Continue to Site

Eng-Tips is the largest engineering community on the Internet

Intelligent Work Forums for Engineering Professionals

Register Log in
  • Congratulations waross on being selected by the Eng-Tips community for having the most helpful posts in the forums last week. Way to Go!

Computer output shows value of real-world judgement 4

Status
Not open for further replies.
youngstructural

youngstructural

Structural
Aug 17, 2004
713
So I was looking at a plant deck for an existing building, and in designing some of it using software I have found quite the humorous computer output. Warning: Probably not funny to anyone other than a Structural Engineer, and that might even be a stretch!

Not to restart the "when is overstress okay" debate, but have a look at the screen capture; Funny what a computer considers a "failure"...

I thought it would be useful to mention, as it reinforces yet again the value of engineering judgement.

Cheers,

YS

B.Eng (Carleton)
Working in New Zealand, thinking of my snow covered home...
 
Some hints on how to read the program's output without studying it very long?
 
YS,

The results part doesn't scroll. I imagine that is where the funny bit is located.
 
That is pretty funny. Is there no way to define a "max overstress" or something similar? Some programs would let one set it to say "OK" if the ratio isn't over 1.03 or whatever.

What I'd like to see if a Ramsteel-designed framing plan with insane sizes due to one button being pushed wrong somewhere. I worked on a job like this years ago: the poor guy had pressed the "Shored" button a couple of menu levels deep and it was an unshored composite beams. Those W12x14, 34' long did't stand a chance when they started putting concrete on there. All of the main beams and girders were undersized by at least 2x and the errors got by everybody working on the job -- designer, cad guy, detailer, shop drawing checker, fabricator, erector...didn't get by Mr. Gravity, though, LOL. I need to get a copy of that plan to show my students to make an impression about mindless use of programs.
 
I think its funniest you were listening to Ashley MacIssac at the time. You really must be missing your snow covered home.
 
Don,

I think it was 1.0000001. Thus NG.
 
Yes, this program is similar in its report. What amused YS was that the report said 1.00, but yet it said no good.
 
Yes, I found that amusing as well, which led to my rounding theory. I tend to play with the numbers in software to see how the software reacts when things are close - like deflection.

Don Phillips
 
Some software spits out all types of warnings. Knowing when to heed or ignore some of the warnings takes a lot of experience and judgment, both with the specific software and the structure. Ignoring the warnings without documenting why can be a trap.
 
Program Output looks like Space Gass:
I am sure this is a case of rounding and display format precision. I am sure if you modify your default output format, you will find that the computed result is something like 0.996 (which is displayed as 1.00 when formatted to two decimal places), but is definitely less than 1 exactly, and so is a "Fail", according to the user-defined test limits.

You have to remember that computers are dumb but obedient servants - if you say you want to test against a pass / fail criterion of greater than or equal to 1.00 exactly is a "Pass", less than 1.00 exactly is a "Fail", then that is what it will do.

If you are prepared to accept a little bit of "overload", you can specify this when you run your Steel Design Check - instead of setting the "Load Factor Limit" equal to 1.00 (which is the default), you could change this to 0.95 say (YOU have to tell the computer what YOU consider to be "a little bit"!), but you would face the same issues then when it reported a Load Factor of 0.951 for one member (which displays as 0.95) as a "Pass", but another member with a load factor of 0.949 (which also displays as 0.95) would be a "Fail". YOU have to draw the line somewhere, and wherever YOU draw the line, the computer will faithfully tell you which side it calculates you are standing on!

Hope this helps!
 
Thanks, Julian. I stand corrected.
 
Yes, it's just a rounding quirk. The computer is evaluating the internal value not the rounded output.
 
Okay all; I had thought that rounding being the issue was obvious... I would be very keen to hear any differing opinion!

And of course Julian is right; The screen display is not the value (proper) that the machine is considering. I hope the issue of machine accuracy and the fact that this number is by nature flawed and inaccurate doesn't need to be pointed out!

I just thought it was funny, and something to keep in mind. The blind acceptance of an output as "failed" or even as "okay" is irresponsible from a Professional point of view in any case, however I did think it was worth posting, even just for a chuckle.

Regards,

YS

B.Eng (Carleton)
Working in New Zealand, thinking of my snow covered home...
 
An excellent point, DonPhillips!

Our computers and calculators enable us to calculate results to 4, 8, 12 or even more decimal places. However, as engineers, we should never forget that we often only know the input parameters to one decimal place accuracy (or less, sometimes!)

Back in my undergraduate years, I had one lecturer who used to deduct marks for every decimal place of of precision you gave in your final solution greater than the lowest precision of the supplied input data. It was a lesson well-learned.

For example, consider the coefficient of internal pressure in a building due to external wind loading. Your design code may indicate a value of +/- 0.3 say. (Think about what this means - a change of just +/- 0.1 - the implied accuracy of the provided figure - would yield a change in total load effect of +/- 33%!) Your wind loading program may tell you the total uplift load is 773.45 kN, say. By all means, carry this precision through in the subsequent calculations, but if someone asks you, all you REALLY know is that the total uplift is about 800 kN - not 770, not 773, and definitely not 773.5!

In some cases, 0.7 = 1 = 1.3 - to the required engineering accuracy!
 
That is an interesting point, however significant figures are not the cause here; And if you extended this argument to the logical conclusion, nothing we do in Metric design would be any more accurate than two digits. I prefer four figure engineering accuracy and continual rounding during steps in calculations.

That said, it's not just the accuracy of the input, but also the floating-point number error creep. You cannot expect laser accuracy from computers, any more than you can expect engineering judgement. The computer cannot multiply, divide, or do any of the higher mathematical operations on decimal figures without automatically causing errors. The problem is that everything is stored in "pseudo" numbers, as actually represented by a string of base two numbers.

Errors occur in the most routine of computer calculations, and only by the grace of random error cancellation and rounding do the answers come out "okay". There are some very famous examples, including a Patriot missile which failed to stop a Scud attack in the middle east. If memory serves the error was so bad that the missiles were seriously far apart... But the computer thought they were spot-on!

As soon as you get into Matrix Methods, you get into even more expansive realms of errors, and very smart people who try to make them not matter. But they do, and trusting the black box blindly is the fastest way to get people killed that I can think of in our line of work...

Cheers,

YS

B.Eng (Carleton)
Working in New Zealand, thinking of my snow covered home...
 
Status
Not open for further replies.

Similar threads

MyCupboard
  • Question
Value in post-design inspection work?
Replies
16
Views
3K
XR250
XR250
southard2
  • Locked
  • Question
Floor Vibration Analysis - Engineering Judgement
Replies
9
Views
800
bones206
bones206
Henry94606
  • Locked
  • Question
Building tiedown negative value
Replies
4
Views
2K
driftLimiter
driftLimiter
bIockhead
  • Question
Hem Fir (N) Design Value Reductions
Replies
3
Views
3K
ANE91
ANE91
Martino8
  • Locked
  • Question
The world of AASHTO
Replies
9
Views
449
JoelTXCive
JoelTXCive

Part and Inventory Search

Sponsor

Back
Top Bottom