3D Design

3D Design

Axisymmetric model – Rubber seal

    • Pierre Drut
      Subscriber

      Good morning,

      I'm trying to simulate a seal going over a small steel protrusion as per picture below. The green cylinder will be inserted inside the rubber seal (in red).

      I have represented this assembly using an axisymmetric model as represented below.

       

      The issue I have is to go over the small steel protrusion where the elements of the rubber seal become highly distorted.

      The seal material is isotropic elastic linear (for simplification).

      I have tried a lot of different set-up without success to go over the small steel protrusion: Different mesh size, contact/target set-up, type of contact, formulation, behaviour, contact stiffness behaviour etc...

      I would be grateful if someone could help me? I can even share the Ansys model if this is possible?

       

      Thanks in advance,

       

    • peteroznewman
      Subscriber

      Use smaller elements around the two circled areas of the rubber part.

      Mesh the rubber part with Linear elements, not Quadratic.

      Add a small radius to the edge of the steel part and have frictional contact between the radius and the rubber face in the area circled in red.
      Is the flat steel ring bonded to the rubber or is there frictional contact?

      Use a Hyperelastic material model such as Neo Hookean, rather than a Linear Elastic material model.

      While the above may solve the highly distorted element issue, other problems may arise such as convergence failure in a Static Structural model. This is because after the seal reaches the tip of the steel protrusion, it will have a lot of strain energy and will want to snap past the tip. That is a dynamic event which can’t easily be captured in a Static Structural model. There is no static equilibrium just a tiny bit past the tip of the protrusion. This could be overcome by solving in Transient Structural. There are some tricks that can be played in a Static Structrual model to temporarily switch to an explicit dynamics solver to get past the dynamic event.

      The other issue is that you are enforcing axisymmetric response, but if you modeled this in 3D, you might find the rubber cylinder buckles into a non-round shape.

      If you want to share your model, use File Archive and put the .wbpz file on a File Sharing site such as Google Drive or OneDrive and paste the link to the file in your reply and say what version of Ansys you are using. Members like me will be able to look at your model, but ANSYS Staff are not permitted to download files, so you can only get their help if you post details in your reply of the error messages.

    • Pierre Drut
      Subscriber

      Good afternoon/morning,

      Thanks a lot for your advises! I have tried using the Transient Structural modul and it did work a lot better, I managed to pass the tiny bit part with the seal. I will now try in 3D to assess if this is making any difference.

      I have tried setting up an hyperelastic model in the past (even on very simple model), and I had a lot of convergence issues.

       

      Shall I expect any big differences in the calculation's results (such as train within the rubber part for example) using Transient Structural or static structural?

      Many thanks again

    • Pierre Drut
      Subscriber

      Hi Peter,

      Would you be able to assist on my last questions above?

      Many thanks again

    • peteroznewman
      Subscriber

      Hi Pierre,

      The differences between the strain achieved in a Transient Structural vs a Static Structural are problem dependent. They could be big or small, but in this case, there is no data from Static Structural because there is no solution past the tip of the bump, so Transient is all you have.

      There may be differences in strain when changing from a Linear Elastic material to a Hyperelastic material model.  However you need to perform a suite of hyperelastic tests on the material and fit the model constants to the material data to get accurate predictions from the model.

Viewing 4 reply threads
  • You must be logged in to reply to this topic.