This paper addresses the problem of discovering a Petri Net (PN) from a long event sequence representing the behavior of discrete-event processes. A method for building a 1-bounded PN able to execute the events sequence S is presented; it is based on determining causality and concurrence relations between events and computing the t-invariants. This novel method determines the structure and the initial marking of an ordinary PN, which reproduces the behavior in S. The algorithms derived from the method are efficient and have been implemented and tested on numerous examples of diverse complexity. Note to Practitioners-Model discovery is useful to perform reverse engineering of ill-known systems. The algorithms proposed in this paper build 1-bounded PN models, which are enough powerful to describe many discrete-event processes from industry. The efficiency of the method allows processing very large sequences. Thus, an automated modeling tool can be developed for dealing with data issued from real systems. Index Terms-Model discovery, Petri nets (PNs), t-invariants. I. INTRODUCTION D ISCOVERING formal models from external observation of systems behavior is an interesting and challenging approach for reverse engineering of discrete-event processes which are unknown or ill known. Although the problem is relatively recent, it deserves the attention of several research groups in the fields of discrete-event systems (DESs) and workflow management systems (WMSs). A. Model Discovery Pioneer works on the matter, named language learning techniques, appeared in computer sciences in the late 60s. The aim was to build fine representations (finite automata or grammars) of languages from samples of accepted word [1], [2]. In the field of DES, where the problem is usually named identification, several approaches have been proposed for
In this paper the problem of discovering a Petri net (PN) from sampled events sequences representing the execution of industrial or business processes is addressed. A method for building a 1-bounded PN from a single event sequence S composed of numerous execution traces is presented; it is based on determining causal and concurrency relations between tasks. A technique for computing the t-invariants of the PN from S is proposed; the obtained invariants allow determining the structure of a PN that executes S. The algorithms derived from the method have been implemented and tested on numerous examples of diverse complexity.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.