Explain ASD vs LRFD to a dumb ME 20

[jdog1]

Can someone explain in really simple terms the difference between ASD and LRFD? ASD seems easier but is it going away?

I was reading through the new AISC Steel Construction Manual about the two methodologies and, aside from the different load combinations, was having a hard time seeing a clear distinction between the two.

 

[Qshake]

Agree with JAE on all points.

Uniform safety and a reliability based rational design is LRFD.

Used both ASD, the not so adopted LFD, and now LRFD and I don't see the big deal.

Same issue we had with metric, you're an engineer for cripes sake, shouldn't matter what units you use.

Regards,
Qshake

Eng-Tips Forums:Real Solutions for Real Problems Really Quick.

 

[zerosum]

jdog1, if you're still here: You sure stirred up the structurals with dancing cows and all, and peaked my interest, too. For a nice concise description of LRFD go to

http://www.iccsafe.org/ps/pdf/9008s2.pdf

 

[miecz]

I've lost my old LRFD introductory seminar notes, but as I remember it, LRFD was based on a lot of statistical analysis to arrive at scientifically based load and resistance factors. Fair enough, but I seem to remember that the factors, arrived at scientifically, were then "adjusted" so that real world designs came out closer to designs done by ASD. If that's the case, then it appears to me that the reliability theory was abandoned.

Also, the reliabiltiy of a design system depends on how it is used. With multiple load cases to consider, LRFD demands computer based designs. You've heard it yourself. At the end of any LRFD seminar, attendees with glaxed eyes start asking for the software that can perform all these calculations. Well, for my money, as soon as the designs are done by engineers who cannot design the element by hand, reliability goes down.

 

[JAE]

I use LRFD by hand all the time. There are just as many load combinations in ASD.

LRFD factors are indeed adjusted as you say.

What is done is that statistical analyses are performed to determine the weighted variabilities of different load cases. (i.e. live load is more variable and thus has a higher load factor).

Then, a measure of probability of failure is determined based upon historical design methods like ASD.

This probability is a [β] factor that is based upon the combination of load factors / phi factors. Then the LF/Phi factor is adjusted down or up to reflect the overall probability of failure desired.

The statistics are still there (i.e. the load factors are of different values based on variability of loads and the phi factors are relatively different depending on the material and the mode of failure).

They haven't abandoned the reliability theory at all. That is incorrect.

 

[miecz]

Then the LF/Phi factor is adjusted down or up to reflect the overall probability of failure desired.

Correct me if I'm wrong, but I thought the LF/Phi factor was also adjusted down or up to more closely match designs done using ASD.

 

[Calif]

Bows humbly in the direction of Master JAE (Yoda). He is correct.

The resisant virtues of the structure that we seek depend on their form; it is through their form that they are stable, not because of an awkward accumulation of material. There is nothing more noble and elegant from an intellectual viewpoint than this: to resist through form. Eladio Dieste

 

[JAE]

miecz - you're correct in my view. That is what I meant to say - that the overall prob of failure is adjusted to match traditional ranges of safety found in ASD. But you can't match it exactly because ASD doesn't provide a truly consistent level of safety. Just in general is the [β] factor adjusted.

 

[COEngineeer]

Hmm.. I need to look through the new NDS. It is for ASD and LRFD combined. How do they list the strength of the wood? Do they use the actual stress or do they still use the lowered stress for ASD?

Never, but never question engineer's judgement

 

[miecz]

JAE- As you say, ASD provides a range of safety levels and so the LRFD folks had a range of possible levels to pick from. Economy suggests that they could have selected the lowest level provided by ASD. As I remember it though, they selected a relatively high safety level. So, the result of switching to LRFD was to provide greater safety, rather than greater economy. Was the driving force behind LRFD that ASD provided insufficient safety? I don't think so. So, what problem were we trying to solve here?

 

[271828]

"So, what problem were we trying to solve here?"

I think that's the wrong way to look at it.

I'm sure an engineer or builder 8-9 decades ago would ask the same thing when faced with newfangled materials, design methods, and construction methods. Would we be better off today if no advances were made since the 1920s? 1960s? Do we think structural engineering knowledge advanced to a practical maximum level for all time in the 1980s (a pretty arrogant claim if you ask me)? I don't claim to know what structural engineers will be able to do with advances made between now and, say 2050. It's not about "solving a problem" in most cases. It's about "moving forward with new and better information." Is Beam 437 in a given building directly influenced? Probably not, in most cases. That's not the point.

 

[miecz]

271828-Point well taken. We're all for advancing the state of the art. If that was the driving force behind the LRFD movement, then it has achieved it's goal.

But a benefit of LRFD often cited is increased reliability. In that regard, I'm not so sure I agree. As I said earlier, the multiple load cases inherent in LRFD demands more use of computer aided designs.* I believe that, as more designs are carried out by computer, engineers will lose the feel for what constitutes a safe design. In that regard, I believe that overall reliability suffers. Is advancing the state of the art worth a decrease in reliability?

*Note: While it's true that IBC and ASCE7 require multiple load cases for ASD design, ASD traditionally used one or two load cases, and no load factors. Including load factors in ASD, as IBC has done in Article 2.4, essentially converts ASD into a Load Factor Design.

 

[JAE]

I don't think the safety was set higher with LRFD.

If you take a design - say a steel beam spanning 30 feet, and design it with ASD for a set load. Then take that load and design it with LRFD assuming it is all dead load. Then design it again with LRFD assuming it is all live load.

Here's an example:
Per AISC 13th Edition

Span: 30 feet
Load: 3 kips/ft (service)
Unbraced length: 0 ft
Fy = 50 ksi

[red]ASD[/red]
Moment = wL^2/8 = 3(30)^2/8 = 337.5 ft-kips
[Ω]b = 1.67

Mn = Mp = FyZx

Zx (required) = 337.5(12)1.67/50 = 135.3 in^3

W21 x 62 (Zx = 144)

[red]LRFD with all dead load[/red]
Mu = 1.4(337.5) = 472.5 ft-kips
[φ] = 0.9

Zx(required) = 472.5(12)/.9/50 = 126 in^3 (which is less than the ASD requirement)

W21x55 (Zx = 126)

[red]LRFD with all live load[/red]
Mu = 1.6(337.5) = 540 ft-kips
[φ] = 0.9

Zx(required) = 540(12)/.9/50 = 144 in^3 (which is more than the ASD requirement)

W21x62 (Zx = 144)


So LRFD sort of brackets ASD depending on the DL vs. LL ratio.

 

[271828]

"I believe that, as more designs are carried out by computer, engineers will lose the feel for what constitutes a safe design. In that regard, I believe that overall reliability suffers. Is advancing the state of the art worth a decrease in reliability?"

I am totally with you on this one, as I've typed around here many times. There *will* be another major collapse at some point due to mis-use of a black box.

I doubt that LRFD is to blame, however. The advent of the cheap personal computer and the currently popular, un-named automated steel design system, combined with tight structural fees combine to cause the danger, IMO.

Structural fees and schedules are so tight that it is very tempting to press "Run" then "Export to dxf" (or worse yet, "Export to SDS/2") and then never look at it again. I'm sure most of us can point to a pet case or two of that. I have one in mind that actually caused failure.

I'm going out on a limb to come up with an example of possible future analyses that might be carried out with LRFD, but not ASD. Here goes. Imagine that I have a load that is genuinely better represented by a probability distribution than a deterministic design load. If I have separate load and resistance factors, then this can be dealt with. If one factor of safety exists for both sides, there's no way to approach the problem. I can think of several good examples: earthquake, wind, footfall forces, etc.

 

[DRC1]

The factors for LRFD were adjusted based on statistical data to achieve on average a combined load factor of 1.5 which matches the .67 Fb in allowable design because everone felt comfortable with that number. So there is really nothing new except some racing stripes. Same engine. The LTB formulas are somewhat more refined but use the same basic curves that are 30 or 40 years old. Im not saying there has not been improvement in our analysis, its just not the quantum leap every one says it is. As for the basis of ASD stresses, it is the culmination of quite a bit of research. Most of it focuses in the significant variability of the strength of the structural steel. I do not think LRFD addresses this with any greater accuracy than ASD, as LRFD assumes uniform plastification of the member as the stress inceases. As for the allowable loads, For some work, they provide a significant margin of safety. When was the last time you saw a 50 psf floor loading? However for a lot of work you are actually working with measured loads and are designing based on actual loads. Thus I believe the point remains that AISC LRFD is simply a restatement of ASD and LRFD does not give you any significant information about your structural member except that it is big enough.

 

[JAE]

ASD stresses havent' changed in many many years. How that can be proof of the "culmination of quite a bit of research" is beyond me.

With all due respect, DRC1, ASD was the method one of my mentors used from the 1930's. He started his career in 1927 and was very comfortable with ASD. Fine and dandy. OK with me.

LRFD isn't a quantum leap...no one ever said it was. It's just based upon statistical balancing of variabilities. I did my thesis on Phi factors for metal deck slabs and read multitudes of papers on how LRFD is developed. Yes, it balances with ASD for general loading but does allow you to consider the variability of loading AND of the material variations.

From that research, you would know that your reference to the "variability of the strength of structural steel" has very little to do with the safety factors. 0.6Fy isn't about whether your Fy equals 36 ksi or 38 ksi. The 0.6 factor is simply, historically, a feel good level of safety for design. It's not about variability of steel strength - that only accounts for about 3 to 5% of the total variability of our level of safety.

LRFD does NOT assume "uniform plastification" of a member...it is based upon a limit state within the elastic range (i.e. at yield, not beyond it).

I agree with you that LRFD doesn't give you any magic information better than ASD other than accounting for variability (thus my design example above).

 

[miecz]

JAE-

So LRFD sort of brackets ASD depending on the DL vs. LL ratio.

True, but it shouldn't be that way.


Here's how I understand the sequence. LRFD theory started out with a set of assumptions and goals. Correct me if I'm wrong, but I believe one of those goals was to provide the same reliability as ASD. After all, we didn't have a reliability problem. If I understood the seminars, the initial result turned out to be more economical than ASD. Then, LRFD's [β] factor was "adjusted" to bracket ASD, or bring it more in line with traditional designs. So, instead of a design approach that had the same reliability as ASD and better overall economy, LRFD was adjusted to provide the same overall economy and better reliability. It's at that point that one of the initial goals (same reliability) was abandoned.

 

[JAE]

miecz - perhaps your history is correct. But LRFD has better reliability in that, for a specific combination of dead and live loads, there is a better adjustment in the factor of safety to account for the higher, or lower, variability of the loads. It also sets us up for the future as more data is researched to better understand variability of materials, formulae, construction, loading, etc.

I am with others in that the codes are getting into overkill mode with the attempt to get more and more precise with design methods - chapter 10 of the ACI code comes to mind - but my only main points in this discussion are:

1. LRFD isn't that hard to use (as some have suggested)
2. LRFD provides a better rationale for safety factors and produces a better level of reliability than ASD.
3. ASD is fine to use...just has a variable level of reliability that may or may not matter.

Having ASD and LRFD in the same steel manual is fine, I guess...[blue]but when I first saw it my mind immediately thought of a scenario where some confused, inexperienced engineer would wrongly use different tables for a single design and screw up.....like using a non-factored dancing cow load on an LRFD log.[/blue]

 

[miecz]

JAE-I agree with all of your main points, except that I differentiate between what I'll call the target reliability and the effective reliability. I agree that the target reliability of LRFD is higher, but I still believe the effective reliability is lower than traditional ASD methods. I see that most engineers disagree with that belief, and that's comforting, to a degree.

 

[COEngineeer]

JAE, another ASD and LRDF question. When I went to college I dont remember doing any deflection analysis when I took steel class (LRFD). So when I design a beam with LRFD, do you factor the load to find the deflection? On ASD I just add up the live load and the dead load and check the deflection and I dont think they reduce the modulus of elasticity for ASD right? So how do you do deflection analysis on ASD vs LRFD? Usually on timber the deflection controls (not the stress). I am one of those new inexperienced PEs :-(

Never, but never question engineer's judgement

 

[frv]

Deflection is calculated based on Nominal (not factored) loads. You should get the exact same deflection with ASD and LRFD for the same beam size.

The reduction in E is something entirely different. This is done when you check stability of frames under Appendix 7 of the 13th Edition of AISC.