The performance evaluation of manufacturing systems is a critical and difficult task to be addressed throughout the factory life-cycle phases, including the early design, detailed design, ramp-up, reconfiguration, and monitoring. An efficient and effective performance evaluation may have a relevant impact on the profitability of an industrial company. This paper addresses the application of a data model for virtual factories to the performance evaluation problem, aiming at exploiting the interoperability with other software tools to continuously update the virtual representation of a manufacturing system, so that accurate estimations can be obtained. A test case is described and then used to check the viability of the proposed approach in the case of Discrete Event Simulation (DES) based on a commercial software tool like Arena.
The design of manufacturing systems is a complex task strictly related to
manufacturing strategy decisions having an impact on a long time horizon (usually
more than two years) and involving a major commitment of financial resources [
Usual performance indicators in a manufacturing context can be the production volumes, the quality of the output, the incurred cost, etc. In addition, more detailed performance indicators may be calculated, e.g. the utilization of production resources, the average flow time of products, the average level of the work in progress. Different models can be used to address specific types of analysis and levels of detail while modeling a manufacturing system to evaluate its performance. In the field of discrete part manufacturing, two main approaches are in common use:
Analytical models using mathematical or symbolic relationships to provide a
formal description of the system [
Simulation models represent the events occurring in a manufacturing system in
its operation by a sequence of steps that are executed in a computer program
[
Simulation models enable the representation of an higher level of details, thus providing more accurate estimates of the manufacturing system behavior compared to analytical models. However, to reach this level of details, also a more detailed formalization of the manufacturing system is needed. Simulation modeling of manufacturing systems usually relies on commercial software tools (e.g. Arena, Simio, Plant Simulation, Visual Components, etc.) providing an integrated environment to describe the system and its behavior in terms of relationships and rules and, in addition, to deal with the generation of random values and the collection of the statistics.
Performance evaluation tools can be more effective if they are based on a virtual representation of the manufacturing system that is continuously updated during both the design and operational/execution phase, thus guaranteeing an overall coherence of the obtained results. Moreover, the generation of a simulation models and/or analytical model can be time-consuming and it would be beneficial if this activity could be as much automated as possible. The resulting need of interoperability between performance evaluation tools and tools supporting the design and management of real industrial systems can be met by an extended framework enabling: the cooperation among different actors with different competences and expertise in the design and management of a factory based on common definitions and a shared virtual representation of its components linking different manufacturing domains while guaranteeing their coherence; the management and update of a huge amount of manufacturing data made available through standard and interoperable interfaces.
The development of a framework for the interoperability between software tools
supporting factory processes is currently carried out by the European project “Virtual
Factory Framework” [
Virtual Factory Data Model (VFDM), i.e. a coherent, standard, extensible, and
common data model for the representation of factory objects related to
production systems, resources, processes and products [
Virtual Factory Manager (VFM), i.e. the manager of a shared data repository
containing factory data that can be accessed and modified by all the software
tools integrated in the framework [
This paper focuses on development and enhancement of the VFDM for modeling a generic manufacturing system and then evaluating its performance. Section 2 gives an overview of the current state of the art on data models for manufacturing systems and presents the VFDM solution. Section 3 describes a test case representing a production line. Section 4 delves into the problem of evaluating the performance of a manufacturing system formalized according to the VFDM; in particular Discrete Event Simulation by means of the commercial software tool Arena is addressed. Finally, conclusions are drawn in Section 5. 2.
Several scientific contributions and proposed technical standards have faced the problem of developing a holistic and complete data model for representing manufacturing systems, both considering tangible (e.g. machine tools, part types to be produced, etc.) and intangible (e.g. process plans, production logics, etc.) aspects.
Among the available technical standard, ANSI/ISA-95 [
According to ANSI/ISA-95 standard, a manufacturing process can be modeled using the ProcessSegment class. The ProcessSegment class can represent a single step in a manufacturing process or a whole process through composition. The ProcessSegment class is linked to further classes to characterize the process, e.g. the needed equipment (EquipmentSegmentSpecification class), the personnel (PersonnelSegmentSpecification class) and the material (MaterialSegmentSpecification class). Furthermore, precedence relations between different process steps can be defined thanks to the ProcessSegmentDependency class. The EquipmentSegmentSpecification class allows the user to specify the pieces of equipment needed for the execution of a process step and how the equipment. ANSI/ISA-95 standard enables the user to freely define customized properties that can be attached to most of the classes representing processes and production resources. However, such flexibility can be a major drawback from the interoperability point of view. Indeed, if two users adopt ANSI/ISA-95, they still have to agree on the definition of the object properties before being able to exchange data characterized by a proper semantic. Furthermore, ANSI/ISA-95 does not provide a complete support for modeling physical data such as the placement and shape representation of objects in the manufacturing system (e.g. a machine tool).
A different approach in the modeling of manufacturing process is offered by the
Process Specification Language (PSL) standard [
The Industry Foundation classes (IFC) standard by buildingSMART [
The Virtual Factory Data Model (VFDM) of the VFF project is based on already
existing technical standards and extends their definitions to represent the characteristics
of a manufacturing system in terms of the products to be manufactured, the
manufacturing process they must undergo and the resources entitled to operate the
different manufacturing operations [
The Entities in the IFC standard are mapped to OWL Classes in the VFDM. Most of the classes derived from IFC are specializations of two fundamental classes named IfcTypeObject and IfcObject, both being subclasses of IfcObjectDefinition. The former class is the generalization of any thing or process seen as a type, the latter seen as an occurrence. OWL individuals of class IfcObject may be linked with a corresponding individual of class IfcTypeObject.
IfcTypeObject has the following subclasses: IfcTypeProduct, IfcTypeProcess, IfcTypeResource. IfcTypeProduct represents a generic object type that can be related to a geometric or spatial context (e.g. manufactured products, machine tools, transport systems, etc.). IfcTypeProcess defines a generic process type that can be used to transform an input into output (e.g. assembly operation, machining operation, etc.). IfcTypeResource represents the information related to resource types needed to execute a process. A resource represents the “use of thing”.
IfcObject has the three main subclasses (i.e. IfcProduct, IfcProcess, IfcResource) that represent an occurrence of the corresponding type modeled by the subclasses of IfcTypeObject.
The previously described generic classes can be exploited to model a wide range of manufacturing systems while taking into consideration both physical and logical aspects. The subclasses of IfcTypeObject can be used to specify the designed characteristics of a manufacturing system, e.g. the part types to be produced (as individuals of IfcTypeProduct), the process plans (as individuals of IfcTypeProcess), the required type of production resources (as individuals of IfcTypeResource). On the other hand, the subclasses of IfcObject can be used to represent the execution phase of a manufacturing system by defining the workpieces in process (as individuals of IfcProduct), the actually executed operations (as individuals of IfcProcess), and the usage of production resources (as individuals of IfcResource).
The relations between the processes and resources can be formalized as shown in
Figure 1 where the boxes represent classes and the arcs represent property restrictions
linking classes according to the Manchester OWL Syntax [
operatesOn only hasRelatedObjects only isResourceOf only hasRelatedObjects only IfcTypeProcess
IfcTypeResource
IfcObjectDefinition IfcRelAssignsTo
Process
IfcRelAssignsTo
Resource hasRelatingProcess only hasAssignments only hasRelatingResource only hasAssignments only hasRelatingType only IfcRelDefines
ByType isTypedBy only operatesOn only hasRelatedObjects only isResourceOf only only hasRelatedObjects only IfcProcess
IfcRelAssignsTo
Process
IfcRelAssignsTo
Resource
IfcObjectDefinition hasRelatingProcess only hasAssignments only hasRelatingResource only hasAssignments only During the manufacturing system design/planning phase, the resource types needed by a process type can be specified by means of the objectified relationship class IfcRelAssignsToProcess, whereas the resource providers (as individuals of class IfcObjectDefinition) can be linked to a resource type thanks to the class IfcRelAssignsToResource.
During the manufacturing system execution phase (both real and simulated), occurrences of processes and resources can be created while referring to specific types defined during the design phase thanks to the class IfcRelDefinesByType.
As described by Terkaj et al. [
VffProcessType and VffProcess are defined as subclass of IfcTypeProcess and IfcProcess, respectively, to model generic transformation processes that, provided a given input, obtains a certain output according to certain rules and using a specified set of resources, i.e. a recipe. A process can be described as a whole or can be decomposed into subprocesses thanks to the class IfcRelNests. VffProcessType and VffProcess are further specialized to represent manufacturing, assembly, maintenance and handling processes. Moreover, precedence constraints between the processes can be defined by means of the objectified relationship class IfcRelSequence, whereas input and output entities of a process can be linked by using the classes IfcRelAssignsToProcess and IfcRelAssignsToProduct, respectively.
VffProductionResourceType and VffProductionResource are subclasses of IfcTypeResource and IfcResource, respectively, modeling a generic resource used in a factory (and its production systems). These classes are further specialized to represent equipment resources, material resources, and human resources, respectively.
In the VFDM the classes VffMachineryElementType and VffMachineryElement have been defined as subclasses of IfcTypeProduct and IfcProduct, respectively, to represent generic pieces of machinery equipment.
Finally, specific property classes (e.g. VffProcessProperties, VffMachineryElementProperties) have been created to properly characterize processes, resources and machinery elements.
This section presents a test case representing a production line to show how the VFDM can be employed to create factory projects and use them with different digital tools. The test case consists of four ontologies that instantiate the VFDM classes, thus exploiting the data distribution empowered by the Semantic Web approach: three factory libraries (i.e. VffLibrary01, VffLibrary02, VffLibrary17) and one main factory project (i.e. VfProductionLine04). All these ontologies import the set of VFDM ontologies.
VffLibrary01 ontology defines a production site and a building.
VffLibrary02 ontology defines five machine types (as individuals of class VffMachineryElementType (i.e. MtA, MtB, MtC, MtD, MtE). Each machine type is associated with two possible shape representations in VRML and 3DS format.
VffLibrary17 ontology defines a part type as individual of class VffWorkpieceType (i.e. a subclass of IfcTypeProduct) and a possible process plan to obtain a final product from a raw piece. The process plan named processPlan01 is defined as an individual of class VffManufacturingProcessType (i.e. a subclass of VffProcessType) and decomposed into five process segments (as individuals of class VffManufacturingProcessType) characterized by a processing time and a predefined sequence. Moreover, each process segment requires a specific type of production resource and the processing time is modeled as an exponential distribution (see Table 1). VfProductionLine04 ontology contains the factory project that imports and enriches the data provided by the three libraries. The factory project defines the units of measurement, the representation context and world coordinate system where the production site and the building imported from VffLibrary01 are placed. One production line is designed and placed in the building of the factory. The production line consists of seven machines (as individuals of VffMachineryElement) that are typed by the machine types defined in VffLibrary02 (see Table 2) and characterized by a shape representation and a placement. The production line is designed to process the part type defined in VffLibrary17 and is thus organized into five production stages. Each needed production resource type can be provided by one or more machinery element as shown in Table 2. An example of relations between the individuals defined in the test case is shown in Figure 2 where the boxes represent individuals (identified by their local URI and class) and the arcs represent object properties linking the individuals. In particular, it is shown that the process segment ML02 requires the resource type MillingRes02 that can be provided by the machine type MtC (i.e. MS02 or MS03) or by the specific machine MS04.
The presented test case has been serialized in RDF/XML files [
The applicability of the VFDM to model a manufacturing system and its behavior, aiming at evaluating its performance, has been validated focusing the attention on the case of Discrete Event Simulation (DES) and taking as a reference the test case presented in the previous section.
The capability of generating simulation models in an automatic (or
semiautomatic) way has been often considered one of the great challenges in the simulation
of manufacturing systems [
Within the area of the simulation of manufacturing systems, similar issues have
been also addressed in the literature, e.g. by Lorenz and Schulze [
The design of the VFDM was driven by the need of providing a way to unambiguously describe a generic manufacturing system regardless the specific application and, hence, to address some of the lacks of the approaches already proposed in the literature. 4.1
Discrete Event Simulation using Arena
Among the great number of available general-purpose commercial off-the-shelf
(COTS) simulation packages, Arena by Rockwell Automation [
An Arena model is built by dragging modules into the model window and connecting them to define the flow of entities through the model. An example is shown in Figure 4 where parts are generated in the Generate Parts block on the left and then flow to the Machine Part block representing the execution of a certain set of operations.
As shown in Figure 5, a set of resources can be invoked to operate the defined operations. In this case, each part entering the block asks for the resource Machine 1 that has been already defined in the Arena model. A different and more flexible way of defining a process or a sequence takes advantage of the capability of defining sequences. A sequence consists of an ordered list of stations that an entity will visit. For each station in the sequence, values may be assigned to attributes and variables. Moreover, using sequences it is possible to assign different routing to different type of entities, i.e. part types. Each station in the sequence is referred to as a step (or jobstep) and can be characterized by specific attributes (e.g. the processing time).
A DES simulator based on Arena can exploit the interoperability enablers offered by
VFF only if it becomes a VF module (see Sect.1), thus being able to access and
understand the contents of the shared data repository where factory projects and
libraries are formalized according to the VFDM. The information stored into a
VFDM-compliant project can be used to automatically generate an Arena simulation
model only if the Arena data structures are properly mapped to the VFDM classes. This
mapping has been implemented by a software component named Arena-VF Connector
that is attached to Arena and works as a client exploiting the services offered by the
VFM. The Arena-VF Connector has been developed in C++ language and can
import/export ontologies serialized in RDF/XML format. The Arena-VF Connector
makes use of the VF Connector C++ Library that is based on the Redland C libraries
[
The development of a Arena-VF Connector requires a deep analysis because Arena represents a manufacturing system according to proprietary data structures and the relationships between the processes and resources are formalized in a different way compared to the VFDM. In particular, Arena requires each step of a process to be explicitly assigned to a station, thus preventing (grounding on the traditional modeling) the opportunity of defining a group of entities that can be used as resources and postponing the actual assignment of the parts to different stations at run time. The definition of a sequence in Arena allows the modeling of the process steps, but each process steps must be directly linked to a station or a group of station of the same type. To cope with this limitation, the assignment of parts to stations at run time must be explicitly managed in Arena.
Taking as a reference the example in Figure 2, the manufacturing system model can be translated into an Arena model as shown in Figure 6. The resource type (e.g. millingRes02) required by a specific process step (e.g. the milling operation ML02) in the sequence is mapped to a Station block and the assignment to the available objects providing the needed resource (e.g. the machines MS02, MS03, MS04) is explicitly managed through a tree of Branch blocks with as many leaves as the number of machine occurrences that can execute the process step. On the leaves of the branch tree the parts to be manufactured are routed to the specific machine using a Route block after recording the destination in an attribute of the Assign block. Instead of Branch blocks, different methods can be used to model specific assignments policies as well. Finally, the machine occurrences (i.e. MS02, MS03, MS04) are mapped to Station blocks (see the bottom of Figure 6) that are followed by a traditional sequence of Arena blocks, i.e. a Seize block allotting the machine, a Delay block initialized with the processing time and a Release block freeing the machine. Then the manufactured part can be routed to the following process step in the sequence (SEQ keyword in the Route block).
Arena also offers the opportunity of modeling identical machines (e.g. MS02 and MS03) as a single resource with cardinality greater than one. However, by adopting this option the dispatching of parts to the identical machines is internally managed by Arena according to predefined policies that cannot be directly controlled. Further blocks can be added to an Arena simulation model for collecting specific statistics regarding the involved resources. These statistics can be formalized according to the VFDM by using the definitions imported from the IFC standard. For instance, the IfcResourceTime class offers a set of attributes to store the time-related information associated to a resource, e.g. the start and finish time for the assigned workload and the percentage usage during the considered time horizon. The time-related attributes can specify scheduled or actual values, thus showing how the VFDM can be used to support factory planning, performance evaluation and factory monitoring activities.
The performance of the considered manufacturing system has been evaluated in terms of utilization of the different resources and flow time of the parts. The results have been also validated against a simulation model built manually and representing the same manufacturing system (Figure 7). Table 3 reports the results for 10 simulation runs of length 10 days with a warm up of one day, for both the automatically and the manually generated simulation models. The last column in Table 3 reports the 99% confidence intervals for the mean of the difference between the results of the two simulation models. All the confidence intervals contain the value 0, hence, the difference can be considered equal to 0 and, consequently, the two simulation models provides the same results demonstrating the validation of the automatically generated simulation model.
This paper has presented a data model for representing virtual factories, in particular aiming at modeling the complex relationships between physical and logical entities of a manufacturing system. It was shown how the adoption of a shared data model can enhance the interoperability between software tools supporting the design, management and performance evaluation of the factories.
Further developments of the data model are needed to better represent the production logics characterizing a manufacturing system so that the generation of a simulation model can be automated as much as possible. Moreover, the accuracy of the generated simulation models will be improved if the common data model is used to formalize the data coming from the shop-floor, thus closing the loop between the real factory and its virtual representation.
In this paper the VFDM has been used mainly to support interoperability, however further research can be carried out to exploit the enablers of the Semantic Web approach to perform reasoning and enrich the knowledge about specific manufacturing contexts.
Finally, the applicability of the VFF approach needs to be further tested by integrating more software tools for performance evaluation into the framework. Such integration will be supported by the development of programming libraries helping the implementation of customized versions of VF Connector.
The research reported in this paper has been funded by the European Union Seventh Framework Programme (FP7/2007-2013) under the grant agreement No: NMP2 2010228595, Virtual Factory Framework (VFF) and the grant agreement No: 262044, VISION Advanced Infrastructure for Research (VISIONAIR). The authors would like to thank COMPA S.A. (Sibiu, Romania) for kindly providing information for representing the test case.