Chapter 4. Processing Activities and Processing Networks

A Processing Activity is a resource-constrained activity that takes one or more objects as inputs and processes them in some way (possibly transforming them). The processed objects have been called "transactions" in GPSS and "entities" in SIMAN/Arena, while they are called processing objects in DPMN.

Ontologically, there are one or more objects participating in an activity, as shown in Figure 4-1. Some of them represent resources, while others represent processing objects. For instance, in the information and process models of a medical department shown in Figure 3-11 and Figure 3-14, there are two processing activity types: walks to room and examinations. In walks to room, since nurses are walking patients to examination rooms, nurses and rooms are resources, while patients are processing objects. In examinations, doctors and rooms are resources, while patients are processing objects. If patients would walk to an examination room by themselves (without the help of a nurse), patients would be the performers of walks to a room, and not processing objects, and, consequently, walks to a room would not be processing activities.

Figure 4-1. Resource-constrained activities involving processing objects are processing activities.
???

Processing activities typically require immobile physical resources, like rooms or workstation machines, which define the inner nodes of a Processing Network (PN). A Processing Object enters such a network via an Arrival event at an Entry Node, is subsequently routed along a chain of Processing Nodes where it is subject to Processing Activities, and finally exits the network via a Departure event at an Exit Node.

The nodes of a PN define locations in a network space, which may be based on a two- or three-dimensional Euclidean space. Consequently, OEM-PN models are spatial simulation models, while basic OEM and OEM-A allow to abstract away from space. When processing objects are routed to a follow-up processing activity, they move to the location of the next processing node. The underlying space model allows visualizing a PN simulation in a natural way with processing objects as moving objects.

Each node in a PN model represents both an object and an event type. An Entry Node represents both an entry point (e.g., a reception area or an entrance to an inventory) and an arrival event type. A Processing Node represents both a resource object (e.g., a workstation or a room) and a processing activity type. An Exit Node represents both an exit point and a departure event type. A flow arrow connecting two Processing Nodes represents both an event flow and an object flow. Thus, the node types and the flow arrows of a PN are high-level modeling concepts that are overloaded with two meanings.

A PN modeling language should have elements for modeling each of the three types of nodes. Consequently, DPMN-A has to be extended by adding new visual modeling elements for entry, processing and exit nodes, and for connecting them.

In the field of DES, PNs have often been characterized by the narrative of “entities flowing through a system”. In fact, while in basic DPMN and in DPMN-A, there is only a flow of events, in DPMN-PN this flow of events is over-laid with a flow of (processing) objects.

PNs have been investigated in operations management and the mathematical theory of queuing (Loch 1998, Williams 2016) and have been the application focus of most industrial simulation software products, historically starting with GPSS (Gordon 1961) and SIMAN/Arena (Pegden and Davis 1992). They allow modeling many forms of discrete processing processes as can be found, for instance, in the manufacturing industry and the services industry.

It is remarkable that the PN paradigm has dominated the discrete event simulation market since the 1990’s and still flourishes today, mainly in the manufacturing and services industries, often with object-oriented and “agent-based” extensions. Its dominance has led many simulation experts to view it as a synonym of DES, which is a conceptual flaw because the concept of DES, even if not precisely defined, is clearly more general than the PN paradigm.

The PN paradigm has often been called a “process-oriented” DES approach. But unlike the business process modeling language BPMN, it is not concerned with a general concept of business process models, but rather with the special class of processing process models for discrete processing systems. A processing process includes the simultaneous handling of several “cases” (processing objects) that may compete for resources or have other interdependencies, while a “business process” in Business Process Management has traditionally been considered as a case-based process that is isolated from other cases.

For PN models, a simulator can automatically collect the following statistics, in addition to the resource-constrained activities statistics described in Section 3.2:

  1. The number of processing objects that arrived at, and departed from, the system.
  2. The number of processing objects in process (that is, either waiting in a queue/buffer or being processed)
  3. The average time a processing object spends in the system (also called throughput time).

During a simulation run, it must hold that the number of processing objects that arrived at the system is equal to the sum of the number of processing objects in process and the number of processing objects that departed from the system, symbolically:

arrived = in-process + departed