Significant Figures & and code interpretation 5

[Robvh1984]

I am a stickler to sound logic, e.g. significant figures, etc. However, My question arises from maybe a more philosophical standpoint (or poorly assumed to be). I see structural engineering as having 4 main components where variability may or will arise. 1.Loading 2.Engineering Analysis 3. Construction 4. Material Strength/general behavior.

All else equal, as engineers we have most control over the 1st and 2nd part of those 4 items as loadings may vary greatly with application (requiring our judgement) and analysis is literally performed by us. Therefore, as a young engineer, I prefer to use the code verbatim (i.e., if f'c=3000 I take it to mean that f'c has been measured to be precise to (1) significant figure(ignoring scatter of points). Then, for example, if I calculate the rupture strength (7.5f'c^.5) one calculates something on the order of 410.7919181288745850927273371006. I would say that the rupture strength is most nearly 400 psi, and I shall carry as many digits as I can from this intermediate calculation through my remaining calculations. I think my physics/chemistry professors would agree based on the standard mathematical laws or error propagation for arithmetic and single variable functions that this is valid.

However, My boss does not follow this logic when he performs calculations. He crudely & significantly rounds down whenever numbers dont seem to 'feel' right. Then in the end he selects sections with capacities as much as 20% greater than demanded by his rounded calculations. He justifies his methods of computing broadly based upon personal liability, loadings, and construction & material variability.

I find his methods to be unacceptable, as he applies this conservatism very broadly to everything he does. I understand our field is mostly experimental, but I feel that many of us (structural engineers) are so afraid of variability, we fly through analysis with grossly over estimated values and thus structural members.

In, short, how does everybody frame their logic in performing calculations as it relates to precision?



-Robert Miller, E.I.T.
KPA Structural Engineers

 

[Teguci]

Sorry but I have to keep coming back to it. If the analysis doesn't "feel right" there must be a reason. If an experienced engineer doesn't "feel right" ("trust your feelings Luke") that should throw up a red flag. But, the response, except in rare situations, is not to add 20%. The contractor gets to cut with an axe, not us.

Its the projects where I do add the arbitrary safety factor (there I admit it) that make me not sleep well at night. The reason, my feelings tell me I'm overlooking something. Will the 20% cover me? Not if I forgot to add the closely spaced vertical stiffeners in the torsionally sensitive beam (sorry, I had too!).

@IDS - I suppose there could be some non-conservative issues with our building codes. If there are we should discuss and have them corrected. As for materials, I'm not looking for accuracy, I'm looking for being on the safe side of failure. If I'm missing something, a 20% increase in capacity may not put me there.

 

[IDS]

@IDS - I suppose there could be some non-conservative issues with our building codes. If there are we should discuss and have them corrected. As for materials, I'm not looking for accuracy, I'm looking for being on the safe side of failure. If I'm missing something, a 20% increase in capacity may not put me there.

Sure, I'm not saying apply a flat 20% to everything. For instance, with the tensile strength of concrete as mentioned in the OP I'd suggest reducing by at least 50% if you want an upper bound deflection estimate, or better do a proper analysis including shrinkage, creep and differential temperature effects. The point is that just using code values won't always give reliable results, especially with non-manufactured materials.

Doug Jenkins
Interactive Design Services

http://newtonexcelbach.wordpress.com/

 

[Ron]

Perhaps the greatest advantage to increases in computational ability is the perception of a more efficient design. Is that really true? We seem to want to design, at times, to the literal and figurative "knife edge". What is our gain? Is it a savings to our client? Is it a compromise to the health, safety and welfare of the public that we are entrusted to protect through our licensing process?

Saving one's client a few bucks will probably yield more work in the future. That's a self-serving enterprise. Not wrong, just self-serving. But in that process, are we being truly efficient or is it just an exercise in risk management?

Yes, we have progressed in our understanding of materials science as applied to construction materials. We have materials that were not available a few years ago. But in our rush to "betterment", are we neglecting that sometimes performance history is our best measure of science?

Questions to ponder...individually and as a profession.

 

[SAIL3]

From my own experience, Ron hit the nail on the head.
It is often said that engineering is more of an art than a science.
When I started out in engineering I had a very narrow view of my role in design. It basically boiled down to beating the last oz. of weight out of the structure even if that meant numerous different sizes of members and footings etc.
Over time one learns that there are many other factors that come into play...as mentioned in the above posts ie.
economies of scale, repeatition, simplicity, errors, future changes etc.
Schedules are becoming more compressed leaving less time for engineering analysis and and the vetting process required for this and thus increasing the possibilty of errors.
What concerns me is the increasing complexity of the codes and the time involved chasing down each nit-picking detail. To me this is a dangerous distraction from what is the core importance of my role and that is engineering analysis.
The 10, 15 or 20% overdesign(on paper) takes the wind out of the sails of the idea of beavering the last oz. of weight out of a design performed in splendid isolation.
What is lost in all of this is the crude approximation of the input loads and the load factors ie, 1.4 why not 1.35?, 1.6 why not 1.55? etc. Taking these input loads as gospel and designing a strutcure to evermore refinement is to me misguided.
Progression thru experience:
4 significant figures
3 significant figures
2 significant figures
1 significant figure
5% overdesign
10% overdesign
15% overdesign.....
replaced by computer!!

 

[Teguci]

As an aside:
A young foolish engineer would criticize his boss publicly and sign his name to it.
A devious young engineer would criticize his boss and sign someone he doesn't like's name to it.
An older devious engineer would criticize the short comings of his office and sign someone else's company's name to it.

A wise engineer, having tried the usual methods, would post anonymously (with Nom de Plume), the shortcomings of his boss on a blog he knows his boss uses in a way that the boss will recognize his shortcomings and correct the supposed error but not put his defenses up.

I think we are all wondering which one of these (or other) are you Rob?

 

[steellion]

One of my pet peeves is reviewing calcs with too many sig figs. Don't tell me that the beam moment is 435.2416 ft-k when that number was derived from a load that was a guess, load factors that are arbitrary, material strengths that are a guess and generally quite conservative, etc...

I try to keep intermediate calculated numbers intact as best as I can. It's easy to maintain 17 decimal places when performing calcs on a calculator or Excel, but I would never write down that many sig figs on the calculation pad. If you keep the intermediate numbers intact, then when you get to the final line of Load vs Capacity, you know how close you are to the line. As mentioned in posts above, there are lots of good reasons to be conservative, but I don't know how anyone can determine how conservative they are being if they round every intermediate number along the way.

Rob, I will spare you a lecture but it seems you could use a little more maturity. Instead of being critical of your boss, ask why he is making those decisions and you'll learn something. Least material is not always least cost.

 

[IRstuff]

"but I would never write down that many sig figs on the calculation pad. "

I think that's a serious issue in itself. One ought not be using a tool that requires you to "write down" anything. There are potentially more errors from transcribing numbers or reversing digits than anything else I've read in this thread.

I almost never have a calculation or result that's hand written. The only time that happens is if I'm without my laptop, or I'm caught in the hallway, on the way to the bathroom. Almost everything I do is in Mathcad, which allows you to set the displayable precision to a reasonable amount, AND it requires NO unit conversions ever, for any unit in its database; adding 20 inches plus 5 meters plus 2 furlongs is trivial.

TTFN

FAQ731-376
Chinese prisoner wins Nobel Peace Prize

 

[DWHA]

I feel that whenever possible rounding within an analysis is not the best approach. I perfer to stay consistant, without rounding throughout the analysis. I like 2 sig figs. Then in the end, I will up the beam size, reinforcing, # bolts, etc as I feel compelled. I like to do it this way because if there is ever a situaion where something happens in construction, I can look at the analysis and know how much margin there is. Rather than saying, well I made a conservative round here, and another one here; not knowing how much margin I actually have.

 

[TehMightyEngineer]

I know it's been said before but as a young engineer I too want to "weigh in" on this as well based on my limited design experience.

I've found that it ultimately boils down to two things we must consider as engineers; our responsibilities and our practicalities.

We are responsible for public safety and the economics of the project (both our own economics and the economics of the client). What I mean is that we must "ensure" (statistically) that the structures/devices we design must not "fail" so as to cause excessive economic harm (expensive repairs) or detrimental public safety (loss of life and so on). Practicality comes into play in all the things mentioned above (assumed loads, variables, errors, constructability, time constraints, etc.). As engineers we must balance these two.

So, when I design something I evaluate these two considerations. For a "simple" project with well defined loads, limited time constraints, limited harm if it fails, I would be precise in my calculations and select the smallest size member(s) that meet the minimum design requirement. This satisfies both requirements as it meets my requirement of responsibility as an engineer while making sure that it results in the most money saved for both my own company and the client.

However, in a complex project the lines become fuzzier and I will likely choose to "round up" a lot of things. The loads might be not clearly defined; round them up. The harm if it fails might be high; round up the safety factors. The client and my firm might save more money if I can get him a slightly more expensive design sooner, round things up. The cost of going back and fixing a small error that I made might be more than the cost of the extra material and/or labor, round things up. And so on.

I'd especially like to point out the last thing. Personal errors are my biggest pet peeve. I will never hesitate to point out my own errors and accept the consequences. However, it just looks bad to myself if I was at fault. Even if the design was ultimately still "safe" (well within a safety factor), I feel it reflects poorly on myself if I keep having to revise myself. How can anyone trust me as an engineer if I keep coming back to them saying "I'm sorry, that design wasn't safe, but this one is. Trust me, I got it right this time." That just doesn't sound good in my head.

So, for that reason I will always round numbers in a conservative fashion when it seems appropriate. As long as I "feel" that I've satisfied my responsibilities as an engineer while making sure to keep costs down (remember, we're here to make money too) then I'm satisfied. Researchers are the only people who need to truly care about significant figures in my opinion.

P.S. Rob, I'm hoping that you're showing that you have a lot of spunk signing your post while you criticize your boss. If you really do have the confidence to accept responsibility for your statements in a public forum then I applaud that, too many people hide behind the anonymity of the internet these days. However, if that was unintentional then I really hope you consider the repercussions of this and think more carefully before you speak in the future.

EIT with BS in Civil/Structural engineering.

 

[Ron]

Personal errors are my biggest pet peeve.

Would that include transposing the "e" and "h" in "The" in your handle, or does "Teh" actually mean something?....or have I made an assumption that is not based on authoritative evidence?

Just pullin' your chain a little to point out the human factor in all of us.

 

[TehMightyEngineer]

Heh, "Teh" is a product of my "internet" generation. I kept it more as a relic of my past and as reminder of what the internet lets us get away with these days.

I was wondering how long it would take before my usual internet alias drew a comment. You would laugh at the back story behind my "mighty pirate".

I suppose that's actually a reasonable point which is semi-relevant to the topic. A third requirement to engineering design would be to enjoy it. Sweating over the difference between 4.92185 and 5.0 just takes all the fun out of a nice design, just call it 5.0 and enjoy it. Leave the 4.92185 to the mathematicians and statisticians out there.

EIT with BS in Civil/Structural engineering.

 

[TehMightyEngineer]

Further reading on "teh":

http://en.wikipedia.org/wiki/Teh

(Back on topic now, forgive me).

EIT with BS in Civil/Structural engineering.

 

[vmirat]

One of the things that I've never been able to adopt is the use of kips. For some odd reason, I just felt more comfortable using the whole number. I would be curious to know who started the whole kip thing anyway. I'm thinking that this goes back to using slide rules which had sigfig limits built in (SlideRuleEra could probably speak to that!).

 

[TehMightyEngineer]

Really? I almost always use kips as most loads involve many thousands of pounds. While it's obviously situational I think the only thing I regularly prefer to be in pounds is stresses (psi vs ksi).

I imagine that kips came into favor about the same time the metric starting showing up but perhaps it was a slide-rule thing. A brief internet search on kips yields nothing of real note.

EIT with BS in Civil/Structural engineering.

 

[vmirat]

I suppose my practice of not using kips is a result of the type of projects I design, which are smaller one-story buildings and odd structures. In fact, I do most of my designs by hand. I will occasionally use an old 1998 version of STAAD III 2D for frame analysis or weird indeterminate members. So, I don't usually deal in big numbers.

I admit it's odd. I think it stems from my difficulties in college with figuring out significant figures and getting in trouble using kips in one place and whole numbers in another place, resulting in the wrong answer. I found it easier to just use whole numbers for everything versus keeping track of where I used kips and where I didn't.

 

[dik]

It's been my experience that the economy is not generally a 'tight design', but the structural system selected in the first place. Even with tendered documents, there can often be a scatter of 20% in the tender prices.

I have little confidence in significant places.

Dik

 

[SlideRuleEra]

vmirat - I hadn't though of "kips" being a more or less obsolete term... but you are probably right. When using a slide rule putting the loads in kips made sense, but not because of the significant digit limitations. IMHO, a more likely reason is the OTHER inherent limitation of a slide rule: no decimal point - you had to do that yourself (in your head or on paper). By using kips the numbers (for many calcs) had a reasonable number of intergers and the big advantage was that you could eliminate three decimal points (10^3) from your calculations. That does not seem like much of a problem by today's standards... but things were different then.

http://www.SlideRuleEra.net

http://www.VacuumTubeEra.net

 

[vmirat]

SlideRuleEra,

The hidden meaning behind my post was that most engineers are taught to do calcs in kips, so there is already a built-in sigfig factor.

It's interesting how we continue certain practices as a carry-over from antiquated design, such as the spacing of railroad rails (Snopes.com has an interesting article about that story).

 

[sasa2k]

In structural engineering 3 significant figures are more than sufficient for almost all type of calculations and to carry the precision throughout calculations. Everything more than that is practically waste of time and money. Actually, it reveals inexperience (better case) or ignorance (worse).

Structural engineering involves understanding some of the basic assumptions behind the statics, strength of materials, statistics, etc. Based on these assumptions it is apparent that our calculations are very approximate in nature. Increased number of significant figures will actually give a false sense of accuracy and may be misleading.

Now, to be a bit philosophical - if you already use 17-digit precision - why stop there, why don't you use 20, 30, or 10000 digits?

 

[IRstuff]

"if you already use 17-digit precision - why stop there, why don't you use 20, 30, or 10000 digits? "

Because that would actually be more, or quite, difficult. If you are using Excel or Mathcad or TKSolver, or even a programmable calculator, why would you waste your time truncating intermediate calculation results to 3 sigfigs? For that matter, why even use a calculator when you can quite easily restrict yourself to 3 sigfigs by using a sliderule?

"Everything more than that is practically waste of time and money. "

The cost in time and money is exactly zero, unless you're using pencil and paper and using a sliderule. Even the lowliest calculator can do it, and if you have your calculation routines programmed in, there is no need to transcribe anything, and there is no cost in maintaining the precision. And, no one is suggesting that one litters one's worksheets with 17 digit numbers, if for no other reason than it being visually cluttered. All my worksheets are set for 3 decimal place displayed precision. The inner workings of the math package are completely hidden and transparent to the user.

There are plenty of structual engineers using Mathcad, which does all the unit conversions on the fly, kips or otherwise, documents the calculations, and is fully repeatable and reusable, time and again. The worksheets can be neatly arranged and are suitable for customer documentation, without any transcription or typing, or whatever. And, because you can do calculations directly with the units, unit and dimensionality checking is automatically done.

TTFN

FAQ731-376
Chinese prisoner wins Nobel Peace Prize