Is SOFA tailored for the linear system

[MA]

Hello!

I am dealing with a highly nonlinear dynamic system and results do always diverge from the empirical results, The question is SOFA tailored for the linear system? even if I do linearization it does not capture what happens in reality.

Thanks,

[Hugo]

Hi @ma1991

The first feature to consider is the time step. You need to use a sufficiently small time step so that a solution can be found.

With a highly nonlinear dynamic system, convergence can be difficult to reach, even with a small time step. Especially, if you are doing one Newton step per time step which is the case in most SOFA schemes. Only two schemes allow for defining the number of Newton step in one time step:
– the StaticSolver
– the VariationalSymplecticSolver

The latter should work in your dynamic case.

I hope this helps.
Best wishes

Hugo

[jjcasmar]

Would it make sense to provide some default integration schemes in which you can specify the number of Newton steps done per time step?

[Hugo]

Hi @jjcasmar

I agree with you that default integration schemes do not involve Newton steps. It is a pitty. However, the VariationalSymplecticSolver or the StaticSolver are doing it.

Best

Hugo

[MA]

Hi Hugo,

I am quite afraid whether the SOFA would be the right tool for my problem, simply I am going to do a benchmark to see whether the response could suffice. In the following scene, I am having a single s-yarn where I apply a force. The only thing I can add the young modulus and Poisson ratio. Most software for textiles is commercial and my institute does not have access to that software.

I am sharing the video, could you please tell me whether this could be worthwhile to continue with SOFA to model fabric, basically the analytical model is a set of springs( different stiffness), I am confused how would be the answer, I am expecting strain rate 1 up to 4 %.

https://gofile.io/?c=ny8A1i

https://gofile.io/?c=hG1l9y

[Hugo]

Hi @ma1991

I am sure SOFA can be your holly grall!
I guess you don’t want to limit to the simulation of one yarn, right? you would like to have several yarns interacting, colliding etc. right ?
A volumetric modeling as the one in your video is fine for one yarn but it will quickly become too computationally demanding. Unless you use some numerical optims, like a BeamModel (see BeamFEM) or with model order reduction.
https://github.com/sofa-framework/sofa/blob/master/modules/SofaGeneralSimpleFem/BeamFEMForceField.inl
https://project.inria.fr/modelorderreduction/
Note that recently a plugin implementing the Cosserat model has been released. This formulation could nicely fit your needs as well.
https://github.com/SofaDefrost/plugin.Cosserat

Best

Hugo

[Author]

Posts