instability of simulation - hyperelastic models

[anyCli0]

Hi , I m trying to run a simulation studying clipping of vessels. I m using the 22.06 version of SOFA. I m trying to module the behaviour of the vessel as a hyperelastic material. I use Mooney Rivlin, St Venant, Ogden, NeoHookean, but everyone is unstable under large deformation during clipping? can someone help me understanding at least the cause of this instability? Is there an energetic analysis to lead? are there any tips to improve the stability of the simulation. I tried using also Corotational model and I didn't have problems, so I supposed the issue is linked with the Hyperelastic behaviour. THank you in advance

[bakpaul]

Hello !

The v22.06 is more than 2 years old and thus deprecated, so the only help I can give you is on the numerical aspect : is your simulation well set ?

Hyperelastic models are non-linear models. But to solve the mechanics in one timestep, the stress-strain relationship is linearized. This has lot of impact when you are using the dynamic ODE solver EulerImplicitSolver -> the model is kept constant during the full timestep and may diverge too much from the real stress-strain relationship in one timestep, which is also directly related to the timestep size.

This is even more true if big displacements are imposed using either Projective or Lagrangian constraits : the first will project the position, and if they are far from the relaxing position, it might make the system totally ill-conditionned, the later is perform under the timestepping paradigm, meaning that the resolution fo those constraint might lead to deformation that are huge, leading to big instability in the non linear model.

I have thus two suggestions:

1. Use a very small timestep
2. If you are imposing displacement, add springs with the constraint (it doesn't matter if the constraint is projective or Lagrangian), this will help with the numerical conditioning and help not to go too far from the final position during the computation of the motion of the object (before the enforcement of the constraints)

[anyCli0]

thank you! I ll follow your suggestions

[anyCli0]

Do you know works or papers where the same problem has been addressed?

[hugtalbot]

Hi @anyCli0

A former topic (#3019) was addressing this topic. The instabilities observed in this topic could have been fixed following Paul's suggestions above ☝️

We can continue providing support on your own case. Providing simplified scenes can help and make our support more efficient