That looks strange. Your first constraint tells (provided it's linked to animal
) that the type of aninmal
must be mamifere
, but mamifere
inherits from animal
. That does not make sense.
The second variant does not make sense either. Provided the constraint applies to anything on the diagram, each instance must be named mamifere2
. So you can have only instances with name == mamifere2
. Especially strange with a vivipare2
instance.
I don't see where you model any child relation at all. So I'm just guessing you mean this:

A child has two parents (well, for humans there now can be more). And there can be *
children which must have the same type as the parents (so you can't model mules or the like).