Model too stiff!!

[DrReinerKlimpke]

Hey everyone,
I am currently trying to re-simulate some bolt bearing analyses described in a paper I found recently (

https://journals.sagepub.com/doi/abs/10.1177/0021998310391047

). Unfortunately my model seems to be too stiff, approximately by a factor of 2, which I cannot explain up to date.

Therefore, I would like to address this issue to you guys, so maybe you have a hint for me concerning what could be the error of my model with respect to the referece simulations given in the above paper. To enable you to analyse my simulation in detail, you can find the corresponding ABAQUS input file in the appendix of this post. Furthermore, I illustrated all the details of my model together with some results and the way I determined them on two slides also given in the appendix of this post.

Keep in mind that I don't took account for any FRP damage in my model yet, since I am only interested in the initial model stiffness so far.

I'm looking forward to your suggestions.

Greetings,
DrReinerKlimpke

 

[Erik Panos Kostson]

In the paper they talk about Ffem=Fexp/4, while you might be using Fexp/2. Not sure if that accounts for the difference though, but it does result in a factor of 1/2 just like you show.

 

[DrReinerKlimpke]

Hey Erik, I know what you mean. But I think I already took care of this "symmetry condition" topic in a correct manner. When you look at the graphs illustrated in the paper you can see that they obviously relate to the full model, since a full model description is the only manner in which you can directly compare your results with experimental measurements. You can proof this by looking first of all at one of the given force-displacement graphs given in the paper. Dividing the force values of these graphs by the full projected area of the bolt, eg. A = 3.175 mm x 1.472 mm, will give you exactly the stress values that are illustrated a bit later in the paper as a function of the bearing strain.

To make my results comparable with the full model, I multiplied the force values obtained by my model by 4 but related it to the full projected bolt area at the same time, as you can see on slide two of my presentation ("Compute the bearing strain and stress values"). This calculation step directly relates to equation 5 of the paper.

Is that what you meant? Or did I get something wrong?

Thanks for your help!

 

[Erik Panos Kostson]

It looks like the difference comes from there (since if you divide stress by half you get similar results), so I would make sure that you are doing exactly the same thing (symmetry, and applied loads and how these are used to obtain stress,..) as the authors. I would recommend to write to them and ask how they do it, so you can then do the same thing.

Just some general feedback about the model. Mesh is really nice! Bricks (C3D8), do not have rotational dof so you do not need to restrain rotations. Bolt brick faces on symmetry planes do not have symmetry BC on them, they should.

 

[JBlack68]

can't access the paper. What is the bearing area you are using to calculate the bearing stress? cant see a 0.25*PI*D^2 on the slides

 

[DrReinerKlimpke]

Hey,

the bolt bearing stress is computed using only the projected area of the bolt/laminate contact area. So you don't have to use pi times something but only bolt diameter times laminate thickness. That is just a general agreement within the "bolt bearing community" so to say.

@ Erik: Thanks for the feedback. Actually I am in contact with the first author of the paper at the moment but since he's also quite bussy he did not have the time to look at the informations I presented you guys in the appendix of this post. I also hope that he will reply in the upcoming days.

Concerning the rotational dofs: How would you suggest to change the model BCs? Eliminate the rotational constraints from the symmetric boundary conditions?

Thanks to all of you for your help so far!

 

[Erik Panos Kostson]

No worries.

Well it is just that they (rotational restraints) do not do anything since there are no rotational degrees of freedom in 3D elements, so you do not need to do anything , it was just to point it out.

 

[Erik Panos Kostson]

In the mean time, I would recommend to compare the raw FEA results (rather then the stress calculation you do), of the reaction force at the application point and the displacement there, and compare that (load-displacement curve) to the graphs (FEA) in the paper (rather than stress and strains).

 

[DrReinerKlimpke]

Hey Erik,

that I also tried, but here my model is also far too stiff. An additional difficulty that must be taken into account when considering the force-displacement graphs is the effect of probe length that directly affects the amount of force you need for a certain displacement of the bolt. The longer the probe, the less force you need to reach a given displacement value of the bolt.
That would not be a problem, if the length of the model would be clearly defined in the paper. Although measurement values are given in the paper, which I also used for my model measurements (see slide 1, 101,6 mm - 19.1 mm = 82.5 mm), thinks are not that clear when considering the illustrations of the model that was used in the paper. You can see clearly, that the model in the paper is significantly shorter than the 82.5 mm. To make it short: I am not completly clear what length he used in his paper, which is an issue that is not of interest when looking at the bearing strains.

This is why I concentrated on the stress-strain graphs in the last view days.

Do you get my point?

 

[Erik Panos Kostson]

Then you definitely need to get some information from the authors, in order to have the same model.

 

[DrReinerKlimpke]

That I definitely want to do.

But I can already tell you that also for the force-displacement relation my model will overestimate the reference stiffness of the paper's model by about a factor of two. This is, because I already use a model, that is either of equal length (in case the authors stuck to their defined measurements, so L = 82.5 mm) or longer (in case the authors used a model that is indeed significantly shorter like the illustrations in the paper indicate) then what is used in the paper. Knowing that with increasing length the amount of force required for a given simulation decreases.

I can tell you, it's annoying as hell not to know where to look next to fix a problem, since you double-checked almost all details that might be of relevance.