This paper proposes a method to build formal
models of networked automation systems, in the form of sets of
communicating timed automata, which are reduced enough to
avoid (or limit) combinatory explosion, but accurate enough to
provide meaningful proof results, when they are checked.