Modeling bolts 3

[SkyD]

Hi All!

Has anybody got any experience with modelling bolts in FE? We're currently looking at the Main Landing Gear rib attachment to the spar and covers of our wing. We are using shell elements to model the parts (which are all composites) but we would like to assess bolt loads on the various attachments?

For the moment we are using an offset bettween the 2 faces and an RBE2 elements that constrain/links all 6 degree of freedom between the nodes at each bolt location. I don't think this is suitable...

Somebody suggested to use 3 springs to constrain translations in x,y and z and an RBE2 to constrain all 3 rotations...

What do you guys think?

Also, do you think we should model the contact between the faces? And if yes is there any way to do that other than using gap elements and a non-linear FE code?

Anyway if you have any suggestion, that will be much appreciated!

Cheers,
SkyD.

 

[Dinosaur]

I have been attempting to solve a problem similar to this involving bridge design. In our use, the bolts are tightened so they create a substantial normal force against the connected parts. The loads acting on the parts are in-plane forces so the loads are transfered as friction through the faying surface. Since the parts cannot move relative one to another the location of each bolt could be treated as a node on the edge of a plate bending element. However, the problem of eccentricity of the two plates may effect the individual bolt loads, so I want to include this in my model. To capture this, without introducing twice as many nodes and other problems, I am seeking a plate bending element derived with the nodes on one face of the plate rather than at midplane as is usually the case. With an element like this, I could study the bolt-plate interaction as the load moves through the parts.

I'm sorry that I don't have any specific help but if you get an answer I can use, I'll go back to work on my problem as well. Good Luck.

 

[rb1957]

i'm not sure about copying thread links, but here goes ... there're several threads like this one (from the "mechanical engineering other topics" forum, thread is "tightening bolts to 90% yield")

javascriptpenindex(450,450,'

http://www.tipmaster.com/includes/refinfo.cfm?w=450&h=450')

but there a some good references regarding bolted joint design.

I would analyze the bolts outside of the FEM, using the FEM to tell me the maximum load on the bolt, then stressing the bolt depending on the preload chosen. the hand calc would also consider the other effects you've already noticed, the shear loads at different planes causing moments in the bolt.

you can model the bolt as a single node, considering the bolt to be effectively inifinitely stiff. i've also modelled them with two co-incident nodes and model the bolt as a zero length spring (i had two elements, one for each (X, and Y) direction ... i was mostly interested in the shear loading of the bolt).

you could always preload the bolt in FE, then apply the external loads, but this seems way too much work !

good luck

 

[Aerodesign]

Hi SkyD,
The approach we use is consistent with Airbus UK techniques and is used for current A400M and A380 stress analysis. The bolt is represented as a discreet CBUSH (3 way spring type element) element, this should be zero length (ie coincident nodes used forboth eends), the stiffness is calculated using one of a number of ways (the HUSH or HARTS SMITH methods appears to give a good simulation of reality), the bolt stiffness calc takes into account the adjacent material flexibility. This takes care of the shear and axial stiffness.

FE post-processing in Nastran is easy, it is best to have a local coordinate system if your bolts are aligned away from cartesian system.

You will find that the flexibility of the bolt will soften the load transfered from part to part.

Hope this helps.

 

[SkyD]

Thanks Guys,

I think all this is pretty useful. Aerodesign, I think your method is appropriate for what I want to do. I believe that some people also add an RBE2 to the springs with only RX,RY and RZ as connected degrees of freedom. I would argue that RZ (being the axis of the bolt) may not be required.

Dinosaur, I would think that this method is not so appropriate for what you are trying to achieve though. In aircraft design, we tend to assume that the shear loads between 2 plates are transfered through the stiffness of the bolts. We do not take into account the pretension load and assume that no loads is transfered by friction. This fits with the use of springs between the plates...

As I understand it in your case, most of the load is transfered through friction. I've got the feeling that this may be a non linear case.

As for the eccentricity, that's another problem and I must admit I don't quite know how to deal with it. I think I simply won't model it.

Thanks again for your help.

Cheers,
SkyD

 

[rb1957]

"We do not take into account the pretension load ..."

i think this statement is at best partially correct.

we might not consider the effect of preload in a staatic analysis, on the assumption that the bolt gaps under ultimate load (and so the preload is irrelevant to the strength of the bolt). i would argue (and who doesn't like a good argument, so long as we degenerate to contradiction) that we should check the bolt load state with preload, 'cause whilst the applied load might be less than the bolt ultimate allowable, it not cause the joint to gap, and so the load in the bolt would be higher than we think, particularly if (as in tension applications) the preload is a significant portion of bolt ultimate capability.

in fatigue load applications, of course we consider the bolt preload.

 

[franck]

Hi Aerodesign,

Please could you point to an article or a book, which describes the "Hush or Harts Smith methods". I have tried on the web for some reference with no success.

Best regards,

Franck

 

[SWComposites]

It is Huth, not Hush. The reference is:

Huth, H., Influence of Fastener Flexibility on the Prediction of Load Transfer and Fatigue Life for Multiple-Row Joints, ASTM STP 927, 1986, pp. 221-250

Other fastener flexibility references are:

Tate and S. J. Rosenfeld, Preliminary Investigation of the Loads Carried by Individual Bolts in Bolted Joints, NACA-TN-1051, May 1946

S. J. Rosenfeld, Analytical and Experimental Investigation of Bolted Joints, NACA-TN-1458, October 1947

The Hart-Smith reference is:

L. J. Hart-Smtih, Design Methodology for Bonded-Bolted Composite Joints, Volume 1, AFWAL-TR-81-3154, Fegbruary 1982

 

[SkyD]

Thanks rb1957 for correcting my answer there on the pre-tension of bolts. When I said we didn't consider the pre-tension loads in our calcs, I was referring to a static analysis of a shear joint... I mainly wanted to emphasise the fact that we do not take into account any load transfer through friction.

Of course in the case of tension bolts like you would find say in a root joint, pre-tension is critical especially from a fatigue point of view.

Thanks SWComposites for the useful references...

Cheers guys,
SkyD.

 

[vonlueke]

SkyD: What advantage does using spring and/or RBE2 elements, or coincident nodes, give over just using a stiffened beam element, as described in my post at thread727-138868, other than being more complicated, which is no advantage? I agree, you should (can) release the torsional dof, unless you aren't neglecting friction.

Dinosaur: The load quickly migrates to the plate midplane, so I don't see an advantage in using an unusual element formulation, which misrepresents the predominant loading. If your loads are truly transferred by only friction, then you probably don't need to model a bolt, since it takes no applied load (unless you are also modeling preload and contact, which is more complicated).

 

[rb1957]

i think you may be missing something on the bolts ...
depends on how conservative you want to be.

if you're loading your bolts in shear, then the preload should be somewhat smaller than a tension loaded bolt (say 40%-50% of fty) but the bolt will still have a tension load and the applied shear load means that you should be considering load interaction in the bolt.

this is a little "belts and braces" ... if there's preload then there's a good friction loadpath so the bolt's not carrying the shear; but the ultimate condition could exceed the friction and so load the bolt ...

 

[Dinosaur]

In my work in the past, some gusset plates engaged the elements in single shear and were thus eccentric to the applied load. In order to capture the effect of this eccentricity, the thickness of the plates must be included in the model. Rather than connect the two pieces using plate elements with their nodes at midplate, I would like an element with the nodes on one face. Without this, I have to create two independent planes of nodes, and additional elements connecting the planes. Does anyone else see the value in decreasing the size of the stiffness matrix by half its nodes and eliminating the unnecessary "bolt" elements?

 

[rb1957]

i'd have thought that a simple model of the fastener, a zero length spring for example, would be good enough to get the shear reaction of the bolt. then do a hand calc on the bolt to include the effect of 2ndry bending.

alternatively, assuming you've modelled the gussets with 2D plate elements at their mid-plane, then a short beam with section properties of the shank diameter, could work. i think you'd need to check how the end moments get into the bolt (ie, look into the pressure distribution on the head of the bolt). you probably should still do a hand calc on the bolt, to include pre-load.

if you want bolt stiffness numbers, Niu's book has some (pg 235 of "Airframe Structural Design") or google "Swift" to get his formulas. i think you'll find that these two approaches give very different results, and maybe it makes just as sense to try a bunch of different values to see how sensitive the load transfer is.

good luck

 

[vonlueke]

Dinosaur: Your further explanation helped me understand your idea better. Unfortunately I currently don't know of an element with that formulation. I guess inputting a nonzero "z offset" value (in your shell element options) wouldn't work because, although it would induce the correct moment due to eccentricity on the first plate, it would unfortunately induce the opposite (wrong) moment on the support plate.

 

[Dinosaur]

" would unfortunately induce the opposite (wrong) moment on the support plate. "

This is one of the problems I came up with when I tried to derive the element myself. It seems it is also difficult, even after you have a mathematical element, to arrange the elements so that they know the nodes are on top of this element, but on the bottom of the other. I have been shocked to learn how intricate a rectangular element derivation can be.

If you find a plate like this, let me know!

 

[yli]

Hello SWComposites,

I am very interested in the reference you mentioned: "Huth, H., Influence of Fastener Flexibility on the Prediction of Load Transfer and Fatigue Life for Multiple-Row Joints, ASTM STP 927, 1986, pp. 221-250". However, I don't have access to this article, is it possible you can scan and post it somewhere so we all can access the reference?

Thanks,

yli