 1.01, then you're going to have just as difficult a time adding anything to it as an LRFD-designed member that has a unity check of 1.01.

Importantly, the other option is just as true. If your unirty check in LRFD is 1.1, it is just as easy to add some load to the member as an ASD member designed to 1.1

 

[RFreund]

frv - I agree but if serviceability does not control and your LL to DL ratio is less than 3 there is a difference. I am saying that difference is small (comparable to a gnats ass). But if you trying to make something work than that gnat's ass may make the difference...that's all.
Not that I would want to go around designing this way.

EIT

http://www.HowToEngineer.com

 

[dcarr82775]

I agree with frv,

You get about the same answer. I can make certain conditions work for LRFD and not ASD, and the opposite is also true for other loading conditions. A gnats ass is a gnats ass and if you think otherwise you are fooling yourself.

I also don't agree with having a better 'feel' for something with ASD. Having a better feel with respect to an arbitrary reference capacity value (say 0.66Fy) at which absolutely nothing actually happens is no better that having a feel for an inflated load relating to a value at which something might actually happen (i.e. failure)

 

[DST148]

Some of the posts above suggest that "ASD is easier to use since you only have to deal with one set of load combinations for strength and serviceability".
Ref ASCE 7-05, Appendix C - Serviceability Considerations: Although this Appendix is not a mandatory part of the standard, it does provide guidance for design of serviceability. The load combinations suggested in the commentary of the Appendix C are not the same as the load combinations given in Chapter 2. Moreover, where allowable stress design is used , the drift is computed using the strength level seismic forces. The return period for wind could be different for strength(ASD) and serviceability. Therefore, you still have to write separate load combinations for checking serviceability. Given an option there will always be preferences, but the number of load combinations should not be the criterion in choosing between ASD and LRFD.

 

[weab]

Having lived and worked through the whole ASD/LRFD mess, here is my observation. AISC made a mistake of abandoning "stress" design and ignoring those who preferred it. Finally (?), AISC decides to reintroduce ASD, but strength design is not the same as stress design. Of course, if new ASD is used, it is possible to divide by the area of a shape to determine stresses.

The problem is twofold. First, while 20 years pass and AISC ignores most (yes most) structural engineers desire to use stress design, the code has advanced. That's a good thing. However, the formulas have become out of control and few formulas appear to be what we remember. So what some of us attribute to LRFD is actually just code advancement. Has anyone really looked at formula F4-8? Really!? In the end saving up to 5% of steel is not important to many of us. Saving 5% of cost for something that is only 10% of the overall project cost is not worth anyone's time. Let's face it, opening a building/structure/facility one day earlier pays for the extra steel in my projects.

The second problem is also AISC's making. They get us interested again by reintroducing new ASD. They either think we won't notice that stresses are not used, or they actually believe that we just didn't want factored loads. We use factored loads in concrete design.

In the end, I find that the black book is only better because it addresses issues not addressed in the green book. It however does not provide a better end steel product, just longer formulas. Anyone want to try to convince me that there isn't a shorter formula F4-8 that is 99% as good.

 

[271828]

I don't think Eq. F4-8 is a very good example, weab. That's in F4 which only applies for noncompact and/or monosymmetric beams. If someone's dealing with those, he's probably into metal building systems or in some other special area. They'd program that equation immediately and wouldn't care if it was twice that size and had a few more square roots.

Surely, I'll be pointed to Eq. F2-6, so I'll go ahead and answer. That Lr is tabulated in Part 3 for every standard shape, so who cares if it's a couple inches long on the page? If you're in F2, and are dealing standard shapes, you're good to go. If you're using doubly symmetric built-up shapes, you're in a special area and don't care if the equation is large.

Besides, [nerd hat on]I timed myself punching Eq. F2-6 through my calculator three times and finished each try in around 25 sec. with no mistake.[/nerd hat off] LOL

Those short ASD89 equations came at a price: clarity.

Take Eq. F1-7 and F1-8 for example. The older Salmon and Johnson texts included derivation of those. There were so many steps and approximations to go from the theoretically exact elastic buckling moment that the new equations bear no resemblance to what they stand for. I wonder how many times they've been misapplied over the years. Huge numbers, I'm sure.

Here's another one: 0.66Fy vs 0.6Fy. Earlier in this thread, someone typed "I guess I disagree that old and new ASD are, at the root, the same thing. New ASD is Limit State Design, right?" which revealed that he or she didn't know what the 0.66Fy was all about in the 1989 Specification. I've heard numerous people reveal that over the years.

 

[SAIL3]

By reading the responses one pretty much figure out who is a practicing engineer in the real world and who is operating in the so-called ideal world of academia.
Also, who are the most recent graduates and those of us who are more experienced(old)..
I generally agree with frv and weab about all this focus on saving 5% of steel weight(if it really exists) when other factors in a project can blow this 5% savings away...steel is cheap in comparion..
Before history is revised, the fact is that the majority of practicing engineers rejected the use of LRFD for many years.
If there was a clear, distinct advantage in using LRFD over the old ASD, these engineers, me included, would have embraced it.

 

[271828]

http://www.nizkor.org/features/fallacies/ad-hominem.html

 

[JAE]

Maybe I'm odd SAIL3, - I've been practicing engineering since 1979 and use LRFD today - after a number of years using the light blue, red, and green manuals with ASD.

I'm not living in academia.

Engineers (people in general) just don't like change.

 

[dhengr]

JAE hit the nail on the head with his last sentence. And, I would add, they dislike change all the more if the rate of change gets so fast as to make it impossible to absorb and apply without greater chance of error, and the more so when the reasons for the change or the benefits gained are almost impossible to appreciate, except for the complexity they add to the entire process.

 

[ron9876]

Well said dhengr. Like Appendix D in ACI and the latest changes in ASCE7 wind design. After all is said and done what has been accomplished except adding more complexity which brings a greater risk of a mistake.

 

[a7x1984]

In general:

Economics aside, I would tend to lean towards this: If I am very concerned about quantifying unique probabilities of loading and material behavior, ductility and post-yield strength, I would rather use LRFD.

If the structure is not critical for progressive collapse or overloading, ASD all the way. With that said, AISC has unified the two methodologies pretty well. I doubt my preference really matters in that case.





"Structural engineering is the art of modelling materials we do not wholly understand into shapes we cannot..."...ah...screw it, we don't know what the heck we are doing.

 

[ToadJones]

271828-
You surely must not evaluate existing structures in your work.

Limit States were also built-in, which means we didn't have to check 3 equations but rather just determine which equation applied and use it.


I also suspect that most posting here do most of their work on new steel buildings and use software for design and dont actually go thru the equations anyway.

 

[ToadJones]

Yea, those equations are really simple when you are choosing shapes from the book for which all of the nasty terms are given in tables....when they aren't your life starts to get a little more difficult.
It's a very narrow minded approach/argument.

 

[dcarr82775]

I have not run across an LRFD or ASD equation, no matter how long that can't be made short work of with a properly created spreadsheet. I do much more remodel/rehab than new work and have no trouble with LRFD or ASD. One is no easier than the other.

 

[271828]

Toad, I was taught LRFD in school, used ASD89 for three years, then LRFD for six. Designed lots of different types of buildings ranging from small and simple to large and difficult, including ones with existing structures.

The way I see it, there are three main differences between ASD89 and the 2005 or 2005 Manuals/Specifications:

1. The 2005/2010 Specification provides codified equations for many, many situations that are not addressed in the 89 Specification.

2. The 2005/2010 Specification leaves equations closer to their original forms, such as leaving some J's and Cw's in there rather than simplifying as in the 89 Specification Ch. F. This makes it easier to see what's in the equation. Because everybody uses computers today, I fail to see why this is a problem. Heck, we've had programmable calculators for a long, long time.

3. The 2005/2010 Specification is a LOT better organized.

Say you had a column that's an I-shape, but not a hot-rolled shape. Do you know where the flexural-torsional guidance is located in the 89 Spec? (Commentary to Ch. E) How about slender compression element stuff? (Appendix B) In other words, if you had just about any column out of the ordinary, you had to pull together pieces and parts from the Specification, an appendix, and the Commentary.

ALL those are in Ch. E of the 2005/2010 Spec., where they belong.

To keep you from having to look at them if you don't want to, the FTB and slender element parts are tucked away after the routine parts in Ch. E.

Yea, those equations are really simple when you are choosing shapes from the book for which all of the nasty terms are given in tables....when they aren't your life starts to get a little more difficult.
It's a very narrow minded approach/argument.

Those who do regular types of designs will be pulling values from tables almost every time. In the Specification, those regular design provisions are right up front where they're quick to find. If the pages behind E3 in your Ch. E offend you, you're welcome to tear them out LOL.

If someone finds himself in Sections E4, E7, F4, or F5 (a handful of the sections that people seem to point at as examples), he's probably in some specialty area. If so, then he's an expert on those provisions anyway, and doesn't care about the complexity. If a non-specialist finds himself in there occasionally and the calcs take a little while, then is it that big of a deal? Is this the only time in most SEs lives that they have such a thing happen? Of course not.

This entire issue is completely obvious to me. People will naturally fight against almost any change. That's logical, seriously. If you successfully designed a building in 1990 with the 89 Spec., that building is OK, so what's the point in changing methods? That's a good enough reason to have a personal wish or desire that the steel timeline would've frozen itself in 1989.

However, I don't think anybody has any real leg to stand on in saying that the 89 Spec./Manual is superior to the modern ones, unless you consider ad hominem attacks or genetic fallacies as valid.

 

[ToadJones]

271828-
I give up man....You're 100% right.

I use the unified code every day.

For the specific problems it causes me, please refer to the following website:

http://www.cranerepairengineer.com

I am in no way affiliated with this company, but occasionally do the same type of work.

Please read, at your leisure of course, through the sections referring to "AIST Torsion", "SDC Torsion" "Torsional Warping Constant", "Plastic Section Modulus" and all of the excellent and extensive PDF's available on the site. This will clue you into just exactly how difficult a task using parts of the new code has become for those of us without a personal relationship with Dr. Galambos and an extra 500 hrs per year for research.

I don't mean a word of this facetiously either.


Excerpt from website:

The analysis of unsymmetrical built-up shapes is the most difficult part of the new AISC Code to understand. The Fcr appearing in equations F12-3 and F12-4 represents two different types of buckling/critical stress. There is no formula for either one of the Fcr’s except that the code requires that they be "determined by analysis". The Fcr in Chapter F should not be confused with the Fcr noted in other section(s) of the Code dedicated to the critical stress due to flexural-torsional buckling.

The Code generalizes the critical stress resulting from all buckling modes as "elastic buckling stress". As stated in the Code Commentary: "the stresses are to be limited by the yield stress or the elastic buckling stress. The stress distribution and/or the elastic buckling stress must be determined from principles of structural mechanics, text books or handbooks, such as SSRC (Galambos, 1998), papers in journals, or finite element analyses". Much as we tried per AISC recommendation, SDC has not been able to acquire any literature dealing with "lateral torsional buckling" for unsymmetrical sections. In lieu of referencing papers in journals, our last resort is to honor the AISC yield stress limit by keeping all the calculated stresses below the yield stress. SDC is undertaking major modification of our automated crane girder design tools including the following:

Accept user-defined "effective girder component" based on width-thickness ratio of each element up to the AISC non-compact limit.
Use "gross section properties" to calculate flexural shear- and all torsion-related stresses.
Use "effective section properties" per AIST guideline to calculate flexural fiber stresses.
For unsymmetrical sections, use SRSS combination of all ASD (bending plus warping normal) fiber stress with (flexural horizontal shear plus pure torsion plus warping torsion) shear stress and then limit the SRSS value to the smaller of: (a) material allowable stress of Fy / Ω or 0.6 Fy and (b) flexural torsional buckling stress.

This interim scheme has the concurrence of Prof. Galambos.　

 

[ToadJones]

In this specific case (large runway girders) saving steel will only create job security......for some other engineer.

 

[271828]

TJ,

Unsymmetric shapes can buckle in various modes, but that fact was not invented by the authors of F12. If people didn't have similar difficulties prior to F12, then does that indicate that they didn't realize that these different modes and checks needed to be made?

I have a criticism of F12 and some parts of Ch. F, a technical writing error (IMO) that I think makes it a LOT harder to understand quickly.

Look at Ch. E, which is clear. One computes Fe which is an idealized elastic buckling stress. Purely linear elastic, no residual stresses, no initial imperfections, etc. This is apples-to-apples to an elastic buckling stress one would get from a finite element analysis. That little "e" makes it pretty clear, I think. Then one goes to subscript cr for Fcr which is for either elastic or inelastic. I'm pretty sure Ch. E is consistent everywhere.

Now look at Ch. F. Now we have subscript "cr" on idealized elastic buckling stresses such as in Eq. F2-4, F4-5, F5-3 and 4 (even worse), and so on. The Ch. F authors (different people?--I don't know) do not keep the "e" elastic buckling stress clearly separated from the "cr" which is buckling stress for inelastic or elastic buckling.

They really need to have a little "e" on every idealized elastic buckling stress everywhere in the Spec. and Manual, and a little "cr" on a buckling stress which can be inelastic or elastic. "e" and "cr" subscript then banned from every other use anywhere in the document. I think people would get it pretty quickly without being steel code reading experts.

This is going the right direction, although slowly. The 13th Ed. had a "cr" in the single plate shear connections part in a place that had nothing to do with stability. (wretch LOL) At least that's gone now.

 

[271828]

TJ, oops, hit Submit too quickly. I meant to poke the 89 Spec. one more time LOL! In it, about everything in Ch. F had a "b" subscript, whether it's flange local buckling, yielding, or lateral-torsional buckling. Those all existed in 1989, of course, but it's hard to see it from looking at the Spec.

Maybe I'm the weird one, but it seems to me that more transparency is better. If AISC would've jumbled all of the provisions together, then I could see the problem. They put the more esoteric ones behind the more commonly used ones.