Bolt Holes: Accuracy vs Computational Speed?

[PhilipFry]

I'm new to the stress world as of a few months ago. We are currently using Patran/Nastran to do our FEM on brackets and tubing. One of the big questions we are having at the moment is how accurate our current modeling techniques are. We usually make an RBE2 inside the bolt hole and then connect a constrained bar element to the independent MPC node to simulate the bolt. After thinking about it for a while, I realized that we have been neglecting the stiffness contribution of the structure against the underside of the bracket. To fix this, I used gap elements to simulate contact between the two surface. The drawback is that it dramatically increased the running time...which was expected I've connected every single node that makes contact to the surface with a gap element. That's a lot of gap elements! The question is, how much of this is overkill, or is it necessary to properly model the system?

 

[johnhors]

It depends on how much you need or want an accurate analysis, if your organization has been using RBE2's then it's either been quite happy to accept (or blissfully unaware) the nonsense results obtained by using these entities. The real question should be do you want results that have a chance of being realistic or no chance at all.

 

[PhilipFry]

Personally, I'm interested in realistic results, but I am also aware of the need to get results in a timely manner. Is there a compromise in between?

 

[rb1957]

mae1778,

how are you accounting for the preload in the bolt ? there is going to be some pressure between the brkt and the underlying structure, which is probably going to account for as big an effect as you're detecting by allowing the underlying strcuture to gap. the modelling issues with this are one headache, the next one would be how much preload to apply ... analyze both extreme values just in case something "goes south" somewhere along the line ??

i think too that maybe you are (fairly uniquely) plaed to answer your own question. how much difference does modelling the gap make ?

consider the problem from a different viewpoint. consinder your brkt is just about to fail, this would be highly gapped (no?). if your loads can be reacted acceptably in this configuration that they'll be ok in any less critical condition, albeit, your margins will be (slightly?) higher. I think this approach also works for an "optimal/minimal" design, if that's where you're working.

so i guess the short answer is that i think modelling gaps is probably overkill.

 

[Dinosaur]

The proof of the pudding is in the eating.

How does your original ungapped model compare with the later gapped model? Are the answers substantially different or are they nearly the same? Can you put the difference to practical use?

I wouldn't be so concerned with the actual run-time, but with the input and checking time. Does the difference in answers justify the labor?

 

[PhilipFry]

All good points. The input is relatively painless, but the computing time could be very long if I did this to every bracket in my system.

As to whether or not this makes a difference? It does. However, how big of a difference depends on how you interpret your results. I end up getting a big stress concentration on the bracket where the head of the bolt ends. That's to be expected, but how much of the stress concentration comes from modeling? I'll probably need to refine my mesh again around the hole to know for sure. If I simply ignore the elements with the peak stress, it lowers it by 5 ksi...if I leave it in, it's higher by about 20 ksi.

Also, using the gap elements means that the brackets will actually exhibit a heal and toe loading pattern that it did not previously show.

In summary: Inconclusive. I probably need to try it with several load cases to know for sure.

 

[Kwan]

how complicated is this bracket? how many mounting fasteners are there? How many loadcases? Lastly, how much does weight cost you?

 

[PhilipFry]

Some of them can be complicated...others are simple L and Z brackets. Usually, we have three mounting fasteners, but that increases with the complexity of the bracket. We run 5-16 load cases. Weight is always an issue.

 

[vonlueke]

mae1778: Note that if the plates are relatively thick (stiff), and the bolt pattern has a sizable footprint (i.e., more than one line of bolts) resisting significant prying moment in that direction, then not using gap elements will still give reasonable bolt tensile loads and will be slightly conservative. If you use this method, make the bolts very stiff, and also create a stiffened washer face region around the bolt node approximately the size of a large washer (or slightly larger), to resist prying rotation, thereby keeping the plates parallel. Note especially that each washer face region can be square instead of circular, giving essentially the same results. Release a torsional dof on each bolt beam element; and typically ignore moments on the beam. Pay attention to the von Mises stress just outside your stiffened washer face region (if the plates are actually prying apart there instead of compressing together) to ensure your plate can handle the bending stress.