design accuracy 3

[brinkley]

I am planning on modeling an existing structure via staad|pro.

Knowing that it's a model and not the construction documents, is it acceptable to round your models dimension up to the next whole inch or even foot in some cases? I know that most FEA software's out there have the ability to go to that precision, and I've done so in the past.

Do any of you round the dimensions of your model to save time?

 

[COEngineeer]

I model everything according to my engineering scale ruler on the plan. I dont dimension it from CAD. I am sure it is only accurate to 3 inches.

 

[whyun]

3 inch increment is a good estimate. Set the unit to feet and use the closest .00 .25 .50 .75 value.

 

[MikeHalloran]

Rounding dimensions when building a CAD model does NOT save time.



Mike Halloran
Pembroke Pines, FL, USA

 

[IRstuff]

>> Given dimensions with already limited accuracy, why possible justification could you have to further reduce the accuracy?

>> This is compounded by your erroneous assumption that reducing precision speeds up the program. It does not. That's because the program cranks through the same 32-bit or 64-bit floating point calculations, regardless of what the actual value is.

TTFN

 

[MikeHalloran]

It's worse than that.

Even customary rounding, e.g. using .63 instead of 5/8, adds up throughout the building process, and it doesn't average to zero. You end up with corners that look closed but aren't, lines that appear to touch end to end but really overlap, stuff like that. CAD programs get confused when presented with discontinuous geometry like that, and you'll waste so much time tracking down those little errors that any gains from using fewer keystrokes at input are more than consumed.



Mike Halloran
Pembroke Pines, FL, USA

 

[crossframe]

We went to the moon and back with three significant figures...

 

[Dinosaur]

Assuming you are modeling a building in a FEA program, there is no standard of practice I am aware of that would require you to base your calcs on floor heights that were "exact". If the floors were 13.79 feet high, no one could say the structure was underdesigned if you used 13.8 or even 14 feet. On the other hand, a particularly tall building could be modeled more accurately by giving the first and last nodes and then auto generating the others. The computer does all the hard work.

One quick comment for the novice engineers out there though, do not mistake precision for accuracy. Precision is the number of digits you carry through your calculations. Accuracy is the degree of correctness of an answer or process. Many of our loadings are only accurate to two figures although most engineers I work with strive for three figures. When the loadings are only accurate to three figures (at best), don't waste resources preparing calculations with 5 significant figures. You are fooling yourself.

 

[COEngineeer]

I thought he was talking about modeling for structural analysis. I am sure 3 inches doesnt make a big difference espcially with all the Factor of Safety we consider,

 

[whyun]

Sane architects provide grid systems with bay lengths in even round numbers such as 30'-0", 28'-0" and so on. Where you have somewhat "odd" bay widths like 16'-5" or 16'-4 1/2", it makes no significant difference in the output if you simplify the dimension to 16.5 feet.

Structural engineering is not "rocket science". I agree with COEngineeEr.

 

[brinkley]

And that's the point I'm shooting for. I see some of the reply's are gearing toward the CAD side, but i'm focusing on the modeling side of the analysis for only a (3) story building. I know that a high rise building is a hole different animal.

 

[lkjh345]

FWIW:

I do a lot of 2 to 5 story buildings. I round everything off to the next greater even foot dimension in my calculations and computer analysis models.

We hardly know the loading withing an order of magnitude, so extreme precision in the length of meembers is complete overwehlmed by the inaccuracies inherent in our loading assumptions.

 

[RichBridge]

I'll Add my 2 cents.

It might make more of an impact on how it is rounded...meaning with some experience, one can round up or down depending on what dimension is being dealt with.

For example: for finding a moment in a beam, round the span up and round the loading up. For the structural depth of a slab design, round the dimension down so slightly more steel is required. Knowing which way to round for each situation is always helpful.

I also agree that for CAD, don't round. It will just cause problems down the road.

 

[RareBugTX]

My 2 cents:

I never draw or model anything, call it AutoCAD, SAP, Staadt etc to fractions of an inch. I just imagine the iron workers at the job site reading 13'-1-5/16" from a framing plan..... it has got to be hilarious.

I have found very sane to work with round numbers and agree to stay to the quarters of a foot numbers, when modeling. 12.25', 12.50,12.75' etc.

Also remember, most of us deal with buildings up to three stories high. 93% of construction is of that size.

Regards

 

[dik]

Just because you can model stuff and get answers to 10 decimal places, it doens't mean the 'real' answer is that correct. You should treat any output as being an approximate of the true answer.

When I program, for example, I may have pi defined to 17 decimal places only because the programming language defaults to a 'double' variable and there is no overhead in doing so. The answer is not likely any more correct than if I had used pi defined to 5 decimal places.

And with CAD, I draw stuff to the 'exact' dimension just to prevent closure issues, and although I usually set the precision to 1/64, I realize that on site, the precision may not be any better than 1/8...

Dik

 

[SlideRuleEra]

brinkley - I understand that you are modeling a three story building. Let me give you a hypothetical, but realistic example of what could happen on a larger, more complicated structure.

Say that the column line locations are all established by dimensions from a common base line, such as this:

Because it is a large structure, you round off to the closest foot:
83'7" becomes 84'
84'5" becomes 84'
Your model has just merged two distinct column lines. In industrial buildings it is common to have unusually spaced column lines in different parts of the structure - what I just described can really happen.

You can do what you want, and I know that the above scenario is unlikely on your current project. However, I suggest that you use good practice now to avoid "unintended consequences" later.

http://www.SlideRuleEra.net

 

[grandcasita]

Whenever I build a computer model while neglecting any available precision, at some point, I wish I had started with precise numbers for one reason of another.

If I start with the best precision available, I always end up more satisfied, and usually with less rework, often for the reason to which SlideRuleEra points.

 

[JLNJ]

Use your engineering judgement. That's why we get the big bucks.

 

[JEmH]

RareBug: I once had the occaision to be on site while the carpenters built the forms for an odd shaped concrete member. I had dimensioned one corner angle to the exact odd minutes and seconds.

Carpenter holding plans: Okay, turn that 43 degrees 18 mintues and 43 seconds

Carpenter holding speed square: Alright, 45 degrees. Shoot it.

It was hilarious for everyone but me.

 

[DaveAtkins]

I was taught in college to round everything to three significant figures, and that is what I do.

DaveAtkins