General Mechanical

General Mechanical

Convergence issue of interference fit for hyperelastic material

    • Haiquan
      Subscriber

      Hello everyone

      I faced convergence issue in 2D axisymmetric interference fit for hyperelastic model as shown below. I am using Yeoh 3rd Order. My test data for curve fitting are uniaxial tension and biaxial tension data. Data curve fitting is up to strain 80%, and my simulation strain range is 50%.

      I tried many techniques , including fined mesh, large substeps, u-P method, introducing small Incompressibility Parameters, contact setting and analysis setting, but I still failed to get convergence.

      If I use displacement load instead of interference fit, it is very easy to get a convergence as shown below, even almost all the settings are default value. So the hyperealstic can sustain a 50% of compression load, but why is it difficulty to get convergence in the interference fit case? I attach my simulation cases in version 2020 R1.

      Thank you so much in advance!  

    • Bhargava Sista
      Ansys Employee
      How far along does your interference fit solve to? Please make sure that you have the following settings defined correctly:nPinball radius is more than the amount of interferencenInterface Treatment is set to Add offset, ramped effectsnHyperelastic part is chosen as Contact side nIn addition to that, it may help tonSet contact stiffness factor to 0.1nUse first-order elements for hyperelastic part if incompressibility parameter is zeronIf you haven't watched it already, then check out this video for a few more pointers.n
      target=blank>
      n
  • Haiquan
    Subscriber
    Thank you very much for your comments and the information! I have tried all your proposals, contacts has been established, but it still didnot work. The digvergence issue occurred at the beginning of the calculation as shown belownnnnnn
  • Haiquan
    Subscriber
    Any comments on my case? Thank you !n
  • peteroznewman
    Subscriber
    nSince Displacement load is easy to converge and Interference Fit method is failing to converge, you have a way to get results.nYou are only asking why the Interference Fit method is having difficulty. I would say it has to do with the size of the interference.nI don't know exactly how the ramped effect works, Does it offset the contact element normal to each face on an element-by-element basis? If so that would bunch all the contact elements into a ball near the center of the part. That could make convergence difficult. I do know that Adjust to Touch moves the entire set of contact elements together along the vector of the maximum interference. Maybe Ramped Effects does the same, I don't know. nIn either case, only the contact element is moving, not the shell element sharing nodes with the contact element at the surface of the circle. That means huge moments are present between each contact element and its underlying shell element. This may be the reason for convergence difficulty. In the Displacement method, the contact element remains coincident with the nodes on the shell elements and no moments are present.nOne thing you can try is to flip the Contact-Target pair and see if the Ramped Effect works any better.n
  • Haiquan
    Subscriber
    Thank you very much for your comments! Yes, I found that large interference will cause convergence issue. Actually I need to apply temperature in second laod step, that is why I use interference fit for the first load step. If I use the displacement load as the first load step, it is not easy to apply temperature in second load step. I tried your proposal to flip contact-target, it did not workn
  • Bhargava Sista
    Ansys Employee
    As Peter mentioned, the direction of loading can get messy at large interference which poses convergence issues unlike in displacement loading case where the direction of loading is consistent.nTwo additional suggestions that I have:nUse Normal Lagrange formulation (this does not use contact stiffness so that option disappears)nChange contact detection to Nodal Normal to Target since the target surface remains relatively flat.nRegarding the interface treatment, Ramped Effects increments the total interference over the time steps so the load is applied gradually as recommended in any nonlinear FE simulation. If you set it to No Ramping, the solver tries to resolve the full interference in the very first sub-step which causes convergence issues.nWith Adjust to touch, the contact detection points (which are not the same as nodes) are offset such that the interference between the two surfaces is zero. The nodes are not physically moved so you may still see some geometry penetration but the contact forces will be calculated assuming there's no interference. It's a crude explanation but I hope it helps paint the picture.nCheck out this video from the time I indicated, it explains the difference between these options.n
    target=blank>
    n
  • Haiquan
    Subscriber
    Thank you very much for your comment !n
  • Viewing 7 reply threads
    • You must be logged in to reply to this topic.