Petri Nets have essential strengths in capturing a system's static structure and dynamics, its mathematical underpinning, and providing a graphical representation. However, visual simulation models of realistic systems based on Petri Nets are often perceived as too large and too complex to be easily understood. This constrains stakeholders in participating in such modeling and solution finding, and limits acceptance. We address this issue by considering a structured approach for guiding the analyst in creating more insightful models. Key elements are a domain-related reference architecture that supports conceptual modeling coupled with uniform rules for mapping high-level concepts onto low-level Petri Net components. The proposed approach is implemented and illustrated in the manufacturing domain.
Petri Net formalisms have proved to be successful tools for the modeling and simulation of manufacturing systems [6], [8], [23] and [41]. Their success is underpinned by their basic strengths in accurately describing a system's static structure and its dynamics, the availability of mathematical analysis methods, and their graphical nature.
Many authors, however, argue that the use of Petri Nets in decision support for the design of manufacturing systems could be further improved if the model's size and complexity could be reduced. These characteristics are seen as hindering more active stakeholder understanding, participation, and acceptance of solutions [2], [10], [14] and [26]. In this article, we address this issue by developing guidance that could assist the analyst in defining more insightful visual models.
Stakeholders' understanding of visual model elements, and their workings, starts from a recognition of high-level concepts [12] and [21]. Examples of such concepts within the manufacturing domain relate to machines, buffers, planners, and goods. The selection of concepts and their visual representation follows from the analyst's creativity. This creativity is bounded and guided by implicit or explicit guidelines, that is by good modeling practices and principles [20], domain-related insights [30] and, last but not the least, the logic and libraries that underlie simulation software [11].
Typically, Petri Nets start from a few basic low-level components used to define high-level constructs that resemble manufacturing entities. These low-level beginnings explain model size on the one hand, and the great efforts that the analyst has to put into insightful model structuring on the other. Basically, we assume model structuring to refer to both sound conceptualization, i.e., the choice of entities, their activities, and their relationships considered characteristic of a domain; and formalization, i.e., the mapping of the respective system elements onto model components. In this paper, we consider guidance for the analyst related to both activities.
In this article, we have addressed the creation of insightful visual simulation models based on Petri Nets. While Petri Nets are known for their qualities in accurately modeling and representing real systems, Petri Net models of real situations are often found to be too large and complex to be understood by non-experts. In response, we propose a structured approach for guiding the analyst in developing a model that provides greater insight to stakeholders.
Key elements of the approach are a reference architecture that captures the essential object classes of a domain, and a set of mapping rules for representing the respective objects as Petri Nets. Here, the reference architecture builds on elementary decomposition principles that characterize the field of interest. As an illustration, the approach was implemented for the manufacturing domain by linking a manufacturing reference architecture to a specific Petri Net formalism, i.e., ExSpect™, using mapping rules, and its use illustrated with a case example.
The reference architecture allows analysts to start from a clear and appealing – and shared – idea of the essential object classes within a domain, rather than having to create a vision of their own. Further, the mapping rules help to bridge the perceived “gap” between the high-level manufacturing concepts and the low-level components of the Petri Net in a uniform way. This enhances model transparency both for stakeholders and for analysts. We view transparency as the basis for model effectiveness in terms of better quality and credible solutions, as well as for modeling efficiency, i.e., minimizing the analyst's efforts in building, adapting, and re-using models.
