Working PaperPDF Available


Content may be subject to copyright.
Monitoring and Automating Factories
Using Semantic Models
Niklas Petersen1,2, Michael Galkin1,2,3, Christoph Lange1,2, Steffen Lohmann2,
and S¨oren Auer1,2
1University of Bonn, Germany
2Fraunhofer IAIS, Germany
3ITMO University, Saint Petersburg, Russia
Abstract Keeping factories running at any time is a critical task for ev-
ery manufacturing enterprise. Optimizing the flows of goods and services
inside and between factories is a challenge that attracts much attention
in research and business. The idea to fully describe a factory in a dig-
ital form to improve decision making is called a virtual factory. While
promising virtual factory frameworks have been proposed, their semantic
models lack depth and suffer from limited expressiveness. We propose a
stronger semantic model of a factory, which enables views spanning from
the high level of supply chains to the low level of machines on the shop
floor. The model includes a mapping to relational production databases
to support federated queries on different legacy systems in use. We eval-
uate this model in a production line use case, demonstrating that it can
be used for typical factory tasks such as assembly line identification or
machine availability checks.
1 Introduction
The Industry 4.0 vision [2] aims at digitizing engineering, production and manu-
facturing with the goal of (i) a seamless integration of devices, sensors, machines,
as well as software and IT systems, (ii) increased flexibility thanks to pushing
more intelligence from centralized planning systems to the edge, (iii) increased ef-
ficiency thanks to automated data exchange and analysis within the value chain.
Currently, much information is isolated within different applications, which pre-
vents efficient access for real-time analytics [9]. The ultimate goal of Industry
4.0 (and related initiatives with different names in different regions, such as In-
dustrie du Futur in France or Industrial Internet in the US) is the creation of a
Smart Factory [7].
A Smart Factory is defined as a factory that supports people and machines
in executing their tasks by providing context-aware information. As an example,
the location of information about orders, products, machines, the available work
force and the overall factory are rarely available in a unified database and a
uniform format. The related idea of a Virtual Factory [15] proposes a framework
2 Petersen et al.
Figure 1: Virtual Factory Framework as proposed by [5]
that links all this information together, providing a mirror of the real factory (see
Figure 1) and thus paving the way towards more innovative factory prototyping,
assembly line optimization, product design and mass customization [13].
In order to realize such a virtual factory, a number of interoperability chal-
lenges need to be solved. These include the identification of relevant data and
information, their representation, unified access and interlinking. Of particular
importance in this regard is the support of different views on the data (logistics
and supply chain, manufacturing, quality control, etc.), the support of different
levels of granularity of information representation (operational, strategic, etc.) as
well as the access and integration of various data models and structures as used
by the existing systems and applications (XML, relational, enterprise models,
An integrated approach that provides a holistic view on an enterprise and
its assets (such as factories) has not yet emerged. To fill this gap, we develop
the notion of a Semantic Factory, employing semantic knowledge representation
formalisms and technologies. The rationale is to employ a network of ontologies
and vocabularies as a semantic fabric to represent, interlink and integrate the
heterogeneous information being distributed in a variety of systems and infor-
mation sources (e.g. manufacturing execution, quality management, enterprise
resource planning systems and sensors). For creating the ontology, in a first
step, we represent the static assets of an enterprise, such as its factories, as-
sembly lines, workforce, etc. We then integrate dynamic information including
business processes, shift plans, orders, etc. Finally, we map our model to pro-
duction databases that contain the respective data for the dynamically changing
The mappings are performed in minimally-invasive way, equipping existing
systems of record with semantic interfaces (e.g. using the W3C R2RML stan-
dard for mapping relational data to RDF). As a result, information and data
in a factory can be integrated in a pay-as-you-go fashion, where mappings as
Monitoring and Automating Factories Using Semantic Models 3
well as the network of Semantic Factory ontologies evolve as required by specific
use cases and application scenarios. Examples of use cases are (1) energy man-
agement, so that the energy consumption can be allocated to specific machines,
work orders or customers, or (2) tool management, so that the time required to
equip machines with the required tools is minimized. In addition to introducing
the semantic factory model, we demonstrate with SPARQL queries how deci-
sions within an enterprise can be based on data currently hidden in different
legacy systems. This includes the performance of supply chains, the detection of
suitable assembly lines and the analysis of assets on a map.
The rest of the paper is structured as follows: We provide motivating exam-
ples and derive requirements for an integrated virtual factory representation in
section 2. In section 3, we give an overview of related work. The overall archi-
tecture and the factory ontology as its core model are presented in section 4. In
section 5, we describe the implementation. We evaluate the performance of our
approach in section 6. The paper is concluded in section 7 with an outlook on
future work.
2 Examples and Requirements
A key motivation of our semantic factory model is to establish a holistic and
integrated view on an enterprise in order to reduce the overall complexity and
improve decision making. This includes the workforce, business processes, ma-
chines, shift plans, supply chains, etc. While a lot of this information is already
captured by different IT systems, it is rarely accessible in a combined way with-
out investing significant manual effort. Thus, the goal of this work is to make all
data that is currently stored in various systems available in a unified model to
support users with different roles in decision making.
2.1 Motivating Examples
An example is a factory planner who requires diverse information about order
plans, workforce availability and machine maintenance dates. Another example
is a machinist who needs to know which tools are to be mounted into which
machine, where these tools are located, where the material is stored and what
quality control standards are required during the production process. A con-
troller, on the other hand, wants to keep track of the productivity of a factory
and get an overview of certain key performance indicators (KPIs). These com-
prise, in particular, information on the production time and effort required by
each machine for each product, such as employee effort and energy consumption
To provide all those stakeholders with the information needed to perform
their tasks and optimize decision making, we aim at semantically describing
as many assets of a factory as possible, taking into account information from
different manufacturing systems.
4 Petersen et al.
2.2 Requirements
We elicited the following requirements in the context of a research project for a
global manufacturing company. The company’s objective to gain a better picture
of its assets (e.g. machines and factories) led us to develop an ontology that serves
as the core element in the overall architecture.
From descriptions of the assets provided by the company and from inter-
viewing domain experts, we gained an overview on typical tasks, processes and
problems of each stakeholder. The interviews took place at the company site
and in multiple meetings, where the company’s current IT infrastructure was
described in detail. That way, we gathered requirements for the Semantic Fac-
tory step by step:
Semantic Multi-Modality. The types of data found in a factory context are
diverse. Hence, the representation of various information, including attribute
trees, relational, sensor, tabular, graph and entity data, must be supported.
Multi-Dimensionality. Information along several dimensions must be repre-
sented and captured, such as:
Business processes: Temporal views on diverse business activities are re-
quired to judge the success of an enterprise.
Spatial hierarchies: The exploration of assets from a geographical perspec-
tive must be possible to increase the findability of said assets and related
Lifecycle: Product and business lifecycles must be represented to support
strategy management and business innovation.
Multi-Granularity. Views on different levels of detail must be provided:
Components: Instant access to sensor and component data must be enabled
to support possible intervention measures.
Factories: To decrease the complexity of factories, master and operational
data needs to be accessible in a singular view.
Organization: A big picture of all business units is needed, including their
hierarchies and responsibilities.
Traceability and Integration. Data and information is currently spread
across various systems, such as manufacturing execution systems, quality as-
surance systems, enterprise resource planning systems, etc. It is important to in-
tegrate all relevant information from these systems, while preserving the systems’
record-keeping character. When integrating information from these systems, the
provenance of the data must be preserved, and changes to information in the
source systems must be reflected in the integrated views, wherever possible in
real time.
Monitoring and Automating Factories Using Semantic Models 5
3 Related Work
There are two categories of related work: i) existing frameworks aim to describe
factories as completely as possible, and ii) existing ontologies representing as-
sembly lines.
Terkaj et al. [14] propose a Virtual Factory Data Model represented as an
OWL ontology based on the Industry Foundation classes (IFC)4standard. The
purpose of their ontology is to describe business processes that involve ma-
chines requiring specific resources. However, the advantage of modeling each
concept twice, once as a class whose instances represent real occurrences (e.g.
IfcProduct,IfcProcess), and then as a class whose instances are sup-
posed to describe generic objects and types (e.g. IfcTypeProduct “describes
a generic object type that can be related to a geometric or spatial context”,
IfcTypeProcess “describes a generic process type to transform an input into
output”) is not clearly justified. A significant part of the ontology employs such
a logical duplication which is misleading for non-experts. Furthermore, the ratio-
nale of proposing property classes such as VffProcessProperties to “char-
acterize processes” instead of using object or data properties is not described.
Chen et al. [4] propose a multi-agent framework to monitor and control dy-
namic production floors. The ontology, serialized in XML, is optimized for the
communication between different agents. It describes radio-frequency identifica-
tion (RFID) tags [17] attached to factory objects and addresses requirements
specific to a bike manufacturing use case. Although RFID sensors are an impor-
tant component of Industry 4.0, the purely XML-based ontology without logical
formalisms behind, such as RDF or OWL, lacks semantics and does not allow
for universal and convenient querying.
uscher et al. [3] introduce the Virtual Production Intelligence platform
based on the Condition Based Factory Planning (CBFP) approach. The authors
developed an OWL-based CBFP ontology advocating “the decoupling of domain
business logic and the technical implementation of a planning system” [3]. The
ontology is relatively small, consisting of only five classes. As it is not available
online, we consider the CBFP ontology rather abstract and superficial. Detailed
evaluation and experiments are not provided, making it therefore hard to assess
the practical contribution of the work.
Kim et al. [8] propose an OWL ontology and an information sharing frame-
work to allow collaborative assembly design. The heart of the ontology is the
assembly line and its direct environment. The ontology defines assemblies and
constraints leveraging capabilities of SWRL and OWL. However, the lack of a
published online version prevented us from reusing it. Nevertheless, the concep-
tual design influenced the one of our ontology, i.e., several concepts in the classes
hierarchy and a few properties have been retrofitted.
Ameri et al. [1] propose the Digital Manufacturing Market (DMM), a se-
mantic web-based framework for agile supply chain deployment. DMM employs
the Manufacturing Service Description Language (MSDL) at a semantic level.
4 overview
6 Petersen et al.
MSDL is an upper-level ontology expressed in OWL-DL. Description Logic is
extensively used to characterize supply and demand entities on several levels,
such as the supplier, shop, machine and process levels. However, the granularity
and ramification (especially for an upper-level ontology) impose restrictions on
the usability, i.e., only a domain expert would have enough expertise to create a
working model with accompanying queries. Furthermore, the ontology is again
not available online, which prevented us from performing a thorough semantic
analysis and considering an adaptation of concepts.
Zuehlke [18] introduces the SmartFactory initiative, which comprises best
practices from the technical, architectural, planning, security and human di-
mensions. The initiative is envisioned to define and elaborate on the concept of
factory-of-things as a vision of future manufacturing. Although semantic services
involving ontologies and knowledge bases are claimed to be a part of the concept,
the author does neither provide any examples nor references of such ontologies.
Therefore, the presented concept is rather an implementation roadmap than a
technical contribution.
4 Semantic Factory Architecture and Ontology
Based on the requirements presented in section 2, we designed an architecture
for a semantic factory application (see Figure 2). Its core is a factory ontology,
which describes real world objects, such as employees, machines and factories,
locations of assets, and their relations with each other. Operational data, such as
information about work orders, machine sensor and process data, is also covered
by the factory ontology; in practice, this data is dynamically mapped from the
respective databases to the ontology.
The ontology is made available through an RDF triple store. Different appli-
cations can execute queries on the data which is expressed as RDF or available
in relational databases. Finally, for geospatial data, an external map provider
service is used for drawing, among others, factories on a world map.
The following subsections explain how we developed the factory ontology
following the methodology proposed by Uschold et al. [16]. We first defined the
purpose and scope of the ontology; then, we captured step-by-step the domain
knowledge, conceptualized and formalized the ontology and aligned it with exist-
ing ontologies. Finally, we evaluated the ontology by measuring the performance
of certain queries for different sized datasets (see section 6).
4.1 Purpose and Scope
The purpose of the ontology is to provide a holistic view of an enterprise. This is
realized by implementing the requirements specified in section 2. The intended
users are different stakeholders of an enterprise, such as managers, machine op-
erators and controllers. Each of them needs different information to effectively
and efficiently perform the corresponding tasks and duties. Thus, the ontology
enables viewing the factory from different perspectives to support each class of
stakeholders in their decision making.
Monitoring and Automating Factories Using Semantic Models 7
Figure 2: Semantic Factory architecture
4.2 Capturing Domain Knowledge
We captured the domain knowledge in three ways:
1. The client provided us with descriptive material of the domain, including
maps of factories, descriptions of machines and work orders, process infor-
mation, sensor data and tool knowledge. The types of input material ranged
from formatted and unformatted text documents to spreadsheets and SQL
2. A live demonstration of a particular machine execution was given, includ-
ing a discussion of further contextual information which was missing in the
material. In subsequent meetings, further open questions were clarified and
concrete use cases of the ontology were discussed.
3. We reviewed relevant existing ontologies with the intention to build upon
available conceptualizations and formalizations of domain knowledge.
4.3 Conceptualizing and Formalizing
The resulting ontology comprises 86 classes, 73 object properties and 142 datatype
properties. Since it has been designed to support an industrial project with sen-
sitive business logic descriptions, not the entire ontology could be made publicly
8 Petersen et al.
Figure 3: Core concepts of the Factory Ontology
available. However, the part of the ontology that can be published is accessible
via a permanent URL5. In the following, we describe the ontology, starting from
the high-level organizational layer to the low-level machine and sensor layer. The
core concepts of the ontology are depicted in Figure 3.
In any concrete scenario, the class Enterprise is instantiated to represent
the organization to which all further resources belong. Possible instances may
be “Volkswagen”, “General Electric” or “Samsung”. Inter-organizational supply
chains could involve multiple enterprises. Different production locations of an
enterprise are described using the class Plant. A plant can comprise one or
many Buildings, such as an office building, a factory building or a warehouse
building. Typically, we can assume that each building serves a single function,
though other configurations can also be represented, as OWL ontologies support
multi-membership. Therefore, the subclasses of Building are not defined to be
disjoint in the ontology.
Each Factory building may have one or multiple AssemblyLines. An
assembly line usually consists of a sequence of multiple machines. The classes
Monitoring and Automating Factories Using Semantic Models 9
MillingMachine and MetalLathe are examples of subclasses of the abstract
class Machine. Machines can be configured to use certain Tools to produce
specific work pieces. As an example, a milling machine may be equipped with
different lathes or end mills of varying granularity. WorkPieces represents ev-
erything which is an output of a machine. Once a work piece reaches its final
stage of production, it becomes a product and is ready for shipping. Plants,
factories and machines can have a representation of their geographical location
(ngeo:Geometry6), which front-end applications can use to display them on a
As a next step, we describe the part of the ontology that represents the every-
day operation of a factory. Employees are instances of the class foaf:Person.
Each employee is qualified to operate specific machines. The class WorkOrder
describes orders driven by customers. Each such order contains one or more
Processes, which need to be executed to fully complete the order. Typical
processes may be the configuration of a machine, the execution of a machine,
quality control, etc. Each order has a due date and a sales price. The proper-
ties requiresMachine,hasInput,needsConfig,canOperate are used
together with Process class. For example, they define which Machine is re-
quired for that process together with the needed configuration (tools assembled)
and the input Material. Finally, it is described which employees have the skills
to operate the respective machine.
4.4 Aligning with Existing Ontologies
The ontology includes concepts and properties from well-known ontologies. The
Semantic Sensor Ontology7provides us with a rich description of sensors, their
measurements, devices and related concepts. The workforce and employees are
described based on definitions by the Friend-of-a-friend (FOAF)8ontology. Co-
ordinates of factories and machines are based on the latitude and longitude def-
initions of the W3C Geo Vocabulary 9. Finally, we reused geometrical concepts,
such as the representation of polygons, from the NeoGeo Geometry Ontology 10
5 Implementation and Application
We developed a software system that implements the presented semantic factory
architecture and ontology and applied it to industry data. In this section, we
first describe the front-end and back-end implementation. Then, we illustrate
its usage by presenting various SPARQL queries that retrieve information for
different production management tasks.
6Prefixes are defined according to
10 Petersen et al.
5.1 Front-end
The front-end is realized as a web application to facilitate access from different
devices. The decision is motivated by the diversity of IT systems, platforms and
devices usually deployed in a factory. The application uses the web framework
AngularJS 11, which follows the model-view-controller design pattern to separate
logic from representation. We created multiple views to address the collected
The map view (Figure 4a) projects all instances (e.g. buildings, machines)
with a geographical representation on a map by making use of the map provider
MapBox API 12 . This API offers map tiles based on the open geographical
database OpenStreetMap13 . The projection itself is realized using the leaflet.js14
JavaScript library. Coordinates in the factory ontology are represented using the
NeoGeo Geometry Ontology15 concepts and properties translated into leaflet
geographical objects to be drawn on the map.
(a) Map View (b) Machine View
Figure 4: Implementation of the Semantic Factory (geographic views).
Further information, such as the person currently operating a machine, which
order is executed or the status of a machine, is provided in the machine view
that can be opened from the map view (see Figure 4b) or independently by the
machine operator. Besides static information, the pop-up contains also links to
operational views and services. For example, the machine operator can follow
the order link to retrieve additional information of that order. Furthermore,
based on the tools required for the next machine operation, the “Find available
Monitoring and Automating Factories Using Semantic Models 11
tools” button points to the respective geographical location of the tools in the
factory. The links “Visualize”, “Analyze” and “Predict” point to external pages
that provide additional graphical content about the machine, machine usage
indicators and prediction dates when machine parts are worn out and need to
be replaced.
5.2 Back-end
The back-end consists of a Python web server16 that supports REST API calls
from the front-end. Each request triggers the generation of SPARQL queries
executed either on the factory ontology or the production databases. To provide
access to the ontology, the Python library rdflib17 is used.
Access to the relational production databases is realized using the d2rq18
system. d2rq provides a generator for creating an RDF mapping file for the
database tables and columns. Using the mapping file, incoming SPARQL queries
are translated ad-hoc into SQL queries and are executed on the respective
database. Thus, d2rq acts as a gateway between the web server and relational
Once the data is obtained, it is returned in JSON data format19 and processed
by the respective front-end controller.
5.3 Factory Queries
In the following, we provide a set of SPARQL queries that demonstrate the usage
of the factory ontology.
Order Feasibility Check. Listing 1.1 shows a query that determines if certain
machines in the factory are free to use or already scheduled for other production
plans. Each factory work order contains a list of tasks to be completed by a
different machine. Thus, each needed machine is checked for its availability. Only
if all machines are available, the query returns a positive answer such that the
work order can be started. The query always checks the current state of the
3# get tasks
4?order a:WorkOrder .
5?order :requiredMachines ?machineList .
6?machineList rdfs:member ?machine.
12 Petersen et al.
8# check if the needed machines are free
9EXISTS { ?machine :isFree false } .
10 }
Listing 1.1: Order feasibility check
Retrieve Geographical Coordinates. Listing 1.2 shows a query to retrieve
the machines, their name and their coordinates. The outline of a machine is
conceived as a polygon, represented as an rdf:List of geographical points,
each with latitude and longitude, which is linked to the machine using the
ngeo:posList datatype property. This information is returned to the front-
end to be projected on the world map.
1SELECT ?machine ?label
2(GROUP_CONCAT( ?lat ; separator=";")AS ?lats)
3(GROUP_CONCAT(?long ; separator=";")AS ?longs)
5?machine rdfs:label ?label .
6?machine ngeo:posList/rdf:rest*/rdf:first ?point .
7?point geo:lat ?lat .
8?point geo:long ?long .
9}GROUP BY ?machine ?label
Listing 1.2: Retrieve geographical coordinates of machines
Suitable Assembly Lines. Listing 1.3 shows a query to find suitable assembly
lines with regard to their sequence. Assembly lines that contain more machines
than required but fulfill the correct order are still considered suitable. Thus,
certain stations may be skipped within an assembly line.
Suppose, for example, that the machines 2and 4are required for an or-
der. Suitable assembly lines include those having the following sequences of ma-
chines: 2,4 or 1,2,3,4,5. Sequences such as 4,2 or 4,3,2 are considered
The query itself works as follows: First, assembly line candidates are filtered
(MINUS) based on whether they contain the required machines in the work order.
Second, of those assembly lines, the position of the needed machines is calculated.
This is achieved by preparing the needed order (?reqSequence) and then
retrieving the position of each machine in that order (?machineSeq). Third,
these sequences are concatenated into strings and finally checked by a regular
expression if the sequence is increasing.20
20 The query is limited to sequences of up to 9 machines but may be extended.
Monitoring and Automating Factories Using Semantic Models 13
1SELECT ?assemblyLine {
2?assemblyLine :machineList ?lists .
4# filter all assembly lines with the wrong order
5FILTER REGEX(?seq,"ˆ0*1*2*3*4*5*6*7*8*9*$")
7# concatenate order of the lists into a string
8{SELECT ?lists (GROUP_CONCAT(?machineSeq; separator="")
9AS ?seq)
11 #Machine Sequence, _sorted_ by required order Sequence
12 {SELECT ?lists ?machineSeq {
13 {SELECT ?lists ?machineInstance ?machines
14 (STRAFTER(STR(?memberProp), "_")AS ?machineSeq)
15 {?lists rdfs:member ?machineInstance .
16 ?lists ?memberProp ?machineInstance .
17 ?machineInstance a?machines . }}
19 # get required Sequence of the Work Order
20 {SELECT ?machines (STRAFTER(STR(?prop), "_")
21 AS ?requiredSeq) {
22 :sampleOrder :requiredMachines ?orderList .
23 ?orderList ?prop ?machines .
24 FILTER (STRSTARTS(STR(?prop), STR(rdf:_)))}
25 ORDER BY ?requiredSeq}
27 # Identify Assembly Lines Candidates
28 {SELECT ?lists ?assembly
29 {?assembly d:machineList ?lists.}}
31 {SELECT ?lists {
32 ?lists ardf:Seq .
33 ?workOrder :requiredMachines ?neededMachineList .
34 ?neededMachineList rdfs:member ?machineType .
35 FILTER NOT EXISTS{?lists (rdfs:member/a) ?machineType .}}}
36 }ORDER BY ?lists ?requiredSeq
37 }GROUP BY ?lists }}
Listing 1.3: Find suitable assembly lines
6 Evaluation
We evaluated our factory ontology and software application by testing the per-
formance of the SPARQL queries introduced in subsection 5.3. For that, we
prepared multiple datasets, consisting of 10K, 100K, 1M, 2M and 5M triples.
The datasets contain generated test data based on our factory ontology, such as
order information, workforce details, assembly lines, etc.
14 Petersen et al.
The queries were executed using the ARQ SPARQL processor version 2.13.021.
The machine we used for the experiment contains 8GB of RAM, 256GB SSD
and an Intel i7-3537U CPU with 2.00GHz.
Figure 5 depicts the results of the performance evaluation. While the growth
for the “Retrieve machine coordinates” query of Listing 1.2 is linear, a response
time of 25 seconds in larger datasets is not satisfactory for a front-end applica-
tion. Thus, large datasets should be split to keep the execution time end-user
10K 100K 1M 2M 5M
Dataset size in triples
Execution time in seconds
Retrieve machine coordinates (1.2)
Order feasibility check (1.1)
Suitable assembly lines (1.3)
Figure 5: Query execution performance
Similarly, as in the previous query, the execution time for the “Order Feasi-
bility Check” query of Listing 1.1 grows linearly. While the overall performance
is slightly better, one should nevertheless keep machine status data in an isolated
Finally, for the large “Suitable Assembly Lines” query of Listing 1.3, it is
rather surprising that the execution time is quite similar to the short previous
queries. As before, with a linear growth, the performance evolution becomes
predictable and it is recommended to split instance data depending on certain
Monitoring and Automating Factories Using Semantic Models 15
Overall, ontology-centered web applications are feasible with a satisfactory
performance. As a common rule of thumb, the acceptable response time for
complex operations is less than 10 seconds in order to keep the user’s attention
on the task [10]. Thus, certain crucial instance data should be kept in different
triple stores to stay below that threshold.
7 Conclusion and Future Work
The use of data-centric approaches in engineering, manufacturing and produc-
tion is currently a widely discussed topic (cf. Industry 4.0, smart manufacturing
or cyber-physical systems initiatives). The complexity of data integration in gen-
eral is perceived to be one of the major bottlenecks of the field. A key issue in
engineering, manufacturing and production is to be able to efficiently and effec-
tively manage factories.
This paper described an ontology and an integration infrastructure to obtain
a holistic view of the status of a factory from different perspectives. We see the
work presented in this paper as a first step towards establishing an ontology-
based integration approach for manufacturing, which is centered around a com-
mon information model, but at the same time supports the management of data
in a decentralized manner in the existing systems of record. The integration
follows a loosely-coupled architecture, where the decentralized data sources are
mapped on demand to the factory ontology. The factory ontology is not supposed
to be a fixed, monolithic schema, but rather a flexible, evolving and interlinked
knowledge fabric. For this purpose, we have developed the collaborative vocab-
ulary development methodology and support environment VoCol [6].
We see a number of directions for future work. In particular, the seman-
tic factory approach could be expanded from single factories to an integration
approach covering the entire enterprise as well as supply networks (e.g. based
on the SCOR model [11]) involving a large number of organizations. Another
promising direction of future work is the exploitation of the integrated data for
advanced analytics and forecasting [12].
Acknowledgments. This work has been supported by the German Ministry
for Education and Research under grants 01IS14019C for the project LUCID
and 01IS15054 for the project Industrial Data Space.
1. Ameri, F., Patil, L.: Digital manufacturing market: a semantic web-based frame-
work for agile supply chain deployment. Journal of Intelligent Manufacturing 23(5),
1817–1832 (2012)
2. Brettel, M., Friederichsen, N., Keller, M., Rosenberg, M.: How virtualization, de-
centralization and network building change the manufacturing landscape: An in-
dustry 4.0 perspective. International Journal of Mechanical, Industrial Science and
Engineering 8(1), 37–44 (2014)
16 Petersen et al.
3. uscher, C., Voet, H., Krunke, M., Burggr¨af, P., Meisen, T., Jeschke, S.: Semantic
information modelling for factory planning projects. Procedia CIRP 41, 478–483
4. Chen, R.S., Tu, M.A.: Development of an agent-based system for manufacturing
control and coordination with ontology and rfid technology. Expert systems with
applications 36(4), 7581–7593 (2009)
5. Ghielmini, G., Pedrazzoli, P., Rovere, D., Terkaj, W., Bo¨er, C.R., Dal Maso, G.,
Milella, F., Sacco, M.: Virtual factory manager of semantic data. In: Proceedings of
7th International Conference on Digital Enterprise Technology (DET ’11) (2011)
6. Halilaj, L., Grangel-Gonz´alez, I., Coskun, G., Auer, S.: Git4voc: Git-based version-
ing for collaborative vocabulary development. In: 10th International Conference on
Semantic Computing (ICSC ’16). pp. 285–292. IEEE (2016)
7. Hermann, M., Pentek, T., Otto, B.: Design principles for industrie 4.0 scenarios: a
literature review. Technische Universit¨at Dortmund, Dortmund (2015)
8. Kim, K.Y., Manley, D.G., Yang, H.: Ontology-based assembly design and infor-
mation sharing for collaborative product development. Computer-Aided Design
38(12), 1233–1250 (2006)
9. Newman, D., Gall, N., Lapkin, A.: Gartner defines enterprise information archi-
tecture. Gartner Group (2008)
10. Nielsen, J.: Response times: The 3 important limits. Usability Engineering (1993)
11. Petersen, N., Grangel-Gonz´alez, I., Coskun, G., Auer, S., Frommhold, M., Tramp,
S., Lefran¸cois, M., Zimmermann, A.: SCORVoc: Vocabulary-based information in-
tegration and exchange in supply networks. In: 10th International Conference on
Semantic Computing (ICSC ’16). pp. 132–139. IEEE (2016)
12. Petersen, N., Lange, C., Auer, S., Frommhold, M., Tramp, S.: Towards federated,
semantics-based supply chain analytics. In: 19th International Conference on Busi-
ness Information Systems (BIS ’16). pp. 436–447. Springer (2016)
13. Sacco, M., Pedrazzoli, P., Terkaj, W.: VFF: virtual factory framework. In: Pro-
ceedings of 16th International Conference on Concurrent Enterprising (ICE ’10).
pp. 21–23 (2010)
14. Terkaj, W., Urgo, M.: Virtual factory data model to support performance evalu-
ation of production systems. In: Proceedings of the Workshop on Ontology and
Semantic Web for Manufacturing (OSEMA ’12). CEUR-WS, vol. 886, pp. 24–27
15. Upton, D.: The real virtual factory. Harvard Business Review 74(4), 123–133 (1996)
16. Uschold, M., Gruninger, M., et al.: Ontologies: Principles, methods and applica-
tions. Knowledge engineering review 11(2), 93–136 (1996)
17. Want, R.: An introduction to RFID technology. IEEE Pervasive Computing 5(1),
25–33 (2006)
18. Zuehlke, D.: SmartFactory—towards a factory-of-things. Annual Reviews in Con-
trol 34(1), 129–138 (2010)
... It is a technology enabler that enables data translation into both machine-readable and human interpretable formats to link the different objects in an automated way. Various researchers have studied the Semantic Web combined with the supply chain through different approaches [6][7][8]. The Semantic Web contributes to limiting the foreseen volatile nature of the virtual supply chain [9]. ...
Conference Paper
Full-text available
Supply Chain Collaboration is a promising approach for network members to exchange information in a trustworthy environment in order to obtain better supply chain performance. Driven by technology development, there are many technology candidates related to information sharing for supply chain management to obtain better supply chain performance. Semantic Web technologies have received increased attention from industry and academic domains. However, there is a lack of research that focuses on possible application of Semantic Web technologies. This work presents a practical approach investigating impact of Semantic Web as an enabler for the semiconductor supply chain. In this study, we show how Semantic Web could improve the supply chain collaboration to increase Supply Chain Performance in Semiconductor chain.
... • A factory planner needs input from diverse sources regarding order plans, machine maintenance schedules, workforce availability, and so on. 3 • A field technician must quickly troubleshoot an onsite industrial asset, and is seeking a solution that combines a summary of the problem, including difficulty and time estimates; links to relevant manuals and necessary parts; additional physical tools to resolve the problems and the current location of these tools; and, if the problem is difficult to resolve, additional support from people with the necessary expertise. • During production, a machinist needs to know which tools are required to perform the task at hand, the location of these tools and materials, and quality control standards to be adhered to. 3 ...
Full-text available
AI techniques combined with recent advancements in the Internet of Things, Web of Things, and Semantic Web-jointly referred to as the Semantic Web-promise to play an important role in Industry 4.0. As part of this vision, the authors present a Semantic Web of Things for Industry 4.0 (SWeTI) platform. Through realistic use case scenarios, they showcase how SweTI technologies can address Industry 4.0s challenges, facilitate cross-sector and cross-domain integration of systems, and develop intelligent and smart services for smart manufacturing.
... The information model aims at a holistic description of the company, its assets and information sources. The core of the model is based on a factory ontology we developed in a previous project[18], which describes real world objects from the factory domain, including factories, employees, machines, their locations and relations to each other, etc. In addition, the information model comprises the mappings between ontologies that represent the data sources (i.e., SD, BOM, MES) and their corresponding schemes. ...
Conference Paper
The digitization of the industry requires information models describing assets and information sources of companies to enable the semantic integration and interoperable exchange of data. We report on a case study in which we realized such an information model for a global manufacturing company using semantic technologies. The information model is centered around machine data and describes all relevant assets, key terms and relations in a structured way, making use of existing as well as newly developed RDF vocabularies. In addition, it comprises numerous RML mappings that link different data sources required for integrated data access and querying via SPARQL. The technical infrastructure and methodology used to develop and maintain the information model is based on a Git repository and utilizes the development environment VoCol as well as the Ontop framework for Ontology Based Data Access. Two use cases demonstrate the benefits and opportunities provided by the information model. We evaluated the approach with stakeholders and report on lessons learned from the case study.
... Petersen et al. [17] present a semantic model for representing smart factories as ontology instances. However, their system is limited to monitoring applications. ...
Conference Paper
Full-text available
In this applied research paper, we describe an architecture for seamlessly integrating factory workers in industrial cyber-physical production environments. Our human-in-the-loop control process uses novel input techniques and relies on state-of-the-art industry standards. Our architecture allows for real-time processing of semantically annotated data from multiple sources (e.g., machine sensors, user input devices) and real-time analysis of data for anomaly detection and recovery. We use a semantic knowledge base for storing and querying data ( and the Business Process Model and Notation (BPMN) for modelling and controlling the process. We exemplify our industrial solution in the use case of the maintenance of a Siemens gas turbine. We report on this case study and show the advantages of our approach for smart factories. An informal evaluation in the gas turbine maintenance use case shows the utility of automated anomaly detection and handling: workers can fill in paper-based incident reports by using a digital pen; the digitised version is stored in metaphacts and linked to semantic knowledge sources such as process models, structure models, business process models, and user models. Subsequently, automatic maintenance and recovery processes that involve human experts are triggered.
Conference Paper
Full-text available
Supply Chain Management aims at optimizing the flow of goods and services from the producer to the consumer. Closely interconnected enterprises that align their production, logistics and procurement with one another thus enjoy a competitive advantage in the market. To achieve a close alignment, an instant, robust and efficient information flow along the supply chain between and within enterprises is required. However, less efficient human communication is often used instead of automatic systems because of the great diversity of enterprise systems and models. This paper describes an approach and its implementation SCM Intelligence App, which enables the configuration of individual supply chains together with the execution of industry accepted performance metrics. Based on machine-processable supply chain data model (the SCORVoc RDF vocabulary implementing the SCOR standard) and W3C standardized protocols such as SPARQL, the approach represents an alternative to closed software systems, which lack support for inter-organizational supply chain analysis. Finally, we demonstrate the practicality of our approach using a prototypical implementation and a test scenario.
Full-text available
Nowadays, one of the main challenges in factory planning is the consistent and coherent information processing along planning processes. Despite the current efforts in the fields of Virtual Production and Digital Factory, planning and simulation applications mostly support only analyses and optimizations of single planning aspects. To match nowadays challenges, planners require solutions that provide an integrated view on all data generated along planning processes to evaluate planning scenarios in advance and to achieve increasing production quality and efficiency. We have developed an essential solution by combining the ‘Condition Based Factory Planning’, a flexible planning approach which decomposes the process into standardized planning modules, and the ‘Virtual Production Intelligence’. This fusion creates a basis for the integration and analysis of data aggregated along production (planning) processes. The current information model provides factory planners to perform integrated analyses of process characteristics on the bases of module parameters to increase transparency of information flows.
Conference Paper
Full-text available
Advanced, highly specialized economies require instant , robust and efficient information flows within its value-added and Supply Chain networks. Especially also in the context of the recent Industry 4.0, smart manufacturing or cyber-physical systems initiatives more efficient and effective information exchange in supply networks is of paramount importance. The Supply Chain Operation Reference (SCOR) is a cross-industry approach to lay the groundwork for this goal by defining a conceptual model for Supply Chain related information. Semantics-based approaches could facilitate information flows in supply networks, and enable to analyze, monitor and optimize Supply Chains (in particular for robustness). This paper first reviews existing formalizations of the Supply Chain Council's SCOR standard. It then introduces the SCORVoc RDFS vocabulary which fully formalizes the latest SCOR standard, while overcoming the identified limitations of existing work. SCORVoc is operationalized by a set of SPARQL queries, that enable to evaluate metrics and key performance indicator (KPIs) defined by SCOR, on-the-fly, in an information systems that adheres to the vocabulary. Finally, we define concrete test scenarios and implement a synthetic benchmark to demonstrate the practicality of SCORVoc.
Full-text available
The growing importance of manufacturing SMEs within the European economy, in terms of Gross Domestic Product and number of jobs, emphasizes the need of proper ICT tools to support their competitiveness. Major ICT players already offer one-does-all Product Lifecycle Management suites, supporting several phases of the product-process-plant definition and management. However, these do also show consistent shortcomings in terms of SME accessibility, degree of personalization and they often lack of an acceptable level of interoperability. These problems are being addressed by the development of a Virtual Factory Framework (VFF), within an EU funded project. The approach is based on four pillars: 1) Semantic Shared Data Model, 2) Virtual Factory Manager (VFM), 3) Decoupled Software Tools that lay on the shared data model and can interact through the VFM, 4) Integration of Knowledge. This paper will focus on the Virtual Factory Manager, proposing an evolution of the former VFF second Pillar (Sacco et al, 2010), that acts as a server supporting the I/O communications within the framework and its stored knowledge for the decoupled software tools needing to access its repository.
Collaborative vocabulary development in the context of data integration is the process of finding consensus between experts with different backgrounds, system understanding and domain knowledge. The complexity of this process increases with the number of people involved, the variety of the systems to be integrated and the dynamics of their domain. In this paper, we advocate that the usage of a powerful version control system is the heart of the problem. Driven by this idea and the success of the version control system Git in the context of software development, we investigate the applicability of Git for collaborative vocabulary development. Even though vocabulary development and software development have much more similarities than differences, there are still important obstacles. These need to be considered in the development of a successful versioning and collaboration system for vocabulary development. Therefore, this paper starts by presenting the challenges we are faced with during the collaborative creation of vocabularies and discusses its distinction to software development. Drawing from these findings, we present Git4Voc which comprises guidelines on how Git can be adopted to vocabulary development. Finally, we demonstrate how Git hooks can be implemented to go beyond the plain functionality of Git by realizing vocabulary-specific features like syntactic validation and semantic diffs.
Conference Paper
Collaborative vocabulary development in the context of data integration is the process of finding consensus between the experts of the different systems and domains. The complexity of this process is increased with the number of involved people, the variety of the systems to be integrated and the dynamics of their domain. In this paper we advocate that the realization of a powerful versioning control system is the heart of the problem. Driven by this idea and the success of Git in the context of software development, we investigate the applicability of Git for collaborative vocabulary development. Even though vocabulary development and software development have much more similarities than differences there are still important differences. These need to be considered within the development of a successful versioning and collaboration system for vocabulary development. Therefore, this paper starts by presenting the challenges we were faced with during the creation of vocabularies collaboratively and discusses its distinction to software development. Based on these insights we propose Git4Voc which comprises guidelines how Git can be adopted to vocabulary development. Finally, we demonstrate how Git hooks can be implemented to go beyond the plain functionality of Git by realizing vocabulary-specific features like syntactic validation and semantic diffs.