Identification of industrial automation systems: Building compact and expressive Petri Net models from observable behavior

The paper deals with black-box identification of industrial automated discrete manufacturing systems. The problem of obtaining Petri net (PN) models from the observable behavior, expressed as a sequence of input-output vectors, is addressed. First the problem is stated: important issues to handle in systems automated by Programmable Logic Controllers that cannot be dealt by other methods are detailed. Then a novel method is presented; it focuses on building a compact and expressive representation of the observable part of the model which allows consequently the construction of a reduced complete Interpreted PN describing both observable and unobservable behavior.