Minor calculations as a structural engineer 9

[thomastheman]

Hi,

I'm a structural engineer working in Norway and have been working 6 years in oil & gas and 6 years with civil engineering. I've worked for different companies and what I see is that the level of hand calculations for minor calculations is quite different in each company. Someone have excel-spreadsheets for environmental loads, Mathcad sheets for the same, whilst others are making the minor calculations from scratch for each project. In oil & gas, the engineering companies are using the same one-span beam software when doing simple steel and column calculations, which also includes a large section library with section properties. These calculations can easily be done in the excel, mathcad or similar, but this particular software is still used across the industry for these particular calculations.

For civil engineering, its common to use a software package for environmental loads, but it also contains opportunity to calculate typical concrete beams, columns, consoles, foundations. The same software package is used across the industry and seems like everyone is doing it the same way.

So my question to the forum is, does this sound familiar with you coming from other countries than Norway?

 

[human909]

I think it's better to utterly flush the idea that the factor of safety covers design errors. Especially when the group includes newer engineers who are trying to make sense out of the design process:

Agreed.

For LRFD, the addition amount over unity, I've name to myself and anybody who asks the "I want to sleep soundly at night factor".

I've seen some engineers design things all the way to 1.0 or 1.01. That is their prerogative however I prefer a healthier margin and so would my clients. (Mostly industrial processing and manufacturing facilities. They are forever getting upgraded with additional and heavier equipment.)

 

[Tomfh]

I think it's better to utterly flush the idea that the factor of safety covers design errors

Yes, the stated intent of load factors/phi factors/safety factors isn’t to catch design errors. However, in practice, these factors do catch design errors, regardless of their intended purpose.

They typically catch anny design errors up to 50-100%, this being the overall amount of fat that these safety factors add, assuming otherwise normal conditions (no gross overload, no defective materials). While they won’t catch gross errors of say 200%, they regularly catch smaller design errors.

When an engineer slightly underestimates tributary width and nothing goes wrong, it’s safety factors catching a design error. For instance, we recently found purlins spaced at 1800 mm where the engineer assumed 1200 mm.

Every time an engineer inadvertently designs for too low a live load, such as 3 kPa instead of 5 kPa, safety factors compensate here.

Every time an engineer miscalculates the dead loads and safety factors cover it, that’s another instance of safety factors catching design error.

While safety factors won’t protect against gross errors, they catch countless minor ones. I’ve seen many structures only standing because of these applied safety factors.

 

[Smoulder]

Agree with Tomfh. The factor of safety sits right about where things start to look wrong so get fixed. The ones that don't look wrong are close enough most of the time.

 

[271828]

On the safety factor discussion, we're just talking past each other. In the interest of not further hijacking, I'll drop it from my end.

Have a good weekend, folks.

 

[Just Some Nerd]

Isn't it kind of self-evident what safety factors are intended for in LRFD? Load factors aim at unexpected differences in loading, resistance factors aim at the materials?

----------------------------------------------------------------------

Why yes, I do in fact have no idea what I'm talking about

 

[DTS419]

Safety factors are simply NOT there for design errors by the engineers, they are there for other things, variations in the materials, loading, and construction tolerances. I'm not sure I've seen any comprehensive listing of what exactly the phi factors are accounting for, but back when I was going through Reliability with Ted Galambos, the monte-carlo simulations we did all involved various coefficients of variation of the sections, yield stresses, etc, so even something like tolerance on rebar placement wasn't in that calculation (which, admittedly, was for structural steel, I mean it's Ted Galambos teaching it). I'd expect some measure of tolerance on rebar placement in a flexural member to be in that sort of reliability calculation behind the phi factor, but I can't confirm or deny it's existence as I just don't know the background on concrete design that deeply. They rolled out their LRFD before I was born.

Simply put, safety factors are there to account or all variation and mistakes from design, manufacturing, and construction. This means the engineer who didn't calc the right wind load, didn't model the connection with enough fixity, etc. to the concrete mix that had too much water to the column that was set out-of-plumb. As most of us here have experienced, there can be just as much slop in engineering as there is in the field.

 

[271828]

DTS419, design errors can be of any magnitude.

If the engineer calculates a wind load that is 5x too low, how do safety factors help with that?

 

[DTS419]

If the engineer calculates a wind load that is 5x too low, how do safety factors help with that?

Obviously standard safety factors don't protect against gross errors in engineering any more than they protect against a post-installed concrete anchor being installed 2" deep when it needed to be 8". The point is, given the fact that any three engineers will provide four different solutions to the same problem, the safety factors provide some margin from variations in design judgement, minor errors, etc.

 

[human909]

Calling them safety factors in LRFD is somewhat misleading and diminishing what those factors are there for. 'Safety factors' used in other contexts that are not LRFD are pretty darn obscure and often arbitrary.

"Safety factors" in their traditional sense are nice and easy and give you a warm and fuzzy feeling but they are pretty black box in nature. LRFD tries to dispel the black box and approaches things in a probabilistic manner. (The only problem here is we humans don't normally think in probabilistic distributions so safety factor notion is often better understood.)

If you design to a factored CAPACITY/LOAD=1.001 then I'd argue you don't have a 'safety factor'.

You have designed to the bare minimum acceptable probability of failure. If you design to a CAPCITY/LOAD=0.90 then your structure will almost certainly still stand but it inherently has a higher probability of failure. It isn't still standing due to "safety factors", it is still standing because you have not lost the probability lottery.

 

[lexpatrie]

EFFORT POST WARNING......

I will preface this that it is perhaps a bit abrupt in the wording. I don't mean to deliver some kind of ideological smackdown, I'm under the usual swath of deadlines but feel it is important to reply in a substantive fashion, promptly, lest this kind of either (misread?) statement catch on, even if it's just misconstrued by the various readers far down the line. Please keep in mind the context here. I have a tendency to be involved in various construction defect and peer reviews and while these comments may be harsh, or tersely worded, they are ultimately meant the way Neil went through Terrence Howard's thesis on math, or physics, I'm not sure what that thesis was supposed to cover. I haven't read it and I don't recall the Discussion from Neil all that well at the moment.

If we shadows have offended....

Simply put, safety factors are there to account [f]or all variation and mistakes from design, manufacturing, and construction

NO.

NO. NO. NO.

NO.

They are there to provide a reasonably low (or "acceptably low) probability of failure, and to accommodate material variations and tolerances. Normal variations in materials, tolerances, etc. NOWHERE in there is an engineer's mistake (or for that matter, a "junk" material, or wildly deficient construction, half the required rebar, half the required concrete strength) factored into this.

Find the literature if you think this is factual. I am not aware of any. Apologies if I misconstrue your comment.

The betas (reliability indexes) [indicies?] for beams are all based on targets of 2.0, 3.0 or thereabouts (this is a wide range in probability of failure as I'm working off recall here and don't want this conversation to devolve into irrelevant minutia). (That's maybe harshly worded more than I mean it to be).

Source: Galambos, 1981.

This is a probability of failure of 1 in 1,000, 1 in 10,000 or less for a component, (to be clear, speaking off the cuff, here I remember seeing these once and don't recall where). Connections, as I recall are "safer" meaning they are higher reliability indexes as we want to prevent them in a probability standpoint because "failure" in a connection can be substantial whereas some beam "failures" involve slight plastification of the section, slightly excessive deflection, and so forth. AISC commentary is available on this subject.

This isn't what I'm looking for, but it's in the neighborhood.
Reliability Analysis of Simply Supported Steel Beams, Idris and Edache, Australian Journal of Basic and Applied Sciences, 2007.

More to the point.....
Galambos, Load and Resistance Factor Design, TR Higgins lecture, 1981. (cited by 153)
Bartlett, F. Michael; Dexter, Robert J.; Graeser, Mark D.; Jelinek, Jason J.; Schmidt, Bradley J.; Galambos, Theodore V. (2003). "Updating Standard Shape Material Properties Database for Design and Reliability," Engineering Journal, American Institute of Steel Construction, Vol. 40, pp. 2-14.

The point is, given the fact that any three engineers will provide four different solutions to the same problem, the safety factors provide some margin from variations in design judgement, minor errors, etc.

This is a totally separate argument and the two aren't related.

This is a more reasonable view, there will be some variations in an engineering design between various engineers, this is where Standard of Care comes into effect. Reasonably safe. There will be variations in results, particularly when one engineer uses, say, the envelope procedure for wind loads, another uses the SEAW rapid solutions methodology (and it's ideological descendants), and another engineer uses, say, the analytical procedure. One engineer using width x height for C&C loads, another uses effective area, LRFD versus ASD, "Old ASD" (meaning C_c, old C_b, etc), and so forth. While these answers "differ", they do not differ in the sense that they are all generally accepted principles of mechanics and the load standards, still use a consistent mathematical base (1+1=2 and so on). All of these variations (when correctly performed) result in a code-conforming design that is deemed "acceptably safe" (if you would.).

HOWEVER:
Every non-conservative error a design engineer makes erodes the safety factor. Period. There is no splitting this hair. The question is does the error matter?

Every non-conservative error increases the probability of failure, and makes the structure less safe. The deeper question is at what point does a non-conservative design error so impact the structure or element's safety as to make it unsafe, which is relatively undefined. At which point is is no longer "reasonably safe" or "acceptably safe." Again, there is a pretty fuzzy boundary between "conforms to the code" and "meaningful impact to safety due to a deviation," (like, for example, a roof that's good for a ten year event, not a 700 year event, a floor that's accepatably safe for 40 psf but not for 100 psf. (if you agree that "meaningful impact to safety" is even a category, and "unsafe" and "hazardous", and "negligence" and "incompetence" are separate questions that also lie along this domain, farther along, and somewhere in there lies "doomed to collapse during construction."

We have a number of situations and data points on this subject, including Harbour Cay, where a combination of calculations that weren't done, construction errors, bad rebar depth due to the wrong chairs, too aggressive(?) shoring and reshoring procedures doomed several construction workers to their deaths. The Hard Rock Hotel's "underdesigned" beams, along, apparently, with some issues involving misreading the allowable construction spans of the metal deck, with similarly fatal results for construction workers.

These are not pathological examples. I know a lot of us might be tempted to write this off as exclusively a construction error, and surely there are structures with engineering errors that survived construction. My point is there is a "hard" upper limit where an engineer's mistake, no matter the construction procedures, cannot be safely done, and therefore, is an unsafe design. Given the nature of construction, I would be surprised if there's ever a situation where a fatal collapse is ever "purely" a result of an engineer's design error. But the boundary exists.

[I will dismiss the Hyatt regency walkway, partly because it's so OVERUSED, and also, there was no calculation, both parties believed the other was performing an analysis, and there was no testing, so this is a different situation, IMHO, plus the potential for the unforeseen rhythmic dancing and crowd being a factor, i.e. a deficiency in the code regarding live load for the walkways, I will leave open. That was perhaps considered and dismissed, I don't recall at this point), The FAQ on Cantilever roof framing (disclosure: I wrote that) has information on several more: The Bolivar, Tennessee, (Magic mart) collapse, and the Burnaby, BC (Station Square) collapse, with about forty little pieces including some really poor engineering decisions made by an engineer and a peer reviewer, a transcription error in the beam size, extra concrete added by the contractor, etc. etc. etc).

Even Citicorp in NYC (cross bracing bolted when it was designed to be welded) and that unnamed(?) 47 story building in Texas that had a deep wind design flaw, (article in CTBUH) can all be reference points.

Additionally, the Dexter and Galambos article is on point, regarding "eroding a safety factor" as I recall, as they did not consider increasing the phi factor to be "acceptably safe" and left it where it was. So that could give you a sense of where, exactly, the P.E./Ph.D double domes and others behind the development of the Steel code consider the factor of safety as subject to revision or debate or "shaving" (for lack of a better term,.

Moving back to the Standard of Care, [ In the U.S. we are basically "witch doctors with slide rules and pocket protectors" (my phrase). We practice an art and/or profession, similar to a Doctor, Lawyer, Accountant, or other Professional. We are trusted (in fact) with protecting the life-safety of the public in performing work that they CANNOT asses is done competently done (unless, I suppose, they want to repeat our schooling and apprenticeship).

Our standard of care is to practice within the parameters of our "local peer group" (if you would). So, if you live (as a Doctor) in an area where EVERY doctor leaves surgical instruments inside the patients body cavity, along with five or six surgical sponges, then you're practicing within the standard of care. (I suggest you not get surgery in this area, however, go to Thailand for surgery).

CACI instructions 600

What is the Standard of Care, Kardon, Forensic Engineering 2018: Forging Forensic Frontiers ( I don't seem to actually have this, hmmm).

Ok, here:

Source: Kardon, Testifying Regarding the “Standard of Care”, ASCE 2009 Forensic Engineering.

I might as well include the various tables from Galambos in the development of LRFD (LRFD has a "rational" safety factor based on probability and statistics, ASD has a non-rational safety factor based on the development of the codes prior to the statistical and probability methods developed and applied in LRFD, the two are not that dramatically different, fortunately, but the point being that ASD had no known probability of failure, even if LRFD did not produce dramatic changes, the work was still necessary and foundational in establishing a baseline "safe" probability of failure.

Disclosure: I took about eight classes from Ted Galambos, maybe more. He is an all around swell fellow, very generous and patient with his time, and WAY more gifted in the mathematical and engineering sense than I will ever be. I have a vantage point that not many have available to them, hopefully this was helpful.

Thank you for coming to my TED talk!

Hopefully this was a conversation you perhaps didn't want, but a conversation (ok, diatribe or lecture depending on your mood) you needed.

TLDR - "Safety" factors are not intended to cover for an engineer's mistakes, large or small. They are for normal material property variations, load variations, section property variations, modeling variations, [normal construction, erection and fabrication tolerances], slight misalignments or floor-to-floor deviations (maybe), etc. They won't make an incompetently designed structure, or an incompetently or fraudulently built structure "safe" (not that anybody SAID that), though it may limp along for some time, then collapse, or the errors may have minimal effect on the overall safety of the structure and never be noticed (which is more what I think DTS491 is actually trying to put forth). Not all structural errors are fatal, not all are consequential, but some ARE.

Sidebar:
The issue with "beta" is that it's NONLINEAR, and probably a bit exponential, small changes will produce small changes in the probability of failure, but larger changes may produce dramatically more changes in the probability of failure. Think of your various small angle approximations for sine and cosine, for example, or all the structural theories you learned in College where "H.O.T." (Higher order terms) are thrown out due to small deflection theory.

The first 10% is perhaps not that consequential, maybe even only increasing the probability of failure by 10%, however, going to 20% will probably produce and increase in probability of failure of 30%, or more. At some point, the "informed" engineer (practicing in our area, using the same skills and care that we use, etc, etc, etc, ) makes the cutoff, BUT, understand where they tend to make the cutoff is NOT based on the beta or the probability of failure, they make the call based on the "overstress" ratio. Some will go to 1.05, others 1.1. NOBODY (based off what I see here on eng-tips) is making this cutoff based on a reliability analysis, a beta index, or a monte carlo simulation.

I would expect that four out of five dentists would find a 50% increase in probability of failure unacceptable. Where in the "overstress" calculation does that happen? Probably somewhere above 1.05 on the actual/allowable strength/stress/deflection calculation.

 

[Tomfh]

The point is, given the fact that any three engineers will provide four different solutions to the same problem, the safety factors provide some margin from variations in design judgement, minor errors, etc.

The Galambos paper Lexpatrie linked to seems to agree with this, stating that in addition to covering material and load variations, these factors also account for variations in the accuracy and precision of the analysis. They won't catch major blunders, but they do address the minor mistakes. Galambos:

they account for the uncertainties
inherent in the determination of the nominal strength and
the load effects due to natural variation in the loads, the
material properties, the accuracy of the theory, the precision
of the analysis, etc[/quote]
​

Calling them safety factors in LRFD is somewhat misleading and diminishing what those factors are there for. 'Safety factors' used in other contexts that are not LRFD are pretty darn obscure and often arbitrary. "Safety factors" in their traditional sense are nice and easy and give you a warm and fuzzy feeling but they are pretty black box in nature. LRFD tries to dispel the black box and approaches things in a probabilistic manner

Traditional Safety Factors weren't arbitrary; they were refined over centuries to provide acceptable probabilistic protection for structures, as LRFD does. The reliability indices in LRFD factors were specifically calibrated to match those of Safety Factors, to maintain the same general level of probabalistic reliability. LRFD simply builds upon the safety factor approach, tweaking it to account for slight differences in sub-factors. Galambos:

The fundamental difference between LRFD and the
allowable stress design method is, then, that the latter
employs one factor (i.e., the Factor of Safety), while the
former uses one factor with the resistance and one factor
each for the different load effect types. LRFD, by
employing more factors, recognizes the fact that, for example, beam theory is more accurate than column theory
(e.g., in Ref. 1,0 = 0.85 for beams and 0 = 0.75 for columns), or that the uncertainties of the dead load are smaller
than those of the live load (e.g., in Ref. 2, yD = 1.2 and 7^
= 1.6). LRFD thus has the potential of providing more
consistency, simply because it uses more than one factor​

At the end of the day, safety factors and LRFD factors provide a buffer between load and resistance, and this has always covered a certain level of design imprecision, even if this isn't often explicitly stated.

 

[DTS419]

At the end of the day, safety factors and LRFD factors provide a buffer between load and resistance, and this has always covered a certain level of design imprecision, even if this isn't often explicitly stated.

Well said. It’s really just common sense.

 

[human909]

Way to take one word and run with it Tomfh. I said often arbitrary, not always. And you completely missed the point about it being a black box.

Traditional Safety Factors weren't arbitrary; they were refined over centuries to provide acceptable probabilistic protection for structures, as LRFD does.

Yes safety factors are so nice and refined. That they are often nice whole digits, clean and easy. That isn't probability. That is "Yep 4x seem to work, good enough for me."

And centuries? Most of the modern materials and methods haven't been around for centuries. There was far more variability in material output 150 years ago than there is now.

The reliability indices in LRFD factors were specifically calibrated to match those of Safety Factors

We are beyond that now. LRFD is the basis of academic theory and in most countries the only design choice.

 

[Tomfh]

Safety factors weren't arbitrary or a "black box"; they were developed over time to provide rational and acceptable level of structural safety. The factors of safety (or allowable stress) provided relatively consistent structural reliability, and this level of reliability eventually formed the basis of the reliability targets used to calibrate the newer LRFD methods we use today. Essentially, we're still using those safety factors, just refined with additional adjustments. That's why, when designing a steel beam, you multiply dead load by ~1.2, live load by ~1.5, and steel yield stress by ~0.9. These numbers maintain the same level of reliability (B=3.0) that the safety factor method did.

 

[human909]

That's why, when designing a steel beam, you multiply dead load by ~1.2, live load by ~1.5, and steel yield stress by ~0.9. These numbers maintain the same level of reliability (B=3.0) that the safety factor method did.

That emphasises my point. It is a black box. If you multiply those numbers together and get 3 then you can't disentangle those numbers.

And lets not forget that the ratio of dead load to live load varies considerably. So you example is hardly consistent.

 

[Euler07]

I'm not an expert in material/prototype testing, but I have worked on a few projects where prototypes were tested and the capacity factor determined. In these calculations there is a minimum reliability index that needs to be achieved and in doing so both the material capacity factor and load safety factor directly relate to the reliability index. See attached for an example (this is old and hasn't been checked so it is an example only). The example is for Australia using the NCC.

In this case, a safety factor is chosen (eg. 1.2 or 1.5 etc), then the capacity factor is adjusted until an appropriate reliability index is achieved. If you want you can go in reverse and decide a capacity factor, then calculate the appropriate safety factor. In either case, the safety factor and capacity factor are directly related in order to determine the reliability index of a given material/construction.

 

[human909]

Thanks Euler for a good example. I believe my comment safety factors are often arbitrary has been the trigger, of the angst.

I haven't tried to suggest that they are always useless, or not determined with some or even plenty of rigour. I use them myself on some of the engineering I do.

But the point of LRFD is to disentangle the probabilistic behaviour of loads and materials because these change and the resultant multiple required to achieve the same outcome changes...

Maybe that could have been more clearly worded in my previous posts.

 

[EngDM]

Oh how far we have strayed from the original post question