=Paper=
{{Paper
|id=Vol-2941/paper19
|storemode=property
|title=Requirements for an Ontology of Digital Twins
|pdfUrl=https://ceur-ws.org/Vol-2941/paper19.pdf
|volume=Vol-2941
|authors=Claudio Barros,Rebecca Salles,Eduardo Ogasawara,Giancarlo Guizzardi,Fabio Porto
|dblpUrl=https://dblp.org/rec/conf/i-semantics/BarrosSOGP21
}}
==Requirements for an Ontology of Digital Twins==
https://ceur-ws.org/Vol-2941/paper19.pdf
Requirements for an Ontology of Digital Twins
Claudio Barros1 , Rebecca Salles2 , Eduardo Ogasawara2 ,
Giancarlo Guizzardi3,4 , and Fabio Porto1
1
DEXL Lab, National Laboratory for Scientific Computing, Brazil
2
Federal Center for Technological Education of Rio de Janeiro, Brazil
3
Free University of Bozen-Bolzano, Italy
4
University of Twente, The Netherlands
Abstract. Digital twin connects concrete systems to digital representa-
tions, encoding the real world using software systems, tools and models.
Therefore, digital twins should comprise abstractions, formal namings
and definitions of categories, properties and relations between concepts,
data and entities substantiating one, many or every element of some do-
main of interest. Considering the possible synergies between digital twins
and ontology, and the growing demand for connecting the physical and
the virtual world through explicit ontological grounding, our work pro-
poses preliminary discussions about requirements to build an ontology of
digital twins. We outline some relevant topics both in the field of digital
twins and ontology that are important for the proposal of core reference
ontology in the field. We also explore these requirements in detail, from
the conception and creation of the virtual environment and the digital
twins, to the synchronization between digital world and real world, in
addition to computational services, including visualization, prediction,
and prescription . Finally, we present topics for future work. ∗
Keywords: Digital twin · Ontology · UFO · Requirements
1 Introduction
Digital twins connect real systems to virtual representations, codifying the phys-
ical world using computer systems, tools and models, bringing insight about the
current state of a real entity, providing tools to analyse and monitor physical
entities from different perspectives, and predicting future states by combining
machine learning models and simulations based on first principles. These digital
models create the information necessary for decision-making in the real world,
assisting applications in the industry, healthcare domain, agriculture and en-
ergy [17].
Ontology is the branch of metaphysics that studies different modes of exis-
tence, thus, providing theories comprising systems of categories, properties and
relations. Foundational ontologies define domain-independent categories about
∗ Copyright © 2021 for this paper by its authors. Use permitted under Creative Commons
License Attribution 4.0 International (CC BY 4.0).
2 C. Barros et al.
the real world, constituting toolboxes of reusable information modeling primi-
tives for building application ontologies in specific domains [19].
The Unified Foundational Ontology (UFO) is a well-established foundational
ontology composed of three main parts: (i) UFO-A [11], an ontology of en-
durants, describing entities including particulars, universals, substances, mo-
ments, quality, relations, and situations; (ii) UFO-B [15], an ontology of per-
durants, defining concepts such as events and time; and (iii) UFO-C [13], an
ontology of social entities, built on top of UFO-A and UFO-B, to systematize
social concepts, such as plan, action, goal, agent, intentionality, and commit-
ment, among others. Grounded in the UFO, several ontologies were proposed to
bring notions towards specific domains of applications, such as an ontology of
artifacts [23] and ontologies in the context of software development, execution
and process [13, 3].
Therefore, it is possible to map systems to their digital twin counterparts
through explicit ontological grounding, capturing: (i) fundamental aspects of
their existence, such as conception, creation, prototyping, functionalities and dis-
posal; (ii) computational aspects, including development and deployment, data
processing, and model execution; and (iii) application-oriented aspects, consider-
ing the impact of digital twins in the market, in the business, and in the lifecycle
of products and processes. The paper proposes preliminary discussions on re-
quirements to build an ontology of digital twins, mainly focused on foundational
aspects, in addition to presenting computational aspects at a higher level.
The paper is organized as follows: Section 2 outlines the fundamentals of
digital twins and ontology, presenting topics which are relevant to further discuss
the requirements for an ontology of digital twins. Section 3 explores in details
these requirements, from the conception of the virtual environment and the
digital twins, and their creation, to the synchronization between digital twin
and real world, and computational services, including visualization, prediction,
and prescription . Section 4 offers an insight into related work. Finally, Section
5 concludes and points to further future work.
2 Fundamentals of Digital Twins
A digital twin is a virtual representation of the real world, whose data flow
between the physical object and its digital counterpart is integrated and auto-
mated in both directions. Therefore, it reflects the conditions of the physical
object through data captured from the physical world, and provides feedback by
offering computational services including visualization, control, prediction and
what-if scenario simulation in the virtual world, existing along the life-cycle of
the physical entity [24].
The use of twins to monitor physical objects and test different conditions on
them dates back to NASA’s Apollo program, simulating the conditions of a ve-
hicle in the space using another vehicle on Earth, mirroring the flight conditions
as accurately as possible [24]. In the 1990s, Gelernter [7] described the idea of
recreating real world inside a virtual environment, and Grieves first proposed
Requirements for an Ontology of Digital Twins 3
a digital twin as a model in the early 2000s as a concept for Product Lifecycle
Management (PLM), asserting that all real-world systems are dual in nature,
having both a physical embodiment and a virtual representation [9].
2.1 Digital Twin Models and Platforms
Earlier digital twin models in the literature followed Grieves proposal [8], being
composed of three dimensions: (i) a physical environment, containing physical
entities in real space; (ii) a virtual environment, consisting of virtual objects
and computational tools in digital space; and (iii) the data and information
connection, linking the physical and the virtual environments. Furthermore,
Grieves and Vickers [10] classifies digital twins into three sub-types: (i) a digital
twin prototype, which includes requirements and models related to the concept
of a physical entity; (ii) a digital twin instance, describing a specific physical
entity that the instance remains linked to throughout the life-cycle of that entity;
and (iii) the digital twin environment, an integrated, multi-domain physics
application space for operating on digital twins, including the laws of physics
and rules that every twin instance in this environment must respect.
A more sophisticated model proposes five dimensions for digital twins, includ-
ing [17]: (i) a physical environment; (ii) a virtual environment, regarded
as the sum of all models representing physical entities; (iii) services to be ex-
ecuted by the Digital Twins, such as model execution, visualization, machine
learning prediction, prescriptive task allocation and maintenance; (iv) the data
to which the Digital Twin has access, coming both from the physical and the vir-
tual environments; and (v) connections between every other dimension. Unlike
the three-dimensional model, the 5D model separates data and communications,
and further includes services within digital twin domain, in general being of-
fered externally to the virtual environment where the digital twin settings and
properties are defined, such as in a cloud platform.
2.2 Digital Twin Platforms
The development of a digital twin is a computational workflow composed of
computational services representing models for the process stage, and its interac-
tions. Therefore, digital twin platforms should provide an orchestration of these
different independent services to provide both flexibility and computational per-
formance. A digital twin platform should provide different levels of abstraction,
including [2]: (i) the digital twin user level, where the user gains an access
to the available digital twins in the form of applications; (ii) the digital twin
developer, where the platform provides resources for the development of digi-
tal twins; (iii) the computational service developer level, with the platform
providing an API for the development of computational services; and (iv) the
infrastructure provider level, where instances of the computing services are
mapped to computing resources.
4 C. Barros et al.
2.3 Digital Twin Maturity Levels
A digital twin is the product of an orchestration of computational services and
available data connected to their physical counterparts in real-time throughout
their life-cycle, and therefore it can be used to monitor the current state of the
objects, predict future states, prescribe desired states, and remotely correct real-
object states in order to fulfill some real-world requirement. Verdouw et al. [22]
proposes five different maturity levels of a Digital Twin according to its func-
tionalities: (i) an imaginary twin is a conceptual entity capable of describing an
object that does not yet exist, containing information necessary to materialize
it, i.e. a digital twin prototype; (ii) a monitoring twin is a digital represen-
tation of the real state, behaviour and trajectory of a physical object present
in the system, similar to a digital twin instance; (iii) a predictive twin a dig-
ital projection of the future states and behaviours of the physical object using
predictive analytics such as statistics, simulation and machine learning methods;
(iv) a prescriptive twin is an intelligent digital object capable of recommending
corrective and predictive actions on the physical object based on optimization
algorithms and specialized heuristics; and (v) an autonomous twin operates
autonomously and fully controls the behaviour of real objects without human
intervention, becoming self-adaptive systems able to learn about the environ-
ment, perform self-diagnosis and adapt to user preferences.
3 Requirements for an Ontology of Digital Twins
We identified several Competency Questions that ontologies for digital twins
should address. For each competency question a possible answer is advanced
and brought up for discussion (see Table 1).
3.1 Models and Multi-Level Theory Requirement
Since the virtual environment of a digital twin is composed of models of the
physical environment, the first step in developing an ontology of digital twins is to
elaborate the concept of a model in detail. Following the definition of a model as
an artifact that abstracts a system or a process from a particular perspective [3],
one could notice the distinction between: (i) the conception of modeling, which is
regarded as a model universal; (ii) a model as a representation encompassing
information, rules, methods and premises, such as a 3D model or a class model;
(iii) a model of a type (a prototype), including a 3D model of a car; and (iv)
a model of an individual, i.e. a 3D model of John’s car.
Note that a model universal is a type whose individuals are models, as every
model instantiates a particular concept of modeling (e.g., 3D Model, 2D Model).
General model universals can be specialized in particular subtypes of model uni-
versals that instantiate particular prototypes, (i.e. specifications of properties,
dispositions and modes) that a model of an individual must conform to. The
same entity in reality can be represented by models instantiating different pro-
totypes (i.e., specific subtypes of model universals). Therefore, a multi-level
Requirements for an Ontology of Digital Twins 5
Table 1. Competency questions for an ontology of digital twins.
# Question Sec.
Q1 What is the difference between a model universal, a prototype, and a 3.1
model of an individual?
Q2 What is a model of a substantial? 3.1
Q3 How does a model relate to real world entities? 3.1, 3.2
Q4 How to model an event in the real world bringing about a situation in 3.1
which a substantial in the real world is present?
Q5 How to represent a moment inhering in a substantial in the real world 3.1
in a model of that substantial?
Q6 What is a twin instance? 3.2
Q7 Given a substantial having parts and otherwise related individuals, how 3.2
does a twin instance of the former relate to twin instances of the latter?
Q8 Who decides which substantials in the real world are going to be mod- 3.2
eled by twins and how?
Q9 What is a twin prototype? 3.2
Q10 What is the difference between hardware twins and digital twins? 3.3
Q11 How are hardware twin instances or digital twin instances created? 3.3
Q12 How does an execution of a digital twin instance give information about 3.3, 3.5
situations in the real world where modeled substantials are present?
Q13 What is a twinning disposition of a digital twin instance? 3.4
Q14 What is a twinning process and how does it manifests a twinning dis- 3.4
position?
Q15 What is a feedback brought by a digital twin instance and how is it 3.5
triggered?
Q16 What are the temporal references in a digital twin context? 3.6
Q17 What is a possible situation? 3.6
Q18 What is a digital sibling instance? 3.6
Q19 How does a digital sibling instance bring feedback about a possible 3.6
situation?
theory [5] is a fundamental requirement to delineate the description of a model
within an ontology of digital twins, embracing topics including powertypes and
categorization schemes which are themselves part of subject matter [12].
Following Guizzardi et al. [12], a prototype may be regarded as a powertype
whose instances are variable embodiments which classify instances of models of
an individual. A variable embodiment is an individual f that, at each world
w, picks up a particular rigid embodiment - a set of individuals standing on a
relation - according to a given principle F . As such, a prototype can hold gen-
eral properties that characterize that variable embodiment including: resultant
properties from instances of models of an individual classified by it; regularity
properies that capture regularities over the instances of a particular type; and di-
rect properties of the type, but not of any individual instance [12] (see Figure 1).
From now on, we focus on models and prototypes of substantials, i.e. existen-
tially independent individuals. Even though it is possible to model events and
6 C. Barros et al.
«kind» «powertype»
3D Model of an Individual 3D Model Prototype
Volume
is classified by Number of Instances
Number of Components
Average Volume
Program Configuration
«subkind» «subkind»
3D Model of a Car 3D Model of a Plane
instance of instance of
instance of instance of
3D Model of a Car 3D Model of a Plane
3D Model of a Specific Car 3D Model of a Specific Plane Number of Instances = 35 Number of Instances = 35
Volume = 2567L Volume = 28425L Average Volume = 3400L Average Volume = 34000L
Number of Components = 15 Number of Components = 78 Program Configuration = car.json Program Configuration = pln.json
Fig. 1. A UML class diagram showing the relationship between instances of a prototype
and instances of models of an individual.
moments, digital twins are usually related to objects (car, spacecraft or an oil
platform) or agents (person, company or a society).
An event in the real world would bring about a situation in which the
substantial of interest participates. Such an event must be mapped to a cor-
responding event in the digital world that would bring about a situation
involving the model, which reflects the real situation. Similarly, a moment in-
hering in a substantial can be abstracted into a moment that inheres in
the model of that substantial.
The term real-world in our context refers to the portion of reality which
refers to the substantial and other individuals of interest. We take here the
real-world to be a complex situation composed by situations in which these
individuals participate.
3.2 Twin Instances, Mereology, Relationships and Twin Prototypes
The first use of something to monitor real objects and test possible situations
was by the use of an object, a hardware twin assumed identical to the one
studied, whose goal was to simulate the exact conditions of the object of interest
to obtain a representation of reality. Therefore, before specifying whether a twin
is digital or non-digital, a more fundamental definition of a twin, regardless of
its physical manifestation, should be conceived.
Henceforth, the difference between a twin and a model becomes quite sub-
tle. Considering that a twin is intended to be expandable, i.e. being able to
integrate, add or replace models [18], another essential requirement is that a
twin can be understood as a composition of models. More specifically, a twin
instance is a complex model of individuals composed of one or more mod-
els of individuals. As an example, a twin instance of an specific car could be
composed of a 3D model of that car, a database table containing values of prop-
erties including speed, acceleration and direction of that car, and a predictive
model which receives, as an input, images captured from a camera attached on
the front of the car, and, as an output, detected objects with their positions and
categories.
Requirements for an Ontology of Digital Twins 7
It is noteworthy that a twin instance could contain models of different in-
dividuals, such as in our previous example, where the twin instance of a car
contains a model of a camera attached to the car. Nevertheless, the twin in-
stance refers to the car, implying that the camera is, according to the twin, a
part of it. Moreover, a twin instance may incorporate other twin instances (i.e.
a twin of the camera, which is part of the twin of the car), and an conceptual
theory of mereological relations is required to implement such particularities
of a twin [11], in addition to ensuring the twins are scalable, providing insights
at different scales [18].
With such discussion, it is not mandatory to create twins of every substantial
present in a system, since a model is not necessarily a twin. The decision of which
substantials are going to be modeled or even projected into a twin is reflected in
the intention of a stakeholder (who is responsible for the twin instance) whose
interest is to represent the system these substantials compose. For instance, if
the intention of creating the twin of a car is to control speed, acceleration and
direction, one would not necessarily instantiate a twin of the driver. However, if
the objective includes modeling impacts from a possible car accident, probably
that additional twin would be needed.
One question, however, that arises from this example is: is the driver twin
part of the car twin? As the aforementioned substantials have no mereological
connection, instead have relationships that range from physical contact to own-
ership and control (i.e., the driver owns and controls the car), such aspects need
to be reflected in their twins as well. Consequently, a theory of relations must
be encompassed by an ontology of digital twins in order to faithfully abstract
interactions, physical and social bonds in the real world [6].
Similarly, a twin prototype can be discussed from the view of a complex
prototype. In the aforementioned example, a twin prototype of a car could con-
tain the 3D prototype of a car, the schema describing the properties mentioned
above (speed, acceleration and direction), and the specifications of a predictive
model for the camera. A twin instance, therefore, is a particular exemplification
of a twin prototype.
Next, it is essential to discuss how to physically manifest these twins, giving
models the capacity to bring information to a stakeholder about some part
of the reality where the real substantial participates, instantiating the modeled
specifications, properties, and dispositions into an object that is considered, by
some stakeholder, a reliable representation of the substantial of interest. Due to
the history of the development of digital twins, it is interesting to distinguish a
hardware twin from a digital twin.
3.3 Physical Manifestation of a Twin: Hardware Twins and Digital
Twins
The creation of a hardware or a digital twin instance is described as an activity
occurrence: (i) requiring resources, such as every physical resource to build
the substantial of the same type and a hardware twin prototype (for a hardware
twin), or a machine to execute software and a digital twin prototype in the case
8 C. Barros et al.
of a digital twin, along with their corresponding twin instance, which refers to
the substantial whose representation is intended; (ii) adopting procedures, such
as a document template containing the description of the twin that is going
to be implemented either physically or virtually, and a method, which are the
plan description for implementing the model in its physical manifestation. This
activity occurrence receives, as an input, a twin instance, and outputs a hardware
twin instance or a digital twin instance, depending on the chosen resources. These
steps require elements of a process ontology, which is more precisely defined in
the context of software [13], and of significant interest to the digital twins area.
Although it is easier to check that the hardware twin instance is an artifact
composed of (i) a substantial of the same type of the one to which the twin
instance refers and (ii) the twin instance, a digital twin instance is more
subtle to define. In fact, it may be understood as an artifact composed of a twin
instance and a program, implementing a program specification intending
to satisfy some higher-level system specification, in this case regarding the
representation of the real substantial. The existence of digital objects and their
role in ontologies are important for a fuller definition of digital twins [16].
The Software Ontology (SwO) [3] proposes several details about how a pro-
gram has physical dispositions to be copied and loaded into a machine, and the
execution of a loaded copy is an event which physically manifests the afore-
mentioned disposition. Moreover, the execution brings about an observable
state, a particular situation involving the qualities and values of the machine
in which the loaded program copy inheres, as well as of entities residing in that
machine. Hence, the execution of a loaded copy of a digital twin instance should
bring about an observable state where the twin instance (the models) and its
moments (properties and dispositions) participate.
In addition, there is an agent who is involved in the activity occurrences
and becomes responsible for the physically-manifested twin, being regarded as
a twin developer (further specialized into hardware twin developer or a digital
twin developer). In fact, one can argue that a twin developer is a role played by
a stakeholder when he/she becomes responsible for these physically-manifested
twins. Therefore, these intentionally created objects require an ontology of arti-
facts [23] to fully describe their context of development and use.
Focusing the discussion from now on digital twins, it is necessary to under-
stand in more detail how an observable state brought about by an execution of a
loaded copy of a digital twin instance refers to a situation in the real world where
the real substantial participates. We need to discuss how, after the creation of
a digital twin instance, the state of the real-world substantial and the state of
its digital counterpart are synchronized, a process known in the literature as
twinning.
3.4 The Synchronization of the Real-World Substantial and the
Digital Twin Instance
A digital twin instance inheres a disposition to update its current state in order
to match one or more moments of the twin instance composing it (such as prop-
Requirements for an Ontology of Digital Twins 9
erties, relations with other substantials, modes or dispositions) with the current
state of the substantial it refers to. That disposition is manifested by events as-
sociated with the data lifecycle in a digital twin. Thus, a more complete view of
this synchronization step is to describe it as software processes [13], whose asso-
ciated activities relate to the different steps from data collection, transmission,
storage and integration, to data processing, cleaning, analysis and mining [1],
and the result is an updated state of the digital twin instance.
Twinning may be also interpreted as a commitment that the digital twin has
to reliably represent the real substantial. However, as digital twins are typically
non-agentive objects, they cannot bear intentional moments (including commit-
ments). They are programmed by a user, which, in turn bears these intentional
moments. Instead, we take twinning as a complex disposition reflected in the
digital twin disposition to reproduce the state of the real substantial, and the
disposition of the real substantial to provide data to help the digital twin update
its own state. The different computational steps to perform such synchronization
can be expressed through twinning processes.
After defining how the real world affects the digital world, the next step is
to define different computational services to which the digital twin must have
access, and more specifically how the results obtained in the digital world have
an impact in the real world. In fact, the state of the digital twin reflects the state
of a substantial, so it is essential to describe in detail how such a reflection is
perceived by the stakeholder responsible for the digital twin, and what kind of
information is inferred from that computational state.
3.5 Feedback about the Real World
A digital twin instance is composed by a program whose execution of its loaded
copy brings about an observable state in the computational environment which
refers to a situation in the physical environment where the substantial referred
by the twin instance participates. Although theoretically this observable state
gives a complete information about the real world, in practice, we have: (i) uncer-
tainties of the twinning process that do not guarantee that the observable state
of the digital twin instance is mapped identically to the situation in which the
substantial participates; and (ii) situations that satisfy the same propositional
content of interest as the observable state. For instance, if the execution of the
digital twin instance of a specific car brings about the following state: ‘the car
is moving with a speed of 50km/h’, ideally there is a situation ‘the real car is
moving with a speed of 50km/h’. Nevertheless, other situations could be: (i) the
car is moving with a speed between 48km/h and 52km/h (uncertainty), and (ii)
the car is below the speed limit allowed by the urban road where it is travelling.
Therefore, one may define the feedback brought by a digital twin instance as
an event that creates a belief in the digital twin user. A digital twin user is
a role played by a stakeholder who is interested in a situation where a particular
substantial is present (i.e., the observable state given by the execution of the
digital twin refers to a situation in which the substantial that twin models is
10 C. Barros et al.
present). As a belief, it can be justified if the situation referred by that observable
state obtains.
To trigger a feedback, a digital twin user must run or schedule the execution
of programs within a software product that participate in software processes
as software resources, resulting in the updated status of the digital twin in-
stance and the feedback event. This requires elements of the Software Ontology
(SwO) [3], and Software Process Ontology (which includes activities, resources
and scheduled actions [13]).
3.6 Computational Services, Temporal Structure and Possible
Situations
To detail the role of computational services in a digital twin domain, it is
necessary to grasp the importance of time, which is addressed in details in
UFO-B [15]. There are three temporal references in a problem involving digital
twins: (i) the user temporal reference; (ii) the temporal reference of the
digital twin instance, built from the data captured and stored in the computer
system; and (iii) the temporal reference of the modeled substantial, in
terms of the user interest. More specifically, the temporal reference is associated
with the time point or the time interval both of the observable state of the digital
twin and of the situation in which a substantial is present and about which a
stakeholder intends to obtain feedback.
Regarding the observable state of the Digital Twin instance, it is possible
both to refer to data obtained in real time (in this case, the time reference of
the digital twin is similar to that of the user), as well as to stored historical data
(or that is, the digital twin represents an observable state with data obtained
in the past) or to simulations (which can either represent a possible observable
state in the past or in the future, or even in counterfactual situations). As for the
situation of the real entity, it is possible to obtain feedback that leads to a belief
that an observable state infers a situation in real time, infers a situation that
happened in the past or even a situation that may happen in the future. Several
tasks, including real-time monitoring, historical data analytics, real-time
analytics, simulation, prediction, and prescription, may be executed as
computational services within the virtual environment. Moreover, when the goal
behind a task is to get feedback from a situation in the real world that it is not
known whether it happened or not, or that might happen in the future, we can
consider the existence of possible situations. These concepts may be further
explored with the help of ontologies addressing simulations aspects [14].
Virtual representations of reality that intend to give feedback about possible
(but non-actual) situations cannot, by definition, be called digital twins, despite
such a term being used even in this context: These are typically named a sibling,
a spin-off of the reality encompassing a possible represented world. A digital
sibling instance may be regarded, ontologically, as a special kind of digital
artifact, hence an instance of a digital artifact prototype, which is created by
using a digital twin instance as an input.
Requirements for an Ontology of Digital Twins 11
3.7 A Preliminary Conceptual Model of Some Digital Twins
Notions
We here propose a very preliminary conceptual model of some Digital Twin
notions, connecting these concepts to those of UFO and some of its extensions.
This model is show in Figure 2.
Universal Individual Activity
Substantial Real-World
Endurant Event
Universal Event
instance of ◀
triggers▲ triggers▲
brings about▼ brings about▼
refers to▲ Substantial Moment
Real-World
Situation
Situation
infers ▶ participates in▲
Artifact Stakeholder
Belief Disposition Real-World
Substantial
Document Model Program inheres in▼
Digital Twin Observable
User observation
State
of ▶ brings
interested about▲
Model of a in ▼ activates ◀
Prototype classified by ◀ Feedback
Substantial
Loaded DT
materialized by ▶
Program Copy
DT Program
Twin Prototype Twin Instance
execution of ▶ Copy Execution
measures
state of ▲
Digital Twin classified by ◀
Digital Twin Twinning
Twinning
Prototype Instance inheres in ◀ Process
manifested by ▶
updates state of ◀
represents ▶
Fig. 2. A digital twin conceptual model encompassing requirements and discussions
aforementioned where (i) blue notions are from UFO and ontologies grounded in UFO;
(ii) red ones are specific to digital twins, and (iii) green ones refer to real-world elements.
4 Related Work
To the best of our knowledge, no attempts have been made so far to discuss
requirements to develop an ontology of digital twins focusing on fundamental
aspects of their existence and use. Nevertheless, some authors propose ontologies
and conceptual models aimed at applications or more specific computational
aspects, such as managing digital twin databases. Barth et al. [1] structures
digital twin applications by developing a domain-specific ontology summarized
by their data resources, external value creation and internal value creation. Singh
et al. [20] capture and model the conceptual knowledge of the digital twin domain
by an ontology model further transformed into a mininum data model structure
to map, query and manage databases for digital twin applications. Steinmetz
et al. [21] propose an ontology for digital twins of devices in an IoT system,
providing them with application programming interfaces (APIs) and human-
machine interfaces.
12 C. Barros et al.
5 Conclusion and Future Directions
This work discusses several requirements to build an ontology of digital twins,
encompassing different steps from models to the conception of the digital twin,
creation of the prototype and materialization of its instances, in addition to some
of its main functionalities, such as real-world synchronization, visualization, data
analysis, prediction, and simulation. Possible future directions based on this work
include:
– Design and implement a core reference ontology of digital twins, using on-
tologically well-founded languages such as OntoUML [11], and following the
Systematic Approach for Building Ontologies (SABiO) [4].
– Further elaborate the relation between ontology, which describes reality, and
phenomenology, which seeks to understand causes and effects, in addition to
simulate possible worlds using laws of nature modelled by mathematical and
computational tools.
– From agent-based ontologies, extend the discussion to include autonomous
digital twins.
We expect that our paper motivates the development of novel ontologies of
digital twins, specially those grounded in foundational ontologies.
Acknowledgements. This research is partially supported by Accenture Israel
Cyber R&D Lab (RiskGraph Project).
References
1. Barth, L., Ehrat, M., Fuchs, R., Haarmann, J.: Systematization of digital twins:
Ontology and conceptual framework. In: Proceedings of the 2020 The 3rd Interna-
tional Conference on Information Science and System. pp. 13–23 (2020)
2. Borodulin, K., Radchenko, G., Shestakov, A., Sokolinsky, L., Tchernykh, A., Pro-
dan, R.: Towards digital twins cloud platform: Microservices and computational
workflows to rule a smart factory. In: Proceedings of the10th International Con-
ference on Utility and Cloud Computing. pp. 209–210 (2017)
3. Duarte, B.B., de Castro Leal, A.L., de Almeida Falbo, R., Guizzardi, G., Guizzardi,
R.S., Souza, V.E.S.: Ontological foundations for software requirements with a focus
on requirements at runtime. Applied Ontology 13(2), 73–105 (2018)
4. Falbo, R.A.: Sabio: Systematic approach for building ontologies. In: Proc. of the
Proceedings of the 1st Joint Workshop ONTO.COM/ODISE on Ontologies in Con-
ceptual Modeling and Information Systems Engineering. CEUR (2014)
5. Fonseca, C.M., Almeida, J.P.A., Guizzardi, G., Carvalho, V.A.: Multi-level concep-
tual modeling: Theory, language and application. Data & Knowledge Engineering
134, 101894 (2021)
6. Fonseca, C.M. et al.: Relations in ontology-driven conceptual modeling. In: Inter-
national Conference on Conceptual Modeling. Springer (2019)
7. Gelernter, D.: Mirror Worlds: Or the day software puts the Universe in a shoebox...
How it will happen and what it will mean. Oxford University Press (1993)
8. Grieves, M.: Digital twin: Manufacturing excellence through virtual factory repli-
cation. White Paper 1, 1–7 (2014)
Requirements for an Ontology of Digital Twins 13
9. Grieves, M.: Virtually intelligent product systems: Digital and physical twins. In:
Flumerfelt, S., Schwartz, K.G., Mavris, D., Briceno, S. (eds.) Complex Systems
Engineering: Theory and Practice. American Institute of Aeronautics and Astro-
nautics (2019)
10. Grieves, M., Vickers, J.: Digital twin: Mitigating unpredictable, undesirable emer-
gent behavior in complex systems. In: Transdisciplinary Perspectives on Complex
Systems, pp. 85–113. Springer (2017)
11. Guizzardi, G.: Ontological Foundations for Structural Conceptual Models. Ph.D.
thesis, University of Twente, The Netherlands (2005), published as the book “Onto-
logical Foundations for Structural Conceptual Models”, Telematica Institute Fun-
damental Research Series No. 15, 05–74, ISBN 90-75176-81-3 ISSN 1388-1795; No.
015; CTIT Ph.D. Thesis, ISSN 1381-3617
12. Guizzardi, G., Almeida, J.P.A., Guarino, N., de Carvalho, V.A.: Towards an onto-
logical analysis of powertypes. In: JOWO@ IJCAI (2015)
13. Guizzardi, G., de Almeida Falbo, R., Guizzardi, R.S.: Grounding software domain
ontologies in the unified foundational ontology (ufo): The case of the ode software
process ontology. In: CIbSE. pp. 127–140. Citeseer (2008)
14. Guizzardi, G., Wagner, G.: Dispositions and causal laws as the ontological foun-
dation of transition rules in simulation models. In: 2013 Winter Simulations Con-
ference (WSC). pp. 1335–1346. IEEE (2013)
15. Guizzardi, G. et al.: Towards ontological foundations for the conceptual model-
ing of events. In: International Conference on Conceptual Modeling. pp. 327–341.
Springer (2013)
16. Hui, Y.: On the existence of digital objects, vol. 48. U of Minnesota Press (2016)
17. Qi, Q., Tao, F., Hu, T., Anwer, N., Liu, A., Wei, Y., Wang, L., Nee, A.Y.: Enabling
Technologies and Tools for Digital Twin. Journal of Manufacturing Systems (2019)
18. Schleich, B., Anwer, N., Mathieu, L., Wartzack, S.: Shaping the digital twin for de-
sign and production engineering. CIRP Annals - Manufacturing Technology 66(1),
141–144 (2017). https://doi.org/10.1016/j.cirp.2017.04.040
19. Schneider, L.: How to build a foundational ontology. In: Annual Conference on
Artificial Intelligence. pp. 120–134. Springer (2003)
20. Singh, S., Shehab, E., Higgins, N., Fowler, K., Reynolds, D., Erkoyuncu, J.A.,
Gadd, P.: Data management for developing digital twin ontology model. Proceed-
ings of the Institution of Mechanical Engineers, Part B: Journal of Engineering
Manufacture p. 0954405420978117 (2020)
21. Steinmetz, C., Rettberg, A., Ribeiro, F.G.C., Schroeder, G., Pereira, C.E.: Inter-
net of things ontology for digital twin in cyber physical systems. In: 2018 VIII
Brazilian Symposium on Computing Systems Engineering (SBESC). pp. 154–159.
IEEE (2018)
22. Verdouw, C., Tekinerdogan, B., Beulens, A., Wolfert, S.: Digital twins in smart
farming. Agricultural Systems 189, 103046 (2021)
23. Weigand, H., Johannesson, P., Andersson, B.: An artifact ontology for design sci-
ence research. Data & Knowledge Engineering 133, 101878 (2021)
24. Zhang, H., Zhang, G., Yan, Q.: Digital twin-driven cyber-physical production sys-
tem towards smart shop-floor. Journal of Ambient Intelligence and Humanized
Computing 10(11), 4439–4453 (2019)