"Nested" parts and no penetration

[AbeFM]

Sorry if this question is a little vague...

I've got a customer who wants to stiffen a long aluminum rectangular tube with a short length of smaller bit of steel channel centered within it. I thought I'd run a quick sim to just see how it behaves... And I'm having trouble setting up the correct boundary conditions such that the outer tube bends until it contacts the steel center as one would expect. They start off without contact, so I added a "no penetration" condition to them, and they just move as if bonded. I can only assume there's something to do with the "gap" feature, but being that I set that down to a few mils and the parts are separated internally by a hundred times that, I'm at a loss on how to make it behave.

Tweaking a few random options (like global non-penetration) and trying h-adaptive is making it run quite slowly.... Just looking for a quick pointer, thanks!
-Abe.

 

[mmar2087]

It sounds strange. Do you have a picture on this setup?
The global bonded condition (default) only works if the parts have initial contact so this shouldn't be the problem in your case.
If you manually add a no penetration contact (also works with bonded) the part don't have to have initial contact.

 

[AbeFM]

Thanks for getting back to me - I was starting to give up.

I was hoping with this to get an idea for when my central beam would overstress it's shell and break through. With this simulation I get none of that.

I have convinced myself that the times where the shapes flex independently and appear to go through one another is actually the scaled deformation - the beam may bend only 0.100", but it translates 0.400". When the deformed plot exaggerates this, the central beam "leaves" the envelope of the other part. That's a working theory. So far I can't figure out why sometimes they bend together and sometimes they move relative to each other.

This shows both my result (scales at 1x), the whole system (with a split line on the external face for fixtures), and a close up of how I defined the force and fixture. Everything's taken at a pretty steep angle since the aspect ratio (2%) won't allow a face on view with any detail on a reasonably sized image.

Right now my suspicions lean towards gap handling and bonding. In this case, I have no global bonding defined (suppressed) and I have the four inner faces of the outer tube as well as the 5 outer faces of the inner channel set up in a series of 1:1 (or 2:1 in the case of the open end of the C) 'no penetration' contact sets.

I'd truly love to get this working, especially since for my own edification we have a physical test rig set up and I would love to test the results. Thanks in advance for helping or any publicity you can give my problem.

 

[AbeFM]

Small PS - The bar seems to move the wrong way - it's as if the arrows show the force 'coming out' of the surface, not going in. I get similar results when I leave the bar touching one side for initial conditions, FYI.

 

[AbeFM]

 

[badboggs]

Hi Abe. I'm having a hard time understanding your boundary conditions. How is the inner steel piece attached to the AL piece in real life? Would be happy to help but I need to understand the setup a little more.

As far as global settings go you want to use Allow Penetration and then specifically set your contact conditions between the parts. Also, this will have to be run as a non-linear analysis. Linear static analysis cannot do what you are trying to do. Let me know.

 

[AbeFM]

Oh, interesting. I thought non-linear wouldn't let me assign no penetration!

In real life, the beam is vertically (between windows in a tall building), and a bolt or otherwise will be slipped thought the outer channel to prevent the channel from falling. It really does "free float" in there as best as can be accomplished. I was thinking perhaps some weak spring or some such to better define it might be appropriate, but I really just want it to sit loose in there. I was thinking of bonding just the very ends of the bar to one side but I didn't want that force of compression showing up preventing the aluminum to stretch.

I can certainly add a global Allow Penetration condition.

As to the conditions I have defined - all inner surfaces (4) of the outer channel are "no penetration" with the closest outer surface of the inner channel (if you shook the whole assembly, that's what keeps the channel from popping out the sides). The only oddity is the open side of the channel (with it's two small ~0.135" faces, so for that relation I have the two of them "no penetration" with the inner wall of the alum channel facing it.

Let me know if that's still not clear, and if I'm mistaken about the non-linear sim not allowing "no penetration". I read it in help and didn't see the option in the software (2012)

 

[badboggs]

Sure you can use no-penetration in non-linear. Just look at the tutorial example Nonlinear Contact Analysis of a Pipe Holder. Anytime you are looking at large deflection like this you have to use non-linear. Remember that non-linear does not exclusively mean below yield. Let me work up an example model assy for you.

 

[AbeFM]

If you'd prefer I'd be happy to send you mine so you don't have to work anything up.

 

[badboggs]

Looks like I won't have time to do that right now. One other suggestion. Try using shell models as well. They should work well for your situation.

 

[badboggs]

Is that something you can post?

 

[AbeFM]

Sure. If someone manages to take over the world by stealing THIS design, I'd be happy to be their subject.

This is just the basic model with two configurations for inner beam centered or resting against one wall. If you need more PM me your email and tell me what you need.

Thanks!!
-Abe.
P.S. On another deadline for tonight, tomorrow I will likely be able to experiment with non-linear again. Not sure why I didn't see no penetration before.

 

[AbeFM]

Shoot, non-linear keeps crashing out. With shells, I can't really get it started. Without them, it runs for an hour and complains things aren't tied down. This all makes sense, but not sure how to tackle it.

Attached is the same assembly, with the non-linear study files included.

 

[mmar2087]

You don't need to be in the non-linear module to handle contacts.
In SW, contacts, large displacement and "in plane effects" (stress stiffening) are available in the static analysis.

I would try to add a fixture on one node on the inner structure and lock it only in the axial direction to prevent RBM.
Don't use the soft spring option.
I also prefer to manually add BC instead of the global one for "no penetration", but it should still work if you stay on global.

See if this helps.

 

[AbeFM]

Thanks. From what I read, the global no penetration is just a lot slower. Mostly I've been staying away from it.

I did a non-linear real quick and saw a difference in deformation between the two bars, so I feel I'm on the right track. That said, I like the idea of of the single bonded contact.

What's wrong with soft spring? You're right it seems I wouldn't need it if I had it bonded - otherwise it fails.

 

[AbeFM]

Slightly related question - does this seem like something worth using symmetry on?

 

[badboggs]

Here is a sample model set I made based on your geometry. I'm still not 100% clear on how your inner part relates to your outer part. Note that I'm fixing both models and am using draft mesh for speed. Take a look and let me know if this shows some insight into what you could possibly do.

http://files.engineering.com/getfil...5d-4f2b-a1e2-fd84de5300bb&file=_fea_tubes.zip

The real key is figuring out how to properly restrain the inner part and still observe the stiffening effects you are after.

Cheers

 

[AbeFM]

Thanks! That is certainly in the vein of what I am doing (was doing, am doing again - had to back burner that project for a while).

Perhaps try to download the files I attached earlier? There are a couple changes:

1) My beam is fixed on both ends, bending in the middle
2) My beams have some/more cross sectional thickness

The issue I'm having with what you posted is that it matches what I was getting earlier - a bar in the middle flexing for no reason. There's nothing really pushing on it. What I ultimately am after is what is the pressure/force/etc on the inside of the outer tube due to the stiff inner support pushing on it.

Even time I try shells things fail - though I'll admit I've never used them before.

I expect a result something like this (I've shown the red stiffener as unbent, in reality there would be a SMALL bend but in no way should it have the same shape as the outer bar)

The issue is the simulation puts them both out with the same shape, or not interfering at all. I did try bonding one edge of the steel to the inner surface of the alum, and this works but the simulation is taking many hours to run - likely due to the small mesh size to capture the thickness.

 

[AbeFM]

Oh! Actually making good progress. I got an error saying "use surface to surface" contact instead of node to surface, and after that and changing the solver, things seem to be going well. I still feel shells might be more appropriate (every element spans the thickness of the tube), but at least I'm getting something reasonable. 1.8" deflection on 8,000 lbs of force which is also ballpark reasonable.

I'd like to tell it to push as hard as it takes to get me a specific deflection, is there an easy way to set that up?

 

[loki3000]

how will the original girder handle the stiffener with regards to corrosion?