Question: Tuning solver setting for the simulation of CKC multibody systems

Hello,

I could obtain the simulation of my multibody with kinematic closed chain (CKC).

However, it seems that from a specific time (around 12s) in my model I believe that I have some numerical instabilities. Indeed, I could compare my simulation results with another mulbody software. I obtain the same simulation until 12s and after in MapleSim, it appears many perturbations as you can see on the figures belows.

So, I think that I tune the numerical solver. This numerical solver must solve DAEs equations since my model contains 4 kinematic closed loops.

If i read correctly the help menu, there are the following methods to solve the DAEs :

- use specific DAE numerical solver (3 differents solvers are used : ck45 method, RKF45 method and Rosenbrock method

- use reformulation equations techniques (Baumgarte, Projection) which can be associated (I believe) with a classic solver like (RK4).

For the moment, I have obtained my results with the rosenbrock solver with error absolute : 1.0*10^(-4) and eror relative :error absolute : 1.0*10^(-4) 

Do you have some ideas or advices so as to find a better method to solve my multibody systems with kinematic closed loops ? This method should  prevent the creation of numerical instabilities.

Thanks a lot for your help