# Computer model vs actual test

Hi,
I have a problem!

I don't have any formal training in engineering, but have been building and testing all sorts of spanning methods over the past 10 or so years, with some earlier professional help.

My main interest at present is in open web trusses/bridges. I have developed a double web/three chord planar truss system that I am presently testing... The main top chord of the truss is only 25x25x2.5 (1"x1") with a span (17m radius arch) of 27.2m (90 ft).

I have had a couple of engineers do a simple analysis for me, but their' analyses do not seem to take the arch itself into account... (where I am suggesting the extra strength is coming from due to the control of horizontal deflection by using two sets of diagonals instead of the conventional one set).

The computer models so far suggest that the arch would have trouble even supporting its own self weight... and at the most 1 ton (1 small car)... So far however, I have managed 1.2 tonnes (4x300kg) before one of my footings buckled (an oversight).

Will I have to do an actual test and match it with a computer model to get a true reading before the computer model can be used to design larger spans?

I am suggesting that the truss will handle at least 4 and possibly even up to 10 tonnes... so long as my footings hold up and lateral deflection is fully controlled.

Hope someone here can help.

Dave

chroot
Staff Emeritus
Gold Member
Do a finite-element analysis.

- Warren

Clausius2
Gold Member
Do you mind posting a picture of that?. Maybe I know how to solve the problem analytically (via Navier-Bresse equations), but I don't know many english words of structural engineering, so I'm a bit lost in your literature.

Thanks Clausius2... I'll try to post a picture here but I haven't had much luck in the past as I'm basically computer numb... I have managed to (kind of) set up my own msn site though.

This is what I call the 'ultimate' using 25mm tube... Please be aware that these designs are very hard to imagine as being 'workable' as your first impression will be... NO WAY!
http://groups.msn.com/TrussTesting/trussesandbridges.msnw?action=ShowPhoto&PhotoID=18 [Broken]

I've put a few photos together here to show where I'm up to so far... including any 'blunders'. I can send you a better pic if you want as the site is very limited
http://groups.msn.com/TrussTesting/trussesandbridges.msnw?action=ShowPhoto&PhotoID=28 [Broken]

And this one shows what a computer model predicts... which I have already gone past... Although it has been suggested that you can increase the value of "Y"(?)... But I believe you would still have to do the actual test so that it can be matched with first... before going on to other sizes?
http://groups.msn.com/TrussTesting/trussesandbridges.msnw?action=ShowPhoto&PhotoID=27 [Broken]

See what you think anyway and thanks for your kindness as I will be needing a lot more (paid) help later on. Right now though (as I have often wondered in the earlier stages) I really am not certain that a computer simulation is able to accurately predict what will happen with this type of truss configuration... which would have saved me a bulk amount of work and expense if it were able.

Best regards,
Dave

Last edited by a moderator:
Clausius2
Gold Member
Well Dave, I'm not an expert at this. I have coursed three (semestral) courses on Structural Engineering as a part of my studies in Mechanical Engineering. But this seems to overwhelm my knowledge. The structure is not treatable analytically (maybe it is possible but it will take you a lot of time). The unique thing you can do is some gross estimations and computer it before building one of them.

I have understood somehow your computer model doesn't fit with your experimental testing. While your computer model predicts the breaking of the structure, you have looked experimentally that your structure doesn't collapse, isn't it?

Well, I have dealt with that sort of computer programs (TriDim, Ansys). They are based on the Elasticity Equations (Navier-Bresse) or Energetic Theorems (Castigliano). My advice is take into account the computerized results. Why? Well, although you haven't seen your structure collapsing after loading it, maybe, inside the arched beam there are plastificated zones. If the tension on some part is higher than Von Misses stress (i.e plasticity limit) the computer program is pointing you where is the possible starting of the eventual failure. So that, try to interpret the computer results or claim help to some engineer nearby. Experimentally, you're not seeing what is happening at interior material zones, but computing it you are able.

As far as I know, the validation of this codes are really difficult. I have experience in validating Fluid Mechanics codes, because they are visible and comparable with experimental photographs or velocimetry tests. But a structural code does not predict the failure except that caused by plastification. If you want a serious validation you will have to use extensometric bands, measure the deformations and check it with the computed data applying a factor of correction. Also the deformations provided by your code (I have seen your figure) are exagerated. Take this into account. It is an usual characteristic of the code postprocessor.

Regards.

Javier.

PerennialII
Gold Member
From the looks of it I wouldn't start treating it analytically .... as Clausius2 said is may be possible up to a point, but treatment of plasticity and possible buckling (local) of a structure like yours goes beyond the scope of such methods (you could get a simple estimate by reducing the structure to an arched beam for global displacements). However, using finite element analysis the problem ought to be solvable with "relative ease", even if it were to include advanced features such as plasticity and large deformations (like buckling). I suppose the problem of giving a definite answer about this is that you'd have to have some sort of measurement results to compare the numerical results against, like displacements from positions relevant to the failure mode / where they overall are large & characteristic to the problem. Details of the numerical analysis, like how they've applied boundary conditions and what sort of material and FE formulations they've used would be a somewhat of a necessity in qualifying the result as well.

Thanks Javier and PerennialII.

Hmmm, Sounds like I won't be able to get out of doing a full test... I did wonder about FE, whether it also would take the extra strength of the arch itself into consideration.

I'm sure that with conventional trusses (single web layer) a computer analysis would give a good comparison as they have been continually put to the test and so are very accurate... Not so with this one though until it has initially been matched up properly.

Thanks again, you have both been a great help... Incidently, I just happened to be loading the demo truss up again today. I reached the 1200kg (4x300), (the same as the last time), only this time there was a lot less deflection... 60mm compared to 100mm at one point just off centre... I've finished for the day now and hope to continue again tomorrow.

So far as predictions go (if something else unforseen doesen't delay things) I feel now that the truss is only just 'starting' to do the work... It has been pushing outwards on one end more than the other, due to the off centre loading, but is also starting to creep upwards toward the centre as anticipated...

So now is the 'real' test... whether or not the secondary webbing is controlling the horizontal deflection... This test is not what I feel will be the ultimate as it is only made up of older 'rough' sections (and a bit so so in places), but it should at least prove my original theory and control a big percentage of the buckling... I can still see it handling at least four tonnes.

With thanks,
Dave

By the way... I have a simple design for a tower/mast if anyone here is interested in playing around with it... I believe that it will (should at least) handle near full compression (without buckling). That is the aim anyway... Good for lifting cranes etc'.

With a tower you do not need to worry about if the computer model will compensate for an arch/deflection as there are no arches, so should be an easier test... I believe there are a few free program downloads available on the internet... I tried a couple of times but it got too confusing... values etc'.
Cheers!

PerennialII
Gold Member
Hmmm, Sounds like I won't be able to get out of doing a full test... I did wonder about FE, whether it also would take the extra strength of the arch itself into consideration.
A properly done FE analysis ought to produce really well comparable result to your experiment. Once the actual failure starts to occur (if there is excessive local plasticity & buckling, then coupling of these two as the failure progresses) that begins to cause some trouble for the FE analysis, but up to that point the effects of all structural elements on overall stiffness and load bearing capacity are included.

Thanks again, you have both been a great help... Incidently, I just happened to be loading the demo truss up again today. I reached the 1200kg (4x300), (the same as the last time), only this time there was a lot less deflection... 60mm compared to 100mm at one point just off centre... I've finished for the day now and hope to continue again tomorrow.
I'm starting to feel envious for you carrying out your demo tests ... that's some real work !

So now is the 'real' test... whether or not the secondary webbing is controlling the horizontal deflection... This test is not what I feel will be the ultimate as it is only made up of older 'rough' sections (and a bit so so in places), but it should at least prove my original theory and control a big percentage of the buckling... I can still see it handling at least four tonnes.
At least to my eye the structure looks quite harmonious ... wouldn't be suprised if it fared well.

By the way... I have a simple design for a tower/mast if anyone here is interested in playing around with it... I believe that it will (should at least) handle near full compression (without buckling). That is the aim anyway... Good for lifting cranes etc'.

With a tower you do not need to worry about if the computer model will compensate for an arch/deflection as there are no arches, so should be an easier test... I believe there are a few free program downloads available on the internet... I tried a couple of times but it got too confusing... values etc'.
Cheers!
Sounds promising ... and yeah, there are free software available that'll handle these kinds of beam/rod problems well, in my opinion don't lose much to the commercial alternatives.

I've just returned from testing again... CATASTROPHE!
I left the weights overnight and today checked deflections again where more sag was noted... up to 50mm in places!... I put it down mainly to the rain we had last night (85 points or 20mm) which may have loosened the footings (which were pretty meagre anyway). So I started loading again... reached 150kg, then the whole thing twisted and collapsed...

Where I had the flat sheet over the top chord for lateral stability a where the same 8 ft lengths overlapped/joined, I had them tek screwed together with 4 screws on either side... It appears that the screws tore through the sheet, allowing the truss to push out to the side and buckle... The lateral was always a concern, but I overlooked the joins themselves... Too bad!... If there had been no trouble I fully expected to at least load the drums right up... which would have been convincing enough... at least 4 tonnes.

At least to my eye the structure looks quite harmonious ... wouldn't be suprised if it fared well.
Next time perhaps?... At this stage I am contemplating either building another truss (the 'ultimate' design I stated) a tower, or perhaps even a trebuchet...

By building another truss, I can reconnect the older pieces to use them as the rear of two parrallel trusses with battens and sheeting (a bit more secure with the lateral stability).

With a tower, it is something that I've always wanted to do, but the biggest problem would be in lifting up the sections plus working height... I'm not quite as agile as I used to be!

The trebuchet... Less materials... I could do it in my own back yard... I already have most of the materials and it'd be a lot of fun!... Very seriously considering it!

Ideally, I would like a university to take over with all testing... computer modelling... the whole lot!... while I play around with a trebuchet.

PerennialII
Gold Member
Some really tough luck, atypical environmental effect if there is one ... the trebuchet sounds like a fun project. And think integrity wise its easier to handle. Hope you have a big backyard thinking about neighbors and all .

Hello. I am in my senior year in mechanical engineering at the University of Arizona. I have spent the previous 25+ years hanging iron, installing mining equipment, building oilfield trucks and rigs, etc. I don't mean to hijack this thread, but a couple of thoughts have crossed my mind when reading about your spans.

Failure of a structure is not when it catastrophically falls to the ground. Rather, failure is defined as plastic deformation, ANY PLASTIC DEFORMATION, even when the structure remains standing. I have been in buildings and around equipment that remained intact but were considered failed due to the bent members. Are you planning on testing to failure or to destruction?

The cycling of a structure will weaken the material of the structure if the stresses are high enough. In other words, a maximum load applied once may not plasticly fail a part, but apply that same load repeatedly and failure is very quick. The data given by the steel manufacturers is conservative. They have to certify that their steel will carry X amount of stress without plastic failure.

Any engineering help you get will most likely factor in a margin of safety. This margin of safety is to insure the engineer that he can continue designing structures and not having to spend his days in court answering to charges that he built an unsafe structure. Where did the margin of safety come from? Testing to destruction!!

By the way, how are you loading the span? Remotely? Filling barrels of water with a hose?

Good luck, and stay safe.

Bill

Clausius2
Gold Member
arizonian said:

Failure of a structure is not when it catastrophically falls to the ground. Rather, failure is defined as plastic deformation, ANY PLASTIC DEFORMATION, even when the structure remains standing. I have been in buildings and around equipment that remained intact but were considered failed due to the bent members. Are you planning on testing to failure or to destruction?
Don't worry, you don't hijack at all. . I commented the same as you about plasticity. The fact you are seeing an structure collapsing implies the structure is failing, but the fact you are not seeing the structure doing so does not imply it is not failing. It can be achieving plastic conditions inside the beams and you are not able to see it, on the contrary the computer is able to check the whole volume.

arizonian said:
Rather, failure is defined as plastic deformation, ANY PLASTIC DEFORMATION, even when the structure remains standing.
I'd replace 'ANY' with something like 'significant'. How big is 'significant' will depend on the structure but localised plastic deformation isn't necessarily a problem and is, indeed, inevitable at the tip of the small cracks that exist in every structure. I'd agree with the need for FEA but, as has already been said, with this you have to decide in advance what type of failure mode is likely. Buckling, plastic collapse, fatigue, etc?

Good luck!

Ron.

Thanks, Clausius2 and rdt2. I still think I am in the field, and any plastic yield is significant.

Bill

Are you planning on testing to failure or to destruction?
It's already destroyed... What I wanted to do was to load it until whatever happened. I should have kept blocks about 1/2" under the drums so that I could correct, if it was interfering with the aim of the test, which was to see how much of horizontal deflection was being controlled... It probably wouldn't have saved it in this case though as it was the lateral control that gave out and the truss slipped sideways... unless I had rails for the drums to slide on... Too bad, as I really would have liked to have gotten up to the 4 tonne or more so that I could see what 'should' have happened... bottom or middle chord snapping, strut buckling or the top chord going into compression and popping through the arch... or another thought... one of the diagonals breaking... It was all about the horizontal deflection anyway, which was not put to the test due to my own oversight.
In hindsight now, I would have screwed some strapping (25x3) along the outer (vertical) return of the top sheet.

I'd agree with the need for FEA but, as has already been said, with this you have to decide in advance what type of failure mode is likely. Buckling, plastic collapse, fatigue, etc?
In this case I'd say anything was possible

http://groups.msn.com/TrussTesting/cartoons.msnw?action=ShowPhoto&PhotoID=69 [Broken]

Thanks all. I'll have some photos soon as well as presently working on a chart of the deformations...
I'll be in touch

Last edited by a moderator:
PerennialII
Gold Member
Cool picture ... :rofl: !

In this case I'd say anything was possible
Since this is essentially a (quasi-)static test the analysis approach that can capture all the details (plasticity + buckling) would be a material & geometric nonlinear FEA ... up to a point where you start to have ductile tearing itself, but at that point it's way beyond just deforming it. Sure, a generic FEA of that level starts to have some work in it, but I wouldn't call it overly difficult.

Even so, with an FEA model, unless it showed some pretty ectraordinary results, I probably wouldn't be happy until it was confirmed by the sound of things anyway. I'm not convinced yet.

Here are some pics of the collapse:
http://groups.msn.com/TrussTesting/trussesandbridges.msnw?action=ShowPhoto&PhotoID=77 [Broken]
http://groups.msn.com/TrussTesting/trussesandbridges.msnw?action=ShowPhoto&PhotoID=76 [Broken]
http://groups.msn.com/TrussTesting/trussesandbridges.msnw?action=ShowPhoto&PhotoID=75 [Broken]

http://groups.msn.com/TrussTesting/trussesandbridges.msnw?action=ShowPhoto&PhotoID=78 [Broken] (81mm first test...then 60mm on second)

http://groups.msn.com/TrussTesting/trussesandbridges.msnw?action=ShowPhoto&PhotoID=18 [Broken] Persons with FEA skills are welcome to use this design if they wish.

Also, If anyone wants to suggest a Phd thesis or other by a university... go for it!

Best regards,
David

Last edited by a moderator: