General Contact (Nonlinear, Hard Contact) in CCX

VMHVMH
edited January 2016
Hi, anyone has a suggestion on modeling general type contact (open and close without penetration and able to slide with friction coefficient defined) without doing trial and error the CCX contact parameters? The trial and error take alot of time waiting for the model to solve then do it all over again until a reasonable solution is reached. Thanks
«1

Comments

  • Hi VMH, look for the clip example in the Mecway installation folder, is a good example to start. I have to tune a little the first coefficient to work on soft parts.

    Regards
  • Yea. I also tried a few from bconverged calculiX beam contact examples. Adjusting the first coefficient to about 9 to 10 times the young modulus usually give a hard contact. Smaller values yield softer contact (with penetration). The hard contact take alot more time to solve. I tried the one of my contact models in Ansys using its general contact feature and it solved in seconds to minutes compared to CCX solved in minutes to hours and still didn't get a truly hard contact.
  • Using a fixed multiple of Youngs modulus, which is also suggested in the CCX manual, only really works when the geometry scale is constant.

    Here are two models with the same material properties, the same contact stiffness and about the same strains. The 1000x smaller one has about the same absolute penetration distance than the larger one, but since it's much smaller, one element goes about a third of the way through the other.

    So perhaps a better rule of thumb might be 10 * Youngs modulus / thickness. This formula also has the correct units (force/length^3) so it's not limited to newtons and meters.
  • I agree with Victor's comment above.

    Also, in my (limited) experience with CCX, I often find it's useful to start with NODE_TO_SURFACE contact and no friction to get things working, as NODE_TO_SURFACE seems 3x-5x faster and more efficient at multithreading than SURFACE_TO_SURFACE, at least on my problems (v2.8p2 on Windows).

    I usually add friction last, as it can cause convergence problems and sometimes requires really small steps. It's much easier if the penalty is already in the right range.

    I think with most codes, you can plan 2-3 iterations to adjust contact parameters for the first solves with a new model involving complex contact.
  • Thanks. I usually add friction last too.
  • VMHVMH
    edited January 2016
    Wow, there's a new feature (contact CCX only) I didn't realize (attached). This is nice!
  • Victor, this feature probably not in the current version 4.0 yet, correct? I searched and couldn't find it.
  • It only appears in the menu when nonlinear or dynamic response analysis types are selected. I guess this is quite confusing and other people have had trouble finding it too. I did that because it triggers a nonlinear solve in CCX so it would introduce side effects if you put it in a linear static model without knowing that.
  • Victor, would we see a side effect or odd solution if we use that in nonlinear type analysis with just 1 step instead of multiple step?
  • I wouldn't think so. The side effects I'm thinking of are geometric nonlinearities appearing when they would have been hidden in a linear analysis, and bonded contact behaving in a more restrictive way. In nonlinear, it becomes *TIE which doesn't allow large gaps between parts, whereas in linear static analysis it uses *EQUATION which allows large gaps but not large rotations.
  • VMHVMH
    edited January 2016
    Thanks Victor. Would the next version has something like material CCX only simliar to the contact CCX only so we don't have to type them in again and again each time. In either case, I'm really hoping that components can be included in the results when using CCX. This is to look at the results at the components interface. The current slicer is hard to do this because of possible geomety difficulties or simply cant align the model nicely with the slicer. So far that the only things I'm waiting for now. Thanks alot!
  • Components - yes. Materials - almost certainly.
  • Victor,
    This new contact CCX only feature and your suggested initial value are great! I've been trying to this example for some time and your feature is very easy to use. See YouTube video in the link below. Thanks!


  • Attached are the files if anyone interested.
  • Excellent video! I've added it to the Mecway YouTube channel.
  • Looking forward to the next release!
  • Excellent video, thanks!
  • VMHVMH
    edited February 2016
    See YouTube below for beam clip contact test.
  • Hi VMH, very good example and nice video also! What are you using to capture this videos?

    By the way, I see the same issue when you postprocess as me. The first thing that you look for is the magnitude displacement (to check that the model is displacing a razonable value), but, you must press the select face or element button to remove the node "emphasis" representation that make hard to see the model. Would be great that in the postprocessing mode this balls would not be pressent, as they almost hide the model result colors, or at least that the default selection mode for postprocess be element or faces (to present to the user a clear image of the results).

    Yes, I undertand that normally when postprocessing the more used feature is check values at nodes, but for complex meshes this balls almost complete hide the colors of the model.

    Regards!
  • VMHVMH
    edited February 2016
    Sergio, I used ActivePresenter which is free for noncommercial uses. I'm still looking for a better and clearer one that doesn't cost too much but haven't found one yet that is better than ActivePresenter n it's free.

    I do agree with your comment on all the nodes are selected cluttering the view. Basically we can't see anything useful there until click on select face icon to turn off the nodes display.
  • VMHVMH
    edited January 2016
    .
  • VMHVMH
    edited February 2016
    .
  • VMHVMH
    edited February 2016
    .
  • Updated YouTube video of Beam Clip Contact Test CCX in link below.


  • Very good example VMH, just one point to notice, when you solve in CCX, you have two error messages about bonded contact converted to TIE. What is the differences between a bonded contact and TIE? On the other hands, should be this messages changed to warnings? As far as I know in every solver an error will not allow to solve, but a warning yes, as in this case.

    Have you tried with bolt preloading? I has made some examples very like you did but still not the bolt preloading in CCX.

    Is very good your way to define the sets for contacts, first you define a set and rename it and then select the oposite side of the contact and define the contact feature. In that way you save to define and rename one of the sets (I made examples with 18 bolts with contact in both side, upper and lower side... and define/rename every set!!! I would save the half of sets in this way).

    I always make a set of contact (or tie) for every bolt, but in case that have as you several bolts in the same plane, can be made only one contact set, with the lower sides of the several heads and only one face of the plate? My procedure is generate a partition on the plate surface for every bolt, then make an individual contact, but now I'm in doubt if is necessary or not.

    Regards!
  • VMHVMH
    edited February 2016
    Sergio, I haven't tried bolts pretension. Just a thought: I dont really see a need for them unless doing slip critical connection where the connections can not allow to slip. But even with that, we can calculate the connection capacity with equations from AISC Manual. Bolts pretension also use for eliminating cyclic stresses in the bolts. If the max. tension in the bolts are less than the min. bolt pretension, then there are no cyclic stresses in the bolts, so we also don't need to model the bolts pretension in FEA.

    In the beam clip video, I could have used manual CCX feature to add the TIE but that require some typing to define the contact. Bonded contact in Mecway and TIE contact in CCX is the same except that TIE contact in CCX does not allow for large gap between the surfaces and Bonded contact in Mecway does (Victor told me that in another thread recently). This is just a thought: hopefully in the next release we also have something like "New Tied Contact (CCX Only)" and the original "New Contact (CCX only)" could be renamed as "New General Contact (CCX only)". You already requested for these and much more in your other post.

    Victor,
    Can you address Sergio question on selecting multi-surfaces and group them together to define contact. They seem to work for my test case but not sure if I also missing something. Thanks
  • VMH, I work for the automotive markets and sometimes we are request to perform FEA on bolted unions under extreme loads, in order to address slip and gaping problem. We do a quick hand calculation but at the end we perform FEA for some sensible parts, not only for the bolt stress but also to see how the load patern change due to the rotation due to sliplage (sometimes parts before breacking rotate and the bolts section is then in contact with the part... this change all your boundary conditions)

    Best regards!
  • VMHVMH
    edited February 2016
    Sergio, very interesting works! I would do the same too for that application.
  • Using a single bonded contact for many contact pairs should be perfectly OK. The solver breaks the slave surface down into nodes and the master surface into element faces so it has no concept of whether they are continuous or many separate parts.

    Sergio, good point about errors vs warnings. I'll try to distinguish them more clearly. That one is only a warning and it's safe to do as long as you're aware of the limitations of *TIE. for coincident aligned surfaces, it should behave the same as Mecway's bonded contact.
Sign In or Register to comment.

Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!