Dear All,
I have the big plant to model and is the first time for me.
there are many lines and is not possible to think to do only one model.
Therefore, I've thought to divide the plant into sub-models. Is there any criteria to do it?
I can fix some semi-anchor points(guide+stop), then I can end the model at these points and restart from here with new model...

IMO, you can split the model in sub-models with the boundary conditions as "semi-anchors" (terrible noun!) provided that the possible rotation there has limited effects in both sub-systems.

Dear Mariog,
Regarding the possibility to split the model in more sub-models, If I end the first model in correspondence of stop support, I have to consider the forces on this support and apply them at the second model( that it starts at the same stop support) ?

In my opinion, the best approach is those discovered by experiments on relatively simple systems, split in two parts. You can see the effect of ignoring rotations in the split point.
If I were you, I would validate the split point by experimenting with rotations in both subsystems and see what it's happening; in case the subsystems are sensitive to that rotation I would change the split point.

Regarding your question, I think it is enough to place the stop in both subsystems; you obtain Force_Left and Force_Right on that stop and you can vector sum them.