inquiry in ETAB program 1

[swelm]

I have question in this program


1- when i make design for building by hand and check the result by e tab i found big differet in the value for the momenet , shear and deflection for the beams

so do you think to use the value from the program and use it in my design

 

[swivel63]

that'll always happen. in any program you use.

depends on the use of the beam. if it's for framing only, i'd just use hand calcs. if it's essential to the lateral force system, i'd use the etabs and coupled with effects from the hand calcs.

 

[271828]

I'll stray from swivel63 a little here.

If you can't come pretty close to the program's output, then you have no idea what the program is actually doing, so it would be irresponsible to use whatever it spits at you.

Of course "pretty close" depends on what type of problem you're looking at.

If it's a static analysis of a framed structure, then you should be able to get pretty close to ETABS' answer. Make simplifying assumptions, use the portal method, cross-check with another program. You should be able to get within a few percent.

As a harder example, I was looking at a nasty vibration problem the other day using SAP. Even so, I was able to perform a manual calc with many simplifying assumptions and get within about 30% of SAP's answer.

 

[swivel63]

like i said, it depends on usage....i'd hardly spin my wheels on the forces for a beam framing a door opening in a stair well (just columns and beams) the way i would for a beam framing in a shearwall at the base floor (opening only at the ground).

 

[civilperson]

I agree with 271828, hand calcs that do not agree with software results make the software unusable! The GIGO rule and the fact that a simplified analysis is easily calculated makes it imperative that you can get identical results in modeled problems.

 

[271828]

Agreed, but it sounds like he has a framed structure and is not getting anywhere near what ETABS is spitting back at him.

I'm envisioning a moment frame where he calcs a D+L moment of wL^2/10 (or whatever) and is getting 3x larger or smaller. There's no way it should be very different.

I'm not talking about checking each and every beam, just doing enough to make sure that the program is working correctly and that the model is correct.

I have a great example. I was working on another vibration problem the other day and arrogantly didn't do a manual calc. My frequency was 4 Hz. The next day, when I was going over it with someone, he said it seemed high to him. I went back and did a very simplified manual calc and got 1.7 Hz. Sure enough, I'd computed line masses, but had forgotten to apply them at 2:00am. The answer was about 1.3 Hz, 30% off from my manual calc.

Manual calcs to verify overall behavior and model accuracy are absolutely mandatory, in my opinion.

 

[271828]

swelm, one idea to start with is to create problems for which you can get the exact solution. If you can't get a match within 1%, then you have a problem. Make the problem extremely easy, a simply-supported beam with a uniform load for example. Be sure to turn off shear deformations if you want an exact match with deflection equations such as those in the AISC Manual.

If necessary, work your way up through a few problems of increasing complexity, but still having exact solutions.

If you can get that to work, then try to figure out what's wrong with your model.

 

[swelm]

I WANT TO THANK ALL OF YOU FOR THIS VALUABLE INFORMATION

 

[ash060]

swelm

Are you using shells to model the floor slab? If so how fine of mesh are you using?

ETABS will place a point load at each node of the shell element. If your mesh is not very fine the program will place only a couple of point loads along the length of the beam, and it will give you bad answers.

You might want to try making the mesh finer near the beams or using the one-way distribution check box that ETABS has when defining floor slab objects

 

[UCFKnights]

I think this has to do with the fact the ETABS is a finite element program. You hand calcs probably assume that you have a pin - pin condition at the ends. Therefore as far as your hand calcs are considered your vertical deflection at both ends of the beam is zero. As far as Etabs is concerned the elements supporting your beam will vertically deflect. (I'm assuming that you have girders and not columns supporting the beams.) This condition is more like a beam supported by springs at the ends (k1 and k2). Both of the supporting girders will probably deflect differently. This difference will induce moments that your hand calcs will not account for.

 

[271828]

UCFKnights, I agree that this might be the case.

I don't want to sound disagreeable, but I think we should be clear here that he can just assume this is the cause and trust what ETABS is spitting back at him.

He should be able to do _something_ to make sure he knows for sure what ETABS is doing. Even if it is a FE program, he should be able to modify that model or create other ones that can be checked by hand.

He also didn't specify _how_ different the results were. I don't care if it is a FEA program, if it's off by 50%, then it's wrong--either the input or understanding of what the program is doing.

Anyway, sorry for sounding disagreeable, but I think this is an important point.

 

[ryansteel]

very valuable Information out-here thanks for the posts.... surley cannot directly compare hand calcs with ETABS design. Specially that Etabs has lots of extra features in it(compared to other FEA softwares). the object in a model is "PART OF THE MODEL", each and every object in that model interacts.


I would just like to aks this question... Is the END_LENGTH_OFFSET zone very rigid if set to automatic?

thanks

 

[ryansteel]

if you'll be doing(modeling) an lone object(i.e. beam), it should be same result with hand calc.

 

[swelm]

thank you for all engineer and for this valuable information

my question is

what can i do for the module to get the same reult with hand calc?

the question by anther way

do you have any facilties in this programm to get the same result with hand calc?

 

[271828]

swelm, create a simpler model. For example create a single beam and put a uniform load or point load on it. Disable shear deformations, give it a simple support on one end and a roller on the other and you should be able to duplicate the answer. Enable shear def and the shear and moment should be the same, but the deflection should change slightly.

If that works, then try one a step above that, like create a single bay with 4 corners, with beams around the perimeter and one down the middle, all simple supports. Make it a lot longer in the beam direction than the girder direction (like 10x) to minimize 2-way shell bending. Now check the middle beam. You should still get pretty close. Keep in mind that the load will be as if the shells are continuous over the middle beam, so the middle beam gets 1.25 its normal trib width load. The edge beams will get 0.375 instead of 0.5.

Work your way up to your real bldg in 2-3 models.

You can't expect to exactly duplicate ETABS results for a complex model, but you should still be able to get fairly close for a framed structure.

As an aside, some of the reply posts in this thread really scare me, seeming to indicating willingness to just accept the fact that it's FEA so can't be verified, but still use those mystery results in design!? Perhaps I'm reading them wrong.

 

[ryansteel]

Very Well said... It can't be more clearer than the above..

Mabuhay ka!

 

[UCFKnights]

I agree that it would be irresponsible to blindly accept results from any program, especially Etabs. There's nothing worse than being burned by a program because you didn't understand what it was doing. My first post assumed that all the input was done correctly. My experience is that if you use one of the simpler models mentioned by 271828 the answers should be very close to your hand calcs. But if you have a fully completed complex multistory model (for example a 15 story model). You cannot blindly pick a beam (supported by girders) from the 10th story and expect that beam to match your hand calcs (because of the reasons I mentioned above). Just look at the deflection of the beam given by ETABS and you will see that the deflections at the end of that beam are not zero. Therefore the results from the program will not match your "pin-pin" (wl^2)/8 ---- 5wl^4/384EI handcalc assumptions.

 

[271828]

ETABS results for a real bldg surely won't match simply-supported calcs. The question is: can we come up with estimates? For a framed structure, we absolutely can.

Pick an interior beam in a moment frame, gravity load combo, and the negative moment should be somewhere in the neighborhood of wL^2/10. Could it be 1/12 or 1/14--maybe, but it's not wL^2/4 or wL^2/40.

We can come up with examples like this all day for most types of problems.

 

[271828]

Sorry UFCKnights, for perhaps seeming abrasive with my last reply--not trying to say that you don't use programs responsibly. I have no clue what level of QA you do.

I just don't think it's possible to overstate the importance of being 100.00% sure of what the program is doing.

Mis-use of these programs usually comes from incorrectly assuming that the program is right OR incorrectly assuming that one knows how to use the program. If one can't get pretty close using hand calcs, then the program (or input) must be assumed wrong.

Mark my word--there will be a major collapse someday (if there hasn't been already) attributed to somebody running the program, not really knowing what it's doing, and drastically under-designing something.

 

[hokie66]

It happened in (I think) 1978 with the collapse of the Hartford Coliseum space frame roof. A classic case of 271828's point. Worth looking it up.