Learning Forum

[javat33489]

I am a young engineer. And I often face the limitation of the power of my PC and calculations. Please tell me this question. I have a rubber ring that I'm trying to squeeze. I know the material and I digitized it. The ring cannot be made symmetrical or axially symmetrical, because the cutouts are not symmetrical. It has to be counted in its entirety. With a good grid, at least 1 mm, this is 500k elements. My PC cannot do this. Maybe there are life hacks to simplify the calculation yet?

[Mike Rife]

Hi javat33489

Define compressibility for the material so that mixed uP elements are not needed, and an iterative solver (PCG) can be used.  Mike

[javat33489]

I know about the direct solver.

>>Define compressibility for the material so that mixed uP elements are not needed

It is possible here in more detail?

[Mike Rife]

javat33489

What kind of details?

[javat33489]

Define compressibility for the material so that mixed uP elements are not needed What do you mean? Please describe in more detail?

[peteroznewman]

What material model are you using for your rubber?  What material constants are you using for that model?

[javat33489]

I am using Ogden 1 order.

Constants built using curve fitting.

[peteroznewman]

Please reply with a screen shot of the material constants.

[Mike Rife]

The 1st order Ogden has 1 incompressibility term. If this is not defined then the material is incompressible and mixed uP elements are used.  For incompressible hyperelastic materials the mixed uP elements enforce the uP via Lagrange multipliers.  Iterative solvers like PCG have a hard time solving when stiffness matrix has Lagrange Multipliers (or cannot solve it at all).  So the sparse solver must be used.  Or define some compressibility for the material and then don't use the mixed uP formulation (this is done automatically in WB Mechanical).  The PCG solver uses about an order of magnitude less RAM than the sparse (direct) solver does.  So in a hardware limited situation, we try to use an iterative solver if possible.

Mike

[javat33489]

Thank you all is clear. Tell me, can I experimentally substitute small incompressibility values? Start for example with 1E-8 and gradually increase

[javat33489]

Tell me, can I experimentally substitute small incompressibility values? Start for example with 1E-8 and gradually increase

[Mike Rife]

Hi javat33489

Why not?  Have you tried running a test?  

[javat33489]

Yes, I tried with the incompressibility parameter 1E-8, the result is much better, the rubber began to compress. The calculation is on my PC. The iterative solver + enabled incompressibility 1E-8 helped me. Decides of course long but faster than it was. Should I gradually decrease the incompressibility parameter or is 1E-8 small enough?

[Mike Rife]

Hi javat33489

Deciding what to do from here depends on 1) the intent of the analysis and 2) whether a more accurate incompressibility is required for that intent and 3) if you have enough information to make a better choice of incompressibility.