Peer review for bolted joint example

I have created an FEA example set for analyzing a simple bolt joint. The goal of the example is to help those interested in learning more about analyzing bolted connections and/or those interested in examples with contact or prestressing. The examples are not particularly difficult, but do demonstrate several advanaced FEA techniques as mentioned above, and thus I think may be useful. I would appreciate having some peer review before I make them available to anyone with a web-browser. The examples were modeled in FEMAP and solved with NE/NASTRAN.

I realized that many of the participants in this group probably don't use these packages, but if you would at least read the info/summary document at:

http://www.probasci.com/index_files/FEA_examples/Simple_bolt/bolt_example_2.doc

and provide comments it would most likely help to improve the quality of the example set. If anyone is interested in trying to run the models, the .NAS files are availabe in a zipped file at

http://www.probasci.com/index_files/FEA_examples/Simple_bolt/NAS_FILES.zip

(~3 MB)

and the full FEMAP neutral format files are available at

http://www.probasci.com/index_files/FEA_examples/Simple_bolt/ALL_FILES.zip

(~235 MB! ).


I will cross post to some different groups, so please forgive me should you recieve multiple postings.


I look forward to hearing comments and suggestions and thank you for helping.

Bryan Kirking

 

[mloew]

Bryan,

Please avoid cross-posting to ensure that the posts are not removed.

Best regards,

Matthew Ian Loew


Please see FAQ731-376 for tips on how to make the best use of Eng-Tips Fora.

 

[corus]

The model in the first link would have been better if you'd taken advanatage of axial symmetry and used only a 2D mesh. You can also see that the results are, in some places, mesh dependent. You would have produced better results using quadrilateral elements, particularly for such simple geometry.

corus

 

[johnhors]

I agree with Corus. With the solid mesh you have a very unrealistic 90 degree sharp corner, which is producing erroneous stress values. In 2D with a quad mesh you can quite easily and accurately model this corner with a true fillet radius. The size of model would still be very much smaller that your current 3D version. Also why do the mid side nodes not follow the curvature of the geometry?

 

[ProbaSci]

corus and johnhors,

Thanks for your insightful comments.

I used the tet because I originally had wanted to keep the examples such that someone (such as a student) who may have limited-functionality software could play along, and I thought tets would be a more "universal" option. I remember using Mechanica previously and that version did not have anyother option when starting from a solid. Of course one could build a model without using starting geometry - especially when the geometry is this simple - so this is due to my "proceedure rut".

I kept the full geometry to facilitate the comparisons with calculations to stiffness and the length deformation due to the preload. In the final format however, I left most of that out because I didn't think the increased complexity added enough to outweight the increased educational value - and since the models run pretty quickly anyway thought it was a better choice.

The midside nodes and fillet were oversights on my part. I was more focused on getting the average bolt stress to compare with the analytical calculations that I assumed that users would realize that the purpose wasn't to simply find the maximum stress, but that would be the obvious focus from most design criteria.

I will add comments to the write up to highlight these concerns. Maybe it will be something like "if you were doing this analysis for an actual product design keep in mind...."

Anyone else?